article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
free will ( fw ) , as we informally understand it in daily life , is the power of a rational agent to pick her / his own choice from among various alternative possbilities .
but what exactly is fw ?
the age - old question has provoked much debate among philosophers and scientists @xcite .
our outlook on the world implicitly assumes that human behavior is governed by fw : we choose , we plan and we normally hold people responsible for what they say or do , either because we imagine that they are at liberty also not to act so , or because we do nt have the freedom to believe otherwise ! neuroscience has been uncovering causal chains that appear to explain our choices , emotions and even beliefs , in terms of neurophysical events extending back to the pre - natal stage . as much as it is interesting and important to understand whether we have fw
, there is a more elementary and pressing problem to deal with : namely to try to define fw rigorously and scientifically in a way that agrees with the above intuitive notion .
this last requirement is important , as one can define almost anything one wants in a theory of one s own design @xcite .
having defined fw , one might ask whether such a fw is compatible with known scientific knowledge or can be demonstrated experimentally as a new kind of resource in nature .
philosophers over the millenia have proposed different responses to the problem .
two prominent positions are _ compatibilism _ , according to which determinism is compatible with fw , and _ incompatibilism _ , according to which the two are incompatible
. there are shades of intermediate positions .
compatibilism is espoused by , among others , calvinists who believe that personal freedom to choose does not preclude foreknowledge ( possessed by a all - powerful intelligence ) of future choices .
two broad incompatibilist positions are _ ( metaphysical ) libertarianism _ and _ hard determinism_. according to the former , in a situation of alternative possibilities ( ap ) which is ontologically ( metaphysically ) available , an agent has the liberty to choose one or other option , which is not pre - determined .
this position upholds fw , and rejects determinism .
by contrast , _ hard determinism _
upholds determinism and rejects fw , relegating it to an illusory feeling . in western philosophical thought ,
this is represented by , among others , armenianism .
recently , a number of physicists have studied fw mainly ( but not exclusively ) in connection with quantum mechanics ( qm ) @xcite .
the important contribution of quantum mechanics to this debate is in introducing a concrete instance of fundamental indeterminism via the @xmath0 born rule .
whether this indeterminism is epistemic or ontic still remains a moot issue and is related to the notion of realism in the sense of the bell - kochen - specker theorems . in the quantum approach to fw , some have assumed that quantum indeterminacy gives ` elbow room ' for fw to act , while others identify fw with unpredictability , or independence from all past information ( with ` past ' defined in an appropriately relativistically invariant way ) .
we call this approach _ soft incompatibilism _ , according to which fw is incompatible with determinism but compatible with indeterminism
. however , this approach , which also finds favor in the neuroscientific and artificial intelligence ( ai ) communities , fails to capture the sense of control , of having liberty and making a deliberate choice , implicit in the notion of fw . in this work
, we aim to define libertarian fw , which encompasses such a liberty , and to study its relation to the laws of physics .
libertarian fw is arguably the most intuitive version of fw , and yet , as we will find , the most elusive to define .
we do not claim that it is somehow a better variety of fw , or that it can be experimentally demonstrated at this time .
it is our preferred understanding of fw simply because it feels right !
the article is arranged as follows .
section [ sec : mackay ] discusses a specific compatibilist version of fw , due to the cognitive scientist mackay .
section [ sec : wfwp ] presents a logical paradox that would afflict attempts to define libertarian fw , and presents a resolution .
section [ sec : sfwp ] explains another basic logical paradox about libertarian fw .
this logical barrier is surmounted in section [ sec : fun ] through a new model of fw .
certain biological and philosophical aspects of the model are considered in section [ sec : funhash ] .
sections [ sec : fun+ ] and [ sec : fun++ ] may be omitted by readers not interested in quantitative aspects of the model .
section [ sec : fun+ ] presents a simple mathematical representation of the proposed model , while section [ sec : fun++ ] applies it to quantum mechanics and considers some unusual consequences .
neurological tests , implications and medical applications of our model are considered in section [ sec : neuro ] .
we finally conclude with section [ sec : end ] , with a discussion on fw in connection with darwinian evolution and the subject of consciousness .
we mention two flavors of what are arguably compatibilist positions : calvinist - like ( mentioned above ) and mackayan . in the mackayan position @xcite ,
the idea here is that even in deterministic dynamics , self - predictions lead to indeterminacy , which is interpretable as fw . according to this argument , given prediction @xmath1 about agent @xmath2 ,
if @xmath2 believes @xmath1 , then she can falsify it by acting contrary to @xmath1 .
one might suppose that a more detailed algorithm to arrive at @xmath1 would be able to take into consideration @xmath2 s reaction , but @xmath2 can simply choose to falsify _
it then follows that only if @xmath2 disbelieves @xmath1 , then @xmath1 holds with certainty , but then this would make @xmath2 a kind of inconsistent agent , for disbelieving a truth .
thus , if @xmath2 ( and hence the universe ) is consistent , then @xmath2 must believe @xmath1 . to avoid the first inconsistency , that she falsify @xmath1 if she knew it ,
we conclude that believing @xmath1 affects @xmath2 s brain state , invalidating the premise that went into reaching @xmath1 .
thus , @xmath2 s falsification of @xmath1 does not imply inconsistency .
but it does imply that any consistent predictive algorithm @xmath1 is unable to encompass her reaction to @xmath1 .
therefore , according to mackay , it is upto @xmath2 how @xmath2 chooses to act ; she can make @xmath1 or @xmath3 true .
this indeterminability is interpreted as fw .
a caeful scrutiny reveals the similarity of the structure of the above argument analogous to gdel s celebrated incompleteness theorem @xcite , when the notion of _ provability _ is replaced with that of _ belief _ , though , strangely , mackay denied the connection . to recast the mackayan position in a form closer to gdel s theorem : in a consistent sufficiently rich , deterministic world
, there will be unpredictable situations .
not surprisingly on account of the presence of self - reference , mackay s argument can also be compared readily to turing s proof of the unsolvability of the halting problem , where the putative algorithm that outputs @xmath1 is analogous to the would - be halting algorithm @xmath4 and the nay - saying agent to the contradictive turing machine that uses the @xmath4 as a sub - routine .
mackayan free will , like turing uncomputability , comes from the fact that the cardinality of the set of outcome situations is greater than proofs of prediction of an agent s choice .
thus , in the mackayan argument , unpredictability of a free agent s choice is identified with gdel incompleteness and turing uncomputability . but does incompleteness or uncomputability constiute a kind of fw ?
intriguing though the idea is , we think that it is , ultimately , philosophically unsatisfactory , because it identifies free will with unpredictability rather than with genuine freedom , or even indeterminacy .
likewise , from a philosophical perspective , we think that the soft incompatibilist approach from quantum mechanics does not go far enough , because sense of control in fw is not explained .
fw in its various quantum avatars looks just like free whim !
the reason libertarian fw resists definition is that attempting to define it leads to logical paradoxes , which we discuss in this and the next section .
although these paradoxes were not , to our knowledge , explicitly mentioned before , yet they implicitly crop up in past attempts to define fw , thwarting a definitive resolution to the problem for over two millenia . it is not surprising that in the face of this daunting logical barrier , scientists tend to adopt a position other than libertarianism .
if we accept the libertarian position , then by assumption of incompatibilism , one rules out determinism as the universally correct description of the physical laws .
but neither does indeterminism ( as governed by some fixed probability rule , @xmath1 ) leave enough room for fw to act .
consider the sample mean @xmath5 over @xmath6 trials @xmath7 by the weak law of large numbers .
this implies that there is a ` probability pressure ' _ not _ to choose _ atypical _ sequences , which would cause deviations from the sample mean .
thus , there is a kind of long - run determinism , and hence a restriction on fw as we intuitively understand it .
for example , given a coin with outcome space @xmath8 and the corresponding probability vector @xmath9 , the coin is not free to indefinitely choose outputs hhhhhhhhh@xmath10 .
thus liberatarian fw is compatible with neither determinism nor indeterminism , and thus belongs to the position of _ hard incompatibilism_. this situation creates a paradox for the libertarian position , which we call the weak fw paradox ( wfwp ) , whereby fw is compataible with neither of the available alternatives .
the other mentioned positions on fw are unaffected by wfwp .
therefore , if we accept libertarian fw , the free - willed choice will potentially intefere with the underlying physical dynamics @xmath11 .
this interference will take the form of ( a ) overriding causality , if @xmath12 is deterministic ; or ( b ) causing deviations from the relevant probability rule @xmath1 , if @xmath12 is indeterministic . clearly , it is immaterial whether the underlying physics is classical or quantum .
the arugment sometimes made , that quantum indeterminism gives ` room ' for free - willed action is thus not found to be persuasive .
fw itself ca nt be part of the dynamics @xmath11 , for in that case it could not produce the required deviations .
therefore , we require a cartesian dualism with a physical and _ extra - physical _ component making up a free - willed agent
. physical dynamics @xmath12 governs the former while fw comes from the latter .
the above resolution of wfwp says that if libertarian fw exists , it must be an extra - physical resource whose intervention causes deviations from @xmath11 .
it still leaves open the question what libertarian fw is or how it can be consistently defined . here
we will show how trying to pin down its nature leads to another paradox , which we call the strong free will paradox ( swfp ) .
the paradox is obtained essentially by extending the traditional incompatibilist claim for the incompatibility of determinism and fw , to that of _ indeterminism _ and fw .
the idea is that the truth of determinism or indeterminism would mean that we do nt control actions in a way one would expect of self - determining , free - willed agents .
since determinism and indeterminism are the only available causal primitives , there is no such thing as libertarian fw , according to this line of reasoning .
there is a similarity to wfwp , but the focus has shifted from `` where ? '' to `` what ? '' , or even `` whether ? '' .
we present two slightly different versions of sfwp . in the first of them
, one assumes that fw exists and tries to locate it in terms of properties of the agent . according to this argument : * [ i ] * an agent has free will only when her intentional actions emerge from the agent herself . *
[ ii ] * therefore , they are deterministic functions of her volitions , beliefs , desires etc , which we denote by _ variables of intent _ , @xmath13 . * [ iii ] * if there are no prior causes of her volitions , desires , etc . , then the values @xmath13 take on must be random , making the agent s volitions , desires , etc . , and hence her actions , whimsical .
* [ iv ] * since whimsicality undermines the notion of control or intent , @xmath13 must not be random , but deterministic functions of some other second order intent variables @xmath14 , which , by virtue of * [ i ] * , must belong to the agent herself . *
[ v ] * as before , the variables @xmath14 themselves are caused or uncaused .
if the latter , then @xmath14 are random , making @xmath13 , and hence also her actions , ultimately whimsical .
but if the former , then @xmath14 are not free but depend , by a similar reasoning , on higher order variables @xmath15 , and so on similarly to even higher orders indefinitely .
* [ vi ] * if this pattern of recursion terminates at some finite depth @xmath6 , then @xmath16 either has no causes or external causes of unspecified origin . in the former case , we obtain capricious , indeterministic behavior , in the latter case , unfree , deterministic behavior .
neither connotes fw as we recognize it . in brief ,
when we try to incorporate the notion of will or intent into the action , the action becomes deterministic and hence unfree .
putting freedom back means removing determinism , which undermines intent by making the action random and the agent whimsical .
thus the ` free ' and the ` will ' in ` free will ' are at loggerheads , making the word an oxymoron .
this is the strong free will paradox ( sfwp ) , according to which randomness and determinism seem to be the only fundamental causal primitives in nature , with libertarian fw a figment of imagination .
it is of interest to note that certain classical accounts of fw fall prey to swfp , illustrating its elusiveness to define .
according to thomas hobbes , `` a free agent is he that can do as he will , and forbear as he will .... '' .
david hume characterizes it thus : power of acting or of not acting , according to the determination of the will : that is , if we choose to remain at rest , we may ; if we choose to move , we also may .... this hypothetical liberty is universally allowed to belong to everyone who is not a prisoner and in chains . " to paraphrase in terms of our discussion above , they both are essentially saying @xmath17 , but were nt taking the argument farther .
arthur schopenhauer does take it one step farther but says : `` you can do what you will , but you can not will what you will . in any given moment of your life you can will only one definite thing and absolutely nothing other than that one thing . ''
he is thus effectively responding to swfp by characterizing @xmath18 and denying the existence of libertarian fw .
one can try other variants , such as @xmath19 being a probabilistic function , whereas @xmath20 being deterministic , so that there is mix of determinism ( will ) and indeterminism ( freedom ) at _ different _ levels of the agent s personality .
however , the core of sfwp remains .
a related , but different version of sfwp , tries to identify fw as a particular kind of influence over the act of making a choice . according to this argument : * [ a ] * suppose that from a set of alternative possibilities ( ap ) , a choice is eventually made by an agent . *
[ b ] * from the principle of excluded middle , the choice was made either according to a rule or it was not made according to a rule . * [ c ] * in the former case , there is no genuine ap situation .
hence we have determinism , and there is no fw . *
[ d ] * in the latter case , we have pure randomness , and hence , again , no fw in the sense of the agent s control and voluntary choice .
we thus find again that the agent s choice is deterministic or random , with no apparent room for ( libertarian ) fw .
this is the basis of the _ pessisimist _ view that there can be no such thing as fw @xcite .
implication * [ c ] * lies historically at the heart of the incompatibilist argument . here , if the rule is _ prescriptive _ , we have _ causal determinism _ , as a law of physical dynamics in nature .
if the rule is only _ descriptive _ , we have _ logical determinism _ , with outcomes of future choices being known to an all - powerful intelligence and assigned definite truth values .
we propose the following hard incompatibilist model as a way out of sfwp , drawn from an interpretation of eastern philosophy . according to the model , the conscious personality of the free - willed agent , which may be called the _ ego _ ( after the father of modern psychoanalysis , sigmund freud )
, is influenced in an ap situation by three structural elements .
this tripartite division is at the heart of our model .
two of the elements , and their functions , are as follows : * nature ( n ) .
* imposes mental constraints in the form of desires , instinctive drives and emotional tendencies .
* understanding ( u ) .
* offers guidance in action through a rational capacity to model the world , and understand the ( ethical , social , financial , etc . ) implications of each choice .
when a situation with alternative possibilities is encountered , nature * n * and understanding * u * present their respective recommendations to the ego on selecting one of the alternatives . while the recommendation of nature appears as a _ desire _
, that of understanding appears as a _ thought _ or _
feeling_. as a specific example , the desire may be oriented towards seeking pleasure ( freud s pleasure principle ) , while the thought or feeling may be about conscientious restraint .
the dilemma sometimes experienced by one about to make a choice is the possible conflict between the two recommendations .
it is convenient to designate the _ mind _ as the seat of nature * n*. in freudian psychoanalytic terminology , the mind may be identified with the _
id_. similarly , the seat of @xmath21 is posited to be the _ intellect _ , which may be identified with freudian _
superego_. to use a crude but visually helpful computer analogy , we may picture the ego as the cpu ( central processing unit ) of a compter , with * u * and *
n * being softwares uploaded onto it during the waking period from their respective seats , which are like hard disk memory locations . for non - libertarian accounts of fw , including ai , the two elements * u * and * n * suffice .
they are both _ causal resources
_ , i.e. , systems that can drive the agent s behavior independently , and thus aspects of the dynamics @xmath11 .
the eventual choice will be some deterministic or probabilistic outcome of their interplay .
philosophically speaking , the main drawback with such accounts is that they lack the sense of having liberty to make a deliberate choice .
crucially , for ( libertarian ) fw , * u * is not a causal resource , but a _ guidance resource _ , unlike * n*. when the input from * u * and * n * are experienced as possibly conflicting tendencies by the ego , the guidance by itself is powerless to influence the agent s choice .
a further element is required , the third and last in our model , which expresses the idea of empowering the ego to deliberately choose between the recommendations of desire and guidance .
this is the faculty of volition , or simply : * freedom of will ( f ) . * the _ extra - physical _ freedom to orient or align the choice in line with the understanding @xmath22 , by overcoming , if necessary , the constraints imposed by nature @xmath23 .
the extra - physicality follows from the resolution of the wfwp .
swfp also implies it because if it were not extra - physical , then * f * would be part of the underlying dynamics @xmath11 as a deterministic or indeterministic aspect , and not libertarian , as required . as to the question of
whether such an extra - physical resource is sanctioned by physics or can be tested , we return to it in section [ sec : fun++ ] . here
we are concerned only with obtaining a suitable _ definition_. we call our model ` fun ' in recognition of the fact that it involves the triad of elements * f * , * u * and * n*. faced with an ap situation , nature and understanding may present conflicting recommendations to the ego under some description a desire or drive vs a thought , feeling or belief . if the nature - induced desire is aligned away from the understanding - proposed thought , then supported by fw , the thought acts as a restraint , whereas if the desire is attuned to * u * s recommendation , then fw acts to that extent effortlessly . together
* f * and *
u * constitute a causal resource , in which * f * contributes the magnitude or `` energy '' while * u * the direction . recognizing this is crucial in resolving sfwp , and its oversight is the cause of much confusion in existing accounts of fw .
thus fw is the expression of * f * in a situation where the ego is exposed to the potentially conflicting recommendations of * u * and *
n*. fw as defined above is the new _ causal primitive _ or _ principle of causation _ , apart from determinism and indeterminism . by its fundamental character , it is empowered to override physical causality ( of nature ) under the intellect s guidance .
fw thus enables a state of affairs wherein logical determinism holds whereas causal determinism fails .
we return to this point elsewhere to use this as a basis for a ( meta-)logical framework to characterize fw .
we think that our definition of fw , and a quantification of it below , agree with the notion of _ volitional causation _ proposed in ref .
@xcite , as distinguished from _ physical causation _
the agent s freedom to control , together with the understanding , as we have proposed , bears some similarity to the _ ultimate responsibility _ posited by the philosopher kane @xcite .
however , the role of * f * and * u * need not be confined to moral issues alone , and may extend to other walks of life .
for example : agent @xmath2 is suffering from a medical problem , and has the understanding that @xmath2 needs to undergo treatment .
however @xmath2 s natural tendency , governed by constraints * n * , would be to avoid the treatment because it would cause discomfort .
@xmath2 exercises fw to endure the discomfort in the interest of later cure and longterm benefits .
we will now prove that the fun model resolves sfwp .
in particular , the above example highlights how the _ incompatibilist _ implication * [ c ] * of sfwp fails .
if one happens to be familiar with the _ weltanschauung _ , and hence the understanding of @xmath2 , and @xmath2 has high free will , then the _ freely _ chosen alternative will be _ predictable_. proposition [ pro : saint ] below , whose truth is clear in this light , is rigorously validated by our definition of fw .
we define a _ saint _ as an agent whose understanding is ethical and whose fw is near maximal . presented with the choice between the good and the evil , the saint freely chooses the good .
[ pro : saint ] by definition , the saint , equipped with maximal fw , can if necessary override any constraints like desire and instinctive drives imposed by his human nature , and align his choice in tune with his understanding .
thus , the deterministic behavior and predictability of the saint does not preclude his free will , contrary to the _ incompatibilist _ implication * [ c ] * in swfp .
knowing that a person is a saint , we have sure _ foreknowledge _ of the ethicality of all his future choices , _ because _ he is endowed with high fw .
we have here an instance of a descriptive rule ( logical determinism ) that is strictly non - prescriptive ( i.e. , strictly not causal determinism ) .
intuitively , this is clear , since one should be able to choose freely , even to act predictably . not to be able to choose to act
predictably would imply a constraint on freedom . the instance of saintly determinism does not make our model compatibilistic , as it must be clear . in particular ,
other situations are allowed to exist by our model , that resist compatibilistic interpretation , as noted below .
suppose that an agent s fw is not maximal .
then his choice will fluctuate randomly between choosing according to the dictates of his nature and the recommendation of his understanding .
this is illustrated in the following proposition , whose intuitvely apparent truth is validated rigorously by our definition of fw . presented with the choice between the good and the evil , the conscientious criminal vacillates .
[ pro : cc ] this criminal , being conscientious , has a clear understanding of the virtue of ethical behavior , but , owing to lack of sufficiently high fw , can not always overcome the compulsion of his criminal nature .
his choice is random , being good sometimes and being evil at other times .
thus the randomness of his choice does not imply lack of free will . here
it only implies low free will .
the _ pessimist _
@xcite implication * [ d ] * in swfp is also therefore falsified .
the recognition of fw as a new form of causation also clarifies why the first version of swfp fails .
for example , implication * [ vi ] * is incorrect for reasons stated above .
in contrast to proposition [ pro : saint ] , predictable behavior can arise also because of low degree of fw .
consider : presented with the choice between the good and the evil , the hardcore criminal chooses evil .
[ pro : hcc ] although the hard - core criminal may possess an ethical understanding , yet , because he almost entirely lacks fw , his choice is deterministically decided by his evil nature . knowing that a person is a hardcore criminal , we have foreknowledge of the iniquity of all his future choices because he _ lacks _ any fw .
if we theoretically empower him with some fw , we obtain the scenario of proposition [ pro : cc ] .
in other words , adding a degree of fw removes determinacy . clearly , this prediction of the model is not in accord with compatibilism .
therefore , our model is neither incompatibilist nor compatibilist .
the fun model as it stands does not explain the notion of moral responsibility .
when the given an ap situation is a moral one , responsibility arises arguably because a person has the power * f * to follow through the guidance provided by * u * by overcoming the dictates imposed by * n*. the model does not explain how the differences in the degree of freedom , eg .
, between that of a saint and a conscientious criminal , arise .
unless an agent can somehow be construed as having played a role in the current level of freedom , the judgment of moral responsibility must be relative ( to a given degree of freedom ) .
further assumptions are needed , which will be taken up in a subsequent work , to develop a more complete notion of moral responsibility in the model . for the triad of elements of the fun model described in section [ sec : fun ] ,
we suggest the following plausible brain correlates . for convenience
, we refer to this neurologically adapted version of fun as the fun # model . the element * n * , and hence the mind s brain correlate , is probably the limbic system , which , comprising brain structures like the hippocampus , amygdala , etc . ,
is known to support a variety of functions , including behavior , emotion and long term memory .
the element * u * , and by extension the intellect s brain correlate is located presumably in the brain prefrontal cortex .
this region of the brain is believed by neuroscientists and psychologists to be responsible for many higher cognitive functions , such as planning , differentiating between the good and bad , determining consequences of actions , planning actions , and so on .
the fun # model posits that , while these two regions participate in decision making processes , they do not correspond to the _ ultimate control module _ ( ucm ) , the brain correlate of fw .
this could be a localized region , such as the pineal gland .
the claim for this gland is mainly historical .
the fun # model accepts this as a working hypothesis on the strength of the observation that this gland is centrally located in the brain , as befits a ucm .
physiologically , this would mean that brain areas corresponding to voluntary muscle movement ( the motor cortex ) will be functionally linked to the frontal cortex , the limbic system and the region of the pineal gland , during the execution of planned actions . in the fun model , * f * , * u * and * n * are independent elements that interact when a choice is made . in actual fact , they influence each other at the _ subconscious _ level in complicated situation - dependent ways , so that the elements as they appear at the conscious level are ` dressed versions ' rather than the ` bare versions ' ( to use terminology borrowed from optics or particle physics )
. the agent s personality is thus a nonlinear functional of the elements , where each element influences the others and also itself indirectly through its effect on the others .
an instance where * u * affects * n * is in consumers brand loyalty , which manifests as preferences , detectable as neural correlates , that humans display towards products that are otherwise similar ( cf .
an interesting study reported in ref .
@xcite with regard to drinks ) .
an instance where * n * affects * u * is the familiar case of people showing reduced objectivity when dealing with those they have a strong emotional connection ( whether negative or positive ) with .
an instance where * n * affects ( diminishes ) freedom is when a drug addict ( resp . ,
smoker ) relentlessly seeks his next shot ( resp .
, puff ) even though he knows it is not good for him ; and so on . for this reason , from a behavioral perspective , it would be more convenient to talk not in terms of interaction between the elements , but instead in terms of interaction between three broad traits : that dominated by nature *
n * ( the animalistic " ) , that dominated by freedom ( the saintly " ) , and that wherein these two are in balance ( the human " ) .
the philosophy of _ yoga _ @xcite terms these three basic traits as _ tamas _ , _ rajas _ and _ sattva_. in this line of thought , a human character is constituted by these three basic traits and determined by their relative dominance .
the _ sattvic _ type of humans have the highest degree of fw , while the _ tamasic _ have the lowest . to complete the model
, we offer some thoughts on the extraphysical agency to which the will is ascribed . here
we will appeal to the philosophies of _ vedanta _ @xcite and yoga , which recognize and supply such an entity the _ self _ ( _ tman _ in sanskrit ) , which is quite distinct from the ego ( _ ahamkra _ ) and conceived of as the essential individual . at a deeper level of consciousness , the self fans out to serve as the overarching _
substratum _ on which the agent s faculties of ego , understanding and nature rest . at the deepest level of consciousness ,
there is no structure and the individual selves resolve into an unindividuated absolute .
contradictory as these descriptions seem , they can be readily interpreted geometrically , cf .
figure [ fig : isbergr ] . as in the freudian
_ structural iceberg model _ , the constraints of * n * are communicated from the subconscious mind to the conscious ego , where they are experienced as desires . the understanding and
will are exercised _ preconsciously _ , in that although having deeper roots , they are supplied from within the conscious level .
the self cognizes the faculties of ego , mind and intellect through a mode of consciousness called _ witness consciousness _ ( _ skshi bhva _ ) .
this is subtler than and different from the agent s ego consciousness .
figure [ fig : isbergr ] depicts a simple geometric interpretation of these modes of awareness .
the basic information about the the above three elements are summarized in table [ tab:3 ] .
.the three structural elements of libertarian fw according to the fun model , their corresponding place of origin ( with a freudian near - equivalent term ) , the function of the elements , possible physical correlates in the brain , and ( in the terminology of yogic philosophy ) the corresponding character trait . [ cols="^,^,^,^,^",options="header " , ]
we develop a minimal mathematical model , denoted fun+ , to quantify the qualitative ideas developed above as part of the fun model . given an ap situation , and a choosing event @xmath24 , let the choice be represented by a random variable @xmath25 over the sample space @xmath26 , representing the possible @xmath27 choices that can be made .
if @xmath28 involves a material particle , which presumably lacks any libertarian fw , the outcome @xmath29 at event * e * would occur with probability @xmath30 , and there is no distortion of @xmath1 .
the dictates imposed by the constraints of nature * n * are represented , over @xmath31 , by the probability vector @xmath32 , normalized so that @xmath33 .
the vector could be pure ( @xmath34 ) or mixed .
the recommendation due to understanding * u * is represented , over @xmath31 , by the probability vector @xmath35 , which is normalized and could be pure or mixed .
for example , if @xmath36 , then we would have @xmath37 according to the moral criterion , whereas by nature , @xmath38 .
if @xmath39 , the understanding could use a health criterion to return @xmath40 , whereas it may be that @xmath41 .
fw * f * , as freedom to drive the choice from the prescriptive @xmath1 towards the recommended @xmath42 , is represented by the scalar quantity @xmath43 ( where @xmath44 ) such that greater freedom connotes larger @xmath43 .
the probability vector @xmath45 representing the eventual choice of an alternative , is obtained by distortion of @xmath1 towards @xmath42 in the measure of strength of @xmath43 . a particularly simple form to capture
this idea is a convex combination of @xmath1 and @xmath42 , parametrized by @xmath43 : @xmath46 thus the random variable @xmath47 representing the selected option is a weighted mean of random variables @xmath48 and @xmath49 , representing , respectively , understanding and nature . when @xmath50 ( maximal fw ) , @xmath51 , i.e. , the choice is aligned with the understanding .
when @xmath52 ( vanishing fw ) @xmath53 , i.e. , the choice is entirely determined by nature .
for example , if @xmath1 is the probability for obtaining an outcome 0 or 1 when measuring the pauli observable @xmath54 on a quantum two - level system ( qubit ) in the state @xmath55 , then @xmath56 . since the qubit arguably lacks fw ( @xmath52 ) , @xmath57 according to eq .
( [ eq : ccfw ] ) .
the nonlinear functional @xmath58 $ ] must be viewed as a set of ontic probabilities , and thus a fundamental limitation on the predictability of an agent s choice , and not epistemic probabilities arising from ignorance about the details of an agent .
the unpredictability @xmath59 of an agent may be quantified by the shannon entropy of @xmath60 : @xmath61 where @xmath62 is the shannon entropy of @xmath45 for a given value of @xmath43 .
the following two results quantitatively present our earlier resolution of sfwp . in each case
, we provide an example in the following two theorems , that contradicts each of implications * [ c ] * and * [ d ] * of swfp . the predictability of a saint s behavior does not imply his lack of fw .
[ thm : saint ] * proof .
* we construct an explicit instance of predictable behavior with ( high ) fw .
let @xmath63 . by definition ,
the saint of proposition [ pro : saint ] is ethical in his understanding , so that @xmath64 , and his will is free , so that @xmath65 .
even if he may bodily not be attuned to perfection , still by dint of his high fw , he is always able to choose in accord with his ( ethical ) understanding , i.e. , @xmath51 by eq .
( [ eq : ccfw ] ) . from eq .
( [ eq : unpred ] ) , we have @xmath66 bit , implying almost complete predictability , irrespective of @xmath1 .
@xmath67 this implies , as noted earlier , that our model is not incompatibilist .
more generally , we note that @xmath62 is not a monotonous function of @xmath43 .
the regime where @xmath68 and thus increase in fw leads to certainty , may be regarded as a zone of disagreement with the incompatibilist position . for example , given a situation involving two choices , with @xmath69 and @xmath70 and @xmath71 , @xmath72 for @xmath73 . the randomness of the conscientious criminal s choice does not imply his lack of fw .
[ thm : cc ] * proof .
* we construct an explicit instance of random behavior with non - vanishing fw . in the case of the conscentious criminal of proposition [ pro : cc ] , who has non - vanishing but
not maximal fw , let @xmath74 , and let @xmath75 , corresponding to the constraint imposed by his evil nature . from eq .
( [ eq : ccfw ] ) , we have @xmath76 , and hence @xmath77 , implying near maximal randomness in choice . however , he does not lack fw , as @xmath78 .
@xmath67 in the case of the hard - core criminal of proposition [ pro : hcc ] , we have the same @xmath31 , @xmath1 and @xmath42 as in the example used in theorem [ thm : cc ] , but @xmath79 , corresponding to this criminal s low fw . from eq .
( [ eq : ccfw ] ) , we have @xmath80 , and thus from eq .
( [ eq : unpred ] ) , @xmath81 , implying very little unpredictability in choice , similar in this respect to the saint of proposition [ pro : saint ] . however , unlike with the saint , whose predictability comes from high fw , here the predictability is due the criminal s sure surrender to his natural instinctive constraints . augmenting his fw
, we find that @xmath82 that is , increase in fw leads to increase in uncertainty ( when @xmath83 ) .
thus our model is also not compatibilist .
together , eqs .
( [ eq : sigless ] ) and ( [ eq : sigmore ] ) imply that our model is neither compatibilist nor incompatibilist .
the fun model implies that @xmath43 must be at least partially extra - physical and can not be described by any purely physical theory , even a theory - of - everything ( toe ) , that encompasses only physical phenomena . in particular , artificial intelligence ( ai ) , which is fundamentally built on ( quantum ) physical rules ,
can not capture true cognitive behavior .
these considerations entail that a human agent , and by extension any sentient agent , could not be considered merely as a sufficiently complex robot , but a qualitatively distinct class of entities . here
penrose s interesting thesis is worth noting , according to which conscious processes are fundamentally non - algorithmic @xcite .
wfwp shows that quantum indeterminism is not better for fw than classical determinism .
however , the fact remains that the world is fundamentally a quantum mechanical place .
further , there is in a sense a lesser departure from the physical dynamics @xmath11 through free - willed intervention if @xmath11 were indeterministic rather than deterministic , in that there is only a _
statistical _ violation of causality in the former case , rather than the _ logical _ violation of causality , as it is with the latter . accordingly ,
if @xmath2 is the random variable corresponding to a fw - influenced quantum measurement @xmath84 on system @xmath85 ( presumably a suitably small brain element ) in state @xmath86 , then @xmath2 would deviate from the born probability rule .
free - willed intervention will therefore manifest as _ statistical deviations from the born rule_. by contrast , the freedom or ` free will ' of quantum particles , which is plain quantum randomness , will conform to the born rule . in the model described below ( called fun++ ) , which is a particular realization of the fun+ model , we propose that fw intervenes by controling the collapse of the wave function . to begin with
, we represent the ap situation as a _ uniform _ quantum superposition , and the willful choice by a directed collapse of the wavefunction .
this quantum model thus presumes objective collapse of the wave - function , and is compatible with interpretations of quantum mechanics that admit it .
the exercise of fw is broadly divided into three stages as follows : attention .
: : faced with an ap situation of alternatives @xmath87 , the brain creates a quantum superposition that reflects the dictates of nature * n * : @xmath88 in a suitable subneuronal system @xmath85 in an appropriate basis @xmath89 , where the states @xmath90 correspond to the base choice space @xmath31 .
the main requirement is that it should be possible to shield @xmath85 indefinitely from decoherence or other noise , while the process of making a choice is under way . according to ref .
@xcite , brain microtubules may be the seat of such superpositions .
: : the ego surveys the recommendations of the mind and intellect .
this event is mathematically represented by the preparation of a nonlinear positive operator - valued measurement ( povm ) @xmath91 which satisfy the completeness condition @xmath92 . in the absence of fw , @xmath52 , and
the povm reduces to ordinary projective measurement . collapse . : : application of the povm to the state @xmath93 causes the transition : @xmath94 with probability @xmath95 .
the last stage in the model is tantamount to controlling and directing the wavefunction collapse to produce an outcome probability distribution that is closer ( as quantified by trace distance , relative entropy or any other suitable distance measure ) to @xmath96 than @xmath97 , if fw is available .
if fw is absent , then nature alone determines the outcome probability distribution , which , moreover , conforms to the born rule .
when @xmath98 , the resulting probabilities of outcomes will violate the born rule .
in general , such violations can give rise to various non - standard effects , among them violation of energy conservation , possibility of signaling at superluminal speed @xcite and the possibility of counter - intuitive models of computation that allow efficient solution of hard problems @xcite .
however , these non - standard effects will be confined to a small sub - cellular region of the brain , where it will not be easily discernible from measurment errors , decoherence effects , neural noise and statistical fluctuations .
moreover , it can be facilely masked behind the remaining features of the brain s physiology , which can be described in terms of deterministic , classical mechanisms ( e.g. , metabolic changes leading to arousal potentials in motor neurons , initiating familiar physical movements of the body ) .
finally , it is worth stressing that the extra - physical agency posited to determine the outcome of wavefunction collapse should not be confused with hidden variables in the sense of bohm @xcite .
the latter represents a device to turn qm into a deterministic theory , and thus restore classical causality , whereas in our approach requires , fw is a new form of causation .
in the conventional view , the brain is a complex input - output device , with apparently voluntary actions arising from a mix of deterministic and indeterministic mechanisms .
so it is an interesting question what would constitute a falsifiable proof of fw .
for example , in research reported in ref .
@xcite , tethered fruit flies ( drosophila melanogaster ) in a visually impoverished environment had their flight maneuvers recorded . lacking any external input ,
their random behavior should have resembled a poisson process .
however , the analysis showed that the temporal structure of fly behavior follows a lvy - like probabilistic behavior pattern . at first blush
, it seems possible that this may be due to spontaneous behavior arising from ( a primitive version of ) libertarian fw of the fruit flies .
however , it turns out that the observed pattern can be simulated via intrinsic noise amplified by suitable nonlinearity @xcite .
experiments that purport to _ disprove _ the existence of fw can also find alternate explanations . based on a study of volunteers wearing scalp electrodes , libet and collaborators @xcite showed in 1983 that a ` readiness potential ' ( rp ) was detected a few tenths of second before the subjects , in their own reckoning , made the decision to perform an action ( to flex a finger or wrist ) .
their result was interpreted as indicating that the motor cortex was preparing for the action , that unconscious neural processes predetermine actions and hence that fw was illusory . in a recent comment on the experiment , miller and
trevena @xcite asked subjects to wait for an audio tone before making a decision to tap a key or not . if the activity detected by libet et al . really was the making of the decision prior to any conscious awareness of doing so ,
then that activity ought to occur only if the subjects decided to act . but no such correlation was found .
miller and trevena conclude that the rp may only indicate that the brain is paying attention and does not indicate that a decision has been made . under the circumstances , one suggestion would be to trace backwards along a deterministic observable causal chain of neurons to the specific area in the motor cortex responsible , perhaps a single neuron , that may be considered as initiating a free - willed action at the physiological level , and hence a candidate site for carrying signatures of the putative new physics , such as the claimed deviation from the born rule .
identifying such single neurons will be difficult , given the high density of neuronal packing , and the weakness of excitatory synapses , which would make the role of individual neurons in the brain cortex hard to identify .
further , the extreme complicatedness of brain dynamics and attendent decoherence , measurement errors , noise and statistical fluctuations , will play an adverse role .
a study where the twitching of a mouse s whisker has been traced to single pyramidal neurons in the cortex is reported in ref .
these neurons could be examined for unusual behavior that might be considered as appropriate signatures , after suitable allowance has been made for noise and innervations from neighboring neurons : for example , there may be fluctuations in the whisker - twitch causing potentials that could somehow not be ascribed to any neighboring influences .
the neuron may then be examined carefully for a sub - cellular site to which the initiation of the twitch causing potentials may be attributed . another promising area to look for evidence suggestive of fw is psychiatry and neurology , which are startlingly revelatory on how , from a physiological perspective , the human mind works @xcite .
the fun # model suggests that the decision making ability in a human is affected , possibly leading to anti - social behavior , in broadly three _
different _ ways : when the limbic system , or the frontal cortex , or the pineal gland area ( or any other candidate ucm ) is impaired .
if in some case such behavior does not coincide with impairment to the limbic system or the frontal cortex , then there is a reasonable case to attribute the behavior to diminished fw .
studying such cases , if any , would help locate the ucm and verify whether the pineal gland hypothesis holds good .
there are potential therapeutic and medical applications based on the above observations .
the fun # model can help clarify what abnormalcy of behavior means and can help classify abnormal depending of which faculty is affected . if a particular behavioral disorder can be attributed to one of the three above causes , then the therapy prescribed to the afflicted patient could also be varied with better efficacy .
a conservative , darwinian view regarding the origin of fw would be that it enables an organism to deviate from the local gradient in the brain s energy landscape , and to better than locally optimize in the struggle for self - presevervation , which would require improvisation under novel situations ; once this was done , fw brought along with it potential by - products , like altruism , self - harm , etc . , behaviors which are not necessarily conducive to self - preservation , and hence not ` intended ' by the evolutionary process that gave rise to fw .
another possibility is that fw is an emergent phenomenon that is a product of self - organization in complex quantum physical systems .
these views are not necessarily incompatible with the fun model
. if anything , one can be adventurous and suggest this extra - physical agency guides or conduces to darwinian evolution to make the human physiology optimal for the physical experience ! for example , it is known that the higher the oxygen concentration ( up to a point ) , the better it is for the metabolism and life of organisms .
it is also known that the lower the altitude , the higher the oxygen concentration .
if there is fw , an organism in a mountainous area may feel the urge to explore farther by climbing higher , thereby improving its chances of reaching a valley that is deeper , due to which the oxygen concentration is higher , than in the valley reached by following the local gradient .
finally we offer our opinion on fw being , together with awareness , aspects of consciousness .
to be aware of a theory is to be able to talk about it and to understand its implications and limitations . to this end ,
one requires a meta - theory , which is equipped with a langauge in which to formalize propositions about the theory .
an axiomatic system that is rich enough to encompass the meta - theory will , in general , be more powerful than the one that axiomatizes the theory .
an instance of this is provided by gdel s theorem @xcite , which provides a metamathematical proof of mathematically undecidable propositions .
if fw did not exist , and the behavior of observers were entirely determined by the rules of the base theory , then the observers algorithmic complexity @xcite would not be greater than that of the theory , making them incapable of encompassing the meta - theory .
equipped with fw , human experimenters can be _ meta - entities _ in the theory .
it allows them to freely pose questions , perform tests and draw inferences about the theory as external agents whose choices are not entirely determined by the theory . from this perspective , fw is an aspect of the consciousness of sentient observers that is necessary to make them aware of a physical theory such as qm and to experience the ( quantum ) world as humans do .
99 t. oconnor .
free will .
free will .
l. zagzebski .
foreknowledge and free will .
wikipedia entry of free will . `
en.wikipedia.org/wiki/free_will ` s. esposito
. a personal point of view about scientific discussions on free will .
eprint arxiv:1202.3395 .
j. conway and s. kochen .
the free will theorem . found .
36 1441 ( 2006 ) .
arxiv : quant - ph/0604079 . j. conway and s. kochen .
the strong free will theorem .
math . soc .
56 , 2 ( 2008 ) ; arxiv:0807.3286 .
h. stapp . quantum theory and the role of mind in nature .
quant - ph/0805.0116 .
g. t hooft .
the free - will postulate in quantum mechanics .
quant - ph/0701097
. h. nikoli . closed timelike curves , superluminal signals , and `` free will '' in universal quantum mechanics .
arxiv:1006.0338 .
a. suarez .
quantum randomness can be controlled by free will- a consequence of the before - before experiment .
a. suarez .
the general free will theorem .
a. suarez .
the free will theorem and the flash ontology implicitly assume the before - before experiment and thereby imply free will .
comment on a note by nicolas gisin in arxiv:1002.1392v1 .
v. vedral , is the universe deterministic ?
new scientist , 18 november 2006 , p. 55 . n. gisin . the free will theorem , stochastic quantum dynamics and true becoming in relativistic quantum physics .
are there quantum effects coming from outside space - time ?
nonlocality , free will and no many - worlds .
quant - ph/1011.3440 .
m. j. w. hall .
local deterministic model of singlet state correlations based on relaxing measurement independence .
lett . * 105 * , 250404 , 2010 . s. hossenfelder
. the free will function .
. how free will could will .
http://arxiv.org/abs/1202.2007 .
m. mackay .
_ the clockwork image : a christian perspective on science _ inter - varsity press , london ( 1997 ) .
ber formal unentscheidbare stze der principia mathematica und verwandter systeme , i. monatshefte fr mathematik und physik 38 ( 1931 ) .
( various translations / commentaries are available on the web . )
a. turing .
on computable numbers , with an application to the entscheidungsproblem .
london math .
soc 42 , 230 ( 1936 ) .
j. d. barrow .
_ impossibility : the limits of science and the science of limits_. oxford university press , usa ( october 28 , 1999 ) .
r. srikanth .
srimadbhagavad gita in the new age ( 2012 ) : + http://poornaprajna.org/srik/gitopanishad d. hodgson .
constraint , empowerment , and guidance : a conjectural classification of laws of nature .
philosophy 76 , 341 - 70 ( 2001 ) . ` http://users.tpg.com.au/raeda/website/constraint.htm ` r. kane .
http://en.wikipedia.org/wiki/robert_kane_(philosopher ) s. hameroff and r. penrose . orchestrated objective reduction of quantum coherence in brain microtubules : the `` orch or '' model for consciousness .
+ http://www.quantumconsciousness.org/penrose-hameroff/orchor.html
s. m. mcclure , j. li , d. tomlin , k. s. cypert , l. m. montague and p. read montague .
neural correlates of behavioral preference for culturally familiar drinks .
neuron 44 , 379 ( 2004 ) .
h. b. puryear .
the edgar cayce primer : discovering the path to self - transformation ( bantam books , 1982 ) .
r. penrose .
shadows of the mind ( oxford university press , 1994 ) .
r. srikanth .
entanglement , intractability and no - signaling .
physica scripta 81 ( 2010 ) 065002 .
d. bohm and b. hiley .
the undivided universe : an ontological interpretation of quantum theory .
( london : routledge 1993 ) .
a. maye , c.- h. hsieh , g. sugihara and b. brembs .
order in spontaneous behavior .
plos one issue 5 , e443 ( may 2007 ) .
available freely online from www.plusone.org .
b. libet , c. a. gleason , e. w. wright and d. k. pearl .
time of conscious intention to act in relation to onset of cerebral activity ( readiness - potential ) : the unconscious initiation of a freely voluntary act .
brain ( 1983 ) 106 ( 3 ) : 623 - 642 .
doi : 10.1093/brain/106.3.623 . j. trevena and j. miller .
brain preparation before a voluntary action : evidence against unconscious movement initiation .
consciousness and cognition volume 19 , issue 1 , march 2010 , pages 447 - 456 .
m. brecht , m. schneider , b. sakmann , t. w. margrie .
whisker movements evoked by stimulation of single pyramidal cells in rat motor cortex .
nature 427 , 704 ( 2004 ) .
|
the basic question in the long - standing debate about free will ( fw ) is not whether fw can be demonstrated to exist nor even whether it exists , but instead how to define it scientifically .
if fw is not dismissed as an illusion nor identified with a variety of unpredictability , then logical paradoxes arise that make fw elusive to define .
we resolve these paradoxes through a model of fw , in which fw is a new causal primitive empowered to override physical causality under guidance .
we develop a simple mathematical realization of this model , that when applied to quantum theory , suggests that the exercise of fw corresponds to a nonlinear povm causing deviations from the born rule . in principle
, these deviations would stand in conflict with known conservation laws and invariance principles , implying that the brain , the presumed seat of fw , may be an arena of non - standard physics .
however , in practice it will be difficult to distinguish these deviations from quantum and neural noise , and statistical fluctuations .
we indicate possible neurobiological and neurological tests , implications and applications of our proposed model .
|
abell 2218 is a very rich galaxy cluster ( richness class 4 , according to abell et al .
1989 ) , characterized by a large velocity dispersion of the galaxy population ( @xmath1 km / s , le borgne et al .
1992 ) , and a high x - ray temperature and luminosity ( @xmath2 kev , see e.g. markevitch 1997 , @xmath3 erg / s km / s / mpc and @xmath4 in the energy band 0.54.4 kev , see e.g. kneib et al .
1996 ) . these properties , coupled to a relatively small distance ( @xmath5 , le borgne et al .
1992 ) , made this cluster an attractive target for studies of the sunyaev - zeldovich effect ( birkinshaw & hughes 1994 ) , and one of the closest clusters where gravitational arcs are detected .
the amazing concentration of gravitational arcs and arclets has stimulated a huge observational effort first to get sub - arcsec imaging ( from the ground , see , e.g. kneib et al .
1995 , and from space with hst , kneib et al .
1996 ) , and then to get spectra of the arcs ( ebbels et al .
1997 ) .
the optical and near - ir observations have allowed a very detailed modelling of the mass distribution within the cluster ( kneib et al .
1996 ) , confirmed later to a great level of accuracy , ( ebbels et al .
it was found that the cluster mass distribution is bi - modal , with the main concentration centred on the cd galaxy .
the x - ray surface brightness , as obtained via rosat observations , does not trace the gravitational potential as derived from the lensing analysis , a possible indication that the x - ray emitting gas is far from hydrostatic equilibrium ( kneib et al . 1996 ) .
this would also explain the discrepancy in the x - ray and lensing mass estimates ( markevitch 1997 ) .
in this paper we report on the mid - ir observations of abell 2218 done with isocam on - board esa s _ infrared space observatory _ ( iso ) satellite .
the high sensitivity of isocam allows us to determine the photometric properties of lensed and cluster galaxies in the mid - ir band , thus widening the wavelength coverage of their spectra .
mid - ir observations are critical in understanding the intrinsic nature of these distant galaxies , since a starburst is known to emit a large fraction of its energy in the mid - ir , due to dust grain re - processing of the emitted radiation .
abell 2218 was observed on march 3 , 1996 and again on february 28 , 1997 as part of an iso program of observations of lensing galaxy clusters with arcs and arclets : abell 370 , abell 2218 , cl2244 - 02 and ms2137 - 23 .
an important criterion for target selection was to choose the clusters with the brightest arcs in the optical , spatially extended and the farthest from the lens , due to the limited spatial resolution of iso .
+ the observations were performed by rastering in microscanning mode yielding to very deep isocam images of the galaxy cluster .
the pixel field - of - view was 3@xmath6 and the microscanning steps were 7@xmath6 in both directions of the 5x5 raster .
the cluster was imaged in the 4 filter bands , lw1 [ 4 - 5@xmath0 m ] , lw2 [ 5 - 8.5@xmath0 m ] , lw7 [ 8.5 - 11@xmath0 m ] , and lw3 [ 12 - 18@xmath0 m ] , at 5 or 10 sec integration times for a total observing time of 5.5 hours , covering a field of more than 2x2 arcminutes .
the data reduction was done partly within the idl based ( iso)cam interactive analysis ( cia ) package ( ott et al . , 1997 ) and partly through the usage of
c++ based multi - resolution median transform ( _ wavelet _ ) techniques ( starck j .- l .
et al , 1997 ) .
cross - correlation of the two methods allows us to increase the reliability of source detection .
the lw1 , lw2 and lw3 maps are presented in figures 1 & 2a , as overlays on top of the hst image ( courtesy j .-
kneib ) of abell 2218 . in the cluster core
, no arclet is detected at a significant level at any wavelength in the mid - ir .
this non - detection of arcs or arclets contrasts with our recent mid - ir imaging of abell 370 ( metcalfe et al .
1997 ) , where the a0 giant arc is clearly detected as the main feature and emitter in the cluster core at 15@xmath0 m .
but this giant arc is already a prominent feature in the optical , ( first giant arc discovered ) , whereas arcs and arclets are fainter in abell 2218 .
+ the cd galaxy which is apparently centred on the cluster potential is clearly detected in the lw1 & lw2 filter bands . at 4.5@xmath0 m ( rest - frame wavelength 3.8 @xmath0 m )
the cd emission is extended , covering part of the optical halo ; at 7@xmath0 m the emission is much more confined to the centre , but with an extension along the optical major axis , probably contaminated by the neighbouring merging dwarf galaxies . at 10@xmath0 m and 15@xmath0 m
it lies just above the detection limit of isocam and no statement can be made on its extension .
the mid - ir spectral energy distribution up to 15@xmath0 m seems to follow a simple rayleigh - jeans tail of the cold stellar component as found in optically - selected normal early - type galaxies ( e , s0 , s0a ) in the virgo cluster ( boselli et al .
1998 ) . but due to the contamination by close - by galaxies it is very difficult to estimate the 4.5@xmath0 m flux of the cd .
+ the brightest cluster member galaxies are also detected at 4.5 @xmath0 m and 7 @xmath0 m . at the redshift of the cluster the 4.5 @xmath0 m emission is mostly coming again from cold stellar photospheres .
the 7 @xmath0 m ( lw2 band ) ( rest - frame 5.7 @xmath0 m ) emission includes a small fraction of pah emission but comes mainly again from the cold stellar photospheres . at 10 & 15 microns
most of cluster members have vanished ( ie . are below or very close to our detection limit ) , as expected for normal early - type galaxies ( ellipticals ) where there is only the rayleigh - jeans tail mid - ir emission from cold stellar photospheres , in dust & gas poor cluster core ellipticals .
+ still one cluster member galaxy # 373 ( numbering system from le borgne et al .
1992 in the following ) is detected at 15@xmath0 m , but also at 4.5 @xmath0 m and 7 @xmath0 m , it has a cart - wheel like aspect , from the ring surrounding it and is probably a big face - on spiral galaxy .
the 4.5@xmath0 m and 7@xmath0 m emission comes mostly from the nucleus whereas the 15@xmath0 m originates apparently from the ring structure .
it could be pah and small hot grain emission enhanced by star formation in tidal shocks in the ring. object # 323 is detected at 15 @xmath0 m .
this object was suspected to be an arclet candidate in the first deep multi - filter ground - based imaging ( pello et al .
1992 ) , since it appeared as a fine unresolved extended structure , together with another 32 arclets in total .
it is classified as one of the 235 arclet candidates in the first hst observation of abell 2218 ( kneib et al .
however , as noted by kneib et al . , it was also suspected not to be a strongly lensed object when comparison is made with shear orientation .
this was again confirmed by ebbels et al .
( 1997 ) , object # 323 being very elongated but at an angle of 45 degrees from the shear direction .
identification of several absorption features in its spectrum reveal it to be a cluster member at a redshift of z=0.179 , a spiral seen edge - on . to our surprise
the brightest source by far from 7@xmath0 m to 15@xmath0 m is object # 395 , an apparently insignificant z=0.1032 ( le borgne 1992 ) foreground sb galaxy , showing an ultraviolet excess and a strong h@xmath7 emission line , both indicating strong star formation of massive stars .
most of this mid - ir emission could then originate from the reprocessing by the small hot dust heated by a strong uv radiation field .
the second strongest emitter at long wavelength ( from 10@xmath0 m to 15@xmath0 m ) , but much less at shorter wavelengths is object # 317 , a lensed galaxy at z=0.474 ( ebbels et al .
1997 ) , a rather red object from the visible to the mid - ir .
first indications show that this type of lensed object is common on our larger field imaging towards lensing clusters ( metcalfe et al .
1998 in preparation ) birkinshaw m. , hughes j.p .
1994 , 420 , 33 boselli a. , lequeux j. , sauvage m. et al .
, in press , 1998 .
ebbels t. et al . 1997 , astro - ph/9703169 , in press kneib j .-
, mellier y. , pell r. et al .
1995 , 303 , 27 kneib j .-
, ellis r.s .
, smail i. , sharples r.m .
1996 , 471 , 643 le borgne j .- f . ,
pell r. , sanahuja b. 1992 , 95 , 87 markevitch m. 1997 , astro - ph/9704106 , in press metcalfe et al .
1997 , to appear in 1997 moriond conf .
mamon g. ed .
pell r. , le borgne j.f .
, sanahuja b. , mathez g. , fort b. , 266 , 6 ott s. et al .
1997 , in hunt g. , payne ( eds ) asp conf . series vol .
125 starck j .- l .
et al . , `` imaging processing and data analysis in the physical sciences : the multiscale approach '' , cambridge univ
|
we report on the mid - infrared imaging at 5 , 7 , 10 & 15 @xmath0 m of the galaxy cluster abell 2218 obtained with the isocam instrument onboard esa s infrared space observatory ( iso ) , as part of an on - going program to image gravitational arcs and arclets in distant clusters .
several cluster galaxies as well as field galaxies are detected .
we discuss their mid - ir flux properties .
|
fast nuclear gas outflows appear to be a relatively common phenomena in active galactic nuclei ( agns ) . they have been detected - using optical , uv and x - ray observations - in seyfert galaxies ( see for some examples crenshaw et al .
2000 ; aoki et al . 1996 ; capetti et al . 1999 and
refs therein ) , quasars ( see e.g. turnshek 1986 , krongold et al.2003 and refs therein ) , but also in starburst galaxies ( veilleux et al .
2002 , 2003 ) .
these gas outflows can have different origins , such as being jet - driven ( if the object is radio loud ) or related to starburst and agn winds .
why should we expect gas outflows also in radio galaxies ?
in these objects all the potential drivers for powerful outflows ( a highly luminous ionising agn , relativistic jets and , often , a starburst ) are present to varying degrees . in particular , the activity - inducing processes ( e.g. merger , interaction ) are believed to involve the injection of substantial amounts of gas / dust into the central nuclear regions .
thus , we can expect that in the first phase ( or in a re - started phase ) of activity , strong interactions with the relatively dense medium will occur .
indeed , gas outflows are now increasingly detected in radio galaxies and here we present a summary of the results we have obtained so far in our study of both the ionized , as well as the neutral , components of such outflows .
the first clear example of an outflow of ionized gas that we have found is in the southern _ starburst _ radio galaxy pks 154979 ( i.e. a radio galaxy that spectroscopically shows a young stellar population component in addition to the old stellar component typical of elliptical galaxies ) . in pks 154979 , different redshift systems associated with both
the low- and high - ionisation emission lines were found ( tadhunter et al .
the high ionisation lines are broader , with fwhm@xmath2 @xmath0km@xmath0s@xmath1 while the low - ionisation lines are narrower .
the absorption detected in this object ( morganti et al .
2001 ) has the same redshift as the low ionisation gas , while the high ionisation gas appears blueshifted with respect to it .
the absorption and the low ionisation gas are associated with a cocoon of material surrounding the tiny ( @xmath3 pc in size ) radio source . on the other hand ,
the high ionisation ( and blueshifted ) gas is believed to be associated with an outflow from a region close to the radio jet and disturbed by it . taken together , all these results are consistent with the idea that pks 154979 is a young radio source in which the cocoon of debris left over from the event that triggered the activity has not yet been swept aside by circumnuclear outflows .
more recently , bellamy et al . (
2003 ) have detected broad pa@xmath4 in pks 154979 confirming that this object indeed contains a quasar nucleus that is moderately extinguished , despite evidence that its radio jet points close to our line - of - sight .
an even more extreme case of gas outflow has been found in the radio galaxy 4c 12.50 ( pks 1345 + 12 ) . this is a particularly interesting object as it is a prime candidate for the link between ultraluminous infrared galaxies ( ulirgs , sanders & mirabel 1996 ) and radio galaxies ( evans et al .
the radio source is confined to a region @xmath5 arcsec ( @xmath6 pc ) and has all the characteristics of young radio sources ( @xmath7 yr ) .
the ism of this radio galaxy is extremely rich : it is the brightest far - ir radio galaxy and has a high molecular gas mass ( evans et al .
long slit spectra were taken ( using the wht , see holt et al .
2003 for details ) in order to investigate the impact of the nuclear activity on the circumnuclear interstellar medium .
the spectra show extended line emission up to @xmath8 kpc from the nucleus , consistent with the presence of an asymmetric halo of diffuse emission as observed in optical and infrared images . at the position of the nucleus ,
complex emission line profiles are observed and gaussian fits to the [ oiii ] emission lines ( see fig . 1 _ left _ ) require three components ( narrow , intermediate and broad ) , the broadest of which has fwhm@xmath9 @xmath0km@xmath0s@xmath1 and is blueshifted by @xmath9 @xmath0km@xmath0s@xmath1 with respect to the halo of the galaxy and the deep and narrow absorption .
this component is interpreted as material in outflow .
a large reddening and high density ( @xmath10 4200 @xmath11 ) has also been found for the most kinematically disturbed component .
4c 12.50 , like pks 1549 - 79 , appears , therefore , to be a young radio galaxy with nuclear regions that are still enshrouded in a dense cocoon of gas and dust .
the radio jets are now expanding through this cocoon , sweeping material out of the nuclear regions ( holt et al .
extremely intriguing is the discovery of a number of radio galaxies where the presence of fast outflows is associated not only with ionized but also with _ neutral _ gas .
this finding gives new and important insights on the physical conditions of the gaseous medium around an agn .
the best example found so far of this phenomenon is the radio galaxy 3c 293 ( morganti et al .
2003a ) . in this galaxy
, very broad absorption has been detected against the central regions - this is in addition to the known and much narrower absorption ; haschick & baan 1985 , beswick et al .
the absorption profile , obtained using the new broad band ( 20 mhz ) system available at the westerbork synthesis radio telescope , has a full width at zero intensity of about 1400 @xmath0km@xmath0s@xmath1 and most of this broad absorption ( @xmath12 @xmath0km@xmath0s@xmath1 ) is blueshifted relative to the systemic velocity .
the broad absorption is shallow ( the optical depth is only @xmath13 % ) and corresponds to a column density of the of @xmath14 k @xmath15 .
this is likely to be a lower limit to the true column density as the @xmath16 associated with such a fast outflow can be as large as a few 1000 k ( instead of 100 k which is more typical of the cold , quiescent in galaxy disks ) .
this absorption represents a fast outflow of _ neutral _ gas from the central regions of this agn .
new optical spectra ( see fig . 2b , from emonts et al . in prep .
) show that the optical emission lines also contain a broad component that is very similar to the broad absorption .
this suggests that the and ionized gas outflows may be coming from the same physical region .
the radio galaxy 4c 12.50 also shows broad absorption ( shown in fig .
1b ) when observed using a broad radio band .
the absorption appears complex and _ extremely broad_. the full range of velocities covered by the absorption is @xmath9 @xmath0km@xmath0s@xmath1 , _ the broadest detected so far in _ .
the peak optical depth of the broad component is only @xmath17 and the column density of the full system of shallow absorption ( assuming a covering factor is 1 ) is @xmath18 @xmath15 . the central question is how _ neutral _ gas can be associated with such fast outflows .
as discussed in morganti et al .
( 2003a ) for 3c 293 , a number of possible hypotheses can be made about the origin of such outflows .
they include , e.g. , starburst winds ( veilleux et al . 2002 ) , radiation pressure from the agn ( dopita et al .
2002 ) , jet - driven outflow and adiabatically expanding broad emission line clouds ( elvis et al .
2002 ) .
the model that we favour is jet - driven outflow , although we can not rule out that one of the other mechanisms is also at work at some level .
this model assumes that the radio plasma jet hits a ( molecular ) cloud in the ism . as a consequence of this interaction ,
the kinematics of the gas are disturbed by the shocks and the gas is ionized by it . once the shock has passed , part of the gas may have the chance to recombine and become neutral while it is moving at high velocities .
one problem with this model ( considered in detail by mellema et al .
2002 ) is that it may not be possible to accelerate the clouds of gas to the high velocities that we observe , as indicated by e.g. the simulations presented by van breugel ( these proceedings ) . to understand which is the more likely mechanism ,
we now need high spatial resolution observations to find the exact position of the absorption .
so far this information is available only for the seyfert galaxy ic 5063 where it is providing evidence that the interaction between the radio jet and a molecular cloud is indeed the cause of the gas outflow .
the southern seyfert galaxy ic 5063 was the first agn where a fast outflow of neutral hydrogen was detected ( morganti et al .
vlbi observations have confirmed that the broad absorption is detected against the bright radio lobe ( i.e. _ not _ against the nucleus ) situated at @xmath19 kpc from the nucleus , as shown in fig .
3a ( see also oosterloo et al .
2000 ) .
deep optical spectra have been taken with the eso - ntt to compare the kinematics of the ionized gas with that of the neutral hydrogen component in order to study the mechanism that could drive this outflow .
preliminary results are presented in morganti et al .
( 2003b ) .
the data reveal extremely complex gas kinematics as can be seen in fig .
this includes the presence of an outflow of ionised gas ( with velocities of several hundred km / s ) at the location of the brighter radio lobe .
the amplitude of this outflow is strikingly similar to that of the .
because of its location , we consider the interaction between the radio jet and the ism to be the most likely mechanism for the extreme kinematics .
our observations show that gas outflows of both ionized gas and neutral hydrogen are found in some radio galaxies .
so far the indications are that these outflows are present in objects that are either in the early - stage of their evolution ( like 4c 12.50 ) or , perhaps , in a phase of re - started activity ( as we think it is the case for 3c 293 ) . moreover ,
another characteristic that appears to be common among these objects is the presence of a young stellar population component ( tadhunter et al .
+ although it is difficult to attribute the outflow to a starburst wind ( as the young stellar population is usually relatively old , 0.5 - 2 gyr ) , the presence of such a young stellar component could be an indication that the galaxy is indeed in a particular stage of its evolution , where large amounts of gas / dust - likely from the merger that triggered the activity - are still present in the inner region and the radio jet is strongly interacting with it . +
a more systematic search for fast gas outflows in radio galaxies is now in progress .
aoki k. , ohtani h. , yoshida m. , kosugi g. 1996 , aj , 111 , 140 bellamy m.j .
, tadhunter c.n .
, morganti r. , wills k.a .
, holt j. , taylor m.d . , watson c.a .
2003 , mnras in press ( astro - ph/0308345 ) beswick r.j .
, pedlar a. , holloway a.j .
2002 , mnras 329 , 620 capetti et al . 1999 , apj 516 , 187 crenshaw d.m .
2000 , aj 120 , 1731 dopita m.a .
2002 , apj 572 , 753 elvis m. , marengo & karovska , 2002 , apj 576 , l106 ( astro - ph/0202002 ) evans a.s .
et al . 1999 ,
apj 521 , l107 haschick a.d . &
1985 , apj 289 , 574 holt j. , tadhunter c. , morganti r. , 2003 mnras in press ( astro - ph/0302311 ) krongold et al . 2003
( astro - ph/0306460 ) mellema g. , kurk j.d .
, rttgering h.j.a .
2002 , a&a 395 , l13 morganti r. , oosterloo t.a . ,
emonts b.h.c . , van der hulst j.m .
, tadhunter c.n .
2003a , apjletter 593 , l69 morganti r. , oosterloo t. , holt j. , tadhunter c. , van der hulst j.m . , 2003b eso messanger 113 , 69 ( http://www.eso.org/gen-fac/pubs/messenger/ ) oosterloo t.a . , morganti r. , tzioumis a. , reynolds j. , king e. , mcculloch p. , tsvetanov z. 2000 , aj 119 , 2085 morganti et al .
2001 , mnras 323 , 331 morganti r. , oosterloo t. , tsvetanov z. 1998 , aj , 115 , 915 sanders d.b .
& mirabel i.f .
1996 , ara&a 34 , 749 tadhunter c.n .
, wills k.a .
, morganti r. , oosterloo t. , dickson r. 2001 mnras 327 , 227 tadhunter et al .
2003 mnras , submitted turnshek d.a .
1986 , in _ quasars _ , swarup g. , kapahi v.k .
dordrecht , d. reidel publishing co. , 1986 , p. 317
veilleux et al .
2002 in _ emission lines from jet flows _ , eds .
henney et al .
revmex aa 13 , 222 ( astro - ph/0101075 ) veilleux et al .
2003 , in _ `` neutral ism in starburst galaxies '' _ , eds .
aalto et al .
, pasp in press ( astro - ph/0309119 )
|
we present a summary of our recent results on gas outflows in radio galaxies .
fast outflows ( up to 2000 @xmath0km@xmath0s@xmath1 ) have been detected both in ionized and neutral gas .
the latter is particularly surprising as it shows that , despite the extremely energetic phenomena occurring near an agn , some of the outflowing gas remains , or becomes again , neutral .
these results are giving new and important insights on the physical conditions of the gaseous medium around an agn .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in
|
a radio labeling of a graph is a generalization of vertex coloring . given a graph @xmath1 ( for us
, @xmath1 is simple and connected ) with vertex set @xmath9 , any function @xmath3 is a labeling of the vertices .
if that labeling satisfies the inequality @xmath10 for all distinct vertices @xmath11 , then @xmath12 is a coloring of @xmath1 .
this inequality is exactly the condition needed to guarantee that adjacent vertices are labeled with a different color " .
we can generalize this notion of graph coloring by changing the inequality to the following : @xmath13 for some @xmath14 , @xmath15 . a labeling satisfying this inequality for all vertices
@xmath16 is called a @xmath2-radio labeling @xcite . in particular , it guarantees that any two vertices that are @xmath2 apart or less in distance have different labelings ( which is the reason that @xmath2 is bounded above by @xmath17 ) .
when @xmath8 , we call a @xmath2-radio labeling simply a radio labeling , and we call inequality the radio condition
. radio labeling is connected to the channel assignment problem , which aims to assign frequencies ( channels ) to radio transmitters in a way that minimizes interference .
so , if two transmitters are geographically close , it is optimal to have their frequencies be far apart .
this directly relates to the idea of radio labeling of a graph by thinking of the vertices as radio transmitters and the labels as radio frequencies ( hence the name ) @xcite , @xcite , @xcite .
just as with vertex coloring , we are interested in minimizing the largest label given to a vertex . for any particular labeling , the largest label used is called the span of the labeling . in vertex coloring , the minimum possible span of a graph
is called its chromatic number . in radio labeling , the minimum span
is called the radio number , denoted @xmath18 . finding a closed formula for the radio numbers of different types of graphs is the goal , but it remains unknown for all but a handful of them @xcite , @xcite , @xcite , @xcite , @xcite .
even the discrete problem of finding the radio number of a specific example @xmath1 can be difficult because of the size of the computation .
one way of finding the radio number involves analyzing all possible permutations of the vertices .
given a radio labeling @xmath12 of a graph with vertices @xmath19 , we can rename ( order ) the vertices as @xmath20 so that @xmath21 if and only if @xmath22 . on the other hand , we can produce a radio labeling of a graph by picking an ordering of the vertices and then forcing the labels to increase with respect to that order . since we are interested in minimizing span , we always choose the smallest label that works at the given moment .
therefore , an ordering of the vertices uniquely determines a radio labeling if we enforce this increasing condition .
it is this radio labeling we refer to when we speak of the radio labeling `` induced '' by the ordering .
so , one brute force way of computing the radio number of a graph is to consider all radio labelings induced from all possible orderings ( permutations ) of the vertices , and then choose the smallest span of all of them .
even after an ordering is fixed , producing the induced radio labeling is nontrivial , so producing @xmath23 labelings is computationally very intense . unlike all other @xmath2-radio labelings of a graph ,
if @xmath8 , then no two distinct vertices can have the same label ; this tells us that @xmath24 .
if @xmath25 , then there is a radio labeling of @xmath1 of consecutive integers @xmath26 .
such a labeling is called a consecutive radio labeling .
the complete graph @xmath27 of @xmath28 vertices is an example of a graph that has such a labeling .
since @xmath29 , the radio condition is trivially satisfied , so any labeling with consecutive integers will work .
a more interesting example is that of the peterson graph ( figure [ fig : petersongraph ] ) .
its diameter is 2 , so the radio condition is not satisfied for all possible consecutive labelings , but it does have a consecutive radio labeling . the qualities that the peterson graph possesses that allow it to have a consecutive radio labeling can be boiled down to the following : it is a diameter 2 graph that has an ordering of its vertices such that each vertex is diameter away from the next vertex in the ordering .
the existence of such an ordering of the vertices is a rare property for a graph to have , but this alone will not always be sufficient to produce a graph with a consecutive radio labeling .
it is only because we also had a graph of diameter 2 in this case that the qualifications were so simple . in general , for a graph @xmath1 with vertices
@xmath30 , we would need an ordering of the vertices @xmath31 such that @xmath32 this gives an idea of how special graphs that have consecutive radio labelings are , and how difficult it can be to produce examples of them , particularly examples of high diameter . a tool we will be using in our construction is the cartesian product of graphs .
given two graphs @xmath1 and @xmath33 , their cartesian product , denoted @xmath34 , is a graph defined by the following : * the vertex set @xmath35 is given by @xmath36 .
* two vertices @xmath37 are adjacent if the vertices in one coordinate are the same , and the vertices in the other coordinate are adjacent in their respective original graph .
that is , if * * @xmath38 and @xmath5 is adjacent to @xmath39 in @xmath33 , or * * @xmath40 and @xmath4 is adjacent to @xmath41 in @xmath1 .
see figure [ fig : k3boxp3 ] for an example . ]
one very nice property of the cartesian product of graphs is that distance is inherited from the original graphs : @xmath42 this allows us to learn about @xmath43 by looking at @xmath18 and @xmath44 .
we will be considering the cartesian product of @xmath45 copies of a graph @xmath1 , denoted @xmath46 : @xmath47 vertices of @xmath46 can be represented as @xmath45-tuples with entries in @xmath9 , and there is an edge between two vertices @xmath48 and @xmath49 if , for some @xmath50 , @xmath51 is adjacent to @xmath52 in @xmath1 , and @xmath53 for all @xmath54
. the graph of @xmath46 can be quite complicated , even if @xmath1 itself is simple .
( the number of vertices of @xmath46 is @xmath55 and , if @xmath56 denotes the set of edges in @xmath1 , then @xmath46 has @xmath57 edges . )
distance is given by @xmath58 diameter is then easily computed : @xmath59 .
when @xmath1 has a consecutive radio labeling , we will be looking at @xmath46 for new examples of graphs with such a labeling .
since @xmath60 increases as @xmath45 does , we have some hope of finding high diameter examples by looking here .
actually , one example is well known .
we have already seen that the peterson graph ( which we will hereafter denote @xmath61 ) has a consecutive radio labeling ( figure [ fig : petersongraph ] ) .
a consecutive radio labeling can also be exhibited for the diameter 4 graph @xmath62 .
this finding is exciting for the cause for finding higher diameter examples , and it can spark some ( very optimistic ) conjectures .
for example , it is tempting to hope that @xmath63 has a consecutive radio labeling for all @xmath64 .
this is desirable because if @xmath45 is unbounded above then so is @xmath65 .
it is nt true , however , as we will see in @xmath66[highpowers ] .
nevertheless , we were able to prove an even better result in the following section .
better because it produces a family of examples with consecutive radio labelings in which there is an example for every possible diameter ( whereas @xmath63 has even diameter ) .
we construct a family of examples of arbitrary diameter that have consecutive radio labelings by taking cartesian powers of the complete graph @xmath27
. we will begin by giving an ordering and @xmath67 , then this recursive definition is an ordering on @xmath68 : @xmath69 where we use the convention that @xmath70 . ] of the vertices of @xmath71 , and then prove that this ordering induces a consecutive radio labeling .
let @xmath73 be the vertices of @xmath27 , @xmath74 .
consider @xmath71 where @xmath75 .
we describe the desired ordering of the vertices of @xmath71 in groups of @xmath28 vertices at a time .
the first @xmath28 vertices are the @xmath45-tuples @xmath76 for the sake of easier reference , we will think of these @xmath28 vertices as an @xmath77 matrix @xmath78 where @xmath79 is the @xmath80th coordinate of @xmath81 .
that is , @xmath82.\ ] ] let @xmath83 be the @xmath28-cycle @xmath84 .
we produce a second matrix @xmath85 where @xmath86 we then form subsequent matrices by the following rule . to produce @xmath87 , we first determine @xmath88 , the largest integer such that @xmath89 . then @xmath87 is a matrix made up of entries @xmath90 using this definition we create @xmath91 matrices of dimensions @xmath77 : @xmath92 .
each matrix @xmath87 is identical to the one that came before ( @xmath93 ) except for a single column which differs by an application of @xmath94 .
the @xmath28 rows of each matrix ( which total @xmath95 rows over all @xmath91 matrices ) correspond to vertices of @xmath71 , and tell us our ordering , @xmath28 vertices at a time . formally ,
if @xmath96 , @xmath97 , then @xmath98 the @xmath99th row of matrix @xmath100 . our first goal is to show that this definition of @xmath102 is actually an ordering of the vertices of @xmath71 .
certainly @xmath103 . this coupled with the fact that @xmath104 means we only need to show that our proposed ordering has no repetition : @xmath105 for all @xmath106 .
we can make this task a bit easier by noticing some structure of the @xmath87 s .
@xmath78 has a structure that is inherited by @xmath87 for all @xmath2 : @xmath107 .
therefore , any row of @xmath87 determines all of @xmath87 . for this reason
, it suffices to show that no two matrices have the same first row .
since we are narrowing our scope to the first rows of each matrix , let s make a new matrix @xmath108 comprised of them .
let @xmath108 be the @xmath109 matrix defined by @xmath110 .
we can give an equivalent definition by using the definition of @xmath87 .
to define @xmath111 , let @xmath112 be the largest integer such that @xmath113 .
then @xmath114 so now we show that @xmath108 has no identical rows .
we will start by studying the specific and repetitive structure of @xmath108 . to do this
, we break up each column into `` blocks '' what we ll call a certain collection of column vectors that , when adjoined , produce the original column .
fix a column @xmath80 .
we wish to break up the column into groups of @xmath115 entries which will form the blocks .
so , the vectors @xmath116,\left[\begin{array}{c } a_{n^{t - j}+1,j}\\ \vdots\\ a_{2n^{t - j},j } \end{array}\right],\dots,\left[\begin{array}{c } a_{(n^{j}-1)n^{t - j}+1,j}\\ \vdots\\ a_{n^{t},j } \end{array}\right]\ ] ] are the blocks of column @xmath80 .
note that the blocks are of uniform size , but that size is dependent on the column .
the first column of @xmath108 has only one block , while in the last column of @xmath108 each entry is itself a block . in order to easily reference
which column a block comes from , we will commonly call a block from column @xmath80 a @xmath80-block . in summary : a * j - block * is any one of the vectors listed in line .
we will call the set of rows that are associated to the entries of a block the scope of the block .
the * scope * of a @xmath80-block @xmath117 $ ] is the set with rows @xmath118 through @xmath119 of @xmath108 as its elements : @xmath120_{1\leq h\leq t},\dots,[a_{(c+1)n^{t - j},h}]_{1\leq h\leq n}\}$ ] .
the scope of multiple @xmath80-blocks is the union of the scopes of the individual @xmath80-blocks .
[ blockshaveidenticalentries ] given a @xmath80-block of @xmath108 that has entries @xmath121 , then @xmath122 . based on the definition of @xmath108 , @xmath123 unless @xmath124 where @xmath112 is the largest integer such that @xmath125 .
equivalently , @xmath123 unless @xmath126 .
the only way to get any distinct entries in the block is for one of the entries ( excluding the first ) to be the image under @xmath94 of the previous entry .
so the question is , is it possible for @xmath126 when @xmath127 ?
one condition for @xmath126 is that @xmath128 .
it s clear that this can not be the case for any of the possibilities for @xmath129 : @xmath130 therefore , since @xmath131 for all @xmath127 , each entry of the vector must be the same as the first entry @xmath132 . as a result of this claim , along with the definition of @xmath108 , we know that adjacent blocks in a column can only have two possible relationships : either they are identical , or the second block has entries which are the image under @xmath94 of the first block s entries .
we d like to know when adjacent blocks are identical .
[ blockrepetition ] the adjacent blocks in column @xmath80 of @xmath108 @xmath133 , \left[\begin{array}{c } a_{(c+1)n^{t - j}+1,j}\\ \vdots\\ a_{(c+2)n^{t - j},j } \end{array}\right ] $ ] are identical if and only if @xmath134 divides @xmath28 .
it follows from claim [ blockshaveidenticalentries ] that the blocks are identical if and only if @xmath135 .
we know from our definition of @xmath108 that this happens only when @xmath131 , where @xmath112 is computed for @xmath136 .
so what is @xmath112 in this case ?
certainly @xmath137 is a candidate since @xmath138 .
in fact , @xmath139 if and only if @xmath140 , which is equivalent to requiring that @xmath134 divides @xmath28 .
the blocks have now given us a really detailed picture of what s happening with our matrix @xmath108 .
if we look down an arbitrary column @xmath80 , the column is partitioned into @xmath141 blocks , each with @xmath115 entries .
the entries of any fixed block are identical .
the first block in each column has all entries of @xmath142 .
when changes do happen down a column , they change according to @xmath94 ; so @xmath142 can be followed by @xmath143 then @xmath144 and so on .
and we now know by claim [ blockrepetition ] exactly when those changes will not occur due to the repetition of adjacent blocks
. now we want to examine how the structure of the different columns relate .
it happens that the scope of a ( @xmath145)-block is equal to the scope of a collection of @xmath80-blocks the appropriate @xmath28 of them in a row .
this is immediate because of the size of the blocks .
a ( @xmath145 ) block has @xmath146 entries , and @xmath80-blocks have @xmath115 entries .
so , for every block in column @xmath145 , there are @xmath147 blocks in column @xmath80 .
the interesting thing is that those @xmath28 @xmath80-blocks with the same scope as the single ( @xmath145)-block must be distinct ( in the sense that no two of these @xmath80-blocks can have the same entries ) .
we will prove this in the following claim .
note that the blocks listed in the claim are in column @xmath80 , of the correct size ( @xmath115 entries ) , they are adjacent , and there are @xmath28 of them .
also note that they cover rows @xmath148 through @xmath149 .
this is significant because @xmath150 $ ] is a ( @xmath145)-block .
[ distinctblocks ] the following @xmath80-blocks in @xmath108 are all distinct : @xmath151,\left[\begin{array}{c } a_{bn^{t - j+1}+n^{t - j}+1,j}\\ \vdots\\ a_{bn^{t - j+1}+2n^{t - j},j } \end{array}\right],\dots,\left[\begin{array}{c } a_{bn^{t - j+1}+(n-1)n^{t - j}+1,j}\\ \vdots\\ a_{(b+1)n^{t - j+1},j } \end{array}\right ] .\ ] ] let s begin by proving that the first two @xmath80-blocks @xmath152,\left[\begin{array}{c } a_{bn^{t - j+1}+n^{t - j}+1,j}\\ \vdots\\ a_{bn^{t - j+1}+2n^{t - j},j } \end{array}\right]$ ] are distinct . in order to easily reference claim [ blockrepetition ] , we rewrite the indices of the entries in the form that appears there : @xmath151=\left[\begin{array}{c } a_{(bn)n^{t - j}+1,j}\\ \vdots\\ a_{(bn+1)n^{t - j},j } \end{array}\right],\left[\begin{array}{c } a_{bn^{t - j+1}+n^{t - j}+1,j}\\ \vdots\\ a_{bn^{t - j+1}+2n^{t - j},j } \end{array}\right]= \left[\begin{array}{c } a_{(bn+1)n^{t - j+1}+1,j}\\ \vdots\\ a_{(bn+2)n^{t - j+1},j } \end{array}\right].\ ] ] now we are already done because ( by claim [ blockrepetition ] ) showing these two blocks are distinct is equivalent to showing @xmath28 does not divide @xmath153 , which is immediate . with the same argument we can show that , in this collection of @xmath28 blocks , adjacent blocks are distinct .
we proceed by similarly rearranging the remaining indicies to be compatible with the previous claim .
@xmath154&=\left[\begin{array}{c } a_{(bn+2)n^{t - j}+1,j}\\ \vdots\\ a_{(bn+3)n^{t - j},j } \end{array}\right]\\ & \
\vdots\\ \left[\begin{array}{c } a_{bn^{t - j+1}+(n-1)n^{t - j}+1,j}\\ \vdots\\ a_{(b+1)n^{t - j+1},j } \end{array}\right]&=\left[\begin{array}{c } a_{(bn+n-1)n^{t - j}+1,j}\\ \vdots\\ a_{(bn+n)n^{t - j},j } \end{array}\right].\end{aligned}\ ] ] showing that adjacent blocks are distinct is equivalent to showing that @xmath28 does not divide @xmath155 , which is again immediate . because of this we can now say the following : if the first block has entries equal to @xmath156 , then the second has entries of @xmath157 , and the third has entries of @xmath158 , @xmath159 , and the @xmath28th has entries of @xmath160 . since @xmath94 is an @xmath28-cycle , @xmath161 , which confirms that the @xmath28 blocks are distinct .
we are now able to prove matrix @xmath108 has no repeated rows by showing that an arbitrary row can only exist in one place in @xmath108 .
the matrix @xmath108 has no two rows with identical entries .
we begin by proving the following claim using an induction argument .
+ * claim : * if two rows share their first @xmath2 entries , then they both belong to the scope of the same @xmath2-block .
_ pf _ : as we ve already mentioned , the first column of @xmath108 is itself one block , so the @xmath7 case is trivial .
suppose the claim is true for @xmath2 , and let the @xmath162th and @xmath163th rows share their first @xmath164 entries .
then they must share their first @xmath2 entries , so by assumption rows @xmath162 and @xmath163 belong to the scope of a single @xmath2-block . by claim 3 , this scope is equal to that of a collection of @xmath28 distinct ( @xmath164)-blocks . since rows @xmath162 and @xmath163 also share the ( @xmath164)st entry , they can only be in the scope of one of those ( @xmath164)-blocks , which proves the claim .
this result tells us that if rows @xmath162 and @xmath163 share all @xmath45 of their entries , then they belong to the scope of the same @xmath45-block . but
a @xmath45-block consists of @xmath165 entry , so its scope contains only one row .
then @xmath166 and no two rows of @xmath108 can have identical entries .
[ orderingofknt ] let @xmath167 , @xmath168 .
as defined in line , @xmath169 is an ordering of the vertices of @xmath71 .
over the course of this section , we have already made the arguments proving theorem [ orderingofknt ] .
let s review .
we defined the @xmath81 s to be the rows of the matrices @xmath170 .
we needed only to show that no two of those rows are identical .
for each @xmath2 , @xmath87 has the characterisitic that @xmath171 , which implies that fixing a row fixes all of @xmath87 .
( it also implies that @xmath87 can not have any repeated rows itself . )
so if any two rows across all the rows of @xmath170 are identical , that would imply two of the @xmath87 s are identical .
then of course the first rows of those two matrices are identical , which makes two rows in @xmath108 identical , but we just proved that this can not happen .
so our definition of @xmath169 is indeed an ordering of @xmath68 . to prove that our ordering induces a consecutive radio labeling
, we need to keep track of instances where close " vertices have common entries .
let @xmath172 be the number of coordinates where @xmath81 and @xmath173 agree .
if @xmath174 and @xmath175 , then @xmath176 let @xmath12 be the radio labeling induced from our ordering .
the radio condition becomes @xmath177 equivalently , @xmath178 .
in order for @xmath12 to be consecutive , we need @xmath179 for all @xmath129 , so we need @xmath172 to satisfy @xmath180 since @xmath172 counts coordinates , @xmath181 .
so @xmath182 trivially satisfies if @xmath183 .
also , if @xmath184 , then is equivalent to requiring that @xmath81 and @xmath185 are not identical , which we ve already proven must be the case .
we will show that @xmath182 satisfies for all @xmath186 .
[ kntconsecutive ] let @xmath167 , @xmath168 .
then @xmath187 has a consecutive radio labeling .
we use the ordering @xmath169 of the vertices of @xmath71 from theorem [ orderingofknt ] , and we let @xmath188 be defined as @xmath179 . we will prove that @xmath12 satisfies the radio condition . from the above discussion , we need to show that @xmath182 satisfies @xmath189 for all @xmath190 .
we will be referring again to the @xmath87 s the matrices from @xmath66[orderingdef ] used to define @xmath169 .
let @xmath81 be a row in matrix @xmath191 . if @xmath185 is also a row in matrix @xmath191 , then @xmath192 since @xmath193 .
suppose @xmath185 is not a row in @xmath191 . by assumption @xmath194 .
since the @xmath195 s each have @xmath28 rows , @xmath185 must be a row in @xmath196 .
note that , because @xmath197 , @xmath81 and @xmath185 can not have the same placement as rows in @xmath191 and @xmath196 ( i.e. , if @xmath81 is the third row of @xmath191 , then @xmath185 can not be the third row of @xmath196 ) .
we have seen that @xmath191 and @xmath196 are identical except for a single column , and that column differs by an application of @xmath94 .
these last two observations together imply that @xmath198 . with this
, there is only one case left to consider : @xmath199 .
we need to show @xmath200 .
in this situation , @xmath201 must be the @xmath28th row of @xmath191 and @xmath185 must be the first row of @xmath196 .
so @xmath202 is identical to the first row of @xmath191 except for one coordinate , let s call this the @xmath99th coordinate .
that is , @xmath203 we also know that @xmath204 .
we can observe then that @xmath81 and @xmath202 share no common coordinates ( because @xmath94 is an @xmath28-cycle , @xmath74 ) .
in particular , we check that the @xmath99th coordinates are different : indeed , @xmath205 . we have shown @xmath200 .
therefore , @xmath206 for all @xmath186
, so @xmath71 has a consecutive radio labeling .
it s natural to now wonder if @xmath71 has a consecutive radio labeling for all @xmath64 .
it turns out that we can find a high enough power @xmath45 such that @xmath71 no longer has a consecutive radio labeling . with this next theorem
we can do better we can find such a @xmath45 for an arbitrary graph @xmath1 . given a graph @xmath1 , there is an integer @xmath186 such that @xmath46 does not have a consecutive radio labeling for any @xmath207 . in particular , if @xmath1 has @xmath28 vertices , @xmath208 is such a value . let @xmath209 .
let @xmath210 , and let @xmath211 , @xmath207 . in search of contradiction ,
suppose @xmath169 is an ordering of the vertices such that @xmath212 defined by @xmath213 is a radio labeling
. then @xmath12 must satisfy the radio condition for all @xmath214 : @xmath215 we can get a get an upper bound for @xmath216 by counting the number of coordinates for which @xmath81 and @xmath173 have common entries . if we define @xmath172 to be this number of coordinates where @xmath81 and @xmath173 agree , then @xmath217 combining these two bounds on @xmath218 yields @xmath219 then the maximum possible number of coordinates with common entries between @xmath81 and @xmath173 is @xmath220 fix @xmath81 .
the maximum number of coordinates that @xmath81 can have in common with any prior vertex is @xmath221 .
then the total number of coordinates in which any of the first @xmath222 vertices in the ordering can agree is @xmath223 we can obtain with some computation that @xmath224 since @xmath45 is larger than the number of coordinates with this property , there is at least one coordinate in which none of the vertices @xmath225 agree .
this is impossible to accomplish , however , as we have only @xmath28 possible entries : @xmath226 .
so there is no ordering of the @xmath95 vertices of @xmath46 that induces a consecutive radio labeling .
[ kntforhight ] @xmath71 does not have a consecutive labeling for any @xmath227 .
this is a simple computation : @xmath228
together , theorem [ kntconsecutive ] and corollary [ kntforhight ] give us a lot of information about @xmath71 .
if @xmath67 , then @xmath71 has a consecutive radio labeling , so @xmath229 . if @xmath227 then @xmath71 does not have a consecutive radio labeling , so @xmath230 .
this is illustrated in figure [ fig : kn_graph ] .
|
for @xmath0 and @xmath1 a simple connected graph , a @xmath2-radio labeling @xmath3 of @xmath1 requires all pairs of distinct vertices @xmath4 and @xmath5 to satisfy @xmath6 . when @xmath7 , this requirement gives rise to the familiar labeling known as vertex coloring for which each vertex of a graph is labeled so that adjacent vertices have different colors " .
we consider @xmath2-radio labelings of @xmath1 when @xmath8 . in this
setting , no two vertices can have the same label , so graphs that have radio labelings of consecutive integers are one extreme on the spectrum of possibilities .
examples of such graphs of high diameter are especially rare and desirable .
we construct examples of arbitrarily high diameter , and explore further the tool we used to do this the cartesian product of graphs and its effect on radio labeling .
|
intel s tera - scale computing vision ( _ a parallel path to the future _ ) is to aim for hundreds of cores on a chip . in their white paper
@xcite intel puts `` programmability '' at the top of the list of challenges for the multi - core era , and considers the development of multi - core software to be amongst the greatest challenges for tera - scale computing .
however , hill and marty @xcite propose that in exploiting multi - cores we should not just aim for _ speedup _ but also _ costup _ , i.e. an increase in performance that is greater than the increase in cost , be it measured in money or energy . and that is our target , to maximise costup .
we were presented with the following challenge .
the second author had just completed an empirical study of exact algorithms for the maximum clique problem and a variety of algorithms had been implemented in java @xcite .
being summer , a university environment and the students away , we had access to over 100 teaching machines that we could use but not modify ( limiting us to using ssh , nfs and java ) .
we did nt have a shared memory system with hundreds of cores but we did have access to a hundred networked pcs , but only for four weeks .
could we take one of our programs , make minimal changes to it , distribute it over the machines available and solve some big hard problems quickly ? to be more precise , starting mid - week ( wednesday 4th july )
could we do this by friday ( the 6th ) , use the available resources over the weekend ( 7th and 8th ) , solve some big hard problems and analyse the results on monday ( the 9th ) ?
the problem studied was the _
maximum clique problem_. a * simple undirected graph * g is a pair ( v(g),e(g ) ) where v(g ) is a set of vertices and e(g ) a set of edges .
an edge @xmath0 is in e(g ) if and only if @xmath1 v(g ) and vertex @xmath2 is adjacent to vertex @xmath3 .
a * clique * is a set of vertices c @xmath4 v(g ) such that every pair of vertices in c is adjacent in g. clique is one of the six basic np - complete problems given in @xcite .
it is posed as a decision problem [ gt19 ] : given a simple undirected graph g = ( v(g),e(g ) ) and a positive integer k @xmath5 @xmath6v(g)@xmath6 does g contain a clique of size k or more ?
the optimization problems is then to find a _
maximum clique _ , whose size is denoted @xmath7(g ) .
a graph can be _ coloured _ by assigning colour values to vertices such that adjacent vertices take different colour values .
the minimum number of different colours required is then the _ chromatic number _ of the graph @xmath8(g ) , and @xmath7(g ) @xmath9(g ) .
therefore a colouring of the graph g can be used as an upper bound on @xmath7(g ) . finding the chromatic number is np - hard , but fast approximations exist such as @xcite . in the next section
we describe our starting point , the algorithm mc .
we then show a simple modification to the algorithm that allows distribution at various levels of granularity .
implementation details are then given followed by a report of our computational study .
we then reflect , considering what we might have done differently given more time , and then conclude .
we can address the optimization problem with an exact algorithm , such as a backtracking search @xcite .
backtracking search incrementally constructs the set c ( initially empty ) by choosing a _ candidate vertex _ from the _ candidate set _ p ( initially all of the vertices in v(g ) ) and then adding it to c. having chosen a vertex the candidate set is then updated , removing vertices that can not participate in the evolving clique .
if the candidate set is empty then c is maximal ( and if it is a maximum we save it ) and we then backtrack .
otherwise p is not empty and we continue our search , selecting from p and adding to c. there are other scenarios where we can cut off search , e.g. if what is in p is insufficient to unseat the champion ( the largest clique found so far ) then search can be abandoned .
that is , an upper bound can be computed .
graph colouring can be used to compute an upper bound during search , i.e. if the candidate set can be coloured with @xmath10 colours then it can contain a clique no larger than @xmath10 @xcite .
there are also heuristics that can be used when selecting the candidate vertex , different styles of search , different algorithms to colour the graph and different orders in which to do this .
for our study we use mc ( * m*aximum * c*lique ) , algorithm [ mc ] .
mc is essentially algorithm mcsa1 in @xcite and corresponds to tomita s mcs @xcite with the colour repair step removed .
mcsa1 is a state of the art algorithm and a close competitor to san segundo s bbmc @xcitepat / maxclique ] .
the algorithm performs a binomial search ( see pages 6 and 7 of @xcite ) and uses a colour cutoff .
vertices are selected from the candidate set p and added to the growing clique c. the graph induced by the vertices in p is coloured such that each vertex in p has an associated colour .
vertices are then selected in non - increasing colour order ( largest colour first ) for inclusion in c. assume a vertex v is selected from p and has colour k. the graph induced by the vertices in p , including v , can be coloured with k colours and can therefore contain a clique no larger than k. consequently if the cardinality of c plus k is no larger than that of the largest clique found so far search can be abandoned .
crucial to the success of the algorithm is the quality of the colouring of the candidate set and the time taken to perform that colouring .
empirical evidence suggests that good performance can be had with any of the three colour orderings studied in @xcite .
algorithms 1 and 2 are presented as procedures that deliver a result , possibly void ( line 12 ) and possibly a tuple ( line 22 ) .
we assume that a * set * is an order preserving structure such that when an item @xmath3 is added to a * set * @xmath11 , i.e. @xmath12 , the last element in @xmath11 will be @xmath3 and when @xmath13 the elements in @xmath14 will occur in the same order as they appear in @xmath15 .
mc , algorithm [ mc ] , takes as parameter a graph @xmath16 and has three global variables ( lines 3 to 5 ) : integer @xmath17 the number of vertices in @xmath16 , @xmath18 the set of vertices in the largest maximal clique found so far and integer @xmath19 the size of that clique .
mc then calls expand ( line 8) to explore the backtrack tree to find a largest clique in @xmath16 .
procedure @xmath20 is called ( in line 8) with three arguments : the candidate set @xmath21 ( line 6 ) and the growing clique @xmath22 ( line 7 , initially empty ) and the graph @xmath16 .
initially the candidate set contains all the vertices in the graph , v(g ) , and is sorted in non - increasing degree order ( the call to @xmath23 in line 6 ) and this order is then used for colouring the graph induced by @xmath21 .
procedure @xmath20 starts by colouring the graph induced by @xmath21 ( step 12 ) , delivering a pair @xmath24 where @xmath11 is a stack of vertices and colour is an array of colours ( integers )
. if vertex @xmath3 is at the top of the stack then vertex @xmath3 has colour @xmath25 $ ] and all vertices in the stack have a colour less than or equal to @xmath25 $ ] .
the procedure iterates over the stack ( while loop of line 13 ) , selecting and removing a vertex from the top of the stack ( line 14 )
. if the colour of that vertex is too small then the graph induced by @xmath3 and the remaining vertices in the stack and the vertices in the growing clique @xmath22 will be insufficient to unseat the current champion and search can be terminated ( line 15 ) .
otherwise the vertex @xmath3 is added to the clique ( line 16 ) and a new candidate set is produced @xmath26 ( line 17 ) where @xmath26 is the set of vertices in @xmath21 that are adjacent to the current vertex @xmath3 ( where @xmath27 delivers the _ neighbourhood _ of @xmath3 in @xmath16 ) , consequently each vertex in @xmath26 is adjacent to all vertices in @xmath22 . if the new candidate set is empty then @xmath22 is maximal and if it is larger than the largest clique found so far it is saved ( line 18 ) .
but if @xmath26 is not empty @xmath22 is not maximal and @xmath22 can grow via the recursive call to @xmath20 ( line 19 ) .
regardless , when all possibilities of expanding the current clique with the vertex @xmath3 have been considered that vertex can be removed from the current clique ( line 20 ) and from the candidate set ( line 21 ) .
procedure @xmath28 , line 22 , corresponds to tomita s number - sort in @xcite .
the vertices in @xmath21 are sequentially coloured ( assigned a colour number ) and sorted ( the vertices are delivered as a stack such that the vertices in the stack appear in descending colour order , largest colour at top of stack ) . in line 24
an integer array of colours is created and in line 25 an array of sets is produced , such that the set @xmath29 $ ] contains non - adjacent vertices that have colour @xmath10 , i.e. @xmath29 $ ] is an _
independent set_. the candidate set is iterated over in line 28 , and this will be in non - increasing degree order as a consequence of the initial sorting in line 6 .
line 30 searches for a colour class for the current vertex @xmath3 : if any vertex in @xmath29 $ ] is adjacent to @xmath3 we then look in the next colour class ( increment @xmath10 ) ( lines 37 to 41 ) delivers @xmath30 if in the graph @xmath16 vertex @xmath3 is adjacent to a vertex in the set @xmath11 . ] . in line 31
we have found a suitable @xmath29 $ ] ( possibly empty ) and we add @xmath3 to that @xmath31 , assign that colour to the vertex ( line 32 ) and take note of the number of colours used ( line 33 ) .
once all vertices have been added to colour classes we sort the vertices using a pigeonhole sort ( lines 34 and 35 ) : for each colour class we push the vertices in that colour class onto the stack .
the procedure then finishes by returning the colour - sorted vertices with their colours ( line 36 ) as a pair .
@xmath32 @xmath33 @xmath34)~coloursort(\textbf{set}~p,\textbf{graph}~g)$ ] @xmath35
there are a number of ways we might distribute mc across @xmath36 processors .
we could split the problem into @xmath36 parts , run each part individually and merge the results , as a map - reduce style implementation .
but we can not partition the problem into @xmath36 roughly - equally sized chunks .
we could instead split the problem into @xmath17 parts , where @xmath17 is the number of vertices in the graph , and then use a worker pool model of execution .
each job would expand a backtrack tree rooted on a node at level 1 where the current clique contained a single vertex .
that is , we kick off @xmath17 jobs each with a different clique of size one . but given the uneven size of the search trees , @xmath17 is likely still too small to give well - balanced workloads . more generally , we could divide at level 2 potentially kicking off @xmath37 jobs where each process has an initial clique containing two adjacent vertices .
more generally we might kick off @xmath38 jobs where each job expands @xmath39 backtrack trees rooted at specified nodes at depth @xmath10 .
mcdist allows us to do that . in algorithm [ mcdist ]
mcdist ( line 1 ) takes the following arguments : the graph @xmath16 , a set of sets @xmath40 where each element of @xmath40 is of size @xmath41 , and integer @xmath42 the size of the largest clique reported by other processes .
elements of the set @xmath40 describe the nodes in the backtrack tree to be expanded by this process .
for example , if @xmath43 arity would equal 1 and a call to @xmath44 would explore the backtrack tree rooted on the clique @xmath45 . if @xmath46 a call to @xmath44 would be equivalent to the call @xmath47 . ] . if @xmath48 with @xmath49 and @xmath50 @xmath51 would start expanding search from all triangles ( level 3 ) looking for cliques of size greater than @xmath10 . and
finally , we might divide the set of edges @xmath52 equally amongst the sets @xmath53 to @xmath54 and distribute calls to @xmath55 , for @xmath56 , across the available processors .
each call to @xmath55 would occur within a separate job and the current best clique size @xmath42 would be used at dispatch time . in algorithm [ mcdist ]
we replace expand with its distributed counterpart distexpand .
the essential difference between @xmath20 and @xmath57 is the call to @xmath58 in line 16 , i.e. the lines 16 to 20 in algorithm [ mc ] are now executed conditionally .
procedure @xmath58 determines if the current clique can be considered for expansion .
if the clique is below or above the critical size then expansion can proceed , otherwise search can proceed only if @xmath59 and @xmath22 corresponds to a specified node of interest , i.e. @xmath60 . @xmath61 @xmath62 @xmath63 procedure @xmath57 is similar to the search state re - computation technique used by @xcite and it can be made more efficient .
assume the argument @xmath41 equals @xmath64 .
a call to @xmath28 is made on each call to @xmath57 as the clique @xmath65 is incrementally constructed ( repeatedly passing the test @xmath66 at line 25 ) although it might ultimately be rejected when @xmath67 ( failing the test @xmath60 at line 25 ) . in the worst case ,
if @xmath40 was empty or contained a single tuple that did not correspond to any node in the backtrack tree , @xmath28 would be called @xmath68 times to no effect .
this is the cost of making a simple modification to @xmath69 to give us @xmath51 .
however in practice much of this cost is easily avoidable and in our studies this overhead has always been tolerable . a second improvement is to enhance @xmath58 such that rather than delivering true if @xmath70 we deliver true if @xmath66 and there exist a set @xmath71 such that @xmath72 .
this will reduce redundant search .
a further improvement is to remove elements of @xmath40 after they are expanded and put a test immediately after line 11 that makes a * return * if @xmath40 is empty .
possibilities for implementation were constrained by the available resources . we intended to reuse an existing implementation of the algorithm , which was written in java .
this was convenient , since the available machines all had a jvm installed . for communication
we were limited to nfs ( with file - level locking only ) and ssh .
the existing code was modified in line with the differences between mc and mcdist . rather than passing @xmath40 explicitly , for a graph with @xmath17
vertices an integer @xmath73 between @xmath74 and @xmath75 was used to parameterise subproblems by splitting on the second level of the binomial search tree , as if @xmath41 was @xmath76 .
a simple implementation is evident : we may number the top level nodes from @xmath77 down to @xmath74 , from right to left ( corresponding to the order in which they are popped from @xmath11 ) . then for a given @xmath73
, every element of @xmath40 contains the element we numbered @xmath78 . on the second level of the tree
, we again label elements of @xmath11 from right to left , from @xmath79 down to @xmath74 , and for a given @xmath73 , take those where the node s number divided by @xmath17 equals @xmath73 , modulo @xmath80 . for each value of @xmath73 , a file named for that number was created in a `` pending '' directory .
these files were split by the last two digits between 100 subdirectories , to reduce contention and to keep each directory sufficiently small to avoid nfs scalability problems ( initially no split was used , and lock contention prevented scalability beyond 10 machines ) .
the restriction to @xmath75 jobs was necessary to avoid having too many files , but still have significantly more jobs than machines so that they can be distributed using a worker - pool model .
in addition a global `` best so far '' file was used to hold the value @xmath42 for mcdist .
this was set to contain @xmath74 initially .
the worker programs were launched by ssh ( using public key / keychain logins to avoid having to repeatedly enter passwords ) , with one worker program per machine .
each worker program reads in the graph , performs the initial colour ordering on the vertices , and then starts running subproblems .
the worker randomly shuffles the 100 directories ( to reduce contention ) , and then for each directory in turn , while that directory is not empty , picks a job file from that directory and moves it to a `` running '' directory .
the worker then reads in the `` best so far '' , runs the subproblem , saves the result to the job file and moves it to a `` results '' directory .
the `` best so far '' is then updated , and another problem is attempted .
note that the `` best so far '' is only read in before starting any individual subproblem .
two sets of locking are required to avoid race conditions .
firstly , a lock file is associated with each subproblem directory , to ensure that two machines can not both start running the same problem . here exclusive locking is used . secondly , the `` best so far '' file needs to be locked to avoid races when it is being updating , and to avoid the possibility of inconsistent reads .
initially , a shared lock was used when reading in the value before launching a problem , and an exclusive lock was used when checking and updating the file afterwards .
it was observed that this was a serious limiting factor on scalability ( on smaller problems , more than 50% of the total runtime was being wasted waiting for a lock ) .
thus , a more sophisticated mechanism for updating the file was introduced .
the value in the `` best so far '' file may only increase over time and its largest possible value is typically much smaller than the number of subproblems .
this means most of the exclusive locks we were obtaining were in fact not being used to change the value .
this suggests a better strategy : when a worker finishes a problem , it obtains a shared lock on the `` best so far '' file , and compares the existing value to the value it calculated
. the lock is then released .
if the newly calculated value is better than the existing value , then an exclusive lock is obtained ( avoiding the possibility of deadlock , since the shared lock is already released ) .
the value is then re - compared ( in case it has been updated in between releasing the shared lock and obtaining the exclusive lock ) and written if necessary .
we observed that in practice , this mechanism substantially reduced overhead . after execution , obtaining the results is a simple matter of checking every file in the `` results '' directory . ( the `` best so far '' file only contains the size of the best clique , not its members . ) the existing implementation of the algorithm produced data on the number of nodes ( i.e. calls to @xmath57 ) and the time spent working as well as the clique found as part of its results .
our study was performed over hard dimacs instances , instances from the bhoslib suite ( benchmarks with hidden optimum solutionskexu / benchmarks / graph - benchmarks.htm ] ) and erds - rnyi random graphs .
these instances were selected because they took between minutes and weeks on a single machine and were hard enough to cover the start up costs of distribution .
approximately 100 student lab pcs were used , running fedora 13 with an amd athlon 64 x2 5200 + , 4gbytes ram and access to an nfs server .
machines were largely idle , but there was no guarantee regarding availability ( several machines were switched on and off or became unavailable whilst experiments were being run ) . in all cases
a problem was split into @xmath81 jobs and these were then distributed across the machines , e.g. frb35 - 17 - 1 has 450 vertices , 3,600 jobs were produced and these were dispatched over 25 machines , then 50 machines and finally 100 machines .
run time was measured in seconds and is the difference between the wall clock time at the start of the first job and the wall clock time at the end of the last job .
clocks on the machines were loosely synchronised , sometimes differing by a couple of seconds .
the single machine runtimes are for the sequential ( undistributed ) algorithm and exclude the read - in times for the problem instance as in @xcite , whereas the distributed results include program startup and read - in times . table [ bigtable ] shows the run time in seconds to find and prove optimality using 25 , 50 and 100 machines compared to the undistributed time .
we list the number of vertices in the graph ( @xmath17 ) , the size of the maximum clique ( @xmath7 ) and when more than one machine was used the speed up ( gain ) .
the instances mann - a45 , brock400 and p - hat500 - 3 are from dimacs and frb * from bhoslib .
.large hard instance : run time in seconds , using 1 to 100 machines .
an entry of corresponds to job that did not terminate after one week . [ cols="^,^,^,^,^,^,^,^,^,^",options="header " , ] table [ randomtable ] gives results on 30 random graphs @xmath82 all with @xmath83 vertices .
the first 10 graphs have edge probability of @xmath84 , @xmath85 for the next 10 graphs and @xmath86 for the last10 .
the @xmath87 instances are easy on a single processor , taking less than a second .
therefore the run time on many machines is largely overhead of dispatching @xmath88 jobs and collecting their results .
this demonstrates that our approach is only applicable to large hard instances where that overhead can be amortised .
the @xmath89 instances are relatively hard ( about 20 minutes on a single machine ) and the overhead pays off modestly with a speedup between 1.8 ( 1000 - 50 - 09 with 50 machines ) and 5.6 ( 1000 - 50 - 00 with 100 machines ) .
why so modest ?
analysing instance 1000 - 50 - 00 shows that none of the jobs took more than two seconds runtime yet incurred the same overhead as the easier @xmath87 instances .
so again , the overhead remains significant for these instances .
for @xmath86 only 50 machines were available to us , nevertheless we see speedups from a minimum of 19.8 ( 25 machines ) up to 42.3 ( 50 machines ) . on several occasions during execution of some of the larger problems , a machine was either powered off or rebooted .
here the nfs implementation was an advantage : the lost task could easily be identified and restarted .
the `` results '' directory also provides an easy check that every task was in fact executed .
the system is in no way tolerant of failures of the nfs server , but experience suggests that the nfs server is far more reliable than the lab machines .
on reflection , there are many things we could do to improve performance if we were given more time .
[ [ order - of - task - execution ] ] order of task execution : + + + + + + + + + + + + + + + + + + + + + + + + we use a random order of task execution as a simple way of reducing contention .
this is unlikely to be optimal , as the initial ordering of vertices has a large effect upon performance .
this also significantly affects reproducibility of results .
although reproducibility could be improved by using fixed per - machine orderings , a more sophisticated implementation that did not have to rely upon nfs would be able to dispatch jobs in the order in which they would be executed in the non - parallel version of the algorithm .
[ [ slow - startup ] ] slow startup : + + + + + + + + + + + + + with the current implementation it can take up to a minute for all 100 machines to start running .
this is because ssh login attempts are deliberately rate limited .
we could avoid this cost with a more sophisticated startup mechanism .
[ [ re - reading - the - best - so - far ] ] re - reading the best so far : + + + + + + + + + + + + + + + + + + + + + + + + + + + we only read and write the file containing the best clique found so far at the start and end of a task respectively .
it would be better to do this more frequently .
however , due to locking , this has considerable overhead on nfs .
this also affects reproducibility of results : in some cases , a small delay before starting a job would lead to it having a better `` best so far '' value , which in turn would vastly reduce runtimes .
however , some of the long tails in other problems would not be removed by this method . in some cases , a better initial `` best so far ''
is of no help for the most time - consuming subproblems .
[ [ finer - granularity - vs - work - stealing ] ] finer granularity vs work - stealing : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + to reduce the long tail in cases where re - reading `` best so far '' does not help , we could split the second level fully into ( less than ) @xmath90 jobs , or split on the first three levels .
this was not possible with the nfs server available to us , but would be an option for better implementations .
an increase in splitting is still not enough to remove every long tail , however . in some cases there
are a small number of areas deep down in the tree that contribute most to the runtime . dealing with these
would need some kind of work stealing mechanism , which in turn requires much more sophisticated communication than was available .
[ [ threading - the - workers ] ] threading the workers : + + + + + + + + + + + + + + + + + + + + + + each of the lab machines we had available was dual core .
memory limitations prevented us from running two instances of the worker program per machine ; threading the client ( and adding a second set of in - process locking , since file locks are per - program rather than per - thread ) could possibly give us the equivalent of doubling the number of machines , at the expense of a more complicated implementation .
[ [ not - using - java ] ] not using java : + + + + + + + + + + + + + + + java was used due to an existing implementation being available , and because of the ease of running java programs on non - identical systems . a reimplementation in a faster language and
the use of bit encoded sets ( such as in @xcite ) would produce a substantial speed - up , at the cost of increased development time and complexity of implementation .
so , how did we do ?
did we get a _ costup _ ?
did we get an increase in performance greater than the increase in cost ?
we think so .
just looking at the frb35 instances , each instance typically takes weeks to solve on a single machine .
we solve frb35 - 17 - 5 in 11 hours , frb35 - 17 - 2 in under 7 hours ( and over 13 days on a single machine ) , frb35 - 17 - 1 in about 9 hours ( and more than a week on a single machine ) , frb35 - 17 - 4 in under 5 hours and frb35 - 17 - 3 in under 4 hours .
we have spent less than a week to do this , less than a week to get more than a week s speedup .
many of our difficulties were down to performance limitations with nfs , and these would be vastly reduced if we were using a shared memory multi - core or even just a numa multi - processor system .
in other words , we believe that things are going to get better for this kind of costup in the future , not worse .
|
we take an existing implementation of an algorithm for the maximum clique problem and modify it so that we can distribute it over an ad - hoc cluster of machines .
our goal was to achieve a significant speedup in performance with minimal development effort , i.e. a _
maximum costup_. we present a simple modification to a state - of - the - art exact algorithm for maximum clique that allows us to distribute it across many machines .
an empirical study over large hard benchmarks shows that speedups of an order of magnitude are routine for 25 or more machines .
|
the dynamics of a quantum coherent capacitor connected via a single contact to an electron reservoir have attracted experimental and theoretical interest .
a capacitor connected via a quantum point contact ( qpc ) to an edge state shows mesoscopic capacitance oscillations and a quantized charge relaxation resistance @xcite .
in addition a recent experiment demonstrated an `` electron gun '' emitting and absorbing a single electron in every oscillation cycle @xcite .
the emission process @xcite injects an electron into states above the fermi level , whereas absorption of an electron leaves a hole below the fermi energy .
the invention of lasers revolutionized optics .
similarly , single electron injectors either using capacitors or quantized electron pumps @xcite provide novel , coherent sources for electronics .
it is a challenging task to detect the electrons with the speed they were emitted with . in modern experiments the dynamics of single electron transport through a mesoscopic system is often explored experimentally using as a charge detector , either a radio - frequency single - electron transistor @xcite or a qpc @xcite .
however the speed of these detectors is not sufficient to detect electrons with a nanosecond resolution . to circumvent this problem ,
we propose as a fast detector a device which is analogous to the emitter : a quantum capacitor , such as used in @xcite .
such a detector is able to register particles as fast as an emitter can inject them into the quantum circuit .
we therefore consider a system consisting of two quantum cavities coupled in series by a single edge state and modulated by in general different , periodically - varying potentials , both with frequency @xmath0 , as shown in fig .
[ fig_model ] . and @xmath1 act homogeneously on the regions of the two cavities .
] the charge emitted by the first cavity is detected by nullifying the _ total _ current with the use of the modulation of the second cavity .
namely , the potential @xmath1 can be chosen in such a way that the total current vanishes .
in general the current @xmath2 consists of a series of pulses corresponding to electrons and holes emitted by either of the cavities .
however if the time when an electron was emitted by the first cavity coincides with the time when a hole was emitted by the second cavity the total current @xmath2 is suppressed .
this electron - hole annihilation process can be viewed as the reabsorption by the second cavity of an electron emitted by the first cavity and it can be used to count electrons
. if the counting efficiency is perfect the total current vanishes completely @xcite .
since the capacitor system generates an ac current it is convenient to investigate the degree of the current suppression by studying the square of the current integrated over one period @xmath3 , @xmath4 note that @xmath5 is different from the noise , see ref . .
we develop the conditions for nullifying the total current and investigate the measuring precision . alternatively ,
being driven in phase such a double - capacitor system can serve as a two - electron ( two - hole ) emitter .
the system consists of two cavities with edge states of circumference @xmath6 and @xmath7 connected via qpcs with the reflection ( transmission ) amplitudes @xmath8 ( @xmath9 ) and @xmath10 ( @xmath11 ) to an edge state of length @xmath12 and modulated by time - dependent potentials @xmath13 and @xmath1 respectively .
a particle with energy @xmath14 entering the cavity @xmath15 picks up a phase @xmath16 , which is the kinetic phase of the guiding center motion @xcite .
the time @xmath17 that a particle spends for one revolution in the cavity @xmath15 is related to the cavity s level spacing , @xmath18 . due to the time - dependent potential @xmath19
an additional time - dependent phase @xmath20 is accumulated in the cavity during @xmath21 revolutions .
the separate cavities can be described by a time - dependent scattering matrix for a particle with incoming energy @xmath14 , leaving the system at time @xmath22 , given by the fabry - perot like expression @xmath23 with the time @xmath24 which the particle spends in the connecting edge state , the scattering matrix of the full system is @xmath25 a floquet scattering matrix approach @xcite , used to deal with quantum pumping @xcite , enables us to investigate the dynamics of the system beyond the linear - response regime and adiabatic approximations .
the full time - dependent current response to a periodic modulation with frequency @xmath0 is @xmath26\\ & & \cdot\frac{\omega}{2\pi}\int_0^{2\pi/\omega } dt ' e^{in\omega(t - t ' ) } s_\mathrm{tot}^*(t',e)s_\mathrm{tot}(t , e)\ .\nonumber\end{aligned}\ ] ] in the following we analyze the conditions to achieve efficient particle counting in the double - capacitor system by nullifying the total current and discuss the precision .
inserting the total scattering matrix , given in eq .
( [ eq_full_s ] ) into the current formula of eq .
( [ eq_current_general ] ) , we obtain a general result for the total current due to a harmonic modulation of the system .
we first investigate the adiabatic regime , specifying results at zero and at high temperatures .
subsequently corrections to the adiabatic results and the strongly nonadiabatic limit are considered . in the following
we calculate the current response to two potentials @xmath27 . in the adiabatic limit , @xmath28 , where the time scale set by the modulation is much larger than the time particles spend in the cavities and the connecting edge state , we expand eq .
( [ eq_current_general ] ) in first order @xmath0 .
the current @xmath29 is related to the instantaneous densities of states @xmath30 of the two cavities @xmath31.\end{aligned}\ ] ] with the transmission @xmath32 the density of states is @xmath33 the phase @xmath34 can be written as the sum of a time - dependent and a time - independent contribution @xmath35 with @xmath36 the phase of the respective qpc s reflection coefficient is given by @xmath37 , with @xmath38 .
the detuning @xmath39 defines the position of the quantum level in the cavity @xmath15 with respect to the fermi level @xmath40 at zero driving amplitude , @xmath41 .
thus to lowest order in frequency the current consists of a sum of separate contributions of the two cavities .
this means that the total time - dependent current can be nullified , whenever the phase difference @xmath42 of the two harmonic modulations is equal to @xmath43 and when the amplitudes together with the cavity parameters are adjusted in an appropriate way .
we now choose a harmonic time - dependence , @xmath44 , with @xmath45 . from eq .
( [ eq_lowest_order ] ) as a function of the phase difference of the modulation potentials .
the ratio between modulation amplitude and the level spacing of the first cavity is given by @xmath46 .
the transmission probabilities of the qpcs of the two cavities are @xmath47 .
( a ) different detuning and @xmath48 .
( b ) fixed detuning @xmath49 and different values of @xmath50.,title="fig:",width=288 ] from eq .
( [ eq_lowest_order ] ) as a function of the phase difference of the modulation potentials .
the ratio between modulation amplitude and the level spacing of the first cavity is given by @xmath46 .
the transmission probabilities of the qpcs of the two cavities are @xmath47 .
( a ) different detuning and @xmath48 .
( b ) fixed detuning @xmath49 and different values of @xmath50.,title="fig:",width=288 ] in fig .
[ fig_current_average ] we plot the time integral of the squared current , eq .
( [ eq_lowest_order ] ) , as a function of the phase difference of the potentials given for different choices of detuning @xmath52 , @xmath53 and for different values for @xmath50 . to understand these plots
, we consider the limit of very low transmission at the qpc s , @xmath54 .
then the instantaneous density of states , eq .
( [ eq_dos ] ) , of the cavities takes the form of a sum of breit - wigner resonances , around the zeroes of the quantity @xmath34 , defined in eq .
( [ eq_energy_phase ] ) , to be taken mod@xmath55 .
we take the modulation amplitude @xmath56 to be smaller than half the level spacing @xmath57 and larger than the detuning @xmath58 , such that one electron and one hole are emitted per cycle .
we consider particles with energies equal to the fermi energy . when periodically driving the potentials @xmath19 , the densities of states have a peak at the fermi energy around resonance times @xmath59 and @xmath60 .
to lowest order @xmath0 , the current pulse generated , eq . [ eq_lowest_order ] , is expressed in terms of the resonance times , @xmath61 , and the half - widths of the pulses , @xmath62 , @xmath63^{-1/2}\ .\end{aligned}\ ] ] we are interested in a situation where during the driving process an electron and a hole are fully emitted , separately from each other , and therefore the distance between the resonance times @xmath64 is much larger than the width of the current pulse , @xmath65 .
we find @xmath66 + \frac{2e^2}{\pi}\left [ l(t_1^+-t_2^+ ) \right.\nonumber\\ & & \left .
+ l(t_1 ^ --t_2 ^ -)-l(t_1^+-t_2 ^ -)-l(t_1 ^ --t_2^+ ) \right]\ .\end{aligned}\ ] ] where we introduce the lorentzian @xmath67 $ ] .
its arguments @xmath68 are taken mod@xmath55 , in the interval @xmath69 $ ] .
the four lorentzians contribute only if the respective resonance times are close to each other compared to the width of the current pulse .
if the first two lorentzians contribute , two particles are emitted by the system at the same time , either two electrons or two holes and @xmath70 is maximized .
we are instead interested in the situation where both of the last two terms contribute , meaning that one cavity emits a hole approximately at the same time as the other emits an electron and vice versa .
the conditions for nullifying the current exactly are [ eq_conditions ] @xmath71 experimentally these conditions can be obtained by tuning the phase @xmath72 , the amplitude of the time dependent part and the phase difference of the potentials .
close to these conditions , @xmath70 as a function of the phase difference has a pronounced dip @xmath73 where @xmath74 is taken mod@xmath55 on the interval @xmath75 $ ] . in fig .
[ fig_current_average ] ( a ) we show @xmath70 as a function of @xmath76 at finite transmission probability of the qpcs for @xmath77 . whenever the maximum is at @xmath78 , both the two electrons and the two holes are respectively emitted at the same time ( solid and dashed - dotted curve ) .
whenever the minimum is at @xmath79 , any emitted electron is annihilated by a hole at the same time ( solid and dashed line ) . if the current pulses of an electron of one cavity and a hole of the other are both coinciding but the width of the pulses are different , the distance of the minimum of @xmath70 from zero is @xmath80 showing a smooth dependence on the system parameters .
it guarantees the robustness of the dip against small deviations from the ideal conditions .
this minimum for the more general case of eq .
( [ eq_lowest_order ] ) is shown in fig .
[ fig_current_average ] ( b ) . in this regime
the quantized emission is destroyed .
however the current nullification can still be achieved and , e.g. , be used to tune the parameters of the cavities . at high temperatures we use @xmath82 in eq .
( [ eq_lowest_order ] ) . then from eq.([eq_scip ] ) we find that the time integral of the square of the low - frequency current takes a particularly simple form @xmath83 it shows a cosine - like behavior as a function of the phase - difference , in contrast to the zero - temperature result , where the width of the dips ( peaks ) is determined by @xmath62 .
independently of the detuning of the two cavities and the transmission of the qpcs , @xmath84 is exactly zero when @xmath85 and @xmath79 and deviates from zero at @xmath79 , by @xmath86 .
the response in second order in frequency @xmath87 \right]\ , \end{aligned}\ ] ] contains mixed terms in the densities of states of the cavities and the connecting channel , @xmath88 , as well .
comprising information about the entire system , it can lead to non - vanishing contributions in the regime where the adiabatic current response vanishes .
it is interesting to consider corrections in higher order @xmath0 , eq .
( [ eq_current_second ] ) , which when @xmath70 vanishes are dominant . independently of the temperature regime ,
the correction to @xmath89 in second order in @xmath0 , for the parameters where @xmath70 in first order in @xmath0 vanishes is always zero .
the leading term in @xmath0 of @xmath89 is then at least of third power in @xmath0 . at zero temperature and small qpc transmission under the conditions given in eqs .
( [ eq_conditions ] ) we find that @xmath90 is of the order @xmath91 , where @xmath92 is the dwell time for an electron in the first cavity . in comparison , the contribution stemming from the first order in frequency current away from resonance is @xmath93 , see eq .
( [ eq_breit_wigner ] ) .
we find that in the non - linear regime the adiabaticity condition is @xmath94 , which differs strongly from the one in linear response , @xmath95 . at high temperature and at the parameters nullifying eq .
( [ eq_hight_first ] ) , @xmath96 is in general not zero and it depends additionally on the transmission of the qpcs and the inverse of the density of states of the connecting edge state , @xmath97 .
however , in contrast to the low - temperature regime @xmath98 can be nullified by introducing further transmission - dependent conditions .
these conditions are directly obtained from eq .
( [ eq_current_second ] ) and read for cavities of unequal lengths , @xmath99 . in an experimental set - up , instead of by a sinusoidal modulation ,
the system is often driven by a square - pulse potential , where a treatment in the highly nonadiabatic regime is required .
for the following analysis we start from eqs .
( [ eq_full_s ] ) and ( [ eq_current_general ] ) , and limit ourselves to the high - temperature regime .
we are interested in a step potential which is the limit of a periodic square - pulse modulation with infinitely long period . in principle
, the potential at the two cavities can have different amplitudes and can be switched on at different times @xmath100 and @xmath101 , with @xmath102 , the sum of the switching time of the first cavity @xmath100 , the time a particle needs to pass through the connecting edge state @xmath24 and a time - delay @xmath103 , where here we choose @xmath104 .
the step potentials at the two cavities read @xmath105 if @xmath106 and @xmath107 otherwise .
the cavities response to the potentials decays with a characteristic time given by the bigger value of @xmath108 . after a waiting time which is much bigger than the decay time
, the charge emitted by the system equals the sum of the charges that would be emitted by two completely independent cavities and is given by @xmath109 . while this charge is nullified for @xmath110 , the nullifying of the integral of the squared current can in general not be reached , meaning that an ac current is generated . to find some simple analytical results ,
let us restrict ourselves to the limit of identical ( @xmath111 and @xmath112 ) , weakly coupled cavities ( @xmath113 ) and consider the interesting case where the total charge is nullified and @xmath89 is suppressed , i.e. @xmath114 . for a single cavity as well as for the double - cavity system we find the time - integral over the squared current to be of the form @xmath115 the function @xmath116 for a system with a single cavity is given by @xmath117 .
for the double - cavity system with equal lengths , @xmath118 oscillates in the potential difference with a phase @xmath119 .
we find @xmath120 $ ] , if @xmath121 , @xmath122 if @xmath123 and @xmath124 , if @xmath125 .
the time integral of the squared current is of the same order for the system of a single and a double cavity , showing that the coupling between the two cavities is important in the highly nonadiabatic regime .
this is indicated already in eq .
( [ eq_current_second ] ) , where in second order @xmath0 mixed terms in the densities of states of the two cavities appear .
we investigated the ac current response of a two - particle emitter consisting of a double - cavity system , and propose it as an efficient tool for counting electrons emitted at high speed .
the square of the total current integrated over one period shows a pronounced dip when the two cavities are synchronized .
we extract the conditions for perfect counting by complete current nullification and show that in the adiabatic regime the counting efficiency is maintained at small deviations from the obtained conditions . in the highly nonadiabatic regime ,
current nullification can in general not be obtained . however , in principle , pulsed cavities can be used to analyze single events .
j. s. and m. b. acknowledge the support of the swiss nsf , the national center of competence in research manep and the european strep project subtle .
|
we consider two driven cavities ( capacitors ) connected in series via an edge state .
the cavities are driven such that they emit an electron and a hole in each cycle .
depending on the phase lag the second cavity can effectively absorb the carriers emitted by the first cavity and nullify the total current or the set - up can be made to work as a two - particle emitter .
we examine the precision with which the current can be nullified and with which the second cavity effectively counts the particles emitted by the first one . to achieve single - particle detection we examine pulsed cavities .
|
over the last years , effective field theory ( eft ) methods have become the standard tool to deal with strong interactions in the non - perturbative regime . in the form of chiral perturbation theory ( chpt ) , they have been used with some success in both the mesonic and single nucleon sectors .
the eft of nuclear forces based on a chiral expansion was originally suggested by weinberg @xcite , and since then a lot of work has been devoted to gain a better understanding of the two - nucleon interaction at low and intermediate energies @xcite@xcite .
while at the beginning , these studies did not aim at substituting the highly successful _ realistic _
potentials built from meson exchanges ( bonn - jlich , nijmegen , argonne , @xmath1 potentials ) , the importance of uncovering such an eft can not be ignored , since this theory will allow for rigorous calculations of both elastic and inelastic processes in systems with two or more nucleons , in a framework consistent with the standard model of strong and electroweak interactions .
furthermore , the latter eft works provide an accurate description of nucleon - nucleon ( nn ) phase - shifts for several partial waves and in a wide range of energies @xcite .
weinberg s original proposal was to determine the nn potentials using the organizational principles of chpt and then to insert these potentials into the lippman - schwinger equation ( lse ) to solve for nn scattering amplitudes .
though such a scheme has proved to be successful @xcite , @xcite , @xcite , it suffers from formal inconsistencies , in particular , divergences that arise at a given order in the chiral expansion can not be absorbed by terms of the same order in the lagrangian @xcite,@xcite .
the trouble arises , since weinberg s power counting assumes that the coefficients of the higher order operators were set by a _ typical _ quantum chromo dynamics ( qcd ) scale , but in practice it turns out that there appear explicit @xmath2 factors ( being @xmath2 the nucleon mass ) in the needed counter - terms to renormalize the amplitudes , which formally decrease the degree of the operator in the weinberg s counting .
this was partially solved by kaplan , savage and wise ( ksw ) who introduced a new scheme where pions are treated perturbatively @xcite .
however , it turns out that the ksw expansion converges slowly in the @xmath0 ( @xmath3 ) channel and does not converge at all in the @xmath4@xmath5 coupled channels @xcite , @xcite .
very recently , a so called _ expansion about the chiral limit _ has been proposed in ref .
@xcite , which seems to be equivalent to ksw power counting in the @xmath0 channel and to weinberg power counting in the @xmath4@xmath5 coupled channels . in this work
we adopt the following point of view . for a scattering process involving external momenta @xmath6 , one should only consider a lagrangian / potential which explicitly includes light degrees of freedom for which @xmath7 .
the effects of heavy virtual particles appear as an infinite number of non - renormalizable operators suppressed by some mass scale relevant to the degrees of freedom excluded from the theory . at low - intermediate momenta , below twice the pion mass
, it should be enough to include explicitly one - pion - exchange , and simulate the rest of the physical contributions by a tower on non - renormalizable contact interactions , organized as a derivative expansion . given a potential , which includes contact terms up to some order , the lse performs an infinite sum of diagrams , which in general would require to be renormalized .
after renormalization the coefficients of the contact terms would contain , in general , contributions from all orders in @xmath8 ( pion mass ) .
most of the formal problems appearing within the weinberg s scheme are linked to the use of a somehow restrictive scheme to renormalize the lse @xcite@xcite .
such a restricted conception of the renormalization of the lse would also lead to unexpected consequences in other scenarios , as for example @xmath9 scattering in the @xmath10channel @xcite . among the abundant literature on the subject , the framework presented in ref .
@xcite constitutes a first step towards the renormalization scheme proposed in this work .
there , a subtracted lse is derived , and the numerical values of the fitted low energies constants ( lec s ) are not translated to values for any ultraviolet ( uv ) cutoff . the aim of this paper is to present a renormalization scheme of the lse for nn scattering by a combination of known long - range and unknown short - range potentials .
for simplicity , we restrict the discussion to the @xmath0 channel , where we show that the proposed scheme describes phase - shifts up to center of mass ( cm ) nucleon momenta of the order of 260 mev .
this renormalization scheme is based on previous findings on @xmath9 @xcite and meson - baryon scattering @xcite@xcite , it leads to renormalized amplitudes which fulfill exact elastic unitarity , and it is easily generalized to study higher waves and/or number of derivatives in the contact part of the interaction .
after projecting into the @xmath0 partial wave , the nn lse for a cm nucleon kinetic energy @xmath11 , reads : @xmath12 with @xmath13 mev the nn reduced mass , for which we take that of the neutron - proton system , @xmath14 and @xmath15 are the initial and final relative momenta of the two nucleons , and @xmath16 the @xmath0 nn potential .
the normalization of the scattering amplitude @xmath17 is such that on the mass shell , @xmath18 , it is related to the phase shifts , @xmath19 , by : @xmath20 the potential consists of the one - pion - exchange potential,@xmath21 , plus contact terms,@xmath22 , with up to two derivatives @xcite : @xmath23 with @xmath24 , @xmath25 mev and @xmath26 mev the pion mass and weak decay constant and finally @xmath27 the axial coupling constant .
note that the contact term includes the @xmath28 contribution from one pion exchange , and it also does the leading and sub - leading contributions in the derivative expansion of all shorter distance effects , such as two pion exchange ( tpe ) , tpe contributions would have to be treated in the same footing as the one - pion exchange ones . ] , intermediate @xmath29 s , @xmath30 exchange , etc @xmath1 .
such a procedure suffers from some limitations , for instance , since the tpe potential has not been explicitly included one can not relate nn scattering to other processes as pion - nucleon , pion - deuterium , etc scattering .
the lec s @xmath31 and @xmath32 are not fixed by chiral symmetry and have to be determined by a fit to the phase shifts , as we will see below .
scattering by short - range interactions in the presence of a known long - range potential,@xmath21 , can be treated by distorted wave theory ( dwt ) @xcite .
we write the full scattering matrix as wave and taking into account that do not lead to uv divergences , in what follows we will assume that the measure in momentum space and the free nucleon propagator are given by @xmath33 and @xmath34 , respectively . ]
@xmath35 with @xmath36 the identity , @xmath37 the free nucleon green function and @xmath38 the scattering matrix for @xmath21 alone .
besides , @xmath39 describes the scattering between distorted waves of @xmath21 .
it satisfies the lse @xmath40 where @xmath41 is the nucleon green function in the presence of @xmath21 , i.e. , @xmath42 . to solve the lse of eq .
( [ eq : dwt ] ) , the full off - shell scattering matrix @xmath43 is required , which is determined by solving the lse of eq .
( [ eq : lsedef ] ) with the obvious substitution of @xmath44 .
this equation is finite ( it corresponds to the scattering by the usual yukawa force studied in most books of quantum mechanics ) and does not require to be renormalized , being then possible a numerical evaluation .
we have obtained @xmath38 by discretizing the momentum space and using the inverse matrix algorithm .
some results are shown in the top panel of fig .
[ fig:1s0 ] . to obtain @xmath45 ,
we solve @xmath46 the above equation can be reduced to a linear algebraic system of equations , which solution is straightforward , @xmath47 with the energy dependent functions @xmath48 the uv divergent integrals @xmath49 are defined by : @xmath50 and the pair of integers @xmath51 take the values ( 2,2),(4,2 ) and ( 4,4 ) for the @xmath52 and 4 cases respectively .
we use the dimensional regularization ( dr ) scheme , since it preserves chiral and gauge symmetry and galilean invariance , which makes the integrals relatively simple to evaluate .
dr discards all power - law divergences of the type @xmath53 , which makes finite all the @xmath54 integrals define above , @xmath55 to simplify the @xmath56 integrals it is useful to realize that in dr the linearly uv divergent integral @xmath57 is finite ( @xmath58 ) and independent of @xmath14 @xcite .
making use of the lse satisfied by @xmath38 to get expressions involving the above integral , the @xmath59 and @xmath60 integrals can be related to the @xmath61 one , @xmath62 for @xmath63 .
besides , @xmath64 is logarithmically divergent and it only requires one subtraction , i.e. , @xmath65 is finite and can be numerically evaluated .
plugging eq .
( [ eq : short ] ) into eq .
( [ eq : dwt ] ) one gets for the on - shell scattering amplitude @xmath66 : @xmath67 with @xmath68 and @xmath69 the long and short range on - shell matrices and @xmath70 the integral @xmath71 is finite and the uv divergent integral @xmath72 in dr becomes finite , i.e. , @xmath73 @xmath74 . with all above results one gets @xmath75 with @xmath76 .
elastic unitarity is exactly fulfilled thanks to a watson s type relation satisfied by @xmath77 , i.e. , @xmath78 .
] , @xmath79 and that @xmath80 indeed , the on shell scattering matrix , @xmath81 , can be re - written in a form where unitarity is manifest , @xmath82 above threshold , the function @xmath83 is real , and @xmath84 are the phase shifts deduced from @xmath68 .
the functions @xmath85 and @xmath86 and the phase shifts @xmath87 are plotted in the top panel of fig .
[ fig:1s0 ] .
note that in the complex @xmath11 plane , @xmath83 does not have right - hand - cut , but it does have a left - hand - cut due to pion exchange . before going ahead
we should study how to get rid of the uv divergences .
the dr scheme has led to a drastic reduction of uv divergences and thus we are just left with the logarithmic divergent integral is given by the integral @xmath88]:@xmath89 , which should be absorbed into a redefinition of the coupling constants of the potential , provided one included in the potential all terms consistent with the symmetry principles .
however , the amplitude can not be made finite by simply redefining the couplings @xmath31 and @xmath32 .
this problem arises because we have not included operators with more than two derivatives in the potential , which are needed as counter - terms to render multiple insertions of the two derivative operator finite .
indeed , the divergent constant @xmath89 appears in the function @xmath83 defined above , and it seems natural to define a renormalized contact potential @xmath90 which leads to an infinite series of powers of @xmath91 for @xmath92 . since the lse performs a non - perturbative resummation of diagrams , in principle one should also add a series of counter - terms to the bare amplitude such that the sum of both becomes finite . at each order in the @xmath91 ( or derivative )
expansion , the divergent part of the counter - term series is completely determined .
however , the finite part remains arbitrary .
it means that the coefficients of each of the terms in the @xmath91 series , implicit in the definition of @xmath93 , become undetermined and have to be fitted to data or , if possible , evaluated in qcd .
thus , the scenario might look like pessimistic , and because of the appearance of divergences , @xmath83 turns out to be a completely undetermined function .
this situation has some resemblances to chpt in the meson - meson and meson - baryon sectors . for simplicity ,
let us pay attention to elastic @xmath94 @xmath9 scattering .
it is well known @xcite that the divergences arising at one loop , @xmath95 , can not be absorbed into a redefinition of the leading terms , @xmath96 , of the lagrangian .
new counter - terms , @xmath97 , higher in the chiral expansion are needed to absorb the divergences , which finite parts @xmath98 , remains undetermined , can not be fixed by chiral symmetry and have to be fitted to data .
these lec s are fundamental parameters of the eft , which contain the contribution at low energies of higher degrees of freedom , which have been integrated out .
the evaluation of the divergent parts of the amplitudes with an uv cut - off does not necessarily provide a reasonable estimate for them is employed , the @xmath99 contributions in all isospin - angular momentum elastic @xmath9 scattering channels will be determined just by one parameter , @xmath100 , while at this order there are four independent parameters , @xmath101 , which , for instance , incorporate the effect of the @xmath102 and other resonances in the amplitudes ( see for instance discussion in section 3 ( pages 6370 ) of second entry of ref .
] . to be predictive , one can adopt here a renormalization scheme , similar to that used in refs .
@xcite@xcite to renormalize the bethe salpeter equation for meson - meson and meson - baryon scattering , which reduces the enormous freedom discussed above .
since for small momenta , higher derivative operators should have a tiny influence , we choose to reduce all this proliferation of lec s , by imposing relations , among all lec s associated to counter - terms with a number of derivatives higher than the terms included in the potential , in such a way that the renormalized amplitude can be cast , again , as in eq .
( [ eq : t_res ] ) and therefore it fulfills elastic unitarity .
this amounts in practice , to interpret the previously divergent quantity @xmath89 as a renormalized undetermined parameter . after having renormalized
, we add a superscript @xmath103 to differentiate between the previously divergent , @xmath104 , and now finite quantity , @xmath105 , etc ... , contributions to that of the @xmath106 one ( note that , the @xmath106 contribution shows up at the one - loop level induced by contact terms quadratic in momenta : @xmath107 and @xmath108 ) , which is determined by @xmath109 .
the relation is such that elastic unitarity is restored , and thus this scheme differs from those where the higher order terms are set to zero . indeed , setting to zero the higher order terms is as arbitrary as setting them to any other value . at next order in this derivative expansion , it is to say when terms of the type @xmath110 , @xmath111 or @xmath112 are included explicitly in the potential , the lse would generate terms of order @xmath113 and higher in the amplitude .
thus , the expansion based on this renormalization scheme is systematically improvable .
] , @xmath114 .
this parameter and therefore the renormalized amplitude can be expressed in terms of physical ( measurable ) magnitudes .
the estimate given by means of a _
reasonable _ uv cut - off for the numerical value of @xmath114 might not be good .
for instance , eqs .
( a15 ) of second entry of ref .
@xcite illustrate this point , if an uv cutoff is employed , the divergent integrals appearing there will be independent of the @xmath115 channel , and to get a reasonably value for @xmath116 unrealistic scales or cut - offs of the order of 300 gev will be needed .
this is due to the special nature of the @xmath102-resonance , and that is not the case , for instance , for the @xmath117waves .
thus , we are proposing an expansion of the short range part of the potential in powers of @xmath91 , i.e. , delta function and its derivatives in coordinate space , and for a given potential , to compute @xmath81 to all orders in the @xmath91 expansion to restore exact unitarity .
thus , the contact terms effectively account for degrees of freedom higher than the pion and one should expect the scheme to work up to the first threshold , likely two pion production ( @xmath118 ) , which is around the normal nuclear matter fermi momentum .
a word of caution must be said here ; this expansion is not equivalent to a chiral expansion in @xmath8 , since the coefficients of the contact interaction would contain contributions from all orders in @xmath8 , as it was firstly pointed out in ref .
: phase shifts and functions @xmath85 and @xmath119 defined in eqs .
( [ eq : defj ] ) and ( [ eq : defl0 ] ) , respectively . from the energy dependence of @xmath84
, we deduce the effective range parameters ( see eq .
( [ eq : ere ] ) ) @xmath120 fm and @xmath121 fm .
bottom panel : @xmath0 @xmath122 phase shifts in degrees plotted versus cm momentum ( log scale ) .
solid line stands for the results of this work ( eq . ( [ eq : fit ] ) ) , while other two approaches with @xmath123 and @xmath124 respectively , are also shown .
data are taken from the phase - shift analysis of ref .
@xcite , and data for momenta above the vertical line ( 260 mev ) have not been included in the fit.,height=521 ] after the above discussion , it is clear we have three undetermined lec s : @xmath31 , @xmath32 and @xmath125 , which we obtain from a @xmath126 fit to nijmegen phase - shifts @xcite . to perform the fits , we assume errors on the phase shifts , @xmath19 ( in degrees ) , given by @xmath127\right)/100 $ ] , where @xmath128 is the kinetic energy in the laboratory frame .
we fit data from threshold up to a certain value , @xmath129 , of the cm nucleon momentum .
we determine @xmath129 , by studying the dependence of @xmath130 on it , and fix it in the region of 260 mev , since the inclusion of simply a new datum of higher energy would double the value of @xmath130 .
thus , we have fitted 62 phase shift values for which @xmath128 has varied from 0.01 to 142 mev , and results are shown ( solid line ) in the bottom panel of fig .
[ fig:1s0 ] .
best fit parameters are @xmath131 with @xmath132 . in the above equation ,
statistical errors have been shown in brackets , affect to the last digit of the parameters and are given by the square root of the diagonal elements of the covariance matrix , @xmath133_{ij}$ ] , with @xmath134 any of the three fitted parameters .
it is worth to mention that from the above matrix , one also learns that the pairs @xmath135 and @xmath136 are highly anticorrelated , with correlation coefficients ( @xmath137 ) smaller than @xmath138 .
this means that effectively there is just one independent parameter .
we will come back to this point below .
if the derivative terms of the contact interaction are set to zero , @xmath139 , the uv divergence can be absorbed in @xmath31 @xcite and thus one is just left with one parameter , which can be fitted to data . yet , though @xmath140 , one can arbitrarily set the renormalized coefficients of the higher order terms to zero , i.e. , take @xmath141 , and fit the non - zero lec s , @xmath31 and @xmath32 , to data .
both procedures are also shown in fig .
[ fig:1s0 ] , the first one leads to the scheme developed in ref .
@xcite and the second one was studied for the very first time in ref .
as seen in the figure , these two schemes do not work for momenta higher than 10 and 40 mev respectively , while the solid line provides a good description of data up to momenta of the order of 260 mev .
indeed , from eq .
( [ eq : fit ] ) we have @xmath142 mev , while such ratio takes a value around of 70 mev when @xmath109 is set to zero .
the scattering length , @xmath143 , and the effective range , @xmath144 , are defined from the effective range expansion , @xmath145 fitting the inverse of our amplitude , up to @xmath146 mev , to the above formula , we get @xmath147 fm and @xmath148 fm , in reasonable agreement with the nijmegen results @xcite ( @xmath149 fm and 2.73 fm , respectively ) also obtained in this way . to finish , we would like just to summarize the results obtained in this work .
we advocate for an expansion of the short range part of the potential in powers of @xmath91 and computing @xmath81 to all orders in the @xmath91 expansion to restore exact elastic unitarity .
dwt techniques and dr have been used to solve the @xmath150 nn lse with a potential , which consists of pion exchange and contact terms with up to two derivatives .
the procedure of solving the lse is quite simple and an explicit expression , where exact elastic unitarity is manifest , has been given as well ( see eq .
( [ eq : t_res ] ) ) .
a particular renormalization scheme has also been discussed .
it is based on previous findings on meson - meson and meson - baryon systems and its main ingredient is to realize that an eft is not a renormalizable field theory in the sense of qcd , i.e. , with a finite number of counter - terms .
thus , to keep finite the amplitude , obtained after performing the non - perturbative resummation implicit in the lse , would require , in general , the addition of an infinite set of counter - terms in the short distance potential ( @xmath91 series ) .
the finite parts of the coefficients of the series would remain undetermined ( lec s ) and encode the contribution of higher degrees of freedom , not explicitly included .
an uv cut - off can effectively account for this freedom only up to some momentum scale , which will depend on the terms explicitly included in the potential . for a contact potential including up to @xmath91 terms
, we find a good description of phase shifts up to cm nn momenta of the order of 260 mev with the inclusion of just one additional parameter , @xmath109 . in ref .
@xcite , an uv cut - off is used to regularize the amplitudes , and with reasonable values of it , in the range @xmath151 gev , a good description of data is found .
note , however , that this does not have always to be the case and depends , as we mentioned above , of the physical system and of the order of the expansion included in the potential .
thus , to describe the @xmath10resonance in @xmath9 scattering , one is left to deal with uv cut - off s of the order of 300 gev @xcite,@xcite .
unrealistic values of the uv cutoff are also needed to account for the @xmath152resonance in meson baryon scattering @xcite,@xcite , but this is not the case , for instance , for the @xmath153 and @xmath154 resonances in @xmath9 scattering @xcite or the @xmath155 resonance in meson baryon scattering @xcite .
besides , we find extremely big anticorrelations between @xmath109 and the coefficients , @xmath31 and @xmath32 , of the iterated short distance potential and among the two later ones , as well .
this indicates that the higher derivative operators in the eft , generated in this scheme by the inclusion of @xmath109 , are actually highly correlated .
this might support the idea that though , the higher derivative operators are controlled by a scale that diverges as @xmath156 , thanks to these high correlations , the effects that diverge with @xmath143 cancel , as it was pointed out in ref .
@xcite . to include more derivatives is trivial and would result , besides the left hand cut due to pion exchange and accounted for by means of @xmath38 , in a higher order _ pad _ approximant for the function @xmath83 in eq .
( [ eq : t_res ] ) .
the extension of the procedure to higher partial waves is also straightforward although cumbersome and will be presented elsewhere @xcite .
s. weinberg , phys .
b251 * ( 1990 ) 288 ; nucl . phys .
* b363 * ( 1991)2 ; phys .
* b295 * ( 1992 ) 114 . c. ordez and u. van kolck , phys
* b291 * ( 1992 ) 459 ; u. van kolck , phys . rev . * c49 * ( 1994 ) 2932 ; prog .
* 43 * ( 1999 ) 337 ; c. ordez , l. ray and u. van kolck , phys . rev . lett . * 72 * ( 1994 ) 459 ; phys . rev .
* c53 * ( 1996 ) 2086 .
t.park , d.min and m. rho , phys .
rev . lett .
* 74 * ( 1995 ) 4153 ; nucl
a596 * ( 1996 ) 515 .
d. kaplan , m. savage and m. wise , nucl .
b478 * ( 1996 ) 629 .
d. kaplan , m. savage and m. wise , phys .
b424 * ( 1998 ) 390 ; nucl . phys .
* b534 * ( 1998 ) 329 .
t. cohen and j. hansen , phys . lett . *
b440 * ( 1998 ) 233 ; phys . rev .
* c59 * ( 1999 ) 13 ; phys . rev .
* c59 * ( 1999 ) 3047 ; nucl - th/9908049 .
e. epelbaum , w. glckle and u. meiner , nucl . phys .
* a637 * ( 1998 ) 107 ; nucl . phys .
* a671 * ( 2000 ) 295 .
t. frederico , v.s .
timteo and l. tomio , nucl
a653 * ( 1999 ) 209 .
d. r. phillips , g. rupak , m. savage , phys . lett .
* b473 * ( 2000 ) 209 .
s. fleming , t. mehen and i. stewart , nucl . phys . *
a677 * ( 2000 ) 313 ; phys . rev . * c61 * ( 2000 ) 044005 . m. lutz , nucl . phys .
* a677 * ( 2000 ) 241 .
j. gegelia , phys .
b463 * ( 1999 ) 133 ; j. gegelia and g. japaridze , phys .
b517 * ( 2001 ) 476 .
s. beane , p. bedaque , m. savage , u. van kolck , nucl.phys .
* a700 * ( 2002 ) 377 .
d. eiras and j. soto , nucl - th/0107009 .
t. barford and m.c .
birse , hep - ph/0206146 .
d.r . entem and r. machleidt , phys . lett .
* b524 * ( 2002 ) 93 .
j. oller , nucl - th/0207086 .
j. nieves and e. ruiz arriola , phys .
b455 * ( 1999 ) 30 ; nucl .
a679 * ( 2000 ) 57 .
j. nieves and e. ruiz arriola , phys .
* d64 * ( 2001 ) 116008 ; c. garca - recio , j. nieves , e. ruiz arriola and m. j. vicente - vacas , phys .
* d67 * ( 2003 ) 076009 .
t. inoue , e. oset and m.j .
vicente - vacas , phys . rev . * c65 * ( 2002 ) 035204 .
|
' '' '' we use standard distorted wave theory techniques and dimensional regularization to find out solutions of the nucleon - nucleon lippman schwinger equation with a potential that includes one pion exchange and additional contact terms with derivatives . though for simplicity , we restrict the discussion to the @xmath0 channel and to contact terms containing up to two derivatives , the generalization to higher waves and/or number of derivatives is straightforward .
the undetermined low energy constants emerging out of the renormalization procedure are fitted to data .
|
an array of identical oscillators has been used to model a wide range of systems , such as neural networks , convecting fluids , laser arrays and coupled biochemical oscillators .
these systems exhibit rich collective behaviors including synchrony and spatiotemporal chaos @xcite .
most of the earlier theoretical works on these systems assume either local coupling ( nearest - neighbor interactions ) or global coupling ( infinite - range interactions ) ; a third type named non - locally coupling began to be explored in the past years , which is somewhere between local coupling and global coupling .
in non - local coupled systems , the oscillators interact with all others and the strength between oscillators varies with the distance between them .
chimera state is a spatiotemporal pattern in which some of the identical oscillators are coherent and synchronous while others remain incoherent @xcite .
they usually appear in systems with non - local coupling and could only be built for proper initial conditions @xcite .
chimera state does not relate to the partially synchronized states observed in populations of nonidentical oscillators with dispersive frequencies in which the splitting of the population roots in the inhomogeneity of the oscillator themselves and the intrinsically fastest or slowest oscillators remain desynchronized . its emergence can not be ascribed to a supercritical instability of the spatially uniform oscillation , because it occurs even if the uniform state is stable .
chimera may be a paradigm to study unihemispheric sleep in neurology which states a fact that many creatures sleep with only half of their brain while the other half is still active at the same time .
@xcite . in the year 2002
, chimera state was first reported by kuramoto and his colleagues @xcite when simulating the non - locally coupled complex ginzburg - landau equation .
they showed that identical oscillators with non - locally symmetrical coupling could self - organize into chimera states .
soon , spiral wave chimera @xcite was discovered in two - dimensional arrays of non - locally coupled oscillators . in succession , abrams and
strogatz @xcite found an exact solution for this state in a ring of phase oscillators coupled by a cosine kernel .
recently , two interesting findings on chimera state are reported .
firstly , in the study of the non - locally coupled oscillators with time delay @xcite , clustered chimera state that has spatially distributed phase coherence separated by incoherence with adjacent coherent regions in antiphase , was found .
the observed clustered chimera state in these systems is stationary and its pattern does not change with time .
secondly , abrams and strogatz @xcite found a breathing chimera state in a model consisting of two interacting subpopulations of oscillators .
pikovsky and rosenblum @xcite considered oscillators ensembles consisting of several subpopulations of identical units , with a general heterogeneous coupling between subpopulations , through which they acquired quasiperiodic chimera states .
laing @xcite summarized chimera states in several heterogeneous networks of coupled phase oscillators , in the mean time , he analyzed chimera state applying the ott - antonsen ansatz @xcite in one - dimensional and two - dimensional systems .
he pointed out that , in one - dimensional system , when parameter heterogeneity is introduced , a breathing chimera state exists . in this work ,
we study a one - dimensional array of non - locally coupled identical phase oscillators .
we find that , in the absence of time delay , a two - cluster chimera state could exist .
we also find that , in the absence of parameter heterogeneity , the two - cluster chimera state is not stationary but oscillating .
different from laing s results @xcite , we find that the oscillation of the two - cluster chimera state only exists for the system with a finite number of oscillators . both the two - clustered chimera state and finite size oscillations of the chimera state in this model are analyzed based on the ott - antonsen ansatz .
the array of non - locally coupled phase oscillators can be described in a concise form as @xmath0dx^\prime.\end{aligned}\ ] ] here , @xmath1 is the phase of the oscillator at position @xmath2 at time @xmath3 .
the space variable @xmath2 is in the range @xmath4 $ ] @xmath5 .
the periodic boundary condition is imposed for @xmath6 , otherwise the no - flux boundary condition is imposed on the system .
@xmath7 is the natural frequency ( same for all oscillators ) , which plays no role in the dynamics . without losing generality
, we can set @xmath8 .
the angle @xmath9 @xmath10 is a tunable parameter .
the kernel @xmath11 provides non - local coupling between oscillators .
@xmath12 is non - negative , even , decreasing with @xmath13 along the array , and normalized to have unit integral .
following abrams and strogatz @xcite , we make use of the cosine kernel @xmath14 where @xmath15 . when @xmath6 , abrams and strogatz found a chimera state with only one coherent cluster @xcite .
however , we find a novel chimera state which has two coherent clusters and is oscillating for any finite number of oscillators . as mentioned above , clustered chimera state is only found in the time - delay coupled systems @xcite and oscillating chimera state is found in the systems with subpopulations @xcite or the systems with parameter heterogeneity [ i.e. , @xmath16 @xcite .
let @xmath17 denotes the angular frequency of a rotating frame whose dynamics are simplified as much as possible , and let @xmath18 denotes the phase of an oscillator relative to this frame .
the key idea behind the analysis of chimera stats is the introduction of a mean - field - like quantity , namely , a complex order parameter @xmath19 @xcite which is defined as @xmath20 then eq ( 1 ) becomes @xmath21 for stationary state , @xmath22 and @xmath23 are time - independent and only depend on space variable @xmath2 . let @xmath24 where @xmath17 is the angular velocity for the oscillators in coherent regions , the oscillators with @xmath25 are in coherent regions and are phase - locked to @xmath26 @xcite .
( color online ) ( a ) phase pattern for two - cluster chimera state when the steady state is reached .
( 1 ) is integrated using the runge - kutta method with fixed time step @xmath27 with oscillator number @xmath28 , @xmath6 , @xmath29 and @xmath30 .
( b ) the triangle ( green ) symbol is the modulus @xmath22 of the complex order parameter , the square ( black ) symbol is the distribution of @xmath31 of individual oscillators averaged over 200 time units .
the circle ( red ) symbol is the fluctuation @xmath32 of @xmath33 .
( c ) three types of distribution of the phase @xmath23 of the order parameter at different time .
( d ) , ( e ) and ( f ) show the contour graphs of @xmath34 , @xmath35 and @xmath36 , respectively .
the horizontal axis is position @xmath2 and vertical axis is time @xmath3 . , width=576 ] for parameters @xmath37 , @xmath30 , @xmath6 , the system with @xmath28 phase oscillators could evolve to a two - cluster chimera state under the initial conditions as follows : @xmath38 where @xmath39 is a random variable from a uniform distribution of @xmath40 $ ] . as shown in fig .
1(a ) , there exist two coherent clusters in which all oscillators are synchronized .
oscillators in the same cluster are nearly in phase yet in different clusters are in antiphase . on the other hand ,
the oscillators between the two coherent clusters are de - synchronized and their phases are randomly distributed in @xmath41 $ ] .
( color online ) ( a ) phase pattern for two - cluster chimera .
the parameters @xmath28 , @xmath42 , @xmath29 and @xmath30 .
( b ) the triangle ( green ) symbol is the modulus @xmath22 of the complex order parameter , the square ( black ) symbol is the distribution of @xmath31 of individual oscillators averaged over 200 time units .
the circle ( red ) symbol is the fluctuation @xmath43 of @xmath33 .
( c ) the distribution of the phase @xmath23 of order parameter .
( d ) , ( e ) and ( f ) show the contour graphs of @xmath34 , @xmath35 and @xmath36 , respectively.,width=576 ] then we consider three quantities characterizing a chimera state : the modulus @xmath22 of the complex order parameter at an arbitrary time , the angular velocity @xmath31 averaged in a time interval of 200 units , and the fluctuation of the instantaneous angular velocity which is defined as @xmath44 .
the quantities against the locations of oscillators are presented in fig .
1(b ) . clearly , there are two plateaus on the curve of @xmath45 which refer to the coherent clusters in the chimera state .
the zero @xmath46 in the coherent clusters means that the oscillators in the coherent clusters all move on the same instantaneous angular velocity .
further , nonzero @xmath46 outside the coherent clusters refers to the fluctuation of angular velocities for the oscillators outside the coherent clusters and indicates desynchronization .
1(b ) reveals two features on @xmath22 for the two - cluster chimera state .
firstly , the oscillators can be divided into two domains which join at the minimum of @xmath22 and the curve of @xmath22 against the locations of oscillators does not show symmetry about the minimum of @xmath22 .
further explorations show that @xmath22 is a function of time . as shown in fig .
1(d ) where the spatiotemporal evolution of @xmath22 is featured , @xmath22 is oscillating in each coherent cluster .
especially , when @xmath22 reaches its minimum in one domain , @xmath22 in the other one reaches its maximum .
secondly , the boundaries of the coherent clusters are not determined by the condition of @xmath47 and the coherent regimes are narrower than those expected according to @xmath47 in most of the time , which are different from the stationary chimera state .
the reason for this observation roots in the time - dependent order parameter @xmath22 .
actually , for a forced phase oscillator obeying @xmath48 , the synchronization of the phase oscillator by the force just requires the phase of the oscillator to be confined within @xmath49 but not to a fixed value , which leads the onset of the synchronization of the oscillator not to obey the condition of @xmath50 and the onset of synchronization strongly depends on the details of the functional form of @xmath51 . to be noted , even though @xmath22 is time - dependent
, the oscillators in the coherent clusters still have the same angular velocity which does not fluctuate as exhibited by zero @xmath46 .
the spatiotemporal evolution of @xmath22 in fig .
1(d ) shows another feature : the two - cluster chimera state displays an irregular motion along the ring , for example , the locations of coherent clusters vary with time .
similar phenomenon is also observed for the chimera state with a single cluster @xcite .
furthermore , the snapshot and the time evolution of @xmath23 presented in fig .
1(c ) and ( e ) show that @xmath23 is almost uniform in each domain except for those near the junction between domains and there is a phase difference of @xmath52 for @xmath23 in different domains . to be stressed , the features on @xmath23 revealed by fig .
1(c ) and ( e ) could be used as a more general scheme for the initial conditions to generate an oscillating two - cluster chimera state .
that is , the initial conditions in eq .
[ intial condition ] could be changed to be @xmath53 for @xmath54 and @xmath55 for @xmath56 . to get a better illustration
, we present the evolution of @xmath57 in fig .
1(f ) which shows that , resulting from the oscillation of @xmath22 , the territories of the coherent clusters alter with time .
furthermore , it should be pointed out that , though a two - cluster chimera state could be generated for the system characterized by eq .
[ eq:1 ] in the absence of time delay under proper initial conditions , we have not detected other chimera states with more than two coherent clusters .
( color online ) the maxima of @xmath22 in one domain varies with time for different number of oscillators .
( a , b , c , d ) are for @xmath58 , @xmath59 , @xmath60 and @xmath61 , respectively .
other parameters : @xmath30 , @xmath62 and @xmath42 .
the ( red ) lines denote the growth of @xmath22 in the transient and the upper or lower bound of the maxima of @xmath22 in the steady state .
the inset in ( d ) shows the amplitude of the oscillation of @xmath63 , @xmath64 , against the system size @xmath65 , which indicates a power law of @xmath66.,width=480 ] the two - cluster oscillating chimera state can exist when @xmath67 . in comparison with the case with @xmath6 where oscillators locate on a ring , here the oscillators locate on a chain .
we take @xmath42 as an example .
the results are given in fig .
[ fig2 ] . the differences with those in fig . 1
are that the pattern of two - cluster chimera state becomes stationary in space and , no matter where the chimera state is initialized , it will adjust its pattern to be symmetrical about the center of the chain where the minimum of @xmath22 appears .
the analysis above are made on the system with @xmath28 .
one question is how the observed two - cluster oscillating chimera state depends on the number of oscillators . for this aim ,
we focus on one domain and record the maximum value of @xmath22 in this domain at any time instance ( we denote it as @xmath63 ) . the time - dependent behavior of the two - cluster chimera state can be reflected by the time evolution of @xmath63 .
for example , a constant @xmath63 indicates a stationary chimera state , otherwise an oscillating one . to avoid the influences induced by the irregular motion of the chimera state along the ring in the system with periodic boundary condition
, we let @xmath42 .
the results for different @xmath65 are presented in fig .
one remarkable feature revealed by the figure is that , before the oscillating chimera state is established , the system first evolves to a two - cluster chimera state which looks like a stationary one due to the weak oscillation of @xmath63 .
however , the `` stationary '' state is not stable and its instability leads to the appearance of an oscillating chimera state .
interestingly , the stationary value of @xmath63 is independent of the size of the system and the time consumed for the system to build an oscillating state becomes much longer as the number of the oscillator @xmath65 increases .
another feature in fig.[rmax ] is that larger @xmath65 seems to weaken the oscillation of @xmath63 in the two - cluster oscillating chimera state , which is prominent by the comparison between fig .
[ rmax ] ( a ) ( @xmath58 ) and ( b ) ( @xmath28 ) .
since the transient time to build an oscillating two - cluster chimera state for large @xmath65 becomes extremely long , we just give a rough estimate based on the data presented in the inset of fig .
[ rmax](d ) that the oscillation amplitude @xmath66 , which means that the two - cluster oscillating chimera state becomes a stationary one in the thermodynamic limit , which is different from the laing s results @xcite .
( color online ) the modulus @xmath22 and the phase @xmath23 of the order parameter and @xmath68 by solving of eq .
( [ kbfinal ] ) via an iterative scheme as presented in the context .
( a ) the square ( black ) symbol denotes the distribution of @xmath22 and the circle ( red ) the value of @xmath68 .
( b ) distribution of @xmath23 .
other parameters : @xmath30 , @xmath62 and @xmath6 .
, width=384 ] the above results can be understood theoretically .
first , the two - cluster stationary chimera state in the thermodynamic limit can be explained in terms of kuramoto - battogtokh self - consistency equation@xcite as follows : @xmath69 note that there are three unknown quantities ( the real - valued functions @xmath34 ,
@xmath70 and the real number @xmath68 ) in terms of the assumed choices of @xmath71 and the kernel @xmath11 .
we take @xmath6 as an example . considering the patterns in figs .
1(b ) and ( c ) , the modulus @xmath22 and the phase @xmath23 of the order parameter approximately satisfy @xmath72 we substitute eqs .
( [ symmetry ] ) and ( [ eq : kernel ] ) into eq .
( [ kb ] ) and get @xmath73 @xmath74e^{i\theta(x^\prime)}h(x^\prime)dx^\prime,\end{aligned}\ ] ] where @xmath75 . under the transformation @xmath76 in the second term of the right hand side of eq .
( [ kb2 ] ) , the self - consistency equation changes into @xmath77 ( color online ) phase pattern for two - cluster chimera when the steady state is reached .
( [ eq:12 ] ) and ( [ eq:13 ] ) is integrated using the runge - kutta method with fixed time step @xmath78 and the oscillator number @xmath28 .
( a , c ) are the modulus @xmath22 and the phase @xmath79 of the complex order parameter with @xmath6 .
( b , d ) are the modulus @xmath22 and the phase @xmath79 of the complex order parameter with @xmath42 .
other parameters : @xmath62 , @xmath30 .
, width=480 ] to solve eq .
( [ kbfinal ] ) , we first determinate the value of @xmath68 . because eq . ( [ kbfinal ] )
is left unchanged by any rigid rotation @xmath80 , we can specify the value of @xmath70 at any point @xmath2 we like .
we set @xmath81 .
now we can get @xmath68 .
then we take @xmath34 and @xmath70 obtained from the dynamical simulations as initial guesses and use an iterative scheme to determinate @xmath34 and @xmath70 in function space , behind which the idea is that the current estimates of @xmath34 and @xmath70 can be entered into the right - hand side of ( [ kbfinal ] ) , and used to generate the new estimates appearing on the left - hand side .
figure [ tukb ] shows the results obtained from eqs.(9 ) and ( 2 ) . to be stressed , without the requirement of eq.(7 ) , the self - consistency equation for any finite system always yields to a one - cluster chimera state , which also evidences that the stationary two - cluster chimera state here is not stable . from fig .
[ tukb ] , we could notice the order parameter of stationary state , which has a little difference from fig .
[ pi](b ) and ( c ) in the vicinity of the junction between two domains . the transition of the former case performances like a step function while the latter one is continuous .
the difference probably originates from the finite size effect in fig .
[ pi ] .
the oscillating characteristic could be interpreted with the assistance of the ott - antonsen ansatz @xcite . following the line in @xcite
, we assume that there is a probability density function @xmath82 characterizing the state of the system .
this function satisfies the continuity equation @xcite @xmath83 where @xmath84 with @xmath85 and @xmath86 .
the complex order parameter can be formulated as @xmath87 in terms of complex order parameter @xmath88 , eq .
( [ eq:6 ] ) can be rewritten as @xmath89\end{aligned}\ ] ] where @xmath90 denotes the complex conjugate of @xmath88 and @xmath91 . following ott and antonsen @xcite , we have ( color online ) the boundary of the two - cluster chimera state plotted on @xmath92 parameter plane .
below the curves , the two - cluster chimera state is stable .
the square ( black ) symbol and circle ( red ) symbol are plotted at the parameter values @xmath30 and @xmath93 , respectively .
, width=384 ] @xmath94\}\end{aligned}\ ] ] where @xmath95 is the complex conjugate of the previous term and @xmath96 is the distribution of natural frequency . in this work
, we assign all oscillators a same natural frequency ( @xmath8 ) , so @xmath97 . substituting eqs.([eq:8 ] ) and ( [ eq:9 ] ) into eqs .
( [ eq:5 ] ) and ( [ eq:7 ] ) , we obtain @xmath98\end{aligned}\ ] ] @xmath99 letting @xmath100 , we have @xmath101\end{aligned}\ ] ] @xmath102 ( color online ) two - cluster chimera under the condition of different types of nonlocal coupling kernel @xmath12 .
the top panels are the phase patterns for two - cluster chimera and the bottom are the modulus @xmath22 ( black ) and the phase @xmath23 ( red ) of the order parameters .
( a , b ) is computed under the assumption that @xmath12 takes exponential form with parameters : @xmath103 , @xmath104 and @xmath29 while @xmath12 in ( c , d ) is steplike function with parameters : @xmath105 , @xmath104 and @xmath29 .
all above take no - flux boundary condition and the initial conditions are as following : @xmath55 for @xmath106 and @xmath53 for @xmath56,width=576 ] by numerically simulating these two equations , we have the time evolutions of @xmath34 and @xmath107 .
the results for @xmath6 and @xmath42 are presented in fig .
[ fig3 ] , respectively . clearly , the main features in figs .
( 1 ) and ( 2 ) , such as the oscillating nature of the two - cluster chimera state , the movement of the pattern ( or the frozen pattern ) of @xmath22 and @xmath23 for @xmath6 ( or for @xmath42 ) , and the uniform distribution of @xmath107 in different domains , are reproduced in fig .
[ fig3 ] .
using eqs .
( [ eq:12 ] ) and ( [ eq:13 ] ) , we may probe into the regime for the existence of the two - cluster chimera state on @xmath92 parameter plane .
the results are presented in fig .
[ fig4 ] at @xmath30 and @xmath93 .
as shown in this plot , the two - cluster chimera state is not favorable at large @xmath71 , and either small @xmath108 or large @xmath108 tends to be harmful for the two - cluster chimera state . and
the smaller @xmath109 is , the smaller the domain of two - clustered chimera exists .
if @xmath110 , the domain does not exist any longer .
beyond the stable regime for the two - cluster chimera , i.e. , above the curves in fig .
[ fig4 ] , the initial two - cluster chimera state tends to become a normal chimera state with only one coherent cluster .
the two - cluster oscillating chimera state may be detected in coupled oscillators with other types of non - locally coupling kernel @xmath12 which could be exemplified by two cases . in the first one ,
@xmath12 takes an exponentially decaying function , @xmath111 . in the other case
, @xmath12 takes a steplike function , @xmath112 for @xmath113 and @xmath114 for @xmath115 .
figure 7 shows the results for the systems with these two types of non - locally coupling and , clearly , the two - cluster oscillating chimera states are reproduced .
in summary , we study a one - dimensional system consisting of non - locally coupled phase oscillators , which is a prototype for studying chimera states . by numerically simulating this simplest system , we find the existence of a two - cluster oscillating chimera state in the absence of time delay coupling and parameter heterogeneity .
the numerical results are confirmed by the theoretical analysis based on the self - consistency treatment and the ott - antonsen ansatz .
the work was supported by national natural science foundation of china under grant no .
90921015 and no . 10775022
y. braiman , j. f. lindener and w. l. ditto , nature * 378 * , 465 ( 1995 ) .
l. kocarev and u. parlitz , phys .
lett . * 77 * , 2206 ( 1996 ) . h. g. winful and l. rahman , phys .
rev . lett . * 65 * , 1575 ( 1990 ) .
l. kocarev , u. parlitz , t. stojanovski and p. janjic , phys .
e * 56 * , 1238 ( 1997 ) .
y. kuramoto and d. battogtkh , nonlinear phenomena in complex systems * 5 * , 380 ( 2002 ) .
d. tanaka and y. kuramoto , phys .
e * 68 * , 026219 ( 2003 ) .
d. m. abrams and s. h. strogatz , phys .
* 93 * , 174102 ( 2004 ) .
y. kawamura , phys .
e * 75 * , 056204 ( 2007 ) .
o. e. omelchenko , y. l. maistrenko and p. a. tass , phys .
lett . * 100 * , 044105 ( 2008 ) .
a. e. motter , nat . phys .
* 6 * , 164 ( 2010 ) c. g. mathews , j. a. lesku , s. l. lima , and c. j. amlaner , ethology 112 , 286 ( 2006 ) . s. i. shima and y. kuramoto , phys .
e * 69 * , 036213 ( 2004 ) .
e. a. martens , c. r. laing and s. h. strogatz , phys .
lett . * 104 * , 044101 ( 2010 ) .
j. h. sheeba , v. k. chandrasekar and m. lakshmanan , phys .
e * 81 * , 046203 ( 2010 ) .
g. c. sethia , a. sen and f. m. atay , phys .
lett . * 100 * , 144102 ( 2008 ) .
d. m. abrams , r. mirollo , s. h. strogatz and d. a. wiley , phys .
* 101 * , 084103 ( 2008 ) .
a. pikovsky and m. rosenblum , phys .
lett . * 101 * , 264103 ( 2008 ) . c. r. laing , physica d * 238 * , 1569 ( 2009 ) . c. r. laing , chaos * 19 * , 013113 ( 2009 ) .
e. ott and t. m. antonsen , chaos * 18 * , 037113 ( 2009 ) .
o. e. omelchenko , m. wolfrum , and y. l. maistrenko , phys .
e 81 , 065201 ( 2010 ) .
|
we investigate an array of identical phase oscillators non - locally coupled without time delay , and find that chimera state with two coherent clusters exists which is only reported in delay - coupled systems previously .
moreover , we find that the chimera state is not stationary for any finite number of oscillators .
the existence of the two - cluster chimera state and its time - dependent behaviors for finite number of oscillators are confirmed by the theoretical analysis based on the self - consistency treatment and the ott - antonsen ansatz .
|
throughout the years many studies on galactic star formation have been conducted ( e.g. kennicutt & evans 2009 ; mckee & ostriker 2007 ; zinnecker & yorke 2007 ) .
it has been observed and theorized that within giant molecular clouds ( gmcs ) there are clumps and small cores .
this resulted in the idea that gmcs are hierarchical in nature with substructures spanning a large range of sizes .
clumps are dense regions with masses @xmath210@xmath3 - 10@xmath4m@xmath5 and sizes of a few parsec , whereas cores are smaller ( @xmath20.1pc ) and denser .
it is here where individual star formation occurs ( williams et al .
2000 ) .
the physical processes that determine how molecular clouds fragment , form clumps / cores and then stars depends strongly on the past history of star formation .
this occurs through both recent radiative and mechanical feedback from massive stars and , on longer term , through enhanced cooling due to the buildup of metals .
radiative and mechanical energy input from stellar populations can alter subsequent star formation over a large part of a galaxy and is hence relevant to the evolution of galaxies .
much of our knowledge of star formation on galaxy wide scales is based on scaling laws and other parametric descriptions of star formation ( e.g. kennicutt 2012 ) . but to understand the overall evolution of star formation in galaxies we need to watch the feedback processes at work on giant molecular cloud ( gmc ) scales . by doing this
we can begin to answer how strong feedback environments change the properties of the substructure in gmcs .
applying the theory of galactic star formation to other galaxies has been a challenging process due to the lack of resolution with current instruments .
however , only the nearest galaxies allow us to resolve gmcs and their substructures .
the large magellanic cloud ( lmc ) , is one of the closest galaxy ( d@xmath2 50kpc ) and is close enough that current instruments can resolve the sub - structure of its gmcs to @xmath61pc .
molecular gas is seen across the disk of the lmc , with previous studies revealing a dominant ridge of bright co emission extending @xmath2 2kpc south of 30 doradus , the most extreme star forming region in the local group ( e.g. columbia 1.2 m millimeter wave telescope , cohen et al .
1988 ; nanten , fukui et al .
1999 ; magma survey with the australia telescope national facility [ atnf ] mopra telescope , ott et al .
2008 , pineda et al .
2009 , muller et al .
2010 , hughes et al .
2010 , wong et al .
2011 ) . despite its proximity ( @xmath2 20pc ) to the young star cluster powering 30 doradus , r136 , and its intense radiation field ( @xmath7500@xmath8 in units of draine [ 1978 ] ; pineda et al .
2009 ) , a molecular cloud , 30dor-10 is observed towards the hii region ( e.g. johansson et al .
co ( 1 - 0 ) observations of this gmc reveal that it is about 40pc in size , fainter , more compact and has broader linewidths than other lmc clouds ( johansson et al .
1998 ; heikkil et al .
1999 ; rubio et al .
recent atacama large millimeter array ( alma ) observations of 30dor-10 in @xmath9co , @xmath10co , and c@xmath11o probe the highest resolution view of molecular gas in this extragalactic star forming region ( @xmath20.4pc @xmath12 0.6pc ) ( indebetouw et al .
the @xmath9co is mostly associated with clumps and small pillars of creation. this makes 30dor-10 a perfect target to study extragalactic gmc dense gas properties toward a sub - solar metallicity environment under the influence of an intense starburst .
other active star forming gmcs have been observed at lower resolution along with 30dor-10 using sest , such as the n159 hii complex ( johansson et al .
1994 ; chin et al .
1997 ) in the lmc .
these observations report the detection of eight molecules ( @xmath9co , cs , so , cn , hcn , hnc , hco@xmath0 , and h@xmath1co ) .
heikkil et al .
( 1999 ) detected nine molecules ( @xmath10co , @xmath9co , cn , cs , hnc , hnc , hco@xmath0 , c@xmath1h , and c@xmath13h@xmath1 ) and wang et al .
( 2009 ) detected several molecules toward the star forming regions n113 , n44 , and n214 .
complementary high resolution interferometric observations of a subset of these gmcs were carried out in the dense gas tracers hcn(1 - 0 ) , hco@xmath0(1 - 0 ) , and nh@xmath13 ( 1,1 ) and ( 2,2 ) towards selected non-30doradus gmcs in the lmc using the australia telescope compact array ( atca ) ( wong et al .
2006 ; ott et al .
2010 ; seale et al .
2012 , hereafter s12 ) . in this paper
, we extend the s12 sample ( e.g. n159 , n113 , n105 and n44 ) to include 30dor-10 , to probe dense gas that is embedded in the most extreme radiation field in the lmc . in section 2
, we describe our observational strategy and data reduction . in section 3
, we analyze the physical properties of the hcn and hco@xmath0 clumps in 30dor-10 , and compare them to the properties of hcn and hco@xmath0 clumps detected in other lmc gmcs by s12 . in section 4 , we discuss a potential origin for the filamentary morphology of the dense gas tracers in 30dor-10 , and compare our maps to the higher resolution alma data recently published by indebetouw et al .
our key results and conclusion are summarized in section 5 .
we observed the dense gas cloud , 30dor-10 with the atca .
hco@xmath0 ( j=1@xmath140 ; 89.1885ghz rest frequency ) and hcn(j=1@xmath140 ; 88.6318ghz rest frequency ) are used to probe the structure of the gmc .
ten separate observations were taken during 2011 september 16 - 28 .
we observed 28 pointing centers , which were mosaicked together to cover the inner 3of the 30dor-10 region .
the compact array broadband backend ( cabb ) correlator was used to observe the redshifted frequencies of hco@xmath0 and hcn simultaneously , so observations from both transitions have consistent sensitivities and ( @xmath15 ) coverage .
a bandwidth of 2ghz with 2048 x 1mhz channels and up to 16 zoom bands in each if band were used for the cabb correlator setup . for a given zoomband ,
the native spectral resolution of the correlator configuration corresponds to a channel width of 0.5khz ( 1.6 m @xmath16 ) , but we binned the channels to a velocity width of 0.84 km s@xmath17 to increase the signal to noise prior to analysis .
typical clump linewidths are @xmath2 3 km s@xmath17 ( see table 1 ) , so the clumps remain spectrally well - resolved at this spectral resolution .
the center of the band was tuned to a redshifted velocity of 255 km s@xmath17 ( lsrk ) , which is the average velocity of the area .
all observations were conducted using the most compact array configuration ( h75 ) with baselines of 31 - 48 m giving a synthesized beam fwhm of 6.8@xmath12 5.8(1.6pc @xmath12 1.4pc at the assumed distance of @xmath2 50kpc to the lmc ) .
complex gain calibration was performed using pks 0637 - 752 for all observations and pks 2223 - 052 was used to calibrate the bandpass .
pointing was measured every hour on the phase calibrator with an accuracy of 2 - 5 .
flux calibration was performed using uranus .
the absolute flux calibration uncertainty is estimated to be @xmath250% . however , relative calibration between spectral lines should be significantly better , @xmath610% .
data were reduced , imaged and cubes created using atnf miriad software package ( sault et al .
the root mean square ( rms ) of the brightness fluctuations due to noise for both hco@xmath0 and hcn cubes is 10mjy / beam in a 0.84 km s@xmath17 channel .
we examine the internal structure of the extragalactic gmc in the star forming cloud 30dor-10 at 1.45pc resolution .
figures 1a and 1b show maps of the integrated intensity of hco@xmath0 ( 1 - 0 ) and hcn(1 - 0 ) data while figure 1c shows the peak brightness temperature map of hco@xmath0 ( 1 - 0 ) with clump regions in ellipses overlaid , clump region detection is discussed in more detail in the section below .
figure 2 displays an image of 30dor-10 indicating where intense star formation is occurring .
h@xmath18 is in green , 8@xmath19 in red , and x - ray emission in blue .
the dense gas morphology of 30dor-10 together with h@xmath1o masers and yso counterparts ( gruendl et al .
2009 ) are overlaid on figure 2 .
we observe a roughly linear structure extending outward from the star cluster r136 in the south - west , which bifurcates near the center of the field .
this structure is resolved into a collection of dense clumps .
we detect 13 clumps in hco@xmath0(1 - 0 ) , the four brightest of which are also detected in hcn(1 - 0 ) and are associated with previously identified ysos . properties of individual clumps are determined using cprops ( see section 3.1 ) and can be found in table 1 .
h@xmath1o masers are also coincident with the brightest hco@xmath0 peak . determined either from virial masses or clumps masses under local thermal equilibrium ( lte ) , the masses span from 10@xmath3 to 10@xmath4m@xmath5 .
cprops resolves 4 of the 13 clumps with sizes of 0.9 - 1.6 pc , the remaining clumps are unresolved with radii @xmath200.7pc .
these sizes are consistent with the other gmcs in the lmc discussed in s12 .
the gmcs are significantly brighter in hco@xmath0(1 - 0 ) than hcn(1 - 0 ) , with typical line ratios hcn / hco@xmath0@xmath2 0.20 for the gmcs with m@xmath2110@xmath4m@xmath5 .
we use hco@xmath0 as the primary clump tracer , since hcn is much weaker . to identify hco@xmath0 clumps in the atca data we used the cprops procedure presented by rosolowsky & leroy ( 2006 ) ( r06 ) .
we follow the method of wong et al .
( 2011 ) hereafter w11 and s12 , respectively to facilitate direct comparison with these studies .
we briefly review the basic properties of the algorithm here .
cprops uses a dilated mask technique to isolate clump regions of significant emission within spectral line cubes .
clump regions of significant emission in the atca data are identified by finding volume pixels ( `` voxels '' ) with emission greater than a 3@xmath22 threshold .
cprops is run on position - position - velocity data cubes where the cloud is continuous and bordered by an isosurface of brightness temperature , t@xmath23 .
clump regions are determined when the pixel temperature in the cloud is @xmath21t@xmath23 and that the pixels outside the cloud have t@xmath20t@xmath23 . once the clump regions are determined , we apply various criteria to estimate radii , line widths and other parameters , as follows .
we require that each region has an area larger than two beam sizes and a velocity width of more than a single channel .
also , the intensity contrast between the peak and the edge of the region must be greater than a factor of two . if it meets the above criteria then the cloud can be decomposed , otherwise it is considered noise . local maxima
are then determined by identifying the elements in the data cube that are larger than their neighbors .
once the local maxima are found then these are labeled as a clump and the parameters are then extracted for each clump as described in r06 .
the quoted radius is determined by the geometric mean of the second spatial moments along the minor and major axis .
also , any flux detected below the threshold is assumed to be properly modeled by the extrapolation algorithm .
the clump radius is determined by assuming a brightness profile and distance .
if the extrapolated clump size is smaller than the synthesized beam , it can not be deconvolved .
marginally resolved clumps suffer from substantial radii uncertainties ( s12 ) .
following , s12 , we derive the same clump properties i.e linewidth ( @xmath24 ) , virial mass ( m@xmath25 ) , lte mass ( m@xmath26 ) , and radius .
we compare our values to what s12 obtained for the other gmcs in the lmc in table 2 .
the velocity dispersions are found in the same way as the radii are , except they are extrapolated to what would be measured in zero - noise observations and then deconvolved from the channel width .
assuming gaussian line profiles , the linewidth is determined from the full - width at half maximum as in s12 .
the flux is determined in a similar manner but by using the zeroth moment .
the hco@xmath0 luminosity is determined from the integrated flux scaled by the square of the distance ( w11 ) : @xmath27 = d^{2 } \displaystyle \sum_{\rm i } t_{\rm i}\delta v,\ ] ] where @xmath28 = @xmath29 is the volume of the voxel , t@xmath30 denotes brightness temperatures of individual voxels within the region , d is the distance to the lmc in parsec .
the virial mass is determined by : @xmath31=1040\sigma_{\rm v}^{2}\rm r\ ] ] where @xmath32 is the velocity dispersion in kms@xmath17 and r is the radius in pc .
this assumes a cloud with a radial density profile of @xmath33 .
virial masses are shown in table 1 along with the h@xmath34 gas mass traced by hco@xmath0 under lte conditions , which is determined by following equation ( 9 ) from barnes ( 2011 ) : @xmath35 where n@xmath36 is the column density determined by assuming optically thin emission and lte , shown in equation ( 4 ) , x@xmath36 is the hco@xmath0 abundance relative to h@xmath1 , m@xmath37 is the mass of molecular hydrogen , and r is the radius of the individual clumps in pc .
@xmath38 where t@xmath39 is the excitation temperature .
the hco@xmath0 column density is derived with molecular parameters from the splatalogue database .
an excitation temperature ( t@xmath40 ) of 30k is adopted which is typical for massive clumps observed in dust continuum ( e.g. fandez et al .
2004 ; beltrn et al .
2006 ) and a hco@xmath0 abundance relative to h@xmath1 of x@xmath41 is assumed ( barnes et al .
if the hco@xmath0 abundance relative to h@xmath1 is increased , the h@xmath1 gas mass traced by hco@xmath0 will decrease by a corresponding factor . on the other hand ,
an increase in the excitation temperature to t@xmath42 50k will increase the mass by a factor of two .
all clump parameters discussed above are found in table 1 .
atca observations similar to ours were performed on the star forming regions n159 , n105a , n44 , and n113 by s12 , who detected individual clumps with the same dense gas tracers hco@xmath0(1 - 0 ) and hcn(1 - 0 ) .
s12 used the same compact antenna array configuration of the atca , cabb correlator settings which result in a very similar spatial and spectral resolution and sensitivity compared to our observations .
compared to these gmcs , 30doradus exhibits a much higher star formation rate ( sfr ) , more intense and extended uv flux , and is forming a super - star cluster .
n159 is located south of 30dor-10 along the molecular ridge ( pineda et al .
2009 , 2012 ) and has a high local radiation field due to individual star forming regions .
n113 , n105a and n44 have progressively weaker interstellar radiation field , respectively .
the individual gmc properties are recorded in table 2 . in the following sub - sections we describe both the clump properties of 30 dor-10 and how they compare to the other lmc gmcs .
we investigate dense - gas fraction , peak temperature - linewidth , rms centroid velocity - co linewidth , total co mass to total hco@xmath0 mass , and hcn / hco@xmath0 line ratio relations as shown in figure 3 .
since cprops only obtained deconvolved radii for four of the 13 clumps , the size - linewidth and virial mass - linewidth could not be reliably determined for the entire sample .
we compare the h@xmath1 mass traced by hco@xmath0 of all the clumps in each star forming region ( n159 , n113 , n105a and n44 ) to the molecular gas masses estimated from the gmcs co luminosities from magma data .
the magma co data ( figure 2 ) covers our entire observed region unlike the alma map and we therefore use magma for calculations of the bulk properties of the entire gmc .
m@xmath43 is calculated from w11 equation ( 2 ) : @xmath44=4.4\frac{x_{\rm co}}{2.2\times 10^{20}cm^{-2}(\rm k\,km\,s^{-1 } ) } l_{\rm co},\ ] ] where the galactic co - to - h@xmath1 conversion factor @xmath45 is adopted ( strong et al .
1988 ) but see bolatto et al .
( 2013 ) for other conversion factors .
@xmath46 is the co luminosity and values are shown in table 2 for each cloud .
@xmath46 comes from the magma co data which is measured directly from the integrated intensity map .
the co allows us to examine how total molecular gas compares to the dense gas in each individual gmc . the dense gas fraction for 30dor-10 ( as measured by hco@xmath0(1 - 0)/co(1 - 0 ) ) is quite similar to the other gmcs in the lmc .
30dor-10 s dense gas fraction is @xmath21/12 whereas the other mapped gmcs in the lmc have a dense gas fraction @xmath2 1/10 .
each clump s virial mass and lte molecular mass , traced by hco@xmath0 , is found in table 1 .
if the lmc clumps are composed entirely of dense gas and are in gravitational equilibrium then we would expect m@xmath25/m@xmath26 @xmath2 1 .
the spatially resolved clumps in 30dor-10 , however , exhibit a ratio of m@xmath25/@xmath47 @xmath26 .
such a high ratio may be attributed to several possibilities , including 1 ) the adopted t@xmath40 is underestimated , 2 ) the hco@xmath0 abundance of x@xmath48 10@xmath49 could be overestimated , 3 ) the dense gas traced by hco@xmath0 accounts for less of the total clump mass , or 4 ) that the virial mass estimate is incorrect and clumps are slightly super - virial . on the other hand ,
compared to 30dor-10 the other gmcs also have elevated @xmath50m@xmath25/@xmath51 , but by a smaller amount ( a factor of three ) .
the lte mass determined is dependent on the chosen t@xmath40 and hco@xmath0 abundance and so is likely the largest source of uncertainty .
therefore we consider options 1 ) and 2 ) the most likely cause of the overall difference between m@xmath25 and m@xmath26 in the lmc .
however the larger discrepancy between m@xmath25 and m@xmath26 in 30 dor-10 suggests that some extra reason is needed to explain the factor of three difference . given 30 dor-10 s more intense uv radiation field and dynamical impact of star formation , both higher gas excitation and clumps being super - virial likely are the causes of the increase ( see section 3.2.3 ) .
figure 3c , displays the rms of the velocity centroids of the individual dense clumps versus the co linewidth of the parent gmcs and provides an indication of how the dense clumps are moving relative to the more diffuse molecular material of the whole gmc . within the errors the dense clumps in 30 dor-10
fully sample the gmc s co velocity distribution and so the dense gas is not dynamically segregated from the less dense gas . for a given brightness ,
the clumps in 30dor-10 have a larger line width relative to the clumps in the other four gmcs ( figure 4 ) .
the clumps in 30dor-10 lie at the bottom of the distribution having a weaker temperature and broader linewidths .
the clump line widths of the rest of the four gmcs exhibit similar behavior with respect to each other . a weak trend for increasing clump brightness with line width in 30 dor-10 is also observed .
evidently 30dor-10 is significantly more turbulent than the other gmcs .
the l(hcn)/l(hco@xmath0 ) line ratio relates the two commonly used dense gas tracers .
figure ( 3d ) shows the line luminosity ratio determined from cprops for comparison between these five gmcs in the lmc .
the gmcs are significantly brighter in hco@xmath0(1 - 0 ) than hcn(1 - 0 ) , with typical line ratios l(hcn)/l(hco@xmath0)@xmath2 0.20 for the gmcs with m@xmath2110@xmath4m@xmath5 .
30dor-10 has a similar line ratio as the other gmcs .
we compare our atca hco@xmath0 dense gas clumps to the alma co data , to investigate the physical properties of 30dor-10 .
compared to the alma clump decomposition , we detect 7 hco@xmath0 clumps over the mutually observed field of view ( fov ) while the alma data resolve 103 co sub - clumps . comparing the morphology of co to that of hco@xmath0
we observe that the two are very similar in structure . over the alma field
, hco@xmath0 is detected from all prominent co clumps and the brightest co clumps tend to be the brightest co clumps .
the only exception is clump 10 , which is brighter in hco@xmath0 .
however the integrated intensity map of @xmath9co shows more diffuse , continuous emission than what we see in the hco@xmath0 data ( figure 5 ) .
the virial masses of the detected clumps in the alma and atca data are both @xmath210@xmath4 , providing a good indication of the amount of mass that is gravitationally bound in the associated clumps . on a clump by clump basis the ratio of @xmath9co
/ hco@xmath0 is nearly constant for the filament being around @xmath2 12 - 14.5 .
only clump 3 , which is one of farthest from r136 covered in the alma field , has a significantly higher ratio ( @xmath230 ) .
the @xmath9co / hcn ratios are much higher across the whole gmc ( @xmath260 - 140 for the filament , up to 300 for clump 3 ) , consistent with the low hcn / hco@xmath0 ratio discussed in section 3.2.4 .
clearly from these ratios the filament is the densest part of the cloud .
the clump by clump ratios of 12 - 14.5 is very similar to what we see with the @xmath9co / hco@xmath0 magma data ( section 3.2.1 ) , giving further evidence that the large scale measurements are being dominated by clump / filament properties and not by a diffuse co medium ( see section 3.2.3 ) .
the hco@xmath0 velocity profile of the linear feature can be determined by taking a position - velocity slice along its structure ( figure 6 ) .
the position - velocity diagram shows almost no change in velocity , hence the clumps are either moving uniformly along the structure with the same speed and direction or its in the plane of the sky . the dense gas in 30dor-10 may potentially be a portion of a filament extending outward from the star cluster , r136 .
a tight connection between filaments and star formation in galactic dense cores has been revealed by the herschel gould belt survey ( andr et al .
this survey supports an emerging picture of star formation , in which thin , long ( @xmath21pc scales ) filaments form first a molecular cloud and then the densest part of the filaments fragment into protostellar cores via gravitational instability ( inutska et al . 1997 ) . for non - star forming regions , like polaris , all filaments have subcritical mass per unit length .
this gives an explanation for the extinction threshold for the formation of prestellar cores of av @xmath2 7 observed in aquila and polaris .
andr et al .
( 2010 ) were able to determine that the extinction threshold corresponds to the threshold above which interstellar filaments are gravitationally unstable and can collapse to form stars .
they found that @xmath21 60% of the bound prestellar cores and class 0 protostars in aquila are concentrated in gravitationally unstable filaments where the mass per unit length exceeds the critical mass per unit length ( eqn .
6 ) required for hydrostatic equilibrium . while 30dor-10 is significantly more distant than the molecular clouds studied by andr et al . ( 2010 ) , the structure of the bulk of hco@xmath0 emission in 30dor-10 bears a resemblance to a filament ( figures 1 and 2 ) .
this structure has a y - shaped morphology , which could be one bifurcated filament or two filaments interacting .
if the structure is a filament , we consider whether it is dense enough to be gravitationally unstable .
the clump critical mass per unit length depends only on the temperature ( ostriker 1964 ) , @xmath52 where @xmath53 is the isothermal sound speed , and @xmath54 is the gas temperature in units of 10k .
we estimate the filament s observed clump mass per unit length . adopting @xmath55k for dense clumps ( section 3.1.1 ) we get a clump critical mass per unit length for the filamentary structure detected in 30dor-10 of @xmath56/pc . from our observations and the cprops algorithm we know the length of the filamentary structure and the total calculated molecular gas mass for each clump . with these values the virial mass per unit length of m@xmath57/pc which is much larger than the critical clump mass per unit length . even adopting instead , m@xmath58 and t@xmath6100 k as extremes ,
the filamentary structure would still be above the critical mass per unit length .
this confirms the filamentary structure is gravitationally unstable and can collapse to form stars .
the linear feature is indeed forming stars and there is some evidence that star formation appears to evolve along the structure .
as mentioned in indebetouw et al .
( 2013 ) , the compact star cluster k1 in the lmc is located at the tip of what they refer to as `` pillars of creation '' ( clump 4 , figure 1(d ) ) . in hco@xmath0(1 - 0 ) , we observe the dense gas associated with this pillar as a filament extending outward from the star cluster , r136 , continues back into the co cloud ( figure 2 ) . toward 30dor-10 ,
h@xmath1o masers and ysos tends to correlate with the positions of hco@xmath0 clumps along the filament .
additional ysos are observed to extend back toward r136 along the same line .
the presence of ysos and h@xmath1o masers demonstrates that hco@xmath0 cores are currently forming stars .
this suggests that the 30dor-10 clumps may be slightly younger than those near r136 and we are seeing the dispersal of molecular gas in that part of the filament . the clumps in figure 1 ,
do not appear to be randomly spaced .
the clumps appear to be spaced sparsely but roughly uniformly along the filament .
the idea that clumps along a filament are uniformly spaced is discussed in jackson et al .
( 2010 ) for the galactic infrared dark cloud ( irdc ) known as the `` nessie '' nebula .
these authors suggest stellar cluster formation arises from the fragmentation of filaments due to a varicose fluid instability . in the jackson et al .
( 2010 ) picture , the clumps should have spacings equal to the wavelength ( @xmath59 ) of the fastest growing unstable mode of a self - gravitating fluid . for the case of an infinite isothermal gas cylinder @xmath60
corresponds to 22 h@xmath61 , where h@xmath61 is the effective scale height , @xmath62 and @xmath63 is the linewidth .
a self - gravitating fluid usually assumes thermal pressure to be the dominant source of gas pressure .
however , in our case turbulent pressure dominates over thermal ( @xmath64c@xmath65 ; figure 3c ) .
the average value of the linewidth for the clumps in 30dor-10 is 3.72 km s@xmath17 , and assuming the same central volume density as in the `` nessie '' filament , @xmath66@xmath210@xmath67@xmath68 we get a @[email protected] .
observationally , the clumps along the filament are spaced @xmath2 4.8 - 7.0pc apart which is consistent with the implied @xmath59 . in summary ,
the morphology and dynamics appear consistent with the string of clumps in 30dor-10 being a gravitationally unstable filament with some evidence that the phase of star formation occurring in it evolves .
changes in the l(hcn)/l(hco@xmath0 ) intensity ratios from unity found in high metallicity starburst galaxies have been variously interpreted as due to changes in dense gas density ( meijerink et al .
2007 ; baan et al . 2008 ; krips et al . 2008 ; meier & turner 2012 ) or to changes in the nature of the ionization rate or source ( e.g. uv , x - rays , cosmic rays ) ( kohno et al .
2001 ; graci - carpio et al .
2006 ; meijerink et al .
2007 ) . in higher metallicity environments ,
the l(hcn)/l(hco@xmath0 ) ratio varies from 0.5-@xmath694 with a median value of @xmath70 ( e.g. graci - carpio et al .
2006 ; baan et al . 2008 ;
krips et al . 2008 ) .
the fact that the l(hcn)/l(hco@xmath0 ) ratio is @xmath21 in high metallicity starbursts likely results from both transitions being optically thick and thermalized . since hco@xmath0 has a critical density almost an order of magnitude lower than hcn ( graci - carpio et al .
2006 ; krips et al . 2008 ) , if the density drops below @xmath210@xmath71@xmath68 the l(hcn)/l(hco@xmath0 ) ratio can drop below unity even if optically thick . the effect will be more pronounced if the opacity of hcn and hco@xmath0 are lower in the lmc than in these high metallicity starbursts .
thus the small l(hcn)/l(hco@xmath0 ) may suggest that both gas densities and opacities of hcn and hco@xmath0 are significantly lower in these lmc gmcs compared to higher metallicity systems .
this is reasonable given the lower column densities and lower metallicities of the clumps .
furthermore , there is evidence that the n / o elemental abundance is @xmath22 - 3 times lower in the lmc than in solar metallicity galactic environments ( e.g. hunter et al .
if hcn and hco@xmath0 abundances reflect the n and o abundances , respectively , this may further explain a portion of the low hcn found in the lmc .
compared to the l(hcn)/l(hco@xmath0 ) ratio in higher metallicity starburst environments , the ratio of the entire 30dor-10 gmc is much lower ( @xmath2 0.2 - 0.3 ) .
the @xmath24 times lower value in the 30dor-10 ( as well as the other lmc gmcs ) , suggests that the dense gas physical or chemical conditions in the lmc differ from those in massive starburst galaxies
. this could be due to the lower density or n / o abundance ratio in the lmc .
interestingly , there is also no significant difference between the l(hcn)/l(hco@xmath0 ) ratio in 30dor-10 and the other sampled lmc gmcs ( fig 3d ) .
evidently the l(hcn)/l(hco@xmath0 ) ratio is not strongly sensitive to radiation field conditions .
this clearly argues against explanations that solely depend on the strength of the radiation field environment .
atca observations of high dense gas tracers hco@xmath0(1 - 0 ) and hcn ( 1 - 0 ) in 30dor-10 were compared to four other star forming gmcs ( n159 , n113 , n105a and n44 ) in the lmc . along with the comparison of alma co and magma co data .
our results are discussed below . * applying cprops we detect 13 hco@xmath0 ( 1 - 0 ) bright clumps four of which are detected in hcn(1 - 0 ) .
the clumps exhibit similar masses and sizes but broader linewidths , compared to clumps in the other gmcs in the lmc . for 30dor-10
the ratio of @xmath50m@xmath25/@xmath50m@xmath26 @xmath2 6 which is three times the value found for the other four gmcs . * all of the results discussed suggest that substructure in 30dor-10 is not significantly different in mass , dense gas fraction or flux from the other lmc gmcs despite the extreme uv radiation field that surrounds 30dor-10 .
it appears that pdrs may not have a dramatic effect on the molecular cloud even with the presence of the super - star cluster , r136 .
* we detect a linear structure that is a potential gravitationally unstable filament and likely collapsing to form stars .
the morphology of the filament is in agreement with predictions by the varicose fluid instability where the clumps have roughly uniform spacings .
morphology and dynamics suggest that the filament is lying approximately in the plane of the sky . toward 30dor-10 ,
h@xmath1o masers and ysos tends to correlate with the positions of hco@xmath0 clumps along the filament .
additional ysos are observed to extend back toward r136 along the same line .
the presence of ysos and h@xmath1o masers demonstrates that hco@xmath0 cores are currently forming stars .
this suggests that the 30dor-10 clumps may be slightly younger than those near r136 and we are seeing the dispersal of molecular gas in that part of the filament .
part of this research was conducted at the jet propulsion laboratory , california institute of technology under contract with the national aeronautics and space administration .
part of this project was funded by nsf grant ast-1009620 to dsm .
this research has made use of the simbad database , operated at cds , strasbourg , france .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . 1 & - & 05:38:55.09 & -69:04:16.9 & 2.9(0.6 ) & @xmath201.0 & @xmath202.3 & @xmath201.5 & @xmath208.4 & 2.1(0.5 ) & 0.4(0.5 ) & 240.6 + 2 & - & 05:38:52.17 & -69:04:06.1 & 2.5(0.6 ) & @xmath202.0 & @xmath202.6 & @xmath203.7 & 6.7(4.8 ) & 1.4(0.5 ) & 0.2(0.2 ) & 243.0 + 3 & 83,85,87 - 94 & 05:38:50.73 & -69:04:21.9 & 3.1(0.4 ) & @xmath60.7 ( 0.5 ) & @xmath62.7(1.2 ) & @xmath200.3 & 14.6(7.4 ) & 1.7(0.4 ) & 0.1(0.06 ) & 246.2 + 4 & 4 - 7,9 - 10 & 05:38:45.24 & -69:05:05.7 & 4.5(0.3 ) & @xmath200.9 & @xmath65.5 ( 4.2 ) & @xmath202.4 & 28.2(4.1 ) & 3.0(1.0 ) & 0.1(0.04 ) & 248.9 + 5 & 11 - 14,21 - 28,30 - 32,34 & 05:38:47.02 & -69:04:58.9 & 3.4(0.2 ) & @xmath60.7 ( 0.2 ) & @xmath63.2 ( 1.0 ) & 0.3(0.1 ) & 30.3(3.2 ) & 5.4(1.3 ) & 0.2(0.05 ) & 250.3 + 6 & 35,50 - 55,57 - 58 & 05:38:49.03 & -69:04:43.8 & 6.1(0.1 ) & 1.1 ( 0.2 ) & 7.6(2.5 ) & 1.5 ( 0.1 ) & 81.4(6.8 ) & 16.9(2.4 ) & 0.2(0.03 ) & 251.0 + 7 & 44 - 47 & 05:38:45.08 & -69:04:39.0 & 2.7(0.3 ) & @xmath201.6 & @xmath202.1 & @xmath204.7 & 21.4(3.4 ) & 3.0(0.8 ) & 0.1(0.04 ) & 251.4 + 8 & 36 - 43,49 & 05:38:46.80 & -69:04:42.6 & 4.4(0.3 ) & 0.9 ( 0.3 ) & 3.3(2.6 ) & 0.5(0.2 ) & 25.8(3.8 ) & 5.5(1.1 ) & 0.2(0.05 ) & 254.0 + 9 & - & 05:38:51.34 & -69:03:15.4 & 1.8(1.6 ) & @xmath200.7 ( 0.6 ) & @xmath201.5 & @xmath200.3 & @xmath2010.1 & 0.1(0.1 ) & 0.02(0.02 ) & 247.8 + 10 & 71 - 79 & 05:38:52.87 & -69:04:36.8 & 6.9(0.1 ) & @xmath60.7 ( 0.2 ) & @xmath612.9(1.9 ) & 0.7(0.1 ) & 90.3(6.0 ) & 26.3(3.7 ) & 0.3(0.05 ) & 253.9 + 11 & - & 05:38:47.47 & -69:04:05.1 & 3.0(0.2 ) & 1.4 ( 0.4 ) & 2.4(1.1 ) & 0.9 ( 0.1 ) & 33.9(4.6 ) & 7.0(1.5 ) & 0.2(0.05 ) & 248.1 + 12 & - & 05:38:48.27 & -69:03:35.5 & 3.1(0.5 ) & 1.6 ( 0.9 ) & 2.8(3.6 ) & 0.7 ( 0.2 ) & 18.9(6.8 ) & 1.6(0.6 ) & 0.1(0.04 ) & 247.8 + 13 & - & 05:38:56.40 & -69:04:17.4 & 4.1(0.2 ) & @xmath200.8 & @xmath64.5 ( 2.9 ) & @xmath202.2 & 33.1(4.0 ) & 8.3(1.9 ) & 0.2(0.06 ) & 250.6 + r@xmath43 & pc & 28.7(0.05 ) & 14.0(0.19 ) & 18.3(0.12 ) & 9.6(0.50 ) & 12.5(0.33 ) + @xmath72v@xmath43 & kms@xmath17 & 4.6(0.07 ) & 2.4(0.17 ) & 2.6(0.08 ) & 2.5(0.19 ) & 2.2(0.16 ) + l@xmath43 & 10@xmath67kkms@xmath17pc@xmath3 & 1.7(0.03 ) & 1.9(0.25 ) & 1.7(0.03 ) & 0.6(0.18 ) &
1.8(0.23 ) + m@xmath43 & 10@xmath67m@xmath73 & 7.5(0.1 ) & 21.0(1.1 ) & 7.6(0.1 ) & 2.8(0.8 ) & 7.8(1.0 ) + # of hco@xmath0 clumps & & 13 & 27 & 14 & 6 & 8 + @xmath74 & 10@xmath3kkms@xmath17pc@xmath3 & 4.0(1.5 ) & 8.1(2.2 ) & 7.1(2.6 ) & 2.6(1.5 ) & 4.5(2.0 ) + @xmath75 & 10@xmath3kkms@xmath17pc@xmath3 & 0.8(0.1 ) & 1.5(0.1 ) & 2.4(0.2 ) & 0.8(0.1 ) & 0.8(0.1 ) + l@xmath76/l@xmath36 & & 0.2(0.1 ) & 0.2(0.1 ) & 0.3(0.1 ) & 0.3(0.2 ) & 0.2(0.1 ) + @xmath77 & 10@xmath67m@xmath73 & 3.4(0.9 ) & 5.0(1.7 ) & 1.6(0.6 ) & 1.4(0.8 ) & 4.4(2.2 ) + @xmath78 & 10@xmath67m@xmath73 & 0.6(0.7 ) & 2.1(0.3 ) & 1.1(0.9 ) & 0.7(0.1 ) & 1.9(0.2 ) + @xmath72v@xmath79 & kms@xmath17 & 3.9 & 2.6 & 2.0 & 2.2 & 2.1 + @xmath80 & & 0.02 & 0.08 & 0.04 & 0.04 & 0.03 + radiation field & @xmath81 & 562 & 156 & 97 & 83 & 76 +
|
we present parsec scale interferometric maps of hcn(1 - 0 ) and hco@xmath0(1 - 0 ) emission from dense gas in the star - forming region 30 doradus , obtained using the australia telescope compact array .
this extreme star - forming region , located in the large magellanic cloud ( lmc ) , is characterized by a very intense ultraviolet ionizing radiation field and sub - solar metallicity , both of which are expected to impact molecular cloud structure .
we detect 13 bright , dense clumps within the 30 doradus-10 giant molecular cloud .
some of the clumps are aligned along a filamentary structure with a characteristic spacing that is consistent with formation via the varicose fluid instability .
our analysis shows that the filament is gravitationally unstable and collapsing to form stars .
there is a good correlation between hco@xmath0 emission in the filament and signatures of recent star formation activity including h@xmath1o masers and young stellar objects ( ysos ) .
ysos seem to continue along the same direction of the filament toward the massive compact star cluster r136 in the southwest .
we present detailed comparisons of clump properities ( masses , linewidths , sizes ) in 30dor-10 to those in other star forming regions of the lmc ( n159 , n113 , n105 , n44 ) .
our analysis shows that the 30dor-10 clumps have similar mass but wider linewidths and similar hcn / hco@xmath0 ( 1 - 0 ) line ratios as clumps detected in other lmc star - forming regions .
our results suggest that the dense molecular gas clumps in the interior of 30dor-10 are well - shielded against the intense ionizing field that is present in the 30doradus region .
|
neutrino physics has become one of the important areas of intense theoretical and experimental efforts .
this is because neutrinos are instrumental in giving answer to some of the basic questions in cosmology , astro- , nuclear , and particle physics .
the @xmath1 scattering processes have been studied by several authors for the quasielastic , 1 pion production and for deep - inelastic scattering processes @xcite-@xcite . however , there are very few works where neutrino induced strange particle production have been been studied @xcite-@xcite .
these processes are important for the analysis of the precise determination of neutrino oscillation parameters . in the few - gev region
it allows the detailed study of the strange - quark content of the nucleon and gives some important information about the structure of the hadronic weak current .
apart from that in the atmospheric neutrino analysis these processes gives @xmath2 backgrounds for nucleon - decay searches .
the antineutrino induced @xmath3 single - hyperon production can give us useful information about the weak form - factors . a precise measurement of the hyperon cross - section specially @xmath4 distribution will be useful in the determination of axial form factors
. a better understanding of these processes will give more strength to the basic understanding of v - a and cabibbo theories .
recently miner@xmath0a is taking data in its first phase of experiment with high statistics to explore the strange physics .
there is also probability of getting events of single kaons in the various beta - beam experiments in the energy region of 1gev .
also t2k is planning to run phase - ii in antineutrino mode and the no@xmath0a experiments where the information about the single hyperon and single kaon production may be obtained .
strange particle production via the weak interaction were initially studied by shrock @xcite , mecklenburg @xcite and dewan @xcite .
shrock @xcite and mecklenburg @xcite independently studied the associated production of charged current ( cc ) reactions by employing the cabibbo theory with su(3 ) symmetry .
amer @xcite used harmonic oscillator quark model to estimate cross section for some of the associated production process .
dewan @xcite studied the cc and strangeness changing ( @xmath5s = 1 ) strange particle production reactions . recently singh and vicente vacas @xcite have studied hyperon production induced by antineutrinos from nucleons and nuclei , rafi alam et al .
@xcite have studied single kaon production and adera et al .
@xcite have studied differential cross section for @xmath0 induced c. c. associated particle production .
most of the earlier neutrino experiments ( 1970 s and 80s ) were performed using bubble chambers where cross - sections for many associated - production and @xmath6 reactions were obtained using bubble chambers filled with freon and/or propane or with deuterium . however , the data are statistically limited with large error bars@xcite-@xcite . in this work ,
we have presented the results of our calculations for single kaon produced induced by neutrino / antineutrino reactions . in sect.ii , we present the formalism in brief and in sect.iii , the results and discussions are presented .
the basic reaction for the @xmath7 induced charged current kaon production is @xmath8 where @xmath9 and @xmath10=n , p .
the expression for the differential cross section in lab frame for the above process is given by , @xmath11 where @xmath12 and @xmath13 are the 3-momenta of the incoming and outgoing leptons in the lab frame with energy @xmath14 and @xmath15 respectively .
the kaon lab momentum is @xmath16 having energy @xmath17 , @xmath18 is the nucleon mass , @xmath19 is the square of the transition amplitude matrix element averaged(summed ) over the spins of the initial(final ) state . at low energies , this amplitude can be written in the usual form as @xmath20 where @xmath21 and @xmath22 are the leptonic and hadronic currents respectively , @xmath23 is the fermi constant and @xmath24 is the gauge coupling . the leptonic current can be readily obtained from the standard model lagrangian coupling the @xmath25 bosons to the leptons @xmath26\ ] ] , scaledwidth=80.0% ] , scaledwidth=80.0% ] in the case of neutrino induced kaon production process , we have considered four different channels that contribute to the hadronic current .
they are depicted in fig .
[ fig : feynman_neutrino ] .
there is a contact term ( ct ) , a kaon pole ( kp ) term , a u - channel process with a @xmath27 or @xmath28 hyperon in the intermediate state and finally a meson ( @xmath29 ) exchange term . for the specific reactions under consideration
, there is no s - channel contributions given the absence of @xmath30 baryonic resonances .
kp term is proportional lepton mass and therefore its contribution is very small . while in the case of antineutrino induced kaon production process , besides the processes mentioned for the neutrino case , there are contributions from s - channel @xmath31 propagator , s - channel @xmath32 resonance terms(fig .
[ fig : feynman_antineutrino ] ) . the contribution of the different terms can be obtained in a systematic manner using chiral perturbation theory ( @xmath33pt ) .
the lowest - order su(3 ) chiral lagrangian describing the pseudoscalar mesons in the presence of an external current is @xcite : @xmath34 + \frac{f_\pi^2}{4}\mbox{tr}(\chi u^\dagger + u\chi^\dagger),\ ] ] where the parameter @xmath35mev is the pion decay constant , @xmath36 is the su(3 ) representation of the meson fields @xcite and @xmath37 is its covariant derivative . and ( right panel )
@xmath38,title="fig:",scaledwidth=50.0% ] and ( right panel ) @xmath38,title="fig:",scaledwidth=50.0% ] the lowest - order chiral lagrangian for the baryon octet in the presence of an external current can be written in terms of the baryon su(3 ) matrix as @xcite : @xmath39 -\frac{d}{2 } \mbox{tr}\left ( \bar b \gamma^\mu \gamma_5 \ { u_\mu , b \ } \right ) -\frac{f}{2}\mbox{tr}\left ( \bar b \gamma^\mu\gamma_5 \left [ u_\mu , b \right ] \right),\ ] ] where @xmath18 denotes the mass of the baryon octet , b is the baryon su(3 ) matrix and the parameters @xmath40 and @xmath41 which are determined from the semileptonic decays . and ( right panel )
@xmath42,title="fig:",scaledwidth=50.0% ] and ( right panel ) @xmath42,title="fig:",scaledwidth=50.0% ] at intermediate energies , we found that the weak excitation of the @xmath43 resonances and its subsequent decay in @xmath44 is also important . to calculate amplitudes associated with @xmath32 we first parameterize the @xmath45 .
for this , we can write the most general form of the vector and axial - vector matrix element as , @xmath46 where @xmath47\gamma_5 \nonumber\\ \gamma^{\alpha\mu}_a ( p , q ) & = & \left [ \frac{c_3^a}{m}\left(g^{\alpha\mu } \slashchar{q}- q^\alpha\gamma^\mu\right ) + \frac{c^a_4}{m^2 } \left(g^{\alpha\mu } q\cdot
p- q^\alpha p^\mu\right ) + c_5^a g^{\alpha\mu } + \frac{c_6^a}{m^2 } q^\mu q^\alpha \right ] .
\label{eq : del_ffs } \end{aligned}\ ] ] in the above expression @xmath48 are the @xmath49 dependent scalar and real vector and axial vector form factors and @xmath50 is the rarita - schwinger spinor .
it is the @xmath51 term which is most dominant and we have considered only the terms with @xmath51 in the present work .
the spin 3/2 propagator in the momentum space is given by , @xmath52 where @xmath53 is the spin 3/2 rarita - schwinger projection operator and @xmath54 is the resonance mass ( @xmath55 ) .
the @xmath32 decay width @xmath56 is around 36@xmath575mev , however , we have taken p - wave decay width which is given as @xmath58 where @xmath59 is callen lambda function and @xmath60 is the step function . @xmath61
is the kn@xmath32 coupling strength taken to be 1 in the present calculation . at @xmath62 for single kaon production induced by neutrinos .
the curves are labeled according to the final state of the process .
( right panel ) cross sections as a function of the neutrino energy for single kaon production vs. associated production obtained with genie @xcite .
, title="fig:",scaledwidth=50.0% ] at @xmath62 for single kaon production induced by neutrinos .
the curves are labeled according to the final state of the process .
( right panel ) cross sections as a function of the neutrino energy for single kaon production vs. associated production obtained with genie @xcite .
, title="fig:",scaledwidth=50.0% ]
the total scattering cross section @xmath63 has been obtained by using eq .
( [ d9_sigma ] ) after integrating over the kinematical variables . in the left panel of fig .
( [ fig : xsec_pp_nu ] ) , we present the results of the contributions of the different diagrams to the total cross sections for the neutrino induced process .
the kaon pole contributions are negligible at the studied energies and are not shown in the figures although they are included in the full model curves .
we observe the relevance of the contact term , not included in previous calculations .
we find that the contact term is in fact dominant , followed by the u - channel diagram with a @xmath28 intermediate state and the @xmath64 exchange term .
as observed by dewan @xcite the u - channel @xmath27 contribution is much less important , basically because of the larger coupling ( @xmath65 ) of the strong vertex .
the curve labeled as full model has been calculated with a dipole form factor with a mass of 1 gev .
the band corresponds to changing up and down this mass by a 10 percent . on the right panel of fig .
( [ fig : xsec_pp_nu ] ) , we have presented the results for antineutrinos : @xmath66 .
we find that the contact term is the most dominant one followed by pion in flight and the s - channel diagram with @xmath32-resonance and @xmath28 as the intermediate states .
the suppression of @xmath27 as the intermediate state is due to the difference in the coupling strength @xmath67 and in the @xmath68 in flight due to @xmath69 .
we also checked the effects of the @xmath70 resonance at the said energies .
we find that unlike the @xmath71 dominance in pion production the contribution of @xmath32 is not too large . in fig .
( [ fig : xsec_nn_nu ] ) , corresponding results for @xmath72 and @xmath73 processes are shown . in fig .
( [ fig : xsec_dq2_genie ] ) , we have shown the results for the @xmath4 distribution for the three studied channels at a neutrino energy e@xmath74 gev .
the reactions are always forward peaked ( for the final lepton ) , even in the absence of any form factor ( @xmath75 ) , favoring relatively small values of the momentum transfer . on the right panel of the fig .
( [ fig : xsec_dq2_genie ] ) , we have compared our results for the @xmath76 process with the values for the associated production obtained by means of the genie monte carlo program @xcite .
we observe that , due to the difference between the energy thresholds , single kaon production for the @xmath77 is clearly dominant for neutrinos of energies below 1.5 gev . for the other two channels associated production becomes comparable at lower energies .
still , single @xmath78 production off neutrons is larger than the associated production up to 1.3 gev and even the much smaller @xmath79 production off neutrons is larger than the associated production up to 1.1 gev .
the consideration of these @xmath3 channels is therefore important for the description of strangeness production for all low energy neutrino spectra and should be incorporated in the experimental analysis .
9 s. boyd , s. dytman , e. hernandez , j. sobczyk and r. tacik , aip conf .
* 1189 * ( 2009 ) 60
. l. alvarez - ruso , arxiv:1012.3871 r. e. shrock , phys . rev .
d * 12 * , 2049 ( 1975 ) .
w. mecklenburg , acta phys .
austriaca 48 , 293 ( 1978 ) .
a. a. amer , phys .
d * 18 * , 2290 ( 1978 ) .
h. k. dewan , phys . rev . * d24 * , 2369 ( 1981 ) .
s. k. singh and m. j. vicente vacas , phys .
d * 74 * , 053009 ( 2006 ) .
m. rafi alam , i ruiz simo , m. sajjad athar and m. j. vicente vacas , phys . rev .
d * 82 * , 033001 ( 2010 ) .
g. b. adera , b. i. s. van der ventel , d. d. van niekerk and t. mart , phys .
c * 82 * , 025501 ( 2010 ) .
h. deden et al .
58b * , 361 ( 1975 ) . o. erriquez _ et al .
_ , nucl .
b * 140 * , 123 ( 1978 ) .
barish et al . , phys .
* 33 * , 1446 ( 1974 ) n. j. baker et al .
, phys . rev .
* d23 * , 2499 ( 1981 ) .
baker et al .
, phys . rev . * d24 * , 2779 ( 1981 ) . v. v. ammosov _ et al .
_ , jetp lett .
* 39 * , 209 ( 1984 ) [ pisma zh .
39 * , 176 ( 1984 ) ] .
mann _ et al .
d * 34 * , 2545 ( 1986 ) .
c. andreopoulos _ et al .
. instrum .
a * 614 * ( 2010 ) 87 .
|
the weak kaon production off the nucleon induced by neutrinos and antineutrinos is studied at low and intermediate energies of interest for some ongoing and future neutrino oscillation experiments .
we develop a microscopical model based on the su(3 ) chiral lagrangians .
the studied mechanisms are the main source of kaon production for neutrino energies up to 2 gev for the various channels and the cross sections are large enough to be amenable to be measured by experiments such as minerva , t2k and no@xmath0a .
|
the concept of rigidity percolation , first introduced about 25 years ago by thorpe @xcite based on work by phillips @xcite , describes how an elastic network goes from floppy to rigid as constraints are added to it .
this theory has been applied with success to many systems , including covalent glasses @xcite and proteins @xcite . in the last decade , however , experimental studies have shown that the rigidity phase diagram could be more complex than initially thought , uncovering the presence of an intermediate phase between the floppy phase and the stressed - rigid phase , with the system in the intermediate phase being rigid but unstressed @xcite . a basic explanation for this new phase was first proposed by thorpe _
et al . _
it was shown that when the network self - organizes in order to minimize the stress , the rigid but unstressed intermediate phase can indeed arise . in the original work by thorpe and collaborators @xcite , as well as subsequent simplified models of chalcogenide glasses by micoulaut and phillips @xcite
, networks were constructed using an `` aggregation '' process , in which bonds or simple network units were added to the network without subsequent equilibration .
more recently , barr _ et al . _
@xcite have considered a thermodynamically proper model with an energy cost associated with stress and showed that in the canonical ensemble , the intermediate phase still exists . in a recent paper , we have confirmed this result for the @xmath0 version of the model of barr _ et al .
_ , but on a more realistic regular lattice , and also have shown that the intermediate phase is entropically feasible in actual physical systems . in both the model by barr _
et al . _ and our model
, the intermediate phase has an unusual property : both percolating and non - percolating networks coexist in the ensemble at all mean coordination numbers within the intermediate phase .
while providing a general support for self - organization , these previous studies did not look in detail at the properties of networks in the intermediate phase . here
we provide a first glimpse at some of these properties .
first , we show that in both percolating and non - percolating networks , the sizes of non - percolating clusters have a power - law distribution . in effect , the system remains in a critical state over an extended range of mean coordinations , corresponding to a self - organized critical phase @xcite , but in an equilibrium system .
second , we find that adding or removing a single bond affects the rigidity of macroscopic parts of the network in the intermediate phase and in particular , can turn a non - percolating network into a percolating one and vice versa .
this property supports the interpretation by wang and co - workers @xcite of the puzzling response of vibrational frequencies to applied pressure that was observed in their experiments .
using our results , we address some intriguing questions that have to do with the unusual coexistence of percolating and non - percolating networks in the intermediate phase . in particular , we show that percolating and non - percolating networks can be considered identical in the intermediate phase , as both stay on the edge of percolation .
this paper is structured as follows . in the next section ,
we briefly review the intermediate phase .
we then present our methodology . in the fourth section
, we present our results on the properties of rigid clusters ( both percolating and non - percolating ) in our model . in section
v , we look at the response of the network to local perturbations . in section
vi , we discuss how our results can help understand the experiments by wang _
et al . _
we finish with conclusions .
using the mean - field approach first introduced by maxwell @xcite and known as maxwell counting , we can define the rigidity of a network in terms of its number of zero - frequency motions , or _ floppy modes _
, @xmath1 . in a @xmath2-dimensional network ,
each atom has @xmath2 degrees of freedom . in an unconstrained network of @xmath3 atoms ,
each degree of freedom corresponds to a floppy mode and thus @xmath4 . assuming that each added constraint takes away one floppy mode , we can write @xmath5 where @xmath6 is the number of constraints in the network . as @xmath6 increases , @xmath1 decreases , until @xmath7 is reached , and then there are no floppy modes left and the network undergoes a rigidity transition from floppy to rigid where a percolating rigid cluster emerges in the network ( rigid clusters are sets of mutually rigid atoms ) disordered networks can be conveniently characterized by the _
mean coordination _
@xmath8 , which is the average number of bonds connecting a site . in the mean - field approximation ,
the location of the rigidity transition only depends on @xmath8 and not on other details , such as fractions of sites with a particular coordination .
the transition occurs at @xmath9 for the triangular lattice of elastic springs and @xmath10 in chalcogenide glasses ; in the latter case , it is assumed that both bond - stretching and bond - bending constraints are taken into account . to go beyond the mean - field theory ,
corrections must be made .
for example , adding a constraint to an already rigid region does not remove a floppy mode .
such type of constraint is called redundant .
redundant constraints introduce _ stress _ into the network .
such constraints do not change the number of floppy modes ( and thus violate the assumption of the maxwell counting ) ; taking this into account , eq .
( [ eq:1 ] ) becomes @xmath11 where @xmath12 is the number of redundant constraints .
another type of correction is due to the fact that even above the rigidity transition there can still be some floppy inclusions and thus @xmath13 at the transition . to find @xmath12 , jacobs and co -
workers have introduced a topological algorithm , the _ pebble game _ @xcite .
this algorithm is based on a theorem by laman @xcite which states that in two dimensions a generic network with @xmath3 sites and @xmath14 bonds does not have a redundant bond if and only if no subset of the network containing @xmath15 sites and @xmath16 bonds violates @xmath17 .
a similar approach works in 3d , but in general only for bond - bending networks like those used to model chalcogenide glasses . the pebble game , described in more detail in ref .
@xcite , characterizes the global rigidity of a network , provides its complete decomposition into rigid clusters and finds stressed regions .
the approach uses only the topology of the network and not its exact geometry . using the pebble game , it was possible to show that the rigidity transition occurs at @xmath18 @xmath190.001 @xcite for the central - force triangular lattice and at @xmath20 @xcite for an amorphous bond - bending 3d network . while early measurements also identified a rigidity transition in these glasses near @xmath10 , recent experiments have shown that there is not one but two transitions @xcite .
starting at low mean coordination , the first transition seems to go from a floppy to a rigid but unstressed phase , and the second one on to a rigid and stressed phase .
this new rigid but unstressed phase is known as the _ intermediate phase _ and is believed to be related to the self - organization of the system in order to minimize the stress in the network .
this interpretation is supported by a number of theoretical works . in their pioneering work , thorpe and
collaborators demonstrated that a network constructed by adding bonds with no stress allowed until it becomes inevitable would go through three phases : a floppy , a rigid - unstressed and a rigid - stressed phases @xcite .
the first transition , between the floppy and the rigid - unstressed phases , is the _ rigidity transition _ ; the second one , between the rigid - unstressed and rigid - stressed phases , is the _ stress transition _ and happens immediately when avoiding stress is no longer possible , the intermediate phase survives when the rigid but stress - free networks are fully equilibrated as was demonstrated by barr and co - workers on bethe lattices @xcite and chubynsky _ et al . _
@xcite on two - dimensional triangular lattices .
this is not the only approach to self - organization , however , and micoulaut and collaborators @xcite have shown that it is possible to recover an intermediate phase in a stressed network if this stress is localized .
the model we study here is the same as the one used in our previous paper @xcite . in our simulations
, we use the pebble game algorithm described in the previous section ; our computer code is based on the original program by d.j .
jacobs and m.f .
thorpe .
we consider 2d triangular bond - diluted central - force networks . while we can not make a direct comparison with experiment
, previous work has shown that the triangular lattice presents an intermediate phase similar to that of covalent glasses with angular constraints ; our results should therefore be applicable , at least qualitatively , to experiments . in the original model of thorpe and collaborators
@xcite , bonds were added one by one and each checked for redundancy ; redundant ( stress - causing ) bonds were rejected .
each new added bond was frozen in the network and was never moved nor removed .
this procedure does not guarantee that stress - free networks are equiprobable , making some networks more likely than others . to eliminate this bias , we introduced a bond - equilibration scheme , allowing the system to rearrange itself by moving bonds around . each time
a new bond is added , bonds are reshuffled throughout the lattice : a bond is picked at random , removed , and then a new bond is inserted in a random place choosing among those where it would not create stress .
this bond - shuffling procedure is repeated until the system is equilibrated .
we find that an equilibration of 10 iterations per added bond below @xmath8=3.5 and 100 iterations above @xmath8=3.5 is enough for convergence ( for more details , see chubynsky _ et al . _
@xcite ) and it is the equilibration scheme we use throughout this paper , unless stated otherwise . in equilibrium , all stress - free networks with a given number of bonds ( or mean coordination ) are equiprobable .
this corresponds to the thermodynamic equilibrium at @xmath21 for any model in which all stress - free networks have equal energy , but the energies of stressed networks are higher . as discussed in our previous paper @xcite , the intermediate phase in the model described above
is associated with a non - trivial probability of finding a percolating network in this equilibrated model : this probability rises linearly from zero at the rigidity transition ( @xmath22 ) to one at the stress transition ( @xmath23 ) , a result similar to that obtained by barr _
_ on the bethe lattice @xcite
. thus there are both percolating and non - percolating networks and these two classes need to be studied separately .
all results presented in this paper ( with the exception of the cluster size distributions given in fig . [ fig : clst_dist ] ) are obtained by running 200 independent simulations and obtaining the quantities of interest at different @xmath8 .
when overall averages are presented , these are obtained by averaging over all these simulations .
when , e.g. , an average for percolating networks is presented , then the averaging is done over only those of these networks that are percolating at the given @xmath8 .
obviously , close to the rigidity transition , very few of the 200 networks are percolating , and so the corresponding quantity will be an average over a very small number of realizations and may contain a bigger error . when we report the results of insertion or removal of a single bond , only one attempt of insertion / removal per network is made , unless stated otherwise .
in this section , we look at properties of rigid clusters ( both percolating and non - percolating ) that exist in self - organized networks in the intermediate phase . some properties of rigid clusters that distinguish them from , e.g. , clusters in usual connectivity percolation , need to be kept in mind . first , unlike in the connectivity case , a site can belong to more than one cluster ( it then serves as a pivot joint between the clusters sharing this site ) . on the other hand
, a bond in 2d always belongs to just one cluster ( not so in 3d , when it can serve as a hinge around which several clusters can rotate ) . for this reason ,
cluster decomposition in 2d is best expressed in terms of bonds and not sites .
while this is not generally so , in the case of self - organized networks in the floppy or intermediate phase , the conversion between cluster sizes expressed in terms of bonds or sites is easy : since there is no stress , there is also no redundancy , and every rigid cluster of @xmath15 sites contains exactly @xmath24 bonds . we start with properties of the percolating cluster . normally , when the probability of percolation is either zero or one in the thermodynamic limit , the exact definition of percolation does not matter . in our case , since the probability of finding a percolating cluster increases linearly with average coordination in the intermediate phase , it is not clear that we can be so cavalier . for example , the probability that a cluster that percolates _ in only one direction _ ( no matter which one ) seems to be non - zero in the intermediate phase , although it is very low , around 0.1 or less .
this means that the results will differ slightly if we define percolation using only one direction or both . to avoid confusion
, we choose here to call percolating those networks in which percolation occurs in both directions , and non - percolating those networks in which there is no percolation in either direction .
networks with percolation in just one direction are ignored whenever we separate our results into those for percolating networks and those for non - percolating ones , but such networks are taken into account when this separation is not done . while we have already studied the probability that a percolating cluster occurs in the intermediate phase , its size had not been characterized .
figure [ fig : perc_clst ] shows the average number of bonds in the percolating cluster as a fraction of bonds _ actually present _ in the network .
the averaging is done over all networks in which percolation occurs .
a remarkable feature is that even at the lowest @xmath8 at which percolation is still ( rarely ) observed , the size of the percolating cluster is well above zero .
the smallest cluster size observed at the onset of the intermediate phase is around 40% .
this behavior is different from that for both connectivity and rigidity percolation on regular lattices in the random ( non - self - organized ) case , where the size of the percolating cluster , considered a good order parameter , grows from zero at the transition , as expected in a second - order transition .
this result in the self - organized case is reminiscent of the _ first - order _ rigidity transition , such as that observed on bethe lattices .
however , our other results , as discussed below , do not support this analogy .
figure [ fig : perc_clst ] presents _ average _ sizes of percolating clusters .
given that the very existence of the percolating cluster is uncertain in the intermediate phase ( since only some networks are percolating ) , it is reasonable to ask about the variation of the percolating cluster size .
the quantity we look at is the standard deviation , or width , of the fraction of bonds in the percolating cluster calculated as @xmath25 , where @xmath1 is the fraction of bonds in the percolating cluster , @xmath26 denotes the average over percolating networks and @xmath15 is the number of percolating networks .
in ordinary percolation , this width , of course , tends to zero as the network size grows ; the percolating cluster size is a self - averaging quantity .
figure [ fig : perc_width ] shows that this is not so in our case .
the width is above zero and is essentially size - independent .
this is yet another difference from non - self - organized percolation ( including that on bethe lattices ) .
note at the same time that the width of the distribution of percolating cluster sizes is much smaller than the average size .
in other words , an overwhelming majority of networks have either a big percolating cluster or no percolating cluster at all there are few ( if any ) `` intermediate cases '' with small percolating clusters . to further characterize the intermediate phase , it is useful to look at the distribution of rigid cluster sizes . in case of a second - order phase transition ,
the correlation radius is finite away from the transition . as a result , at a certain cluster size
there is a crossover from a power - law behavior to an exponential behavior .
the correlation radius diverges as the transition is approached from either side , and so the crossover moves towards bigger sizes as the transition is approached , and exactly at the transition , the power law persists indefinitely . since the divergence of the correlation radius ( or the crossover point )
is governed by the same critical exponent on both sides of the transition , it is expected that for two values of @xmath8 at the same distance from the transition but on opposite sides , the crossover points will be the same by order of magnitude , if the transition is second - order . in case of a first - order transition , the correlation radius does not diverge at the transition and there is always a crossover to the exponential behavior .
figure [ fig : clst_dist ] shows the distribution of rigid cluster sizes at the coordinations of @xmath27 ( below the transition ) and @xmath28 ( above the transition ) .
the points are chosen at about the same distance from the rigidity transition , far enough from it to make sure that the correlation radius is not too big ( if the transition is second - order ) yet not too far to ensure that the points are still within the critical region .
all results are obtained by running 200 independent simulations on networks of 40000 sites , each starting from an already equilibrated network and continuing for 1000 additional equilibration steps .
after each equilibration step , the distribution of non - percolating cluster sizes is obtained .
while networks one equilibration step apart can not be considered truly independent , inserting or removing a single bond often changes the rigidity of the network very significantly , justifying inclusion of the data obtained at every step . for @xmath28 , the distributions for percolating and
non - percolating networks are plotted separately ( at @xmath27 , obviously , there are very few percolating networks and we plot the result for non - percolating networks only ) . to decrease the noise in the tail , all clusters of a given class ( i.e. , percolating and non - percolating ) are binned using a logarithmic scale .
below the rigidity transition , in the floppy phase [ @xmath27 ; fig .
[ fig : clst_dist],(a ) ] , there is a clear crossover between the power - law and exponential behaviors .
the data are fitted using the product of a power law and an exponential : @xmath29 we use the data between @xmath30 and @xmath31 , i.e. , dropping just a few data points at the tail , where noise and finite - size effects are significant , and omitting a region at the low end , as there are big deviations from the behavior described by eq .
[ eq : fit1 ] , probably due to discretization effects from the lattice .
the best fit is obtained with @xmath32 , @xmath33 and @xmath34 .
the dashed line is a power law with the same @xmath35 and @xmath36 , but without the exponential factor .
above the transition [ @xmath28 ; fig .
[ fig : clst_dist],(b ) ] , however , the power - law behavior persists at least for non - percolating networks , with no hint of the exponential tail , even for the largest sizes for which the data are available ( around 30000 , well above the crossover observed for @xmath27 around @xmath34 ) . for percolating nets
, there is some deviation from the power law near the end , but it is likely due to finite - size effects ( there is a percolating cluster taking up most of the network , so only relatively small non - percolating clusters are possible ) .
to fit the data , we use pure power - law functions : @xmath37 where subscripts @xmath38 refer to non - percolating and percolating networks , respectively .
we use data above @xmath30 in the non - percolating case and from @xmath39 to @xmath40 in the percolating case .
the values of the parameters providing the best fits are : @xmath41 , @xmath42 , @xmath43 , @xmath44 .
while the difference between @xmath36 and @xmath45 is probably due to finite size and sampling effects , the difference between @xmath45 and @xmath46 suggests that these two quantities are different , an unusual behavior .
our results indicate that the power - law distribution of cluster sizes is observed in the whole intermediate phase , rather than at a single point ( as would be the case without self - organization ) . in effect , the self - organization , which minimizes the stress in the network , _ maintains the system in a critical state throughout the intermediate phase_. similarly , unlike in the case of the critical point of random percolation , the cluster size critical exponent appears to be different in the non - percolating and percolating cases .
[ panel ( a ) ] and in both percolating and non - percolating self - organized networks at @xmath28 [ panel ( b ) ] .
the details of the simulation and the fits ( lines ) are given in the text.,width=259 ] based on the results of this section , we see that the rigidity phase transition in our model of self - organized percolation is very different from both first- and second - order phase transitions .
the minimization of the stress appears to act similarly to an external driving force , leading to a self - organized critical phase in a thermodynamically equilibrated system .
given that both percolating and non - percolating networks coexist in the intermediate phase , it is interesting to investigate the relation between the two classes . as we show in this section , a local perturbation involving the addition or removal of even a single bond is enough to affect rigidity of huge regions of the network and often converts a non - percolating network into a percolating one and vice versa .
this behavior is not observed in regular random networks : since the probability of percolation in these systems is always either zero or one in the thermodynamic limit , a single bond can only change the percolation property right at the transition . since
a percolating network at that point is fractal and involves only an infinitesimal fraction of bonds , if an infinite cluster is created or destroyed , this can only involve an infinitesimal fraction of bonds and sites ; away from the transition , the size of the affected region is always finite and then the fraction is obviously infinitesimally small .
a few general comments about the consequences of addition or removal of a single bond are in order .
first of all , the removal of a single bond can only break up the cluster to which this bond belongs ; other clusters are not affected .
this is because in 2d , all rigid clusters are always rigid by themselves , i.e. , they remain rigid when taken in isolation from the rest of the network .
conversely , the addition of a bond can merge several rigid clusters into one , but will not affect the clusters outside the resulting cluster . note that even though the self - organized networks are by definition stress - free , a newly inserted bond can be redundant and introduce stress ( always confined to the cluster in which it is inserted ) ; in this case the configuration of rigid clusters is not affected , but the created stressed region may still be macroscopic and percolate . while we will briefly consider this situation at the end of this section , we mostly concentrate on the case when the inserted bond is non - redundant and thus affects the rigidity of the network but does not create stress . for brevity , we will call the places where insertion of a bond does not create stress , as well as a bond inserted in such a place , _ allowed_. note that if we add an allowed bond creating a certain rigid cluster and then remove this bond and insert another allowed bond both ends of which are in the region coinciding with the cluster created by the first bond , exactly the same cluster will be created : indeed , the count of constraints within the region will be the same ( @xmath24 for a region of @xmath15 sites ) , regardless of where in this region the bond is inserted .
we first look at the conversion between non - percolating and percolating networks as a result of bond addition and removal .
there are several related questions here . in fig .
[ fig : ins1 ] , we plot the probability that inserting a bond at a _ randomly chosen _ allowed place makes a non - percolating network percolating . in the floppy phase ( below @xmath22 ) , this probability tends to zero in the thermodynamics limit ; however , it is non - zero everywhere in the intermediate phase .
this probability is close to zero just above the rigidity transition and approaches one close to the upper boundary of the intermediate phase at @xmath47
. we can also ask about the frequency of networks in which conversion from non - percolating to percolating upon a single bond addition is possible . note that because of the above - mentioned property that insertion of a bond into a region made rigid by another bond rigidifies exactly the same region
, we do not need to try each and every allowed bond , which speeds up the simulation enormously .
[ fig : ins2 ] plots the fraction of non - percolating networks that can become percolating with a bond placed judiciously in an allowed position .
note that this is different from the previous figure : instead of choosing a place to insert a bond at random , we now make the best effort to cause percolation with a single bond addition , if at all possible . as is expected , this quantity is higher than that plotted in fig .
[ fig : ins1 ] , but the difference is small ; over most of the intermediate phase , whenever there are _ any _ allowed positions where bond insertion creates a percolating cluster , _ most _ allowed positions will do .
it is also interesting to note that the quantity in fig .
[ fig : ins2 ] is very close to linear in the intermediate phase , and is probably exactly linear , just like the probability of percolation without any bond insertions . in the appendix
, we explain why these two quantities are equal . finally , in fig . [
fig : clst_compare ] , the average size of the percolating cluster arising after bond insertion is compared to the average size of the percolating cluster in those cases when it exists even without bond insertion ( i.e. , the quantity in fig .
[ fig : perc_clst ] ) .
these values are identical : the size of the percolating cluster emerging after inserting a bond in a non - percolating network is the same as in originally percolating networks .
this suggests that a percolating network that arises after bond insertion is a typical percolating network , just like those networks that percolate without insertion . in the next subsection
, we present more evidence in favor of this . also see a more detailed discussion of this and a possible caveat in the appendix to this paper . given the significant probability of conversion of non - percolating networks to percolating ones ( see fig .
[ fig : ins2 ] ) , we can hypothesize that _ all _ non - percolating networks can become percolating after a _ finite _ number of bond insertions ; moreover , the average size of the percolating cluster after the minimal number of insertions needed to create it is again the same as the average size of the percolating cluster in networks that are percolating without bond insertions . an even stronger hypothesis is that after each new bond addition , the fraction of so far non - percolating networks becoming percolating is still the same linear function of @xmath8 as in fig .
[ fig : ins2 ] .
these hypotheses need to be tested in the future .
likewise , we can consider the probability that removing a randomly chosen bond from a percolating network breaks the percolating cluster .
this quantity is shown in fig .
[ fig : rem ] .
just like for bond removal , this probability is non - zero everywhere in the intermediate phase .
interestingly , even at the upper boundary ( @xmath48 ) , where almost all networks are percolating , it is still very easy to break percolation and thus create a non - percolating network .
in fact , the probability is the highest in this limit .
of course , this can be attributed to the fact that the percolating cluster is itself the biggest at this point ( taking up the whole network ) , so there is a greater chance than elsewhere to select a bond that belongs to it ( which is , of course , a necessary condition of its destruction ) .
this effect can be factored out by dividing the quantity in fig .
[ fig : rem ] by the average fraction of bonds in the percolating cluster .
this will then give the probability of the destruction of the percolating cluster , _ given that _
the removed bond belongs to this cluster .
this is plotted in fig .
[ fig : redbonds ] .
it is seen that across the whole intermediate phase , this quantity is nearly constant at around 70% . in other words , everywhere in the intermediate phase ,
the percolating cluster on average contains about 70% of bonds such that removal of any one of them will destroy percolation .
this is true even close to @xmath47 .
this is , of course , just the average ; one could ask if for some networks this quantity is zero ( similar to how in the case of bond insertion , not all non - percolating networks can be made percolating by a single bond addition , as shown in fig . [
fig : ins2 ] ) . to check
if this is the case , we remove ( and then reinsert ) up to 10 bonds one by one ( all chosen within the percolating cluster ) and see if the percolating cluster ever gets destroyed .
this happened for 5398 out of 5429 90000-site networks , or about 99.4% , including 198 out of 200 networks ( 99% ) at @xmath49 ( the highest mean coordination in our simulations ) . since these fractions are so close to 100% , we can hypothesize that , unlike in the case of bond insertion , in fact , _ all _ percolating networks have some bonds whose removal destroys percolation . )
divided by the average fraction of bonds in the percolating cluster ( the quantity in fig .
[ fig : perc_clst ] ) .
this serves as an estimate of the probability that a bond chosen at random among those belonging to the percolating cluster destroys percolation.,width=259 ] to conclude this subsection , our results for the change in the percolation state of the network upon single bond addition or removal again confirm that the system remains critical in the whole intermediate phase . in the previous subsection , we have found that in many cases insertion or removal of a single bond can change the percolation status of the network and affect the rigidity of its significant part .
since we were dealing with percolating clusters only , we could not study the effect of bond insertion or removal in those cases when the percolation status does not change . to do this , we introduce the concept of `` rigid '' and `` floppy '' bonds .
below the rigidity transition , only small rigid clusters are present in the network .
interestingly , a significant fraction of them consist of just a single bond .
the number of bonds belonging to these single - bond clusters decreases as the rigidity transition is approached and eventually crossed , although they are still encountered well above the transition in floppy pockets of the network . since these bonds are associated with the floppy phase and floppy regions of the network in the rigid phase , we call such bonds _
floppy_. all other bonds ( i.e. , those belonging to clusters consisting of more than one bond , or , in other words , rigid with respect to some other bonds ) are called _
rigid_. the fraction of rigid bonds as a function of @xmath8 is plotted in fig .
[ fig : rigbonds ] for both random and self - organized networks .
as expected , it grows in both cases with increasing @xmath8 , but in the self - organized case it reaches 1 at @xmath47 ( when the network becomes fully rigid ) , which in the random case does not happen until the full coordination at @xmath50 .
for the self - organized case , the averages over just percolating networks and over just non - percolating networks are also shown in fig . [
fig : rigbonds ] .
interestingly , the average for non - percolating networks remains nearly constant over the whole intermediate phase at about 75% , even as @xmath48 . in this limit
, the number of floppy modes per site tends to zero in all networks ; yet , as these results show , in those few networks that still do not percolate , many bonds ( about 25% ) are still in single - bond clusters .
just as in the previous section we considered the width of the distribution of percolating cluster sizes , it is interesting to look at the width of the distribution of fractions of rigid bonds . of course
, since the averages are very different for percolating and non - percolating networks , it makes sense to separate these two classes .
the results are shown in fig .
[ fig : rigbonds_width ] .
note that in the intermediate phase , the width is size - independent and so likely remains finite in the thermodynamic limit , for both percolating and non - percolating networks , just as we have seen for the percolating cluster size ( see fig .
[ fig : perc_width ] ) .
but in the floppy phase ( where , of course , only non - percolating networks are present ) , the width clearly decreases fast with size ( it is roughly inversely proportional to the square root of the network size ) .
thus the number of rigid bonds is a self - averaging quantity in the floppy phase but not in the intermediate phase , even when percolating and non - percolating networks are considered separately . at the same time , the widths for both percolating and non - percolating cases are much smaller that the difference between these two cases , so based on the count of rigid bonds , these two classes are clearly distinct . and non - percolating [ panel ( b ) ] self - organized networks for different network sizes.,width=259 ] we now look at the influence of insertion or removal of a bond on the number of rigid bonds and their spatial distribution . in fig . [ fig : bonds_flop2rig2 ] , we show the average change in the number of rigid bonds upon _ insertion _ of a single bond , in both the random and the self - organized cases .
note that in the random case , this change is very small on average , around 10 bonds or less . on the other hand , in the self - organized case
this quantity diverges very fast when the rigidity transition is approached .
this is not surprising : we have seen that infinite percolating clusters can easily form and break in this case .
for this reason , it makes sense to look at the _ fraction _ of bonds that undergo the change . in fig .
[ fig : frac_flop2rig ] , we show the fraction of bonds converted from floppy to rigid when a bond is added to the network at a randomly - chosen allowed position , among all _ floppy bonds _ in the network . in panel
( a ) , we plot the overall average , as well as partial averages restricted to those cases when the network goes from non - percolating to percolating , remains percolating and remains non - percolating ( the curve for the latter case is barely above the baseline ) . in panels ( b ) , ( c ) , and ( d ) , we show these partial averages for several different sizes . based on what we have already learned for networks switching from non - percolating to percolating , it is quite natural that in this case a significant fraction of floppy bonds become rigid ; this fraction approaches 1 when @xmath48 , which is again expected , since in this limit the percolating cluster takes up the whole network ( see fig .
[ fig : clst_compare ] ) .
the other two cases are more interesting . in the floppy phase ,
only the non - percolating @xmath51 non - percolating situation is possible , and as we see from panel ( d ) , in this region of the phase diagram the average fraction of bonds converted from floppy to rigid decreases fast with size . in the inset
, we plot the average _ number _ ( rather than fraction ) of converted bonds , and we see that this number is size - independent .
so in the floppy phase , a _ finite _ number of bonds gets converted . in the intermediate phase
, the situation is different .
looking first at the percolating @xmath51 percolating situation [ panel ( c ) ] , the fraction of converted bonds depends only weakly on the network size , and it is possible that this quantity goes to a constant in the thermodynamic limit . even if , in fact , this fraction decreases to zero as the size @xmath3 goes to infinity , it is clear that the decrease is much slower than @xmath52 , and thus the mean _ number _ of converted bonds diverges when @xmath53 . in the non - percolating @xmath51 non - percolating case , the quality of the data is lower ( there is very much noise due to a very large variation in the number of converted bonds ) , but still it is clear that the decrease with @xmath3 ( if present ) is certainly much slower than @xmath52 . the fact that for the percolating @xmath51 percolating and non - percolating @xmath51 non - percolating cases the average number of converted bonds diverges in the thermodynamic limit in the intermediate phase means that even when the percolation status does not change , the region of the network whose rigidity is affected are macroscopic _ at least in some cases_. this is again consistent with the criticality of the intermediate phase .
the distribution of the sizes of affected regions ( or numbers of converted bonds ) is likely power - law , which needs to be tested in the future . as an illustration of effects of bond insertions
we show two examples in fig .
[ fig : net ] .
the upper panel shows an example for the case when the network switches from non - percolating to percolating after a bond is added .
the lower panel shows an event where the network remains non - percolating , but large - scale rigidification still occurs without percolation . in both cases ,
the added bond is red and pointed with an arrow , thick green bonds are those that are originally rigid ( and , of course , remain rigid after bond addition ) , thin blue bonds are originally floppy and remain floppy , finally , thick black bonds are of most interest : these are the ones that switch from floppy to rigid . in the first ( non - percolating @xmath51 percolating ) case , converted bonds are spread throughout the network .
many bigger rigid clusters separated by floppy `` interfaces '' merge into one percolating rigid cluster ; in essence , the figure illustrates the rigidification of these interfaces . in the second
( non - percolating @xmath51 non - percolating ) example , the affected region is still large , but non - percolating . this example is larger than average , as can be deduced from fig .
[ fig : frac_flop2rig ] , panel ( d ) , but it is not a rare event .
we find overall similar behavior upon _ removal _ of a bond .
the results for the fraction of rigid bonds converting to floppy are in fig .
[ fig : frac_removal ] . similarly to fig .
[ fig : frac_flop2rig ] , in panel ( a ) we have the overall average and partial averages for percolating @xmath51 non - percolating , percolating @xmath51 percolating and non - percolating @xmath51 non - percolating cases , for a single network size ( 90000 sites ) . in panels ( b ) , ( c ) , and ( d ) , we have the same partial averages , but for three different sizes .
again , the conclusions are similar to the case of bond insertion : in the percolating @xmath51 non - percolating case , when the percolation status of the network changes [ panel ( b ) ] , the fraction of bonds that switch from rigid to floppy is expectedly high .
it is much lower in the other two cases , when the percolation status does not change , but still , just as for bond addition , while in the floppy phase the fraction of converting bonds falls rapidly with increasing size and , as the inset of panel ( d ) shows , the average _ number _ of converting bonds remains constant , in the intermediate phase , again , the dependence of the fraction of converting bonds on the size is very slow and the number of converting bonds diverges in the thermodynamic limit thus again , macroscopic regions of the network can be involved .
we finish this subsection with an observation : there exists a symmetry in bond conversions upon addition and upon removal .
namely , the average number of bonds converting upon bond addition in the case when the network transforms from non - percolating to percolating is the same as the average number of bonds converting upon bond removal when the network transforms from percolating to non - percolating .
this is illustrated in fig .
[ fig : backnforth ] . in this figure , we plot the ratios of the above - mentioned numbers and the total number of bonds in the network . we find that these two quantities coincide .
note that these quantities are different from those plotted in figs .
[ fig : frac_flop2rig ] and [ fig : frac_removal ] : in these figures , the numbers of converted bonds were divided by the number of floppy bonds and the number of rigid bonds , respectively , and not by the total number of bonds . the equality is easy to understand , if we recall that we have already seen some evidence ( see fig .
[ fig : clst_compare ] ) that networks that become percolating after bond insertion are , in fact , typical percolating networks , just like those that are originally percolating . by extension
, we can assume that networks that become non - percolating after bond removal are also typical non - percolating networks .
if so , then the average change in the number of rigid bonds should in both cases be the same as the difference in the average number of rigid bonds between percolating and non - percolating networks .
indeed , in fig .
[ fig : backnforth ] we also plot the difference between the average fractions of rigid bonds in the non - percolating and percolating cases ( i.e. , between the dashed and the dot - dashed lines in fig .
[ fig : rigbonds ] ) ; it is seen that this quantity coincides with the other two . in the previous subsections
, we have looked at cases when an `` allowed '' bond insertion is done ; in other words , the bond is inserted in one of those places where it does not create stress .
we now look at the opposite situation , i.e. , we analyze the results of inserting a bond in one of `` disallowed '' places . in this case
, the inserted bond is _ redundant _ , which means that its insertion does not change the configuration of rigid clusters , but some bonds ( including the inserted one ) become stressed .
we are interested in the emerging stressed region and in particular , whether it percolates or not .
similarly to the case of allowed bond insertion , we first look at the probability that the stressed region percolates .
note that the stressed region emerging upon bond insertion is always restricted to the rigid cluster containing the new bond .
for this reason , the stressed region can only percolate if the original network is percolating ( i.e. , contains a percolating rigid cluster ) , and we only need to look at percolating networks . in fig .
[ fig : str_perc ] , we show the probability that in a network in which rigidity percolates , a percolating stressed region emerges after insertion of a bond at a randomly chosen disallowed place ( not necessarily within the percolating cluster ) .
we note that apparently , this probability remains finite ( does not go to zero ) in the thermodynamic limit .
we also look at the sizes of stressed regions arising upon disallowed bond insertion .
the results are in fig .
[ fig : stress ] .
panel ( a ) shows the overall average fraction of stressed bonds and the partial averages for cases when stress does and does not percolate , for networks of 90000 sites ; panel ( b ) gives the partial average for the case when stress percolates ; and panel ( c ) gives the partial average for the case when stress does not percolate .
we do not make distinctions based on the _ rigidity _ percolation status of the network .
again , we see that the size of the stressed region as a finite fraction of the total number of bonds is only weakly size - independent and thus the affected region can be macroscopic in the intermediate phase even when there is no percolation . in the floppy phase ,
on the other hand , the fraction of stressed bonds decays fast with increasing network size and as the inset ( where the _ number _ of conversions , rather than the fraction , is plotted ) shows , the average number of conversions is size - independent and the affected region remains finite .
an atypical response to an external perturbation seen in the intermediate phase has been observed experimentally @xcite .
raman pressure experiments show that applying external pressure to a network blueshifts the frequency of corner - sharing ( cs ) tetrahedral units only once a certain pressure threshold ( @xmath54 ) value is reached .
@xmath54 is found to be zero within the intermediate phase and non - zero outside the intermediate phase .
these experimental results , published in ref .
@xcite , are reproduced in fig .
[ fig : boolchand ] . se@xmath55 glasses .
triangles are results taken from the work of murase and fukunaga @xcite and filled circles are results of the work of wang _ _ et al.__@xcite .
the figure is taken from ref.@xcite , width=259 ] outside the intermediate phase , the presence of a non - zero threshold for the shift in the frequency suggests that inhomogeneities cancel the effect , leaving only a broadening of the peak @xcite . on the other hand
, we have seen that in the intermediate phase , a finite fraction of the network is either in a percolating cluster or in a region that becomes percolating after a few bond insertions .
this region is also stress - free .
this can only be achieved by a precise balancing between constraints and degrees of freedom not just on average , but on all length scales , in other words , by the network being _ homogeneous _ because of self - organized criticality .
we note that the way rigidity and stress propagate through a macroscopic fraction of the network in our simulations is analogous to how stress propagates uniformly when external pressure is applied in experiments .
while the relation between @xmath54 and inhomogeneity still needs to be fully established , self - organized criticality in the intermediate phase is consistent with the observations .
we have studied the structural and response properties of the intermediate phase in the phase diagram of rigidity percolation using a model of self - organization on a 2d triangular network .
we had shown previously that the probability of rigidity percolation in the intermediate phase increases linearly from zero to one as a function of the mean coordination @xmath8 . at any @xmath8 ,
there are both percolating and non - percolating networks in the ensemble . in this paper , we have looked at the properties of both percolating and non - percolating clusters , the latter separately for percolating and non - percolating networks .
it turns out that at the point at which the percolating networks first emerge ( the rigidity transition ) , the percolating cluster takes up about 40% of the network , unlike the case of the usual second - order phase transition , where the emerging cluster is fractal and thus the fraction of bonds belonging to it is zero in the thermodynamic limit .
the size of the percolating cluster and some other characteristics are not self - averaging , but the distributions are rather narrow , if percolating and non - percolating networks are considered separately .
the distribution of sizes of non - percolating clusters is exponential at large sizes in the floppy phase , but power - law for arbitrarily big cluster sizes in the whole intermediate phase and not just at the transition .
the power - law exponents are different for non - percolating and percolating networks .
we have also looked at the changes in the rigidity of the network due to a microscopic perturbation in the form of insertion or removal of a single bond .
it turns out that one bond is often enough to convert a non - percolating network into a percolating one and vice versa ; moreover , when a non - percolating network is turned into a percolating one , the average size of the resulting percolating cluster is the same as that for the initially percolating networks . in a sense
, the percolating cluster is `` hidden '' in the non - percolating network and is revealed upon addition of a single bond .
it appears , further , that _ all _ non - percolating networks can be converted into percolating ones with a finite number of bond additions , with the same size of the resulting cluster as in networks that percolate without bond addition .
this implies that in the thermodynamic limit , there is no difference between percolating and non - percolating networks .
the intermediate phase then can be thought of as being the region of the phase diagram where all networks possess a percolating region that is _ nearly isostatic_. that is , this region can technically be floppy or stressed , but the number of floppy modes or redundant constraints causing stress is negligible in the thermodynamic limit .
the lower boundary of the intermediate phase is then , strictly speaking , not the rigidity percolation transition , but rather the `` near - isostaticity '' percolation transition .
these unusual properties indicate that the intermediate phase is a self - organised critical phase , with the system staying at the rigidity percolation threshold for the whole range of mean coordinations .
the existence of such a self - organized nearly isostatic critical phase explains recent pressure experiments @xcite .
since rigidity percolation occurs on a finite fraction of the network under any microscropic perturbation , the strain associated with the external pressure is immediately transferred to a macroscopic fraction of the sample , leading to a shift in the raman spectrum .
we can certainly expect other surprises associated with this intermediate phase in the rigidity phase diagram . but for that , we will likely need to create more realistic models of the experimental systems .
the authors are supported in part by the fonds qubcois de la recherche sur la nature et les technologies ( fqrnt ) , the natural sciences and engineering research council of canada ( nserc ) and the canada research chair program .
we thank the rseau qubcois de calcul de haute performance ( rqchp ) for generous allocation of computational resources .
numerically , the probability that there exists a place in a non - percolating network such that insertion of a bond at that place causes percolation is the same linear function in the intermediate phase as the probability of rigidity percolation ( see fig .
[ fig : ins2 ] ) . here
we show that this equality is a necessary consequence of the latter quantity being finite ( between zero and one ) and continuous throughout the intermediate phase .
given the probability of percolation among networks with @xmath14 bonds , we can calculate this probability for networks with @xmath56 bonds .
the method is similar to that used to calculate the bond - configurational entropy of self - organized networks in our previous paper @xcite .
suppose there are @xmath57 networks ( or bond configurations ) with @xmath14 bonds that are stress - free .
if the probability of percolation is @xmath58 , then @xmath59 of these are percolating and the rest , @xmath60n_{\rm bc}(b)$ ] , non - percolating .
also suppose there are on average @xmath61 allowed places to insert a bond in percolating networks and @xmath62 in non - percolating networks and the probability that a non - percolating network becomes percolating upon a _ random _ bond insertion is @xmath63 .
since all percolating networks will remain percolating upon bond insertion , then an average percolating network with @xmath14 bonds will produce @xmath61 different percolating networks . on the other hand ,
only a fraction @xmath63 of non - percolating networks will become percolating , so an average non - percolating network will produce @xmath64 percolating networks and @xmath65n_{\rm an}(b)$ ] non - percolating networks .
if we simply multiply these numbers by the number of networks with @xmath14 bonds of each kind , this will not produce the correct count of percolating and non - percolating networks with @xmath56 bonds , because each such network can be obtained in many different ways .
specifically , each network with @xmath56 bonds can be produced by bond insertion from as many different networks with @xmath14 bonds as there are the latter that can be obtained by bond _
removal _ from the former .
this number is always @xmath56 , since removal of every bond will produce a distinct network and all of them are stress - free .
so the count has to be divided by @xmath56 , and we get for the number of percolating networks with @xmath56 bonds , @xmath66 and for the number of non - percolating networks with @xmath56 bonds , @xmath67n_{\rm an}(b)}{b+1}.\ ] ] then the probability of percolation for a network with @xmath56 bonds is @xmath68 note that since the probability of percolation changes continuously , the difference between @xmath69 and @xmath58 is @xmath70 , where @xmath3 is the network size .
neglecting terms that are @xmath70 , we should equate @xmath71 and then , omitting the argument @xmath14 for brevity and using @xmath72 , @xmath73 or @xmath74 now , as we have seen , networks that become percolating after bond insertion are typical , generic percolating self - organized networks .
so we can expect that on average , when a bond is inserted into a non - percolating network and it becomes percolating , the change in the number of allowed bonds is @xmath75 .
but bonds that become disallowed are exactly those that are within the percolating cluster , and , according to our arguments at the beginning of section [ sec : pertur ] , these and only these bonds will create the percolating cluster and cause percolation ( indeed , if insertion of any bond creates a percolating cluster , then insertion of any bond outside the region that turns into the percolating cluster will not cause percolation because the cluster that will form can not overlap with this region , other than over a single pre - existing cluster , and thus can not percolate ) .
so , if for a given network making it percolating after one bond insertion is at all possible , then the probability that percolation will occur after _ random _ insertion is equal to the fraction of allowed bonds within the percolating cluster , or @xmath76 , and then the right - hand side of eq .
( [ eq : s ] ) is exactly the probability that percolation after one bond insertion is possible , and the proof is complete
. there are some caveats in this proof , related to the fact that , as we have seen , many quantities in the intermediate phase are not self - averaging .
because of this , for instance , if fractions of allowed bonds are not strictly deterministic , there can be correlations between the number of allowed bonds in a network and this number in the percolating network created after bond insertion .
for instance , the latter may tend to be above average when the former is below average .
then the average change in the number of allowed bonds may differ somewhat from @xmath75 , which is the difference of averages .
in fact , the very statement that networks obtained by bond insertion can be considered typical needs to be treated with caution . indeed , if all quantities associated with self - organized networks were self - averaging , then , in order to change the values of these quantities by inserting a bond , one would need to upset the balance so badly as to create a finite chance of seeing networks with properties that at equilibrium have probability zero .
such bias after inserting just a single bond can not happen , except in very special cases .
but the situation changes when there is no self - averaging : if the distribution of values of a certain quantity has a finite width , then any bias can change the average .
for this reason , it is possible that both the equality proved above and the statement that any quantities obtained for networks that become percolating after inserting a bond are the same on average as for originally percolating networks ( as demonstrated in figs .
[ fig : clst_compare ] and [ fig : backnforth ] ) are , in fact , only approximate . in any case ,
deviations ( if any ) are very small .
|
experimental results for covalent glasses have highlighted the existence of a new self - organized phase due to the tendency of glass networks to minimize internal stress .
recently , we have shown that an equilibrated self - organized two - dimensional lattice - based model also possesses an intermediate phase in which a percolating rigid cluster exists with a probability between zero and one , depending on the average coordination of the network . in this paper
, we study the properties of this intermediate phase in more detail .
we find that microscopic perturbations , such as the addition or removal of a single bond , can affect the rigidity of macroscopic regions of the network , in particular , creating or destroying percolation .
this , together with a power - law distribution of rigid cluster sizes , suggests that the system is maintained in a critical state on the rigid / floppy boundary throughout the intermediate phase , a behavior similar to self - organized criticality , but , remarkably , in a thermodynamically equilibrated state .
the distinction between percolating and non - percolating networks appears physically meaningless , even though the percolating cluster , when it exists , takes up a finite fraction of the network .
we point out both similarities and differences between the intermediate phase and the critical point of ordinary percolation models without self - organization .
our results are consistent with an interpretation of recent experiments on the pressure dependence of raman frequencies in chalcogenide glasses in terms of network homogeneity .
|
in terms of size , the narrow - line region ( nlr ) of active galactic nuclei ( agns ; including seyfert galaxies and quasars ) represents the link between the inner structure of the agn the dusty torus and the broad - line region ( blr)and the interstellar medium in the host galaxy .
the nlr bridges the region dominated by the gravitational field of the central massive black hole and the region dominated by the gravitational potential of the bulge stars of the host galaxy . to gain a complete understanding of the agn phenomenon , the role of the nlr therefore needs to be understood . unlike the blr , the nlr , or at least the outer parts of the nlr , can be spatially resolved in nearby agns . despite this
, however , the nlr remains remarkably poorly understood and there are many unanswered questions about the structure and kinematics of the nlr and its relationship to the blr / torus .
[ oiii]@xmath0 is one of the most important nlr lines .
its equivalent width ( ew ) in the agn ensemble varies dramatically by a factor of @xmath4 ( from @xmath5 to 157 ; baskin & laor 2005b ) , one of the things showing the most object - to - object variation in agns .
it is a dominant variable in the set of correlations making up the first principal component ( commonly called `` eigenvector 1 '' ; ev1 ) in the principal component analysis of qso properties of boroson & green ( 1992 ) ; ev1 is believed to be linked to certain fundamental parameters of the accretion process .
[ oiii ] luminosity is commonly used as a surrogate for the bolometric luminosity ( e.g. , heckman et al .
2004 ) and its doppler width as a surrogate for the stellar velocity dispersion ( @xmath6 ) of the host galaxy ( e.g. , nelson & whittle 1996 , wang & lu 2001 , komossa & xu 2007 , gaskell 2009a ) . although the spatial distribution of [ oiii]-emitting gas in the nlr of agns is rather inhomogeneous ( see , e.g. , das et al .
2006 , crenshaw et al .
2010 ) , there are some global regularities in its strength and kinematics .
its ew was found to correlate negatively , albeit with a large scatter , with the continuum luminosity ( grindlay et al .
1980 , steiner 1981 , dietrich et al .
2002 , kovacevic et al . 2010
table 3 of dong et al .
this is similar to the famous `` baldwin effect '' ( baldwin 1977 , baldwin et al .
1978 ) first discovered for the broad civ@xmath7 line .
the [ oiii ] line profile can be regarded as consisting of two components : a narrow component ( hereinafter , the ` core ' ) with a velocity close to the systemic redshift of the host galaxy and a low , broad base that is usually blueshifted ( hereinafter , the ` wing ' ) .
there is mounting evidence that the core component is a good tracer of the gravitational potential of the host galaxy bulge ( see , e.g. , gaskell 2009a and references therein ) while the wing component arises from agn outflows ( see , e.g. , crenshaw et al .
2010 , komossa & xu 2007 and references therein ) . scattering could also be a contributor to broadening of the wings ( gaskell & goosmann 2008 ) . besides these general global trends in the agn ensemble , there are some interesting peculiar [ oiii ] phenomena that have been discovered in some particular agn subclasses . for instance , in narrow - line seyfert 1s ( nls1s ) that have broad h@xmath8 line with full width at half maximum ( fwhm ) @xmath9
@xmath10 ( the `` population a '' of the eigenvector 1 domain , zamanov et al . [ 2002 ] ; see xu & komossa [ 2010 ] for a recent review ) , [ oiii]@xmath0 is generally very weak .
moveover , in a small fraction of nls1s , the whole [ oiii]@xmath0 profile has a bulk velocity shift of @xmath11 @xmath10 blueward ( the so - called `` blue outliers '' , zamanov et al .
2002 , boroson 2005 , zhou et al .
2006 , komossa et al . 2008 ) and there seems to be no non - blueshifted component from a canonical nlr as seen in low - ionization lines ( komossa et al .
komossa & xu ( 2007 ) found that although the width of the core component of [ oiii]@xmath0 is generally a good surrogate for @xmath6 , this however does not hold in `` blue outliers '' .
it is an interesting question whether or not these various regularities and phenomena can be unified and this has been looked at by previous authors . for instance , zamanov et al .
( 2002 ) suggested that `` blue outliers '' with extreme blue wing components originate in strong agn outflows , and moreover , are associated with the same outflows as the blue - shifted component of the broad civ@xmath12 emission line .
the origin of the blueshifting of high - ionization broad emission lines such as civ@xmath12 remains a subject of debate .
there is general agreement that there has to be a radial component of motion and some opacity source .
gaskell ( 1982 ) proposed that the blueshifting was due to an outflow of the high - ionization gas and obscuration by the accretion disk .
while this has remained the most popular model , it has the major problem that outflows of the dominant line - emitting gas seem to be strongly ruled out by velocity - resolved reverberation mapping ( gaskell 1988 , koratkar & gaskell 1989 , see gaskell 2009b for a review ) .
gaskell & goosmann ( 2008 ) therefore argue that the blueshifting is due instead to scattering of photons by material with a net inward motion as suggested by corbin ( 1990 ) .
observationally , the civ@xmath12 baldwin effect and blueshifting are related in the sense that higher - luminosity agns have both lower civ ews and larger blueshifts ( corbin 1991 , 1992 ) .
this is further supported by the study of richards et al .
if this is also true for [ oiii]@xmath0 , we will gain important insights into the origin and kinematics of the nlr .
it might point to similar underlying causes for the two effects in both the blr and nlr .
meanwhile , understanding these effects should enable us to calibrate better the doppler widths and luminosities of [ oiii]@xmath0 to serve as proxies for the bulge gravitational potential and agn bolometric luminosity
. motivated by these considerations , we present here the results of a systematic study of the velocity shift of [ oiii]@xmath0 emission line , its connection to the baldwin effect , and connection to possible physical drivers .
this study takes advantage of the unprecedented spectroscopic data from the sloan digital sky survey ( sdss ; york et al .
2000 ) . throughout this paper
, we use a cosmology with @xmath13 = 70 kms@xmath14mpc@xmath14 , @xmath15 = 0.3 , and @xmath16 = 0.7 .
type 1 agns are chosen from the sample of seyfert 1 galaxies and quasars of dong et al .
( 2011 ) taken from the fourth data release ( adelman - mccarthy et al .
2006 ) of the sdss .
we select type 1 agns having the [ oiii]@xmath0 emission line in the sdss bandpass and with continuum and emission lines suffering only minimally from contamination by host galaxy starlight . to ensure reliable analysis of the [ oiii]@xmath0 profile , we also require high spectral quality in the [ oiii]@xmath0 region .
our primary goal is to study the connection between velocity shifts and ews . to ensure the reliable analysis of the velocity shift of [ oiii ] , we first construct a basic sample by taking the [ sii]@xmath17 doublet lines as the fiducial reference for the systematic redshift .
[ sii ] is free from broad line contamination so using it is more accurate than using narrow h@xmath8 ( see section 2.2 ) .
the final criteria used for the basic sample are therefore : ( a ) @xmath18 , to ensure that [ sii ] emission lines , as well as [ oiii]@xmath0 and broad h@xmath8 , are in the bandpass ; ( b ) a median signal - to - noise ratio ( s / n ) @xmath19 per pixel for the optical spectra , and particularly @xmath20 in the [ oiii ] region ( 49955020 ) ; ( c ) the s / n of either [ sii ] @xmath21 or @xmath22 line @xmath23 to ensure [ sii ] to be a reliable reference for velocity shift ; and ( d ) weak stellar absorption features , such that the rest - frame ews of cak ( 3934 ) , cah + h@xmath24 ( 3970 ) , and h@xmath25 ( 4102 ) absorption features are undetected at @xmath26 significance . the last criterion ensures the measurement of the agn luminosity and the emission - line ews suffering little from the contamination of the host galaxy starlight ( see the appendix of dong et al .
2011 ) . after removing duplications and sources with too many bad pixels in the h@xmath8 + [ oiii ] region
, we obtain 565 type 1 agns .
radio jets may enhance [ oiii ] emission ( labiano 2008 ) and thus possibly affect the results .
we therefore use the method of lu et al .
( 2007 ) to exclude sources detected by the first ( faint images of the radio sky at twenty centimeters ) survey by matching with first catalog ( becker et al . 1995 ) . after this ,
our basic sample consist of 383 sources .
because of the redshift cutoff of @xmath18 ( to ensure the presence of [ sii ] as the redshift reference ) , the dynamic ranges of agn parameters is somewhat restricted ( see subsection 3.2 ) .
so in the investigation concerning [ oiii ] ews only , we further define an extended sample by relaxing both the redshift cutoff and the s / n requirement for [ sii ] .
the extended sample comprises 1951 radio - quiet type 1 agns at @xmath27 culled from the dong et al .
( 2011 ) sample , with a median s / n @xmath28 pixel@xmath14 in the [ oiii ] region .
data analysis methods are as described in detail in dong et al .
we only provide a brief description here . following dong et al .
( 2008 ) , we fit simultaneously the agn featureless continuum , the feii multiplets , and other emission lines using a code based on the mpfit package ( markwardt 2009 ) , and if necessary we refine the emission - line fitting with the continuum subtracted spectra using a code described in detailed in dong et al .
the agn continua are modeled by local power - laws for 42005600 and the h@xmath29 regions .
the feii emission is modeled with two separate sets of templates in analytical forms , one for the broad - line system and the other for the narrow - line system , using the identification and measurement of feii lines in i zw 1 from vron - cetty et al .
( 2004 ) , as listed in their tables a1 and a2 . within each system
, we assume that the respective sets of feii lines have no relative velocity shifts and have the same relative strengths as those in i zw 1 .
broad feii lines are assumed to have the same profile as broad h@xmath8 , while each narrow feii line is modeled with a gaussian . during the fitting , the normalization and redshift of every system
are set to be free parameters .
broad hydrogen balmer lines are fitted with as many gaussians as is statistically justified .
all narrow emission lines are fitted with a single gaussian , except for the [ oiii]@xmath30 doublet lines .
each line of the doublet is modeled with two gaussians , one accounting for the line core and the other for a possible blue wing , as seen in many objects .
the gaussian that has a smaller velocity shift relative to [ sii ] , whether blueward or redward , is taken to be the core component .
it also turns out to be narrower than the other gaussian component ( if the other one was present ) .
the velocity shifts of the core component with respect to [ sii]in the basic sample have a mean of @xmath3147 @xmath10 ( negative values denoting blueshifts ) and a standard deviation of 72 @xmath10 , while the mean shift of the wing component is @xmath31225 @xmath10 with a standard deviation of 240 @xmath10 , respectively . on average
the core components comprise 54 per cent of the total emission , with a standard deviation of 0.16 dex .
we list in table 1 the fluxes , ews and velocity shifts of the core and wing components and the whole profile of [ oiii]@xmath0 for all the objects in the basic sample .
the velocity shifts are calculated by assuming that the [ sii ] doublet lines are located at the systematic redshift . in table 2 we list the fluxes and ews of the core and wing components of [ oiii]@xmath0 for all the objects at higher redshifts that are thus not in the basic sample but in the extended sample .
the other spectral parameters of this sample are given in dong et al .
an example is illustrated in fig . 1 of fitting the [ oiii]@xmath0 line profile of sdss spectrum of j080131.58 + 354436.4 ( i.e. , the decomposition of the core and wing components ) . in the basic sample the velocity shifts of narrow h@xmath8with respect to [ sii ]
have a mean of @xmath32 @xmath10and a standard deviation of 82 @xmath10 .
such a scatter is a bit larger than that of the core component of [ oiii ] .
so if we use narrow h@xmath8 as the fiducial reference for the systematic redshift to study the velocity shift of [ oiii ] , it would smear the correlations of interest in this study .
this should be especially true for the sources at relatively higher redshifts in the extended sample .
those sources have lower spectral s / n and a weaker narrow h@xmath8 component than the sources in the basic sample , thus the deblending of narrow h@xmath8 from the broad component has a larger uncertainty .
the use of narrow h@xmath8 is checked in section 3 .
we notice that there is usually a prominent feii emission feature immediately redward of [ oiii]@xmath0 , which is dominated by feii42 @xmath33 and [ feii]20f @xmath34 . to ensure that the [ oiii ] measurements suffer little from poor subtraction of this feii feature , we visually inspect the fits for the entire extended sample and fine - tune the fits for some objects by carefully matching the model to this local feature . in such objects ,
the relative strengths of feii lines are significantly different from those of i zw 1 on which the feii template data are based .
we explore the connection between the velocity shift and strength of [ oiii]@xmath35 in the basic sample .
we calculate the velocity shifts of the core and wing components and the centroid ( i.e. , intensity - weighted average ) of the whole profile ( @xmath36(core ) , @xmath36(wing ) , and @xmath36(centroid ) , respectively ) with respect to [ sii ] .
the line strength is characterized by the rest - frame ew , as measured by dividing the emission line flux by the underlying agn continuum flux density at the central wavelength , for the core and wing components and the whole [ oiii ] emission , respectively .
we perform bivariate spearman rank correlations to test the relationship between these kinds of velocity shifts and ews .
this method tests for not only a linear relation but a monotonic one .
the results are summarized in table 3 , where we report the spearman coefficient ( @xmath37 ) and the probability ( @xmath38 ) that a correlation is not present .
a striking regularity emerges from the correlation matrix that _ the velocity shifts of the core and wing components and the whole profile , correlate significantly with the ew of the whole [ oiii ] emission ( @xmath39 ) and particularly with that of the core component , ew(core ) ( @xmath40 ) . _
the strongest correlation is between ew(core ) and @xmath36(centroid ) , the velocity shift of the whole profile ( @xmath41 and @xmath42 ) .
for any a specific kind of velocity shift , the correlation is most significant with ew(core ) , yet not very significant ( or not significant at all ) with the ew(wing ) ( see table 3 ) . for the ews of both the core component and the whole emission ,
the correlation is more significant with @xmath36(centroid ) rather than with @xmath36(wing ) or @xmath36(core ) , yet their correlations with @xmath36(core ) are still very significant ( @xmath43 ) . in fig .
2 , we show the relationships between the ew of the core component and the velocity shifts of core component and the whole profile . note that our @xmath36(core ) is effectively the same as the velocity shift measured from the peak of the whole [ oiii ] profile ( @xmath36(peak ) ) , because by fitting the core component one is always getting the peak of the whole [ oiii ] profile as the central wavelength of the 1-gaussian model .
@xmath36(peak ) is commonly used in the definition of `` blue outliers '' in the literature ( e.g. , zomanov et al .
2002 , boroson 2005 , komossa et al .
2008 ) . to illustrate the [ oiii ] velocity shift
ew relationship , we divide the entire sample into three subsamples according to @xmath36(core ) [ namely , @xmath36(peak ) ] with @xmath36(core )
@xmath44 @xmath10 , @xmath45(core ) @xmath46 @xmath10 , and @xmath36(core ) @xmath47 @xmath10 .
we then construct a composite spectrum for each subsample .
we first subtract broad h@xmath8 and feii emission from the sdss spectra , normalize every spectrum by the average flux density in the continuum window around 4200 , and then construct the composite in the same manner as richards et al .
( 2002 ) ( see their fig . 4 ) .
the composite spectra are demonstrated in fig .
3 ( panel a ) . the variations in ews and velocity shifts among the composite spectra
can readily be seen to be consistent with our correlation analysis in the sense that _ the more the peak of the line is blueshifted , the more the ews of the total emission and , particularly , the core component , decrease dramatically , while the blue wing changes much less_. this is similar to the case of civ as richards et al .
( 2002 ) found ( see also richards et al .
2010 , wang et al .
in addition , there seems to be a trend of a positive connection between the blueshift of the peak and the strength of the wing component , which is not very significant according to the above correlation result between @xmath36(core ) and ew(wing ) yet ( @xmath48% ) .
it is interesting to explore whether the blueshifting / baldwin effect of [ oiii]@xmath0 are driven by some fundamental physical parameter .
we have therefore investigated the correlations of the velocity shifts and ews of the core and wing components and of the total [ oiii ] emission with broad h@xmath8 fwhm , continuum luminosity @xmath49@xmath50 ) , @xmath2 , and @xmath3 , for the sources in the basic sample .
we calculate the black hole masses based on the fwhm of broad h@xmath8 using the formalism presented in wang et al .
( 2009 , their eqn .
this formalism was calibrated with recently updated reverberation mapping - based masses and assuming blr radius @xmath51 ( bentz et al .
the eddington ratios are calculated assuming a bolometric luminosity correction @xmath52@xmath50 )(elvis et al .
1994 ; kaspi et al .
the correlation results are summarized in table 4 .
we find that there are some significant correlations ( @xmath43 ) .
the ew of the core component has a negative correlation with the continuum luminosity ( i.e. , the classical baldwin effect ) and with the eddington ratio ( see also dong et al .
there is also a positive correlation of the blueshifting with the eddington ratio , but none of these correlations are strong ( @xmath53 ) . in particular , _ no correlation of any ew or velocity shift with these physical agn parameters is more significant ( stronger ) than the @xmath36(centroid)ew(core ) relationship for the same sample .
_ we also divide the 383 sources into three subsamples by @xmath1 and three subsamples by @xmath3 .
we construct composite spectra for each subsample as described in section 3.1 .
the three @xmath1 subsamples have @xmath54 , @xmath55 , and @xmath56 erg s@xmath14 , respectively , and the three @xmath3 subsamples have @xmath57 , @xmath58 , and @xmath59 .
the composite spectra are shown in panels b and c of fig .
all of the already mentioned correlations can be seen .
we also plot the relatively strong correlations of the velocity shifts of the peak ( i.e. , the core ) and centroid of [ oiii]@xmath35 with eddington ratio ( fig .
4 ) . in summary ,
the results worthy of note in the correlation matrix are as follows : * the most significant correlation in the ew - related correlation matrix is the positive one between the intensity ratio of the wing to the core component and the continuum luminosity ( @xmath60 for the 383 sources ) , which hints a positive correlation albeit a weak one to be sure between the ew of the wing component and the continuum luminosity .
this is confirmed by our analysis using the extended sample free from the redshift cutoff required by the presence of [ sii ] ( see below ) , which gives a significant correlation between ew(wing ) and @xmath1 with @xmath61 .
the significance is high because of the large number of sources ( 1951 ) but is still not strong ( @xmath62 ) .
the correlation of the ew of the core component with luminosity is slightly weaker than that with eddington ratio in the basic sample ( and much weaker in the extended sample see below ) .
both are much more significant than those with broad - line fwhm or black hole mass . *
the most significant correlations of all the velocity shifts of the core and wing components and of the total profile are with @xmath3 ( @xmath63 ) , rather than with broad - line fwhm , @xmath1 or @xmath2 .
these correlations are in the sense that the higher the @xmath3 the more the line is blueshifted .
moreover , all the three kinds of velocity shifts have _ no _ significant correlation with @xmath2 at all ( @xmath64 ) .
furthermore , all the objects with large blueshifts have a high eddington ratio , but the reverse is not true ( see fig . 4 ; also komossa et al .
2008 ) .
the dynamic ranges of @xmath1 , @xmath2 and @xmath3 in the basic sample are 2 , 1.5 and 1.5 dex respectively , which are not very large .
it is thus possible that some weak but real correlations are obscured by the significant measurement errors ( see section 4.1 below ; also wang et al .
2009 , dong et al . 2011 ) . to check the correlations concerning ews that do not require [ sii]as the reference for systematic redshift
, we use the extended sample as defined in section 2.1 .
this sample enlarges the dynamic ranges of @xmath1 , @xmath2 and @xmath3 by @xmath65 , 1 , and 0.5 dex respectively .
the results are also summarized in table 4 ( bottom panel ) .
the main conclusions remains almost unchanged with similar correlation strengths , except for the ew of the wing component .
as above mentioned , ew(wing ) now has a very significant ( yet still not strong ) correlation with @xmath1 .
it has a less significant correlation with @xmath2 , and almost no correlation with @xmath3 still .
another notable change is that the significance of the correlation of ew(core ) with @xmath3 now becomes much greater than that with @xmath1 , although the correlation is still not tight . to check for possible effects of the formalism we use to estimated the black hole masses , we re -
examine the correlation analysis with @xmath2 calculated using several other commonly used formalisms ( e.g. , greene & ho 2005 , collin et al .
2006 , vestergaard & peterson 2006 , shen et al . 2010 ) .
all the tests give similar results to those listed in table 4 ( see also table 3 of dong et al .
this is mainly because the dynamic range of @xmath2 covered by our samples is not very large ( @xmath66 dex for the extended sample , centered at @xmath67 @xmath68 ) , and in this range the various formalisms for calculating the black hole masses have only subtle differences from one another ( wang et al .
2009 ) .
furthermore , we consider possible effect that our sample selection and spectra analysis may bring . in section 2.1
, we simply use the limiting radio flux of first to minimize the effect of possible agn jets .
it is possible that many of the excluded radio sources are not radio loud but just nearby and bright . to check for this
, we perform the correlation tests using the 565 sources before the radio exclusion .
it turns out that , the inclusion of the radio sources only increases the correlation coefficients by a few per cent ( rarely by 10 per cent ) . particularly , all the correlations we concern in this paper , such as the blueshift
ew(core ) relationship , stay nearly unchanged .
this is mainly because the radio sources account for only 30 per cent of the 565 sources .
we check the effect of using narrow h@xmath8 as the fiducial reference for the systematic redshift . when it is applied to the basic sample ,
the strengths of the correlations concerning [ oiii ] velocity shifts as listed in tables 3 and 4 are all reduced significantly but the relative strengths among the correlations keep almost unchanged . when it is applied to the extended sample ,
the existing correlations get more statistically significant due to the larger sample size , yet the strengths ( @xmath37 ) get weaker ; e.g. , @xmath69 ( @xmath70 ) for the correlation of @xmath36(centroid ) with ew(core ) , and @xmath71 ( @xmath72 ) for its correlation with @xmath3 .
again , the relative strengths among the correlations keep almost unchanged .
it is , obviously , rather artificial to decompose the [ oiii ] profile into `` core '' and `` wing '' components by using simply two gaussians .
there is , for example , no theoretical justification that the wing component should have a gaussian profile . and , decomposing the line profile into different components certainly depends on the line width at the sdss spectral resolution . for objects with small line widths and small velocity shifts ,
the decomposition would be more difficult .
we do several investigations to check possible decomposition effects .
we find that the width of the decomposed core component , which is the most sensitive one to the spectral resolution , has only weak ( @xmath73 ) correlations with ew(wing ) , ew(total ) , and the velocity shifts of the core and wing components and of the total emission , and no correlation ( @xmath74% ) with ew(core ) at all .
the width of the decomposed wing component has no strong correlations with the ews or velocity shifts of the core component and of the total emission too , and almost no correlations with either the ew or velocity shift ( @xmath75% ) of the wing component .
these mean that the decomposition process would not bring a serious systematic effect into our results .
despite the crude nature of the decomposition , the primary results about the velocity shifts and the ews of [ oiii]@xmath0 ( and its core and wing components ) appear quite robust , according to both the correlation analysis and the composite spectrum analysis .
our main result is that , while there is a fairly strong correlation between the ew of the core component and the velocity shifts , there are no similarly strong correlations of the ews or velocity shifts of either the cores , wings , or of the whole emission with black hole mass , eddington ratio , or agn luminosity .
thus a tentative conclusion is that neither the large object - to - object variation of the ew of [ oiii]@xmath35 nor its blueshift are readily attributable to some fundamental agn parameters such as @xmath1 , @xmath2 and @xmath3(see also baskin & laor 2005b ) . likewise , neither does the velocity shift
ew relationship seem to be primarily driven by some agn parameter .
the weakness of the correlations of [ oiii ] properties with fundamental agn parameters ( including nuclear luminosity , @xmath2 and @xmath3 ) have also been reported in the literature by using samples with much larger dynamic ranges ( e.g. , dietrich et al .
2002 , netzer et al .
2004 ) .
this is in sharp contrast with the results for the ews of broad mgii@xmath76 doublet emission lines and of narrow optical feii emission lines , and the intensity ratios of ultraviolet and optical feii to mgii , as recently discovered by dong et al .
( 2009a , 2011 ) .
they found that those ews and intensity ratios correlate with @xmath3 strongly ( @xmath77 and @xmath78 ; see the above references ) , by using the same data set as the present study .
considering the effects of measurement errors are similar on the correlations of both those lines and [ oiii ] , we believe that the correlations of [ oiii ] properties with @xmath1 , @xmath2 , and @xmath3are intrinsically weak . for those emission lines arising from the blr and inner nlr ( narrow feii lines , dong et al .
2010 ) , dong et al .
( 2009a , 2009b , 2011 ) proposed that the essential physical mechanism is that eddington ratio regulates the global properties ( particularly , the distribution of column density ) of the clouds gravitationally bound in the line - emitting regions .
it appears now that the mechanism operating in the blr and the inner nlr does not play a major role in the canonical nlr located far out in the host galaxies and free from the gravitational influence of the central supermassive black holes .
a useful positive inference is that , at least in a statistical sense , the above - mentioned use of [ oiii]@xmath0 as the proxy of agn luminosity does not depend seriously on @xmath1 , @xmath2 and @xmath3 . _ on the other hand , although the [ oiii ] properties do not depend as strongly on @xmath1 , @xmath2 or @xmath3 as the above - mentioned emission lines arising from the inner nlr and the blr , there are still some significant ( albeit not strong ) correlations between [ oiii ] properties and those fundamental agn parameters .
the ew of the core component has the most significant ( anti-)correlation with @xmath3 , a less significant correlation with continuum luminosity , a much lower correlation with the broad - line fwhm , and almost no correlation with @xmath2 . for the wing component of [ oiii ]
the most significant correlation is with continuum luminosity . the behavior of the total emission results from the combination of the above two factors , and is mainly dominated by the core component . the magnitude of the blueshifting of the core , the wing , and the whole profile has the most significant correlation with @xmath3 , rather than with the other parameters .
because the sdss spectroscopic survey is magnitude - limited and , moreover , because both @xmath2 and @xmath3are constructed from broad - h@xmath8 fwhm and @xmath1 , there are apparent ( likely not intrinsic ) correlations among these four quantities .
for instance , the spearman correlation coefficients of @xmath3with broad - h@xmath8 fwhm , @xmath1 , and @xmath2 for our basic sample of 383 sources are @xmath37 = @xmath79 , @xmath80 , and @xmath81 , respectively . in light of the serious inter - dependence among these four quantities , the correlations of ew(wing ) with broad - h@xmath8 fwhm , @xmath2 , or @xmath3are probably a secondary effect of the stronger ( thus presumably intrinsic ) correlation with @xmath1 , while the correlations of the velocity shifts and ew(core ) with broad - h@xmath8 fwhm , @xmath1 , or @xmath2 are a secondary effect of that with @xmath3 .
this is probable given that there are additional systematic uncertainties plaguing the estimated values of @xmath3 .
one effect is the large uncertainties in virial bh masses , which can be a factor of 4 ( @xmath82 ) statistically , and perhaps as large as an order of magnitude for individual estimates ( vestergaard & peterson 2006 ; wang et al .
another uncertainty comes from the bolometric correction assumed for @xmath1 , which is definitely an oversimplification given the diverse spectral energy distributions of agns ( ho 2008 ; vasudevan & fabian 2009 ; grupe et al .
yet , since the above bivariate correlations of [ oiii ] properties with @xmath1 , @xmath2 and @xmath3 are not strong , being weaker than the @xmath36(centroid)ew(core ) relationship and much weaker even than the inter - dependence among the agn parameters , we refrain from an over - discussion on this issue at present . the weakness of those correlations either reflects true scatters that are caused by multiple ( and somehow independent ) processes and are thus irreducible , or suggests that there are other dominant variable(s ) yet to be identified ( see the discussions on various possible physical variables / processes in baskin & laor 2005b , risaliti et al . 2011 and caccianiga & severgnini 2011 ) .
the fairly tight anticorrelation between the blueshifts and the ew of the core component appears more notable than the weak correlations with fundamental agn properties .
interestingly , as analyzed in subsection 3.1 , this pattern is just like the situation for the broad civ@xmath7 emission line ( richards et al .
2002 , bachev et al .
2004 , richards et al .
2010 ) , which has already been noted by some researchers ( e.g. , zamanov et al .
this suggests that both the blueshifting phenomenon and the classical baldwin effect are tightly linked ( corbin 1991 , 1992 ) .
note that the classical baldwin effect only concerns the core component .
the wing component can even show an inverse baldwin effect the higher the agn luminosity , the larger the ew of the wing component .
this is similar to the situation of civ@xmath7 ( richards et al .
2010 , wang et al .
2011 ) . in light of the remarkable similarity in the velocity shift
ew connection between [ oiii ] and civ , occam s razor would suggest that the mechanism producing the two correlations is the same .
a wind has been proposed as the cause of the civ blueshifting ( gaskell 1982 ) and the correlations discussed above seem to be consistent with a scenario where the core of [ oiii ] comes from the canonical extended nlr where the gas motions are dominated by the gravity of the bulge stars while the wing component arises from an outflow .
this has been discussed by many authors ( see zamanov et al .
2002 , baskin & laor 2005a , richards et al .
2010 , and wang et al .
2011 for details ) . in this model
the `` blue outliers '' could just be extremes ; presumably a higher relative accretion rate drives a stronger outflow .
as pointed out in section 4.1 , the large object - to - object variation in the strength of [ oiii ] emission and the blueshift
ew(core ) connection are not governed primarily by any fundamental agn parameters such as @xmath1 , @xmath2 and @xmath3 .
it is likely that it is the ism conditions of the host galaxies that mainly determine the diversity of the [ oiii ] properties in active galaxies ( particularly its strength ; see netzer et al .
in the outflow scenario , it is natural to assume that the ism gives rise to [ oiii ] emission and decelerates the outflow .
although agn activity determines the launching speed of outflow , the final speed of the outflow depends on the density and density profile of the ism .
meanwhile , the density and the density profile also constrain the amount of [ oiii]-emitting gas and its emissivity . at the same agn activity level
, denser ism will produce more [ oiii ] emission while decelerate the outflow efficiently , leading to higher ew and lower velocity .
on the other hand , the similar blueshift
ew relationship in civ may be linked to the shape of the ionizing continuum ( richards et al .
so it is possible that the above two factors work together to produce the blueshift
ew(core ) connection .
yet , we should also mention that , while the outflow scenario is attractive , there seem to be problems .
for instance , the outflow explanation for the blueshifting of civ(gaskell 1982 ) clashes head on with the blr kinematics implied by velocity - resolved reverberation mapping .
velocity - resolved kinematics have long favored gravitationally - dominated motions ( gaskell 1988 , koratkar & gaskell 1989 , 1991 ; gaskell 2009b , 2010b ) and apparent cases of outflow are unlikely to be real but are instead probably the result of off - axis variability ( gaskell 2010a ) .
gaskell & goosmann ( 2008 ) have proposed instead that the blueshifting is the result of scattering off material with a slight net inflow .
this naturally explains the correlation of blueshifting with accretion rate .
if the nlr and blr blueshiftings have a similar cause , the blueshifting of [ oiii ] too might be caused by scattering , as has been proposed by gaskell & goosmann ( 2008 ) .
finally , we note that while all the agns with large blueshifts have a high eddington ratio , the reverse is not true ( see also komossa et al .
2008 ) . again
, this is very similar to the situation for broad civ@xmath12 ( baskin & laor 2005a ) .
here we must stress once again ( see section 4.1 ) that the correlations between velocity shift and any fundamental agn parameter such as nuclear luminosity , black hole mass , and the eddington ratio , are not strong ( see also boroson 2005 , komossa et al .
thus it is not clear if the velocity shift
ew(core ) connection for [ oiii]@xmath0 can be reduced to the effect of a specific underlying physical process of the agn activity . anyway , whatever the cause(s ) of the blueshifting of high - ionization nlr and blr lines , the correlations discussed here provide further support for an intimate connection between the nlr and the blr .
we thank the anonymous referee for helpful suggestions , jian - guo wang and dawei xu for discussions , and yifei chen for computing support .
this work is supported by chinese nsf grants nsf-10703006 , nsf-10973013 , and nsf-11073019 , a national 973 project of china ( 2009cb824800 ) , and the fundamental research funds for the central universities ( ustc wk2030220004 ) .
the visit of x .- b.d . to carnegie observatories is mainly supported by oatf , ustc ( up to april 2011 ) .
mg is grateful for support from the gemini - conicyt fund through grant 32070017 and from the cento de astrofsica de valparaso .
funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s .
department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england .
the sdss web site is http://www.sdss.org/. dong , x .- b .
, wang , j .-
, wang , t .-
g . , wang , h. , fan , x. , zhou , h. , yuan , w. , & long , q. 2009b , in asp conf . ser .
408 , the starburst - agn connection , ed .
w. wang et al .
( san francisco , ca : asp ) , 83 ( arxiv:0903.0698 ) [oiii ] region of sdss j080131.58 + 354436.4 .
we plot the continuum and feii subtracted spectrum ( black ) , the sum of all the best - fitting components ( red ) , the fitted narrow ( green ) and broad h@xmath8 ( cyan ) , and particularly the decomposed core ( pink ) and wing ( blue ) components of [ oiii]@xmath0 line .
the green dashed vertical line denotes the wavelength for [ oiii]@xmath0 as inferred from the [ sii ] line .
[ fig : fig-1 ] ] ( core ) bins .
the blue , green and red lines are spectra with decreasing @xmath36(core ) .
panel ( b ) : the composite spectrum of different @xmath1 bins .
the blue , green and red lines are spectra with decreasing @xmath1 .
panel ( c ) : the composite spectrum of different @xmath83 bins .
the blue , green and red lines are spectra with decreasing @xmath83 .
[ fig : fig - composite ] ] lccccccc j000410.80 - 104527.2 & -15.02 & 4.39 & -15.00 & 4.64 & -77.52 & -608.9 & -351.3 + j000904.54 - 103428.6 & -14.24 & 33.43 & -14.59 & 14.93 & -1.431 & -88.4 & -28.3 + j001247.93 - 084700.4 & -14.75 & 6.55 & -14.46 & 12.91 & -25.71 & -186.9 & -132.7 + j001340.73 - 111100.6 & -14.56 & 15.29 & -15.13 & 4.10 & -319.6 & -1134.4 & -492.9 + j001416.92 + 145038.4 & -14.94 & 9.24 & -14.85 & 11.26 & -130.5 & -483.8 & -324.9 + j003723.49 + 000812.5 & -14.82 & 13.12 & -14.97 & 9.35 & -37.27 & -422.4 & -197.9 + j004930.90 + 153216.3 & -14.08 & 32.48 & -14.62 & 9.24 & 16.22 & -480.5 & -94.2 + j005118.27 + 135448.0 & -14.07 & 26.38 & -15.26 & 1.72 & 2.772 & 265.9 & 18.8 + j005328.80 - 085755.0 & -14.69 & 10.91 & -14.98 & 5.51 & -32.51 & -250.1 & -105.7 + j005709.93 + 144610.2 & -14.02 & 6.72 & -14.19 & 4.57 & 106.7 & -487.3 & -134.7 lccccccc j000043.95 - 091134.9 & -15.27 & 4.38 & -15.42 & 3.11 + j000102.19 - 102326.8 & -14.39 & 35.43 & -14.66 & 18.9 + j000110.96 - 105247.4 & -14.99 & 5.52 & -14.70 & 10.8 + j000545.61 + 153833.8 & -14.88 & 7.73 & -14.59 & 15.0 + j000945.46 + 001337.1 & -15.73 & 2.77 & -15.03 & 13.7 + j001030.58 + 010006.0 & -13.84 & 82.26 & -14.43 & 21.4 + j001327.31 + 005232.0 & -14.33 & 24.42 & -14.34 & 23.6 + j001630.43 - 093853.4 & -14.98 & 7.48 & -15.00 & 7.18 + j001725.35 + 141132.5 & -14.74 & 5.28 & -14.52 & 8.82 + j001855.22 - 091351.1 & -15.08 & 3.85 & -14.70 & 9.39 + @xmath84 the number of the sources used in every bivariate correlation tests .
+ @xmath85 @xmath86(5100 ) ; the bh masses are calculated using the formalism presented in wang et al .
( 2009 , their eqn .
11 ) ; eddington ratios ( @xmath3 ) are calculated assuming that the bolometric luminosity @xmath87 .
|
we use homogeneous samples of radio - quiet seyfert 1 galaxies and qsos selected from the sloan digital sky survey to investigate the connection between the velocity shift and the equivalent width ( ew ) of the [ oiii]@xmath0 emission line , and their correlations with physical parameters of active galactic nuclei ( agns ) . we find a significant and negative correlation between the ew of the core component , ew(core ) , and the blueshift of either the core ( the peak ) , the wing , or the total profile of [ oiii ] emission ; it is fairly strong for the blueshift of the total profile particularly
. however , both quantities ( ew and velocity shift ) generally have only weak , if any , correlations with fundamental agn parameters such as the nuclear continuum luminosity at 5100 ( @xmath1 ) , black hole mass ( @xmath2 ) , and the eddington ratio ( @xmath3 ) ; these correlations include the classical baldwin effect of ew(core ) , an inverse baldwin effect of ew(wing ) , and the relationship between velocity shifts and @xmath3 .
our findings suggest that both the large object - to - object variation in the strength of [ oiii ] emission and the blueshift
ew(core ) connection are not governed primarily by fundamental agn parameters such as @xmath1 , @xmath2 and @xmath3 .
we propose that the ism conditions of the host galaxies play a major role instead in the diversity of the [ oiii ] properties in active galaxies .
this suggests that the use of [ oiii]@xmath0 luminosity as proxy of agn luminosity does not depend strongly on the above - mentioned fundamental agn parameters .
|
collimated outflows , or jets , with relativistic speeds are observed in active galactic nuclei and galactic microquasars , and are presumably driven in gamma - ray bursts .
possible mechanisms for driving relativistic jets include the electromagnetic extraction of the rotational energy of a central black hole ( bh ) @xcite , the electromagnetic extraction of the rotational energy of an accretion flow around a bh @xcite , and the thermal gas ejection from an accretion flow ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the electromagnetic extraction of the rotational energy of the accretion flow is the same mechanism as pulsar winds .
@xcite showed that when the poloidal magnetic field is penetrating a rotating conductive star , the steady force - free magnetosphere can be established outside the star , where the poloidal electric fields are maintained by the charge distribution and the poloidal currents flow ( i.e. , the unipolar induction process ) . as a result ,
outward electromagnetic energy and angular momentum fluxes continue to be generated .
the origin of the electromotive force ( and the fluxes ) is the matter - dominated rotating star , while the plasma outside the star only supports the electromagnetic field , playing a passive role for keeping the electric potential differences in the whole system . as for the energy extraction of the bh itself ,
@xcite first showed a mathematical solution for the electromagnetic energy and angular momentum fluxes extracted from the bh with the force - free approximation in the slow rotation limit of the bh , in a similar mechanism to the goldreich - julian model .
it has been frequently argued that this electromagnetic process may be the only viable process to extract the rotational energy from a bh ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , the origin of the electromotive force in this process is still being debated .
there is no matter - dominated region in which the poloidal magnetic field is anchored , and all the matter in the bh magnetosphere only plays a passive role .
then it is not a simple question what determines the electric potential difference ( or the so - called angular velocity of the magnetic field line @xmath3 ) and drives the poloidal currents .
the literature on the origin of the electromotive force focuses on either the event horizon @xcite , the wind separation surface @xcite , or the ergosphere @xcite ( see also * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , although the first one is unlikely because the event horizon is causally disconnected from the bh exterior @xcite .
the above literature mainly uses the analytical methods , while the numerical simulations based on the ideal magnetohydrodynamic ( mhd ) approximation have been actively performed to investigate the nature of the bh magnetosphere ( and even its interaction with the accretion flow ) ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
based on the simulation results , it was proposed that the unavoidable rotation of the fluids in the ergosphere generates mhd waves propagating outwards , and its feedback causes the fluids to have negative hydrodynamic energy as measured at infinity .
they plunge into the bh , decreasing its rotational energy @xcite .
however , the long - term simulations have demonstrated that no regions of negative hydrodynamic energy are seen in the steady state @xcite , and thus more studies are required to understand the real mechanism of the extraction of the bh rotational energy @xcite .
furthermore , it has not been discussed what determines the value of @xmath3 and drives the poloidal currents by utilizing the mhd simulation results , as far as we are aware . since the blandford - znajek process is purely electromagnetic , the force - free simulations appear more suitable than the mhd simulations for investigating the nature of its mechanism ( without including the complicated interaction of the bh magnetosphere with the accretion flow ) .
@xcite clearly shows that when the bh exterior is vacuum , the electric field ( the @xmath0 field ) can be stronger than the magnetic field ( the @xmath1 field ) in the ergosphere , and performed the resistive force - free numerical simulations of the plasma - filled bh magnetosphere , demonstrating that such a strong electric field drives the poloidal currents .
a similar conclusion was drawn from simulation results of the force - free electromagnetic field on highly compact regular space - times ( i.e. , with no event horizon ) with an ergosphere @xcite . however , it has not been evident why the electric field is maintained to be strong in the plasma - filled bh magnetosphere and how the steady state with @xmath4 and non - zero poloidal currents is realized .
a related question is whether the steady state with @xmath5 and/or zero poloidal currents is possible or not . in this paper , we extend the argument of @xcite , providing more general insights .
we consider the steady , axisymmetric kerr bh magnetosphere filled with a collisionless plasma satisfying @xmath6 , and show in a fully analytical way that for open magnetic field lines threading the ergosphere , the state with @xmath4 and non - zero poloidal currents is forced to be maintained , i.e. , the outward poynting flux inevitably keeps being generated .
the origin of the electromotive force is ascribed to the ergosphere .
we show a self - consistent electromagnetic structure along an open field line threading the equatorial plane in the ergosphere , and deduce the range of the possible value of @xmath3 for such field lines .
our assumptions for the magnetospheric plasma are stated in section [ sec : bhmag1 ] , which do not rely on a specific ohm s law as done in @xcite , the force - free condition ( @xmath7 ) , nor the ideal mhd condition ( @xmath8 ) .
this paper is organized as follows .
we first review the goldreich - julian model of pulsar winds in section [ sec : pulsar ] and the convenient formulation of the electrodynamics in kerr space - time in section [ sec : bhem ] .
we then study the plasma - filled bh magnetosphere , discussing its general properties ( section [ sec : bhmag ] ) and the origin of the electromotive force ( section [ sec : bhuni ] ) . section [ sec : discussion ] is devoted to summary and discussion .
we review the unipolar induction mechanism of pulsars , first discussed by @xcite , _ which is quite helpful for understanding the unipolar induction of rotating bhs . _ in particular
, it is instructive that the electric field outside the star is stronger in the plasma - filled case than that in the vacuum case ( equations [ eq : pulsare1 ] and [ eq : pulsare2 ] ) and that the star drives the poloidal currents inside itself in the direction of @xmath9 , which generates the poynting flux ( equation [ eq : pulsar_poynting ] ) .
we will refer to these points in section [ sec : active ] and in section [ sec : structure ] , respectively .
let us consider a uniformly magnetized , rotating star with angular velocity @xmath10 .
our assumptions are as follows : ( 1 ) the rotation vector is parallel to the magnetic dipole moment , so that the system is steady and axisymmetric . ( 2 ) at and inside the stellar surface , the matter is highly conductive , and its rotational energy dominates the total energy .
( 3 ) the region outside the star is filled with plasma .
the plasma is dilute and collisionless , but its number density is high enough to screen the electric field along the magnetic field lines , i.e. , @xmath11 is satisfied .
the energy density of the particles is much smaller than that of the electromagnetic field .
the magnetic field is so strong that the gravitational force of the central star is negligible compared with the lorentz force . in this section
, we adopt the units of @xmath12 .
first of all , at and inside the stellar surface , the high conductivity implies @xmath13 where @xmath14 is the fluid velocity , and we have adopted the cylindrical coordinate system @xmath15 . this electric field is produced by the non - zero charge density @xmath16 .
the electromagnetic field structure in the region outside the star , i.e. , magnetosphere , is described as follows . in the axisymmetric configuration
, the poloidal magnetic field can be written as @xmath17 where @xmath18 is called the flux function , representing the toroidal component of the vector potential times @xmath19 , or the magnetic flux threading through a circle of @xmath19 for a given @xmath20 .
the above equation leads to @xmath21 , which means that @xmath18 is constant along each magnetic field line . as for the electric field
, faraday s law in the steady state gives @xmath22 .
the axisymmetric condition implies @xmath23 .
this equation and @xmath24 mean that the electric field can be written as @xmath25 where @xmath3 is a function of @xmath26 . substituting this equation into @xmath27 and using @xmath28 ,
one obtains @xmath29 this means that @xmath3 is constant along each magnetic field line . the electric field is also described by @xmath30 , which means that each magnetic field line is equi - potential , and @xmath3 corresponds to the potential difference between the field lines .
the plasma in the magnetosphere sustains @xmath16 , which produces the electric field in the similar manner to the star .
equation ( [ eq : e ] ) is satisfied in the whole system we now consider including the inside of the star , as long as there is no region where @xmath31 , i.e. , no gap
. comparing it with equation ( [ eq : estar ] ) , one obtains @xmath32 this means that the stellar rotation generates the potential differences in the magnetosphere , i.e. , the origin of the electromotive force for pulsar winds is ascribed to the rotation of the stellar matter . for the rigid rotation of the star
, @xmath3 is constant in the whole system .
it is interesting that the plasma enhances the strength of the electric field across the magnetic field , although it screens that along the magnetic field ( satisfying @xmath24 ) .
in fact , in the plasma - filled case for the dipole magnetic field , which we have considered , one has from equation ( [ eq : e ] ) @xmath33 in terms of the spherical coordinates @xmath34 , where @xmath35 is the magnetic dipole moment . in the vacuum case
, the electric field has quadrupole shape @xcite @xmath36 where @xmath37 is the radius of the star .
it is obvious that the electric field is stronger in the plasma - filled case for @xmath38 .
let us consider the particle motions in the magnetosphere .
since the energy density is dominated by the electromagnetic field , the particles move via the @xmath39 drifts . for the dipole magnetic field and the associated electric field shown above ,
this motion is in the direction of @xmath40 .
it carries the drift currents , but does not produce @xmath41 .
the drift velocity is written by using equation ( [ eq : e ] ) as @xmath42 as long as @xmath43 , one has @xmath44 . because @xmath45 , the particles co - rotate with the star .
the drift current is @xmath46 .
now one can calculate the charge density in the magnetosphere at @xmath47 as @xmath48 this implies that @xmath49 ( @xmath50 ) in the region where @xmath51 ( @xmath52 ) .
the co - rotation of the particles can not be maintained outside the light cylinder at @xmath53 .
the poloidal electric currents must flow .
the poloidal currents produce @xmath54 , which prevents the particle speeds from exceeding the light speed ( see equation [ eq : v_d ] ) .
the particles must flow out along the magnetic field lines which cross the light cylinder .
those magnetic field lines must be open , because of the symmetry with respect to the equatorial plane .
the charge density distribution ( equation [ eq : rho_gj ] ) and the particle outflow imply that the poloidal electric currents flow towards the northern and southern poles of the star and flow out from the light cylinder around the equatorial plane .
therefore one finds @xmath55 ( @xmath50 ) in the main body of the outflow region ( i.e. , the open field line region ) in the northern ( southern ) hemisphere . according to equation ( [ eq : v_d ] )
, one has @xmath56 , and thus it is reasonable that @xmath57 around the light cylinder .
one might consider that such a strong @xmath41 is produced by the poloidal currents flowing due to the effects of the particle inertia around the light cylinder , but this is not the case , because the particle energy density is still much smaller than the electromagnetic field energy density around the light cylinder , as seen by the mhd theory ( cf .
* ; * ? ? ?
@xmath41 must be produced by the poloidal currents which are driven by the matter - dominated star .
the outflows of the positive ( negative ) charges from the high ( low ) potential region in the star result in the net electric field @xmath58 in the direction of @xmath9 , which continues driving the poloidal currents and supplying the positive ( negative ) charges to the high ( low ) potential region .
the currents flowing in the direction of @xmath9 in the star generates the poloidal component of the poynting flux @xmath59 , as described by @xmath60 this equation also means that the lorentz force @xmath61 , being in the direction of @xmath62 , decelerates the stellar rotation as a feedback of the poynting flux generation . in this way
, the rotating star keeps generating the electromotive force ( equation [ eq : f * ] ) , driving the poloidal electric currents , and generating the outward poynting flux .
since @xmath57 is expected around the light cylinder , one may estimate @xmath63 , where @xmath64 is the strength of the poloidal magnetic field at the light cylinder .
the dipole field implies @xmath65 , where @xmath66 is the strength of the poloidal magnetic field at the stellar surface .
then the electromagnetic luminosity is roughly estimated as @xmath67 here the condition @xmath57 around the light cylinder is equivalent to the estimate of the total current @xmath68 flowing towards the star .
in fact , the order of the current density is probably @xmath69 , and then @xmath70 , where @xmath71 is the area of the stellar surface into which the currents flow , and estimated to be @xmath72 @xcite .
the electromagnetic luminosity is also calculated by @xmath73 , where the potential difference @xmath74 , which is consistent with equation ( [ eq : luminosity ] ) .
( the factor 2 comes from the contributions from the northern and southern hemispheres . )
the poloidal currents must cross the poloidal magnetic field lines in the direction of @xmath75 outside the star for making closed circuits , which accelerates the particles via the @xmath76 force and then converts the poynting flux to the particle kinetic energy flux . in this paper
we do not discuss the particle acceleration mechanism in detail ( see * ? ? ?
* ; * ? ? ?
* and references therein ) .
we briefly summarize the geometrical properties of kerr space - time and the electrodynamics in order to define the various physical quantities used in this paper .
we follow the definitions and formulations of @xcite except for keeping @xmath77 in maxwell equations .
we adopt the units of @xmath78 and @xmath79 , where @xmath80 is the bh mass . for such units , the gravitational radius @xmath81 .
the space - time metric can be generally written as @xmath82 where @xmath83 is called the lapse function , @xmath84 the shift vector , and @xmath85 the three - dimensional metric tensor of the space - like hypersurfaces .
the hypersurfaces are regarded as the absolute space at different instants of time @xmath86 ( cf .
for kerr space - time , @xmath87 .
these correspond to the existences of the killing vector fields @xmath88 and @xmath89 . in the coordinates @xmath90 , @xmath91 and @xmath92 .
the local fiducial observer ( fido ; * ? ? ? * ; * ? ? ? * ) , whose world line is perpendicular to the absolute space , is described by the coordinate four - velocity @xmath93 the angular momentum of this observer is @xmath94 , and thus fido is also a zero angular momentum observer ( zamo ; * ? ? ?
note that the fido frame is not inertial , but it can be used as a convenient orthonormal basis to investigate the local physics @xcite
. in the boyer - lindquist ( bl ) coordinates @xmath90 ( see appendix [ sec : app1 ] ) , fidos rotate with the coordinate angular velocity @xmath95 which is in the same direction as the bh .
the bl coordinates have the well - known coordinate singularity ( @xmath96 ) at the event horizon .
the radius of the event horizon is @xmath97 .
the killing vector @xmath88 is space - like in the ergosphere , where @xmath98 .
the radius of the outer boundary of the ergosphere ( i.e. , the stationary limit ) is @xmath99 . at infinity , this space - time asymptotes to the flat one .
the shapes of the event horizon and the ergosphere are shown in fig .
[ fig : fun ] .
the kerr - schild ( ks ) coordinates have no coordinate singularity at the event horizon .
however , the ks spatial coordinates are no longer orthogonal ( @xmath100 ; see appendix [ sec : app1 ] ) , and then one should be cautious for examining the spatial structure of the electromagnetic field by using the ks coordinates . in the bl
coordinates in the order of increasing @xmath101 .
the line of @xmath102 is identical to the outer thick line .
the spin parameter is set to be @xmath103 . ] in order to study the test electromagnetic field in kerr space - time , we adopt the 3 + 1 electrodynamics of the version which was developed by @xcite ( see also * ? ? ?
* ; * ? ? ?
* and references therein ) . or @xmath104 , and showed some of the expressions in this paper , such as equations ( [ eq : psi ] ) and ( [ eq : d ] ) . ] the covariant maxwell equations @xmath105 and @xmath106 are reduced to @xmath107 @xmath108 where @xmath109 and @xmath110 denote @xmath111 and @xmath112 , respectively , and @xmath113 is the levi - civita pseudo - tensor of the absolute space .
the condition of zero electric and magnetic susceptibilities for general fully - ionized plasmas leads to following constitutive equations , @xmath114 @xmath115 where @xmath116 denotes @xmath117 . at infinity , @xmath118 and @xmath119 , so that @xmath120 and @xmath121 . here
@xmath0 , @xmath1 , and @xmath122 are the electric field , magnetic field , and charge density as measured by fidos , respectively ( see appendix [ sec : app1 ] for more details ) .
the current @xmath123 is related to the current as measured by fidos , @xmath124 , as @xmath125 the covariant energy - momentum equation of the electromagnetic field @xmath126 gives us the energy equation as @xmath127 + \nabla \cdot \left(\frac{1}{4\pi}\bmath{e } \times \bmath{h } \right)= - \bmath{e } \cdot \bmath{j } , \label{eq : energy}\ ] ] where @xmath128 denotes @xmath129 , and the angular momentum equation as @xmath130 + \nabla \cdot \frac{1}{4\pi } \bigl[-(\bmath{e}\cdot \bmath{m } ) \bmath{d } - ( \bmath{h}\cdot \bmath{m } ) \bmath{b } \nonumber \\
+ \frac{1}{2 } ( \bmath{e}\cdot \bmath{d } + \bmath{b}\cdot \bmath{h})\bmath{m } \bigr ] = -(\rho \bmath{e } + \bmath{j}\times \bmath{b})\cdot \bmath{m } , \label{eq : am}\end{aligned}\ ] ] where @xmath131 . from these equations
, one can find the energy density , energy flux , angular momentum density , and angular momentum flux . before investigating the plasma - filled magnetosphere in kerr space - time ,
the properties of the electromagnetic field in the vacuum ( i.e. , no plasma ) are summarized .
@xcite derived the solution of a steady , axisymmetric , vacuum test electromagnetic field in kerr space - time for which the magnetic field is uniform , parallel to the rotation axis , at infinity in an elegant way by utilizing the fact that @xmath132 and @xmath133 generate a solution of maxwell equations ( see appendix [ sec : app1 ] ) .
for this solution , the poloidal @xmath0 field is non - zero .
this may be understood by considering @xmath134 . since @xmath135 generally , one has @xmath136 .
this @xmath0 field is produced by the charges at the event horizon and at infinity .
note that @xmath137 for this solution .
@xmath27 implies @xmath23 , and @xmath138 tells us @xmath139 .
then the poloidal components of the energy and angular momentum fluxes are zero .
these properties are the same as those for the pulsar in the vacuum .
now we examine the steady , axisymmetric , test electromagnetic field in kerr space - time in which the plasma is filled . as a preparation for the discussion on the unipolar induction of rotating black holes ( in section [ sec : bhuni ] ) , we summarize the general properties of the electromagnetic field in section [ sec : bhmag1 ] . in section [ sec : particle ] , we study the particle motions as viewed by fidos .
this study provides a conclusion that @xmath140 is the necessary and sufficient condition for driving the electric currents to flow across the poloidal @xmath1 field lines .
we consider the kerr bh magnetosphere under the following assumptions : ( 1 ) the poloidal @xmath1 field produced by the external electric currents is penetrating the ergosphere .
( 2 ) the plasma in the bh magnetosphere is dilute and collisionless , but its number density is high enough to screen the electric field along the @xmath1 field lines , i.e. , @xmath6 .
the energy density of the particles is much smaller than that of the electromagnetic field .
( 3 ) the gravitational force is negligible compared with the lorentz force . ( the gravitational force overwhelms the lorentz force in a region very close to the event horizon @xcite , but the physical condition in that region hardly affects its exterior . )
these assumptions are the same as those for the pulsar case .
one big difference is that there is no matter - dominated region on which the @xmath1 field is anchored ( see section [ sec : intro ] ) . as a result ,
the determination of @xmath3 is not so simple as the pulsar case . in terms of the vector potential ,
one can write @xmath141 @xcite . thus one finds @xmath142 where we have defined @xmath143 .
it is easily shown that @xmath144 , which means that @xmath145 is constant along each @xmath1 field line . the condition @xmath6 and equation ( [ eq : relation_e ] ) lead to @xmath24 .
taking account of @xmath23 , one can write @xmath146 substituting this equation into @xmath27 and @xmath28 , one obtains @xmath147 that is , @xmath3 is constant along each @xmath1 field line .
the @xmath75 field is also described by @xmath148 , which means that each @xmath1 field line is equi - potential , and @xmath3 corresponds to the potential difference between the field lines .
these properties are the same as those in the pulsar case , discussed in section [ sec : pulsar ] . in the steady , axisymmetric state ,
the angular momentum equation ( [ eq : am ] ) is reduced to @xmath149 the energy equation ( [ eq : energy ] ) is reduced to @xmath150 which can be deduced by equation ( [ eq : am2 ] ) together with equations ( [ eq : omegaf_def ] ) and ( [ eq : omegaf_const ] ) .
these equations imply that @xmath151 is generated by the poloidal currents which have the component perpendicular to the poloidal @xmath1 field , and then the poloidal component of the poynting flux @xmath152 is non - zero when @xmath153 .
we note that @xmath154 and @xmath155 are the same in the bl and ks coordinates @xcite .
@xcite argues that the poloidal currents can have the component perpendicular to the poloidal @xmath1 field in the region where @xmath140 , by utilizing a specific ohm s law ( or some effective resistivity ) .
here we prove this to be valid by examining particle motions in the electromagnetic fields more generally , without relying on the ohm s law .
we treat the particle motion as viewed by fidos .
they can use a convenient local orthonormal basis for which the space - time metric is diagonal and one can investigate local , instantaneous particle motions under the lorentz force as special relativistic dynamics .
in fact , the equation of a particle motion as viewed by fidos ( either bl or ks fidos ) is @xmath156 as shown in appendix [ sec : app2 ] .
here @xmath157 , @xmath158 , @xmath159 , and @xmath160 are the four - velocity , three - velocity , charge , and mass of a particle , respectively , and @xmath161 denotes the vector component in respect of the fido s orthonormal basis , i.e. , @xmath162 ( see appendix [ sec : app1 ] ) . and @xmath163 used in @xcite are identical to @xmath0 and @xmath1 .
the fido s orthonormal basis is not utilized for discussion in @xcite . ]
note that the fido frame is not inertial and a particle feels the gravitational force , although we assume that it is negligible compared with the lorentz force .
in this equation @xmath164 and @xmath165 appear as the electric and magnetic fields as viewed by fidos , respectively , as expected .
the assumption @xmath6 is equivalent to @xmath166 , because @xmath167 is a scalar and @xmath168 . when @xmath169 is satisfied ( which is equivalent to @xmath170 )
, the charged particles freely move along the @xmath171 field line and/or drift in the direction of @xmath172 .
effectively one has @xmath173 , where @xmath174 and @xmath175 denote the four - velocities of positively and negatively charged particles , respectively .
this equation is valid as long as the plasma is dilute and any particle collisions which induce @xmath176 or @xmath177 are ineffective .
in the coordinate basis , one has @xmath178 , since @xmath179 is a scalar and @xmath168 .
then by considering @xmath180 , one has @xmath181 . as a result
, the motions of the charged particles can carry only the electric currents satisfying @xmath182 and @xmath183 .
if @xmath140 is realized ( which is equivalent to @xmath184 ) , then the positively ( negatively ) charged particles are forced to move in the direction of @xmath185 ( @xmath186 ) ( cf . * ? ? ? * ) , i.e. , @xmath187 and @xmath188 .
then one has @xmath189 and @xmath190 in the coordinate basis , which lead to @xmath191 and @xmath192 since @xmath193 . therefore
, the particle motions can carry @xmath194 .
this implies @xmath195 .
the force - free approximation @xmath7 ( i.e. , @xmath196 ) is violated in this case , since @xmath23 . in summary
, @xmath140 is the necessary and sufficient condition for driving the electric currents to flow across the poloidal @xmath1 field lines ( and then obtaining @xmath197 ) under our assumptions ( 1)(3 ) listed in section [ sec : bhmag1 ] .
we discuss the unipolar induction process in the kerr bh magnetosphere where the poloidal @xmath1 field lines threading the ergosphere are open , i.e. , crossing the outer light surface .
we mainly utilize the bl coordinates rather than the ks ones in this section .
the light surfaces are defined as follows . in the bl
coordinates , when @xmath198 , the coordinate angular velocity of the drift motion is @xmath3 , as deduced in appendix [ sec : app2 ] . the light surfaces are the surfaces where the four - velocity of a particle which is rotating with the coordinate angular velocity @xmath3 becomes null , i.e. , @xmath199 , where we have defined @xmath200 there can be two light surfaces .
one has @xmath201 at the outer light surface ( where @xmath202 ) , while @xmath203 at the inner light surface ( where @xmath204 ) . from the condition @xmath205 , equations ( [ eq : relation_e ] ) and ( [ eq : omegaf_def ] ) lead to @xmath206 calculating @xmath207 , one obtains in the bl coordinates @xmath208 where @xmath209 .
this equation can be rewritten by using equation ( [ eq : f_def ] ) ( cf . * ? ? ?
* ) @xmath210 where we have used @xmath211 ( see equation [ eq : relation_h ] ) .
this equation is useful for the following discussion .
we show that there is no steady , axisymmetric state with @xmath5 and @xmath139 for the @xmath1 field lines threading the ergosphere .
the condition @xmath5 means @xmath212 , and @xmath213 .
equation ( [ eq : basic ] ) is reduced to @xmath214 in the ergosphere , where @xmath215 , one has @xmath140 .
note that @xmath216 is a scalar , and thus @xmath217 is valid also in the ks coordinates .
the @xmath0 field stronger than the @xmath1 field drives the poloidal currents to flow across the poloidal @xmath1 field lines , as discussed in section [ sec : particle ] .
these poloidal currents generate @xmath151 ( equation [ eq : am2 ] ) .
therefore , the state with @xmath5 and @xmath139 can not be maintained .
the poloidal currents ( i.e. , the charged particle flows ) across the poloidal @xmath1 field lines due to the strong @xmath0 field in the ergosphere change the charge density distribution .
this reduces the strength of the @xmath0 field .
( we show the charge density distribution calculated for the wald @xmath218 field in appendix [ sec : app3 ] , which might be helpful for understanding the reduction of the @xmath0 field strength by the charged particle flows across the poloidal @xmath1 field lines . )
then equation ( [ eq : basic0 ] ) implies that @xmath4 is realized and one finds a non - zero @xmath75 field . from this argument
, we can conclude that the origin of the electromotive force is ascribed to the ergosphere in the unipolar induction of the bh magnetosphere with @xmath6 .
calculated for the wald vacuum solution in the bl coordinates , as functions of @xmath101 for @xmath219 ( solid ) , @xmath220 ( dashed ) , and @xmath221 ( dotted ) .
the spin parameter is set as @xmath103 . for comparison , @xmath222 ( @xmath223 ) in the plasma - filled case with @xmath5 and @xmath139 for @xmath219
is plotted by the dot - dashed line .
the vertical line represents the event horizon radius @xmath224 .
the outer boundary of the ergosphere is @xmath225 for @xmath219 . ]
the generation of such a strong @xmath0 field may be understood as a phenomenon similar to the pulsar case . in the vacuum case
, we can straightforwardly calculate @xmath222 by using the wald solution ( see appendix [ sec : app1 ] ) , and find that the region where @xmath140 is only in the vicinity of the event horizon at the equatorial plane , as shown in fig .
[ fig : dbbl ] .
in contrast , one has @xmath226 in the plasma - filled case with @xmath5 and @xmath139 ( equation [ eq : omegaf0case ] ) , which is larger than unity in the entire ergosphere ( see the dot - dashed line in fig .
[ fig : dbbl ] ) .
the enhancement of the electric field in the plasma - filled case compared to the vacuum case is quite similar to the pulsar case ( see equations [ eq : pulsare1 ] and [ eq : pulsare2 ] in section [ sec : pulsar_potential ] ) .
the charge distribution of the plasma screen the @xmath0 field component along the @xmath1 field but enhances the total strength of the @xmath0 field .
note that the condition @xmath140 is not due to a shortage of the number of charged particles like the gap with non - zero electric field _ along _ the magnetic field lines @xcite , but rather , it arises due to a sufficiency of the charged particles sustaining @xmath6 .
the divergence of @xmath222 near the event horizon in fig .
[ fig : dbbl ] is not physical , just due to the bl coordinate singularity . in appendix
[ sec : app1 ] , we calculate @xmath222 in the ks coordinates , which does not show any divergence ( see fig . [
fig : db ] ) .
we also find that no @xmath1 field lines can have the condition of @xmath5 and @xmath197 .
along such @xmath1 field lines , the poloidal electromagnetic angular momentum flux is non - zero , but the poloidal poynting flux is zero ( see equations [ eq : am2 ] and [ eq : energy2 ] ) .
the current closure requires that such a @xmath1 field line should have a part where the poloidal currents cross this field line .
focusing on the currents crossing the field line at the far zone , one finds that the @xmath76 force acts on the matter , converting the poloidal momentum flux of the electromagnetic field to that of the matter .
the matter should also gain the poloidal energy flux from the electromagnetic field , which conflicts with the condition of @xmath5 .
therefore , the state with @xmath5 and @xmath197 must not be realized . as discussed above ,
the condition @xmath4 is inevitably satisfied for all the @xmath1 field lines threading the ergosphere in the steady , axisymmetric state .
this condition leads to the azimuthal drift motions of the particles at the far zone , similar to the pulsar case .
for the _ open _ @xmath1 field lines that we consider , they must have the condition @xmath227 so that the azimuthal drift speeds do not exceed the light speed .
therefore , the open @xmath1 field lines threading the ergosphere will be forced to have the condition @xmath4 and @xmath197 .
we have seen that @xmath151 is generated by the poloidal currents driven to flow across the poloidal @xmath1 field line .
then the field line should have a part where @xmath140 in the ergosphere ( see also section [ sec : structure ] below ) in order to maintain the poloidal currents , although those currents reduce the strength of the @xmath0 field ( leading to a larger value of @xmath3 ) .
therefore , the value of @xmath3 will be regulated so that @xmath4 and the @xmath1 field line keeps having a part where @xmath140 .
it should be noted that the ergosphere is causally connected to the region around the outer light surface , which is inside the outer fast magnetosonic surface ( cf .
* ; * ? ? ?
* ; * ? ? ?
thus the steady state with @xmath4 and @xmath197 can be established in principle , although more quantitative analyses on the stability of this state are needed .
if we have any @xmath1 field lines which penetrate the ergosphere but are not open , e.g. , field lines having the cylindrical shapes and not crossing the outer light surfaces , they can be in the steady state with @xmath4 and @xmath139 .
such field lines do not have a region where @xmath140 ( i.e. , @xmath3 is larger than the case of @xmath197 ) . field line threading the equatorial plane in the ergosphere .
the direction of the @xmath0 field changes at the point of @xmath228 .
this corresponds to the change of the @xmath229 drift direction .
the current crossing region is the region of @xmath140 , where @xmath151 is not constant , and located below the inner light surface ( @xmath199 ) . ]
let us call the region where @xmath140 and then the poloidal currents are driven to flow across the poloidal @xmath1 field line
current crossing region . " in the case of @xmath230 and @xmath139 , the current crossing region has been found to be the ergosphere in section [ sec : active ] . where is the current crossing region in the case of @xmath4 and @xmath197 ?
here we focus on an open @xmath1 field line _ threading the equatorial plane _ in the ergosphere , showing a self - consistent electromagnetic structure along such a field line and pinning down the region where @xmath140 .
first , since @xmath4 , there is a point where @xmath228 for each field line .
one has @xmath231 below this point , while @xmath232 above this point , as illustrated in fig .
[ fig : structure ] .
the point of @xmath228 is located between the outer light surface ( where @xmath233 ) and the inner light surface ( where @xmath234 ) .
we plot the contour of @xmath235 in fig .
[ fig : fun ] , which shows that @xmath235 is positive only in the ergosphere .
thus @xmath4 requires that the inner light surface is located in the ergosphere . because of the symmetry , @xmath139 on the equatorial plane
the current crossing region must include the equatorial plane , and @xmath151 is generated in this region ( see equation [ eq : am2 ] ) .
equation ( [ eq : basic ] ) implies that @xmath140 is satisfied in the region of @xmath236 .
therefore , _ the current crossing region must be below the inner light surface , which is within the ergosphere .
_ above the current crossing region , @xmath169 is satisfied and @xmath237 .
( here we do not consider the region where the currents flow across the poloidal @xmath1 field lines in the direction of @xmath75 , i.e. , the particle acceleration zone . )
we summarize the electromagnetic structure in fig .
[ fig : structure ] . in the current crossing region , one has @xmath231 , so that @xmath238 and @xmath239 , which generate the poloidal electromagnetic angular momentum and energy fluxes , as seen in equations ( [ eq : am2 ] ) and ( [ eq : energy2 ] ) .
similar to the pulsar case ( discussed in section [ sec : pulsar_current ] ) , the poloidal currents are driven to flow across the poloidal @xmath1 field lines in the direction of @xmath9 in the region where the outward energy flux is generated . in the main body of the open field line region in the northern ( southern ) hemisphere , one has @xmath240 ,
i.e. , the poloidal currents flow downward ( upward ) . the poloidal particle drift motions in the coordinate basis ( which is in the direction of @xmath241 as shown in appendix [ sec :
app2 ] ) are directed towards infinity in the region of @xmath202 , while towards the equatorial plane in the region of @xmath204 .
the particles which drift towards the equatorial plane do not pass through the plane but turn to move parallel to the plane due to the strong @xmath0 field . for the @xmath1 field lines threading the equatorial plane in the ergosphere as functions of @xmath101 on the equatorial plane .
the spin parameter is set as @xmath103 .
the vertical line represents @xmath224 .
the radius of the outer boundary of the ergosphere is @xmath242 . ]
a necessary condition for maintaining the electromagnetic structure along a @xmath1 field line threading the equatorial plane in the ergosphere ( fig .
[ fig : structure ] ) is that the size of the current crossing region is finite . in the limit of the infinitesimally thin current crossing region ,
@xmath243 on the equatorial plane .
since @xmath139 there , equation ( [ eq : basic ] ) implies @xmath244 .
this condition provides the maximum allowed value of @xmath3 , i.e. , we find the range of possible value of @xmath3 as @xmath245 the value of @xmath246 normalized by the bh angular velocity @xmath247 as a function of @xmath101 on the equatorial plane is plotted in fig .
[ fig : omega ] .
since the poloidal current flows reduce the strength of the @xmath0 field , as discussed above , @xmath3 is expected to have a value near @xmath246 .
[ fig : omega ] implies that @xmath3 should be zero for the @xmath1 field line threading the ergosphere at @xmath248 , which we call last ergospheric field line .
" the poloidal currents will not cross this field line , and thus @xmath139 for this field line .
the poloidal current circuit can consist of the current flow in the current crossing region , the outward flow along the last ergospheric field line , and the inward flow in the main body of the open field line region ( see fig .
[ fig : current ] and section [ sec : discussion ] ) .
similar to the pulsar case , it is reasonable that @xmath249 , where we have used the equations valid at the outer light surface @xmath250 , @xmath251 , and @xmath252 .
when the open poloidal magnetic field roughly scales as @xmath253 ( i.e. , monopole - like rather than dipole ) , where @xmath254 is the field strength in the vicinity of the bh and @xmath255 is the radius of the outer light surface , and then one has the luminosity per solid angle as @xmath256 here @xmath257 is expected to be @xmath258 ( fig . [
fig : omega ] ) for the @xmath1 field lines threading the equatorial plane in the ergosphere .
we consider the blandford - znajek process as the steady unipolar induction process in the kerr bh magnetosphere in which a collisionless plasma is filled so that @xmath6 is sustained and the energy density is dominated by the electromagnetic field . the origin of the electromotive force in this process is the ergosphere , unlike in the pulsar case , in which the origin of the electromotive force is the rotation of the stellar matter ( see section [ sec : pulsar ] ) .
all the open @xmath1 field lines threading the ergosphere inevitably keep having a part where the @xmath0 field ( perpendicular to the @xmath1 field ) is stronger than the @xmath1 field , i.e. , @xmath140 , which drives the poloidal currents to flow across the poloidal @xmath1 field lines ( @xmath197 ; see equation [ eq : am2 ] ) and give rise to the electromotive force , i.e. , @xmath4 . in the current crossing region ,
the currents flow in the direction of @xmath9 , which generate the poloidal poynting flux ( see equation [ eq : energy2 ] ) , similar to the pulsar case .
the condition @xmath259 in the current crossing region implies that the force - free approximation can not be assumed ( see section [ sec : particle ] ) .
note that we only assume @xmath260 , not utilizing the force - free condition @xmath261 in our arguments .
we have shown the self - consistent electromagnetic structure along a @xmath1 field line threading the equatorial plane in the ergosphere in fig .
[ fig : structure ] . also for the @xmath1 field lines threading the event horizon , the above conclusion on the origin of the electromotive force
is applicable , and thus the electromagnetic structure will be similar to fig . [ fig : structure ] , having the current crossing region just outside the horizon .
it is not an easy task to solve the whole electromagnetic structure of the bh magnetosphere , but our arguments so far allow us to conjecture that the solution looks like fig .
[ fig : current ] .
we describe the poloidal current flows by the open arrows in fig .
[ fig : current ] .
some fraction of the poloidal currents can flow across the horizon , while the remaining fraction flows across the poloidal @xmath1 field lines just outside the horizon , generating the poloidal poynting flux .
in the current crossing region around the equatorial plane , the negatively charged particles flow across the horizon .
the current circuits will be made for the bh not to charge up in the steady state .
field lines , and the open arrows the poloidal currents .
see the texts in sections [ sec : bhuni ] and [ sec : discussion ] for details . ] in the membrane paradigm @xcite , the force - free condition is assumed to be satisfied except on the membrane ( at @xmath262 ) covering the horizon , so that all the poloidal currents flow into the membrane ( i.e. , flow across the horizon for freely falling observers ) in the region of the @xmath1 field lines threading the horizon .
the poloidal currents are then regarded as flowing along the viscous membrane and generating the outward poynting flux . however , this picture looks unphysical , because the membrane is causally disconnected with its exterior @xcite .
the energy equation ( [ eq : energy2 ] ) implies that a region of @xmath239 has to be causally connected for producing the poynting flux .
such a region is realized outside the horizon ( or the membrane ) and within the ergosphere , where the force - free condition is violated .
we have deduced the maximum value of @xmath3 , @xmath246 , for the poloidal @xmath1 field lines threading the equatorial plane in the ergosphere as shown in fig .
[ fig : omega ] .
this can be deduced easily by utilizing the symmetry condition on the equatorial plane @xmath139 for equation ( [ eq : basic ] ) . in contrast , for the poloidal @xmath1 field lines threading the horizon , @xmath246 is difficult to deduce , because @xmath197 at the horizon in general .
furthermore , the bl coordinates are not useful for analyzing the condition near the horizon .
equation ( [ eq : d ] ) can be rewritten _ in the ks coordinates _ as @xmath263 where @xmath264 .
this implies that @xmath246 , which is obtained for @xmath243 at the horizon , depends on the structure of @xmath218 .
we have assumed that some field lines of @xmath218 produced by the external currents are threading the equatorial plane within the ergosphere in sections [ sec : structure ] and [ sec : range ] and in fig .
[ fig : current ] .
however , a situation could be realized in which the magnetospheric currents around the equatorial plane become so strong that no field lines intersect the equatorial plane within the ergosphere , as demonstrated by some mhd numerical simulations @xcite .
the blandford - znajek process can operate only for the field lines threading the horizon in this case .
the ideal mhd condition ( @xmath265 ) is @xmath266 as viewed by fidos .
@xcite examined the two - fluid effects on the electron - positron plasma in the pulsar magnetosphere with a monopole magnetic field and the number density much larger than the goldreich - julian density , and showed that the one - fluid flow with the ideal mhd condition is a good approximation in the region @xmath267 ( see also * ? ? ?
* ; * ? ? ?
* ) . in this case
, the ideal mhd condition does not imply the infinite isotropic conductivity , but rather , the flow is almost force - free and moving via the @xmath268 drifts , i.e. , @xmath183 .
most region in the bh magnetosphere may be described as the one - fluid with the ideal mhd condition or the force - free plasma in the same sense .
however , this has to be violated in the current crossing region , as stated above .
( the ideal mhd condition requires @xmath269 from @xmath270 . ) for the pulsar case , the fluid at and inside the stellar surface has a high isotropic conductivity owing to the effective particle collisions , i.e. , the particle collision rate is very high compared with the gyration period ( cf . * ? ? ?
* ; * ? ? ?
* ) , in which the poloidal currents flow across the poloidal magnetic field lines and the electromotive force arises .
the ideal mhd condition is a good approximation there .
@xcite examined the one - fluid mhd flow in the kerr bh magnetosphere with a small spin parameter @xmath271 and a monopole magnetic field , and argued that there are many types of solutions corresponding to different values of @xmath151 and @xmath3 ( see their table 1 ) .
the mhd flow in the kerr bh magnetosphere generally consists of an inflow and an outflow , and they assumed that there is a gap between them , in which @xmath137 and the poloidal currents flow across the poloidal @xmath1 field lines .
some types of those solutions have a surface of @xmath272 in the inflow , inflow with shock " ( see also * ? ? ?
* ; * ? ? ?
* ) , which appear to be consistent with our conclusion , while one of those solutions has no surfaces of @xmath272 for which the ideal mhd condition is satisfied in the whole region except the gap region .
whether the latter type of the solution is realized depends on the nature of the gap and possibly on the assumptions on @xmath273 and the magnetic field structure , and this point remains to be further investigated .
the plasma may keep being injected into the whole kerr bh magnetosphere through the electron - positron pair creation by collisions of two photons emitted from the external hot fluid and/or through electron - proton creation by decays of neutrons escaping from the external fluid @xcite .
the electrons accelerated due to the strong @xmath0 field in the current crossing region produce photons , which also can create electron - positron pairs by collisions with the external photons .
however , it will be hard for those electron - positron pairs to be distributed widely in the magnetosphere , because the poloidal drift motions ( @xmath241 ) are downward in the region of @xmath204 ( see fig . [ fig : structure ] ) .
the origin of the electromotive force in the blandford - znajek process is attributed to the ergosphere .
it can not work in the schwarzschild bh magnetosphere , which has no ergosphere .
then it is expected that the rotational energy of the kerr bh decreases by this process .
however , it is not evident how the bh rotational energy is converted to the poynting flux .
this energy conversion mechanism has been discussed so far by utilizing the ideal mhd simulation results , focusing on the importance of generation of negative particle energies @xcite . we have argued that the breakdown of the ideal mhd condition is essential for giving rise to the electromotive force .
further analyses on the energy transport are needed to fully understand the physics of the blandford - znajek process ( see also * ? ? ?
we thank the anonymous referee for his / her comments , which improved the manuscript .
thanks shinpei shibata , hideyuki tagoshi , yuto teraki , and shigeo kimura for useful discussions .
this work is partly supported by jsps research fellowships for young scientists no.231446 .
99 asano k. , & takahara f. , 2009 , apj , 690 , l81 bardeen j.m . ,
press w.h . , & teukolsky s.a . , 1972 ,
apj , 178 , 347 becker p.a .
, das s. , & le t. , 2011 , apj , 743 , 47 bejger m. , piran , t. , abramowicz m. , & hakanson f. 2012 , phys . rev .
, 109 , 121101 bekenstein j.d . , & oron e. , 1978 , prd , 18 , 1809 beskin v.s . , 2010 ,
uspekhi , 53 , 1199 beskin v.s . , gurevich a.v . , & istomin y.n . , 1983 , sov
jetp , 58 , 235 beskin v.s . , &
kusnetsova i.v . , 2000 , nuovo cimento b , 115 , 795 beskin v.s .
, istomin y.n .
, & pariev v.i
. , 1992 , sov .
, 36 , 642 beskin v.s . , &
. , 2000 , mnras , 313 , 433 blandford r.d . , & znajek r.l . , 1977 , mnras , 176 , 465 contopoulos i. , kazanas d. , & papadopoulos d.b . , 2013 , apj , 765 , 113 goldreich p. , julian w.h .
, 1969 , apj , 157 , 869 goodwin s.p . , mestel j. , mestel l. , & wright g.a.e . , 2004 , mnras , 349 ,
213 hirotani k. , & okamoto i. , 1998 , apj , 497 , 563 khanna , r. , 1998 , mnras , 294 , 673 king a.r .
, lasota j.p . , & kundt w. , 1975 , phys .
d , 12 , 3037 koide s. , shibata k. , kudoh t. , & meier d.l .
, 2002 , science , 295 , 1688 komissarov s.s .
, 2004a , mnras , 350 , 427 komissarov s.s .
, 2004b , mnras , 350 , 1431 komissarov s.s . , 2005 , mnras , 359 , 801 komissarov s.s . , 2009 , j. korean phys .
soc . , 54 , 2503 komissarov s.s . , &
barkov m.v . , 2009 , mnras , 397 , 1153 komissarov s.s . , &
mckinney j.c . , 2007 , mnras , 377 , l49 landau l.d . , & lifshitz e.m . , 1975 , the classical theory of fields , course of theoretical physics , vol .
2 ( new york : pergamon ) levinson a. , 2006 , in trends in black hole research , edited by kreitler p.v .
( nova science publishers , new york ) , 119 ( arxiv : astro - ph/0502346 ) lovelace r.v.e . , 1976 , nature , 262 , 649 mckinney j.c .
, 2006 , mnras , 368 , 1561 melatos a. , & melrose d.b . , 1996 , mnras , 279 , 1168 menon g. , & dermer c.d . , 2005 ,
apj , 635 , 1197 menon g. , & dermer c.d . , 2011 , mnras , 417 , 1098 okamoto i. , 2012 , publ .
japan , 64 , 50 paczynski b. , 1990 , apj , 363 , 218 penna r.f .
, narayan r. , & sadowski a. , mnras , 436 , 3741 penrose r. , 1969 , nuovo .
, 1 , 252 punsly b. , 2008 , black hole gravitohydromagnetics , 2nd edition .
springer , berlin punsly b. , & coroniti f.v . , 1989 , phys .
d , 40 , 3834 punsly b. , & coroniti f.v .
, 1990 , apj , 354 , 583 ruiz m. , palenzuela c. , galeazzi f. , & bona c. , 2012 , mnras , 423 , 1300 takahashi m. , nitta s. , tatematsu y. , & tomimatsu a. , 1990 , apj , 363 , 206 tchekhovskoy , a. , narayan , r. , & mckinney , j.c . , 2011 , mnras , 418 , l79 thorne k.s . , macdonald d.a . , 1982 , mnras , 198 , 339 thorne k.s . , price r.h . , & macdonald d.a . , 1986 , the membrane paradigm .
yale univ . press , new haven , ct toma k. , & takahara f. , 2012 , apj , 754 , 148 toma k. , & takahara f. , 2013 , prog .
phys . , 2013 , 083e02 wald r.m . , 1974 ,
d , 10 , 1680
@xcite showed a vacuum test electromagnetic field solution in kerr space - time for which the magnetic field is uniform , parallel to the rotation axis , at infinity with the strength @xmath274 as @xmath275 } + 2a \partial_{[\mu } \chi_{\nu]}),\ ] ] where @xmath273 is the dimensionless spin parameter .
the covariant forms of the @xmath0 and @xmath1 fields are given by @xmath276 @xcite . in this sense ,
@xmath0 and @xmath1 are the electric and magnetic fields as measured by fidos . here
we calculate the strengths of the @xmath0 and @xmath1 fields of the vacuum solution in the bl and ks coordinates . in the bl
coordinates , one has the following non - zero metric components @xmath277 where @xmath278 ( see e.g. * ? ? ? * ) .
fidos can have a local orthonormal basis as @xmath279 which satisfy @xmath280 .
the corresponding basis vectors are @xmath281 note that @xmath282 and @xmath283 .
then we can calculate the vector components @xmath164 and @xmath165 in respect of the orthonormal basis as @xmath284 the result is @xmath285 , \nonumber \\ & & \hat{d}_\theta = \frac{b_0 a^3 \sqrt{\delta}}{\sqrt{\sigma } \rho^4 } 2 r \sin\theta \cos\theta ( 1+\cos^2\theta ) , \label{eq : blfield } \\ & & \hat{b}^r = \frac{b_0}{\sqrt{\sigma } \rho^4 } \cos\theta \left[(r^2+a^2 ) ( \rho^4 - 4ra^2\cos^2\theta ) - 2ra^4 \sin^4\theta \right ] , \nonumber \\ & & \hat{b}^\theta = \frac{-b_0 \sqrt{\delta}}{\sqrt{\sigma } \rho^4 } \sin\theta \left [ r\rho^4 + a^2 ( r^2-a^2\cos^2\theta)(1+\cos^2\theta)\right ] .
\nonumber\end{aligned}\ ] ] this agrees with the result of the same calculation in @xcite and @xcite . because @xmath286 and @xmath287 are scalars and @xmath168 and @xmath288 , one has @xmath289 and @xmath290 .
we calculate @xmath291 and plot it in figure [ fig : dbbl ]
. in the ks coordinates , one has the following non - zero metric components @xmath292 where @xmath293 @xcite . by transferring the ks fido motion into the bl
coordinates , one finds that the ks fido has the same angular velocity as the bl fido but also moves radially towards the true singularity . the orthonormal basis dual - vectors
can be chosen as @xmath294 and the corresponding basis vectors are @xmath295 @xcite .
note that @xmath282 and @xmath296 , similar to those in the bl coordinates .
then by calculating equation ( [ eq : fido_db ] ) we obtain @xmath297 , \nonumber \\ & & \hat{d}_r = \nonumber \\ & & \frac{b_0 a}{\sqrt{\sigma}\rho^4 } \left[2r^2 \rho^2 \sin^2\theta - ( r^2+a^2)(r^2-a^2 \cos^2\theta ) ( 1+\cos^2\theta)\right ] , \nonumber \\ & & \hat{d}_{\theta } = \frac{b_0 a^3}{\rho^5 \sqrt{1+z } } 2r \sin\theta \cos\theta ( 1+\cos^2\theta ) , \label{eq : ksfield}\\ & & \hat{b}^{\varphi } = \frac{b_0 a^3}{\sqrt{\sigma } \rho^5 \sqrt{1+z } } 4r^2 \sin\theta \cos\theta ( 1+\cos^2\theta ) , \nonumber \\ & & \hat{b}^r = \frac{b_0}{\sqrt{\sigma}\rho^4 } \cos\theta \left[(r^2+a^2 ) ( \rho^4 - 4ra^2\cos^2\theta ) - 2ra^4 \sin^4\theta \right ] , \nonumber \\ & & \hat{b}^\theta = \frac{-b_0}{\rho^5\sqrt{1+z } } \sin\theta \left [ r\rho^4 + a^2 ( r^2-a^2\cos^2\theta)(1+\cos^2\theta)\right ] .
\nonumber\end{aligned}\ ] ] calculated for the wald vacuum solution in the ks coordinates , as functions of @xmath101 for @xmath219 ( solid ) , @xmath220 ( dashed ) , and @xmath221 ( dotted ) .
the spin parameter is set as @xmath103 .
the vertical line represents the event horizon radius @xmath224 .
the radius of the outer boundary of the ergosphere is @xmath225 for @xmath219 . ] in fig .
[ fig : db ] we plot @xmath298 calculated in the ks coordinates .
it does not show the divergence at the event horizon , unlike that calculated in the bl coordinates as shown in fig .
[ fig : dbbl ] . note that the @xmath0 fields ( as well as the @xmath1 fields ) in the bl and ks coordinates are not identical , correspondingly to the difference of @xmath299 in equation ( [ eq : bddef ] ) .
thus @xmath300 , @xmath301 , @xmath222 are different in the two coordinates , but @xmath216 is a scalar , providing the same value in the two coordinates .
this can be confirmed directly by using equations ( [ eq : blfield ] ) and ( [ eq : ksfield ] ) . as a result , the region of @xmath302 where @xmath303 ( i.e. , @xmath304 ) are identical in fig .
[ fig : dbbl ] and fig .
[ fig : db ] .
the equation of a particle motion as viewed by fidos is described as @xmath305 where @xmath306 , @xmath159 , and @xmath160 are the proper time , charge , and mass of a particle , respectively . for @xmath307 in the bl
coordinates as an example , one has @xmath308 where we have used @xmath309 ( see * ? ? ?
thus one obtains @xmath310 the same form of equation is obtained also for @xmath311 and @xmath312 . by taking account of @xmath313 and the assumption that the gravitational force is negligible compared with the lorentz force , one obtains equation ( [ eq : eom ] ) .
the same form of the equation is obtained also by using the ks coordinates .
let us examine the drift motion of a particle in the plasma with @xmath6 in the bl coordinates . from equation ( [ eq : d ] )
, one has @xmath314 , @xmath315 , and @xmath316 .
the drift three - velocity as viewed by fidos is @xmath317 , where @xmath318 .
the azimuthal drift velocity is calculated as @xmath319 in general , one has @xmath320 and @xmath321
. then @xmath322 .
therefore , one finds @xmath323 when @xmath324 , one has @xmath325 . in this case , @xmath326 at the light surfaces , where @xmath327 . when @xmath328 , one has non - zero poloidal components of the drift velocity , @xmath329 then one has @xmath330 and @xmath331 .
this means that @xmath332
we derive a description of the charge distribution which is valid for @xmath333 in the case of @xmath205 and @xmath334 in the bl coordinates . by using equations ( [ eq : d ] ) and ( [ eq : relation_h ] )
, one has @xmath335 \nonumber \\ & = & \frac{\gamma_{\varphi\varphi}(\omega_{\rm f}-\omega)}{\alpha^2 \sqrt{\gamma } } ( \partial_r h_\theta - \partial_\theta h_r ) \nonumber \\ & & + \frac{1}{\sqrt{\gamma } } ( h_\theta \partial_r - h_r \partial_\theta ) \frac{\gamma_{\varphi\varphi}(\omega_{\rm f } - \omega)}{\alpha^2 } \nonumber \\ & & -\frac{\gamma_{\varphi\varphi}(\omega_{\rm f}-\omega)\omega}{\alpha^2 } \nabla \cdot \bmath{d } \nonumber \\ & & - \bmath{d } \cdot \nabla \left[\frac{\gamma_{\varphi\varphi}(\omega_{\rm f}-\omega)\omega } { \alpha^2}\right].\end{aligned}\ ] ] the particle drift motions carry the current as measured by fidos , @xmath336 , which is equivalent to @xmath337 . by using @xmath338 ( equation [ eq : jj ] ) , one has @xmath339 then one obtains ( by using equation [ eq : relation_h ] again ) @xmath340\end{aligned}\ ] ] this equation can be reduced to @xmath341 ( b_\theta \partial_r - b_r \partial_\theta)\omega \right .
\nonumber \\ & & \left .
- \frac{\alpha}{\sqrt{\gamma } } ( \omega_{\rm f}-\omega ) ( b_\theta \partial_r - b_r \partial_\theta)\frac{\gamma_{\varphi\varphi}}{\alpha^2 } - \frac{\gamma_{\varphi\varphi}}{\alpha } b_p^2 \frac{d\omega_{\rm f}}{d\psi } \right\}\end{aligned}\ ] ] the charge density @xmath122 can be calculated if @xmath218 and @xmath342 are given .
field with @xmath6 and @xmath5 .
the surface of @xmath343 is represented by the thin black line .
one has @xmath49 under this line and outside the event horizon ( the thick black line ) .
the inner light surface is identical to the outer boundary of the ergosphere ( the dashed line ) . ]
let us calculate @xmath122 by assuming @xmath218 as the wald vacuum solution ( equation [ eq : blfield ] ) .
this corresponds to the goldreich - julian charge density for the kerr bh magnetosphere . for @xmath344
, we confirmed that our calculation result of the surface of @xmath343 is consistent with fig .
3 of @xcite . for @xmath5
, we obtain the result as shown in figure [ fig : cdd11 ] .
field with @xmath6 and @xmath3 given by equation ( [ eq : omegaf_ref ] ) .
the surface of @xmath345 is represented by the solid lines . in the region surrounded by the two solid lines and the event horizon ( the thick solid line )
, one has @xmath346 .
the inner dashed line represents the inner light surface , and the outer dashed line the surface of @xmath228 . ] for the case of non - uniform @xmath3 , the flux function @xmath145 is required to calculate @xmath122 . for the wald solution of the @xmath1 field ,
it is given as @xcite @xmath347 it is confirmed that this form of @xmath145 provides the @xmath1 field in equation ( [ eq : blfield ] ) through equation ( [ eq : psi ] ) .
as an example , we assume the wald solution of the @xmath218 field and @xmath3 given as @xmath348 where @xmath349 is the cylindrical radius at which a @xmath1 field line crosses the equatorial plane .
this form satisfies @xmath350 for the @xmath1 field lines threading the equatorial plane outside the event horizon , where @xmath246 is given by equation ( [ eq : omegaf_cal ] ) ( see also fig . [
fig : omega ] ) .
the calculation result is shown in fig .
[ fig : cddm2 ] .
|
one of the mechanisms widely considered for driving relativistic jets in active galactic nuclei , galactic microquasars , and gamma - ray bursts is the electromagnetic extraction of the rotational energy of a central black hole , i.e. , the blandford - znajek process , although the origin of the electromotive force in this process is still under debate .
we study this process as the steady unipolar induction in the kerr black hole magnetosphere filled with a collisionless plasma screening the electric field ( the @xmath0 field ) along the magnetic field ( the @xmath1 field ) , i.e. , @xmath2 = 0 .
we extend the formulations and arguments made by komissarov , and generally show that the origin of the electromotive force is ascribed to the ergosphere .
it is explicitly shown that open magnetic field lines penetrating the ergosphere have a region where the @xmath0 field is stronger than the @xmath1 field in the ergosphere , and it keeps driving the poloidal currents and generating the electromotive force and the outward poynting flux .
the range of the possible value of the so - called angular velocity of the magnetic field line @xmath3 is deduced for the field lines threading the equatorial plane in the ergosphere .
we briefly discuss the relation between our conclusion and the ideal magnetohydrodynamic condition .
[ firstpage ] black hole physics magnetic fields relativistic processes .
|
consider the following initial - boundary value problem for the defocusing nonlinear schrdinger equation on the positive half - line @xmath0 with given initial data : @xmath1 and with a given ( generally nonhomogeneous ) dirichlet boundary condition at @xmath2 : @xmath3 here @xmath4 is an arbitrary parameter . assuming that @xmath5 , @xmath6 , and that the compatibility condition @xmath7 holds , carroll and bu @xcite have established the existence of a unique classical global solution of this problem that is a continuously differentiable map from @xmath8 to @xmath9 and that is a continuous map from @xmath10 to @xmath11 .
the defocusing nonlinear schrdinger equation is an integrable equation , being the compatibility condition for the existence of a simultaneous general solution @xmath12 of the equation @xmath13 and also of the equation @xmath14 here @xmath15 is a complex spectral parameter , and the compatibility condition is independent of @xmath15 .
these two linear equations for @xmath12 comprise the _ lax pair _ for .
one of the earliest applications of the lax pair representation of integrable equations was the development of a transform technique based on the spectral theory of the spatial equation of the lax pair , the _ inverse - scattering transform _ , for solving initial - value problems posed for @xmath16 with initial data given at @xmath17 ; see @xcite for a pedagogical description .
more recently , a _ unified transform method _ has been developed involving the simultaneous use of both equations of the lax pair to study mixed initial - boundary value problems of various types . as a general reference for these methods that includes the specific details we will need in this paper , we refer to @xcite ;
there is also a website @xcite that summarizes the salient features of the technique and has links to many original references . for the defocusing nonlinear schrdinger equation on the half - line @xmath18 , the unified transform method first advanced in @xcite and also described in @xcite amounts to the following algorithm .
recall the pauli spin matrices @xmath19 and let @xmath20 .
firstly , define the following special solutions of the lax pair : @xmath21 and @xmath22 the spectral transforms of @xmath23 , @xmath24 , and @xmath25 are then given by @xmath26 and by elementary symmetries one also has that @xmath27 the second column of @xmath28 is analytic and bounded in @xmath15 for @xmath29 whenever @xmath30 ( and hence the same is true of @xmath31 and @xmath32 ) . the second column of @xmath33 is analytic and bounded in @xmath15 for @xmath34 whenever @xmath35 ( and hence the same is true of @xmath36 and @xmath37 ) . next , given these functions of @xmath15 , one formulates a riemann - hilbert problem .
let @xmath38 denote the contour @xmath39 with each of the four half - line arcs of @xmath40 assigned an orientation such that the domain @xmath34 lies on the left . on each of the four arcs we define a _ jump matrix
as follows : @xmath41 @xmath42 @xmath43 @xmath44 here , the spectral coefficients in the jump matrix are @xmath45 and all of the dependence on @xmath46 and @xmath47 appears explicitly through the function @xmath48 the riemann - hilbert problem is then the following .
find a @xmath49 matrix @xmath50 with the following properties : * * analyticity : * @xmath51 is analytic and uniformly bounded for @xmath52 , taking boundary values @xmath53 on each of the four rays of @xmath38 from the domain where @xmath54 . * * jump condition : * the boundary values are related on each ray of @xmath38 by the jump condition @xmath55 where @xmath56 is defined by . * * normalization : * @xmath57 as @xmath58 .
[ rhp - original ] from the solution of riemann - hilbert problem [ rhp - original ] , which depends parametrically on @xmath46 , @xmath47 , and @xmath59 , one obtains a solution of the defocusing nonlinear schrdinger equation by taking the limit @xmath60 this procedure is derived assuming the existence of a solution @xmath61 satisfying the initial and boundary conditions in addition to some other technical assumptions .
it produces the solution to the initial - boundary value problem under two conditions : * the given boundary data @xmath62 used to compute the spectral transforms from are consistent .
that is , @xmath63 must agree with ( @xmath59 times ) the neumann boundary value of the solution of the dirichlet problem whose well - posedness was established by carroll and bu @xcite .
* the function @xmath64 must have no zeros in the closed second quadrant of the complex @xmath15-plane .
this is a technical condition as otherwise riemann - hilbert problem [ rhp - original ] must be formulated differently to allow @xmath50 to have poles at these points and their complex conjugates , with prescribed residue relations .
it is conjectured that in fact @xmath64 is nonvanishing for consistent boundary data , but to our knowledge there is no proof of this in the literature .
of course the problem is that if the boundary data functions @xmath24 and @xmath25 are both independently specified as is required to calculate the spectral transforms and hence the jump matrices , then the initial - boundary value problem is overdetermined and the solution of the equation produced by the method can not generally satisfy the initial and boundary conditions ( although it will solve the differential equation in the interior of the domain ) .
on the other hand , the procedure is sadly incomplete if only the dirichlet data ( the function @xmath24 ) is specified in which case the jump matrices are indeterminate as @xmath65 can not be calculated at all from and . a central role in the unified transform theory
is therefore played by the _ global relation _ , an identity satisfied by the spectral transforms of consistent boundary data @xmath62 that encodes in the transform domain the dirichlet - to - neumann map giving @xmath66 in terms of @xmath67 and @xmath68 . under certain conditions on the dirichlet data ,
the global relation can be effectively solved , and hence the unknown neumann data is eliminated .
the class of boundary conditions for which the global relation can be solved by symmetries in the complex @xmath15-plane is called the class of _ linearizable _ boundary conditions .
unfortunately , the only type of dirichlet boundary condition known to be linearizable is the homogeneous boundary condition @xmath69 .
of course , this special case can also be handled via the usual inverse scattering transform on the whole line @xmath16 simply by extending the initial data @xmath70 to @xmath71 as an odd function .
another approach to general nonhomogeneous dirichlet boundary conditions that avoids the global relation entirely may be based on the observation that under mild conditions , given the spectral transforms @xmath72 , riemann - hilbert problem [ rhp - original ] has a unique solution for almost all @xmath73 by analytic fredholm theory ( see ( * ? ? ?
* proposition 4.3 ) ) combined with steepest descent asymptotics for large @xmath46 and @xmath47 .
the exceptional set is the zero locus of an entire scalar function of @xmath74 that does not vanish identically , i.e. , a complex curve in @xmath75 that may or may not have real points but that would be at worst a closed and nowhere - dense union of analytic arcs in the real @xmath76-plane .
this in turn implies ( by the standard arguments of the _ dressing method _ , see also the proof of proposition [ prop : solution - exists ] below ) that for those @xmath76 for which a solution exists the function @xmath61 produced by taking the limit is necessarily _ some _ solution of the defocusing nonlinear schrdinger equation .
the question is whether this solution satisfies also the ( three in total ) initial and boundary conditions that were used to generate the spectral transforms @xmath72 in the first place .
this line of reasoning suggests an iteration procedure for solving the dirichlet initial - boundary value problem for the defocusing nonlinear schrdinger equation on the half - line with general nonhomogeneous data : begin by making an initial guess for the ( unknown ) neumann boundary data , say @xmath77 for @xmath78 .
set @xmath79 and then : 1 .
take @xmath80 , and with the given dirichlet data @xmath81 and @xmath70 calculate the spectral transforms @xmath82 from . 2 .
formulate riemann - hilbert problem [ rhp - original ] with these spectral transforms and solve it .
denote the function obtained from as @xmath83 .
it is necessarily a solution of the defocusing nonlinear schrdinger equation on the quarter plane @xmath18 and @xmath78 .
3 . define @xmath84 for @xmath78 .
4 . set @xmath85 and go to step 1 . no doubt the reader can imagine various other iterative approaches like this one .
it is not the purpose of this paper to study the convergence of this algorithm , but in the spirit of the principle that finite truncations of a convergent iteration ( or infinite series ) can often provide accuracy in various asymptotic limits , we wish to explore the possibility of using _
just one iteration _ of the algorithm ( actually a slightly modified version of the first iteration , see [ sec : formulation ] for details ) to provide an asymptotic approximation of the solution of the dirichlet initial - boundary value problem in the _ semiclassical limit _
the key to the success of this procedure is to make a very good initial guess for the unknown neumann data , one that is asymptotically accurate in the semiclassical limit ( as we will rigorously prove after the fact , see theorem [ theorem : boundary - condition - recover ] ) .
that is , what we need is an explicit approximation of the dirichlet - to - neumann map for .
let us now explain the approximation of the dirichlet - to - neumann map for the defocusing nonlinear schrdinger equation that we plan to study in this article .
without loss of generality , we represent the complex field @xmath61 in real phase - amplitude form : @xmath87 substituting into , dividing by the common factor @xmath88 , and separating real and imaginary parts yields the following system of equations : @xmath89 this coupled system is equivalent to .
it is useful to intoduce notation for the phase gradient : @xmath90 now , in terms of @xmath91 and @xmath92 , the exact ratio between the unknown neumann data and the given dirichlet data at @xmath2 takes the form @xmath93 consider the second equation of the system at @xmath2 along with in the formal semiclassical limit @xmath94 , assuming that @xmath95 .
this means that we simply neglect the terms explicitly proportional to @xmath59 or @xmath96 in each case , yielding the formal approximations : @xmath97 our approach is to assume that the known dirichlet boundary data is specified in the form @xmath98 with @xmath99 and @xmath100 are given real - valued functions independent of @xmath59 .
obviously we then have @xmath101 and @xmath102 , so we may rewrite the approximate relations as @xmath103 assuming further that @xmath104 we solve the first of these relations for the unknown phase derivative @xmath105 at the boundary : @xmath106 that is , @xmath107 is the ( real valued ) formal semiclassical approximation of the exact phase derivative @xmath105 . finally , for dirichlet boundary data satisfying the condition we use the second equation of to approximate the dirichlet - to - neumann map as follows .
suppose that the dirichlet boundary data @xmath81 of the form satisfies @xmath108 .
the _ semiclassical approximation of the dirichlet - to - neumann map _ is defined by @xmath109 where @xmath107 is given in terms of the phase and amplitude of the known dirichlet data by .
[ definition : d - to - n ] the key point of our approach is that by neglecting the formally small dispersive terms in we obtain a system that is first - order in @xmath46 and hence allows the unknown neumann data to be _ explicitly eliminated _ in favor of @xmath47-derivatives that may be computed along the boundary from the given dirichlet data .
this approximation is a purely local relation between the two boundary values , and in particular the approximate dirichlet - to - neumann map is independent of initial data @xmath23 .
we have selected the positive square root in for a specific reason , which we now explain .
differentiation of the second equation of with respect to @xmath46 produces the equivalent system @xmath110 obviously , is a formally small perturbation of a quasilinear system obtained by simply setting @xmath111 to zero .
characteristic velocities _ of the limiting system are the eigenvalues @xmath112 of the coefficient matrix @xmath113 : @xmath114 as the characteristic velocities are real and distinct ( for @xmath115 ) , the limiting quasilinear system is of hyperbolic type .
causality and local well - posedness for the cauchy problem of the hyperbolic approximating system in the quarter plane @xmath18 and @xmath78 requires that the boundary @xmath2 be a space - like curve .
in other words , we require _ both _ characteristic velocities to be strictly positive at the boundary .
this means that we will require that @xmath116 for all @xmath78 . since @xmath117 for @xmath78 it is clear that well - posedness of the limiting hyperbolic boundary - value problem requires in particular @xmath118 .
in fact , we will ensure the condition @xmath116 by imposing the stronger condition @xmath119 ; the latter condition appears to be necessary to recover the dirichlet boundary data at @xmath2 for all @xmath78 ( see remark [ remark : kb - negative ] below ) . for convenience
we restrict our attention to the already nontrivial and physically interesting case of zero initial data : @xmath120 for all @xmath18 .
however , we fix rather general nonhomogeneous dirichlet boundary data of the form for @xmath78 ( and satisfying several additional conditions allowing our procedure to succeed , see assumption [ assumption : data ] below ) , and attempt to solve the corresponding initial - boundary value problem .
the first step is the calculation of the spectral transforms @xmath121 and @xmath122 corresponding to the dirichlet data @xmath81 given by and the formally approximate neumann data @xmath77 given by definition [ definition : d - to - n ] .
the direct spectral analysis is made possible in practice because the parameter @xmath123 is presumed small , so the equations of the lax pair become singularly perturbed differential equations that may be studied by classical methods .
the results of this analysis are summarized in [ sec : spectral - analysis ] , with the corresponding proofs appearing in two appendices . despite the rigor of these results ,
there are certain difficulties that remain with directly formulating the inverse problem for the exact scattering data , so rather than calculate the solution of the defocusing nonlinear schrdinger equation corresponding to the exact spectral transforms of the ( generally incompatible ) dirichlet - neumann pair @xmath124 we modify the spectral functions in an ad - hoc fashion , but one inspired by the rigorous direct spectral analysis of the temporal problem of the lax pair .
this allows us to formulate a simpler and completely explicit version of riemann - hilbert problem [ rhp - original ] for a matrix @xmath125 ; see [ sec : formulation ] .
the simpler riemann - hilbert problem explicitly encodes the given dirichlet boundary data through two integral transforms denoted @xmath126 and @xmath127 ( these are really the semiclassical analogues of the amplitude and phase of the spectral function @xmath128 ; see ) , and its solution produces , for each @xmath123 , a solution @xmath129 of the defocusing nonlinear schrdinger equation .
the rest of the paper is concerned with analyzing this solution , paying particular attention to the semiclassical asymptotic behavior of @xmath130 at the initial time @xmath17 for @xmath18 and at the boundary @xmath2 for @xmath78 .
our first result is the following .
the solution @xmath129 of the defocusing nonlinear schrdinger equation obtained from riemann - hilbert problem [ rhp - m - tilde ] satisfies @xmath131 where the error term is uniform on @xmath132 for each @xmath133 .
[ theorem : initial - condition - small ] thus the function @xmath130 nearly satisfies the given homogeneous initial condition @xmath134 for @xmath18 .
our rigorous proof of this result is based on the steepest descent method for riemann - hilbert problems , combined with a natural generalization of the method involving @xmath135-problems @xcite . after establishing some preliminary results in [ sec : preliminary - results ] ,
we give the proof of theorem [ theorem : initial - condition - small ] in [ sec : initial - condition - small - proof ] .
our next main result is the following .
the points @xmath136 and @xmath137 are defined as part of assumption [ assumption : data ] below .
suppose that @xmath78 and @xmath138 , @xmath139 .
the solution @xmath129 of the defocusing nonlinear schrdinger equation obtained from riemann - hilbert problem [ rhp - m - tilde ] satisfies @xmath140 and @xmath141 where the error terms are uniform for @xmath47 in compact subintervals of @xmath142 .
[ theorem : boundary - condition - recover ] equation shows that the same solution @xmath130 very nearly satisfies the given nonhomogeneous dirichlet boundary condition at @xmath2 for @xmath78 .
moreover , from we see directly that the true neumann data at the boundary is indeed asymptotically consistent with the formal approximation given by definition [ definition : d - to - n ] .
the proof of this result is again based on the steepest descent method , this time augmented with the use of a complex phase function @xmath143 . after describing the general methodology and constructing the function @xmath143 in [ sec : g - function ] , we present the proof of theorem [ theorem : boundary - condition - recover ] in [ sec : boundary - condition - recover - proof ] .
the main point , however , is that the solution @xmath130 is represented also for @xmath76 not on the boundary of the quarter plane @xmath18 , @xmath78 via exactly the same riemann - hilbert problem ( see riemann - hilbert problem [ rhp - m - tilde ] ) .
this means that one may use steepest descent methods to calculate @xmath130 for small @xmath123 for positive @xmath47 and away from the boundary . for @xmath76 close to the boundary of the quarter plane
the analysis is virtually the same as it is exactly on the boundary , with similar results .
for example , a corollary of the proof of theorem [ theorem : initial - condition - small ] is the following .
the points @xmath144 and @xmath145 are specified in terms of the functions characterized by assumption [ assumption : data ] below , and @xmath127 is explicitly given by .
let @xmath35 , and let @xmath146 denote the smallest nonnegative value of @xmath147 for which the inequality @xmath148 holds for all @xmath149 whenever @xmath132
. then the solution @xmath129 of the defocusing nonlinear schrdinger equation obtained from riemann - hilbert problem [ rhp - m - tilde ] satisfies @xmath150 as @xmath151 whenever @xmath152 .
[ corollary - vacuum ] in the case that @xmath153 is convex , we may characterize @xmath146 explicitly in terms of the legendre dual @xmath154 as follows : @xmath155^*(-4t),\quad t>0,\quad\text{where}\quad f^*(p):=\sup_{k_\mathfrak{a}<k
< k_\mathfrak{b}}(pk - f(k ) ) .
\label{eq : legendre - dual}\ ] ] the proof of corollary [ corollary - vacuum ] is given in
[ sec : vacuum - domain ] .
we call the domain @xmath152 the _ vacuum domain _ corresponding to the dirichlet boundary data @xmath156 . in the vacuum domain
the solution is influenced predominantly by the homogeneous initial data rather than the nonhomogeneous boundary data in the semiclassical limit .
a concrete calculation of the vacuum domain for a particular choice of dirichlet boundary data is shown in figure [ fig : vacuumdomain ] .
( solid black curve ) and the @xmath59-independent vacuum domain ( shaded ) calculated from @xmath127 corresponding to the explicit boundary data illustrated in figure [ fig : assumption1 ] below . also shown ( dashed curves )
are explicit asymptotes to @xmath157 for small @xmath47 ( left panel , the asymptote @xmath158 ) and large @xmath47 ( right panel , the asymptote @xmath159 ) .
these asymptotes are described in [ sec : vacuum - domain ] ( see and ) . ]
another result is the following , which is essentially a corollary of the proof of theorem [ theorem : boundary - condition - recover ] . here
@xmath160 and @xmath161 are defined in terms of the dirichlet data by .
each point @xmath162 with @xmath163 has a neighborhood @xmath164 in the @xmath76-plane in which there exist unique differentiable functions @xmath165 and @xmath166 satisfying @xmath167 and @xmath168 and the partial differential equations @xmath169 and such that the solution @xmath129 of the defocusing nonlinear schrdinger equation obtained from riemann - hilbert problem [ rhp - m - tilde ] satisfies @xmath170 uniformly for @xmath171 as @xmath151 , where @xmath172 and @xmath173 . [ corollary : plane - wave ] the proof of corollary [ corollary : plane - wave ] is given in [ sec : plane - wave ] . note that eliminating @xmath174 and @xmath175 from in favor of @xmath91 and @xmath92 using yields @xmath176 which should be compared with , the defocusing nonlinear schrdinger equation written without approximation in terms of amplitude @xmath91 and phase derivative @xmath92 .
therefore , we observe that for small positive @xmath46 , @xmath130 resembles a modulated plane wave of the form for amplitude @xmath91 and phase @xmath177 independent of @xmath59 , and the modulation is described by the dispersionless nonlinear schrdinger system , or equivalently the whitham ( riemann - invariant form ) system .
this shows consistency with , and adds yet more weight to , our approximate formula for the dirichlet - to - neumann map given in definition [ definition : d - to - n ] .
indeed , the latter was formally derived under the _ initially unjustified assumption _ that the solution resembles a modulated plane wave near the boundary @xmath2 .
we call the union of the neighborhoods of the @xmath76 plane for @xmath18 in which @xmath130 is described by corollary [ corollary : plane - wave ] the _ plane - wave domain _ for @xmath130 .
we therefore see that the quarter - plane @xmath18 and @xmath78 is split up into several regions in which the approximate solution @xmath130 of the dirichlet boundary - value problem only approximately satisfies the given initial and boundary conditions , it is an exact solution of the defocusing nonlinear schrdinger equation for every @xmath123 . ] behaves quite differently .
so far we have observed the vacuum domain , in which @xmath130 simply decays to zero with @xmath123 , and the plane - wave domain , in which @xmath130 resembles a modulated plane wave with nonzero amplitude .
it is to be expected that these two domains do not exhaust the quarter plane .
while we do not pursue the topic further in this paper , the methodology presented in [ sec : g - function ] below also allows one to calculate the semiclassical behavior of @xmath130 for @xmath76 in domains not contiguous to the boundary of the quarter plane , in which ( in principle ) more complicated local behavior of @xmath130 can occur , with microstructure modeled by higher transcendental functions ( e.g. , dispersive shock waves described by modulated elliptic functions ) . see remark [ remark - multiphase ] for more information . with
the proofs of our results complete , we conclude the body of our paper with some further remarks , some indicating directions for future work , in [ sec : concluding - remarks ] . our paper represents a further contribution to the literature on the use of the unified transform to study nonlinearizable boundary value problems in various asymptotic limits .
a key observation that was made fairly early in the development of the theory was that regardless of whether the spectral functions @xmath36 and @xmath37 actually correspond to a compatible dirichlet - neumann pair @xmath62 , riemann - hilbert problem [ rhp - original ] yields to asymptotic analysis in the limit of large @xmath47 ( with @xmath178 for some nonnegative velocity @xmath179 ) by the steepest descent method .
general asymptotic properties of the solution that can be observed by such analysis therefore necessarily also describe the physical solutions of the dirichlet problem that simply correspond to the special case that the spectral functions satisfy the global relation . as a representative of this type of analysis ( for the focusing case of the nonlinear schrdinger equation ) , we cite a paper of boutet de monvel , its , and kotlyarov @xcite , where time - periodic boundary conditions are analyzed .
such analysis does not require any preliminary asymptotic analysis of the spectral functions , as they are independent of the asymptotic parameter @xmath47 .
another approach to the asymptotic solution of nonlinearizable boundary value problems is to consider the situation in which the initial and boundary data are small , in which case a perturbation scheme based on the amplitude as a small parameter can be developed in detail , and significantly this allows the global relation to be solved order - by - order .
this means that the asymptotic results obtained are guaranteed to correspond to a compatible dirichlet - neumann pair @xmath62 even though only @xmath24 is given .
the recent papers of fokas and lenells @xcite pursue this approach and obtain new convergence results showing that for small simple harmonic dirichlet boundary data , the solution is eventually periodic with the same period , at least to third order in the small amplitude .
the semiclassical limit is , in a sense , the exact opposite to the weakly - nonlinear small - amplitude limit .
indeed , the formal semiclassical limit is given by the strongly nonlinear hyperbolic system .
there is at least one other paper in the literature on the subject of semiclassical analysis of the dirichlet initial - boundary - value problem on the half - line for the defocusing nonlinear schrdinger equation , namely a paper of kamvissis @xcite , which directly stimulated our interest in this problem .
like we do , kamvissis considers general nonhomogeneous dirichlet boundary data together with homogeneous initial conditions , and he applies the steepest descent methodology for riemann - hilbert problems to deduce general properties of the solution in the semiclassical limit . our corollary [ corollary : plane - wave ] is consistent with theorem 5 of @xcite ( the main result of that paper ) albeit in the simplest case of genus @xmath180 . on the other hand ,
it is less clear whether the vacuum domain @xmath152 described by our corollary [ corollary - vacuum ] is a special case of kamvissis theorem 5 . while we study the same problem , and apply similar methods , the approach in @xcite is fundamentally different from ours , being based solely on the abstract existence result for the unknown neumann data @xmath63 corresponding to the given dirichlet data @xmath81 .
while kamvissis assumption that @xmath81 is independent of @xmath59 is quite reasonable and physically interesting corresponds to taking @xmath181 .
therefore , in a sense our results can not be compared well with those of @xcite , because we require @xmath182 to be strictly negative ( see ) .
] , his subsequent analysis of the direct spectral problem for the @xmath47-part of the lax pair ( theorems 2 and 3 of @xcite , of which our propositions [ prop : noturningpoints ] and [ prop : two - turning - points ] are analogues ) apparently rests upon the additional hidden assumption that the implicitly - defined function @xmath183 is _ also _ independent of @xmath59 ; otherwise the wkb methodology cited in ( * ? ? ? * section iii ) does not apply . since the defocusing nonlinear schrdinger equation involves the parameter @xmath59 in a singular way ,
whether this assumption is justified is certainly not obvious .
indeed one might worry that a slowly - varying dirichlet boundary condition might give rise to a neumann boundary value with rapid variations in amplitude or phase of period proportional to @xmath59 .
for example , our theorem [ theorem : boundary - condition - recover ] shows that some bounded dirichlet data @xmath67 can lead to rapidly oscillatory neumann data @xmath183 that moreover is large of size @xmath184 . our approach is to avoid abstract assumptions , and instead make a very explicit assumption , based on the modulated plane - wave ansatz , concerning the unknown neumann data as described in definition [ definition : d - to - n ] .
this allows us to justify our steepest descent analysis by ultimately tying the solution generated back to the hypothesized initial and boundary data ( theorems [ theorem : initial - condition - small ] and [ theorem : boundary - condition - recover ] ) in an explicit fashion .
this same approach leads to a very concrete description of the semiclassical dynamics of the solution in the full domain @xmath18 and @xmath78 , as in the characterization of the vacuum domain presented in corollary [ corollary - vacuum ] .
another paper that we wish to mention is work of degasperis , manakov , and santini @xcite that presents an alternate approach to the general initial - boundary value problem for the defocusing nonlinear schrdinger equation .
the method described in @xcite avoids using the @xmath47-part of the lax pair to formulate the inverse problem and instead uses the inverse theory of the spatial part of the lax pair only , at the cost of a more implicit nonlinear description of the time evolution of the jump matrices on the real line .
the fact that the inverse problem is ultimately formulated as a riemann - hilbert problem relative to the real axis may be a crucial benefit in our view ( see remark [ remark - imaginary - amplify ] ) . in the future ,
we plan to explore the possibility of using semiclassical asymptotic techniques to analyze this alternate method of studying initial - boundary value problems .
for simplicity . ] , we consider the case of vanishing initial data : @xmath185 this immediately implies that the spectral transforms defined from the differential equation satisfy @xmath186 and @xmath187 for all @xmath188 .
we take the dirichlet boundary data in the form , and for convenience we impose several conditions on the functions @xmath189 and @xmath190 for @xmath35 .
these are specified in terms of an auxiliary function @xmath191 as follows : the functions @xmath192 and @xmath193 satisfy the following conditions : * @xmath192 is real analytic for @xmath78 , strictly positive for all @xmath78 , and @xmath194 as @xmath195 for all @xmath196 and @xmath197 . also , there is a positive number @xmath198 such that @xmath199 and @xmath200 hold as @xmath201 with @xmath202 .
* @xmath193 is real analytic for @xmath78 , satisfying @xmath203 for some @xmath204 , and @xmath205 as @xmath195 for all @xmath196 and @xmath206 . also , there is a positive number @xmath207 such that @xmath208 and @xmath209 hold as @xmath201 with @xmath202 . *
the functions @xmath210 each have precisely one critical point in @xmath211 , corresponding to a nondegenerate maximum for @xmath212 at a point @xmath213 and a nondegenerate minimum for @xmath214 at a point @xmath215 .
nondegeneracy means that @xmath216 and @xmath217 .
[ assumption : data ] the square - root behavior of the amplitude @xmath189 that is specified in assumption [ assumption : data ] evidently violates the conditions for the proof of carroll and bu @xcite to guarantee the existence of a solution of the initial - boundary value problem .
nonetheless this behavior leads to additional smoothness of the integral transform @xmath127 defined in below that is useful in the proof of theorem [ theorem : initial - condition - small ] .
see remark [ remark : phi - smooth ] .
we may avoid this difficulty as follows .
let @xmath218 be a @xmath219 `` bump function '' satisfying @xmath220 for @xmath221 and @xmath222 for @xmath223 .
replacing @xmath189 by @xmath224 , by @xcite there is a unique solution @xmath225 of for each @xmath123 satisfying @xmath226 for @xmath18 and @xmath227 for @xmath78 .
we may view our results as a comparison between @xmath130 and the function @xmath228 , the latter of which exactly satisfies the given boundary condition for every @xmath78 as long as @xmath123 is sufficiently small ( given @xmath47 ) .
9999 [ remark : amplitude - fix ] we now use to define @xmath190 in terms of functions @xmath229 and @xmath191 satisfying the conditions of assumption [ assumption : data ] as @xmath230\,ds .
\label{eq : sprimeuh}\ ] ] note that as @xmath191 is real , the inequality is automatically satisfied .
we introduce the following notation : @xmath231 ( @xmath232 is well - defined as @xmath233 ) , and we set @xmath234 and @xmath235 .
note that the assumption @xmath236 guarantees that @xmath237 , and the assumption that @xmath203 guarantees that @xmath238 .
the points @xmath239 and @xmath240 lie in the interval @xmath241 , and we see that @xmath242 while @xmath243 . these definitions are illustrated for boundary data satisfying assumption [ assumption : data ] in figure [ fig : assumption1 ] .
( see ) . here
the turning point curve is shown for boundary data @xmath244 and @xmath245 consistent with assumption [ assumption : data ] .
the functions @xmath160 and @xmath161 are given in terms of @xmath246 and @xmath247 by .
the dashed ( solid ) curves in the interval @xmath248 correspond to the function @xmath249 ( the function @xmath250 ) . ] with the semiclassical approximation of the dirichlet - to - neumann map given in definition [ definition : d - to - n ] , the direct scattering problem encoding the boundary data is @xmath251 the oscillatory factors @xmath252 can be removed from the coefficient matrix by means of a simple substitution : @xmath253 indeed , making use of , this substitution leads to the equivalent system of equations @xmath254 with @xmath111-independent coefficient matrix given by @xmath255 that we need to solve subject to the boundary condition @xmath256 the corresponding spectral transforms are given for @xmath257 by @xmath258 and @xmath259 where @xmath260 denotes the first column of @xmath261 .
since @xmath123 appears both in the data @xmath124 and in the differential equation , the spectral functions @xmath262 and @xmath263 will also depend on this small parameter .
we now study this dependence rigorously in the limit @xmath151 . given any sufficiently small number @xmath204 ( not necessarily related to the constant in assumption [ assumption : data ] )
we define @xmath264 to be the closed unbounded subset of the @xmath15-plane characterized by the inequalities @xmath265 and one of the three inequalities : @xmath266 or @xmath267 or @xmath268 .
see figure [ fig : qiidelta ] . of the second quadrant of the complex @xmath15-plane . ] the eigenvalues @xmath269 of @xmath270 satisfy @xmath271=(2k - u(t))^2(k-\mathfrak{a}(t))(\mathfrak{b}(t)-k ) .
\label{eq : lambdasquared}\ ] ] given @xmath15 with @xmath257 , a positive real number @xmath78 is called a _ turning point _ for if the two eigenvalues of @xmath270 degenerate ( at @xmath272 ) .
we have the following basic fact .
[ lem : turningpoints ] suppose that assumption [ assumption : data ] holds and that @xmath265 . then there exist turning points @xmath78 precisely when @xmath15 lies in the negative real interval @xmath273 .
moreover , for each @xmath149 there exist precisely two turning points @xmath274 .
the two turning points coalesce as @xmath275 and as @xmath276 : @xmath277 and @xmath278 . also , @xmath279 as @xmath280 while @xmath281 as @xmath282
. finally , given any @xmath204 , the condition that @xmath283 bounds @xmath269 away from zero uniformly for @xmath78 .
we omit the proof as it is a direct and easy consequence of the conditions on the functions @xmath107 and @xmath189 and formula . given a value of @xmath15 , the presence or absence of turning points determines the nature of the spectral functions in the semiclassical limit .
according to lemma [ lem : turningpoints ] , there are no turning points if @xmath284 .
this implies a certain triviality of the spectral functions in this region of the @xmath15-plane .
we have the following result .
assume that @xmath120 .
let a number @xmath204 be given , and suppose that the functions @xmath285 and @xmath193 satisfy assumption [ assumption : data ] . then : * for bounded @xmath123 , any zeros in @xmath264 of the analytic function @xmath286 lie in an @xmath59-independent bounded subset . *
the analytic function @xmath287 has no zeros in @xmath264 if @xmath123 is sufficiently small .
* the function @xmath288 satisfies a bound of the form @xmath289 where the constant implicit in the estimate depends only on the functions @xmath229 and @xmath191 .
[ prop : noturningpoints ] in other words , when @xmath123 is small , then for @xmath290 , @xmath291 has no poles and is uniformly small .
the proof of this proposition is given in appendix [ sec : appendix - noturningpoints ] .
we now study the asymptotic behavior of the function @xmath288 for real @xmath15 in the interval @xmath248 . for each such @xmath15 , as can be seen in figure [ fig : assumption1 ] , the eigenvalues @xmath292 of the coefficient matrix @xmath270 satisfy @xmath293 for @xmath294 , while @xmath295 for @xmath296 and for @xmath297 .
considering @xmath123 as being very small , one is reminded of the language of the wkb method , in which the interval @xmath298 is analogous to a `` classically forbidden region '' separating two `` classically allowed regions '' .
thus we have an analogue of a quantum tunneling problem .
rather than use the wkb method , which is well - known to fail near the turning points @xmath299 , in the proof of the following results we use the method of langer transformations to uniformly handle the neighborhoods of the two turning points while simultaneously maintaining full accuracy when @xmath47 is not close to either turning point .
the presence of turning points leads to nontrivial behavior of the spectral functions in the limit @xmath151 , as the following result shows .
let @xmath149 with @xmath300 and @xmath301 , and suppose that the functions @xmath192 and @xmath193 satisfy assumption [ assumption : data ] . then in the limit @xmath151 , @xmath302\ ] ] and @xmath303\ ] ] where @xmath304\,dt - 2|k^2-k_\infty^2|t_+(k ) , \label{eq : ell - define}\ ] ] @xmath305 @xmath306 and @xmath307 the error terms are uniform in @xmath15 in compact subintervals of @xmath308 .
[ prop : two - turning - points ] the proof of proposition [ prop : two - turning - points ] is given in appendix [ sec : appendix - two - turning - points ] .
an immediate corollary is the following .
suppose that @xmath120 . under the same conditions and with the same characterization of the error terms as in proposition [ prop : two - turning - points ]
, we have @xmath309 [ corollary : gamma ] since @xmath120 we have @xmath310 and hence @xmath311 for all real @xmath312 .
the formula for @xmath313 follows from the identity @xmath314 equivalent to the condition that @xmath315 .
the factor @xmath316 is exponentially close to @xmath317 ( except near @xmath144 and @xmath145 , points excluded from consideration ) and is included in the formula for @xmath318 to ensure that the jump matrix we shall construct from this approximation has unit determinant .
propositions [ prop : noturningpoints ] and [ prop : two - turning - points ] and corollary [ corollary : gamma ] give a rigorous characterization of the spectral functions associated with vanishing initial data @xmath70 and with a class of boundary data @xmath124 given ( in terms of the functions @xmath246 and @xmath247 described by assumption [ assumption : data ] ) by and subject to the equation giving @xmath190 in terms of @xmath246 and @xmath247 . to summarize : * from @xmath319 we have @xmath310 for all @xmath320 .
therefore , riemann - hilbert problem [ rhp - original ] has no jump on the positive real axis , and the remaining jump matrices only involve @xmath291 , which is simply a ratio of the spectral functions @xmath321 and @xmath322 arising from the approximate boundary data . * if the function @xmath291 has any poles in the second quadrant of the complex plane , they must lie very close ( in the limit @xmath151 ) to the negative real interval @xmath323 $ ] .
* on the imaginary @xmath15-axis , as well as on the negative real @xmath15-axis away from the interval @xmath323 $ ] , @xmath291 is small in the limit @xmath151 . * in the interior of the negative real interval @xmath241 and away from the special points @xmath239 and @xmath240 , @xmath291 has an accurate explicit approximation given by corollary [ corollary : gamma ] .
however , this information alone is insufficient to properly formulate and analyze riemann - hilbert problem [ rhp - original ] associated with the exact spectral transforms @xmath121 and @xmath122 corresponding to the approximate neumann boundary data @xmath77 .
indeed , to formulate the riemann - hilbert problem without poles one would need to know a priori that there can not be any poles of @xmath291 whatsoever in the second quadrant , and it is not enough to know that any poles have to move toward @xmath323 $ ] as @xmath151 .
another issue is that our results do not provide approximations for @xmath291 near the real points @xmath144 , @xmath239 , @xmath240 , or @xmath145 .
in fact , the analytical methodology based on langer transformations used in the proof of proposition [ prop : two - turning - points ] either requires substantial modification or breaks down entirely in neighborhoods of these points . these arguments . ]
suggest making a further modification of the first step of the proposed iteration algorithm : we will reformulate the inverse problem by : * assuming that the riemann - hilbert problem can be formulated without poles , * neglecting @xmath291 and @xmath324 completely on the imaginary axis , * neglecting @xmath291 on the real axis for @xmath325 and @xmath326 , and * replacing @xmath291 in the whole interval @xmath241 by the formulae recorded in corollary [ corollary : gamma ] with the @xmath327 error terms set to zero .
the resulting riemann - hilbert problem has the negative real interval @xmath323 $ ] as its only jump contour . for convenience we will re - orient this contour from left to right , which requires the inversion of the jump matrix written in .
we therefore formulate the following riemann - hilbert problem .
let @xmath328 be defined by @xmath329 where @xmath330 denotes the characteristic function of the interval @xmath241 , and where @xmath331 is defined by while @xmath332 is defined by .
it can be shown that @xmath333 is hlder continuous with every exponent @xmath334 .
seek a @xmath49 matrix function @xmath335 with the following properties : * _ * analyticity * _ : @xmath336 is analytic in @xmath337 and @xmath338 and hlder continuous for some exponent @xmath339 in @xmath340 and @xmath341 , taking boundary values @xmath342 on @xmath343 from @xmath344 . * _ * jump condition * _ : the boundary values are related by @xmath345 * _ * normalization * _ : the matrix function @xmath336 satisfies @xmath346 where the limit is uniform with respect to direction in the complex plane .
[ rhp - m - tilde ] the following is a standard result .
riemann - hilbert problem [ rhp - m - tilde ] has a unique solution for every @xmath73 and for every @xmath123 .
the function @xmath347 is infinitely differentiable with respect to @xmath76 , and for every @xmath123 , @xmath129 satisfies the defocusing nonlinear schrdinger equation . [
prop : solution - exists ] to see the uniqueness of the solution of riemann - hilbert problem [ rhp - m - tilde ] ( assuming existence ) , one first notes that @xmath348 necessarily holds as an identity for any solution , and therefore @xmath349 is also analytic for @xmath350 .
given two solutions , say @xmath336 and @xmath351 , one considers the matrix ratio @xmath352 , which is analytic for @xmath350 and satisfies @xmath353 as @xmath58 .
a simple calculation using the jump condition satisfied by both @xmath336 and @xmath351 shows that the continuous boundary values taken on @xmath343 agree : @xmath354 for all @xmath320 .
it follows that @xmath355 is an entire analytic ( matrix - valued ) function of @xmath15 that tends to @xmath356 as @xmath58 , so by liouville s theorem @xmath357 for all @xmath15 , i.e. , @xmath358 holds for all @xmath350 . to establish existence of a solution
, one observes that riemann - hilbert problem [ rhp - m - tilde ] is equivalent to a system of linear singular integral equations for which the relevant operator is fredholm with zero index on an appropriate space of hlder - continuous functions ( see @xcite and @xcite ) .
it therefore suffices to show that the kernel of this fredholm operator is trivial .
triviality of the kernel is equivalent to the assertion that the only solution of the homogeneous form of riemann - hilbert problem [ rhp - m - tilde ] , in which the normalization condition is replaced with a limit of @xmath359 as @xmath58 , is the zero matrix .
zhou s _ vanishing lemma _
* theorem 9.3 ) shows that this latter assertion holds true in the present case because the jump contour is the real axis and the jump matrix has a positive semidefinite real part as a consequence of the inequality @xmath360 holding for @xmath320 .
the infinite differentiability of the matrix @xmath336 with respect to @xmath76 , and hence that of @xmath361 , follows from the compact support @xmath323 $ ] of @xmath362 .
finally , let us show that @xmath363 satisfies .
we begin by defining a matrix @xmath364 from the solution @xmath125 by setting @xmath365 one verifies that @xmath364 is analytic for @xmath350 , and that @xmath366 it follows from the fact that this jump condition is independent of @xmath46 and @xmath47 , that the matrices @xmath367 and @xmath368 have no jump across the real axis and since @xmath369 , @xmath370 and @xmath371 are entire functions of @xmath15 .
moreover , from the asymptotic behavior of @xmath364 near @xmath372 one can check that @xmath370 is a linear function of @xmath15 while @xmath371 is a quadratic polynomial in @xmath15 .
in fact , using one sees that @xmath370 is given by with @xmath363 .
moreover , using the fact that @xmath364 satisfies the differential equation @xmath373 ( by definition of @xmath370 ) , one sees that @xmath371 is given by with @xmath363 .
the fact that @xmath374 is a simultaneous fundamental solution matrix of the lax pair equations @xmath373 and @xmath375 means that these equations are consistent , that is , the zero - curvature condition @xmath376=\mathbf{0}\ ] ] holds , and substitution from and yields the equation for @xmath363 ( and the complex conjugate of that equation ) .
we note that this proof implies that @xmath125 has a convergent laurent series in descending powers of @xmath15 for @xmath377 sufficiently large : @xmath378 and that @xmath130 and @xmath379 can be expressed in terms of the coefficients @xmath380 and @xmath381 as follows : @xmath382 the question that remains is what , if anything , does the family of functions @xmath130 have to do with the exact solution @xmath228 of the dirichlet initial - boundary value problem with @xmath226 for @xmath18 and @xmath383 for @xmath78 ( recall the modified amplitude function @xmath384 defined in remark [ remark : amplitude - fix ] ) .
this is the topic we take up next .
the values of @xmath65 for @xmath15 real and positive are irrelevant to the inverse problem , as the jump matrix for @xmath385 generally only involves the function @xmath386 obtained from the initial data for @xmath18 ( see ) , and in the present case that @xmath120 , @xmath310 .
the condition @xmath119 implied by assumption [ assumption : data ] ensures that @xmath387 , and hence that the full asymptotic support of @xmath128 on @xmath343 ( and hence by definition the exact support of @xmath362 on @xmath343 ) contributes to the jump matrix for the inverse problem .
if on the contrary we had @xmath388 , then some information about the boundary data would be lost from the inverse problem in the semiclassical limit .
it therefore seems that it is possible to reconstruct the boundary data only if @xmath119 . as pointed out earlier , this is a stronger condition than the necessary condition @xmath116 for the boundary to be a spacelike curve for the hyperbolic dispersionless system .
9999 [ remark : kb - negative ] one may observe that riemann - hilbert problem [ rhp - m - tilde ] is of exactly the same form as that which occurs in the treatment of the initial - value problem for the defocusing nonlinear schrdinger equation formulated on an appropriate space of decaying functions of @xmath16 instead of the half - line . the function @xmath328 , here obtained from dirichlet boundary data via the temporal part of the lax pair , plays the role usually played by the reflection coefficient calculated from initial data via the spatial part of the lax pair .
this means that the `` reflection coefficient '' @xmath328 corresponds to some initial data given on the whole line @xmath16 , a fact that has been made quite rigorous in @xcite . in this case , according to theorem [ theorem : initial - condition - small ] , the initial data is very small for @xmath18 ; however to reproduce the nontrivial boundary data described by theorem [ theorem : boundary - condition - recover ] the initial data must not be small for @xmath71 .
thus , the formula for @xmath328 makes a connection in the transform domain between ( i ) a problem on the half - line with zero initial data and nontrivial boundary data and ( ii ) a problem on the whole line with initial data supported on the negative half - line .
the latter initial data is then defined implicitly in terms of the boundary data for the former problem via the solution of riemann - hilbert problem [ rhp - m - tilde ] .
9999 [ remark : half - line - support ]
in light of assumption [ assumption : data ] , the definition of @xmath126 easily implies the following . under assumption [ assumption : data ] , the function @xmath390 is real analytic and it extends by continuity to a function of class @xmath391 satisfying @xmath392 .
[ lemma : tau - continuous - positive ] we will also require detailed information about the behavior of the function @xmath127 near the point @xmath393 . in this direction
we have the following .
if the functions @xmath191 and @xmath229 satisfy assumption [ assumption : data ] , then @xmath394 defined by extends by continuity to a function analytic in @xmath395 , and of class @xmath396 in a neighborhood of @xmath393 .
also , @xmath397 for @xmath248 with equality only for @xmath393 .
[ lemma - phi - interior ] using analyticity of @xmath191 and @xmath229 , which implies that of @xmath214 and @xmath212 , we may write @xmath398 in terms of a contour integral . indeed , @xmath399 where the fractional powers denote the principal branches . the integrand has a branch cut connecting @xmath17 with @xmath249 ( due to the factor @xmath400 when @xmath401 and due to the factor @xmath402 when @xmath403 ) .
the contour @xmath404 is a positively - oriented loop ; it begins at @xmath17 on the lower edge of the branch cut , encloses the cut once passing through the real axis at a point @xmath405 , and terminates at @xmath17 on the upper edge of the cut . in the neighborhood of a fixed value of @xmath15 the contour @xmath404 may be taken to be independent of @xmath15 , and it follows easily that @xmath398 is analytic in @xmath15 separately in the intervals @xmath406 and @xmath407 . for @xmath300
, all derivatives of @xmath127 may be calculated by differentiation under the integral sign .
thus , @xmath408 @xmath409 @xmath410 and @xmath411 we consider @xmath15 to be a real number close to , but not equal to , @xmath239 . assumption [ assumption : data ] ensures that @xmath412 and @xmath413 for small @xmath47 .
this implies that when @xmath414 is small , @xmath415 is proportional to @xmath416 .
based on this observation , we scale the contour @xmath404 as @xmath417 , where @xmath418 is a suitable contour that we will hold fixed as @xmath280 , and we make the substitution @xmath419 in the above integrals . in each case , the integrand considered as a function of @xmath420 has uniform asymptotic behavior on the contour @xmath418 in the limit @xmath280 , being determined from the local behavior of the functions @xmath191 and @xmath229 near @xmath17 as specified in assumption [ assumption : data ] .
thus , uniformly for @xmath421 one has @xmath422 and @xmath423 in the limit @xmath280 , and it follows that @xmath424,\quad k\to k_0,\ ] ] @xmath425,\quad k\to k_0,\ ] ] @xmath426,\quad k\to k_0,\ ] ] @xmath427,\quad k\to k_0,\ ] ] and @xmath428,\quad k\to k_0 .
\label{eq : phi - four}\ ] ] the contour @xmath418 begins and ends at @xmath429 on opposite sides of the branch cut ( all fractional powers of @xmath430 are understood as principal branches ) and encircles the branch point @xmath431 once in the counterclockwise sense .
it is now obvious that @xmath398 tends to @xmath432 while @xmath433 and @xmath434 both vanish as @xmath280 and hence all three extend by continuity to @xmath393 .
it is also obvious that @xmath435 has a finite limit as @xmath280 ; by computing an integral we find the limiting value @xmath436 this completes the proof that @xmath398 is of class @xmath396 near @xmath393 .
we note that the fourth derivative @xmath437 appears to be singular at @xmath393 , but in reality the issue is subtle because the explicit leading term in the square brackets in is an integral that vanishes identically , and therefore the asymptotic behavior of @xmath437 in the limit @xmath280 can not be determined without making further hypotheses on @xmath191 and @xmath229 sufficient to provide leading - order asymptotics for the @xmath438 error term in .
it remains to determine the sign of @xmath433 for @xmath300 .
we go back to the real integral representation for @xmath398 , which admits differentiation by leibniz rule because either @xmath439 or @xmath440 with the result : @xmath441 which can also be written in the form @xmath442 if @xmath403 , then we use the form and factor the quadratic in the numerator as @xmath443 with @xmath444 since @xmath445 and @xmath446 , obviously @xmath447 .
also , since @xmath448 we have @xmath449 , which is nonnegative for @xmath296 because @xmath450 .
hence @xmath451 for @xmath403 . on the other hand ,
for @xmath401 we instead use the form , because in this range of @xmath15 we have @xmath452 and @xmath453 for @xmath296 so combining this with the inequality @xmath454 shows that also for @xmath401 we have @xmath451 .
[ remark : phi - smooth ] although it may seem counterintuitive , assuming that @xmath229 is smoother at @xmath17 , say vanishing linearly rather than like @xmath455 as @xmath456 , leads to less smoothness of @xmath398 at @xmath393 .
linear vanishing of @xmath229 implies continuity of @xmath127 and @xmath457 at @xmath239 , but @xmath458 will have a jump discontinuity .
the point is that it should be the inverse function @xmath415 that is smooth at @xmath393 , not the functions @xmath459 and @xmath460 at @xmath17 .
9999 the part of assumption [ assumption : data ] concerning the nondegeneracy of the extrema of @xmath214 and @xmath212 allows us to obtain the following result .
if the functions @xmath191 and @xmath229 satisfy assumption [ assumption : data ] , then @xmath461 is analytic at @xmath462 and @xmath463 , with @xmath464 and @xmath465 while @xmath466 and @xmath467 .
[ lemma - tau ] using analyticity of @xmath191 and @xmath229 for @xmath47 near @xmath136 and @xmath137 , we can express @xmath468 as a contour integral : @xmath469 where @xmath404 is a closed contour enclosing the interval @xmath470 $ ] once in the positive sense and where @xmath471 is the function analytic in @xmath472 $ ] , where @xmath418 is a domain containing @xmath404 , that satisfies @xmath473 and that the boundary value @xmath474 taken on the upper edge of the branch cut @xmath470 $ ] is negative .
to analyze @xmath126 for @xmath15 near @xmath144 and @xmath145 , we may in each case choose the contour @xmath404 to be fixed , and then since the integrand is analytic in @xmath15 at each point @xmath475 it follows that @xmath126 extends from a function defined for real @xmath15 in a right ( left ) neighborhood of @xmath462 ( @xmath463 ) to an analytic function of @xmath15 at @xmath144 ( @xmath145 ) . since @xmath404 is fixed and since @xmath476 by differentiation of @xmath477 , @xmath478 now near @xmath136 , the function @xmath479 has the taylor expansion @xmath480 , and since @xmath471 is positive imaginary to the right of @xmath481 and negative imaginary to the left of @xmath415 , it follows that when @xmath462 so that the branch cut collapses to a point @xmath215 , we have @xmath482 similar arguments show that @xmath483 in both cases the indicated square roots are positive numbers . in particular , since @xmath484 and @xmath485 are analytic functions of @xmath47 within @xmath404 it follows from that @xmath486 .
we may now use in to compute @xmath487 and @xmath488 by residues : @xmath489 and @xmath490 this completes the proof .
the nondegeneracy of the extrema of @xmath214 and @xmath212 also leads to the following result .
if the functions @xmath191 and @xmath229 satisfy assumption [ assumption : data ] , then @xmath332 has an analytic continuation into the complex plane from a right neighborhood of @xmath144 and from a left neighborhood of @xmath145 and @xmath491 while @xmath492 [ lemma - phi - endpoints ] this can be shown with the help of the contour integral formula . in particular , note that @xmath127 is continuous in the limits @xmath275 and @xmath276 , but formulae and hold in full neighborhoods of the indicated limit point with only local branch cuts of the logarithms omitted .
note that @xmath493 and @xmath494 so both @xmath495 and @xmath496 are positive .
combining lemma [ lemma - phi - endpoints ] with lemma [ lemma - tau ] yields the following result .
the function @xmath497 has an analytic continuation into the complex plane from right and left neighborhoods of @xmath144 and @xmath145 , respectively , and for each @xmath498 and each @xmath204 sufficiently small , @xmath499 and @xmath500 both hold in the limit @xmath501 with @xmath123 , uniformly for @xmath502 .
( of course by schwarz reflection @xmath503 satisfies similar estimates along segments in the lower half - plane with endpoints @xmath144 and @xmath145 . )
[ lemma - tilde - gamma - endpoints ] this follows from the fact that @xmath504 has an analytic continuation satisfying @xmath505 near @xmath144 and @xmath145 as long as @xmath506 .
by lemma [ lemma - tau ] this estimate holds locally near @xmath144 or @xmath145 as long as @xmath507 or @xmath508 holds , respectively , and in each case we may replace @xmath468 in the estimate by either @xmath509 or @xmath510 . using lemma
[ lemma - phi - endpoints ] shows that the problem boils down to estimating functions of a real variable , @xmath511 , having the form @xmath512 this function vanishes as @xmath513 for each @xmath123 , and it has two critical points for @xmath18 , only the smaller of the two being relevant for bounded @xmath46 .
this critical point is the global maximizer on @xmath514 and it satisfies @xmath515 as @xmath151 .
it then follows that @xmath516 as @xmath151 by direct calculation .
the intuition behind this result is that while the exponential decay of @xmath517 in the upper half - plane is not uniform near @xmath144 or @xmath145 ( and in fact there is no decay at all exactly at these two points ) , the factor @xmath504 vanishes at these points , with the result being that the product is locally uniformly small with @xmath123 , albeit exhibiting a very slow rate of decay to zero .
we now give the proof of theorem [ theorem : initial - condition - small ] .
the strategy is to open a single lens about the entire interval @xmath241 based upon the natural factorization of the jump matrix : @xmath518 however , technical modifications of the steepest descent method will be required because @xmath126 has no analytic continuation from the real axis near the points @xmath239 and @xmath240 , and @xmath127 fails to be analytic at @xmath393 .
we will in particular need a way to extend the three - times differentiable but non - analytic function @xmath398 into the complex plane from a real neighborhood of @xmath393 .
let @xmath519 and @xmath520 denote the real and imaginary parts of the complex variable @xmath15 .
we follow the approach of @xcite and first define a nonanalytic extension of @xmath521 by the formula @xmath522 note that @xmath523 is nearly analytic close to the real axis @xmath524 in the sense that @xmath525 according to lemma [ lemma - phi - interior ] . also according to lemma [ lemma - phi - interior ] , we may identify @xmath398 with two distinct analytic functions , @xmath526 denoting the analytic continuation of @xmath398 from @xmath406 and @xmath527 denoting the analytic continuation of @xmath398 from @xmath407 .
note that for each fixed @xmath528 we have @xmath529 and for each fixed @xmath530 we have @xmath531 , with the error terms being uniform by taylor s theorem for @xmath532 in compact subsets of @xmath241 bounded away from @xmath239 .
now let @xmath204 be so small that @xmath533\subset ( k_\mathfrak{a},k_\mathfrak{b})$ ] , and define the smooth bump function @xmath534 $ ] such that @xmath535 is of class @xmath536 and @xmath537
then we define an extension of @xmath127 into the upper half - plane near @xmath241 as follows : @xmath538 by direct calculation , @xmath539 it follows that @xmath540 is @xmath541 uniformly for @xmath532 in compact subsets of @xmath241 .
note that @xmath542 holds , as the nonanalytic analogue of the schwarz reflection symmetry @xmath543 of the real analytic functions @xmath544 .
based on the extension @xmath545 we now make an explicit transformation of @xmath125 to open lenses about @xmath241 .
consider the domains illustrated in figure [ fig : vacuum - lenses ] .
we make the following explicit transformations , defining a new matrix unknown @xmath546 : @xmath547 @xmath548 @xmath549 @xmath550 @xmath551 @xmath552 and in the unbounded domain @xmath553 we set @xmath554 . here
the notation @xmath555 refers to the two distinct analytic functions @xmath556 defined near @xmath462 and @xmath557 defined near @xmath463 ( see lemma [ lemma : tau - continuous - positive ] and lemma [ lemma - tau ] ) .
indeed , the domains @xmath558 and @xmath559 are chosen small enough to exclude both @xmath239 and @xmath240 , the two points of nonanalyticity of @xmath461 . by making @xmath204 smaller if necessary , we also ensure that these domains have no intersection with the vertical strip @xmath560 , in which @xmath546 fails to be analytic .
the matrix @xmath546 has jump discontinuities across a contour @xmath38 illustrated in figure [ fig : vacuum - contour ] .
note that the real segment common to the boundary of the domains @xmath558 and the real segment common to the boundary of the domains @xmath559 are not part of the jump contour @xmath38 as it is easy to check that @xmath546 is continuous across these segments .
is the union of eleven oriented arcs , labeled as shown .
the domain in which @xmath546 fails to be analytic is shaded . ]
we claim that the matrix @xmath546 satisfies the conditions of a hybrid riemann - hilbert-@xmath135 problem of small - norm type .
this problem is the following .
find a @xmath49 matrix @xmath546 with the following properties : * * continuity : * @xmath561 is continuous in each connected component of @xmath562 and takes continuous boundary values @xmath563 and @xmath564 on each oriented arc of @xmath38 from the left and right , respectively . * * jump condition : * on each oriented arc of @xmath38 the boundary values are related by @xmath565 ( see below for the explicit definition of @xmath566 ) . * * deviation from analyticity : * in each connected component of @xmath562 the matrix @xmath561 satisfies @xmath567 ( see below for the explicit definition of @xmath568 ) .
* * normalization : * @xmath569 as @xmath570 in @xmath571 .
[ rhpdbar - small - norm ] the jump matrix @xmath566 is defined explicitly on each arc of @xmath38 simply by using the definitions and the jump condition satisfied by @xmath336 across the segment @xmath241 according to riemann - hilbert problem [ rhp - m - tilde ] .
the result is the following : @xmath572 @xmath573 @xmath574 @xmath575 @xmath576 @xmath577 and , finally , using the fact that @xmath578 , @xmath579 consider the jump matrix defined by . note that since @xmath580 near @xmath581 according to lemma [ lemma - phi - endpoints ] , it follows from lemma [ lemma - tilde - gamma - endpoints ] that for all @xmath73 , @xmath582 holds uniformly on the four contour arcs @xmath583 , provided the lens opens with an acute nonzero angle and the vertical contours @xmath584 are placed close enough to the respective endpoint @xmath585 .
similarly , since @xmath392 is bounded away from zero while @xmath398 and @xmath586 are real for @xmath587 , it is easy to see from that @xmath588 is uniformly exponentially small on @xmath589 in the limit @xmath151 , again independently of @xmath73 . controlling the jump matrix @xmath590 on the remaining arcs of @xmath38 requires conditions on @xmath73 as we will see below
the matrix @xmath568 is defined explicitly by applying the operator @xmath135 to the formulae .
the result is : @xmath591 and @xmath592 and in all other connected components of @xmath562 , @xmath593 . in particular , @xmath594 has compact support .
riemann - hilbert-@xmath135 problem [ rhpdbar - small - norm ] is of small - norm type in the sense that , as a consequence of the conditions @xmath17 and @xmath18 , the jump matrix @xmath590 defined on the compact contour @xmath38 satisfies @xmath595 as @xmath501 , and at the same time the matrix @xmath594 defined on @xmath562 satisfies @xmath596 as @xmath501 .
indeed , @xmath597 according to lemma [ lemma - phi - interior ] .
this immediately implies , by the cauchy - riemann equations applied to the real analytic functions @xmath598 near the real @xmath15-axis , that the exponential factors appearing in the formulae are bounded in modulus by @xmath317 provided the lens is sufficiently thin ( independent of @xmath59 ) . since the factors @xmath599 are exponentially small as they are on the real axis if the lens is thin enough , we conclude that for @xmath18 and @xmath17 , @xmath588 is uniformly exponentially small for @xmath600 . finally , from and the cauchy - riemann equations , we see that @xmath601 a fact which , taken together with for @xmath18 and @xmath17 shows that @xmath588 is uniformly exponentially small in the limit @xmath151 for @xmath602 . combining these estimates yields the claimed @xmath603 bound for @xmath588 .
similarly , for @xmath17 and for each @xmath18 we obtain exponential decay of the exponential factors in , so combining this fact with the fact that @xmath604 holds for @xmath605 , we obtain ( @xmath606 denotes any matrix norm ) @xmath607 where @xmath608 for @xmath609 and @xmath610 is independent of @xmath46 .
the claimed @xmath611 estimate of @xmath594 follows immediately because @xmath612 is uniformly bounded for all @xmath613 ( recall @xmath614 for @xmath615 ) .
one makes use of the estimates @xmath595 and @xmath596 as follows .
the strategy is to solve the hybrid riemann - hilbert-@xmath135 problem by first solving the `` @xmath135 part '' and then using the result to obtain a standard riemann - hilbert problem of @xmath616 small - norm type .
we therefore consider the following auxiliary @xmath135 problem : find a @xmath49 matrix @xmath617 with the following properties : * * continuity : * @xmath618 is continuous . * * deviation from analyticity : * @xmath619 holds in the distributional sense . *
* normalization : * @xmath620 as @xmath570 in @xmath571 .
[ dbar-1 ] we solve for @xmath621 by setting up an integral equation involving the solid cauchy transform : @xmath622 where @xmath623 denotes the area element .
it can be shown that the integral equation is in fact equivalent to the formulation of @xmath135 problem [ dbar-1 ] .
the operator norm of @xmath624 acting on @xmath625 is easily estimated based on the fact that cauchy kernel is locally integrable in two dimensions .
thus , @xmath626 the latter supremum is finite and depends only on the bounded domain @xmath627 as the double integral is continuous and decays as @xmath628 as @xmath570 in @xmath571 . from the bound @xmath629
it follows also that @xmath630 as @xmath151 .
it follows that for @xmath123 sufficiently small , the operator @xmath631 is invertible by neumann series convergent in @xmath625 .
since each term of the series is continuous , so is the sum @xmath617 of the series for @xmath632 applied to @xmath356 .
furthermore , we obtain the important estimate @xmath633 , which in particular implies that @xmath634 exists for sufficiently small @xmath59 as a continuous function on @xmath571 that satisfies @xmath635 .
finally , compact support of @xmath594 ensures that @xmath617 has a convergent laurent series in descending powers of @xmath15 for @xmath377 sufficiently large , and in particular we obtain @xmath636 as @xmath58 , where @xmath637 .
we now use the unique solution @xmath617 of @xmath135 problem [ dbar-1 ] as obtained above to convert the hybrid riemann - hilbert-@xmath135 problem [ rhpdbar - small - norm ] into a standard riemann - hilbert problem that we can show is of @xmath616 small norm type .
indeed , consider the matrix function @xmath638 defined in terms of @xmath546 solving riemann - hilbert-@xmath135 problem [ rhpdbar - small - norm ] and @xmath617 solving @xmath135 problem [ dbar-1 ] by @xmath639 by direct calculation , @xmath640\dot{\mathbf{o}}({k_\mathrm{r}},{k_\mathrm{i}})^{-1}- \mathbf{o}({k_\mathrm{r}},{k_\mathrm{i}})\dot{\mathbf{o}}({k_\mathrm{r}},{k_\mathrm{i}})^{-1}\left[{\overline{\partial}}\dot{\mathbf{o}}({k_\mathrm{r}},{k_\mathrm{i}})\right]\dot{\mathbf{o}}({k_\mathrm{r}},{k_\mathrm{i}})^{-1}\\ & = \mathbf{o}({k_\mathrm{r}},{k_\mathrm{i}})\mathbf{w}({k_\mathrm{r}},{k_\mathrm{i}})\dot{\mathbf{o}}({k_\mathrm{r}},{k_\mathrm{i}})^{-1}- \mathbf{o}({k_\mathrm{r}},{k_\mathrm{i}})\mathbf{w}({k_\mathrm{r}},{k_\mathrm{i}})\dot{\mathbf{o}}({k_\mathrm{r}},{k_\mathrm{i}})^{-1}\\ & = \mathbf{0},\quad ( { k_\mathrm{r}},{k_\mathrm{i}})\in\mathbb{r}^2\setminus\sigma , \end{split}\ ] ] and hence @xmath641 is analytic in each connected component of @xmath562 . in light of this result , we will henceforth use the notation @xmath642 with @xmath643 .
it is a direct matter to calculate the jump conditions satisfied by @xmath641 across the arcs of the contour @xmath38 in terms of the jump matrix @xmath590 for @xmath561 and the function @xmath621 restricted to @xmath38 , and to calculate the asymptotic behavior of @xmath641 as @xmath58 .
we deduce that @xmath641 satisfies the following ( pure ) riemann - hilbert problem : find a @xmath49 matrix @xmath644 with the following properties : * * analyticity : * @xmath641 is analytic in each connected component of @xmath645 , and takes continuous boundary values @xmath646
( @xmath647 ) from the left ( right ) at each non - self - intersection point @xmath15 of @xmath38 . *
* jump condition : * on each oriented arc of @xmath38 the boundary values are related by the jump condition @xmath648 , where @xmath649 * * normalization : * @xmath650 as @xmath58 in @xmath651 . [ rhp - e-1 ] since @xmath617 and @xmath634 are uniformly bounded independent of @xmath59 for @xmath123 sufficiently small , it follows immediately from the estimate @xmath595 that also @xmath652 as @xmath151 .
since @xmath38 is compact , this condition implies the unique solvability of riemann - hilbert problem [ rhp - e-1 ] as a small - norm problem in the @xmath616 sense .
see @xcite or ( * ? ? ?
* appendix b ) for details . in particular , @xmath653 as @xmath58 with @xmath654 . finally , we consider the matrix @xmath125 solving riemann - hilbert problem [ rhp - m - tilde ] for large @xmath377 .
we obtain the exact formula @xmath655 from which we compute @xmath656_{12 } = 2i[\mathbf{e}_1+\dot{\mathbf{o}}_1]_{12 } = \mathcal{o}((\log({\epsilon}^{-1}))^{-1/2}),\quad{\epsilon}\downarrow 0.\ ] ] as the error term is uniform for @xmath657 , the proof is complete . to prove corollary [ corollary - vacuum ] , we simply observe that the only dependence on @xmath73 in the proof of theorem [ theorem : initial - condition - small ] given in [ sec : initial - condition - small - proof ] involved the inequality , and it is not hard to see that this inequality holds also for some nonzero @xmath47 .
more generally , @xmath658 and as @xmath659 is a function with maximum value zero at @xmath393 only and tending to @xmath660 as @xmath275 and @xmath276 , given @xmath78 there will be some finite @xmath661 such that @xmath662 holds strictly on @xmath248 for @xmath152 but fails for some @xmath149 if @xmath663 .
the rest of the proof of theorem [ theorem : initial - condition - small ] then goes through unchanged , with the same result , and the proof of corollary [ corollary - vacuum ] is complete .
we conclude this short section by obtaining explicit and simple asymptotic formulae for the boundary curve @xmath157 valid for small and large @xmath78 .
we note firstly that @xmath664 is certainly locally convex ( i ) for @xmath414 small , because @xmath665 ( see ) and according to lemma [ lemma - phi - interior ] , @xmath666 is continuous on @xmath241 , and also ( ii ) for @xmath667 small , where according to lemma [ lemma - phi - endpoints ] we have @xmath668 where @xmath669 is given in . in the former case , the slope of the tangent line of @xmath664 is small , while in the latter case the slope is large .
therefore , for small or large positive @xmath47 we can apply the following steps to obtain @xmath146 : firstly solve the equation @xmath670 for @xmath671 , and then obtain @xmath672 .
when @xmath78 is small , we expect @xmath414 to be small , and since @xmath673 from @xmath670 we obtain @xmath674 since @xmath675 , integrating and substituting from yields @xmath676 where the asymptote to @xmath146 for small @xmath47 is defined by @xmath677 on the other hand , if @xmath78 is large , then we expect @xmath667 will be small .
the equation to be solved for @xmath671 in this case is then @xmath670 , where now @xmath678 therefore @xmath679 using @xmath680 as @xmath275 , we then have @xmath681 where the asymptote to @xmath146 for large @xmath47 is given by @xmath682 a general strategy to the analysis of the solution of riemann - hilbert problem [ rhp - m - tilde ] in the semiclassical limit @xmath151 follows the basic approach outlined in @xcite , which is based on the introduction of a scalar _ complex phase function _ @xmath684 having the following basic properties : * @xmath143 is analytic for @xmath685 $ ] and takes continuous boundary values @xmath686 on @xmath241 from @xmath344 , * @xmath687 as @xmath58 , * @xmath688 ( schwarz symmetry ) , and very importantly , * @xmath143 is independent of @xmath59 ( although it will generally depend on @xmath46 and @xmath47 ) .
one introduces such a function @xmath143 into riemann - hilbert problem [ rhp - m - tilde ] by making the substitution @xmath689 the basic properties of @xmath143 listed above are by no means sufficient to determine @xmath143 ( this is why we do not formulate them as a proper riemann - hilbert problem ) , and the point is that one should use the freedom of choice of @xmath143 to try to bring @xmath690 into a form amenable for asymptotic analysis in the limit @xmath151 .
the transformation implies that @xmath691 is analytic where @xmath336 is , takes boundary values in the same way , satisfies exactly the same normalization condition as @xmath58 as does @xmath336 , and satisfies a modified jump condition : @xmath692 where @xmath693 the strategy of @xcite is to try to choose @xmath143 so that the interval @xmath241 splits into a finite number of subintervals of three distinct types : * * voids : * intervals in which @xmath694 and @xmath695 . *
* bands : * intervals in which @xmath696 and @xmath697 . *
* saturated regions : * intervals in which @xmath698 and @xmath699 . in each void interval @xmath700 , the modified jump matrix admits an `` upper - lower '' factorization because @xmath694 : @xmath701 and the monotonicity condition @xmath695 suggests that the first ( second ) factor has a continuation into the lower ( upper ) half - plane that is exponentially close to the identity matrix . in each band interval @xmath702 , the modified jump matrix is obviously exponentially close to a constant off - diagonal matrix due to the inequalities @xmath696 : @xmath703 where @xmath704 denotes the constant value of @xmath705 in the band @xmath702 .
finally , in each saturated interval @xmath706 , the modified jump matrix admits a `` lower - upper '' factorization because @xmath698 : @xmath707 the inequality @xmath699 then suggests that the first ( second ) factor can be continued into the lower ( upper ) half - plane , becoming an exponentially small perturbation of the identity matrix @xmath356 .
the function @xmath143 may be constructed by temporarily setting aside the inequalities involved with the voids , bands , and saturated regions .
we suppose that there are @xmath708 bands in @xmath241 that we denote by @xmath709 $ ] with @xmath710 . the complementary intervals are either voids or saturated regions .
the boundary values of the function @xmath711 then satisfy * @xmath712 which implies @xmath713 for @xmath15 in voids .
* @xmath714 for @xmath15 in bands .
* @xmath715 which implies @xmath716 for @xmath15 in saturated regions . therefore , we know the value of @xmath717 everywhere in the interval @xmath241 with the exception of the band intervals , where instead we know @xmath718 . denoting by @xmath719 the function analytic for @xmath720 that satisfies @xmath721 and @xmath722 as @xmath58
, we may consider instead of @xmath723 the related function @xmath724 .
this function is analytic for @xmath685 $ ] and since @xmath725 changes sign across the band intervals and is otherwise analytic , @xmath726 satisfies @xmath727$)}\\ \displaystyle\frac{2\theta'(k;x , t)-2\phi'(k)}{r_+(k)},&\quad \text{$k$ in bands}\\ \displaystyle\frac{2{\mathrm{i}}\tau'(k)}{r(k)},&\quad\text{$k$ in saturated regions}. \end{cases } \label{eq : m - difference - boundary - values}\ ] ] note that @xmath723 must decay as @xmath728 as @xmath58 because @xmath687 in this limit ; this implies that @xmath729 for large @xmath15 , and in particular @xmath730 .
therefore @xmath726 is necessarily given in terms of the difference of its boundary values explicitly written in by the plemelj formula , which implies that @xmath731 where here @xmath702 denotes the union of all bands and @xmath706 denotes the union of all saturated regions .
expanding the cauchy kernel @xmath732 in geometric series for large @xmath15 , we see that the condition @xmath733 as @xmath58 is equivalent to the following _ moment conditions _ : @xmath734 subject to these conditions , @xmath723 is integrable at infinity , and @xmath711 may be expressed as a contour integral : @xmath735 where @xmath723 is explicitly given by .
equations are @xmath736 conditions on the @xmath737 unknown endpoints of the bands @xmath738 . in general ,
additional conditions arise in order to get the integration constants right so that instead of just @xmath713 in voids we actually have @xmath712 , and so that instead of just @xmath716 in saturated regions
we actually have @xmath715 .
since @xmath739 , no additional conditions are required if there is only one band , i.e. , @xmath180 , in which case the moment conditions may determine the band endpoints @xmath740 and @xmath741 .
for the purposes of the proofs of theorem [ theorem : boundary - condition - recover ] and corollary [ corollary : plane - wave ] , we only consider this case . because @xmath742 is entire in @xmath15 , the corresponding integral in the expression for @xmath743
can always be expressed in closed form by a residue calculation at @xmath372 : @xmath744 where @xmath745 is a large , positively - oriented circular contour that encloses all of the bands . with the help of the expansion @xmath746
we therefore obtain in the case @xmath180 that the moment conditions take the form @xmath747 where @xmath748 suppose that @xmath2 .
the equations are satisfied for @xmath78 by @xmath749 and @xmath750 as long as @xmath751 is a void ( saturated region ) if @xmath752 ( @xmath753 ) and @xmath754 is a void ( saturated region ) if @xmath755 ( @xmath756 ) .
[ lemma : endpoints - x - zero ] suppose that for some @xmath757 we have @xmath758 and @xmath759 .
let us evaluate @xmath760 in four cases : * * vbv : * in this case we assume that @xmath761 , @xmath762 , and both intervals @xmath763 and @xmath764 are voids . * * vbs : * in this case we assume that @xmath761 and @xmath765 , and that @xmath763 is a void but @xmath764 is a saturated region . * * sbv : * in this case we assume that @xmath766 and @xmath762 , and that @xmath764 is a void but @xmath763 is a saturated region . * * sbs : * in this case we assume that @xmath766 and @xmath765 , and that both intervals @xmath763 and @xmath764 are saturated regions .
-plane in the vbv case . ]
-plane in the vbs case . left : @xmath767 .
right : @xmath768 . ]
-plane in the sbv case . left : @xmath769 .
right : @xmath770 . ]
-plane in the sbs case .
top left : @xmath770 .
top right : @xmath768 .
bottom : @xmath771 . ]
we may substitute into from ( after differentiating @xmath468 by leibniz rule taking into account that the integrand vanishes at both endpoints ) and from .
we may also simplify @xmath772 for @xmath773 as @xmath774 , while in the integral over @xmath706 we have either @xmath775 if @xmath776 or @xmath777 if @xmath778 .
the first remarkable fact is that the result can be written in a uniform way in all four cases ( including all sub - cases related to where @xmath239 falls with respect to @xmath758 and @xmath759 as illustrated in figures [ fig : vbv][fig : sbs ] ) .
namely , we have @xmath779 where @xmath780 denotes the positive area element and where the domains @xmath781 and @xmath782 are the indicated shaded regions in figures [ fig : vbv][fig : sbs ] .
it is now obvious that the original order of integration is easily reversed in all cases , with the outer @xmath420 integral over the interval @xmath783 and the inner integrals over the intervals with endpoints being the two most negative ( for the domain @xmath781 ) and the two least negative ( for the domain @xmath782 ) among the four values @xmath784 , @xmath785 , @xmath786 , and @xmath787 . carrying out this reinterpretation of the formula
, we may further combine the inner @xmath15 integrals with the introduction of the function @xmath788 that is analytic for @xmath15 in the complex plane with branch cuts lying in the two intervals of integration omitted , whose square is @xmath789 , and that satisfies @xmath790 as @xmath58 .
the result is @xmath791 where @xmath745 is a positively - oriented loop that encloses both branch cuts of @xmath792 .
the inner @xmath15 integral may now be computed by residues for each @xmath793 .
the second remarkable fact is that for @xmath794 the inner @xmath15-integral is independent of @xmath420 : @xmath795 and @xmath796 it therefore follows by inspection that in all four cases , the equations @xmath797 and @xmath798 written in the form are satisfied for @xmath2 and @xmath78 by taking @xmath749 and @xmath750 .
the locations of the voids , bands , and saturated regions are indicated for the boundary data from figure [ fig : assumption1 ] in figure [ fig : intervals ] . and
@xmath799 from figure [ fig : assumption1 ] and the corresponding voids ( v ) , bands ( b ) , and saturated regions ( s ) . ]
we therefore find for each @xmath78 a well - defined candidate for @xmath711 that we will denote by @xmath800 , with corresponding functions @xmath801 and @xmath802 defined by , and it remains only to confirm the inequalities that were dropped earlier . indeed
, we will now prove the following .
let @xmath2 and @xmath78 , and let the function @xmath800 be determined from the values @xmath749 and @xmath750 and the configuration of voids and saturated regions as described in lemma [ lemma : endpoints - x - zero ] .
the functions @xmath801 and @xmath802 given by in terms of @xmath143 satisfy the following inequalities : @xmath803 @xmath804 and @xmath805 where @xmath700 denotes the union of the voids , @xmath706 denotes the union of the saturated regions , and the prime denotes differentiation in @xmath15 for fixed @xmath47 .
[ lemma : inequalities - x - zero ] the proof of these three statements involves the same object , namely the partial derivative of @xmath143 in @xmath47 for fixed @xmath15 , denoted @xmath806 .
we may construct @xmath806 explicitly as follows .
consider differentiation with respect to @xmath47 of the three equations @xmath807 for @xmath808 , @xmath809 for @xmath810 , and @xmath811 for @xmath812 .
since neither @xmath126 nor @xmath127 depend on @xmath47 , we find that @xmath806 is analytic for @xmath813 $ ] , and on the cut the equation @xmath814 holds . keeping in mind the condition @xmath815 as @xmath58 , it therefore follows by similar arguments as led to that @xmath816 where the notation @xmath817 reminds us that the branch points are @xmath749 and @xmath750 , and the integral can be evaluated in closed form by residues at @xmath818 and @xmath819 : @xmath820 , \label{eq : gt}\ ] ] where we used the identity @xmath821 .
now we prove . from and it follows that @xmath822 for a given @xmath823 , we use the fundamental theorem of calculus to write @xmath824 a formula that makes use of the fact that @xmath825 holds for @xmath826 .
but since @xmath249 always corresponds to the boundary between a band and a void , we have @xmath827 .
taking this into account and comparing with completes the proof of , since for @xmath828 we have @xmath829 next we consider together and . from
and we have @xmath830 differentiation with respect to @xmath15 using @xmath831 then gives @xmath832 now we apply the fundamental theorem of calculus to obtain @xmath833 here the turning point @xmath834 is selected so that either @xmath835 or @xmath836 holds for all @xmath420 in the interval of integration . since @xmath15 lies on the boundary of the band @xmath837 , we have @xmath838 , so it remains to determine the sign of the integral .
first observe that @xmath839 and @xmath840 always have opposite signs , regardless of whether @xmath835 or @xmath836 . indeed , if @xmath835 , then @xmath841 , and by the identity @xmath842 one has @xmath843 . on the other hand ,
if @xmath836 , then @xmath844 and by the identity @xmath845 where @xmath846 are defined by one sees that @xmath847 , because @xmath848 which implies @xmath849 , and also @xmath850 which implies @xmath851 . therefore , we conclude that @xmath852 if @xmath853 , which characterizes @xmath15 lying in a void , and @xmath854 if @xmath297 , which characterizes @xmath15 lying in a saturated region .
this completes the proof of the inequalities and .
the expression for @xmath806 from this proof actually leads to a complete characterization of the constant @xmath855 as the following result shows .
let @xmath2 and @xmath78
. then @xmath856 .
[ lemma - phi - b - s ] since @xmath857 as @xmath58 for all @xmath78 ( by ) , it follows that also @xmath815 as @xmath58 .
but if we examine the explicit expression for @xmath806 given by , we observe that there is a constant leading term in the laurent series of @xmath806 for large @xmath377 .
this constant term therefore must vanish , and this gives rise to an identity expressing @xmath858 explicitly in terms of @xmath459 and @xmath460 ( which can be simplified further with the help of and ):
@xmath859 therefore it remains to determine an integration constant .
suppose that @xmath78 is sufficiently small that both @xmath752 and @xmath755 , i.e. , we have a vbv configuration for @xmath800 .
then @xmath860 is analytic for @xmath861 $ ] , and for @xmath773 we have @xmath862 , so using we get @xmath863 it follows from these considerations that @xmath711 must be given by the formula @xmath864\ ] ] where @xmath865 is an additional constant ( @xmath866 as @xmath58 ) .
but @xmath711 is known to be bounded at the band endpoints @xmath867 and @xmath868 , so the expression in square brackets must be made to vanish for these values of @xmath15 , resulting in a system of linear equations for @xmath865 and @xmath704 : @xmath869 the integrals in the coefficient matrix can be calculated by residues at @xmath819 and the system solved for @xmath704 : @xmath870 likewise , the integral involving @xmath871 can be evaluated by a residue at @xmath819 , yielding @xmath872 where we have also used and . finally , we set @xmath2 and consider the limit @xmath456 , in which @xmath749 and @xmath750 converge to @xmath239 .
using the fact , as shown in the proof of lemma [ lemma - phi - interior ] , that @xmath873 as @xmath280 , we replace @xmath874 in the integral by its limiting value and calculate the resulting integral by a residue at @xmath819 . therefore @xmath875 and the proof is complete .
let @xmath2 and @xmath78 .
the function @xmath801 is analytic for @xmath876 and for @xmath877 .
[ lemma : phi - analytic - gaps ] suppose first that the interval containing @xmath15 is a void . then @xmath878 , so @xmath800 is analytic at @xmath15 , and we may write @xmath879 , which is clearly analytic except at @xmath393 , according to lemma [ lemma - phi - interior ] . since @xmath242 , and since @xmath459 is decreasing if @xmath751 is a void while @xmath460 is increasing if @xmath754 is a void , it follows that the point @xmath239 can not lie in a void for any @xmath78 .
next suppose that the interval containing @xmath15 is a saturated region .
then @xmath809 , so we may write @xmath801 in two alternate forms : @xmath880 therefore @xmath801 will be the boundary value of a function analytic in @xmath344 if this is true of the function @xmath881 . from and , we have @xmath882 but the only point of nonanalyticity of @xmath481 in @xmath241 is @xmath883 , so if @xmath301 , then @xmath881 is the boundary value of a function analytic for @xmath15 in the half - plane @xmath344 near @xmath15 .
it follows that if @xmath15 is in a saturated region and @xmath301 then @xmath801 is analytic at @xmath15 , i.e. , it can be continued into both half - planes .
but since @xmath459 and @xmath460 both tend to @xmath240 as @xmath884 and @xmath459 is increasing if @xmath751 is a saturated region while @xmath460 is decreasing if @xmath754 is a saturated region , it is impossible for @xmath240 to lie in a saturated region for any @xmath78 .
we will refer to the analytic function @xmath801 defined in the interval @xmath751 ( respectively , in the interval @xmath754 ) as @xmath885 ( respectively , @xmath886 ) .
finally , we require an analogue of lemma [ lemma - tilde - gamma - endpoints ] .
the functions @xmath887 have analytic continuations into the complex plane from right and left neighborhoods of @xmath144 and @xmath145 , respectively , and along small segments with one endpoint @xmath144 or @xmath145 and the other endpoint having real part in the interior of @xmath241 and nonzero imaginary part of the appropriate sign so that @xmath888 along the segment , the uniform estimate @xmath889 holds .
[ lemma - y - phi - endpoints ] applying the nondegeneracy of the extrema of @xmath214 and @xmath212 guaranteed by assumption [ assumption : data ] to the formula , it is easy to see that @xmath890 always diverges logarithmically as @xmath275 and as @xmath276 . hence , by integration in @xmath15 one sees that up to a nonzero constant factor plus an integration constant , the leading - order behavior of @xmath801 is the same as that of @xmath398 as established in lemma [ lemma - phi - endpoints ] .
the rest of the proof is then exactly the same as that of lemma [ lemma - tilde - gamma - endpoints ] , playing off the exponential decay of @xmath891 into the appropriate half - plane away from @xmath144 or @xmath145 against the linear vanishing of @xmath892 or @xmath893 to establish the claimed uniform estimate . to prove theorem [ theorem : boundary - condition - recover ] , we apply the steepest descent method to riemann - hilbert problem [ rhp - m - tilde ] with the help of the complex phase function @xmath894 introduced in [ sec : g - function ] , that is , we exploit the transformation from @xmath125 to @xmath690 and use to handle the jump condition by opening lenses about the voids and saturated regions .
some minor modifications are required because @xmath504 is not analytic at @xmath393 ( which may lie in a saturated region but not a void ) or @xmath883 ( which may lie in a void but not a saturated region ) , however it will not be necessary to introduce any nonanalyticity or deal with @xmath135 problems as in the proof of theorem [ theorem : initial - condition - small ] . to open the lenses , we define domains of the complex plane as illustrated in figure [ fig : fourconfigurations ] $ ] in the four configurations of the complex phase function @xmath143 .
void intervals can not contain the point @xmath239 , and saturated regions can not contain the point @xmath240 . the various domains labeled with the letter @xmath895
are so small as to exclude both of these points , because only near the endpoints of @xmath323 $ ] do we need to exploit analyticity of @xmath126 in order to obtain decay along the lens boundaries without installing unusual parametrices but rather by using lemma [ lemma - y - phi - endpoints ] . ] and make the following explicit substitution : @xmath896 @xmath897 @xmath898 @xmath899 @xmath900 @xmath901 @xmath902 @xmath903 and in the unbounded domain @xmath553 we set @xmath904 .
the matrix @xmath905 satisfies the conditions of the following riemann - hilbert problem .
find a @xmath49 matrix @xmath905 with the following properties : * * analyticity : * @xmath905 is analytic for @xmath52 , where @xmath38 is the contour illustrated in figure [ fig : fourcontours ] + for riemann - hilbert problem [ rhp : t - positive - open - lenses ] in the four cases for the complex phase function @xmath143 . ] + and @xmath561 takes continuous boundary values on each oriented arc of @xmath38 , @xmath906 from the left and @xmath907 from the right .
* * jump condition : * the boundary values on each oriented arc of @xmath38 are related by @xmath908 ( see below for the definition of @xmath909 ) . * * normalization : * @xmath910 as @xmath58 .
[ rhp : t - positive - open - lenses ] the jump matrix @xmath909 is defined on @xmath38 as follows : @xmath911 @xmath912 @xmath913 @xmath914 @xmath915 @xmath916 @xmath917 @xmath918 and @xmath919 it follows from and in lemma [ lemma : inequalities - x - zero ] , from lemma [ lemma : phi - analytic - gaps ] , and from lemma [ lemma - y - phi - endpoints ] that @xmath920 is uniformly small on @xmath921 ( that is , all non - vertical and non - horizontal arcs of @xmath38 ) omitting only neighborhoods of the band endpoints @xmath749 and @xmath750 .
the rate of decay is determined from neighborhoods of @xmath144 and @xmath145 according to lemma [ lemma - y - phi - endpoints ] , namely @xmath922 , but away from these points one has exponential decay .
likewise , from the fact that @xmath392 is uniformly bounded away from zero on compact subsets of @xmath241 means that @xmath920 is also uniformly exponentially small on @xmath923 , that is , on all vertical arcs of @xmath38 and on all horizontal arcs except the band @xmath702 .
riemann - hilbert problem [ rhp : t - positive - open - lenses ] is _ not _ , however , a small - norm problem in the semiclassical limit @xmath151 , because @xmath920 is not decaying with @xmath59 on the band @xmath924 , nor is the decay on the non - real arcs of @xmath38 that meet at @xmath867 and @xmath868 uniform near these band endpoints .
we will now remedy this situation by constructing an explicit parametrix for @xmath905 in a standard fashion .
first , we exhibit a matrix solving the limiting form of the jump condition on the band @xmath702 : @xmath925 where @xmath926 and @xmath927 denote the principal branches ( and hence the ratio may be considered to be well - defined on the interval @xmath928 common to both branch cuts ) .
it is easy to confirm that this _
outer parametrix _ has the following properties : * @xmath929 is analytic for @xmath861 $ ] , * @xmath929 satisfies the jump condition @xmath930 * @xmath931 , * @xmath932 as @xmath58 , and * for @xmath15 bounded away from @xmath933 , @xmath929 is uniformly bounded independent of @xmath59 .
the outer parametrix will turn out to be a good approximation of @xmath905 away from the points @xmath934 , but it blows up at these two points and fails to even approximately satisfy the non - negligible jump conditions on the complex contours nearby . for
now , we record what will be a useful formula later on : @xmath935 where @xmath936 let @xmath937 and @xmath938 be open disks centered at @xmath867 and @xmath868 respectively , with radius @xmath204 sufficiently small , but independent of @xmath59 .
we shall construct an _ inner parametrix _ in each of these disks , an approximation that will locally be far superior to the outer parametrix .
first consider the disk @xmath937 .
if the interval @xmath939 , @xmath749 , is a void @xmath700 , then we claim that the function @xmath940 defined for @xmath941 near @xmath174 by @xmath942 ( positive @xmath943 power ) can be analytically continued to a full complex neighborhood of @xmath867 as a function that satisfies @xmath944 .
this is a simple consequence of the fact that @xmath945 and that @xmath946 vanishes like a square root and no higher power at @xmath947 .
it follows that @xmath948 defines a conformal mapping from @xmath937 onto a neighborhood of the origin .
the outer parametrix may be represented locally near @xmath867 in terms of @xmath940 as follows : @xmath949 where @xmath950 is a well - defined unimodular matrix function holomorphic near @xmath867 that is obviously uniformly bounded in @xmath937 independent of @xmath59 .
it will be useful later to write this in the equivalent form @xmath951 let an auxiliary matrix function @xmath952 be defined as follows ( @xmath953 ) : @xmath954 @xmath955 and @xmath956 using well - documented formulae involving the airy function @xmath957 and its derivative @xcite , it follows that @xmath952 is analytic in the three sectors of its definition , that across the rays bounding the sectors one has @xmath958 @xmath959 and @xmath960 and that @xmath961 with the asymptotics being uniform with respect to direction in the complex plane , including along the sector boundary rays .
then , in terms of @xmath952 and @xmath950 we define an inner parametrix near @xmath867 by setting @xmath962 assuming without loss of generality that for @xmath963 the contours @xmath964 coincide with radial segments in the @xmath965-plane with angles @xmath966 , one can check that the following facts hold true : * @xmath967 is analytic for @xmath968 ( there is no jump across @xmath969 ) .
* @xmath967 satisfies the jump conditions @xmath970 @xmath971 and @xmath972 * inner and outer parametrices match well on the disk boundary : @xmath973 * @xmath974 and @xmath975 holds uniformly for @xmath963 . to check the jump conditions one should first express the analytic function @xmath885 on the arcs @xmath964 in terms of the conformal
coordinate @xmath940 ( this has already been done for @xmath802 , really by definition ) .
this is accomplished by noting that @xmath976 holds at every point of @xmath241 .
therefore , since @xmath751 is a void @xmath700 , then for @xmath78 , @xmath127 is analytic at @xmath947 and it follows that the function @xmath977 for @xmath876 is the analytic continuation through @xmath344 of the function @xmath978 for @xmath828 .
thus , for @xmath979 we have @xmath980 , and for @xmath981 we have @xmath982 . to confirm the matching between the inner and outer parametrices , one uses the asymptotic condition and the fact that @xmath983 means @xmath984 .
the @xmath985 bound within the disk can be proved similarly .
if instead @xmath939 is a saturated region @xmath706 , then one defines a conformal mapping @xmath986 taking @xmath937 to a neighborhood of the origin by the analytic continuation from @xmath987 of @xmath988 .
one next defines a uniformly bounded unimodular holomorphic matrix function @xmath989 near @xmath867 by writing the outer parametrix in the form @xmath990 finally , one defines @xmath991 assuming that for @xmath963 the contours @xmath992 coincide with @xmath993 , one has the following .
* @xmath994 is analytic for @xmath995 ( there is no jump across @xmath996 ) .
* @xmath994 satisfies the jump conditions @xmath997 @xmath998 and @xmath999 * inner and outer parametrices match well on the disk boundary : @xmath1000 * @xmath1001 and @xmath1002 holds uniformly for @xmath963 .
the proof is virtually the same as before , with the main difference being that the identity implies that if @xmath751 is a saturated region @xmath706 , then the function @xmath1003 for @xmath810 is the analytic continuation through @xmath344 of the function @xmath978 .
now consider the disk @xmath938 .
if @xmath1004 is a void @xmath700 , we define a conformal map @xmath1005 by continuation from @xmath702 of the formula @xmath1006 .
introduce a uniformly bounded , unimodular , and holomorphic matrix @xmath1007 for @xmath1008 by writing @xmath1009 then define an inner parametrix in @xmath938 by the formula @xmath1010 assuming that for @xmath1008 the contours @xmath1011 coincide with segments with angles @xmath1012 , this parametrix satisfies the following conditions .
* @xmath1013 is analytic for @xmath1014 ( there is no jump across @xmath1015 ) .
* @xmath1013 satisfies the jump conditions @xmath1016 @xmath1017 and @xmath1018 * inner and outer parametrices match well on the disk boundary : @xmath1019 * @xmath1020 and @xmath1021 holds uniformly for @xmath1008 . if instead @xmath1004 is a saturated region @xmath706 , then start with the conformal map defined in @xmath938 by the continuation from @xmath702 of the formula @xmath1022 .
introduce the matrix @xmath1023 bounded , holomorphic , and unimodular in @xmath938 , by @xmath1024 then set @xmath1025 assuming that for @xmath1008 the contours @xmath1026 coincide with segments with angles @xmath1027 , this parametrix satisfies these conditions .
* @xmath1028 is analytic for @xmath1029 ( there is no jump across @xmath1030 ) .
* @xmath1028 satisfies the jump conditions @xmath1031 @xmath1032 and @xmath1033 * inner and outer parametrices match well on the disk boundary : @xmath1034 * @xmath1035 and @xmath1036 holds uniformly for @xmath1008 .
now we combine the outer and inner parametrices into a _ global parametrix _ @xmath1037 defined as follows . if @xmath752 and @xmath755 ( case vbv ) , then @xmath1038 if @xmath752 and @xmath756 ( case vbs ) , then @xmath1039 if @xmath753 and @xmath755 ( case sbv ) , then @xmath1040 finally , if @xmath753 and @xmath756 ( case sbs ) , then @xmath1041 the global parametrix is intended to be a good model for @xmath905 in the whole complex plane . to evaluate this claim ,
consider the error @xmath644 defined as @xmath1042 because @xmath905 satisfies the conditions of riemann - hilbert problem [ rhp : t - positive - open - lenses ] while @xmath1037 is known explicitly , @xmath644 solves a riemann - hilbert problem equivalent to that for @xmath905 .
this problem is the following .
find a @xmath49 matrix @xmath644 with the following properties : * * analyticity : * @xmath644 is analytic for @xmath1043 , where @xmath1044 is the contour illustrated in figure [ fig : smallnormcontour ] , + for the error matrix @xmath644 .
all contour arcs are labeled as in figure [ fig : fourcontours ] ( different labels in the four cases ) with the exception of the disk boundaries @xmath1045 and @xmath1046 ( both oriented clockwise ) .
note that the sub - arcs of @xmath964 or @xmath992 within @xmath937 and the sub - arcs of @xmath1011 or @xmath1026 within @xmath938 are absent from @xmath1044 , because the inner parametrices satisfy exactly the jump conditions of @xmath905 on these sub - arcs . ] + and @xmath644 takes continuous boundary values @xmath646 and @xmath647 on each oriented arc of @xmath1044 from the left and right , respectively . *
* jump condition : * the boundary values on each oriented arc of @xmath1044 are related by @xmath1047 ( see below for a complete explicit characterization of the jump matrix @xmath1048 ) . * * normalization : * @xmath650 as @xmath58 .
[ rhp : small - norm ] on all arcs of @xmath1044 with the exception of ( i ) the disk boundaries @xmath1045 and @xmath1046 , ( ii ) the real arcs of @xmath969 or @xmath996 within @xmath937 , ( iii ) the real arcs of @xmath1049 or @xmath1030 within @xmath938 , and ( iv ) the real arc @xmath1050 ( including parts inside and outside the disks ) , the jump matrix @xmath1048 is given by @xmath1051 here @xmath1052 is the jump matrix for riemann - hilbert problem [ rhp : t - positive - open - lenses ] characterizing @xmath905 , and it has already been shown that @xmath1053 is uniformly @xmath922 on these arcs of @xmath1044 .
since @xmath929 is unimodular and bounded independently of @xmath59 away from @xmath867 and @xmath868 , it follows that also @xmath1054 holds uniformly on these arcs .
the discontinuity of @xmath644 across the boundary of the two disks is caused by the mismatch of the inner and outer parametrices , since @xmath905 has no jump .
letting @xmath1055 denote the inner parametrix installed in @xmath937 and @xmath1056 denote the inner parametrix installed in @xmath938 , a calculation shows that @xmath1057 but using or for @xmath983 , and using or for @xmath1008 , we conclude that @xmath1058 holds uniformly on both @xmath1045 and @xmath1046 . if @xmath15 lies on the real segment of @xmath969 or @xmath996 within @xmath937 , then @xmath1059 and similarly @xmath1060 since the inner parametrix has no jump .
in all four cases we have shown that @xmath1053 is exponentially small in the limit @xmath151 , so even though each conjugating factor amplifies this by @xmath1061 , we still have decay beyond all orders in @xmath59 .
if @xmath15 lies on the band @xmath1050 within either disk , then both @xmath905 and the ( inner ) parametrix @xmath1037 are discontinuous .
if the band edge abuts a void , a calculation shows that @xmath1062 the central factor is an exponentially small perturbation of @xmath356 because @xmath1063 is exponentially small and both @xmath468 and @xmath1064 are strictly positive .
the conjugating factors amplify this by @xmath1065 , but this remains beyond all orders small .
similarly , if the band edge abuts a saturated region , then @xmath1066 and again @xmath1067 is small beyond all orders despite the conjugating factors algebraically large size .
finally , consider @xmath1068 outside both disks . again
both @xmath905 and the ( outer ) parametrix @xmath1037 have jump discontinuities across this segment , and a calculation shows that @xmath1069 which is exponentially close to @xmath356 because the outer parametrix and its inverse are uniformly bounded away from @xmath867 and @xmath868 , while @xmath1063 is exponentially small and @xmath802 and @xmath1064 are both strictly positive as @xmath15 lies on a compact subset of the interior of the band @xmath702 .
these considerations prove that @xmath1070 is uniformly @xmath922 on the @xmath59-independent contour @xmath1044 .
therefore , in the @xmath151 limit , riemann - hilbert problem [ rhp : small - norm ] is a small - norm riemann - hilbert problem in the @xmath1071 sense .
this implies that @xmath644 exists for @xmath123 sufficiently small and has a convergent ( because @xmath1044 is bounded ) laurent expansion for sufficiently large @xmath377 of the form @xmath1072 with the first two moments satisfying @xmath1073 at last , we obtain the exact formula for @xmath125 valid for sufficiently large @xmath377 : @xmath1074 taking into account the laurent expansion @xmath1075 as @xmath58 , using with , and using with , it therefore follows from that in the limit @xmath151 , @xmath1076 where we have used and . since @xmath749 and @xmath750 , using and lemma [ lemma - phi - b - s ] we obtain @xmath1077 the error terms are uniform on compact subintervals of @xmath1078 .
indeed , the @xmath1079 estimate on @xmath1070 fails only as @xmath201 or @xmath884 ( in which case the band @xmath924 shrinks to a point ) , as @xmath1080 ( in which case the void or saturated region on the left of the band shrinks to a point ) , and as @xmath1081 ( in which case the void or saturated region on the right of the band shrinks to a point ) .
furthermore , by a slight modification of the preceding arguments in which the disks @xmath937 and @xmath938 are allowed to be slightly off - center from the points @xmath174 and @xmath175 , it can be shown that for @xmath47 in a given compact subinterval of @xmath142 , riemann - hilbert problem [ rhp : small - norm ] may be formulated with a finite number of different contours @xmath1044 .
therefore the norm of the cauchy projection operator @xmath1082 is uniformly bounded ( independent of @xmath59 , of course ) as a maximum over a finite number of values ( see ( * ? ? ?
* proposition 3 of 3.6 , and 4.6 ) for further information and more details about this procedure ) .
this completes the proof of theorem [ theorem : boundary - condition - recover ] .
the main idea of the proof is to observe that the construction of @xmath800 for @xmath2 and @xmath78 presented in [ sec : g - function ] may be subjected to continuation for small @xmath18 .
consider the prospect of continuation of the solution @xmath1083 of the equations for @xmath2 to nonzero @xmath46 by means of the implicit function theorem .
this requires calculating the jacobian of the system .
we first establish the following .
the functions @xmath1084 for @xmath1085 defined by satisfy the identities @xmath1086 [ lemma - moment - derivatives ] firstly , one establishes the related identities @xmath1087 by replacing the integrals in in a standard way by contour integrals over contours locally independent of @xmath174 and @xmath175 that are bounded away from these points , and then differentiating under the integral sign .
then one eliminates @xmath1088 and @xmath1089 in favor of @xmath1090 and @xmath1091 using to finish the proof .
whenever @xmath1092 satisfy the equations , as is the case for @xmath2 and @xmath78 with @xmath1093 by lemma [ lemma : endpoints - x - zero ] , the jacobian is easily calculated with the help of lemma [ lemma - moment - derivatives ] : @xmath1094 when @xmath2 and @xmath78 , we have @xmath1095 .
it can be shown that in this situation also @xmath1096 indeed , it turns out that @xmath1097 if and only if the function @xmath1098 or @xmath1099 vanishes as @xmath1100 to higher order than @xmath1101 depending on whether @xmath939 is a void @xmath1102 or a saturated region @xmath792 .
similarly , @xmath1103 detects higher - order vanishing of either @xmath1098 or @xmath1099 at @xmath868 .
it is easy to see from the explicit construction of @xmath143 given in
[ sec : g - function ] that as long as @xmath138 or @xmath137 , then @xmath1104 and @xmath1105 .
therefore , the jacobian is nonzero , and so the following is true .
suppose @xmath1106 and @xmath138 , @xmath139 , so that the complex phase function @xmath1107 is well - defined in a particular configuration , vbv , vbs , sbv , or sbs .
then there is a neighborhood of the point @xmath162 in the @xmath76-plane in which there is a unique solution @xmath1108 of the equations formulated for the same configuration that satisfies the boundary condition @xmath1109 .
the solution is differentiable and the partial derivatives satisfy the identities .
[ lemma - endpoints - x - small ] it only remains to establish the identities , but this is done simply by implicit differentiation of the equations @xmath1110 with respect to @xmath46 and @xmath47 , using lemma [ lemma - moment - derivatives ] with @xmath1111 .
this result allows us to define a candidate complex phase function @xmath1112 given @xmath174 and @xmath175 as described in
[ sec : g - function ] .
moreover , it is easy to see that the inequalities , guaranteed for @xmath2 in the particular configuration of voids and saturated regions valid for @xmath1113 according to lemma [ lemma : inequalities - x - zero ] , persist for @xmath1114 in the same configuration for small @xmath1115 .
the existence of an appropriate complex phase function @xmath143 was the only essential ingredient in the proof of theorem [ theorem : boundary - condition - recover ] .
therefore , the same result holds , with @xmath1116 and @xmath1117 taking the place of @xmath459 and @xmath460 respectively , and the proof of corollary [ corollary : plane - wave ] is complete .
here we collect together a few additional remarks , some indicating natural generalizations of our results , and some pointing toward directions for future research . the methodology of the complex phase function as described in
[ sec : g - function ] is capable of detecting and describing asymptotic behavior of @xmath130 that neither corresponds to the vacuum domain nor the plane - wave domain .
a typical scenario would be that as one tunes @xmath76 through the plane - wave domain by continuing the complex phase function @xmath143 along some path , at some point @xmath1118 one of the inequalities associated with the complex phase function @xmath1112 fails for some @xmath149 ( a less generic scenario involves the jacobian of the system vanishing ) .
further continuation becomes impossible without resorting to the device of increasing @xmath1119 , where we recall that @xmath708 is the number of bands in @xmath241 .
thus one is witnessing the birth of a new band , void , or saturated region from the point @xmath15 as one continues beyond @xmath1118 .
once @xmath1120 , instead of an asymptotic formula like , one arrives at an asymptotic formula written in terms of hyperelliptic functions of genus @xmath1119 , a formula that exhibits rapid variations of the amplitude on space and time scales proportional to @xmath59 .
thus , if @xmath1120 , @xmath130 no longer resembles a slowly - modulated plane wave , but rather has a more complicated microstructure .
the point @xmath1118 therefore lies along the common boundary between domains corresponding to @xmath180 and ( in the next simplest case ) @xmath1121 . in general
, one expects the quarter plane to be tiled with @xmath59-independent domains corresponding to various values of @xmath1122 , with the only general statement being that the positive @xmath47-axis abuts the ( plane - wave ) domain with @xmath180 and the positive @xmath46-axis abuts the vacuum domain ( which is not associated with any value of @xmath1119 ) .
these phase transitions may in principal be computed , with the details depending on the particular boundary data given at @xmath2 and @xmath78 .
we leave such considerations for future studies .
note , however , that under the assumptions in force in this paper , it is impossible for a dispersive shock wave region with @xmath1120 that is generated from a gradient catastrophe in the solution of the dispersionless system at some time @xmath1123 to reach the boundary @xmath2 in finite time , since such behavior would be inconsistent with theorem [ theorem : boundary - condition - recover ] .
[ remark - multiphase ] 9999 [ remark - imaginary - amplify ] we wish to briefly explain the key obstruction in our opinion that arises in the semiclassical analysis of riemann - hilbert problem [ rhp - original ] posed in [ sec : introduction ] relative to a cross contour @xmath1124 , and hence our actual motivation in introducing the modified approach described in
[ sec : formulation ] .
a calculation starting from in the vbv configuration shows that @xmath1125 as @xmath456 ( for @xmath2 ) , and hence from we obtain @xmath1126 ( this is consistent with our approach for @xmath18 and @xmath17 , which basically assumed @xmath1127 ) . therefore for @xmath2 and @xmath78 we have @xmath1128 using and assuming that @xmath15 is purely imaginary gives @xmath1129 note that @xmath1130 which together with implies that @xmath1131 holds whenever @xmath15 is positive imaginary and of sufficiently large magnitude , because @xmath1132 for @xmath78 by assumption [ assumption : data ]
. consider now the effect of introducing @xmath143 not into riemann - hilbert problem [ rhp - m - tilde ] for @xmath125 having ( a subset of ) @xmath343 as the jump contour , but rather into riemann - hilbert problem [ rhp - original ] for @xmath50 having @xmath1124 as the jump contour . in the latter problem
one has a lower - triangular jump matrix on the positive imaginary axis given by .
thus , if one introduces the complex phase function @xmath800 by means of a formula analogous to , the resulting jump condition for @xmath690 on the positive imaginary axis ( oriented toward the origin ) would read @xmath1133 the factor @xmath1134 is purely oscillatory for imaginary @xmath15 , and by proposition [ prop : noturningpoints ] we have an upper bound for @xmath1135 that decays to zero with @xmath59 .
however , we now see that the factor @xmath1136 is exponentially large in the limit @xmath151 for sufficiently large @xmath377 with @xmath1137 .
therefore , the same complex phase function @xmath143 that works so well to control the modified problem for @xmath125 produces uncontrollable errors if used to study @xmath50 itself .
this difficulty originates in the jump discontinuities along the imaginary axis , which appear to be negligible before @xmath143 is introduced but that seem impossible to either neglect afterwards or include in a parametrix without explicit beyond - all - orders information about @xmath291 .
9999 as corollary [ corollary : plane - wave ] is formulated , the points @xmath215 and @xmath213 appear to present an obstruction to continuation of the plane - wave approximation to positive @xmath46 from the boundary . while the details are not easy to explain , the fact is that the complex phase function @xmath143 can indeed be continued away from the boundary near such points , which become curves @xmath1138 and @xmath1139 for @xmath18 along which one of the band endpoints is fixed to an endpoint of the interval @xmath323 $ ] ( see , for example , @xcite ) .
there still remains , however , a technical issue in that it appears that a new type of inner parametrix is required near the corresponding point in the @xmath15-plane , and to our knowledge this has not been worked out .
although we are therefore prevented from proving convergence to a plane wave for @xmath76 near the curves @xmath1138 and @xmath1139 , one can check that the plane wave formulae that are indeed valid as given in corollary [ corollary : plane - wave ] on either side of these curves actually match on the curves themselves .
thus one does not expect any new leading - order asymptotic behavior for @xmath130 along these curves , which nonetheless present an obstruction to rigorous analysis .
9999 the reader will observe that the very slow rate of decay of the error terms in our results ( proportional to @xmath1140 ) is in all cases due to the truncation introduced in
[ sec : formulation ] of the jump matrix on @xmath343 to one whose difference from @xmath356 is compactly supported in the interval @xmath323 $ ] .
indeed this truncation leads to dominant contributions to the error generated from small neighborhoods of the endpoints of this support interval .
this suggests that it may be possible to further modify the formula for @xmath328 near these two points with the aim of reducing the magnitude of these local contributions to the error without having any other significant effect .
9999 [ remark : nonzero - ic ] in our view , the incorporation of nonzero initial conditions together with a non - linearizable boundary condition presents substantial new difficulties .
physically , one expects a nontrivial interaction between the initial and boundary data , and from the `` hyperbolic '' point of view of the dispersionless nonlinear schrdinger system one may expect competing influences at a given point @xmath76 from boundary and initial data propagating along characteristics . mathematically , the initial and boundary conditions get mixed up in the complicated jump matrix on the negative real axis as indicated in . the fact that at points @xmath312 where both @xmath386 and @xmath65 are nonzero the elements of the jump matrix are differences of oscillatory exponentials in the semiclassical approximation of @xmath1141 and @xmath128 makes analysis by steepest descent methods quite challenging , and we hope to consider such issues in the future .
9999 we may consider several ways to generalize assumption [ assumption : data ] while still maintaining the form of the dirichlet boundary data .
* one could drop the condition @xmath119 in favor of the weaker condition @xmath116 . in this situation
the boundary @xmath2 is still a spacelike curve for the hyperbolic dispersionless defocusing nonlinear schrdinger system , so one has a local solution of this approximate system near the boundary for all @xmath78 . on the other hand
, it seems that it is not possible to recover the boundary data from the riemann - hilbert problem in this case as pointed out in remark [ remark : kb - negative ] .
is this a merely technical issue , or does this imply that a dispersive shock wave generated for positive @xmath46 reaches the boundary at some later time @xmath47 , ruining the local plane - wave microstructure ?
* one could go further and drop the condition @xmath116 .
if some bicharacteristics of the hyperbolic system point outside of the domain at the boundary @xmath2 , is it still possible for the approximate dirichlet - to - neumann map given in definition [ definition : d - to - n ] to be valid ? perhaps should one expect a kind of boundary layer to form near @xmath2 ? * finally , one could even drop the inequality . in this case
the phase gradient @xmath92 is apparently imaginary at the boundary .
what does this mean ?
it would also be very interesting to consider the nonlinear schrdinger equation in the focusing case , in which case the analogue of the dispersionless system is a quasilinear system of elliptic type , so all intuition involving propagation along characteristics is lost .
p. d. miller was partially supported by the national science foundation under grant number dms-1206131 and by a fellowship from the simons foundation , grant number 267106 .
part of this work was done during two visits miller made to fudan university in shanghai , china , and both authors are grateful for the support fudan university offered .
the authors collaboration began while z. qin visited the university of michigan for an extended period under the support of the china scholarship council .
z. qin was also supported by a grant from the key laboratory of mathematics for nonlinear science , fudan university .
we begin with a key lemma .
[ lem : eigenvaluebranch ] suppose that assumption [ assumption : data ] holds and that either @xmath29 and @xmath1142 , or that @xmath325 , or that @xmath1143
. then there is an eigenvalue @xmath269 of the matrix @xmath270 defined by that depends smoothly on @xmath78 and satisfies @xmath1144 , @xmath1145 .
the fact that @xmath269 can be chosen to be a smooth function of @xmath47 follows from lemma [ lem : turningpoints ] since there are no turning points and @xmath1146 is a smooth function of @xmath78 .
if @xmath15 is real and @xmath325 or @xmath326 , then @xmath1146 is real and strictly negative for all @xmath78 , so both eigenvalue functions @xmath1147 will be purely imaginary for @xmath78 . on the other hand ,
if @xmath29 but @xmath1142 , then the lemma can be proven by showing that @xmath1146 is _ not _ negative real for any @xmath78 .
let @xmath1148 and @xmath1149 , and suppose that @xmath1146 is real for some @xmath1150 , some @xmath1151 , and some @xmath78 . using , we therefore have @xmath1152(h(t)^2-k_\mathrm{r}[2k_\mathrm{r}+u(t)])\right]=0 .
\label{eq : im - lambda - squared}\ ] ] if @xmath1153 , then as also @xmath1151 , implies that @xmath1154 , which is a contradiction with @xmath78 by assumption [ assumption : data ] . on the other hand ,
if @xmath1155 , then we may solve for @xmath1156 and hence from we obtain @xmath1157 ^ 2 - 2k_\mathrm{i}^2\right)\left(2h(t)^2-\tfrac{1}{2}[2k_\mathrm{r}+u(t)]^2 + 2k_\mathrm{i}^2\right)+4k_\mathrm{i}^2[2k_\mathrm{r}-u(t)][2k_\mathrm{r}+u(t)]\\ & = \frac{1}{4k_\mathrm{r}^2}[4k_\mathrm{r}^2-h(t)^2]^2[4k_\mathrm{r}^2-u(t)^2 ] .
\end{split}\ ] ] but , at the same time implies the inequality @xmath1158(h(t)^2-k_\mathrm{r}[2k_\mathrm{r}+u(t ) ] ) > 0 , \label{eq : inequality - one}\ ] ] which can be written in the equivalent form ( because @xmath1155 and @xmath118 by assumption [ assumption : data ] ) @xmath1159 obviously , implies that @xmath1160 . hence @xmath295 is not possible for any @xmath78 .
we denote by @xmath1161 the eigenvalue branch characterized by lemma [ lem : eigenvaluebranch ] , so that @xmath1162 for all @xmath78 . if @xmath1163 , then @xmath1164 is ambiguous up to a sign and we choose the sign so that @xmath1165 is continuous .
the matrix @xmath1166 given by @xmath1167 is an eigenvector matrix for the coefficient matrix @xmath270 defined by , i.e. , the identity @xmath1168 holds . since @xmath1169 for @xmath284 and @xmath78 by lemma [ lem : turningpoints ] ,
the eigenvalues are distinct and therefore @xmath1166 is invertible as long as its columns are both nonzero . but
@xmath1170 as @xmath1142 .
it follows from assumption [ assumption : data ] that the off - diagonal entries of @xmath1166 are nonzero for all @xmath78 and therefore @xmath1166 has nonzero columns for all @xmath284 and for all @xmath78 , so for such @xmath15 and @xmath47 we have @xmath1171 .
moreover , @xmath1172 is bounded away from zero on @xmath1173 .
indeed , one can show that @xmath1174 as @xmath58 uniformly for @xmath265 and @xmath78 , and it follows that @xmath1175 also holds as @xmath58 with the same uniform nature of the error term .
on the other hand if @xmath1176 for some ( large ) @xmath1177 , then @xmath15 lies in a compact subset of @xmath1178 , so since @xmath1172 is a continuous nonzero function of its arguments it follows that it is uniformly bounded away from zero for all such @xmath15 and @xmath47 lying in any compact subset of @xmath211 .
it remains to analyze @xmath1172 in the limits @xmath456 and @xmath1179 , in which @xmath1180 by assumption [ assumption : data ] . since @xmath1181 as @xmath456 and @xmath1182 as @xmath1179
, we have @xmath1183 as @xmath456 and @xmath1184 as @xmath1179 .
both limiting values are nonzero for @xmath284 , and the convergence is uniform for @xmath1176 , so the argument is complete .
let @xmath1185 be the continuous function defined by @xmath1186 and the asymptotic condition ( which selects an unambiguous branch of the square root ) that @xmath1187 as @xmath195 . by a homotopy argument taking the function @xmath189 to zero by scaling
it is easy to see that @xmath1188 .
then @xmath1189 is a uniformly bounded nonvanishing function for @xmath78 and @xmath283 that satisfies @xmath1190 as @xmath58 in @xmath1178 uniformly for @xmath78 .
let an eigenvector matrix @xmath1191 be defined for @xmath78 and @xmath284 by @xmath1192 of course @xmath1193 . also , it is easy to see that @xmath1194 let the first column @xmath1195 of the matrix @xmath261 satisfying and be written in the form @xmath1196 for some new unknown vector function @xmath1197 .
it is clear from and that @xmath1198 and from and that @xmath321 and @xmath322 can be expressed in terms of @xmath1197 by @xmath1199 moreover , a direct calculation using shows that @xmath1197 satisfies the differential equation @xmath1200 where @xmath1201.\ ] ] assumption [ assumption : data ] implies that @xmath1202 and @xmath1203 are absolutely integrable on @xmath1204 , and this in turn implies that @xmath1205 for each @xmath284 .
moreover there is a constant @xmath1206 depending only on the functions @xmath229 and @xmath191 satisfying assumption [ assumption : data ] such that @xmath1207 making the rescaling @xmath1208 where @xmath1209\,ds\ ] ] results in the system @xmath1210 and the boundary condition @xmath1211 note that @xmath1212 is well - defined due to the asymptotic behavior of @xmath1213 for large positive @xmath47 and that the condition is a consequence of .
also , from we have that @xmath1214 now we solve and .
writing @xmath1215 and @xmath1216 and introducing an appropriate integrating factor , takes the form @xmath1217= -{\mathrm{i}}e^{2\varphi(0,t;k)/{\epsilon}}\rho(t;k)y(t;k)-{\mathrm{i}}e^{2\varphi(0,t;k)/{\epsilon}}\rho(t;k ) , \label{eq : hsystem - rewrite}\ ] ] where @xmath1218 is defined as @xmath1219 note that since @xmath1220 by lemma [ lem : eigenvaluebranch ] , the factor @xmath1221 remains bounded as @xmath195 , so we may build in the boundary conditions on @xmath1222 by integrating equations from @xmath47 to @xmath1223 : @xmath1224 this system of integral equations is equivalent to the differential equations and boundary conditions .
to solve , substitute the first equation into the second and exchange the order of integration to get a closed equation for @xmath1225 : @xmath1226 where the kernel is given by @xmath1227 again , since @xmath1220 , we have the @xmath59-independent estimate @xmath1228 we first seek @xmath1229 in the banach space @xmath1230 of continuous functions for @xmath35 decaying to zero as @xmath195 , equipped with the supremum norm . note that @xmath1231 because @xmath1205 . defining a sequence of iterates by setting @xmath1232 and
then @xmath1233 one easily sees that @xmath1234 takes the form of a sum of multiple integrals : @xmath1235 where @xmath1236 and for @xmath1237 , @xmath1238 obviously , for all @xmath1239 we have @xmath1240^m\\ & \le\frac{\|\rho(\cdot;k)\|_1^{2m}}{m!}\|f^{\epsilon}\|_\infty , \end{split}\ ] ] where @xmath1241 denotes the norm in @xmath1230 : @xmath1242 it follows by comparison with the exponential series that the infinite series @xmath1243 converges uniformly on @xmath1204 and hence as each partial sum vanishes as @xmath195 it converges in @xmath1230 .
moreover , @xmath1244 it then follows from the first equation of that @xmath1245 also lies in @xmath1230 , and that @xmath1246 once it is known that both @xmath1247 and @xmath1248 lie in @xmath1230 , it follows directly from the integral equations that both functions are continuously differentiable , so @xmath1249 and @xmath1250 constitute the unique classical solution of the differential equations subject to the boundary conditions .
it remains to estimate @xmath1251 .
the part of assumption [ assumption : data ] that describes the local behavior of @xmath191 and @xmath229 near @xmath17 implies that @xmath1252 is bounded by @xmath1253 near @xmath17 .
this integrable singularity determines the rate of decay of @xmath1254 for small @xmath47 . while integration by parts shows that @xmath1255 for @xmath47 bounded away from zero , letting @xmath456 results in a slower uniform rate of decay , namely @xmath1256 .
therefore , @xmath1257 under the conditions on @xmath191 and @xmath229 in assumption [ assumption : data ] . taking into account that @xmath1258 scales as @xmath1259 for large @xmath15 ,
while @xmath1260 scales as @xmath628 for large @xmath15 , a more precise statement is that for some constant @xmath1261 , the inequality @xmath1262 holds for @xmath290 and all @xmath123 sufficiently small . combining , , , and
, we see that ( since the exponential factors are bounded ) , latexmath:[\[|y(0;k)|=\mathcal{o}\left(\frac{{\epsilon}^{1/2}}{1+|k|^4}\right)\quad\text{and}\quad both hold for @xmath290 and @xmath123 .
then , from @xmath1264 and @xmath1265 and we have @xmath1266\quad\text{and}\quad b_0(k^*)^*=-e^{-{\mathrm{i}}s(0)/{\epsilon}}e^{{\mathrm{i}}m(k)/(2{\epsilon})}\mathcal{o}\left(\frac{{\epsilon}^{1/2}}{1+|k|^3}\right).\ ] ] in particular , it follows that @xmath286 has no zeros for @xmath1267 if @xmath123 is sufficiently small , and that for @xmath123 fixed any zeros of @xmath287 in @xmath264 must lie in a bounded subset ( depending only on the functions @xmath229 and @xmath191 satisfying assumption [ assumption : data ] ) . also , it is clear that holds true .
this completes the proof of proposition [ prop : noturningpoints ] .
returning to subject to the boundary condition , we first consider making a gauge transformation ; that is we consider a given invertible matrix @xmath1268 and transform to a new unknown matrix @xmath1269 by the substitution @xmath1270 which transforms into the form @xmath1271\mathbf{y}(t;k).\ ] ] the main idea of the method is to try to choose @xmath1268 so that the leading term on the right - hand side takes a particularly simple form that can be the basis for a perturbation expansion .
unfortunately , conjugation by @xmath1272 does not allow the determinant to be changed ; however an additional scalar factor can be introduced by making a change of the independent variable .
so let @xmath1273 be a smooth strictly monotone transformation of the independent variable ( hence invertible with smooth inverse map @xmath1274 ) , and let @xmath1275 .
the desired scalar factor then comes from the chain rule : @xmath1276 the target form for the leading matrix coefficient on the right - hand side is ( for a problem with a single turning point in the @xmath47-interval of interest ) the coefficient matrix of the first - order form of the airy equation ; therefore one tries to set @xmath1277 taking determinants of both sides of leads to a differential equation for @xmath1273 : @xmath1278 for @xmath248 and assuming that @xmath300 and @xmath301 , the right - hand side has simple roots at the turning points @xmath1279 .
we consider building two different transformations , denoted @xmath1280 , mapping neighborhoods of the turning points @xmath1279 to corresponding intervals of @xmath1248 .
if @xmath1281 is to be a smooth invertible transformation , it is necessary that @xmath1282 , which fixes the integration constant in each case .
the solution to is obtained ( by separating the variables ) as follows : @xmath1283 this formula of the independent variable , along with the linear gauge transformation and its higher - order correction , constitute the langer transformation of the system in a neighborhood of the turning point @xmath1279 . ]
defines @xmath1284 in the interval @xmath1285 , and it defines @xmath1286 in the interval @xmath1287 .
moreover , @xmath1286 is monotone increasing , while @xmath1284 is monotone decreasing on their respective domains .
it is obvious that within their intervals of definition , @xmath1288 have exactly one more continuous derivative than do @xmath246 and @xmath247 , as long as @xmath47 is bounded away from the corresponding turning point @xmath299 .
however , a local analysis of @xmath1288 for @xmath47 near @xmath1279 shows that @xmath1288 has only the same number of continuous derivatives as do @xmath246 and @xmath247 at the turning point .
given that @xmath229 and @xmath191 are analytic according to assumption [ assumption : data ] , we conclude that @xmath1288 are real analytic monotone functions in their respective domains of definition .
note that even though appears to allow @xmath1284 to be defined also for @xmath1289 and for @xmath1286 to be defined also for @xmath1290 , there will be an essential loss of smoothness of @xmath1288 at the `` other '' turning point @xmath1291 , so in fact our approach will be to cover the positive @xmath47-axis @xmath78 with the two overlapping intervals @xmath1292 and @xmath1293 and hence use two different changes of independent variable to obtain the desired smoothness .
( moreover , it is easy to check that @xmath1286 as defined by fails to satisfy for @xmath297 and that @xmath1284 as defined by fails to satisfy for @xmath853 . )
finally , we record here that the image of @xmath1293 under @xmath1294 is the interval @xmath1295 where @xmath1296 while the image of @xmath1292 under @xmath1297 is the interval @xmath1298 where @xmath1299 .
also , @xmath1300 as @xmath195 . with @xmath1288 determined by
, we can solve for the gauge transformation matrix @xmath1268 .
using the fact that @xmath1301 , we obtain the general solution in the form @xmath1302 where @xmath1303 and @xmath1304 are ( at this moment ) arbitrary functions of @xmath47 .
whether @xmath1305 and @xmath1306 or whether @xmath1307 and @xmath1308 , due to and , equation takes the form @xmath1309 where @xmath1310 of course we think of @xmath1311 as a function of @xmath1248 by means of the invertible transformation defined by .
it turns out that the error term @xmath1312 in the coefficient matrix is too large to be controllable directly when @xmath1313 due to the corresponding factor of @xmath59 on the left - hand side of .
however , the terms proportional to @xmath59 in the coefficient matrix can be removed by an explicit near - identity transformation .
for some @xmath59-independent matrix @xmath1314 to be determined , consider the effect of making the substitution @xmath1315 in equation . in order to obtain an equivalent differential equation for @xmath1316 , it will be necessary to invert @xmath1317 , and while this can always be accomplished pointwise for sufficiently small @xmath59 by neumann series , it is especially convenient if the inversion can be carried out explicitly .
hence we assume at this point that @xmath1318 is a nilpotent matrix of the general form @xmath1319 where @xmath1222 is a general vector function of @xmath47 and @xmath15 .
therefore , @xmath1320 , and it follows that the neumann series for the inverse truncates : @xmath1321 . under and , equation becomes @xmath1322 where @xmath1323+\mathbf{p}(t;k),\\ \mathbf{c}_2(t;k)&:= [ \mathbf{p}(t , k),\mathbf{h}(t;k)]-\mathbf{h}(t;k)\mathbf{a}(z(t;k))\mathbf{h}(t;k)- \frac{1}{z'(t;k)}\frac{d\mathbf{h}}{dt}(t;k),\\ \mathbf{c}_3(t;k)&:= \frac{1}{z'(t;k)}\mathbf{h}(t;k)\frac{d\mathbf{h}}{dt}(t;k)-\mathbf{h}(t;k)\mathbf{p}(t;k)\mathbf{h}(t;k ) .
\end{split}\ ] ] here @xmath1324:=\mathbf{a}\mathbf{b}-\mathbf{b}\mathbf{a}$ ] denotes the matrix commutator .
the question now arises as to how the arbitrary functions @xmath1325 and @xmath1326 for @xmath1085 can be chosen to ensure that @xmath1327 .
in fact , the condition @xmath1327 is equivalent to the two conditions on the matrix @xmath1311 ( really , on the two functions @xmath1325 for @xmath1085 ) : @xmath1328 and the two equations relating @xmath1326 , @xmath1085 to the functions @xmath1325 , @xmath1085 in the matrix @xmath1311 : @xmath1329 it turns out to be consistent to assume , in addition to that @xmath1330 in which case implies that also @xmath1331 then , conditions require that either @xmath1332 or @xmath1333 .
taking@xmath1334 conditions reduce to @xmath1335 it follows that the matrix @xmath1318 can be expressed in terms of @xmath1311 as @xmath1336 while @xmath1311 has the form @xmath1337 assuming that @xmath1338 is differentiable with respect to @xmath47 , it follows from and that @xmath1339 where @xmath1340 and that @xmath1341 it follows that under the gauge transformation , the differential equation becomes a perturbed airy equation : @xmath1342 let us consider finding @xmath1325 , @xmath1085 , so that @xmath1343 , @xmath1344 , and also @xmath1345 .
these calculations are based on the formula and the representation .
assuming for the moment that @xmath1346 , it is easy to see that the condition @xmath1345 implies that in fact @xmath1347 must be independent of @xmath47 ; as the conditions on the elements of @xmath1311 are all linear in @xmath1325 , @xmath1085 , we will assume that they are chosen so that @xmath1348 .
it is then easy to see from and that the condition @xmath1343 implies that the ratio @xmath1349 is independent of @xmath47 ; we therefore write @xmath1350 where @xmath1351 is independent of @xmath47 . with the further assumption that @xmath1352 ,
one checks that because @xmath454 for all @xmath78 when @xmath149 , the condition @xmath1344 implies that @xmath1353 this equation is obviously solved by choosing either @xmath1354 or @xmath1355 . finally , we return to the condition @xmath1348 .
this condition yields the identity @xmath1356^{-1},\quad c=\pm{\mathrm{i}}.\ ] ] recalling the definitions of @xmath459 and @xmath460 , this can be written in the form @xmath1357^{-1},&\quad
c={\mathrm{i}},\\ \left[2z'(t;k)(u(t)-2k)(k-\mathfrak{a}(t))\right]^{-1},&\quad c=-{\mathrm{i}}. \end{cases } \label{eq : p1-squared}\ ] ] in each case , @xmath1287 and @xmath1285 , we choose the value of @xmath1358 so that the turning point contained in the corresponding open interval is _ not _ a root of the factor @xmath1359 or @xmath1360 in the denominator of @xmath1361 . for example , if @xmath415 satisfies @xmath1362 , then @xmath1360 will be nonzero for @xmath1305 , so we choose @xmath1355 . a calculation using the fact that @xmath1348 then shows that @xmath1363 ( it is not necessary to resolve the sign of @xmath1303 to determine @xmath1338 uniquely . ) the corresponding function @xmath1364
is then determined from and .
let @xmath248 with @xmath300 and @xmath301 .
then there exists a constant @xmath1365 such that for @xmath1366 , @xmath1367 , while for @xmath1368 , @xmath1364 satisfies an estimate of the form @xmath1369 [ lemma : q - bound ] since @xmath1370 and @xmath1371 are nonzero on @xmath1292 and @xmath1293 respectively , it follows that @xmath1338 is in each case an analytic function on the corresponding interval .
it then follows from that the same is true of @xmath1372 .
since @xmath1281 is in each case a real analytic bijection , @xmath1364 is smooth on either @xmath1298 or @xmath1295 .
therefore it only remains to prove that @xmath1373 as @xmath1374 in the case when @xmath1308 .
but from and assumption [ assumption : data ] it follows that @xmath1375 and @xmath1376 as @xmath195 , so the desired decay follows from by composition with @xmath1300 as @xmath195 .
we now use the estimate of @xmath1377 recorded in lemma [ lemma : q - bound ] to solve the perturbed airy equation separately in the intervals @xmath1366 and @xmath1368 .
the idea in both cases is the same : consider the perturbation proportional to @xmath96 on the right - hand side as a forcing term , and convert the differential equation into an integral equation with the help of the fundamental matrix @xmath1378 of the unforced problem ( we refer to @xcite for the definitions and properties of the airy functions @xmath1379 and @xmath1380 ) . indeed , setting @xmath1381 in where @xmath1382 denotes a new matrix unknown , we obtain the equivalent equation ( after canceling a factor of @xmath59 ) @xmath1383 using the wronskian identity @xmath1384 to invert @xmath1385 , we may write this system in the form @xmath1386 assuming @xmath1382 is known at some value @xmath1387 , we integrate to obtain the volterra equation @xmath1388 this equation may be analyzed with the help of a weight function @xmath895 defined by @xmath1389 the weight @xmath895 is strictly positive , continuous , and nonincreasing on @xmath343 . also , there is a constant @xmath1390 such that the three inequalities @xmath1391 we now introduce the weight @xmath895 into in two different ways .
first , let @xmath1392 define a new unknown @xmath1393 . in terms of @xmath1393 , the volterra equation becomes the equivalent equation @xmath1394 where the matrix kernel @xmath1395 is given by @xmath1396 and where the forcing term is the matrix function @xmath1397 we will use this form when @xmath1398 in which case @xmath1399 holds for the first row of the kernel . alternately , let @xmath1400 define a new unknown @xmath1401 . in terms of @xmath1401 , the volterra equation becomes @xmath1402 where the matrix kernel @xmath1403 is given by @xmath1404 and the matrix - valued forcing term is @xmath1405 we will use this form when @xmath1406 in which case @xmath1407 holds for the second row of the kernel . let @xmath1408 be a fixed number in the open interval @xmath298
we first use along with the mapping @xmath1409 to analyze @xmath261 for @xmath1410 .
then , we use along with the mapping @xmath1411 to analyze @xmath261 for @xmath1412 $ ] . under the real - analytic bijection
@xmath1409 given by , the interval @xmath1413 is identified with the interval @xmath1414 where @xmath1415 , and the mapping reverses orientation , i.e. , @xmath1416 .
we begin by resolving the only remaining indeterminacy in the langer transformation by first noting that in the formula we must take @xmath1417 ( because @xmath1418 for @xmath1419 while @xmath1420 for @xmath1421 ) . taking into account
that @xmath1416 , we then choose the sign of the square root to obtain @xmath1303 as @xmath1422 in both cases , the positive square root is meant , and @xmath1303 is a purely imaginary nonvanishing analytic function of @xmath47 in the interval @xmath1413 . the definition unambiguously determines the matrices @xmath1423 and @xmath1424 .
then , combining the transformations , , and with , we obtain the exact relation linking @xmath261 and @xmath1393 for @xmath1308 : @xmath1425 since @xmath1426 for @xmath297 ( which implies @xmath1427 ) , from the asymptotic normalization condition on @xmath261 , we expect the following limit to exist for each @xmath123 : @xmath1428 this limit does indeed exist , and it may be computed with the help of ( i ) the estimate @xmath1429 , ( ii ) the asymptotic relations @xmath1430 @xmath1431 where @xmath1432 is defined by and @xmath1433 and ( iii ) known asymptotic formulae for airy functions and their derivatives for large negative @xmath1248 . the result is : @xmath1434 where @xmath1435 is defined by .
we now consider the volterra integral equation with @xmath1436 , on the space of matrix - valued functions defined on @xmath1437 equipped with the supremum norm ( and based on some matrix norm ) .
it is a simple consequence of lemma [ lemma : q - bound ] , the inequality @xmath1399 holding for @xmath1438 and the estimates that the volterra integral operator @xmath1439 with matrix kernel @xmath1395 is bounded on this space with operator norm @xmath327 , because @xmath1440 it follows easily that @xmath1441 is invertible for sufficiently small @xmath59 , and that the operator norm of @xmath1442 is @xmath327 . since @xmath1443 for @xmath1436 , it is easy to see that @xmath1444 defined by is a bounded continuous function on @xmath1445
. therefore we learn that ( @xmath606 is a matrix norm ) @xmath1446 because @xmath1447 is bounded independently of @xmath59 according to and the fact that @xmath1448 .
it follows that we may solve for @xmath261 , set @xmath1113 and take into account that @xmath1415 to obtain an asymptotic formula for @xmath1449 ; here we need to use the fact that @xmath1450 along with asymptotic expansions of airy functions for large positive arguments .
the key point is that @xmath1451 combining this with , , and gives @xmath1452 where @xmath1453 now , we consider the interval @xmath1454 , which is mapped by the real - analytic bijection @xmath1455 to the interval @xmath1456 with orientation preserved , i.e. , @xmath1457 . to fully determine the langer transformation in this interval
, we need to take @xmath1458 ( because @xmath1459 for @xmath401 while @xmath1362 for @xmath403 ) . choosing a sign for the square root of @xmath1361
we then obtain @xmath1460 therefore , @xmath1303 is a strictly positive nonvanishing analytic function of @xmath47 for @xmath1454 .
this choice determines the matrices @xmath1268 and @xmath1318 , here denoted @xmath1461 and @xmath1462 .
then , according to , , , and , the exact relation between @xmath261 and @xmath1401 for @xmath1306 is @xmath1463 if we consider in particular @xmath1113 , then @xmath1449 is given by , so using the fact that @xmath1464 to simplify the weight @xmath895 and using asymptotic formula for airy functions of large positive arguments , we obtain the following : @xmath1465 where @xmath1466 we claim that the explicit terms in the formula are independent of @xmath1467 .
indeed , we have the following results .
suppose that @xmath149 with @xmath300 and @xmath301 , and that @xmath1468 .
then : @xmath1469 where @xmath461 is defined by , and @xmath1470 in particular , @xmath1471 is independent of @xmath1408 and piecewise constant in @xmath149 .
[ lemma - t0-independent ] the proof of follows directly from .
to prove one first computes the product @xmath1472 from using @xmath1473 and along with @xmath1474 with @xmath1417 and then one computes @xmath1475 from using @xmath1476 and along with @xmath1474 with @xmath1458 .
the resulting formulae may be simplified using the positive one - fourth root of .
the result then follows by a direct calculation .
henceforth , we will write @xmath1477 for @xmath1478 in light of this calculation . we now may consider the volterra integral equation with @xmath1479 and @xmath1406 .
fixing a matrix norm @xmath606 and defining the corresponding supremum norm over the interval @xmath1480 , it follows from lemma [ lemma : q - bound ] , the inequality @xmath1481 holding for @xmath1482 and the estimates that the volterra integral operator @xmath1483 with matrix kernel @xmath1403 is bounded with operator norm @xmath327 because @xmath1484 hence @xmath1485 is invertible for sufficiently small @xmath59 , and the norm of @xmath1486 is @xmath327 .
therefore , @xmath1487 because @xmath1488 is bounded independently of @xmath59 according to and taking into account the fact that @xmath1489 . finally ,
we set @xmath17 and solve for @xmath1490 taking into account that @xmath1491 because @xmath1492 . using asymptotic expansions of airy functions for large negative arguments , and using the fact that the second row of @xmath1493 is exponentially small compared with the first row due to the factor @xmath1494 as @xmath1464 , we obtain @xmath1495 also with the help of the identity @xmath1496 this completes the rigorous asymptotic calculation of @xmath1497 by the method of langer transformations . from we then obtain @xmath1498 for small @xmath59 , and then from and we complete the proof of proposition [ prop : two - turning - points ] .
99 r. carroll and c. bu , `` solution of the forced nls equation using pde techniques , '' _ appl .
_ , * 41 * , 3351 , 1991 .
l. faddeev and l. a. takhtajan , _ hamiltonian methods in the theory of solitons _ , classics in mathematics .
springer , berlin , 2007 .
a. s. fokas , _ a unified approach to boundary value problems _ , cbms - nsf regional conference series in applied mathematics , society for industrial and applied mathematics , philadelphia , 2008 .
d. a. smith , ` http://unifiedmethod.azurewebsites.net/ ` .
a. s. fokas , a. r. its , and l. y. sung , `` the nonlinear schrdinger equation on the half - line , '' _ nonlinearity _ , * 18 * , 17711822 , 2005
. j. lenells , `` absence of solitons for the defocusing nls equation on the half - line , '' ` arxiv:1412.3495 ` , 2014 .
x. zhou , `` the riemann - hilbert problem and inverse scattering , '' _
siam j. math .
_ , * 20 * , 966986 , 1989 . k. t .-
mclaughlin and p. d. miller , `` the @xmath1499 steepest descent method for orthogonal polynomials on the real line with varying weights , '' _ intern .
res . notices _ , * 2008 * , article i d rnn075 , 166 , 2008 . a. boutet de monvel , a. r. its , and v. kotlyarov , `` long - time asymptotics for the focusing nls equation with time - periodic boundary condition , '' _ c. r. math .
paris _ , * 345 * , 615620 , 2007 .
a. s. fokas and j. lenells , `` the unified method : i. nonlinearizable problems on the half - line , '' _ j. phys .
a _ , * 45 * , 195201 ( 38 pp . ) , 2012 .
a. s. fokas and j. lenells , `` the unified method : ii .
nls on the half - line with @xmath47-periodic boundary conditions , ''
_ j. phys .
a _ , * 45 * , 195202 ( 36 pp . )
s. kamvissis , `` semiclassical nonlinear schrdinger on the half line , '' _ j. math
_ , * 44 * , 58495868 , 2003 .
a. degasperis , s. v. manakov , and p. m. santini , `` initial - boundary value problems for linear and soliton pdes , '' _ theor . and math .
_ , * 133 * , 14751489 , 2002 . n. muskhelishvili , _ singular integral equations _
, second edition , dover , 1992 .
x. zhou , `` @xmath616-sobolev space bijectivity of the scattering and inverse scattering transforms , '' _ comm .
pure appl .
_ , * 51 * 697731 , 1998
. p. a. deift , _ orthogonal polynomials and random matrices : a riemann - hilbert approach _ ,
courant lecture notes 3 , american mathematical society , providence , ri , 2002 .
r. j. buckingham and p. d. miller , `` large - degree asymptotics of rational painlev - ii functions .
noncritical behavior , '' _ nonlinearity _ , * 27 * , 24892577 , 2014 .
p. deift , s. venakides , and x. zhou , `` new results in small dispersion kdv by an extension of the steepest descent method for riemann - hilbert problems , '' _ internat .
notices _ , * 1997 * , 285299 , 1997 .
nist digital library of mathematical functions . ` http://dlmf.nist.gov/ ` , release 1.0.8 of 2014 - 04 - 25 .
online companion to @xcite .
r. j. buckingham and p. d. miller , `` the sine - gordon equation in the semiclassical limit : dynamics of fluxon condensates , '' _ memoirs amer .
_ , * 225 * , 1136 , 2013 . f. w. j. olver , d. w. lozier , r. f. boisvert , and c. w. clark , editors .
_ nist handbook of mathematical functions _ , cambridge university press , new york , ny , 2010
. print companion to @xcite .
|
initial - boundary value problems for integrable nonlinear partial differential equations have become tractable in recent years due to the development of so - called unified transform techniques .
the main obstruction to applying these methods in practice is that calculation of the spectral transforms of the initial and boundary data requires knowledge of too many boundary conditions , more than are required make the problem well - posed .
the elimination of the unknown boundary values is frequently addressed in the spectral domain via the so - called global relation , and types of boundary conditions for which the global relation can be solved are called _ linearizable_. for the defocusing nonlinear schrdinger equation , the global relation is only known to be explicitly solvable in rather restrictive situations , namely homogeneous boundary conditions of dirichlet , neumann , and robin ( mixed ) type .
general nonhomogeneous boundary conditions are not known to be linearizable . in this paper
, we propose an explicit approximation for the nonlinear dirichlet - to - neumann map supplied by the defocusing nonlinear schrdinger equation and use it to provide approximate solutions of general nonhomogeneous boundary value problems for this equation posed as an initial - boundary value problem on the half - line .
our method sidesteps entirely the solution of the global relation .
the accuracy of our method is proven in the semiclassical limit , and we provide explicit asymptotics for the solution in the interior of the quarter - plane space - time domain .
|
since their discovery ( heger 1922 ) , the diffuse interstellar bands ( dibs ) have been a persistent spectroscopic enigma representing unidentified ingredients in the physics and chemistry of the interstellar medium .
the dibs are ubiquitous in the ism with the strongest clearly visible toward even lightly reddened stars though their relative strengths can vary in different sight lines ( e.g. , kreowski & walker 1987 ) .
recent deep surveys have catalogued at least 540 dibs , to equivalent width limits of several m , in two sight lines with @xmath14(@xmath15 ) @xmath7 1.1 ( hobbs et al .
2008 , 2009 ) .
while the dibs are likely molecular in origin , attempts to identify specific carrier molecules by comparing interstellar spectra with laboratory spectra of various candidates have been unsuccessful ( e.g. , snow & mccall 2006 ; salama et al .
. some characteristics of the carriers may be inferred , however , from the observed absorption - line profiles .
high - resolution , high - s / n spectra have revealed substructure in the profiles of a number of the dibs ( sarre et al . 1995 ; kreowski & schmidt 1997 ; galazutdinov et al . 2008 ) . that substructure was initially attributed to either rotational structure ( ehrenfreund & foing 1996 ) or isotopic substitution ( webster 1996 ) in large , carbon - based carrier molecules in the gas phase which could contain a non - negligible fraction of the available carbon .
variations in the separations and relative strengths of the subcomponents presumably reflecting differences in local environmental conditions may be seen for a number of the dibs ( cami et al .
2004 ; galazutdinov et al .
2008 ) . while the variations in subcomponent separation seem to rule out isotopic substitution , modelling of the substructures as rotational contours
can provide constraints on the size , shape , and rotational constants of the carrier molecules ( ehrenfreund & foing 1996 ; schulz et al .
2000 ; cami et al . 2004 ) .
if the dib carriers can be identified ( and their behavior understood ) , they will provide potentially significant , widely available diagnostics of the physical conditions in the ism .
several potentially important clues to the behavior ( and , ultimately , the identities ) of the dib carriers have recently been found in optical spectra of the heavily reddened star herschel 36 ( her 36 ) .
dahlstrom et al .
( 2013 ; hereafter paper i ) discovered both absorption from excited levels of ch@xmath3 and ch and very strong extended tail toward red ( etr ) in the profiles of a number of the dibs toward her 36 .
absorption from excited interstellar ch@xmath3 and ch has not been seen in any other sight line . while the etr can be seen for some dibs in a few other sight lines , they are generally much weaker than those seen toward her 36 . in this paper , we determine that the molecular excitation must be due to strong far - infrared pumping from the adjacent ir source her 36 se , and we propose that ir pumping is also responsible for the strong etrs on the dibs . if that is the case , then some constraints may be placed on the properties of the molecular carriers of the dibs . in sections 2 and 3 ,
we discuss the excitation of ch@xmath3 and ch , respectively .
in section 4 , we compare the observed dib profiles with simulated profiles for linear molecules subject to both collisional and radiative excitation . in sections 5 and 6
, we discuss what can be inferred about the properties and identities of the molecules responsible for the dibs .
the spectrum of interstellar ch@xmath3 was discovered by theodore dunham , jr .
( 1937 ) as an unidentified absorption line at 4232.6 toward @xmath16 persei , @xmath17 orionis , 55 cygni , and @xmath18 aurigae .
douglas & herzberg ( 1941 ) identified the spectrum as due to the _ _
__@xmath0 @xmath1 _ _ x__@xmath2 ( 0,0 ) transition of ch@xmath3 by laboratory emission spectroscopy .
this was the first identification of a molecular ion in interstellar space . since then interstellar
ch@xmath3 has been observed through absorption at rest wavelengths ( in air ) 4232.5 ( 0 , 0 ) , 3957.7 ( 1 , 0 ) , and 3745.3 ( 2 , 0 ) toward a great many stars , demonstrating the ubiquity of ch@xmath3 in the diffuse interstellar medium ( e.g. gredel et al . 1993 ; crawford 1995 ; gredel 1997 ) .
nearly all of the more than 250 detections of ch@xmath3 absorption lines so far have been of the @xmath19(0 ) transitions of ch@xmath3 in the ground @xmath4 = 0 rotational level .
the sole exception is the circumstellar absorption toward hd 213985 reported by bakker et al .
( 1997 ) . because of the very fast @xmath4 = 1 @xmath20 0 spontaneous emission ( see section 2.3 . ) , the excited @xmath4 = 1 level is not sufficiently populated for detection in ordinary diffuse clouds .
highly excited ch@xmath3 lines have been observed in visible emission from the red rectangle ( balm & jura 1992 ; hall et al .
1992 ; bakker et al . 1997 ; hobbs et al . 2004 ) and in far - infrared emission in the planetary nebula ngc 7027 ( cernicharo et al . 1997 ) , the orion bar photon - dominated region ( habart et al.2010 ; naylor et al . 2010 ) and the disc around herbig be star hd 100546 ( thi et al .
model calculations for the emissions observed in ngc 7027 and the orion bar have recently been published ( godard & cernicharo 2013 ; nagy et al .
2013 ) .
_ __@xmath0 @xmath1 _ _ x__@xmath2 ( 0,0 ) and ( 1,0 ) bands of ch@xmath3 from the archive of the fiberfed extended range optical spectrograph ( feros ) with a resolution of 48,000 ( kaufer et al .
1999 ) toward herschel 36 and 9 sgr are compared in fig . 1
. both stars are members of the young cluster ngc 6530 .
the 3 arcmin separation between the two sight lines corresponds to a linear separation of about 1.3 pc at the roughly 1.5 kpc distance to the cluster .
the @xmath19(1 ) and @xmath21(1 ) absorptions of ch@xmath3 in the @xmath4 = 1 excited rotational level , 40.1 k above the ground @xmath4 = 0 level , are clearly seen for both vibronic bands only toward her 36 . the @xmath19(2 ) and @xmath21(2 )
lines for the @xmath4 = 2 level , 120.3 k above the ground level , are not detectable .
the non - detection of the @xmath4 = 1 lines toward 9 sgr suggests that the excited ch@xmath3 is found in a small region near herschel 36 ( paper i ) .
observed transitions , with oscillator strength @xmath22 , observed ( heliocentric ) wavelength in air @xmath23 , measured equivalent width of the absorption @xmath24 , hnl - london factor @xmath25 , derived ch@xmath3 column density in the @xmath4 level @xmath26(@xmath4 ) , and excitation temperature @xmath27 are listed in table 1 . for weak lines ,
the column densities could be derived from @xmath24 by using @xmath26 = ( @xmath28/@xmath29@xmath30)(@xmath31/@xmath32 ) , where @xmath22 = ( 8@xmath33/3@xmath34)@xmath35@xmath36@xmath37 is the oscillator strength and @xmath36 is the transition dipole moment ( e.g. , spitzer 1978 ) . the fits6p profile - fitting program ( welty et al .
2003 ) was used to fit stronger lines ( accounting for any saturation in the absorption ) and to fit multiple lines from a given species .
a @xmath38-value of 1.2 - 1.6 km s@xmath39 yielded consistent @xmath26(0 ) for the ( 0,0 ) , ( 1,0 ) , and ( 2,0 ) bands of ch@xmath3 ; we adopt @xmath26(0 ) = 9.4 @xmath40 10@xmath41 @xmath42 and @xmath26(1 ) = 1.8 @xmath40 10@xmath41 @xmath42 .
the excitation temperature @xmath27 was derived from @xmath26(@xmath4)/@xmath26(0 ) = ( 2@xmath4 + 1)exp(-@xmath43/@xmath44 ) .
uncertainties are 1@xmath45 ; upper limits for undetected lines correspond to 3@xmath45 .
the values of @xmath24 should be identical for @xmath19(1 ) and @xmath21(1 ) ( which have the same hnl - london factors ) , and the values of @xmath26(@xmath4 ) and @xmath27 should be identical for the ( 0,0 ) and ( 1,0 ) bands . the slight differences in those measured and derived values reflect our measurement uncertainties .
we proceed in the following using the excitation temperature of 14.6 k determined from the three ch@xmath3 bands .
the observed high excitation temperature of @xmath27 = 14.6 k is exceptional , in view of the very short life time of ch@xmath3 in the @xmath4 = 1 level due to the fast spontaneous emission @xmath4 = 1 @xmath20 0 .
the energy difference between the @xmath4 = 1 and 0 levels of ch@xmath3 has been known since the work by douglas & herzberg ( 1941 ) but was recently measured directly by amano ( 2010 ) to be 835,137.504 mhz , which is equivalent to 27.86 cm@xmath39 and 40.08 k. the permanent dipole moment of ch@xmath3 has been calculated by _ ab initio _ theory to be @xmath47 = 1.656 debye ( ornellas & machado 1986 ) , 1.679 debye ( follmeg et al .
1987 ) , and 1.804 debye ( sun & freed 1988 ) .
if we use @xmath47 = 1.7 debye , we have the einstein coefficient @xmath48 = ( 64@xmath49/3@xmath50)@xmath51/3 = 0.0065 s@xmath39 , corresponding to the lifetime of the @xmath4 = 1 level of 150 s. the critical density for this decay is estimated to be on the order of @xmath52(h ) @xmath7 @xmath53/@xmath54 @xmath7 7 @xmath40 @xmath55 @xmath56 , where @xmath54 @xmath7 1 @xmath40 @xmath57 @xmath58 s@xmath39 is the langevin rate constant ( anicich & huntress 1986 ) . at such a high density
, carbon will be primarily in the form of co and any ch@xmath3 would be rapidly destroyed by collisions with molecular hydrogen .
the number density in the diffuse interstellar medium where ch@xmath3 abounds is typically on the order of 10@xmath37 @xmath56 or less , which is insufficient to maintain ch@xmath3 in the @xmath4 = 1 level by collisions .
this is the reason why ch@xmath3 in the @xmath4 = 1 level has not previously been observed in ordinary sight lines
. the high excitation temperature of 14.6 k must be due to radiative pumping in the environment , most likely by dust emission from the nearby extended infrared source her 36 se ( goto et al .
the intensity of radiation must depend on the radial position of the sight line and therefore what determines the rotational distribution of ch@xmath3 must be a complicated weighted average of the radiative temperature @xmath59 along the ch@xmath3 column . if the whole column of ch@xmath3 is irradiated by the dust emission , the average radiative temperature @xmath59 due to dust emission would be equal to the excitation temperature , @xmath60 = 14.6 k. @xmath59 needs to be higher if only part of the ch@xmath3 column is irradiated , but we believe this is not the case since ch@xmath3 in the @xmath4 = 2 level is not detected .
@xmath59 can not be much higher than the upper limit on the excitation temperature for @xmath4 = 2 , @xmath60 @xmath61 22.9 k. the logic used in this subsection to determine the radiative temperature @xmath59 is basically the same as that missed in mckellar s near discovery of the 2.73 k primordial blackbody radiation ( mckellar 1941 , 1949 ) but later used by thaddeus and clauser ( 1966 ) and meyer and jura ( 1984 ) to determine the temperature of the blackbody radiation from the @xmath19(0 ) and @xmath19(1 ) lines of cn .
the cases of ch@xmath3 and ch are cleaner than that of cn because their @xmath4 = 1 @xmath20 0 and @xmath4 = 3/2 @xmath20 1/2 spontaneous emissions ( respectively ) are orders of magnitude faster .
the effect of the high radiative temperature is also observed for the _ _
__@xmath5 @xmath1 _ _ x__@xmath6 ( 0,0 ) transition of ch toward her 36 .
the absorption spectrum has previously been observed only for ch in the blended @xmath13 doublet in the ground @xmath4 = 1/2 f@xmath62 fine structure level through the @xmath63(1/2 ) transition . toward the special sightline of her 36 , however , we see a ch line from the excited fine structure level @xmath4 = 3/2 f@xmath64 , 25.6 k above the ground level , through the @xmath65(3/2 ) transition as shown in fig . 2 . to the best of our knowledge
this is the first time that this line has been observed in absorption ( paper i ) .
[ the line has been observed in emission from the red rectangle ( hobbs et al .
2004 ) . ] the observed ch equivalent widths and derived column densities are listed in table 1 .
data for the weaker _ _
b__@xmath66 @xmath1 _ _ x__@xmath6 ( 0,0 ) transitions , which confirm the @xmath4 = 1/2 column density derived from _ _
__@xmath5 @xmath1 _ _ x__@xmath6 ( 0,0 ) , are also listed . with the large permanent dipole moment of @xmath47 = 1.46 debye ( phelps & dalby 1966
; see also kalemos et al .
1999 ) , and large energy separation of 536,795.678 - 532,721.333 mhz ( amano 2000 ) , the spontaneous emission @xmath4 = 3/2 f@xmath64 @xmath20 @xmath4 = 1/2 f@xmath62 is orders of magnitude faster than collisional pumping in a diffuse cloud .
we thus conclude , as for the case of ch@xmath3 , that the excitation temperature is equal to the effective radiative temperature @xmath59 of the environment of the observed ch column .
the small column density ratio @xmath26(3/2)/@xmath26(1/2 ) = 0.043 , determined by using the hnl - london factors of luque & crosley ( 1996 ) , gives @xmath59 = 6.7(1 ) k. this is considerably lower than 14.6 k for ch@xmath3 .
perhaps this is not surprising since it is well known that ch and ch@xmath3 can exist in different parts of a cloud .
although the velocity of ch determined from the unresolved @xmath13 doublet lines is in agreement with that of ch@xmath3 , suggesting that ch and ch@xmath3 are in the same cloud , the lower temperature for ch suggests that only a part ( @xmath7 35% ) of the ch column is contained in the high @xmath59 region .
hereafter we simply write @xmath11 for @xmath59 for brevity .
the dust emission which populates the @xmath4 = 1 excited rotational level of ch@xmath3 and the @xmath4 = 3/2 excited fine structure level of ch affects the velocity profiles of some dibs in a more spectacular manner .
six of the dibs observed toward her 36 are compared with those in the ordinary , adjacent sight line toward 9 sgr in fig .
3 . for all six dibs ,
the velocity profiles toward 9 sgr ( in red ) are more or less symmetric with respect to the peak position .
the three dibs @xmath135780.5 , @xmath135797.1 , and @xmath136613.6 toward herschel 36 ( in black ) shown on the left of fig . 3 , however , are very asymmetric , with a huge tail extending to the red .
we will call this extended tail toward red as etr .
there are some indications of etr for the three dibs on the right of fig . 3 also ,
but we have not been able to explain them quantitatively by the mechanism we consider in this paper .
they may be due to vibrational hot bands , as discussed in section 5.3 . in any case , the qualitative difference between the variations of the three dibs on the left and those on the right of fig .
3 is obvious .
the broad deep absorption observed for 9 sgr near @xmath136379.3 is a stellar feature and should be ignored .
we interpret the huge etr observed for the three dibs on the left in fig . 3 as reflecting the population of many high @xmath4 levels of the large polar carriers of the dibs , due to radiative pumping by dust emission .
while this effect is relatively minor for ch@xmath3 and ch in which only the lowest excited levels are affected , it can be huge for the dibs since a great many high @xmath4 rotational levels are affected because of the low rotational constants of dib carriers . for such small rotational constants , the @xmath67(@xmath4 ) and @xmath19(@xmath4 ) lines are nearly symmetric and linearly spaced by @xmath7 @xmath682@xmath69 ( where @xmath70 is a rotational constant ) at low @xmath4 and the @xmath21(@xmath4 ) lines pile up at the center of the band . as @xmath4 increases , however , the negative quadratic term ( @xmath71)@xmath72 ( see next section ) takes over and causes an @xmath19-branch band - head and the etr .
the rotational constant in the excited state @xmath73 is almost always smaller than that in the ground state @xmath70 since an electronic excitation tends to weaken chemical bonds , which makes the molecule larger and the moment of inertia higher .
the pronounced etrs observed for the three dibs , @xmath135780.5 , @xmath135797.1 , and @xmath136613.6 indicate that their carriers are polar molecules with dipole moments .
the absence of etr for @xmath135849.8 , @xmath136196.0 , and @xmath136379.3 suggests that either the carriers of those dibs are non - polar or they are large polar molecules with very small @xmath71 .
either way , the qualitative difference between the variations of @xmath136613.6 and @xmath136196.0 eliminates the possibility that they are due to the same molecule , in spite of their near perfect correlation ( mccall et al .
2010 ) .
the distribution of molecules in rotational levels is determined as a result of their interaction with the environment through radiation and collision .
they are influenced by the radiative temperature @xmath11 of planck s radiation formula ( planck 1901 ) and the kinetic temperature @xmath74 of maxwell s velocity distribution ( maxwell 1860 ) . if @xmath11 = @xmath74 , molecules are in equilibrium , following the thermal boltzmann distribution ( boltzmann 1871 ) @xmath27 = @xmath11 = @xmath74 . in interstellar space , the two temperatures are generally different , so a non - thermal molecular distribution is more the rule than the exception .
very approximate calculations for simple molecules ( e.g. for h@xmath75 , by oka & epp 2004 ) have been reported , but a calculation for general molecules is beyond the scope of this paper .
here we consider the simplest case of linear molecules , based on an approximate treatment of collisions , in order just to see if we can explain the pronounced etr using realistic molecular parameters .
long linear molecules , more specifically acetylenic carbon chain molecules , have been proposed as candidates for carriers of dibs ( douglas 1977 ; maier et al .
many such molecules are known to have large dipole moments , e. g. , 3.6 debye for hc@xmath76n ( westenberg & wilson ) , 5.6 debye for hc@xmath77n ( broten et al .
1978 ) and 11.9 debye for c@xmath78h@xmath79 ( brnken et al . 2007 ) , and their rotational distributions are sensitive to @xmath11 .
the result of this treatment is also applicable ( approximately ) to nearly linear molecules with a dipole moment along the linear axis , such as cumulene carbon chains h@xmath62c@xmath80
( @xmath81 3 ) , which have also been considered as carriers of dibs ( mccarthy et al .
calculations for general molecules are more complicated because of the three ( instead of one ) rotational constants , quantum numbers , and components of dipole moments , and the lack of spectroscopic and collisional data .
we will keep that as a future project .
nevertheless , the results from our simple case give insight for other more complicated molecules as well .
the first excited levels of light molecules like ch@xmath3 and ch ( 40.1 k and 25.6 k above the ground level , respectively ) have very short lifetimes for spontaneous emission and their rotational distributions are determined solely by @xmath11 . for the large molecules which cause dibs , the rotational constants are smaller by a few orders of magnitude .
therefore , their distributions are governed by collisions ( with @xmath74 ) for low @xmath4 rotational levels . as @xmath4 increases ,
the rate of spontaneous emission increases rapidly ( @xmath82 ) , making radiative effects competitive with collisional effects until the radiative effects ( with @xmath11 ) take over and dominate at high @xmath4 levels .
for example , for hc@xmath83n ( avery et al .
1976 ) with rotational constant 1331.3313 mhz and dipole moment 4.33 debye , the collisional effects dominate for @xmath4 = 1 , become comparable to the radiative effects at @xmath84 , and the radiative effects dominate for high @xmath4 levels . in the following ,
this subtle balance between radiative and collisional effects is considered .
the radiative effects are accurately calculable while the collisional effects are treated approximately . in treating the latter
, we use the principle of detailed balancing ( boltzmann 1872 ) as the guiding principle , as in oka & epp ( 2004 ) . in the laboratory where the number density is high , spontaneous emission between rotational levels ,
@xmath85 is many orders of magnitude slower than collisional processes and is negligible .
then the detailed balancing between @xmath4 and @xmath4 - 1 levels are expressed as , @xmath86 , where @xmath87 is the number density of molecules with rotational quantum number @xmath4 , and @xmath88 and @xmath89 are rates for collision - induced rotational transitions ( oka 1973 ) @xmath90 and @xmath91 , respectively . throughout this paper
we follow the spectroscopists convention , in which the quantum states of higher energy are on the left and exothermic and endothermic processes are expressed by arrows pointing to the right and left , respectively .
the rates satisfy the principle of detailed balancing , that is , @xmath92 where @xmath93 is the number density in thermal equilibrium , @xmath94 is the degeneracy of the level , and @xmath95 is the rotational energy of the molecule in the @xmath4 level ( townes & schawlow 1955 ) . from eqs .
( 1 ) and ( 2 ) , we can derive the thermal boltzmann number density with collisions alone as , @xmath96,\ ] ] after the normalization @xmath97 , where @xmath26 is the total number of molecules .
when radiation is included along with collisions , the detailed balancing takes the form , @xmath98 where the einstein @xmath99 and @xmath70 coefficients express spontaneous emission and induced radiative effect , respectively ( einstein 1916 ) . using einstein s relations @xmath100 , @xmath101 , and planck s formula ( 1901 ) , @xmath102 , we obtain , @xmath103 = n(j-1)\left[a^j\frac{2j+1}{2j-1}\frac{1}{e^{h\nu / kt_r}-1}+c_{j-1}^\uparrow \right],\ ] ] instead of eq . 1 .
a similar equation was used by goldreich & kwan ( 1974 ) for analyzing the radiation trapping by co. in the parentheses after @xmath104 on the left hand side of the equation , the first term represents the number of zero point photons per phase space , which causes spontaneous emission ( dirac 1927 ) , and the second term the number of photons ( bosons ) , which causes induced emission and absorption ( landau & lifshitz 1969 ) . for the @xmath105 transition , @xmath106 and , @xmath107 where @xmath47 is the permanent dipole moment . for collision rates we use approximate formulae ( oka & epp 2004 ) , @xmath108 which satisfy the detailed balancing of eq . 2 , where @xmath109 is a collision rate ( independent of @xmath4 ) and is proportional to the number density of the environment @xmath52 .
using eq .
5 , together with eqs . 6 and 7
, we have , @xmath110,\ ] ] where @xmath111 . unlike eq . 3 ,
@xmath87 depends on both @xmath11 and @xmath74 , as expected
. one obvious shortcoming of this approximate treatment of the collisions is that the effect of collision - induced transitions with @xmath112 is not included .
actually this formalism is a way around dealing with the @xmath112 collision induced transitions for which the rates are not known .
we believe that this omission is somewhat justified from the treatment in the previous section 4.2.2 . , where such transitions are also not included .
. 1 , 2 , and 3 are parallel to eqs . 4 , 5 , and 8 .
the number densities @xmath87 in eq .
8 can not be expressed analytically as in eq . 3 ,
so we calculate them numerically .
they are calculated as a function of the radiative temperatures @xmath11 due to the dust emission and 2.73 k due to the cosmic blackbody radiation , the kinetic temperature @xmath74 , the rotational constant @xmath70 , the dipole moment @xmath47 , and the collision rate @xmath109 . as examples , calculated population fractions for @xmath70 = 1000 mhz , @xmath47 = 4 debye , @xmath74 = 100 k , and @xmath113 s@xmath39
are given in fig .
4 for @xmath11 = 2.73 k , 14.6 k , and 80 k. as @xmath11 increases , the molecular populations in higher @xmath4 levels increase rapidly .
the rotational energy of a linear molecule is given by @xmath114 .
we ignore small higher order effects such as centrifugal distortion and @xmath13 - doubling .
the line positions @xmath35 and the hnl - london factors @xmath25 are given as , @xmath115 where @xmath73 is the rotational constant in the upper state of the transition and @xmath116 and @xmath117 are the hnl - london factors for parallel ( @xmath118 ) and perpendicular ( @xmath119 ) transitions , respectively ( herzberg 1989a ) .
those factors are for transitions with a singlet state @xmath120 in the lower level , but since @xmath4 is high for the carriers of dibs , they give good approximations also for levels with small @xmath13 .
spectra are calculated by using eqs .
9 and the populations @xmath87 calculated in the previous section . in order to take into account uncertainty broadening due to a short emission life time of the upper state or internal conversion by the douglas effect ( douglas 1966 ) ,
spectral lines are broadened by @xmath121 .
the simulated profiles are then convolved with the feros instrumental function ( assumed to be a gaussian with fwhm = 6.25 km s@xmath39 ) . the most crucial molecular parameter for producing the observed etr is the difference between rotational constants @xmath122 .
in order to explain the pronounced etr , we need molecules with large values of @xmath123 .
the experimental values of this constant for carbon molecules ( mostly from maier s group ) are listed in table 2 .
we note that for all molecules except one ( c@xmath83 ) the rotational constant decreases upon electronic excitation , as expected . the reported positive change for c@xmath83 ( motylewski et al .
1999 ) is probably due to a mistake in the analysis ( j. p. maier 2013 , private communication ) .
the large @xmath8 values of 4.3% , 3.4% , 5.7% , 4.9% etc .
are the essence of the observed pronounced etr .
a small increase on the order of 1% can not reproduce the pronounced etr unless @xmath11 is very high .
this indicates that the carriers of dibs with pronounced etr are perhaps relatively small molecules .
the calculated velocity profiles for @xmath135780.5 , @xmath135797.1 , and @xmath136613.6 are shown in fig . 5 , together with parameters which best mimic the observed etrs .
it has been noted in our simulations that , of the seven parameters used in the analysis ( @xmath11 , @xmath74 , @xmath70 , @xmath47 , @xmath109 , @xmath8 , and @xmath124 ) , those related to collisions ( @xmath74 and @xmath109 ) are not very critical .
even the values of @xmath70 and @xmath47 are not that critical .
the most crucial parameters are @xmath11 , @xmath8 , and @xmath124 ( @xmath125 ) .
the essence of the observations and the analysis in this paper is the prominent etr for the three dibs shown in the left column of fig .
3 , in sharp contrast with the three dibs on the right .
the first author has developed the above formalism based on the premise that the etrs result from a single `` local '' cloud near her 36 se .
it is possible instead that the dibs toward her 36 are a composite of those local dibs and ( as seen toward 9 sgr ) foreground dibs , and that the latter need to be subtracted . in that view , the observed profiles reflect an accidental blend of dibs from two clouds : one in the foreground ( without etr ) and the other local to herschel 36 .
the foreground - subtracted profiles shown ( in black ) in the left - hand panel of fig .
8 of paper i are difficult to understand , spectroscopically , however , as the usually observed dib cores are nearly absent . on the other hand , if the foreground absorption were significantly weaker toward herschel 36 than toward 9 sgr ( as shown in the right - hand panel of fig . 8 of paper i ) , then the subtracted profiles would appear to be more consistent with the analysis given above .
that analysis and the discussion and conclusions given below are based on the premise that the beautiful and systematic etrs , which make sense spectroscopically , must be due to a single , local cloud .
the above analysis has indicated that a large fractional decrease in the rotational constant @xmath8 ( @xmath7 5% ) and a high radiative temperature ( @xmath11 @xmath12 50 k ) in the environment of the dib carriers are needed to produce the observed pronounced etrs .
@xmath11 = 14.6 k is not high enough to explain the etr .
the width of an etr is simply proportional to @xmath8 , and high @xmath11 extends the rotational quantum number @xmath4 of the populated levels ( as seen in fig . 4 ) and thus greatly extends the etr .
the requirement of such a high @xmath11 is surprising , but there seems to be no alternative , since the value of @xmath8 is limited ( table 2 ) .
our calculations are only for linear molecules , but it seems that this requirement holds for other types of molecules with three rotational constants , @xmath99 , @xmath70 , and @xmath109 , as long as the etrs are ascribed to the differences between the rotational constants of the excited and ground states . other possible mechanisms , albeit not very realistic , are discussed below in sec .
although our model calculations for polar linear molecules with high @xmath8 and @xmath11 give the observed pronounced etr for the three dibs as shown in fig . 5 , we can not claim that they are the carriers of the dibs .
in particular , a linear molecule can not explain the triplet structure of dib @xmath136613.6 discovered by sarre et al .
a perpendicular transition of a linear molecule gives a triplet structure but the central pileup of the @xmath21 branch lines should be stronger than those of the @xmath67 and @xmath19 branches by a factor of approximately 2 due to the hnl - london factors which contradicts the observation . the claim by schultz et al .
( 2000 ) that @xmath136613.6 is due to a parallel band of a long chain cumulane molecule ( h@xmath62c@xmath80 ) does not work either , since the @xmath21 branch pileup is weak for such molecule .
kerr et al . ( 1996 ) proposed a large planar symmetric top as the carrier of this band .
such a molecule does not have a dipole moment and thus can not be the carrier of @xmath136613.6 ( which is sensitive to @xmath11 ) .
in general , non - polar molecules often considered for dibs ( such as c@xmath80 , hc@xmath80h , h@xmath62c@xmath80h@xmath62 , nc@xmath80n , etc . ) , symmetric pahs ( like pyrene , coronene , ovalene etc . ) , or c@xmath126 and their cations and anions can not be the carriers of dibs exhibiting the etr .
derivatives of pahs , polar pahs such as phenanthrene , tetraphen , pentaphene or recently synthesized olympicene could be the carriers , but most of them probably do not have sufficiently large @xmath8 to produce the etr .
there are no experimental data for @xmath8 for coronene , but a theoretical value @xmath8 = 0.3% ( cossart - magos & leach 1990 ) is an order of magnitude too small to produce the pronounced etr .
cami et al . (
2004 ) discovered small variations in the structure of @xmath136613.6 for different sight lines and interpreted them as due to variations in rotational temperature from 21 k to 25 k. we believe that the rotational temperature is much lower ( because of spontaneous emission ) , and a minute increase of @xmath11 over the 2.73 k blackbody radiation is sufficient to explain the shifts ( due to excitation of @xmath67 and @xmath19 lines of low @xmath4 ) . a similar difficulty is encountered also for @xmath135797.1 .
the high resolution spectrum of this dib shows a sharp @xmath21 branch line .
our model calculation requires a large line broadening which is obviously contradictory .
here we are just content to be able to produce observed huge etrs using realistic @xmath8 and @xmath11 and leave those thorny problems for future studies .
we ascribe the pronounced etr to small ( @xmath7 5% ) variations in the rotational constants , but there may be two other possible interpretations .
it is known that some molecules change their shape from non - linear in the ground electronic state to linear in the excited state .
some examples are ch@xmath62 ( _ _ b__@xmath127
@xmath1 _ _ x__@xmath128 ) , hco ( _ _ __@xmath6 @xmath1 _ _ x__@xmath129 ) and no@xmath62 ( _ _ e__@xmath130
@xmath1 _ _ x__@xmath131 ) ( herzberg 1989b ) .
these molecules are too small to cause dibs , but if there are larger molecules with the same property , the change in the rotational constant @xmath99 by the electronic excitation is 100% and radiative pumping may cause pronounced etr without having to assume a high @xmath11 . while we have not been able to find relevant experimental data for large molecules of this nature , it is a possibility .
such electronic transitions , however , tend to have many vibronic satellites with high franck - condon factors which are not seen for @xmath135780.5 , @xmath135797.1 , and @xmath136613.6 .
alternatively , we may interpret the etr as a result of the excitation of low frequency vibrational modes ( rather than rotational levels in the ground state ) .
the carriers then would not need to be polar molecules .
the radiative effect for a vibrational excitation is typically 100 times less than that for rotational excitation , but the collisional excitations are also small and they may compete .
the bending frequencies of c@xmath76 , 63.4 cm@xmath39 ( schmuttenmaer et al . 1990 ) , and of c@xmath76o@xmath62 , 18.3 cm@xmath39 ( fusina et al . 1980 ) , are too high , but there may be larger molecules with lower vibrational frequencies .
such spectra , however , are likely to exhibit blue as well as red wings . also its effect will look more like the three examples on the right of fig . 3 , @xmath135849.8 , @xmath136196.0 , and @xmath136379.3 .
we have interpreted the pronounced etrs observed for @xmath135780.5 , @xmath135797.1 , and @xmath136613.6 toward herschel 36 as the result of the variation of the rotational constant upon electronic excitation .
our results do not lead to identification of the specific carriers of dibs , but they exclude some possibilities with various degrees of certainty .
the presence or absence of the etr allows discrimination between polar and non - polar carriers .
the carriers of @xmath135780.5 , @xmath135797.1 , and @xmath136613.6 , which show pronounced etr , can not be non - polar molecules .
molecules like pure carbon chains ( c@xmath80 ) or symmetric hydrocarbon chains ( hc@xmath80h , h@xmath62c@xmath80h@xmath62 , nc@xmath80n , etc . ) and their cations and anions can not be the carriers of those dibs .
likewise , symmetric pahs benzene , pyrene , coronene , ovalene etc . and their cations and anions ( cossart - magos & leach 1990 ; see also the many pahs discussed in sharp et al .
2006 ) or c@xmath126 and their cations and anions can not be the carriers of dibs exhibiting strong etr .
the carriers of dibs which show the etr ( on the left of fig .
3 ) can not be the carriers of the dibs without etr ( on the right ) .
in particular , the carriers of @xmath136196.0 and @xmath136613.6 can not be the same molecule , despite the near - perfect correlation of those two dibs ( mccall et al .
2010 ) .
if linear molecules are assumed ( as in section 4 ) , the number of heavy atoms is likely 36 but not much higher , since the values of @xmath8 = @xmath132 are not high enough for heavier molecules to produce the pronounced etr ( unless the radiative temperature is extremely high ) .
many linear molecules of this size have already been studied spectroscopically by maier s group , but cumulene carbene molecules @xmath133-h@xmath62c@xmath134 ( n @xmath12 3 ) may be candidates .
non - linear molecules are outside the scope of this paper .
the high column densities required to produce the strong dibs are problems ( oka & mccall 2011 ; liszt et al .
2012 ) , but this will be a problem for any candidate molecules of this size . the carriers of @xmath135849.8 , @xmath136196.0 , and @xmath136379.3 , which do not show the etr as observed in @xmath135780.5 , @xmath135797.1 , and @xmath136613.6 , are most likely non - polar molecules .
the possibility that the carriers of those dibs exist in different location from other dibs and are not exposed to the dust emission is perhaps safely neglected . however , our simulations have not excluded the possibility that they may be due to large polar molecules with very small @xmath8 . in this paper
we have considered only linear molecules where the etr is created by large @xmath8 = @xmath135 ( @xmath7 5% ) . for non - linear molecules ,
there are other possibilities , such as large decreases in the differences of rotational constants @xmath136 or @xmath137 for prolate or oblate tops , respectively .
either way , the relative variations of these constants are not very likely to be much higher than 5 % , except for the case mentioned in section 5.3 . in most cases , they are 10 times smaller than needed to produce the pronounced etr
. since carriers of dibs with pronounced etrs are polar molecules with modest size , they should be observable by radio astronomy . because of the low density and subthermal rotational distribution in diffuse clouds , the observation need to be by absorption ( liszt et al .
2012 ) , and in relatively long wavelength region for low @xmath4 transitions .
we are indebted to s. federman and a. witt for criticisms of this paper .
t. o. acknowledges helpful discussions with h. liszt , j. p. maier , and p. sarre .
he also thanks j. luque and d. r. crosley for providing theoretical calculations on ch .
we acknowledge support by nsf grants ast 1109014 ( to ) , ast 1008424 ( jd ) , ast 1009639 ( dgy ) , and ast 1238926 ( dew ) .
achkasova , e. , araki , m. , denisov , a. , & maier , j. m. 2006 , j. mol .
237 , 70 amano , t. 2000 , , 531 , l161 amano , t. 2010 , , 716 , l1 anicich , v. g. , & huntress , w. t. , jr .
1986 , , 62 , 553 avery , l. w. , broten , n. w. , macleod , j. m. , oka , t. & kroto , h. w. 1976 , , 205 , l173 bakker , e. j. , van dishoeck , e. f. , waters , l. b. f. m. , & schoenmaker , t. 1997 , , 323 , 469 balm , s. p. & jura , m. 1992 , , 261 , l25 bembenek , z. , kpa , r. , & rytel , m. 1997 , j. mol .
183 , 1 boltzmann , l. 1871 , wiener berichte 63 , 712 boltzmann , l. 1872 , wiener berichte 66 , 275 broten , n. w. , oka , t. , avery , l. w. , macleod , j. m. , & kroto , h. w. , 1978 , , 223 , l105 brnken , s. , gupta , h. , gottlieb , c. a. , mccarthy , m. c. , & thaddeus , p. 2007
, , 664 , l43 cami , j. , salama , f. , jimnez - vicente , galazutdinov , g. a. , & kreowski , j. 2004 , , 611 , l113 carrington , a. , & ramsay , d. a. 1982 , physica scripta , 25 , 272 cernicharo , j. , liu , x .- w . , gonzlez - alfonso , e. et al .
1997 , , 483 , l65 cossart - magos , c. , & leach , s. 1990 , , 233 , 559 crawford , i. a. 1995 , , 277 , 458 dahlstrom , j. , york , d. g. , welty , d. e. et al .
2013 , , in press ( paper i ) dirac , p. a. m. 1927 , proc .
a114 , 243 douglas , a. e. 1966 , j. chem .
45 , 1007 douglas , a. e. 1977 , nature , 269 , 130 douglas , a. e. , & herzberg , g. 1941 , , 94 , 381 dunham , t. , jr . 1937 ,
pasp , 49 , 26 ehrenfreund , p. , & foing , b. h. 1996 , , 307 , l25 einstein , a. 1916 , mitteilungen der physicalischen gesellschaft zrich , 16 , 47 , reprinted in phys .
zeitschrift , 18 , 121 ( 1917 ) follmeg , b. , rosmus , p. , & werner , h .- j .
1987 , chem .
136 , 562 fusina , l. , mills , i. m. , & guelachvili , g. 1980 , j. mol .
79 , 101 galazutdinov , g. a. , lo curto , g. , & kreowski , j. 2008 , , 386 , 2003 gausset , l. , herzberg , g. , lagerqvist , a. , & rosen , b. , 1965 , , 142 , 45 godard , b. , & cernicharo , j. 2013 , , 550 , a8 goldreich , p. , & kwan , j. 1974 , , 189 , 441 goto , m. , stecklum , b. , linz , h. et al . 2006 , , 649 , 299 gredel , r. , van dishoeck , e. f. , & black , j. h. 1993 , , 269 , 477 gredel , r. 1997 , , 320 , 929 habart , e. , dartois , e. , abergel , a. et al .
2010 , , 518 , l116 hall , d. i. , miles , j. r. , sarre , p. j. , & fossey , s. j. 1992 , nature , 358 , 629 heger , m. l. 1922 , lick obs . bull .
337 , 141 herzberg , g. 1989a , molecular spectra and molecular structure i. spectra of diatomic molecules , krieger publishing company , malabar , florida herzberg , g. 1989b , molecular spectra and molecular structure iii . electronic spectra and electronic structure of polyatomic molecules , krieger publishing company , malabar , florida hobbs , l. m. , thorburn .
j. a. , oka , t. et al .
2004 , , 615 , 947 hobbs , l. m. , york , d. g. , snow , t. p. et al .
2008 , , 680 , 1265 hobbs , l. m. , york , d. g. , thorburn , j. a. et al .
2009 , , 705 , 32 kalemos , a. , mavridis , a. , & metropoulos , a. 1999 , j. chem .
111 , 9536 kaufer a. , stahl , o. , tubbesing , s. et al .
1999 , the messenger , 95 , 8 kpa , r. , para , a. , rytel , m. , & zachwieja , m. 1996 , j. mol .
178 , 189 kerr , t. h. , hibbins , r. e. , miles , j. r. et al .
1996 , , 283 , l105 kreowski , j. , & schmidt , m. 1997 , , 477 , 209 kreowski , j. , & walker , g. a. h. 1987 , , 312 , 860 kuhn , r. , maier , j. p. , & ochsner , m. 1986 , mol .
59 , 441 landau , l. d. , & lifshitz , e. m. , 1969 , statistical physics ( new york : pergamon press ) larsson , m. , & siegbahn , p. e. m. 1983a , chem .
76 , 175 larsson , m. , & siegbahn , p. e. m. 1983b , j. chem .
79 , 2270 lien , d. j. 1984 , , 284 , 578 linnartz , h. , vaizert , o. , motylewski , t. , & maier , j. p. 2000 , j. chem .
112 , 9777 liszt , h. , sonnentrucker , p. , , cordiner , m. , & gerin , m. 2012 , , 753 , l28 luque , j. , & crosley , d. r. 1996 , j. chem . phys .
104 , 2146 maier , j. p. , walker , g. a. h. , & bohlender , d. a. 2004 , , 602 , 286 151 , 103 maxwell , j. c. 1860 , phil . mag . , 4 , 19 mccall , b. j. , drosback , m. m. , thorburn , j. a. et al . 2010 , , 708 , 1628 mccarthy , m. c. , traverse , m. j. , kovcs , a. et al .
1997 , science , 275 , 518 mckellar , a. 1941 , pasp , 53 , 233 mckellar , a. 1949 , pub .
dominion astrophys . obs .
vii , no 15 , 251 meyer , d. m. & jura , m. 1984 , 276 , l1 motylewski , t. , vaizert , o. , giesen , t. f. , linnartz , h. , & maier , j. p. 1999
, j. chem .
111 , 6161 nagy , z. , van der tak , f. f. s. , ossenkopf , v. et al .
2013 , , 550 , a96 naylor , d. a. , dartois , e. , habart , e. et al .
2010 , , 518 , l117 oka , t. 1973 , adv . at .
phys . 9 , 127 oka , t. , & epp , e. , 2004 , , 613 , 349 oka , t. , & mccall , b. j. 2011 , science , 331 , 293 ornellas , f. r. , & machado , f. b. c. 1986 , j. chem .
84 , 1297 pfluger , d. , motylewski , t. , linnartz , h. , sinclair , w. e. , & maier , j. p. 2000 , chem .
phys . lett .
329 , 29 phelps , d. h. , & dalby , f. w. 1966 , phys .
16 , 3 planck , m. 1901 , ann .
d. phys . 4 , 564 salama , f. , galazutdinov , g. , kreowski , j. et al .
2011 , , 728 , 154 sarre , p. j. , miles , j. r. , kerr , t. h. et al .
1995 , , 277 , l41 schulz , s. a. , king , j. e. , & glinski , r. j. 2000 , , 312 , 769 schmuttenmaer , c. a. , cohen , r. c. , pugliano , n. , heath , j. r. , & cooksy , a. l. 1990 , science , 249 , 897 sharp , r. g. , reilly , n. j. , kable , s. h. , & schmidt , t. w. 2006 , , 639 , 194 sinclair , w. e. , pfluger , d. , linnartz , h. , & maier , j. p. 1999
, j. chem .
110 , 296 sinclair , w. e. , pfluger , d. , verdes , d. , & maier , j. p. 2000 , j. chem .
112 , 8899 snow , t. p. , & mccall , b. j. 2006 , ara&a , 44 , 367 spitzer , l. , jr .
1978 , physical processes in the interstellar medium , john wiley & sons , new york sun , h. , & freed , k. f. 1988 , j. chem .
phys , 88 , 2659 thaddeus , p. , & clauser , j. f. 1966 , phys
lett . , 16 , 819 thi , w .- f . , mnard , f. , meeus , g. et al .
2011 , , 530 , l2 townes , c. h. & schawlow , a. l. 1955 , microwave spectroscopy , mcgraw - hill book co. , new york webster , a. 1996 , , 282 , 1372 welty , d. e. , hobbs , l. m. , & morton , d. c. 2003 , , 147 , 61 westenberg , a. a. & wilson , e. b. , jr .
1950 , jacs , 72 , 199 zachwieja , m. 1995 , j. mol .
170 , 285 ch@xmath3 & _ _
__@xmath5 @xmath1 _ _ x__@xmath6(0,0)@xmath138 & @xmath139 & 4226.982 & @xmath61 0.9 & 0.4 & @xmath61 0.26 & @xmath61 22.9 + & @xmath22 = 0.00545@xmath140 & @xmath141 & 4229.273 & [email protected] & 0.5 & [email protected] & [email protected] + & & @xmath142 & 4232.472 & [email protected] & 1 & [email protected] & + & & @xmath143 & 4237.481 & [email protected] & 0.5 & [email protected] & [email protected] + & & @xmath144 & 4239.302 & @xmath61 0.9 & 0.5 & @xmath610.26 & @xmath61 22.9 + @xmath41ch@xmath3 & _ _
__@xmath5 @xmath1 _ _ x__@xmath6(0,0)@xmath145 & @xmath142 & 4232.200 & [email protected] & 1 & [email protected] & + & + ch@xmath3 & _ _
__@xmath5 @xmath1 _ _ x__@xmath6(1,0)@xmath138 & @xmath141 & 3955.407 & [email protected] & 0.5 & [email protected] & [email protected] + & @xmath22 = 0.00331@xmath140 & @xmath142 & 3957.618 & [email protected] & 1 & [email protected] & + & & @xmath143 & 3961.994 & [email protected] & 0.5 & [email protected] & [email protected] + @xmath41ch@xmath3 & _ _
__@xmath5 @xmath1 _ _ x__@xmath6(1,0)@xmath138 & @xmath142 & 3958.100 & [email protected] & 1 & [email protected] & + & + ch & _ _
__@xmath5 @xmath1 _ _ x__@xmath6(0,0)@xmath146 & @xmath147 & 4300.227 & [email protected] & 1 & [email protected] & + & @xmath22 = 0.00505@xmath148 & @xmath149 & 4303.861 & [email protected] & 1.5 & [email protected] & [email protected] + ch & _ _
b__@xmath150 @xmath1 _ _ x__@xmath6(0,0)@xmath151 & @xmath147 & 3878.700 & [email protected] & 1/3 & [email protected] & + & @xmath22 = 0.00320@xmath152 & @xmath153 & 3886.341 & [email protected] & 1 & [email protected] & + & & @xmath154 & 3890.148 & [email protected] & 2/3 & [email protected] & + c@xmath76 & @xmath155 @xmath1 @xmath156 & 0.4305 & -0.0183 & -4.3 & c + @xmath157-h@xmath62c@xmath76 & @xmath158 @xmath1 @xmath159 & 0.346783@xmath138 & -0.01182@xmath138 & -3.4@xmath138 & d + c@xmath160 & @xmath127 @xmath1 @xmath161 & 0.166111 & -0.0091 & -5.7 & e + hc@xmath160h@xmath3 & @xmath162 @xmath1 @xmath163 & 0.14690 & -0.00681 & -4.9 & f + c@xmath83 & @xmath155 @xmath1 @xmath156 & 0.0853133 & 0.0002@xmath164 & 0.2@xmath164 & g + hc@xmath165h@xmath3 & @xmath162 @xmath1 @xmath163 & 0.0445943 & -0.0008022 & -1.8 & h + hc@xmath78h@xmath3 & @xmath162 @xmath1 @xmath163 & 0.0190779 & -0.0002106 & -1.1 & i + hc@xmath166n@xmath3 & @xmath6 @xmath1 @xmath6 & 0.0189665 & -0.001934 & -1.0 & j + c@xmath167h@xmath168 & & 0.011123@xmath169 & -0.000033@xmath169 & -0.3@xmath169 & k +
|
absorption spectra toward herschel 36 for the _ _
__@xmath0 @xmath1 _ _ x__@xmath2 transitions of ch@xmath3 in the @xmath4 = 1 excited rotational level and the _ _
__@xmath5 @xmath1 _ _ x__@xmath6 transition of ch in the @xmath4 = 3/2 excited fine structure level have been analyzed .
these excited levels are above their ground levels by 40.1 k and @xmath7 25.7 k and indicate high radiative temperatures of the environment , 14.6 k and 6.7 k , respectively .
the effect of the high radiative temperature is more spectacular in some diffuse interstellar bands ( dibs ) observed toward her 36 ; remarkable extended tails toward red ( etr ) were observed .
we interpret these etrs as due to a small decrease of rotational constants upon excitation of the excited electronic states . along with radiative pumping of a great many high-@xmath4 rotational levels
, this causes the etrs . in order to study this effect
quantitatively , we have developed a model calculation in which the effects of collision and radiation are treated simultaneously .
the simplest case of linear molecules is considered .
it has been found that the etr is reproduced if the fraction of the variation of the rotational constant , @xmath8 @xmath9 , is sufficiently high ( 3 @xmath10 5% ) and the radiative temperature is high ( @xmath11 @xmath12 50 k ) .
although modeling for general molecules is beyond the scope of this paper , the results indicate that the prototypical dibs @xmath135780.5 , @xmath135797.1 , and @xmath136613.6 which show the pronounced etrs are due to polar molecules sensitive to the radiative excitation .
the requirement of high @xmath8 favors relatively small molecules with 3 - 6 heavy atoms . dibs @xmath135849.8 , @xmath136196.0 , and @xmath136379.3 which do not show the pronounced etrs are likely due to non - polar molecules or large polar molecules with small @xmath8 .
|
the treatment of boundary problems in symbolic computation was initiated in the phd thesis @xcite under the guidance of bruno buchberger in cooperation with heinz engl ; see also @xcite and for the further development @xcite .
its implementation was originally carried out within the th@xmath0orem@xmath1project @xcite . up to now
, we have always assumed differential equations without singularity or , equivalently , monic differential operators ( leading coefficient function being unity ) . in this paper
, we develop for the first time an _ algebraic theory _ for treating boundary problems with a ( mild ) _ singularity at one endpoint_. for details , we refer to section [ sec : bp - one - sing ] .
our approach is very different from the traditional analysis setting in terms of the weyl - titchmarsh theory ( limit points and limit circles ) .
it would be very interesting to explore the connections between our approach and the classical treatment ; however , this must be left for future work .
regarding the general setup of the _ algebraic language for boundary problems _ , we refer to the references mentioned above , in particular @xcite . at this point ,
let us just recall some notation .
we start from a fixed integro - differential algebra @xmath2 .
the formulation of ( local ) boundary conditions is in terms of _ evaluations _ , which are by definition multiplicative linear functionals in @xmath3 .
we write @xmath4 if @xmath5 is a _ subspace _ of @xmath6 ; the same notation is used for subspaces of the dual space @xmath3 .
orthogonal _ of a subspace @xmath7 is defined as @xmath8 and similarly for the orthogonal of a subspace @xmath9 .
we write @xmath10 $ ] for the _ linear span _ of ( possibly infinitely many ) elements @xmath11 ; the same notation is used for linear spans within @xmath3 .
the _ zero vector space _ of any dimension is denoted by @xmath12 .
we write @xmath13 for the _ falling factorial _ , where @xmath14 could be arbitrary ( but will be an integer for our purposes ) while @xmath15 is taken to be a natural number .
note that @xmath16 for @xmath17 .
our main example of an integro - differential algebra will be the space @xmath18 of analytic functions on a closed interval @xmath19 $ ] . recall that by definition this is the space of functions that are analytic in some open set containing @xmath20 $ ] .
the development of the new algebraic theory of two - point boundary problems with a mild singularity ( whose treatment is really just broached in this paper ) was triggered by a collaboration between a _ symbolic computation _ team ( consisting of the first , second and fourth author ) and a researcher in _ engineering mechancs _ ( the third author ) .
this underlines the importance and fruitfulness of collaborations between theoretical developments and practical applications .
we present the example that had originally led to this research in section [ sec : appl ] . from a methodological point of view
, this research on the symbolic algorithm for linear boundary value problems has particular relevance and is drawing from the theorema project , see @xcite .
this project aims at supporting a new paradigm for doing ( algorithmic ) mathematical research : in the phase of doing research on new theorems and algorithms , th@xmath0orem@xmath1 provides a formal language ( a version of predicate logic ) and an automated reasoning system by which the exploration of the theory is supported . in the phase in which algorithms based on the new theory should be implemented and used in computing examples , th@xmath0orem@xmath1 allows to program and execute the algorithms _ in the same language_. in the particular case of our approach to solving linear boundary value problems , the fundamental theorem on which the approach is based was proved automatically by checking that the rewrite rules for integro - differential operators forms a grbner basis . in a second step ,
the algorithm for solving linear boundary value problems is expressed again in the th@xmath0orem@xmath1language and can then be called by the users by inputting the linear boundary value problems in a user - friendly notation . in its current version
, the engine for solving boundary problems is bundled in the ` greengroebner ` package of th@xmath0orem@xmath1 . as an example , consider the boundary problem @xmath21 which we shall consider in greater detail in section [ sec : simple - example ] .
this can be given to th@xmath0orem@xmath1 in the form @xmath22\ ] ] leading either to the solution for @xmath23 as @xmath24\,\mathbf{d}\xi - x^2 \int_0^x f[\xi]\,\mathbf{d}\xi - x^2 \int_0 ^ 1 \xi^2 f[\xi]\,\mathbf{d}\xi + \int_0^x \xi^2 f[\xi]\,\mathbf{d}\xi \right)}\ ] ] or to the green s function @xmath25 as @xmath26 at the user s request .
in addition to the ` greengroebner ` package in th@xmath0orem@xmath1 , a maple package named ` intdiffop ` is also available @xcite .
this package was developed in the frame of anja korporal s phd thesis , supervised by georg regensburger and the first author .
the maple package supports also generalized boundary problems ( see section [ sec : simple - example ] for their relevance to this paper ) .
one advantage of the th@xmath0orem@xmath1 system is that both the research phase and the application phase of our method can be formulated and supported within the same logic and software system which we consider to be quite a novel and promising paradigm for the future .
for illustrating the new ideas , it is illuminating to look at a simple example that exhibits the kind of phenomena that we have to cope with in the kirchhoff plate boundary problem . [
ex : simple - example ] let us start with the _ intuitive but mathematically unprecise _ statement of the following example : given a `` suitable '' forcing function @xmath27 on the unit interval @xmath28 $ ] , we want to find a `` reasonable '' solution function @xmath29 such that @xmath30 but note that the differential operator @xmath31 is singular at the left boundary point @xmath32 of the interval @xmath33 under consideration .
hence the first boundary condition @xmath34 should be looked at with some suspicion . and what function space are we supposed to consider in the first place ? if the @xmath35 and @xmath36 are to be taken literally , the space @xmath37 $ ] will clearly not do . on the other hand ,
we need functions that are smooth ( or at least continuous ) at @xmath32 to make sense of @xmath38 . in the rest of this section , we shall give one possible solution to the dilemma outlined above .
of course we could resort to using different function space for @xmath29 and @xmath27 , and this is in fact the approach one usually takes in analysis . for our present purposes ,
however , we prefer to keep the simple paradigm of integro - differential algebras as outlined in section [ sec : intro ] , but we shall modify it to accommodate singularities such as in @xmath35 and @xmath36 . note that these are just poles , so we can take @xmath39 to be the subring of the field @xmath40 consisting of complex - valued _ meromorphic functions _ that are regular at all @xmath41 except possibly @xmath32 . in other words , these are functions that have a laurent expansion at @xmath32 with finite principal part , converging in a complex annulus @xmath42 with @xmath43 .
in fact , we will only use the real part @xmath44 \setminus \{0\}$ ] of this annulus .
note that we have of course @xmath45 .
the ring @xmath39 is an integro - differential algebra over @xmath46 if we use the standard _
derivation _
@xmath47 and the _ rota - baxter operator _
@xmath48 initialized at the regular point @xmath49 .
we have now ensured that the differential operator of has a clear algebraic interpretation @xmath50 $ ] .
however , the boundary condition @xmath34 is still dubious . for making it precise , note that the integro - differential algebra @xmath2 has only the second of the two boundary evaluations @xmath51 with @xmath52 and @xmath53 in the usual sense of a total function
so while we can interpret the second boundary condition algebraically by @xmath54 , the same does not work on the left endpoint . instead of an evaluation at
@xmath32 we shall introduce the map @xmath55 that extracts the _
principal part _ of a function written in terms of its laurent expansion at @xmath32 . here and
henceforth we assume such expansions of nonzero functions are written with @xmath56 . if @xmath57 the function is regular at @xmath32 , and the above sum is to be understood as zero . clearly , @xmath58 is a linear projector , with the complementary projector @xmath59 extracting the _ regular part _ at @xmath32 .
incidentially , @xmath60 and @xmath61 are also rota - baxter operators of weight @xmath62 , which play a crucial role in the renormalization theory of perturbative quantum field theory ( * ? ? ?
1.1.10 ) .
finally , we define the functional @xmath63 that extracts the _ constant term _
@xmath64 of a meromorphic function expanded at @xmath32 . combining @xmath65 with the monomial multiplication operators , we obtain the coefficient functionals @xmath66 that map @xmath67 to @xmath68 .
in particular , the residue functional is given by @xmath69 .
note that for functions regular at @xmath32 , the functional @xmath65 coincides with the evaluation at the left endpoint , @xmath70 .
however , for general meromorphic functions , @xmath71 is undefined and @xmath65 is _ not multiplicative _ since for example @xmath72 .
hence we refer to @xmath65 only as a functional but not as an evaluation .
we can now make the boundary condition @xmath73 precise for our setting over @xmath39 .
what we really mean is that @xmath74 , where @xmath75 is the projector that extracts the principal part together with the constant term . extending the algebraic notion of boundary problem to allow for boundary conditions that are not functionals
, we may thus view as @xmath76^\perp})$ ] . in this way we have given a precise meaning to the _ formulation of the boundary problem_. but
how are we to go about its _ solution _ in a systematic manner ?
let us first look at the adhoc _ standard method _ way of doing this determining the general homogeneous solution , then add the inhomogeneous solution via variation of constants , and finally adapt the integration constants to accommodate the boundary conditions . in our case , one sees immediately that @xmath77 $ ] so that the general solution of the homogeneous differential equation is @xmath78 , where @xmath79 are integration constants .
variation of constants @xcite then yields @xmath80 for the inhomogeneous solution .
note that @xmath81 may also have singularities as @xmath32 or any other point @xmath82 apart from @xmath49 .
now we need to impose the _
boundary conditions_. from @xmath83 we obtain immediately @xmath84 . for the boundary condition at @xmath32 we have to proceed a bit more cautiously , obtaining @xmath85 where we assume that @xmath27 is regular at @xmath86 .
it is clear that the remaining limit can only exist if the integral tends to a finite limit as @xmath87 , and this is the case by our assumption on @xmath27 .
but then we may apply the boundary condition to obtain @xmath88 .
this gives the overall solution @xmath89 which one may write in the standard form @xmath90 where the _ green s function _ is defined as @xmath91 in the usual manner .
how are we to make sense of this in an algebraic way , i.e. without ( explicit ) use of limits and hence topology ?
the key to this lies in the projector @xmath60 and the functional @xmath71 , which serve to distill into our algebraic setting what we need from the topology ( namely @xmath92 as @xmath87 ) .
however , there is another complication when compared to boundary problems without singularities as presented in section [ sec : intro ] : we can not expect a solution @xmath93 to for every given forcing function @xmath81 .
in other words , this boundary problem is _ not regular _ in the sense of @xcite . in the past
, we have also used the term _ singular boundary problem _ for such situations ( here this seems to be suitable in a double sense but we shall be careful to separate the second sense by sticking to the designation `` boundary problems with singularities '' ) .
the theory of singular boundary problems was developed in an abstract setting in @xcite ; applications to boundary problems ( without singularities ) have been presented in @xcite . at this point
we shall only recall a few basic facts .
a boundary problem @xmath94 is called _ semi - regular _ if @xmath95 .
it is easy to see that the boundary problem @xmath96^\perp})$ ] is in fact semi - regular .
since any @xmath97 can be written as @xmath78 , the condition @xmath54 implies @xmath98 and hence @xmath99 .
but then @xmath100 forces @xmath101 and hence @xmath102 . if @xmath76^\perp})$ ] were a regular boundary problem , we would have @xmath103^\perp } = { \mathcal{f}}$ ]
however , it is easy to see that there are elements @xmath93 that do not belong to @xmath104^\perp}$ ] , for example @xmath105 .
hence we conclude that the boundary problem is in fact overdetermined . for such boundary problems
@xmath94 one can always select a _ regular subproblem _
@xmath106 , in the sense that @xmath107 . in our case ,
a natural choice is @xmath108 $ ] .
this is regular since the evaluation matrix @xmath109 is regular , and the associated kernel projector is @xmath110 by ( * ? ? ?
a.1 ) . for making the boundary problem well - defined on @xmath39
we need one more ingredient : we have to fix a complement @xmath111 of @xmath112 , the so - called _ exceptional space_. intuitively speaking ,
this comprises the `` exceptional functions '' of @xmath39 that we decide to discard in order to render solvable .
let us first work out what @xmath113 looks like .
since @xmath114 forces the principal part and constant term to vanish , we may start from @xmath115 , with the additional proviso that @xmath116 .
applying @xmath117 to this @xmath23 yields @xmath118 , which represents an arbitrary element of @xmath119^\perp}$ ] for a suitable choice of coefficients @xmath120 since the additional condition @xmath116 can always be met by choosing @xmath121 .
but then it is very natural to choose @xmath122^\perp}$ ] as the required complement
. clearly , the elements of this space @xmath111 are the laurent polynomials of @xmath123 $ ] without constant term . by prop . 2 of @xcite , the _
green s operator _ of a generalized boundary problem @xmath124 is given by @xmath125 , where @xmath126 is the green s operator of some regular subproblem @xmath106 and @xmath127 is the projector onto @xmath112 along @xmath111 . in our case ,
the latter projector is clearly @xmath128 , while the green s operator @xmath129 by ( * ? ? ?
26 ) , with the kernel projector @xmath130 as above and the fundamental right inverse @xmath131 given according to ( * ? ? ?
23 ) by @xmath132 which is essentially just a reformulation of the inhomogeneous part in .
following the style of @xcite we write here @xmath133 as @xmath134}$ ] to emphasize its role in the integro - differential operator ring .
similarly , we write @xmath135 for the definite integral over the full interval @xmath33 , which we may regard as a linear functional @xmath136 . putting things together
, it remains to compute @xmath137 which may be done by the usual _
rewrite rules _
1 ) for the operator ring @xmath138}$ ] , together with the obvious extra rules that on @xmath139 the residual @xmath140 vanishes and @xmath141 coincides with the evaluation @xmath71 at zero . using the standard procedure for extracting the _ green s function _ , one obtains exactly when restricting the forcing functions to @xmath142 .
we have thus succeeded in applying the algebraic machinery to regain the solution previoulsy determined by analysis techniques .
more than that : the precise form of accessible forcing functions is now fully settled , whereas the regularity assumption in was left somewhat vague ( a sufficient condition whose necessity was left open ) .
let us now address the general question of specifying and solving boundary problems ( as usual : relative to a given fundamental system ) that have only _ one singularity_. the case of multipliple singularities is left for future investigations . using a scaling transformation ( and possibly a reflection )
, we may thus assume the same setting as in section [ sec : simple - example ] , with the _ singularity at the origin _ and the other boundary point at @xmath143 .
for the scope of this paper , we shall also restrict ourselves to a certain subclass of stieltjes boundary problems @xmath144 : first of all , we shall allow _ only local boundary conditions _ in @xmath145 .
this means multi - point conditions and higher - order derivatives ( leading to ill - posed boundary problems with distributional green s functions ) are still allowed , but no global parts ( integrals ) ; for details we refer to ( * ? ? ?
* def . 1 ) .
the second restriction concerns the differential operator @xmath146 $ ] , which we require to be _
fuchsian without resonances_. the latter means the differential equation @xmath147 is of fuchsian type ( the singularity is regular ) , and has fundamental solutions @xmath148 with each @xmath149 having order @xmath86 , where @xmath150 are the roots of the indicial equation @xcite .
in other words , we do not require logarithms for the solutions .
( a sufficient but not necessary condition for this is that the roots @xmath151 are all distinct and do not differ by integers . )
we call @xmath94 a _ boundary problem with mild singularity _ if @xmath146 $ ] is a nonresonant fuchsian operator and @xmath145 a local boundary space . for a fixed @xmath94 , we shall then enlarge the _ function space _ @xmath39 of section [ sec : simple - example ] by adding @xmath152 as algebra generators , where each @xmath153 is the fractional part of the corresponding indicial root @xmath151
. every element of @xmath39 is then a sum of series @xmath154 , with @xmath155 and @xmath156 .
the integro - differential structure on @xmath2 is determined by setting @xmath157 , as usual , and by using for @xmath158 the integral @xmath133 as we did also in section [ sec : simple - example ] . as _ boundary functionals _ in @xmath145 , we admit derivatives @xmath159 and coefficient functionals @xmath160 whose action is @xmath161 . for functions
@xmath142 we have of course @xmath162 .
since the projectors @xmath61 and @xmath60 can be expressed in terms of the coefficient functionals , we shall henceforth regard the latter just as convenient abbrevations ; boundary spaces are always written in terms of @xmath163 and @xmath164 only but of course they can be infinite - dimensional .
for example , in section [ sec : simple - example ] we had the `` regularized boundary condition '' @xmath74 , which is equivalent to @xmath165 and @xmath166 and hence to @xmath167 .
its full boundary space is therefore @xmath168 $ ] .
the first issue that we must now address is the choice of _ suitable boundary conditions _ : unlike in the `` smooth case '' ( without singularities ) , we may not be able to impose @xmath169 boundary conditions for an @xmath169-th order differential equation .
the motivating example of section [ sec : simple - example ] was chosen to be reasonably similar to the smooth case , so the presence of a singularity was only seen in replacing the boundary evaluation @xmath71 by its regularized version @xmath170 . as explained above , we were effectively adding the extra condition @xmath166 to the standard boundary conditions @xmath171 .
in other cases , this will not do as the following simple example shows .
[ ex : ivp ] consider the nonresonant fuchsian differential equation @xmath172 .
note that here the indicial equation has the roots @xmath173 and @xmath174 , which differ by the integer @xmath143 .
nevertheless , we may take @xmath175 as a fundamental system , so @xmath176 is indeed nonresonant .
trying to impose the same ( regularized ) boundary space @xmath177 $ ] as in example [ ex : simple - example ] , one obtains @xmath178 after a short calculation . but
this means that the forcing functions @xmath27 in the boundary problem @xmath179 must satisfy an awkward extra condition ( viz .
the one on the right - hand side of @xmath180 above ) .
this is not compensated by the slightly enlarged generality of allowing @xmath27 to have a simple pole at @xmath32 . in the present case
, we could instead impose _ initial conditions _ at @xmath86 so that @xmath181 $ ] . in this case one
gets @xmath182 , so there is a unique solution for every analytic forcing function . for a given nonresonant fuchsian operator @xmath146 $ ] , a better approach appears to be the following ( we will make this more precise below ) : 1 .
we compute first some boundary functionals @xmath183 that ensure a _ regular subproblem
_ @xmath94 with @xmath184 $ ] .
adding extra conditions ( vanishing of all @xmath185 for sufficiently small @xmath15 ) we obtain a boundary space @xmath186 such that @xmath187 is semi - regular .
2 . if a _
particular boundary condition _ is desired , it may be `` traded '' against one of the @xmath188 ; if this is not possible , it can be `` annexed '' to the extra conditions . after these amendments , the subproblem @xmath189 is still regular , and @xmath94 still semi - regular .
next we compute the corresponding _ accessible space _ @xmath112 .
this space might not contain @xmath18 , as we saw in example [ ex : ivp ] above when we insisted on the conditions @xmath190 .
we determine a complement @xmath111 of @xmath112 as _ exceptional space _ in @xmath191 .
once these steps are completed , we have a regular generalized boundary problem @xmath191 whose _ green s operator _
@xmath192 can be computed much in the same way as in section [ sec : simple - example ] . in detail , we get @xmath125 , where @xmath126 is the green s operator of the regular subproblem @xmath189 and @xmath127 the projector onto @xmath112 along @xmath111 . as we shall see , the operators @xmath192 and @xmath127 can be computed as in the usual setting @xcite .
let us first address step 1 of the above program .
let @xmath146 $ ] be a nonresonant fuchsian differential operator of order @xmath169 .
then there exists a fundamental system @xmath193 of @xmath176 and @xmath169 coefficient functionals @xmath194 ordered as @xmath195 so that @xmath196 is a lower unitriangular matrix .
we start from an arbitrary fundamental system @xmath197 of the fuchsian operator @xmath176 , where we take @xmath198 fracational as before and we may assume that @xmath199 so that each fundamental solution @xmath200 has order @xmath201 . ( the order of a series @xmath202 is defined as the smallest integer @xmath15 such that @xmath203 . )
we order the fundamental solutions such that @xmath204 .
we can always achieve strict inequalities as follows .
if @xmath205 is the first place where @xmath206 we must also have @xmath207 since @xmath208 .
therefore we have @xmath209 , and we can replace @xmath210 by @xmath211 and make it monic so as to ensure @xmath212 . repeating this process at most @xmath213 times
we obtain @xmath195 .
choosing now the boundary functionals @xmath194 as in the statement of the lemma , we have clearly @xmath214 and @xmath215 for @xmath216 as claimed .
in particular we see that @xmath217 has unit determinant , so it is regular . setting @xmath218 $ ]
, we obtain a regular boundary problem @xmath189 .
note that some of the @xmath153 may coincide .
for each @xmath219 let @xmath220 be the smallest of the @xmath201 with @xmath221 .
we expand the @xmath169 boundary functionals by suitable curbing constraints to the full boundary space @xmath222\ ] ] since the inhomogeneous solutions should be at least as smooth ( in the sense of pole order ) as the homogeneous ones .
note that @xmath94 is clearly a semi - regular boundary problem .
this achieves step 1 in our program .
now for step 2 .
suppose we want to _ impose a boundary condition _
@xmath223 , assuming it is of the type discussed above ( composed of derivatives @xmath224 and coefficient functionals @xmath225 for fractional parts @xmath226 of indicial roots ) .
we must distinguish two cases : trading .
: : if the row vector @xmath227 is nonzero , we can express it as a @xmath46-linear combination @xmath228 of the rows @xmath229 of @xmath230 .
let @xmath15 be the largest index such that @xmath231 .
then we may express @xmath232 as a @xmath46-linear combination of @xmath233 and the remaining rows @xmath234 , hence we may exchange @xmath235 with @xmath223 without destroying the regularity of @xmath236 .
: : otherwise , we have @xmath237 . together with @xmath238
this implies @xmath239 \cap { \mathcal{b}}_n\big)}\!^\perp } = { \beta^\perp } + { { \mathcal{b}}_n^\perp } = { \mathcal{f}}$ ] and hence @xmath240 by the identities of ( * ? ? ?
a ) . furthermore , lemma 4.14 of @xcite yields @xmath241 since we have @xmath242 from the regularity of @xmath94 .
but this means that adding @xmath223 to @xmath145 as a new boundary condition necessarily cuts down the space of accessible functions unless @xmath223 happens to be in the span of the curbing constraints @xmath243 added to @xmath244 in . in the sense of the above discussion
( see example [ ex : ivp ] ) , the first case signifies a `` natural '' choice of boundary condition while the second case means we insist on imposing an extra condition ( unless it is a redundant curbing constraint ) . repeating these steps as the cases may be , we can successively impose any ( finite ) number of given boundary conditions .
this completes step 2 . for step 3
we require the computation of the accessible space @xmath112 .
we shall now sketch how this can be done algorithmically , starting with a finitary description of the _ admissible space _
@xmath245 as given in the next proposition .
the proof is unfortunately somewhat tedious and long - wided but the basic idea is simple enough : we substitute a series ansatz into the boundary conditions specified in @xmath145 to determine a number of lowest - order coefficients .
the rest is just some bureaucracy for making sure that everything works out ( most likely this could also be done in a more effective way ) .
[ prop : char - adspc ] let @xmath94 be a semi - regular boundary problem of order @xmath169 with mild singularity such that @xmath246 is a regular subproblem .
let @xmath247 be the fractional parts of the indicial roots for @xmath176 .
then we have a direct decomposition of the admissible space @xmath248 with components @xmath249 + p_{\mu } \big ( c^\omega(i)^m \big ) \qquad ( \mu \in m).\ ] ] here @xmath250 $ ] are linearly independent laurent polynomials and the linear operators @xmath251 are defined by @xmath252 with @xmath253 and laurent polynomials @xmath254 $ ] , almost all of which vanish over the summation range @xmath255 and @xmath256 .
any element @xmath93 may be written in the form @xmath257 where the @xmath258 are _ a priori _ arbitrary . however , because of the curbing constraints of @xmath145 in we may assume @xmath259 for @xmath260 . let us write @xmath261 for the extra conditions that were `` annexed '' at step 2 ( if none were added then clearly @xmath262 ) ; thus @xmath145 is a direct sum of @xmath244 and @xmath261 and the curbing constraints . if @xmath263 denotes the largest integer such that @xmath264 occurs in any condition of @xmath265 and if we set @xmath266 , we can also write @xmath267 as @xmath268 where the @xmath269 are convergent power series .
next we impose the conditions @xmath270 on @xmath29 .
but each @xmath223 is a @xmath46-linear combintions of coefficient functionals @xmath225 with @xmath271 and derivatives @xmath272 whose action on @xmath29 yields @xmath273 hence @xmath274 yields a @xmath46-linear combination of the @xmath275 and the @xmath276 .
we compile the coefficients @xmath277 into a column vector @xmath278 of size @xmath279 , consisting of @xmath280 contiguous blocks of varying size @xmath281 . putting the ( finite )
set @xmath282 of evaluation points occurring in @xmath265 into ascending order , we compile also the derivatives @xmath283 into a column vector @xmath284 of size @xmath285 , consisting of @xmath280 contiguous blocks , each holding @xmath286 derivatives @xmath287 for a fixed @xmath288 and certain @xmath289 and @xmath256 . assuming @xmath290 has dimension @xmath291 , the boundary conditions @xmath292 for @xmath293 can be written as @xmath294 for suitable matrices @xmath295 and @xmath296 that can be computed from the boundary functionals @xmath270 .
regarding the right - hand side as given , let us now put into row echelon form ( retaining the same letters for simplicity ) .
if the resulting system contains any rows that are zero on the left but nonzero on the right , this signals constraints amongst the @xmath297 whose treatment is postponed until later . for the moment we discard such rows as well rows that are zero on both sides .
let @xmath298 be the number of the remaining rows and @xmath299 the pivot positions and @xmath300 the number of free parameters .
then we can solve in the form @xmath301 . here
@xmath302 is the vector with entry @xmath303 in row @xmath304 and zero otherwise , and @xmath305 consists of the non - pivot columns of the corresponding padded matrix ( its @xmath306-th row is the @xmath307-th row of @xmath308 for pivot indices @xmath309 , and @xmath310 for non - pivot indices @xmath306 ) .
writing the free parameters as @xmath311 , we may substitute this solution @xmath312 into the ansatz to obtain @xmath313 where the @xmath314 are @xmath46-linear combinations of the free parameters @xmath315 and the derivatives @xmath287 comprising @xmath316 .
note that we may regard the @xmath317 as row vectors in @xmath318 whose entries are computed from @xmath308 and @xmath319 .
we turn now to the issue of constraints amongst the @xmath297 , embodied in those rows of the reduced row echelon form of that are zero on the left - hand side but nonzero on the right - hand side .
we put the corresponding block of @xmath319 into reduced row echelon form ( again retaining the same letters for simplicity ) .
let @xmath320 the smallest @xmath226 occurring in any such row .
then each of the rows containing a derivative @xmath321 provides a constraint of the form @xmath322 where the @xmath323 are determined by @xmath319 , and with finite sums over @xmath306 .
let the number of such constraints be @xmath324 .
collecting the @xmath325 into a vector @xmath326 of size @xmath327 , we can write the constraints as matrix equation @xmath328 where the coefficient matrix @xmath329 is determined by the @xmath330 of each constraint , and the right - hand side @xmath331 by the corresponding @xmath332 and the @xmath287 for @xmath333 , which for the moment we regard as known .
note that @xmath334 since @xmath335 is in row echelon form .
let its pivots be in the positions @xmath336 and set @xmath337 .
then we can write the solution as @xmath338 . here
@xmath339 is the vector with entry @xmath340 in row @xmath341 and zero otherwise , while @xmath342 consists of the non - pivot columns of the padded matrix ( its @xmath306-th row is the @xmath307-th row of @xmath335 for pivot indices @xmath343 , and @xmath310 for non - pivot indices @xmath306 )
. writing @xmath344 for the corresponding free parameters , we may thus view as providing @xmath345 constraints @xmath346 where the @xmath347 are @xmath46-linear combinations of the free parameters @xmath348 and certain derivatives @xmath287 that we have collected into a vector @xmath349 .
again we may view @xmath350 as a row vector in @xmath351 whose entries can ultimately be computed from @xmath308 and @xmath319 .
we regard now as determining equations for fixing @xmath345 of the coefficients of @xmath352 . indeed , if @xmath306 is the highest derivative order occurring in we may split according to @xmath353 , with an arbitrary power series @xmath354 , and then substitute this into to obtain @xmath355 for fixing one coefficient of @xmath356 .
now we substitute this @xmath357 back into the above ansatz @xmath358 , and we repeat the whole process for all other constraints , each time determining the lowest unknown coefficient @xmath359 of @xmath356 . of course , it may be necessary to expand the splitting to extract a larger polynomial part ( if all the coefficients in the current polynomial part are determined ) .
eventually , we end up with @xmath360 with some break - off index @xmath361 that is specific to @xmath356 , and with @xmath362 enlarged to comprise also the derivatives @xmath363 that were needed in determining the coefficients . substituting into
, we see that the @xmath364 may be combined with the @xmath365 if we adjoin the parameters @xmath348 to the parameters @xmath315 ; then we can also rename the series @xmath354 back to @xmath356 .
of course new terms @xmath366 may be created in the ansatz , and its polynomial part may be expanded but its overall form is not altered . we have now eliminated those constraints amongst the @xmath367 occurring in that involve @xmath368 , where @xmath320 was chosen minimal .
hence we are only left with constraints amongst the @xmath367 with @xmath369 . repeating the elimination process a finite number of times ( at most @xmath280 eliminations are necessary )
, we obtain the generic form of @xmath260 as given in , where the @xmath297 are now arbitrary ( convergent ) power series . since the terms with distinct factors @xmath370 are clearly in distinct direct sum components @xmath371 , the latter may now be described by @xmath372 where we have set @xmath373 . splitting the coefficients @xmath314 into a @xmath46-linear combination of the free parameters @xmath374 and a @xmath46-linear combination of the derivatives @xmath375 comprising @xmath316
, we collect terms whose coefficient is a specific parameter @xmath376 or a specific derivative @xmath377 .
this leads to @xmath378 and hence to by extracting a @xmath46-basis for each @xmath379 $ ] .
we have now established step 3 of our program since the accessible space @xmath380 can be specified by applying @xmath381 to the generic functions of the admissible space @xmath382 . our next goal is to find a _ projector _
@xmath127 onto @xmath380 , which then gives the exceptional space @xmath383 required for step 4 .
this will be easy once we have a corresponding projector onto @xmath245 .
in fact , the operator @xmath384 in proposition [ prop : char - adspc ] is not quite a projector ( for one thing , it is not even an endomorphism ) , but in a sense it is not far away from being one . for seeing this ,
note that we have @xmath385 where @xmath386 denotes the field of laurent series ( converging in the punctured unit disk ) .
the direct decomposition just reflects the fact that the @xmath370 for distinct @xmath288 are linearly independent .
let us write @xmath387 for the _ indicial projector _ onto the component @xmath388 of .
in other words , @xmath389 extracts all terms of the form @xmath390 from a series in @xmath39 .
note that combinations like @xmath391 provide linear functionals @xmath392 for extracting derivatives of the @xmath226-component of a given series in @xmath39 . for writing the projector corresponding to proposition [ prop : char - adspc ] ,
let us also introduce the _ auxiliary operator _
@xmath393 and its inverse @xmath394 .
then we shall see that the required projector is essentially a `` twisted '' version of two kinds of projector : one for splitting off the polynomial part of the occurring series , and one for imposing the derivative terms . for convenience
, we shall use _ orthogonal projectors _ in @xmath386 , where the underlying inner product is defined by @xmath395 for all @xmath396 .
such projectors are always straightforward to compute ( using linear algebra on complex matrices ) .
[ cor : proj - adspc ] using the same notation as in proposition [ prop : char - adspc ] , let @xmath397 be the orthogonal projectors onto @xmath398 $ ] and onto @xmath399 , respectively .
writing @xmath400 and @xmath401 for their twisted analogs , we define the linear operator @xmath402 by @xmath403 is a projector onto @xmath245 .
the action of @xmath130 on a general series ( which we may split at @xmath404 in its @xmath226-components ) @xmath405 is @xmath406 where we have set @xmath407 .
extracting the @xmath370 component of @xmath408 yields @xmath409 as one sees by comparing with the last displayed equation in the proof of proposition [ prop : char - adspc ] .
hence we may conclude that @xmath410 .
it remains to prove that @xmath411 for @xmath412 since this implies @xmath413 and @xmath414 so that @xmath130 is indeed a projector onto @xmath245 as claimed in the corollary .
so assume @xmath415 is arbitrary , and split it as above .
the orders @xmath220 of the series @xmath416 are given by the curbing constraints .
since @xmath23 already satisfies all boundary conditions in @xmath417 , the reduced row echelon form of will contain only zero rows .
the original ansatz is then left intact ; no expansion of the polynomial part is necessary and no @xmath316 are involved in its coefficients .
therefore the original coefficients @xmath275 in coincide with the coeffients @xmath418 in , which constitute the polynomials of @xmath419 $ ] .
now let us consider .
since the series @xmath420 is thus already in @xmath421 , we may omit the action of @xmath422 and since @xmath423 the triple sum in is zero .
but then @xmath424 becomes identical with @xmath23 as was claimed . for accomplishing step 4 of our program
, it only remains to determine a projector @xmath127 onto the accessible space @xmath112 from the projector @xmath130 onto the admissible space @xmath245 provided in corollory [ cor : proj - adspc ] .
this can be done easily since @xmath127 is _ essentially a conjugate _ of @xmath130 except that we use a fundamental right inverse @xmath131 , for want of a proper inverse .
( the formula for @xmath131 in ( * ? ? ?
23 ) and ( * ? ? ?
20 ) may be used but recall that in our case @xmath425 so that @xmath426 . )
[ prop : proj - acspc ] using the same notation as in proposition [ cor : proj - adspc ] , the operator @xmath427 is a projector onto the accessible space @xmath112 .
let us first check that @xmath127 is a projector .
writing @xmath428 for the projector onto @xmath429 along @xmath430 $ ] and using @xmath414 , we will indeed get @xmath431 provided we can ascertain that @xmath432 since then @xmath433 . we know that @xmath434 since @xmath130 is a projector onto @xmath245 along @xmath435 . on the other hand , we have also @xmath238 since
@xmath436 is regular . intersecting @xmath435 in the former decomposition with the latter yields @xmath437 but @xmath438 implies @xmath439 , so intersecting the decomposition @xmath440 with @xmath441 leads to @xmath442 . using the other decompostion @xmath238 one more time we obtain @xmath443 comparing and
, we can apply the well - known rule ( * ? ? ? * ( 2.6 ) ) to obtain the identity @xmath444 and hence the required inclusion @xmath445 .
it remains to prove that @xmath446 .
the inclusion from left to right is obivous , so assume @xmath447 with @xmath260
. then @xmath448 and hence @xmath449 because @xmath130 projects onto @xmath245 and @xmath433 from the above .
but then we have also @xmath450 and in particular @xmath451 .
we have now sketched how to carry out the four main steps in our program aimed at the algorithmic treatment of finding / imposing `` good '' boundary conditions on a fuchsian differential equation with one ( mild ) singularity . at the moment
we do not have a full implementation of the underlying algorithms in th@xmath0orem@xmath1 ( or any other system ) .
however , we have implemented a prototype version of some portion of this theory .
we shall demonstrate some of its features by with example from engineering mechanics .
circular plates play an important role for many application areas in engineering mechanics and mathematical physics .
if the plates are thin ( the ratio of thickness to diameter is small enough ) , one may employ the well - known kirchhoff - love plate theory @xcite , whose mathematical description is essentially two - dimensional ( via a linear second - order partial differential equation in two independent variables ) .
we will furthermore restrict ourselves to _ circular kirchhoff plates _ so as to have a one - dimensional mathematical model , via a linear ordinary differential equation of second order .
however , we shall not assume homogeneous plates .
indeed , the precise manufacture of _ functionally graded _
materials is an important branch in engineering mechanics . in the case of kirchhoff plates ,
the functional grading is essentially the variable thickness @xmath452 or variable bending rigidity @xmath453 of the plate along its radial profile .
( we write @xmath233 for the radius variable ranging between zero at the center and the outer radius @xmath454 . )
let @xmath455 be the displacement of the plate as a function of its radius .
this is the quantity that we try to determine .
it induces the _ radial and tangential moments _ given , respectively by @xmath456 where @xmath457 is the poisson s ratio of the plate ( which we assume constant ) . for typical materials ,
@xmath457 may be taken as @xmath458 or even @xmath86 .
a reasonable constitutive law for the bending rigidity is @xmath459 where @xmath460 is the variable young s modulus of the plate . the _ equilibrium equation
_ can then be written as @xmath461 where @xmath462 is the cumulative load @xmath463 induced by a certain _ loading
_ @xmath464 that may be thought to describe the weight ( or other forces ) acting in each ring @xmath465 $ ] . for the calculational treatment of
it it useful to introduce the function @xmath466 , which represents the ( negative ) slope of the plate profile . in terms of @xmath467 ,
the equilibrium equation is given by @xmath468 a typical example of a useful thickness grading is the _ linear ansatz _
@xmath469 , cut off beyond some @xmath470 close to @xmath471 ; this describes a radially symmetric pointed plate with straight edges ( more or less a very flat cone ) . suppressing the cut - off for the moment and changing the independent variable @xmath233 to @xmath472
, we have thus thickness @xmath473 and from bending rigidity @xmath474 with @xmath475 .
the equilibrium equation becomes now @xmath476 where we have set @xmath477 for simplicity .
note that the right - hand side of this equation , which we shall designate by @xmath478 , is not a fixed function of @xmath479 but depends on our choice of the loading .
hence we consider @xmath480 as a _ forcing function_. for the _ boundary conditions _
( for once we use this word in its original literal sense ! ) we shall use @xmath481 , which translates to @xmath482 in the @xmath483 formulation , with the abbreviation @xmath484 . physically speaking
, this corresponds to a plate that is clamped in the center and left free at its periphery ( this comes from translating suitable boundary conditions for the displacement @xmath455 and taking the appropriate limits ) . in summary
, we have the boundary problem @xmath485 which is indeed of the type discussed in section [ sec : bp - one - sing ] , by a simple scaling from @xmath28 $ ] to @xmath486 $ ] .
its treatment in the ` greengroebner ` package proceeds as follows [ we apologize for the lousy graphics rendering
we plan to fix this as soon as possible ] : the output is the green s operator @xmath192 of applied to a generic forcing function @xmath478 , giving the solution @xmath487 as an integral @xmath488 in terms of the green s function @xmath489 ; the latter can also be retrieved explicitly if this is desired ( note that the expression is clipped on the right - hand side so the two case conditions @xmath490 and @xmath491 labelling the two lines are not visible ) : let us now choose a constant loading @xmath492 . taking the cut - off into account , this leads to the forcing function @xmath493 . in this case , we can compute the solution @xmath494 of explicitly .
for definiteness , let us choose a cut - off at @xmath495 .
in that case , we obtain @xmath496 and the corresponding displacement @xmath497 is given by @xmath498 of course one could use different functional gradings and/or loading functions , and the integrals would not always come out in closed form . in this case
one can resort to numerical integration ( which is also supported by _ mathematica _ ) .
bruno buchberger , adrian craciun , tudor jebelean , laura kovacs , temur kutsia , koji nakagawa , florina piroi , nikolaj popov , judit robu , markus rosenkranz , and wolfgang windsteiger .
theorema : towards computer - aided mathematical theory exploration .
, 4(4):359652 , december 2006 .
issn 1570 - 8683 .
anja korporal , georg regensburger , and markus rosenkranz .
regular and singular boundary problems in maple . in _ proceedings of the 13th international workshop on computer algebra in scientific computing , casc2011 ( kassel , germany , september 5 - 9 , 2011 ) _ , volume 6885 of _ lecture notes in computer science_. springer , 2011 .
markus rosenkranz , georg regensburger , loredana tec , and bruno buchberger .
symbolic analysis of boundary problems : from rewriting to parametrized grbner bases . in ulrich langer and
peter paule , editors , _ numerical and symbolic scientific computing : progress and prospects _ , pages 273331 .
springer , 2012 .
|
we develop a new theory for treating boundary problems for linear ordinary differential equations whose fundamental system may have a singularity at one of the two endpoints of the given interval .
our treatment follows an algebraic approach , with ( partial ) implementation in the theorema software system ( which is based on mathematica ) .
we study an application to graded kirchhoff plates for illustrating a typical case of such boundary problems .
boundary problems , singular boundary problems , generalized green s operator , green s functions , integro - differential operators , ordinary differential equations , kirchhoff plates , functionally graded materials
|
beside simple metallic systems found with b.c.c and f.c.c structures , with few atoms per unit cells , many intermetallic alloys display more or less intricate structures with polytetrahedral type of packing , and possibly a large number of atoms in the unit cell , like in the so - called frank - kasper ( f - k ) phases@xcite . it is standard in that case to focus on the local atomic environments , the canonical @xmath2 cells ( see fig . [
fig : f1 ] ) , the case @xmath4 corresponding to a local icosahedral environment .
a long - standing question in quasicrystal studies is to relate the atomic order present in quasicrystalline phases with that of f - k phases , some of which being identified as rational approximants of the former ; dually , these quasicrystalline phases ( close in the phase diagram ) could then be thought as mere quasperiodic generalizations of the periodic f - k phases .
the four main frank - kasper coordination polyhedra ( top figure ) @xmath7 , @xmath8 , @xmath9 and @xmath6 , represented also by their voronoi cells ( lower figures ) . frank - kasper lines ( or disclination lines ) , shown as black lines , start from the center of the coordination polyhedra and run through the black sites , or from the voronoi cell centres and go through hexagonal faces , defining a `` major skeleton '' in the structure . ]
as will be detailed below , f - k structures can also be generated with an atomic decoration procedure applied on top of a plane tiling made of triangles and/or squares , building simple atomic layers .
to go further , one should then understand how the f - k construction rules can be accommodated with an underlying quasiperiodic tiling ; it was therefore natural to investigate how well dodecagonal quasiperiodic tilings , made of squares and ( equilateral ) triangles , can be used as building frames to generate some f - k - like quasicrystalline structures .
this is the main task of the work presented here .
note that such tilings have been used to model dodecagonal quasicrystals found in metallic alloys related to frank - kasper phases @xcite and , more recently , in dendrimeric supramolecular liquid crystals @xcite .
we show here how to produce two types of such structures , which are quasiperiodic in a plane and periodic in the third direction , and containing ( or not ) @xmath6 sites ( sites with 16 neighbours ) . in a first step ,
we detail the f - k rules for periodic systems .
four parallel atomic layers are disposed with respect to a periodic planar frame made of triangle and squares .
the occurrence of @xmath2 sites and periodic disclinated networks are direct consequences of these rules . in order to generate quasicrystalline f - k - like structures ,
we then start with a particular tiling with square and triangles , the dodecagonal square - triangle quasiperiodic tiling .
we first recall a standard method to produce the 12-fold symmetric quasiperiodic tiling , the inflation / deflation approach ( see ref @xcite ) , for which we propose a new simplified algorithm .
notice that a standard cut and project approach from a higher dimensional space is also possible , but operationally complex because the acceptance domain ( used to select the tiling points ) has a fractal shape in that case .
we then return to the question of atomic decorations leading to close - packed configurations , and extend the f - k rules to this wider quasiperiodic domain .
a particular such decoration , corresponding to the type of atomic order found in the crystalline f - k @xmath10 phase , was already used by gahler @xcite to model the first dodecagonal quasicrystalline alloys found in ni - cr , v - ni and v - si - ni alloys . notice a slight difference in that case since the underlying quasiperiodic tiling contained thin rhombuses in addition to the square and triangles ( in that case the acceptance domain is a regular dodecagon instead of a fractal shape , making the construction simpler ) .
see also ref .
@xcite for further developments . as for the periodic case ,
four atomic layers are generated , with atomic positional order following that of the underlying quasiperiodic tiling , these quasicrystalline atomic layers being then repeated periodically . in the present work , we discuss two different decoration schemes to be applied on the square - triangle quasiperiodic tiling , leading to two types of @xmath9 sites ( in - plane or connecting different planes ) as well as a variable amount of @xmath6 sites . beside the perfect 12-fold quasiperiodic tiling , a rich set of square - triangle tilings
can also be derived , with either lower ( hexagonal ) symmetry or even forming random tilings . from any such tiling
, an atomic decoration procedure could be applied , locally inspired by the f - k rules , leading to a large set of tetrahedrally close packed ( tcp ) structures . among these generalized structures
, we shall discuss an interesting new set which can interpolate between the f - k @xmath11 phase , whose underlying tiling is a triangular one , and the @xmath12 phase with a square underlying tiling .
this set is periodic along two directions , with , along the third one , a sequence of triangle and square rows which can take any form ( periodic , quasiperiodic , disordered ) .
an explicit quasiperiodic example is described in the appendix .
f - k structures are tetrahedrally close - packed periodic structures , first described by frank and kasper @xcite , and later studied extensively by many authors , among which notably d. and c. shoemaker @xcite .
the atoms occupy ( so - called ) @xmath2 sites , where @xmath3 , the site coordination number , is equal to either @xmath0 or @xmath1 ( fig .
[ fig : f1 ] ) , and have interpenetrating triangulated coordination polyhedra ; the @xmath2 sites voronoi cells have @xmath13 pentagons ( the @xmath7 voronoi polyhedron is a pentagonal dodecahedron ) and @xmath14 hexagons . when icosahedral quasicrystals were discovered in intermetallic compounds ,
the large cell structures found nearby in the phase diagram became of parallel interest , as they were now viewed as periodic approximant of the perfect quasicrystalline structure .
the metallurgist community quickly considered that frank - kasper - like phases @xcite or related phases ( like @xmath15-mn or cacu@xmath16 for instance ) would be good models for these large cell periodic structures .
indeed , they already present a high degree of local and medium range icosahedral configurations , like for instance in the bergman structure @xcite .
quite rapidly , decagonal and dodecagonal quasicrystalline structures were also discovered and further related to f - k structures .
more recently , dodecagonal quasicrystals have also been found in soft matter systems , like in dendrimeric supramolecular liquid crystals @xcite , a particularly interesting discovery showing that metallic bonding is not a prerequisite for stable quasiperiodic order .
note in addition that f - k phases were also observed in micellar structures @xcite .
-free structures ( left part ) , or with @xmath6 sites ( right part ) , for the three types of tile interface : triangle - triangle ( a , d ) , triangle - square ( b , e ) and square - square ( c , f ) .
let us first describe the left part ( a , b , c ) .
sites located on the secondary layers at height @xmath17 and @xmath18 sit at the tile vertices and are shown in grey .
we aim to discuss the decoration of primary layer ( at height @xmath19 ) with atoms located inside the triangular and/or square prisms formed by secondary layer sites . as explained in the text ,
there are two sets of edges ( marked as dark or light - grey edges ) , with respect to their angular value ( modulo @xmath20 ) . edges of a given triangle will belong to one of the two sets , while adjacent edges of a square will alternate among the two sets .
there are two different primary layer decoration rules ( called rules @xmath21 and @xmath22 ) , at the vertical of a triangle of the secondary layer , depending on the triangle colour , and whether the primary layer is at height @xmath23 or @xmath19 .
this point will appear more clearly in fig .
[ fig : decoz15 ] .
the figure displays rule @xmath21 , on the primary layer at height @xmath19 , with atoms inside triangular prisms ( with the triangles being of light - grey colour ) , or square prisms .
these atoms are shown in blue ( colour on line ) , with the piece of disclination network to which they belong .
they fall in the middle of the triangles , in the middle of the dark grey edges and displaced from the middle of the light grey edges by 1/4 of the square edge length .
rule @xmath22 applies when the triangles have dark edges , in which case the decoration adds atomic sites on top of the edge centres .
sites located on primary layers at height @xmath23 are also shown ( in red with colour on - line ) .
note that these two rules for triangle decoration drive the decoration on top of the squares , with atomic sites on top of dark edges and slightly displaced for the light ones .
sites at height @xmath19 in the triangular prism middle are @xmath9 sites , connected to @xmath8 sites through a triangle - square interface .
sites falling on the dark grey mid - edge are @xmath7 sites .
finally , grey sites on the secondary layers are all @xmath8 sites , connected by vertical disclination lines .
now , for the primary layer at height @xmath23 , the decoration is similar , but with rules @xmath21 and @xmath22 reversed with respect to the edge colouring .
let us turn now to the layer decoration with @xmath6 sites , shown on the right part ( d , e , f ) . as said in the text ,
some edges are now `` double edges '' separating stripes . across these double edges ,
rules @xmath21 and @xmath22 are reversed .
this leads to connecting the disclination lines between two primary layers ( in blue and red on - line ) .
a formerly @xmath9 ( resp . a @xmath8 ) site is turned into a @xmath6 ( resp .
a @xmath9 ) site.,width=453 ] frank and kasper showed that the structure complexity can be encoded by a network of lines , called by them the `` major skeleton '' , avoiding the @xmath7 sites , and joining the other atomic sites through the voronoi cells hexagonal faces .
the edges of these networks are precisely those edges in the full structure simplicial decomposition sharing six tetrahedral cells . as a whole
, the edge set of a f - k structure is therefore a subtle mixture of edges sharing five tetrahedra , and this network of edges sharing six tetrahedra .
f - k phases provide an intermediate solution for the frustrated problem of sphere packing maximal compactness .
best packing are well known to be given by fcc or hcp packings , with interstices not limited to tetrahedral ones , while a search for a better , but only local , solution would invite to attempt for a maximization of tetrahedral interstices .
the difficulty comes from the fact that a perfect space tiling by regular tetrahedra is impossible , since the dihedral angle of a regular tetrahedron ( @xmath24 ) is not a submultiple @xmath25 , although being close to @xmath26 ) : the problem is said to be geometrically frustrated .
icosahedral local order is naturally generated in polytetrahedral packing , but a perfect packing of regular tetrahedra is only possible in a positively curved space @xmath27 @xcite .
a polytetrahedral packing in @xmath28 will necessarily contain ( curvature carrying ) topological defects , like disclination lines , corresponding to regions in space where the local order differs from a perfect icosahedral one .
starting with an edge sharing five tetrahedra , the simplest disclination line along this edge will place six tetrahedra around this edge .
this corresponds to the f - k major skeleton rules , the latter being therefore identified as such a disclination network@xcite .
this network has two types of sites : ( i ) edge sites , made of @xmath8 polyhedra , threaded by the edge through opposite points , and ( ii ) vertex sites , where disclinated edges meet by three at @xmath29 on @xmath9 sites , of by four at a tetrahedral angle on @xmath6 sites .
one can show that these disclination lines can not be interrupted in the structure : they run throughout the volume , and possibly connect to other lines .
one can distinguish several families of frank - kasper phases .
a first main difference refers to whether or not atoms can be gathered into simple planes ( called , in the f - k terminology , primary layers " , tiled with triangles , hexagons or pentagons , and `` secondary layers '' , with squares and/or triangles ) , see for instance @xcite@xcite .
we shall describe here the atomic positions , and disclination network , by focusing on the three types of tile interfaces in the secondary layers , either triangle - triangle , square - square and triangle - square ( see fig .
[ fig : discli_z15_z16 ] ) . a nice example of a non - layered f - k phase is provided by the bergman structure @xcite , which shows , around symmetrical sites , an interesting nested sequence of clusters with icosahedral symmetry . among the layered f - k phases ,
another criterion distinguishes between principal layers containing only hexagons and triangles ( like @xmath12 , @xmath11 and @xmath10 phases @xcite ) , and those in which there are also pentagons ( leading to the presence of @xmath6 sites , as discussed below , like in the @xmath30 and @xmath31 laves phases @xcite .
we shall mainly be interested here to extend these f - k layered phases to quasiperiodic order .
atomic decoration of the different layers , illustrated for some classical frank - kasper phases , along the two primary layers " at heights @xmath23 and @xmath19 and the two `` secondary layers '' at heights @xmath17 and @xmath18 , the latter two carrying parallel tilings by squares and triangles ( or even solely squares or triangles ) and being such that atoms are placed at the tile vertices . as explained in the text , there are two sets of edges ( marked as dark and light grey edges ) , with respect to their angular value ( modulo @xmath20 ) . atomic positions on the secondary layers are shown as black dots and atomic positions on the primary layer at height @xmath19 are represented as small blue dots . as a whole ,
the atomic positions on this primary layer at height @xmath19 are vertices of a tiling by hexagons and triangles , with the hexagon centres on top of the atomic positions of the secondary layer at height @xmath17 .
the situation is similar on the other primary layer at height @xmath23 , but with decoration rules being reversed .
the atomic positions on this layer are marked with red open circles .
the full structure is generated as a periodic repetition of these four layers ( the height are given in terms of the square and triangle edge lengths ) .
, width=453 ] let us now describe more in details the atomic positions in the four equally spaced different layers , which are then periodically repeated .
two `` primary '' layers are positionned at height @xmath23 and @xmath19 , and two `` secondary '' layers at height at height @xmath17 and @xmath18 .
it is convenient , following j. sullivan@xcite , to start by considering a secondary layer , say at height @xmath17 , which is a tiling of squares and/or equilateral triangles ( the above heigths are given in terms of the tiles edge length ) .
this tiling plays a central role since it determines the whole atomic structure , up to a decoration choice that will allow or not the presence of @xmath6 sites .
in both cases , atoms ( or micelles ) are located at the vertices of the tiles in this secondary layer , as well as for the parallel other secondary layer located at height @xmath18 .
the atomic decoration of the two primary layers is case dependant .
finally , the full structure is obtained as a periodic repetition of this four - layer stacking .
the decoration details are given in fig .
[ fig : discli_z15_z16]-a ,- b ,- c , and in the caption , because it is more easily understood with an eye on the drawing .
let us make the following remark on the edges of a square - triangle tiling , which help understanding the decoration procedure : with edges orientation measured with respect to one arbitrary edge , all remaining edges fall into two sets regarding their orientation modulo @xmath20 .
edges of a given triangle will belong to one of the two sets , while adjacent edges of a square will alternate among the two sets .
the two types of edges are marked respectively with dark and light colours in fig . [
fig : decoz15 ] .
primary layers atomic decoration depends on the underlying edge and triangle colouring on the secondary layer , which we shall therefore call decoration rules @xmath21 and @xmath22 . in summary ,
the atomic decoration add atoms on each layer .
those belonging to the two parallel secondary layers sit on the square - triangle tiling vertices .
those on the primary layers eventually form horizontal tilings of triangles and hexagons . with these decorations rules , many well known f - k phases
are recovered , as illustrated in fig .
[ fig : decoz15 ] for the so - called @xmath12 , @xmath32 , @xmath11 and @xmath10 phases .
a site on a secondary layer has six first - neighbours on each of its two surrounding primary layers , plus two additional neighbors located on its two neighbouring secondary layers at distance @xmath19 ; so this point has coordination @xmath33 .
these @xmath8 sites lie on vertical disclination lines orthogonal to the layers .
sites on primary layers falling , by projection , onto ( triangle or square ) edge centres ( smaller blue dots on dark gray edges or open circles on light gray edges ) have an icosahedral coordination ( @xmath7 ) .
points falling onto triangle centers have @xmath34 ; these @xmath9 sites belong to three half disclination lines lying on a primary plane .
finally , the atomic sites falling entirely inside squares are @xmath8 sites located on horizontal disclination lines .
since this decoration is unique , it is possible to compute the mean coordination number and express the phase composition in terms of the different types of sites . for that purpose
, we express a three - dimensional vector @xmath35 with components @xmath36 , the number of @xmath7 , @xmath8 and @xmath9 sites in terms of a vector @xmath37 with components the number of squares and triangles . the above described atomic decoration ( see fig . [
fig : discli_z15_z16 ] ) can be encoded in the following relation , with a rectangular matrix : @xmath38 with @xmath39 , the average coordination number for the decorated structures reads : @xmath40 .
the @xmath12 phase has only squares in the secondary layer ( therefore with infinite @xmath41 ) , leading to @xmath42 .
the @xmath11 phase has only triangles ( @xmath43 ) , leading to @xmath44 , and the @xmath10 phase corresponds to @xmath45 , leading to @xmath46 decoration of layers in some frank - kasper phases containing @xmath47 sites . as explained in the text , the triangle - square tiling now carry stripes delimited by double edges , and forming closed loops or ending at the boundary . for a given square - triangle tiling
, there are many ways to draw these stripes , even if one aims to maximize their numbers ( or , said differently , to minimize their individual width ) .
one point should be stressed here : while for the @xmath6-free f.k .
layered structures , the secondary layer tiling is done with perfect squares and equilateral triangles , the atomic positions in the present case will relax in such a way that squares will often get distorted into rectangles and triangles turned into being simply isosceles .
notice finally that , with this decoration , sites at the triangle - square tiling vertices on the secondary layers are now @xmath7 sites when belonging to double edges ( which therefore concerns all these sites when a dense set of double edges is used ) .
, width=453 ] we now describe a more complex decoration procedure which allows for the existence of @xmath6 sites .
when the former procedure gave a unique atomic decoration on top of a square - triangle tiling , the present one leads to a large number of possible atomic decorations on top of a given triangle - square tiling . while in the @xmath6-free case , the disclination segments are either in or orthogonal to the layers , the main idea here consists in introducing disclination segments connecting the layers . in order to do that
, one defines stripes covering the triangle - square tiling , and delimited by edges which we shall call ( and draw as , in the figures ) double edges @xcite .
the rule to follow is that no triangle should have two double edges and no squares two consecutive double edges .
the next step is to decide that the previous decoration rules ( @xmath21 and @xmath22 ) associated with coloured triangles , is now flipped while crossing neighbouring stripes . as shown on fig .
[ fig : discli_z15_z16]-c ,- d ,- e , this leads , compared to the @xmath6-free case , to @xmath9 ( resp .
@xmath8 ) site changed into @xmath6 ( resp .
@xmath9 ) sites . with these decorations rules ,
other well - known f - k phases are recovered , as illustrated in fig .
[ fig : decoz16 ] for the @xmath48 , and @xmath30 phases .
note that , although the decoration procedure was done on top of a tiling with equilateral triangles and squares , the actual atomic positions in these f - k phases display slightly distorted tiles , but without changing the above described disclination network topology .
when the double edge limited stripes are of minimal width , it is again possible to compute the mean coordination number and express the phase composition in terms of the different types of sites .
we express a four - dimensional vector @xmath35 with components @xmath49 in terms of the vector @xmath37 with components the number of squares and triangles .
the above atomic decoration is now encoded in the following relation , with a rectangular matrix : @xmath50 with @xmath39 , the average coordination number for the decorated structures reads now : @xmath51 .
this formula applies to the @xmath6-containing frank - kasper phases . as an example , laves phase have only triangles in the secondary layer ( therefore with infinite @xmath41 ) , leading to @xmath52 .
an interesting case is provided by the z phase .
we have seen above that it can be obtained by decoration on top of a triangular tiling , leading to a structure with @xmath7 , @xmath8 and @xmath9 sites . but this phase can also be generated using the present rule , applied on an underlying square tiling ( in which case only @xmath9 sites are generated accross double edges ) . with @xmath41 infinite in the previous formula ,
the z phase @xmath53 averaged coordination number is recovered .
dodecagonal square - triangle quasiperiodic tilings , first discussed by stampfli @xcite , can be generated along two standard methods : \(i ) tilings with self - similar symmetry which can be constructed directly in two dimensions , with a step by step division ( called inflation / deflation method ) of the basic tiles into smaller ( but similar ) tiles , followed by a global rescaling .
the quasiperiodic tiling is obtained by repeating the inflation and decomposition process ad infinitum .
we use here this inflation scheme , for which a decoration procedure has initially been given by schlottmann , and detailed for instance in ref.@xcite , in terms of five basic proto - tile decorations , and for which we propose below a simplified algorithm leading to the same tilings .
in addition , an even simpler rule in that procedure will allow to generate a six - fold symmetric quasiperiodic tiling .
\(ii ) the `` cut and project '' approach .
one starts from a periodic structure defined in a higher @xmath54dimensional space , with a suitable decomposition into a physical " @xmath55dimensional space and a @xmath55dimensional perpendicular " space ; a selection step ( selection of points projected into an acceptance domain " in the perpendicular space " ) is then processed followed by a projection onto the physical space .
this method is quite generic , allowing for example a simple computation of the fourier space structure factor , and a natural definition of the large - cell periodic approximants .
however the cut and project method is here operationnally quite complex because , in the perpendicular space , the acceptance domain has a fractal shape @xcite .
we shall therefore not explicitely use it here , except when showing the acceptance domains in the figures .
iterative decoration of tiles for a self similar dodecagonal triangle - square tiling : ( a ) triangle decoration ; ( b ) square decoration ; ( c ) point set obtained after decoration , forming a tiling of dodecagons ( surrounding the vertices of the previous iteration ) , squares and triangles ( only vertices are represented ) ; there are two possible decorations for the docedagons , as shown in the drawing ; ( d ) the rule for choosing the dodecagon decoration . at a given step in the 12-fold symmetric tiling
, there is a small set of site environments . in order to decorate a given dodecagon : ( i)go back to the site located , in the previous iteration , at the dodecagon centre , ( ii ) determine the edge orientation ( modulo @xmath20 which is dominant ( this is unambiguous ) , ( iii ) use this orientation for the internal triangular shell inside the dodecagon
this is done easilly , except for some points on the boundary for which this counting can not be done due to missing edges .
but the dodecagon to be filled on the boundary is similar to an ( already decorated ) interior dodecagon , to which it is related by the tiling self - similarity .
the rule is then to get the missing information from this interior dodecagon . ]
we are interested here in generating a 12-fold symmetric triangle - square tiling with a self - similar property around its centre .
for that purpose we propose an alternative decoration method , which we believe operationally simple , leading to the same tiling as if obtained with the more standard method proposed by schlottman .
for that purpose we start from a particular seed , with 6 equilateral triangles around the origin . a simple decoration rule , applied to the triangles ( and later to the squares when they will be generated ) is shown on fig . [
fig : dodecarule]-a ,- b , leading at the next step to a new tiling by dodecagons , squares and triangles with an edge length reduced by a factor @xmath56 .
the next step , to get a tiling with only triangle and squares , is non trivial .
each dodecagon will be decorated by a first shell made of 6 triangles ( an orientation choice is to be made here ) , and , as a consequence , by a second shell where squares and triangles alternate .
in fact , we can adopt here different strategies , leading to either a disordered random tiling , a 6-fold symmetric tiling ( discussed below ) , or have a precise rule leading to a perfectly self similar 12-fold symmetric tiling . in that case ,
the orientation degeneracy is lifted by a simple rule , exposed in fig .
[ fig : dodecarule]-d , which amounts to a kind of majority rule applied on the edge orientations before decoration .
there are five or six edges meeting at a vertex at iteration @xmath57 . among these edges
, at least three have the same direction modulo @xmath58 .
the rule is then that the decorated dodecagon surrounding this vertex at the next iteration is decorated such as to have its six central edges oriented with this direction modulo @xmath20 .
there remains ambiguities for sites on the boundary of the construction , which are solved by taking into account the tiling self - similarity .
looking at the decoration procedure , and introducing a vector @xmath59 whose components are the number of squares and the number of triangles at iteration @xmath57 , the vector @xmath60 is obtained from a transfer matrix as : @xmath61 some informations about the asymptotic tiling are readily obtained from the perron root of the matrix , @xmath62 , and its associated eigenvector @xmath63 , which give the asymptotic ratio @xmath64 between the number of squares and triangles ( which already rules out the possibility for this tiling to be periodic ) .
notice however that the area covered by squares equals that covered by triangles .
a ) a piece of a 12-gonal symmetric square - triangle tiling obtained upon two iterations from a dodecagonal seed ; b ) the associated fractal - like acceptance domain if constructed instead by a cut and project algorithm , showing at this step a 6-fold summetry ; c ) the asymptotic 12-fold symmetric fractal domain reached after a third iteration ] figure [ fig : dodec_tiling_and_za ] represents the tiling obtained using this rule starting from a decorated dodecagon after two iterations .
also shown is the approximative acceptance domain ( in the cut and project approach ) that would be reached if this tiling had been constructed along this method . at this step
the symmetry is only 6-fold ( it is strictly 6-fold in the tiling ) .
but upon iteration , the acceptance domain turns into a nice fractal shape with 12-fold symmetry .
frank - kasper decoration of a dodecagonal quasicrystal .
network of disclinations are represented inside two layers , respectively in red and blue ( on - line ) .
there are also disclinations ( not drawn here ) orthogonal to the layers , threading grey sites located at the triangle - square tiling vertices . ]
once the quasiperiodic 12-fold symmetric triangle - square tiling has been constructed , it becomes a simple exercise to use it as a template and build a generalized quasicrystalline f.k .
phase , which is quasiperiodic along the layers , and periodic along the third direction . for the @xmath6-free structure ,
the obtained structure is unique , and is displayed on fig.[fig : dodecfkz15 ] , together with a piece of the disclination network ( this structure was already briefly discussed in @xcite ) .
the disclination network is made of several disconnected parts .
sites on the secondary layers are @xmath8 sites located at the triangle - square tiling vertex positions , leading therefore to vertical disclinations with this geometry .
each primary layer contains a disclination network having the form of a particular hexagonal tiling with @xmath9 sites at its nodes and @xmath8 sites , grouped by pairs , along some ( long ) edges .
sites in the primary layer at height @xmath19 are shown in blue ( colour on - line ) in fig.[fig : dodecfkz15 ] , with those ( blue ) sites not belonging to the disclination network being @xmath7 sites .
the situation is similar on the primary layer at height @xmath23 , with atomic sites shown in red . from the above
computed perron eigenvector associated with the quasiperiodic triangle - square decoration , one gets the asymptotic square to triangle ratio , @xmath65 .
the average coordination for the quasicrystal is therefore @xmath66 and we can extract an ( unnormalized ) composition , in terms of the coordination numbers : @xmath67 with @xmath68 , @xmath69 and @xmath70 .
frank - kasper decoration of a dodecagonal quasicrystal with the rule inducing @xmath6 sites .
the decoration consists first in drawing stripes delimited by double edges , where the decoration rules swap and the disclination networks connect from a layer to another .
this decoration is not unique ; we show here an example with a dense array of concentric stripes .
the drawing does not show the @xmath7 sites on the primary layers and therefore only displays sites on the disclination networks on these layers : the @xmath8 sites , grey , in the secondary layers , and @xmath8 and @xmath9 sites in the primary layers .
grey sites sites on the secondary layers are either @xmath7 sites if belonging to double edges , or @xmath8 sites on simple edges .
disclination networks are drawn in red ( resp . in blue ) , colour on - line , for the primary layer at height @xmath23 ( resp .
@xmath19 ) .
disclination segments crossing double edges ( and connecting the two primary layers ) appear in purple ( colour on - line ) . ]
let us now describe quasiperiodic structures containing @xmath6 sites . as discussed above in the crystalline case ,
the rule consists in defining stripes delimited by double edges , which form the locus where disclination networks connect between different layers .
but , while simple dense periodic arrays of stripes were easy to define in the former case , this task is more complex here , and there are a priori multiple solutions .
here , we display one such solution which consists in having a dense array of concentric stripes around the tiling centre ( see fig . [
fig : dodecfkz16 ] ) .
triangle - square tilings are known to provide very interesting templates for frank - kasper tetrahedrally close packed structures .
we have given a detailed description of the decoration procedure which connect them , and focused in addition on the resulting disclination network geometry .
two types of decorations are discussed , leading , or not , to structures with @xmath6 sites .
the frank - kasper construction can be generalized to quasicrystalline order , as discussed in the second part of this paper for compact structures based on the 12-fold symmetric triangle - square tiling .
this is quite interesting in view of a renewed interest due to the experimental finding of 12-gonal quasicrystalline soft - matter systems .
we have described how to generate a self - similar such tiling , as well as ( in the appendices ) a related 6-fold symmetric tiling , and a square - triangle tiling which is quasiperiodic in one direction and periodic in the other . in all cases ,
the decorations rules discussed in the periodic case can be used and produce quasicrystalline f - k phases with , as expected , a quasiperiodic disclination network .
it is well known that topological disorder can be introduced in square - triangle tilings , via so - called zipper moves " @xcite .
it would certainly be interesting to study these disordered f - k phases more in details j - f sadoc is pleased to thank m. imperor , a. jaganathan and b. pansu for fruitful discussions ( a ) square - triangle 6-fold symmetric quasiperiodic tiling , with all dodecagon decoration equally oriented ( first hierarchical step on top of a decorated dodecagon ) .
this tiling has an exact 6-fold symmetry axis .
( b ) acceptance domain for the same tiling but after a third hierarchical step ; the boundary has a fractal shape . ]
in order to construct the above 12-fold symmetric square - triangle tiling , we have applied an iterative decoration procedure inside the dodecagon generated at each step , with a rule to decide among the two possible orientations modulo @xmath71 of the six central triangles .
but we can also decide to decorate these dodecagons with a unique orientation . as a consequence ,
a six - fold symmetric quasiperiodic tiling is generated , as displayed on the figure [ fig : quasi6fold]-a with one direction choice .
interestingly , the acceptance domain , shown on figure [ fig : quasi6fold]-b , is less complex than that used for the 12-fold case , but still has a fractal boundary ( whose dimension can be calculated in that case ) . a tiling by rows of squares and triangles , which is quasiperiodic in the direction perpendicular to the rows .
it is obtained by a substitution rule on two letters @xmath72 and @xmath73 , respectively for rows of squares and triangles , which reads : @xmath74 and @xmath75 , corresponding to @xmath76 in the matrix given in the text . recall that this matrix has been chosen in order to have an equal area covered by both squares and triangles , a characteristics which already rules out periodicity .
the figure corresponds to three iterations starting from a row of triangles and a row of squares , schematized on the left.,width=377 ]
interesting new structures can be defined , which interpolate between the f - k @xmath11 phase , whose underlying tiling is a triangular one , and the @xmath12 phase with a square underlying tiling .
it amounts to have linear rows of squares and of triangles , which alternate with any sequence and relative densities .
this set is periodic along two directions , with , along the third one , a type of order given by the considered sequence , which can take any form ( periodic , quasiperiodic , disordered ) .
we now describe an explicit such square - triangle tiling , periodic in one direction and quasiperiodic in the orthogonal direction , which also shares with the above 12-fold symmetric tiling the characteristics of equal area covered by the squares and the triangles .
we consider a structure made of two types of rows : an infinite row of squares ( with unit edge length ) , labelled @xmath72 , and an infinite row of triangles , of width @xmath77 , labelled @xmath78 .
we define a substitution rule on the two letters @xmath72 and @xmath73 , such that a square row is replaced by @xmath79 square rows and @xmath80 triangle rows , and a triangle row is replaced by @xmath81 square rows and @xmath79 triangle rows .
this is encoded in the following transfer matrix form to compute the vector @xmath82 , with @xmath83 ( resp .
@xmath84 ) the number of square ( resp .
triangle ) rows at the @xmath85 iteration : @xmath86 to the perron root ( @xmath87 ) of this matrix corresponds an eigenvector @xmath88 ; from this , we get an equal area covered by triangles and squares , for any value of the integer @xmath79 , and non - periodicity in the direction orthogonal to rows resulting from the irrationality of @xmath89 .
a simple example is shown figure [ fig : quasi1d ] corresponding to @xmath76 .
a f - k decoration then leads to a structure which can be viewed as a quasiperiodic mixing of the @xmath12 and the @xmath11 phase .
notice that @xmath90 differs from the above @xmath37 in that it counts row frequencies instead of tile frequencies .
a triangle row contains alternating triangles of opposite directions , and twice as many triangles as squares , per unit row length . taking this into account
, one recovers the same coordination number @xmath66 as for the above 12-fold f - k quasicrystal .
notice that it is also possible , from the rows , to simply define double edges stripes in such a way as to produce @xmath6 sites . in that case , we obtain , using the expression detailed above , an average coordination @xmath91 , quite close to the celebrated value for the coxeter statistical honeycomb " @xcite .
10 sullivan j. m. , in n. rivier and j. f. sadoc , editors , `` foams and emulsions '' , * 354 * of nato advanced science institute series e : applied sciences , kluwer,(1998 ) 379 - 402 . or proc .
eurofoam ( delft ) , 111 - 119 ( 2000 ) .
|
frank - kasper ( f - k ) phases form an important set of large - cell crystalline structures describing many inter - metallic alloys . they are usually described in term of their atomic environments , with atoms having @xmath0 and @xmath1 neighbours , coded into the canonical @xmath2 cells ( with @xmath3 the coordination number ) , the case @xmath4 corresponding to a local icosahedral environment . in addition , the long range structure is captured by the geometry of a network ( called either `` major skeleton '' or `` disclination network '' ) connecting only the non - icosahedral sites ( with @xmath5 ) .
another interesting description , valid for the so - called `` layered f - k phases '' , amounts to give simple rules to decorate specific periodic 2d tilings made of triangles and squares and eventually get the 3d periodic f - k phases .
quasicrystalline phases can sometime be found in the vicinity , in the phase diagram , of the f - k crystalline alloys ; it is therefore of interest to understand if and how the standard f - k construction rules can be generalized on top of an underlying quasiperiodic structure .
it is in particular natural to investigate how well square - triangle quasiperiodic tilings with dodecagonal symmetry , made of square and ( equilateral ) triangles , can be used as building frames to generate some f - k - like quasicrystalline structures .
we show here how to produce two types of such structures , which are quasiperiodic in a plane and periodic in the third direction , and containing ( or not ) @xmath6 sites .
_ it is a pleasure for us to contribute to a special issue of `` structural chemistry '' in honour of alan mackay , for his 90th birthay .
alan s work has been a source of inspiration for most of the researchers working in the field of complex atomic structures . _ to appear in `` structural chemistry '' .
|
the characterisation of exoplanet atmospheres , particularly with ground based telescopes , is a difficult task .
polarimetry offers a useful approach , providing a strong contrast between the star and planet as the star s light is typically unpolarized @xcite .
large and tightly orbiting hot jupiter planets scatter enough light to potentially produce a polarisation signal dependent largely upon the composition of the atmosphere @xcite .
the detection of linearly polarized light from an exoplanet system can provide information about a planet s orbital orientation and about the properties of the particles that scatter the light in its atmosphere .
the technique therefore provides complementary information to other characterisation techniques such as transit and eclipse spectroscopy .
@xcite modelled the expected polarization levels for hot jupiter - type systems and predicted that linear polarization varying over the orbital cycle at the tens of parts - per - million level might be present in the combined light of the star and planet .
since sun - like stars are thought to generally have low polarizations @xcite this technique provides good contrast and should be achievable from ground - based observations .
past attempts at polarized light detection such as those described in @xcite have not detected significant planetary polarization signals .
the hd 189733 hot jupiter system @xcite is one of the brightest and best studied transiting exoplanet systems .
observations of transits and eclipses of hd 189733b using the hubble space telescope have shown strong evidence for a rayleigh scattering haze in its atmosphere , which shows up as increasing absorption towards shorter wavelengths in the transit spectrum @xcite as well as reflected light seen in the secondary eclipse at blue wavelengths @xcite .
this makes hd 189733b a promising target for polarimetric observations .
however , polarization observations of this system to date have produced conflicting results .
@xcite reported an unexpectedly high level of polarization variation of @xmath1200 ppm ( parts - per - million ) from the hd 189733 exoplanet system using the dipol instrument @xcite at la palma s kva ( royal swedish academy of sciences ) telescope .
subsequently @xcite reported no polarization variation with an upper limit of 79 ppm ( 99 per cent confidence ) using the polish instrument .
further observations @xcite using turpol @xcite on the nordic optical telescope in three bands showed a polarization signal in the _ u _ and _ b _ bands but not in the _ v _ band .
they argued that the non - detection by @xcite was due to the use of too red a band .
the amplitude reported in the _ b _ band was 100 @xmath0 10 ppm .
the polarization was attributed to rayleigh scattering from the planet s atmosphere .
more recently @xcite have reported _ b _ band polarization measurements of hd 189733 using the polish2 instrument on the lick observatory 3-m telescope .
they set a limit on the _ b _ band polarization amplitude of 60 ppm ( 99.7 per cent confidence ) .
this paper presents the most sensitive measurements of the system to date using a blue broadband filter in an attempt to distinguish between these conflicting results .
hd 189733 was observed during three observing runs on the 3.9-m anglo - australian telescope ( aat ) at siding spring observatory , new south wales , australia .
the observations were made with the high precision polarimetric instrument ( hippi)@xcite . the dates of observations were 2014 aug 28 - 31 , 2015 may 23 , 26 and jun 26 .
hippi is an aperture polarimeter using a ferroelectric liquid crystal ( flc ) modulator , a wollaston prism analyser and two photomultiplier tubes ( pmt ) as detectors .
the flc provides a 500 hz primary modulation which is used together with two additional stages of slower modulation obtained by rotation of the wollaston prism and detectors , and finally by rotating the whole instrument to four position angles ( 0 , 45 , 90 , 135 degress ) using the aat s cassegrain rotator .
hippi has been shown from repeat observations of bright low polarization stars to deliver a precision of 4.3 @xmath2 10@xmath3 ( 4.3 parts - per - million or ppm ) or better @xcite .
this is comparable to or better than the precisions reported from polarimeters based on photoelastic modulators such as the pine mountain observatory polarimeter @xcite , planetpol @xcite , polish @xcite and polish2 @xcite .
hippi uses hamamatsu h10720 - 210 pmt modules that have ultra - bialkali photocathodes with a peak quantum efficiency of 43 per cent at 400 nm .
the hd 189733 observations were made through a 500 nm short pass filter ( referred to as 500sp ) to exclude any red light from the k 1.5 v star .
the overall bandpass covers 350 to 500 nm ( the polarization optics cut off light below 350 nm ) and the effective wavelength calculated from the bandpass model described by @xcite is 446.1 nm .
this is a similar but somewhat broader band than the _ b _ band used by @xcite and @xcite .
the moonlit sky can contribute a significant polarized background signal within our 6.7 diameter aperture .
the 2014 august observations were all made with lunar illumination phases less than 35 per cent but with the moon near 90@xmath4 seperation from hd 189733b where rayleigh scattering is at a maximum .
the 2015 observations were all made after the moon had set .
any residual background signal is subtracted from the data using a sky observation made immediately after each science observation at each cassegrain rotator position .
the telescope introduces a small polarization ( telescope polarization or tp ) that must be corrected for .
as described in @xcite we determine the tp using observations of a number of stars we believe to have very low polarization either based on previous planetpol observations @xcite or because of their small distances and expected low levels of interstellar polarization ( see * ? ? ?
* ; * ? ? ?
the telescope polarization has been found to be stable during each run , and between the 2015 may and june runs , but changed signficantly from 2014 to 2015 , probably as a result of the realuminization of the aat primary mirror in early 2015 .
determinations of tp in the sdss g@xmath5 filter are reported in @xcite and @xcite giving values of 48 @xmath0 5 ppm in 2014 , and 36 @xmath0 1 ppm in 2015 .
we have also made a smaller number of tp observations directly in the 500sp filter used with hd 189733 as listed in table [ tab_tp ] .
these observations consistently show telescope polarizations of 22 to 25 per cent higher values than the more extensive @xmath6 band observations .
.low polarization star measurements to determine telescope polarization ( tp ) in the 500sp filter .
the adopted tp for the august 2014 and may 2015 runs use only measurements acquired on that run .
the june 2015 run uses measurements from both may and june 2015 . [ cols="<,<,^,<,<",options="header " , ] [ tab_fit ] the less than 2-sigma uncertainty on the polarization amplitude means that this can not be considered as a detection of polarized light from the planet . a model with no polarization variation due to
the planet is still an accepteble fit to the data .
however , it is interesting that our best fit polarization amplitude of 29.4 @xmath0 15.6 ppm is at a level that is in agreement with plausible values for rayleigh scattering from the planet .
for example @xcite estimate an amplitude of 26 ppm for a rayleigh - like multiple scattering model with a single scattering albedo of 0.99 .
this is an optimistic model because in reality we would expect a modest reduction in the amplitude due to ( i ) atomic and molecular absorption features and ( ii ) the rayleigh depolarisation factor of 0.02 @xcite which reduces the polarisation at each scattering event .
considering the geometric albedo from a fractional eclipse depth of @xmath7 ppm from 290 - 450 nm reported in @xcite , and the rayleigh scattering model grid of @xcite which would place an upper limit to the amplitude of the polarised light contribution at about 30 per cent of the geometric albedo , we can estimate an upper bound for a rayleigh scattering atmosphere to be @xmath8
. a more extensive set of observations with an instrument like hippi on a 4-m class telescope might be capable of detecting the expected planetary signal in this and other bright hot jupiter systems .
putting a similar instrument on an 8-m class telescope should enable the measurement uncertainties to be halved from @xmath1 14 ppm to @xmath1 7 ppm , making detection of the expected signals at levels of @xmath1 20 ppm possible .
a clear result of our observations , is that in addition to any polarization from the planet , there is a constant background level of polarization from hd 189733 .
depending on what we assume about the planetary signal , this ranges from around 55 ppm , if we use the values given in section [ sec_results ] for no planetary signal , to 70 ppm for the best fit model of table [ tab_fit ] .
interstellar polarization is the most likely source of polarization in typical solar - like stars @xcite . extrapolating from the trends in figure 4 in @xcite
, we would expect a star at @xmath9 pc @xcite , in the northern hemisphere and in the galactic plane to have a polarization between 1540 ppm .
this is rather less than the values we find for hd 189733 .
circumstellar dust can also produce a constant offset for the polarized light signal from a system .
however , @xcite found hd 189733 was unlikely to have circumstellar dust substantial enough to affect polarisation measurements based upon measurements of infrared excess .
the effect of circumstellar dust on a system s polarized light measurements would be an addition to the offsets @xmath10 and @xmath11 which would not vary over the timescale of the planet s orbit .
there are other possible sources that could be contributing to polarized light in the system .
hd 189733a is an active by draconis type variable star : a star whose brightness varies due to star spots moving across its surface .
star spots can cause linear polarization by breaking the circular symmetry of the limb polarization @xcite .
however @xcite found that starspots would only cause a maximum of 3 ppm contribution to polarized light based on the photometric transit curve deviations from @xcite .
similarly , @xcite estimated the contribution of starspots on the polarized light signal to be only @xmath12 .
the interplay between the planet and starspot symmetry breaking can produce detectable effects during the planet s transit .
however none of our observations are taken during transit .
the linearly polarized light from the starspots themselves , is negligible @xcite under most circumstances .
magneto - optical effects such as the transverse zeeman effect @xcite or the faraday effect @xcite could introduce noise to the linear polarisation measuements if enough lines were present within our bandpass or if significant starspot coverage is present along with extended ionised gases from the planet s atmosphere respectively .
the rotational period of hd 189733a is known to be longer , at 11.8 days @xcite , than the orbital period of hd 189733b . with additional observations it should be possible to determine whether the magnitude of activity associated with starspots is significant , and
disentangle it from that due to reflection from clouds in the planetary atmosphere .
regardless , in any analysis focused on the orbital period , starspot effects will average out over time .
the large polarization amplitudes for hd 189733 reported by @xcite are not seen in the other three datasets now reported for this object , @xcite , @xcite and this work .
the strongest evidence for polarization variation found by @xcite was in the _ b _ band turpol data from the nordic optical telescope on la palma obtained on dates of 2008 apr 18 - 24 , and 2008 aug 2 - 9 .
they note that the august data was affected by saharan dust but argue that this should not cause any effect on the results as `` _ _ turpol enables exact compensation of any background polarization that is not variable ... _ _ '' .
this statement misunderstands how airborne dust impacts on precision polarimetry .
it is not a background sky polarization that could be automatically subtracted by the instrument s sky - subtraction capability .
the dust introduces a spurious polarization into starlight passing through the dust .
the effects were studied in detail by @xcite using observations of a saharan dust event observed with planetpol in 2005 .
data from the carlsberg meridian telescope on la palma show a substantial saharan dust event over 2008 aug 2 to 9 with the _ @xmath13 _ band extinction ranging from 0.212 to 0.377 throughout this period ( compared with a normal clear sky level of @xmath10.09 ) .
this is a larger and more extended event than the one observed in 2005 by @xcite which led to spurious polarization effects up to 48 ppm at 56 zenith distance .
it seems highly likely that the hd 189733 data reported by @xcite are significantly affected by this dust . as the individual observations are not reported , it is not possible to judge the extent of the problem .
however , without exclusion or correction of affected data , this can not be regarded as a reliable dataset for precision polarimetry .
we have reported new polarization observations of the hot jupiter exoplanet system hd 189733 .
the observations have higher precision than any previously reported for this object .
we do not detect the large polarization amplitude ( 100 - 200 ppm ) planetary signals previously reported @xcite .
our results agree reasonably well with the results of @xcite showing generally positive polarization values in q / i and u / i , and at most a small planetary polarization signal .
a least - squares fit of a rayleigh - lambert model gives a polarization amplitude of 29.4 @xmath0 15.6 ppm . while this signal has less than 2-sigma significance and
can not be claimed as a detection of planetary polarization , it is at a level consistent with a plausible polarization amplitude from the planet .
it is consistent with a multiply - scattering atmosphere which could produce the albedo measurements taken by @xcite .
this suggests that a more extensive series of observations , or observations on a larger telescope should enable the planetary polarization to be detected and measured .
hd 189733 has a significant constant background level of polarization that is a somewhat higher than would be expected for interstellar polarization .
this could be due to the non - unformity of the local hot bubble interstellar medium .
we suggest that the polarization data of @xcite , which shows a large polarization amplitude inconsistent with that reported by other groups , may be unreliable as a result of spurious polarization due to a saharan dust event over the la palma observatory in august 2008 .
the development of hippi was funded by the australian research council through discovery projects grant dp140100121 and by the unsw faculty of science through its faculty research grants program .
jpm is supported by a unsw vice - chancellor s fellowship .
the authors thank the director and staff of the australian astronomical observatory for their advice and support with interfacing hippi to the aat and during the three observing runs on the telescope .
the authors wish to thank referee hans martin schmid for constructive criticism of the paper . 99 afram n. , berdyugina s.v . , 2015 , a&a , 576 , a34 bailey j. , ulanowski z. , lucas p.w . , hough j.h . , hirst e. , tamura m. , 2008 , mnras , 386 , 1016 bailey j. , lucas p.w . , hough j.h . , 2010 , mnras , 405 , 2570 bailey j. , kedziora - chudczer l. , cotton d.v . , bott k. , hough j.h . ,
lucas p.w . , 2015 , mnras , 449 , 3064 berdyugina s.v .
, berdyugin a.v . , fluri d.m . ,
piirola v. , 2008 , apjl , 673 , l83-l86 berdyugina s.v . , berdyugin a.v . ,
fluri d.m . ,
piirola v. , 2011 , apjl , 728 , l6 bouchy f. et al .
, 2005 , a&a , 444 , l15-l19 bryden g. et al . , 2009 ,
apj , 705 , 1226 - 1236 buenzli e. , schmid h.m . , 2009 , a&a , 504 , p 259 - 276 calamai g. , landi deglinnocenti e. , landi deglinnocenti m. , 1975 , a&a , 45 , 2 , p 297 - 301 cotton d.v . , bailey j. , kedziora - chudczer l. , bott k. , lucas p.w . , hough j.h . ,
marshall j.p . , 2016 , mnras , 455 , 1607 evans t.m .
et al . , 2013 ,
apjl , 772 , l16 hansen j.e . ,
travis l.d . , 1974 , space sci .
, 16 , 527 hough j.h . ,
lucas p.w .
, 2003 , earths : darwin / tpf and the search for extrasolar terrestrial planets , 539 hough j.h . , lucas p.w .
, bailey j.a . , tamura m. , hirst e. , harrison d. , bartholomew - biggs m , 2006 , pasp , 118 , 1302 huovelin j. , saar s. , 1991 , apj , 347 , 319 - 329 kemp j. c. , barbour m.s , .
1981 , pasp , 93 , 521 - 525 .
knutson h. et al . , 2007 , nature , 447 , 183 - 186 koen c. , kilkenny d. , van wyk f. , marang f. , 2010 , mnras , 393 , 229 kostogryz n.m .
, yakobchuk t.m . ,
berdyugina s.v . , 2015 ,
apj , 806 , 1 , 97 lucas p.w . ,
hough j.h . ,
bailey j.a . ,
tamura m. , hirst e. , harrison d. , 2009 , mnras , 403 , 4 , pp 1949 - 1968 moutou c. et al . , 2007 , a&a , 473 , 2 , 651 - 660 penndorf r. , 1957 , j. opt .
, 47 , 176 piirola v. , 1973 , a&a , vol .
27 , p. 383 - 388 piirola v. , berdyugin a. , mikkola s. , coyne g.v . , 2005 ,
apj , 632 , 576 - 589 pont f. , knutson h. , gilliland r.l .
, moutou c. , charbonneau d. , 2008 , mnras , 385 , 109 pont f. , sing d.k . , gibson n.p . , aigrain s. , henry g. , husnoo n. , 2013 , mnras , 432 , 2917 press w.h .
, teukolsky s.a .
, vetterling w.t . ,
flannery b.p . , 1992 ,
_ numerical recipes in fortran : the art of scientific computing _ , cambridge university press seager s. , whitney b.a .
, sasselov d.d . , 2000 , apj , 540 , 504 torres g. , winn j.n . , holman m.j .
, apj , 677 , 1324 - 1342 triaud a.h.m.j .
et al . , 2009 , a&a , 506 , 377 wiktorowicz s.j . , 2009 , apj , 696 , 1116 wiktorowicz s.j .
matthews k. , 2008 , pasp , 120 , 1282 wiktorowicz s. , nofi , l.a . , 2015 ,
apj , 800 , l1 wiktorowicz s. , nofi l.a . ,
daniel j .-
, kopparla p. , laughlin g.p .
, hermis n. , yung y.l .
, swain m.r .
, 2015 , apj , 813 , 48 winn j.n .
et al . , 2007 ,
aj , 133 , 4
|
we present linear polarization observations of the exoplanet system hd 189733 made with the high precision polarimetric instrument ( hippi ) on the anglo - australian telescope ( aat ) .
the observations have higher precision than any previously reported for this object .
they do not show the large amplitude polarization variations reported by @xcite .
our results are consistent with polarization data presented by @xcite .
a formal least squares fit of a rayleigh - lambert model yields a polarization amplitude of 29.4 @xmath0 15.6 parts - per - million .
we observe a background constant level of polarization of @xmath1 5570 ppm , which is a little higher than expected for interstellar polarization at the distance of hd 189733 .
[ firstpage ] polarization techniques : polarimetric planets and satellites : atmospheres planets and satellites : individual : hd 189733b
|
since the launch in june 2008 of _ fermi _ lat @xcite , a high - energy ( he ) satellite measuring @xmath2-rays in the range 20 mev@xmath3300 gev , two pulsar catalogues ( 1pc , @xcite ; 2pc , @xcite ) discussing the light curve and spectral properties of 117 pulsars have been released . the vast majority of the _ fermi_-detected pulsars display exponentially cutoff spectra with cutoffs around a few gev .
these spectra are believed to be due to curvature radiation ( cr ) , which is assumed to be the dominating emission process in the gev band ( see section [ subsection : radiationmechanisms ] ) . there
exist several physical radiation models that can be used to study he emission from pulsars .
these include the polar cap ( pc ; @xcite ) , slot gap ( sg ; @xcite ) , outer gap ( og ; @xcite ) , and the pair - starved polar cap ( pspc ; @xcite ) models , which can be distinguished from each other based on the different assumptions of the geometry and location of the ` gap regions ' .
the ` gap region ' is where particle acceleration takes place due to an unscreened , rotation - induced @xmath4-field parallel to the local @xmath5-field , as well as subsequent emission by these particles . in pc models emission from he particles
is assumed to originate close to the neutron star ( ns ) surface .
these particles are accelerated by large @xmath4-fields near the magnetic poles ( known as the magnetic pcs ) only up to a few stellar radii . in sg models ,
the radiation comes from narrow gaps close to the last open field lines ( the field lines that are tangent to the light cylinder where the corotation speed equals the speed of light @xmath6 ) , with the gaps extending from the ns surface up to high altitudes . in the og model
, the gap region extends from the null - charge surface , where the goldreich - julian charge density is zero @xcite up to high altitudes , also close to the last open field lines .
the pspc model involves a gap region that extends from the ns surface to the light cylinder over the full open volume @xcite , since the potential is unscreened in this case , so that there are not enough pairs to fully screen the @xmath4-field . to explain he emission in the standard models , one has to take detailed particle transport and radiation mechanisms into account .
these mechanisms include cr , synchrotron radiation ( sr ) , and inverse compton scattering ( ics ) .
cr occurs whenever charged particles are constrained to move along curved paths , e.g. , along curved @xmath5-field lines ( e.g. , @xcite ) , therefore involving a change in their longitudinal kinetic energy .
when the emitted cr photon energy and the local @xmath5-field are high enough , magnetic pair production may occur ( where an he photon converts into an electron - positron pair , @xmath7 ) , leading to a cascade of @xmath7 pairs which may screen the parallel @xmath4-field outside the gaps .
the pair cascade is characterized by the so - called multiplicity , i.e. , the number of pairs spawned by a single primary .
the pairs may radiate sr if they have velocity components perpendicular to the local @xmath5-field so that this process involves a change in the particles transverse kinetic energy .
also , ics occurs due to the relativistic particles which upscatter soft photons ( e.g. , originating at a heated pc ) , which results in the `` boosting '' of the photon energies up to very high energies .
ics photons may also be converted into @xmath7 pairs .
two - photon pair creation ( @xmath8-absorption ) may also occur , in particular in og models .
cyclotron emission combined with subsequent ics has also been considered by @xcite to explain the broadband spectrum of the crab pulsar .
early prediction of the phase - averaged spectrum for the vela pulsar .
the solid lines represent the spectral components for cr , sr , and the thermal surface flux ( kt ) .
the dashed curve represents the tev pulsed spectral component associated with the ics of sr of the primary @xmath7 ( ssc ) .
adapted from @xcite . ]
early modelling , assuming the standard og model , predicted spectral components in the vhe regime when estimating the ics of primary electrons on sr or soft photons .
this resulted in a natural bump around a few tev ( involving @xmath9 tev particles ) in the extreme klein - nishina limit as seen in figure [ romani96 ] and [ hirotani01 ] .
however , these components may not survive up to the light cylinder and beyond @xcite , since @xmath8 pair creation leads to absorption of the tev @xmath2-ray flux .
expected tev spectra for three bright pulsars including the crab ( solid lines ) , psr b0656@xmath1014 ( dashed lines ) , and psr b1509@xmath358 ( dotted lines ) .
the thick and thin curves represent inclination angles ( between the spin and magnetic axis ) of @xmath11 and @xmath12 , respectively @xcite . ]
other studies assumed cr to be the dominant radiation mechanism producing @xmath2-ray emission when performing spectral modelling and found spectral cutoffs of up to 50 gev .
for example , @xcite modelled the cutoffs of millisecond pulsars ( msps ) .
these are pulsars which possess relatively low @xmath5-fields and short periods .
their model assumed a static dipole @xmath5-field and a pc geometry , and predicted cr from the primary electrons that are released from the pc .
their predicted cr spectral component cut off at @xmath13 gev .
the cr photons may undergo pair production in the intense low - altitude @xmath5-fields , and the newly formed electron - positron secondaries will emit sr in the optical and x - ray band ( see @xcite ) .
therefore , they concluded that the he cr from msps occurred in an energy band that was above the detection range of satellite detectors like _ energetic gamma - ray experiment telescope ( egret ) _ and below that of ground - based cherenkov detectors such as the _ high energy stereoscopic system ( h.e.s.s .
i)_. later studies investigated the x - ray and @xmath2-ray spectrum of rotation - powered msps using a pspc model @xcite , and found cr cutoffs of @xmath14 gev ( see @xcite ) .
optical to @xmath2-ray spectra were also modelled by @xcite assuming an sg accelerator and a retarded vacuum dipole ( rvd ) @xmath5-field , for the crab pulsar .
they found spectral cutoffs of up to a few gev .
another study modelled the phase - resolved spectra of the crab pulsar using the og and sg models , and found he cutoffs of up to @xmath15 gev @xcite .
cutoffs around @xmath16 gev were found for the og model using the rvd @xmath5-field @xcite .
in view of the above theoretical paradigm it was not expected that a pulsar should be visible in the vhe regime .
it was therefore surprising when the _ very energetic radiation imaging telescope array system ( veritas ) _ announced the detection of pulsed emission from the crab pulsar above @xmath13 gev @xcite , followed by the detection by the _
major atmospheric gamma - ray imaging cherenkov ( magic ) _ of emission up to @xmath17 gev ( soon after their initial detection of emission at @xmath15 gev ) @xcite .
the _ magic _ collaboration has since reported the detection of pulsed photons with energies up to 1.5 tev @xcite .
ground - based cherenkov telescopes are now searching for more examples of vhe pulsars .
the observed and modelled spectra for phase - resolved @xmath2-ray emission , including both emission peaks p1 and p2 , from the crab pulsar as measured by _ magic _ ( dark red squares ) .
the plot also contains measurements from _ fermi _ lat ( orange diamonds ) and _ veritas _ ( light blue squares and solid line ) .
an og model including emission from pairs are assumed .
the systematic error of the _ magic_-stereo measurement is shown and corresponds to a shift of @xmath1817% in energy and @xmath1819% in flux . from @xcite . ] in figure [ magic ] the phase - resolved @xmath2-ray spectrum for the crab pulsar for both light curve peaks p1 and p2 , as measured by _ magic _ , is shown .
the spectrum also contains measurements by _ fermi _ lat and _ veritas_. the observations by _ magic _ are in good agreement with those by _ veritas_. the _ veritas _
data fit a broken power law , and raised important questions whether the cr component is extended or if a second component is required to explain the observed spectra .
the detection of the crab pulsar above several gev prompted _ fermi _ to search for pulsed emission at he s .
they detected significant pulsations above 10 gev from 20 pulsars and above 25 gev from 12 pulsars @xcite .
the crab pulsar is the first source which have been detected over almost all energies ranging from radio to vhe @xmath2-rays .
more recently , pulsed emission was detected from the vela pulsar above 30 gev with the _ h.e.s.s .
_ @xcite and up to @xmath19 gev ( at the @xmath20-level ) with the _ fermi _ lat @xcite .
_ magic _ furthermore detected no emission from geminga above 50 gev @xcite ; neither did _ veritas _ above 100 gev @xcite .
a stacking analysis involving 115 _ fermi_-detected pulsars ( excluding the crab pulsar ) was performed by @xcite using _ fermi _ data .
however no emission above 50 gev was detected , implying that vhe pulsar detections may be rare , given current telescope sensitivities .
from all these observations there are three effects visible in the energy - dependent pulse profiles : the peaks remain at the same phase , the p1/p2 ratio decreases as energy increases , and the pulse width decreases with increasing energy .
more vhe pulsars may be found by the _ cherenkov telescope array ( cta ) _ which will have a ten - fold increase in sensitivity compared to present - day cherenkov telescopes .
all of the standard pulsar emission models ( see section [ subsection : emissionmodels ] ) predicted he spectral cutoffs between a few gev and up to @xmath13 gev , assuming @xmath5-fields such as the static dipole and rvd solutions .
clearly , refinements to these radiation models and @xmath5-fields are needed to explain the observed vhe emission from the crab .
there are a few ideas for such refinements .
one is a revised og model by @xcite which can produce ic radiation of up to @xmath17 gev due to secondary and tertiary pairs upscattering infrared to ultraviolet photons @xcite . in this og model
the ic flux depends sensitively on the @xmath5-field structure near the light cylinder .
another idea was proposed by @xcite , invoking the ssc radiation process .
this is indeed a promising radiation mechanism , where relativistic particles upscatter the sr photons emitted by the same population .
the ssc radiation mechanism was applied by @xcite to predict optical to x - ray , and @xmath2-ray spectra ( see figure [ ssc ] ) assuming an sg model and a force - free @xmath5-field .
this process relies critically on the assumed electrodynamics and the magnetospheric structure .
they performed simulations for the crab and vela pulsars , as well as two msps , i.e. , b1821@xmath324 and b1937@xmath1021 .
however , the only significant predicted ssc component was for the crab pulsar .
they also found that the pair sr matched the observed x - ray spectrum of the msps .
they furthermore tested the addition of an he power law extension to the pair spectrum ( dashed lines in figure 4 ) whose sr spectrum would account for the observed emission in the @xmath21 mev range
. however , the resulting ssc component exceeded the observed _ magic _ and _ veritas _ points , implying that the observed @xmath21 mev emission is not produced by the same particles that produce the ssc emission .
modelled spectrum of phase - averaged pulsed emission from the crab pulsar .
the spectral components from primary electrons and pairs ( as labelled ) are for a magnetic inclination angle @xmath22 , observer angle @xmath23 and pair multiplicity @xmath24 . from @xcite . ]
the abovementioned detections of the vhe pulsed emission from pulsars and the explanation thereof implies that this emission may yield strong constraints on @xmath2-ray radiation mechanisms , the location of acceleration regions , and the @xmath5-field structure .
there is thus an urgent need for refinements and extensions of standard pulsar emission models and radiation mechanisms , including more realistic @xmath5-fields .
some examples are discussed above , but there are many more .
sg model refinements include photon - photon pair production attenuation within the model and also more realistic @xmath5-fields such as dissipative magnetospheric solutions .
one could also model the emission pulse profiles as a function of energy .
ground - based cherenkov telescopes are now searching for more examples of vhe pulsars .
new pulsar models will assist us in predicting the level of vhe emission expected from them , which would be very important for the upcoming _
cta_. the low threshold energy of _ cta _ will provide an overlap with the _ fermi _ energy range and will help to discriminate between cr and a potentially new spectral component . this work is based on research supported by the national research foundation ( nrf ) of south africa ( grant numbers 90822 , 93278 , and 99072 ) .
any opinions , findings , and conclusions or recommendations expressed are that of the authors , and the nrf accepts no liability whatsoever in this regard .
acknowledges the support from the nasa astrophysics theory program .
29 abdo a a _ et al .
_ 2010 _ astrophys .
j. supp . _ * 187 * 460 abdo a a _ et al .
_ 2013 _ astrophys .
j. supp . _
* 208 * 17 ackermann m _
_ 2013 _ astrophys .
j. supp . _
* 209 * 34 ahnen , m l _ et al .
_ 2016 ( arxiv:1603.00730 ) aleksi j _ et al .
_ 2011 _ astrophys .
j. _ * 742 * 43 aleksi j _ et al .
_ 2012 _ astrophys .
* 540 * a69 aliu e _ et al .
_ 2008 _ science _ * 322 * 1221 aliu e _
_ 2011 _ science _ * 334 * 69 aliu e _ et al .
_ 2015 _ apj _ * 800 * 61 ansoldi s
_ 2016 _ astron . &
_ * 585 * a133 arons j 1983 _ astrophys . j. _ * 266 * 251 atwood w b _ et al .
_ 2009 _ astrophys . j. _ * 697 * 1071 bulik t , rudak b and dyks j 2000 _ mnras _ * 317 * , 97 cheng k s , ho c and ruderman m 1986 _ astrophys .
j. _ * 300 * 522 daugherty j k and harding a k 1982 _ astrophys . j. _ * 252 * 337 frackowiak m and rudak b 2005 _ adv .
space res . _
* 35 * 1152 goldreich p and julian w h 1969 _ astrophys .
j. _ * 157 * 869 harding a k 1981 _ astrophys .
j. _ * 245 * 267 harding a k , muslimov a g and zhang b 2002 _ astrophys .
j. _ * 576 * 366 harding a k , usov v v and muslimov a g 2005 _ astrophys
. j. _ * 622 * 531 harding a k , stern j v , dyks j and frackowiak m 2008 _ astrophys . j. _ * 680 * 1378 harding a k and kalapotharakos c 2015 _ astrophys .
j. _ * 811 * 63 hirotani k 2001 _ astrophys .
j. _ * 549 * 495 hirotani k 2008 ( arxiv:0809.1283 ) mccann a _ et al .
_ 2015 _ astrophys .
j. _ * 804 * 86 leung c y _ et al .
_ 2014 _ astrophys .
j. _ * 797 * l13 lyutikov m , otte n and mccann a 2012 _ astrophys .
j. _ * 754 * 33 lyutikov m 2013 _ mnras _ * 431 * 2580 romani r w 1996 _ astrophys . j. _ * 470 * 469 stegmann c 2014 talk at `` astroparticle physics 2014 '' amsterdam tang a s p _
_ 2008 _ astrophys .
j. _ * 676 * 562 venter c and de jager o c 2005 _ astrophys . j. _ * 619 * l167
|
the vast majority of pulsars detected by the _ fermi
_ large area telescope ( lat ) display exponentially cutoff spectra with cutoffs falling in a narrow band around a few gev .
early spectral modelling predicted spectral cutoffs at energies of up to 100 gev , assuming curvature radiation .
it was therefore not expected that pulsars would be visible in the very - high energy ( vhe ) regime ( @xmath0 gev ) .
the _ veritas _ announcement of the detection of pulsed emission from the crab pulsar at energies up to @xmath1 gev ( and now up to 1.5 tev as detected by _
magic _ ) therefore raised important questions about our understanding of the electrodynamics and local environment of pulsars .
_ h.e.s.s .
_ has now detected pulsed emission from the vela pulsar down to tens of gev , making this the second pulsar detected by a ground - based cherenkov telescope .
deep upper limits have also been obtained by _ veritas _ and _ magic _ for the geminga pulsar .
we will review the latest developments in vhe pulsar science , including an overview of the latest observations , refinements , and extensions to radiation models and magnetic field structures , and the implementation of new radiation mechanisms .
this will assist us in understanding the vhe emission detected from the crab pulsar , and predicting the level of vhe emission expected from other pulsars , which is very important for the upcoming _
|
we acknowledge support by the riken / bnl research center and doe grant de - ac02 - 98ch10886 .
r. baier , y. l. dokshitzer , a. h. mueller , s. peigne and d. schiff , nucl .
b * 483 * , 291 ( 1997 ) ; nucl .
b * 484 * , 265 ( 1997 ) .
b. g. zakharov , jetp lett .
* 63 * , 952 ( 1996 ) .
u. a. wiedemann , nucl .
a * 690 * , 731 ( 2001 ) .
m. gyulassy , p. levai and i. vitev , phys .
lett . * 85 * , 5535 ( 2000 ) ; nucl .
b * 594 * , 371 ( 2001 ) .
p. arnold , g. d. moore and l. g. yaffe , jhep * 0206 * , 030 ( 2002 ) ; s. jeon and g. d. moore , phys . rev .
c * 71 * , 034901 ( 2005 ) .
w. liu and r. j. fries , phys .
c * 77 * , 054902 ( 2008 ) .
r. j. fries , arxiv:0907.4390 [ nucl - th ] .
r. j. fries , b. mller and d. k. srivastava , phys .
* 90 * , 132301 ( 2003 ) .
r. j. fries , b. mller and d. k. srivastava , phys .
c * 72 * , 041902 ( 2005 ) .
d. k. srivastava , c. gale and r. j. fries , phys .
c * 67 * , 034903 ( 2003 ) ; s. turbide , c. gale , d. k. srivastava and r. j. fries , phys .
c * 74 * , 014903 ( 2006 ) . c. gale ,
t. c. awes , r. j. fries and d. k. srivastava , j. phys .
g * 30 * , s1013 ( 2004 ) .
s. turbide , c. gale , s. jeon and g. d. moore , phys .
c * 72 * , 014906 ( 2005 ) .
s. turbide , c. gale , e. frodermann and u. heinz , phys .
c * 77 * , 024909 ( 2008 ) .
s. s. adler _ et al .
_ [ phenix collaboration ] , phys .
lett . * 94 * , 232301 ( 2005 ) .
w. liu and r. j. fries , phys .
c * 78 * , 037902 ( 2008 ) .
s. turbide , c. gale and r. j. fries , phys .
lett . * 96 * , 032303 ( 2006 ) .
r. chatterjee , e. s. frodermann , u. w. heinz and d. k. srivastava , phys .
lett . * 96 * , 202302 ( 2006 ) ; r. chatterjee , d. k. srivastava , u. w. heinz and c. gale , phys .
c * 75 * , 054909 ( 2007 ) .
w. liu and r. j. fries , arxiv:0805.3721 [ nucl - th ] .
|
the suppression of high-@xmath0 particles in heavy ion collisions was one of the key discoveries at the relativistic heavy ion collider .
this is usually parameterized by the average rate of momentum - transfer squared to this particle , @xmath1 . here
we argue that measurements of identified particles at high @xmath0 can lead to complementary information about the medium .
the leading particle of a jet can change its identity through interactions with the medium .
tracing such flavor conversions could allow us to constrain the mean free path . here
we review the basic concepts of flavor conversions and discuss applications to particle ratios and elliptic flow .
we make a prediction that strangeness is enhanced at high @xmath0 at rhic energies while its elliptic flow is suppressed . for the past decade
high momentum particles and jets have been used to probe the quark gluon plasma ( qgp ) phase created at the relativistic heavy ion collider ( rhic ) .
the energy loss of a fast parton suffered in the medium carries information about the typical momentum transfer @xmath2 along the path , more precisely about the transport coefficient @xmath3 @xcite .
we have recently argued that the mean free path @xmath4 of a fast parton could be determined separately by measuring the change in hadro - chemistry induced by the medium @xcite .
here we discuss a model based on conversions of the leading particle of a jet .
just as partons can lose energy through collisions and induced radiation , they can scatter through channels in which the identity of the fastest parton in the initial and final state are not the same .
examples are binary collisions like @xmath5 or @xmath6 which can lead to conversions of quarks into gluons and vice versa . here
the first parton on each side has a large momentum ( the leading jet parton ) and the second parton in the initial state is a thermal parton from the quark gluon plasma .
the rate of flavor conversions depends on the mean free path @xmath4 of fast partons .
conversions between quarks and gluons should obscure their different color factors coupling them to the medium . instead of a relative factor 9/4 in @xmath1
only an average color factor should be observable in a long enough medium .
it was pointed out that a larger quenching for gluons could be reflected in more suppression of protons compared to pions given the preference of gluon to proton fragmentation in modern fragmentation functions @xcite .
jet conversions should soften this effect and increase the proton to pion ration in central collisions @xcite .
fig .
[ fig:1 ] shows the ratio of nuclear modification factors @xmath7 for protons and pions with and without conversions between quarks and gluons taken into account @xcite .
clearly , flavor conversions lead to less proton suppression .
conversions had also been discussed before for photons and dileptons .
fast quarks and gluons can create real or virtual photons through compton and annihilation processes with the medium , @xmath8 , @xmath9 .
it has been realized over the years that this process can make a large contribution to the overall yield of photons or dileptons @xcite .
fig .
[ fig:1 ] also shows the nuclear modification factor for direct photons with and without jet conversions into photons from the computation described in @xcite together with data from phenix .
most recently we predicted an enhancement of strange hadrons from jet conversions at rhic energies .
this is driven by a strong chemical gradient between the jet sample at rhic and the quark gluon plasma .
while strangeness is equilibrated in the latter , it contributes only a small fraction ( @xmath10 ) to the former .
elastic channels like @xmath11 and in particular kick - out reactions like @xmath12 can lead to an enrichment of strangeness in the jet sample .
obviously , for a sufficiently long medium the jet sample would be driven toward chemical equilibrium through interactions with the chemically equilibrated medium .
fig .
[ fig:2 ] shows the expected nuclear modification factor for neutral kaons at rhic . for the perturbative elastic rates chosen in @xcite with a @xmath13 factor of 4 we see a clear enhancement extending above 10 gev/@xmath14 . for neutral kaons with and without conversion processes
allowed .
the strangeness in the jet sample is driven towards equilibrium by coupling it chemically to the quark gluon plasma.,width=264 ] we have also checked the effect of conversions on heavy quark production @xcite .
we do not find any significant yields of charm or bottom quarks at high @xmath0 from these processes . the same is true even at lhc energies .
the reason is that thresholds and low center of mass energies suppress pair production in interactions of jets with the medium , and kick - out reactions suffer from small heavy quark densities in the medium to begin with . even for charm at lhc
we do not expect chemical equilibration and there is no large chemical gradient between heavy quarks in jets and the medium .
we have also studied the effect of flavor conversions on the azimuthal asymmetry @xmath15 .
it was first pointed out in ref .
@xcite that photons from jet - medium interactions should be more abundantly produced in the direction where the medium is thicker , leading to a negative contribution to @xmath15 .
see @xcite for comprehensive calculations of photon elliptic flow .
it is important to realize that this mechanism is generally true for all particles produced in jet - medium interactions @xcite . in particular ,
we expect it to hold for additional strange hadrons produced at rhic .
fig .
[ fig:3 ] shows the effect of flavor conversions on the azimuthal asymmetry @xmath15 of up , down and strange quarks as well as gluons .
as predicted , conversions make light quarks and gluons behave similarly while there is a significant decrease in the @xmath15 of strange quarks due to the additional large yield with a negative contribution .
note that the total @xmath15 is determined from summing up all sources of particles both with positive and negative @xmath15 .
fig .
[ fig:4 ] shows the resulting azimuthal asymmetry @xmath15 that we expect for kaons at rhic @xcite . in summary , we have shown that jet conversions can lead to measurable signatures in hadron production at high @xmath0 .
we predict an enhancement of strange hadrons at high @xmath0 at rhic and a suppression of their azimuthal asymmetry coefficient @xmath15 .
|
let @xmath7 denote the family of functions @xmath8 of the form @xmath9which are analytic in the open unit disk @xmath10 and satisfy the usual normalization condition @xmath11 let @xmath12 denote the subclass of @xmath7 which are univalent in @xmath13 .
also let @xmath14and @xmath15 denote the subclasses of @xmath7 consisting of functions which are starlike and convex in @xmath13,respectively .
these classes can characterize as follows:@xmath16@xmath17 some special functions like bessel functions play an important role in applied mathematics and physics . we know that the bessel functions of the first kind @xmath18 is defined by(see @xcite and @xcite)@xmath19where @xmath20 stands for euler gamma function and it is a particular solution of the second - order linear homogeneous differential equation(see , for details , @xcite and @xcite):@xmath21 y(z)=0 \label{1.5}\]]where @xmath22 now , we consider the function @xmath23 defined by(see @xcite ) @xmath24on the other hand we know that there is following relation in between pochhammer symbol ( or appell ) and euler s gamma function:@xmath25if we use the relation ( 1.7 ) we can obtain the following series representation for the functions @xmath26 given by ( 1.6):@xmath27where @xmath28 in this study
, we will investigate the ratio of a function of the form ( 1.8 ) to its sequence of partial sums @xmath29 when the coefficients of @xmath30 satisfy some conditions .
we will obtain lower bounds for @xmath31 @xmath32 @xmath33 and @xmath34 for interesting developments on the partial sums of some special functions and the some classes of analytic functions , the readers can browse to the works of orhan and yagmur @xcite , alar and deniz @xcite , brickman et al .
@xcite , silverman @xcite , owa et al .
@xcite , sheil - small @xcite , silvia @xcite , akta and orhan @xcite .
if @xmath35 and @xmath36 then the function@xmath37given by ( 1.6 ) satisfies the follwing inequalities:@xmath38@xmath39 _ _ by using the well - known triangle inequality:__@xmath40 _ and the inequalities _
@xmath41 for @xmath42 _ _ _ _ @xmath43 we get@xmath44__if we consider inequality ( 1.10 ) , then we have__@xmath45
let @xmath46 @xmath36 the function @xmath47 defined by ( 1.8 ) and its sequence of partial sums @xmath48 .
if the inequality@xmath49is true , then we have the followings:@xmath50@xmath51 from the inequality ( 1.9 ) in the lemma 1.1 we write that@xmath52which is equivalent to@xmath53where @xmath28 now we may write@xmath54@xmath55 set @xmath56/@xmath57 so that @xmath58 then@xmath59and @xmath60now @xmath61 if and only if@xmath62it suffices to show that the left hand side of ( 2.4 ) is bounded above by @xmath63 which is equivalent to@xmath64to prove the result ( 2.2 ) , we write@xmath65@xmath66 then from ( 2.5 ) we get@xmath67and@xmath68now @xmath61 if and only if@xmath69since the left hand side of ( 2.6 ) is bounded above by @xmath70 the proof is complated .
let @xmath46 @xmath36 the function @xmath47 defined by ( 1.8 ) and its sequence of partial sums @xmath48 .
if the inequality@xmath71is true , then we have the followings:@xmath72@xmath73 from the inequality ( 1.10 ) in the lemma 1.1 we write that@xmath74which is equivalent to @xmath75where @xmath28 now we write@xmath76where@xmath77the last inequality is equivalent to@xmath78it suffices to show that the left hand side of ( 2.9 ) is bounded above by @xmath79 which is equivalent to @xmath80to prove the result ( 2.8 ) , we write @xmath81where @xmath82the last inequality is equivalent to @xmath83since the left hand side of ( 2.10 ) is bounded above by @xmath84 the proof is complated . for some special cases of @xmath85
it is known that(see @xcite)@xmath86and@xmath87now by using some special cases of @xmath88 and @xmath85 in theorem 2.1 and theorem 2.2 , we can obtain following corollaries . if we take @xmath89 @xmath90 and @xmath91 we have @xmath92 @xmath93 and @xmath94 . in view of theorem 2.2
we get the following inequalities:@xmath95@xmath96 if we take @xmath97 @xmath90 and @xmath91 we have @xmath98 @xmath99 @xmath100 and @xmath101 in view of theorem 2.2 we get the following inequalities : @xmath102@xmath103 if we take @xmath104 @xmath105 and @xmath91 we have @xmath106 @xmath107 @xmath108 and @xmath109 in view of theorem 2.1 we get the following inequalities : @xmath110@xmath111
in this section , illustrative examples along with the geometrical descriptions of the image domains of the unit disk by the ratio of normalized dini function to its sequence of partial sums or the the ratio of its sequence of partial sums to the function which we considered in our corollaries in section 2 , have been given in following figures .
|
let @xmath0 be the sequence of partial sums of normalized dini functions @xmath1 where @xmath2 .
the aim of the present paper is to obtain lower bounds for @xmath3 @xmath4 @xmath5 and @xmath6 .
also we give a few geometric description regarding image domains of some functions .
|
duality is one of the most fascinating phenomena in quantum field theory .
it provides an alternative , often non - perturbative , understanding of the theory that is not accessible in the original description .
seiberg duality @xcite , under which weakly coupled gauge theory is mapped into strongly coupled one with different rank , and vice versa , is a prominent example .
although the mapping of local operators under seiberg duality is known to be rather simple @xcite , the transformation properties of non - local operators , such as wilson loops , are more non - trivial and less studied . by its strong - weak nature
, any checks of seiberg duality must involve a non - perturbative approach .
the localization method @xcite is a powerful technique applicable to supersymmetric field theory and reduces the infinite dimensional path integral of quantum field theory to a finite integral which can be often regarded as a matrix model integral .
this method allows for exact computation of quantities such as partition function and wilson loops at strong coupling , and is an ideal tool for studying non - perturbative physics such as seiberg duality . among various supersymmetric gauge theories calculable by the localization method , we focus on the three - dimensional @xmath12 supersymmetric @xmath13 chern - simons - matter ( csm ) theory , known as the abj theory @xcite , where @xmath14 is the chern - simons level . in the special case with @xmath15 , this theory
is called the abjm theory @xcite .
it is believed that the @xmath16 abj theory gives the low - energy description of the system of @xmath17 m2-branes and @xmath18 fractional m2-branes probing @xmath19 , and its holographic dual geometry is ads@xmath20 in m - theory or ads@xmath21 in type iia superstring @xcite .
it is also conjectured that this theory has another dual description in terms of the @xmath22 supersymmetric , parity - violating version of the vasiliev higher spin theory in four dimensions @xcite . applying the localization method to the abj(m ) theory on @xmath23 yields @xcite the so - called abj(m ) matrix model , which has been extensively studied recently and led to much insight into the non - perturbative effects in the theory .
for instance , applying the standard large @xmath17 technique to the abj matrix model reproduced the @xmath24 scaling of the free energy expected from the gravity dual @xcite .
for the abjm theory , the @xmath25 corrections were summed to all orders using holomorphic anomaly equation at large t hooft coupling @xmath26 in the type iia regime @xmath27 and an airy function behavior was found @xcite .
furthermore , the powerful fermi gas approach was developed and reproduced the airy function in the m - theory regime ( large @xmath17 , finite @xmath4)@xcite .
based on the fermi gas approach , instanton corrections were examined @xcite and a cancellation mechanism between worldsheet and membrane ( d2 ) instantons was discovered .
more recently , the fermi gas approach was generalized to the abj theory in @xcite and further studied in @xcite . in a different line of development , in @xcite , the abj matrix model was rewritten in a form generalizing the `` mirror description '' for the abjm theory @xcite .
this was achieved by starting with the @xmath28 lens space matrix model @xcite which is exactly computable and analytically continuing it by sending @xmath29 to reach the abj matrix model . as a result
, partition function was written as a @xmath30-dimensional contour integral and the residue it picks up can be given an interpretation of perturbative or non - perturbative contribution , according to its @xmath4 dependence . later
, this mirror expression was reproduced in a more direct way in @xcite .
this formulation is particularly suitable for studying the duality with the higher spin theory @xcite .
it was conjectured in @xcite that the abj theory has seiberg duality , which states that the following two theories are equivalent : @xmath31 where we assumed @xmath32 .
this duality can be understood in the brane realization of the abj theory @xcite as moving 5-branes past each other and creating / annihilating d3-branes between them by the hanany - witten effect @xcite .
it is a special case of the giveon - kutasov duality @xcite for more general @xmath33 cs - matter theories , which can be regarded as the three - dimensional analog of the four - dimensional seiberg duality @xcite .
the equality of the partition function ( up to a phase ) between the dual theories in was proven in @xcite , although one mathematical relation was assumed . on the other hand , in the `` mirror '' framework of @xcite , the equality of the dual partition function
is more or less obvious @xcite and , moreover , it was observed that the perturbative and non - perturbative contributions get exchanged into each other under the duality .
wilson loops are the only observables in pure cs theory @xcite and , even in cs - matter theory , they are very natural objects to consider .
it is known that there are two basic circular supersymmetric wilson loops for the abj(m ) theory on @xmath23 ; the wilson loop carrying a non - trivial representation with respect to one of the two gauge groups preserves @xmath34 of supersymmetry @xcite while an appropriate combination of @xmath34-bps wilson loops for two gauge groups preserves @xmath35 of supersymmetry @xcite .
the bulk dual of the @xmath36-bps wilson loop is a fundamental string while the bulk dual of the @xmath37-bps wilson loop is not completely understood @xcite . for more discussion . ] by the localization method , these wilson loops can be computed using the abj(m ) matrix model @xcite , and the techniques developed for partition function to study non - perturbative effects can be generalized to wilson loops , such as the large @xmath17 analysis @xcite , the fermi gas approach @xcite , and the cancellation between worldsheet and membrane instantons @xcite . in this paper
we study the @xmath34 and @xmath35-bps wilson loops and their seiberg duality in the abj theory , based on the approach developed in @xcite for partition function .
we mainly focus on the representations with winding number @xmath38 . starting with wilson loops in lens space matrix model and analytically continuing it , we obtain a new expression for the supersymmetric wilson loops in the abj theory in terms of @xmath39-dimensional contour integrals . as an application and as a way to check the consistency of our formula , we study seiberg duality on the abj wilson loops . the duality rule for wilson loops
is highly non - trivial because they are non - local operators and not charged under the global symmetry .
how to find given - kutasov duality relations for wilson loops in general @xmath33 cs - matter theories , of which the abj theory is a special case , was proposed in @xcite based on quantum algebraic relations .
we derive the duality rule of the supersymmetric wilson loops using our integral expression and confirm that it is consistent with the proposal of @xcite .
we also discuss how the perturbative and non - perturbative effects are mapped under seiberg duality .
moreover , we provide a heuristic explanation of the duality rule from the brane realization of the abj theory . in an appropriate duality frame
, the wilson loop is interpreted as the position of the branes and , carefully following how the hanany - witten effect acts on it , we reproduce the correct duality rule .
this paper is organized as follows . in section [ mainresults ] , we present the main results of the paper without proof : the integral expression for the @xmath34-bps and @xmath35-bps wilson loops and the their transformation rules under seiberg duality . in section [ derivation ] , we explain how to derive the integral expression for the abj wilson loops by analytically continuing the lens space ones . in section [ sdsection ] ,
we first give a heuristic , brane picture for the seiberg duality rule , and then present a rigorous proof using the integral expression .
section [ summarydiscussions ] is devoted to a summary and discussion .
appendices contain further detail of computations in the main text , as well as a discussion of wilson loops with general representation in appendix [ wl_gen_rep ] and an alternative derivation of the seiberg duality rule using the algebraic approach of @xcite in appendix [ kapustinwillett ] .
the wilson loops we are concerned with are those preserving fractions of supersymmetries in the @xmath40 @xmath1 csm theory , also known as the abj(m ) theory , saturating the bogomolnyi - prasad - sommerfield ( bps ) bound .
more specifically , we consider two types of circular bps wilson loops on @xmath23 , one preserving a @xmath41-th and the other preserving a half of supersymmetries @xcite . in the main text of this paper ,
we restrict ourselves to the wilson loops with winding number @xmath38 , where @xmath42 ( or @xmath43 ) corresponds to the fundamental ( or anti - fundamental ) representation of @xmath44 and/or @xmath45 gauge groups , and the wilson loops in more general representations will be discussed in appendix [ wl_gen_rep ] .
the abj(m ) theory consists of two 3@xmath46 @xmath47 vector multiplets @xmath48 with @xmath49 which are the dimensional reduction of 4@xmath46 @xmath50 vector multiplets and four bifundamental chiral multiplets , an @xmath51 vector , @xmath52 with @xmath53 in the representation @xmath54 and their conjugates .
the @xmath41-bps wilson loops of our interest are constructed as @xcite ^i_16(n_1,n_2 ; r)_k:=_r p(ia^1_^ + 2k||m^i_jc_i|c^j)ds , where the path of the loop specified by the vector @xmath55 is a circle , and the matrix @xmath56 is determined by the supersymmetries preserved and one can choose it to be @xmath57 .
these are the wilson loops on the first gauge group @xmath44 , but the @xmath58 , those on the second gauge group @xmath45 , can be constructed similarly by replacing @xmath59 , @xmath60 and @xmath61 ( @xmath62 ) .
the @xmath63-bps wilson loop is constructed in @xcite and can be conveniently expressed in terms of the supergroup @xmath64 as
_ 12(n_1,n_2;r)_k:=_rp(ids ) , where @xmath65 is a super - representation of @xmath64 and @xmath66 is the super - connection = ( cc a^1_^ -i2k||n^i_jc_i|c^j & ||_i|_i + ) with the matrix @xmath67 , the ( super-)circular path @xmath68 , and @xmath69 .
as mentioned , we are concerned with wilson loops with winding number @xmath38 rather than in generic representations in this paper . focusing on this class of wilson loops ,
the application of the localization technique @xcite for the @xmath41-bps wilson loops with winding @xmath38 reduces to the finite dimensional integrals of the matrix model type @xcite @xmath70 where the vev is with respect to the eigenvalue integrals @xmath71 with the shorthand notations for the one - loop determinant factors defined by _ sh()=_1i
< jn_1(2(_i-_j2)),_sh()=_1a < bn_2(2(_a-_b2 ) ) and _ ch ( , ) = _ i=1^n_1_a=1^n_2(2(_i-_a2 ) ) .
the coupling constant is @xmath72 . ]
@xmath73 is related to the cs level @xmath4 by @xmath74 while the factor @xmath75 in front is the normalization factor @xcite _
2-n_2 ^ 2)n_1!n_2 ! , : = k .
meanwhile , the @xmath63-bps wilson loop localizes to the supertrace @xcite @xmath76 that yields , for the @xmath38 winding wilson loop , a linear combination of @xmath41-bps wilson loops @xcite _ 12(n_1,n_2;n)_k = w^i_16(n_1,n_2 ; n)_k -(-1)^nw^ii_16(n_1,n_2 ; n)_k .
note that the integral is a well - defined fresnel integral even for non - integral ( but real ) @xmath4 and thus gives a continuous function @xmath4 , although in the physical abj theory the cs level @xmath4 is quantized to an integer .
we now present the results of our analysis of these matrix eigenvalue integrals .
the @xmath41-bps wilson loops are only on the first @xmath44 or the second @xmath45 gauge group . depending on whether @xmath32 or @xmath77 , their formula takes rather different forms .
[ [ bullet - the-1over-6-bps - un_1-wilson - loop - with - n_1le - n_2 ] ] @xmath78 the @xmath41-bps @xmath44 wilson loop with @xmath32 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in the case of @xmath32 , introducing the normalized wilson loop @xmath79 , we find the @xmath41-bps wilson loop with winding number @xmath38 on the first gauge group @xmath44 to be @xmath80 where @xmath81 and @xmath82 q^{-ns_l+n(l-2)}\prod_{\substack{i=1 \\
i\ne l}}^{n_1 } { \left(q^{s_i - s_{l}-n}\right)_1\over \left(q^{s_i - s_{l}}\right)_1 } \label{integrali}\\ & \quad\times\prod_{i=1}^{n_1}\left[{(-1)^{m}\left(q^{s_i+1}\right)_m \over\left(1+q^{n\delta_{il}}\right)\left(-q^{s_i+1+n\delta_{il}}\right)_{m}}\prod_{j=1}^{i-1}{\left(q^{s_i - s_j}\right)_{1}\over\left(-q^{s_i - s_j+n\delta_{il}}\right)_{1 } } \prod_{j = i+1}^{n_1}{\left(q^{s_j - s_i}\right)_{1}\over\left(-q^{s_j - s_i - n\delta_{il}}\right)_{1}}\right ] \nn\end{aligned}\ ] ] with @xmath83 and the symbol @xmath84 is a shorthand notation for the @xmath85-pochhammer symbol defined in appendix [ qanalogs ] .
the choice of the integration contour @xmath86 will be discussed in detail in section [ intrep ] .
we note that the integral expression @xmath87 without winding agrees with that of the partition function in @xcite .
defined in @xcite is @xmath88 . ]
[ [ bullet - the-1over-6-bps - un_1-wilson - loop - with - n_1ge - n_2 ] ] @xmath78 the @xmath41-bps @xmath44 wilson loop with @xmath77 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in the case of @xmath77 the formula turns out to be slightly more involved and takes the form @xmath89 that is only valid for @xmath90 as will be elaborated in the comments below ) , on the other hand , does not have this restriction . ] and we defined @xmath91}{\pi ds_a\over\sin(\pi s_a)}\right ] q^{n(2c - m)}{\left(q^{1-n}\right)_c\left(q^{1+n}\right)_{m-1-c}\over ( q)_c(q)_{m-1-c } } \prod_{a=1}^{n_2}{\left(q^{s_a+1}\right)_m\over 2\left(-q^{s_a+1}\right)_{m}}\nn\\ & \times \prod_{a=1}^{n_2}\left[{\left(-q^{s_a+1+c}\right)_1\left(q^{s_a+1+c - n}\right)_1 \over \left(q^{s_a+1+c}\right)_1\left(-q^{s_a+1+c - n}\right)_1}\prod_{b=1}^{a-1}{\left(q^{s_a - s_b}\right)_{1 } \over\left(-q^{s_a - s_b}\right)_{1 } } \prod_{b = a+1}^{n_2}{\left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a}\right)_{1}}\right]\ , \label{integralii1}\end{aligned}\ ] ] and @xmath92 q^{-ns_d+n(d - m-2 ) } \prod_{\substack{a=1 \\
a\ne d}}^{n_2}{\left(q^{s_a - s_{d}-n}\right)_1\over \left(q^{s_a - s_{d}}\right)_1 } \label{integralii2}\\ & \times\prod_{a=1}^{n_2}\left[{\left(q^{s_a+1+n\delta_{ad}}\right)_m \over\left(1+q^{n\delta_{ad}}\right)\left(-q^{s_a+1}\right)_{m } } \prod_{b=1}^{a-1}{\left(q^{s_a - s_b}\right)_{1 } \over\left(-q^{s_a - s_b+n\delta_{ad}}\right)_{1 } } \prod_{b = a+1}^{n_2}{\left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a - n\delta_{ad}}\right)_{1}}\right]\nn\end{aligned}\ ] ] with @xmath93 .
note that @xmath94 and the normalization in ( [ 1/6wilsoni2 ] ) differs from that in ( [ 1/6wilsoni1 ] ) in that the ranks of the gauge groups @xmath2 and @xmath3 are exchanged .
the choice of the integration contours @xmath95\}$ ] ( @xmath96 ) and @xmath97 will be discussed in detail in section [ intrep ] .
we would , however , like to make a remark concerning the contour @xmath98 $ ] : there are subtleties in evaluating the integrals with the contour @xmath98 $ ] . in order to properly deal with them
, we shall adopt the @xmath99-prescription shifting the parameter @xmath100 with @xmath101 and the contour is placed between @xmath102 and @xmath103 .
related comments will be made in section [ abjmlimit ] below and appendix [ cancellation ] .
[ [ bullet - the-1over-6-bps - un_2-wilson - loops ] ] @xmath78 the @xmath41-bps @xmath45 wilson loops : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + it follows from the definition and symmetry that the @xmath41-bps wilson loop on the second gauge group @xmath45 with winding number @xmath38 is related to that on the first gauge group @xmath44 in a simple manner : @xmath104 [ [ bullet - the-1over-2-bps - wilson - loop ] ] @xmath78 the @xmath63-bps wilson loop : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + meanwhile , the ( normalized ) @xmath63-bps wilson loop is given by a linear combination of two ( normalized ) @xmath41-bps wilson loops , one on the first and the other on the second gauge group , and turns out to take a rather simple form @xmath105 \displaystyle q^{-{n^2\over 2}+n}{i^{(1)}(n_1,n_2;n)_{k}\over i^{(2)}(n_1,n_2;0)_k } & \mbox{for } & n_1 \ge n_2\ , \label{1/2wilson } \end{array } \right.\end{aligned}\ ] ] where the two terms @xmath106 and @xmath107 cancel out , as we will show later in section [ 1/2bpsproof ] .
[ [ bullet - comments - on - the - zero - winding - limit - nto-0 ] ] @xmath108 comments on the zero winding limit @xmath109 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + there are a few subtleties to be addressed in the expressions @xmath110 in ( [ integralii1 ] ) and @xmath111 in ( [ integralii2 ] ) . they are linked to the comments made below ( [ s1 ] ) and ( [ s2 ] ) concerning the ranges of the summations and the discussions in appendix [ cancellation ] . here
we focus on the subtlety in the range of the sum @xmath112 in ( [ integralii1 ] ) . in its original form
, the sum over @xmath113 is taken from @xmath114 to @xmath115 as derived in ( [ s1original ] ) in appendix [ analcontdetail ] . however , when @xmath116 , we can rewrite the sum ( [ s1original ] ) @xmath117 ( [ s2original ] ) , after passing it to the integral representation discussed in section [ intrep ] , by the sum ( [ integralii1 ] ) @xmath117 ( [ integralii2 ] ) .
in particular , the upper limit @xmath115 of the sum over @xmath113 can be replaced by @xmath118 . for @xmath119
this relies on the fact that the factor @xmath120 when @xmath116 and @xmath121 . since the factor
@xmath122 does not vanish when @xmath123 , it implies that the @xmath124 limit of ( [ integralii1 ] ) that violates the bound @xmath116 would not coincide with ( @xmath18 times ) the integral representation for the partition function in @xcite .
in fact , ( [ integralii1 ] ) in the @xmath124 limit simply vanishes , whereas ( [ integralii2 ] ) reduces to ( @xmath3 times ) the integral representation for the partition function .
[ [ bullet - comments - on - the - bound - on - winding - number - n ] ] @xmath108 comments on the bound on winding number @xmath38 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we observe that the integrals ( [ integrali ] ) and ( [ integralii2 ] ) diverge when @xmath125 . observed in @xcite . see more comments on the @xmath63-bps case . ]
as we will elaborate later , the contour @xmath86 for the integrals goes to imaginary infinity . for large imaginary @xmath126 and @xmath127 ,
the integrands become asymptotically @xmath128 and @xmath129 , respectively , that grow exponentially when @xmath130 and approach a constant when @xmath131 .
this implies that @xmath132-bps wilson loops are only well - defined for @xmath133 . in the @xmath63-bps case
, it might appear that the wilson loop is well - defined for all values of @xmath38 , since the integral ( [ integralii1 ] ) converges for any @xmath38 .
note , however , first that since ( [ integralii1 ] ) is periodic under the shift @xmath134 owing to the properties , @xmath135 and @xmath136 , the winding @xmath38 is restricted to the range @xmath137 .
moreover , as indicated in , a closer inspection shows that the @xmath63-bps wilson loop , in the abjm limit , diverges at the winding @xmath138 due to the factor @xmath139 that comes from the pole at @xmath102 .
this implies that in the case of @xmath140 , the only allowed winding is @xmath124 and thus the @xmath63-bps wilson loop does not exist . [ [ bullet - comments - on - the - abjm - limit ] ] @xmath108 comments on the abjm limit : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in the abjm limit the ranks of two gauge groups are equal , i.e. , @xmath141 , and the two results ( [ 1/6wilsoni1 ] ) and ( [ 1/6wilsoni2 ] ) coincide as they should .
however , the way they coincide turns out to be very subtle .
naively , it may look that @xmath142 vanishes and @xmath143 and @xmath144 coincide in the abjm limit .
however , a careful analysis of the abjm limit reveals that @xmath145 remains finite , but the sum @xmath146 coincides with @xmath143 in the limit . as will be elaborated in section [ abjmlimit ] , this is rooted in the difference of the integration contours @xmath86 in ( [ 1/6wilsoni1 ] ) and @xmath147 in ( [ 1/6wilsoni2 ] ) . since @xmath148 ,
the integral @xmath144 differs from @xmath143 even in the abjm limit , but this difference is canceled by @xmath142 .
note that it is very important that @xmath149 , since the @xmath63-bps wilson loop ( [ 1/2wilson ] ) exists nonvanishing in the abjm limit .
as we discussed in the introduction , the abj theory is conjectured to possess seiberg duality as given in . in our previous paper @xcite
, we have explicitly checked that the partition functions of a dual pair are equal to each other up to a phase factor .
the precise form of the phase factor was first conjectured by @xcite and later derived in @xcite .
the difference of the phase factors in dual pairs is now understood as an anomaly in large gauge transformations @xcite .
we would like to emphasize that not only does our formula for the partition function in @xcite confirm the proof in @xcite but it also allows us to understand how seiberg duality of the abj theory works in detail .
in particular , it was observed explicitly in @xcite that the perturbative and nonperturbative contributions to the partition function are exchanged under the duality . in section [ sdsection ]
we will see the same property in the duality of wilson loops .
[ [ bullet - the - duality - map - of-1over-6-bps - wilson - loops ] ] @xmath78 the duality map of @xmath41-bps wilson loops : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + using our formulas , we find the maps between wilson loops in the original and seiberg dual theories : @xmath150 for the @xmath41-th bps wilson loops , and the rank of the dual gauge group is denoted by @xmath151 .
[ [ bullet - the - duality - map - of-1over-2-bps - wilson - loops ] ] @xmath78 the duality map of @xmath63-bps wilson loops : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for the @xmath152-bps wilson loops , the duality map turns out to be very simple w_(n_1,n_2 ; n)_k = ( -1)^n w_(_2,n_1 ; n)_k .
[ 1/2duality ] note that these three relations are not independent , but one of them can be derived from the other two . in the following sections , we will refer to ( [ 1/6duality ] ) , ( [ flavorduality ] ) , and ( [ 1/2duality ] ) as the @xmath41-bps wilson loop duality , the flavor wilson loop duality , and the @xmath152-bps wilson loop duality , respectively .
sector is inert and can be regarded as a flavor group .
the wilson loops in ( [ flavorduality ] ) are only on the @xmath44 flavor group , hence the name , the flavor wilson loop duality . ]
we will vindicate these maps by a heuristic yet very useful argument based on the brane configuration in section [ branepicturesection ] and an alternative proof that is an application of the proof by kapustin and willett @xcite in appendix [ kapustinwillett ] .
we follow the same strategy as that employed in the computation of the abj partition function @xcite .
the outline of the derivation is as follows : \(1 ) we first compute wilson loops in the @xmath28 lens space matrix model , where the wilson loops are defined in analogy to those in the abj theory .
the matrix integrals are simply gaussian and can be done exactly .
\(2 ) we then analytically continue the rank of one of the gauge groups from @xmath3 to @xmath6 .
this maps the lens space matrix model to the abj matrix model @xcite . as in the case of the partition function
, the result so obtained is expressed in terms of formal series that is not well - defined in the regime of the strong coupling we are concerned with and only sensible as perturbative expansion with the generalized @xmath153-function ( polylogarithm ) regularization assumed .
\(3 ) similar to the case of the partition function , we can render the formal series perfectly well - defined by means of the sommerfeld - watson transform and the resultant expression is given in terms of @xmath154-dimensional integrals .
this is a nonperturbative completion in the sense that what renders the formal series well - defined is an inclusion of nonperturbative contributions , as will be elaborated later .
we define the ( un - normalized ) wilson loop on the first gauge group @xmath44 with @xmath38 windings in the lens space matrix model by @xmath155 where we have defined the expectation value of @xmath156 by @xmath157 where the normalization constant @xmath158 is given by _ lens = i^-2(n_1 ^ 2+n_2 ^ 2)n_1!n_2 ! .
as will be shown in detail in appendix [ lenswl ] , the eigenvalue integrals are simply gaussian and can be carried out exactly . introducing the normalized wilson loop , the end result takes the form w_lens^i(n_1,n_2;n)_k:=w_lens^i(n_1,n_2;n)_k_lens^i(n_1,n_2;0)_k = q^-n^22+ns(n_1,n_2;n)_ks(n_1 , n_2;0)_k , [ lensspacewilsonloop ] where the function @xmath159 is given by @xmath160\ , \nn\end{aligned}\ ] ] where @xmath161 .
we note that this is actually a finite sum : there is no contribution from @xmath162 , since the factor @xmath163 for @xmath164 vanishes .
we have deliberately rewritten the result in the form of an infinite sum that is suitable for the analytic continuation in the next section .
the @xmath41-bps wilson loop in abj theory can be obtained from the lens space wilson loop ( [ lensspacewilsonloop ] ) by means of the analytic continuation @xmath29 .
a little care is needed for the analytic continuation .
namely , the analytic continuation requires a regularization : we first replace @xmath3 by @xmath165 and send @xmath166 in the end .
as mentioned in section [ wilsonloopresults ] , we need to treat the cases @xmath32 and @xmath77 separately .
we leave most of the computational details in appendix [ analcontdetail ] . [ [ bullet - the-1over-6-bps - un_1-wilson - loop - with - n_1le - n_2 - 1 ] ] @xmath78 the @xmath41-bps @xmath44 wilson loop with @xmath32 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the function @xmath167 in ( [ lensspacesfunction ] ) is continued as @xmath168 where we have defined @xmath169 \nn\end{aligned}\ ] ] with @xmath83 .
note that in contrast to the lens space case , there is no truncation of summations and this sum is really an infinite sum .
in fact , this is not a convergent sum and becomes ill - defined for the value of @xmath170 that is of our actual interest .
thus this expression is at best a formal series and we shall render it well - defined on the entire @xmath85-plane by means of a type of sommerfeld - watson transform in the next section .
hence the analytic continuation yields an expression in terms of formal series _
naive:= _ 0w_lens^i(n_1,-n_2+;n)_k = q^-n^22+ns^abj(n_1,n_2;n)_ks^abj(n_1 , n_2;0)_k [ formalabj1/6wilsonloop ] for @xmath32 . after implementing the generalized @xmath153-function ( polylogarithm ) regularization
, this result agrees with the final result ( [ 1/6wilsoni1 ] ) only in perturbative expansion in the coupling constant @xmath171 .
until sommerfeld - watson like transform is performed , this result is not nonperturbatively complete . [ [ bullet - the-1over-6-bps - un_1-wilson - loop - with - n_1ge - n_2 - 1 ] ] @xmath78 the @xmath41-bps @xmath44 wilson loop with @xmath77 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + this case is slightly more involved than the previous case .
the analytic continuation of @xmath167 consists of two terms @xmath172 where we have defined @xmath173\ , \nn\\ s^{\rm abj}_{(2)}(n_1,n_2;n)_k&:={1\over n_2 ! }
\sum_{d=1}^{n_2}\sum_{\substack{1\le d_1,\cdots , d_{d-1},d_{d},\cdots , d_{n_2}\\-n+1\le d_d}}q^{-nd_d+n(d - m-1 ) } \prod_{\substack{a=1 \\
a\ne d}}^{n_2}{\left(q^{d_a - d_{d}-n}\right)_1\over \left(q^{d_a - d_{d}}\right)_1 } \label{s2}\\ & \times\prod_{a=1}^{n_2}\left[{(-1)^{d_a}\left(q^{d_a+n\delta_{ad}}\right)_m \over\left(1+q^{n\delta_{ad}}\right)\left(-q^{d_a}\right)_{m}}\prod_{b=1}^{a-1}{\left(q^{d_a - d_b}\right)_{1 } \over\left(-q^{d_a - d_b+n\delta_{ad}}\right)_{1 } } \prod_{b = a+1}^{n_2}{\left(q^{d_b - d_a}\right)_{1}\over\left(-q^{d_b - d_a - n\delta_{ad}}\right)_{1}}\right]\nn\end{aligned}\ ] ] with @xmath93 and we replaced the sum @xmath174 by @xmath112 provided that @xmath175 in the first line of ( [ s1 ] ) .
see the remark below ( [ integralii2 ] ) for the explanation .
the caveat on convergence and well - definedness of the sum noted in the previous case applies to this case as well .
note that when @xmath124 , ( [ s1 ] ) vanishes by definition .
an important remark is in order : in the first line of ( [ s1 ] ) , we replaced the sum @xmath174 by @xmath112 provided that @xmath175 and extended the range of @xmath176 s from @xmath177 to @xmath178 . in sync with this replacement and extension
, we extended the range of @xmath179 from @xmath180 to @xmath181 in the first line of ( [ s2 ] ) .
as shown in appendix [ cancellation ] , the added contributions conspire to cancel out in the sum @xmath182 , justifying the replacement and extensions we have made in ( [ s1 ] ) and ( [ s2 ] ) .
similar to the previous case , the analytic continuation yields an expression in terms of formal series _ naive&:= & _ 0w_lens^i(n_1,-n_2+;n)_k + & = & q^-n^22+ns^abj_(1)(n_1,n_2;n)_k+s^abj_(2)(n_1,n_2;n)_ks^abj_(2)(n_1 , n_2;0)_k [ formalabj1/6wilsonloop ] for @xmath77 .
note that @xmath183 as inferred from ( [ s1 ] ) and the denominator @xmath184 . again
this result is not complete as yet but agrees , after the generalized @xmath153-function regularization , with the final result ( [ 1/6wilsoni2 ] ) in perturbative expansion in the coupling constant @xmath171 .
[ [ bullet - the-1over-6-bps - un_2-wilson - loops-1 ] ] @xmath78 the @xmath41-bps @xmath45 wilson loops : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as stated in the summary of the main results , it follows from the definition and symmetry that the wilson loop on the second gauge group @xmath45 is obtained from that on the first gauge group @xmath44 as @xmath185_{\rm naive } = \left[w^{\rm i}_{\rm abj}(n_2,n_1 ; n)_{-k}\right]_{\rm naive}\ .\end{aligned}\ ] ] [ [ bullet - the-1over-2-bps - wilson - loop-1 ] ] @xmath78 the @xmath63-bps wilson loop : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the @xmath63-bps wilson loop is given by a linear combination of two @xmath41-bps wilson loops , one on the first and the other on the second gauge group : @xmath186_{\rm naive } & : = w^{\rm i}_{\rm abj}(n_1,n_2 ; n)_{k}-(-1)^nw^{\rm ii}_{\rm abj}(n_1,n_2 ; n)_{k}\nn\\ & = q^{-{n^2\over 2}+n}{s^{\rm abj}_{(1)}(n_1,n_2;n)_k \over s^{\rm abj}_{(2)}(n_1,n_2;0)_k } \label{formal1/2wilson}\end{aligned}\ ] ] for @xmath77 where we used the fact @xmath187 that we will show in section [ 1/2bpsproof ] . in the next subsection we discuss a nonperturbative completion of the above naive results that were given in terms of the formal series @xmath188 , @xmath189 and @xmath190 .
the analytic continuation in the previous section yielded tentative results for the abj wilson loops that involve formal series ( [ abjsfunction1 ] ) , ( [ s1 ] ) and ( [ s2 ] ) .
these are non - convergent formal series , since the summands do not vanish as @xmath191 s and @xmath176 s run to infinity . if we are only interested in perturbative expansion in the coupling @xmath192 , implementing the generalized @xmath153-function regularization _
s=0^(-1)^ss^n=\ { ll li_-n(-1)=(2^n+1 - 1)(-n ) = -2^n+1 - 1n+1b_n+1 & ( n1 ) , + 1+li_0(-1 ) = = -b_1 & ( n=0 ) , . [ genzeta ] where @xmath193 is the polylogarithm and @xmath194 are the bernoulli numbers , the formal series can be rendered convergent as in the case of the partition function @xcite .
the @xmath153-function regularized wilson loops so obtained indeed reproduce the correct perturbative expansions in @xmath73 .
however , when @xmath195 is a root of unity with @xmath196 that is the value we are actually interested in and beyond the perturbative regime , the sums ( [ abjsfunction1 ] ) , ( [ s1 ] ) and ( [ s2 ] ) diverge and require a nonperturbative completion . fortunately , as we have done so for the partition function @xcite , these problems can be circumvented by introducing an integral representation similar to the sommerfeld - watson transform : @xmath197\left(\dots\right)\ , \label{integral1}\\ \sum_{1\le d_1,\cdots , d_{n_2}}\prod_{a=1}^{n_2}(-1)^{d_a+1}\left(\dots\right ) & \quad\longrightarrow\quad \prod_{a=1}^{n_2}\left[{-1\over 2\pi i}\int_c{\pi ds_a\over\sin(\pi s_a)}\right]\left(\dots\right)\ , \label{integral2}\end{aligned}\ ] ] where @xmath191 is replaced by @xmath198 and @xmath176 by @xmath199 . as we will see shortly , _ this is a transformation that adds nonperturbative contributions missed in the formal series .
_ note , however , that this is a prescription that lacks a first principle derivation and needs to be justified . in the case of the partition function , this prescription has passed both perturbative and nonperturbative checks . in particular , the latter has confirmed the equivalence of seiberg dual pairs up to the aforementioned phase factors @xcite .
more recently , a direct proof of this prescription for the partition function was given by honda by utilizing a generalization of cauchy identity @xcite .
in the wilson loop case , although we are missing a similar derivation , the proof of seiberg duality in section [ sdsection ] provides convincing evidence for this prescription .
the contour @xmath86 of integration in ( [ integral1 ] ) and ( [ integral2 ] ) is chosen in order that ( 1 ) the perturbative expansion is correctly reproduced for small @xmath73 ( corresponding to large @xmath4 ) and , ( 2 ) as we decrease @xmath4 continuously , the values of integrals remain continuous as functions of @xmath4 .
these requirements yield the contour @xmath86 parallel and left to the imaginary axis . to elaborate on it ,
we look into the pole structure of integrands : [ [ bullet - the - pole - structure - for - n0 ] ] @xmath78 the pole structure for @xmath124 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + it is illustrative to first review the pole structure for the partition function , i.e. , the @xmath124 case . in this case , without loss of generality , we can assume @xmath200 .
it is very useful to classify the poles into two classes , perturbative ( p ) and nonperturbative ( np ) poles .
we are interested in the summand in ( [ abjsfunction1 ] ) with @xmath124 . by multiplying the factor
@xmath201 $ ] , this becomes the integrand with the replacement @xmath202 .
the p poles come from the factor @xmath203 $ ] , whereas the np poles are from the factors @xmath204 and @xmath205 for _ generic _ ( real non - integral ) values of @xmath4 .
hence the p poles are at s_j = , -m-2 , -m-1;0 , 1 , 2 , [ ppoles ] with the gap between @xmath206 and @xmath114 as shown in the left figures of fig .
[ contourfig1 ] . in , we organized poles into groups separated by a semicolon to clarify this gap structure . in the large @xmath4 limit , these are the only poles .
the contour @xmath86 parallel to the imaginary axis can be placed anywhere in the gap .
indeed , enclosing the contour with an infinite semi - circle to the right in the complex @xmath207 plane , the residue integral reduces to the sum ( [ abjsfunction1 ] ) with the generalized @xmath153-function regularization ( [ genzeta ] ) implemented automatically by the integral formula -12i_cds(s)s^n = -2^n+1 - 1n+1b_n+1 ( n0 ) , and the perturbative expansion in @xmath73 is correctly reproduced .
it should now be clear why this class of poles are called p poles .
the np poles are at @xmath208 this class of poles are called np poles because @xmath209 and thus the residues are of order @xmath210 . on the right panel of fig .
[ contourfig1 ] shown is the case @xmath211 and @xmath212 .
note that for _ integer _ @xmath4 as opposed to _
( non - integral ) values of @xmath4 , there are extra cancellations of the zeros and poles in the factor @xmath213 $ ] since the zeros s_j = k - m , k-(m-1 ) , , k-1k can coincide with some of the p poles in ( [ ppoles ] ) .
more precisely , the gap between @xmath206 and @xmath114 repeats itself periodically modulo @xmath4 and thus the p poles for an integer k appear at s_j = 0 , 1 , , k - m-1k .
[ ppoleskint ] now more important is the fact that for a given @xmath214 as we decrease @xmath4 continuously , the np poles on the positive real axis move to the left . for a sufficiently large @xmath215 ,
the np pole closest to the origin is at @xmath216 .
as @xmath4 is decreased from @xmath215 to @xmath217 , this pole crosses the imaginary axis to the left .
as we decrease @xmath4 further , more np poles cross the imaginary axis . for the partition function to be continuous in @xmath4 , these poles should not cross the contour @xmath86 and therefore we need to shift the contour @xmath86 to the left so as to avoid the crossing of these np poles that invade into the real negative region .
more precisely , the contour @xmath86 has to be placed between @xmath218 and @xmath219 when @xmath220 .
this is a prescription that needs to be justified .
in @xcite it was checked that seiberg duality holds with this contour prescription , vindicating our integral representation as a nonperturbative completion . for the partition function : ( a ) the left figure corresponds to the large @xmath4 limit where only perturbative ( p ) poles indicated by red `` @xmath117 '' are present .
( b ) the right figure is an example of the finite @xmath4 case ( @xmath211 and @xmath212 ) where nonperturbative ( np ) poles indicated by blue `` @xmath221 '' are also present .
the green dotted line corresponds to @xmath222 mod @xmath4 with some @xmath223 . note that some of p poles and zeros of integrands coalesce and cancel out for integer @xmath4 .
, title="fig:",height=172 ] for the partition function : ( a ) the left figure corresponds to the large @xmath4 limit where only perturbative ( p ) poles indicated by red `` @xmath117 '' are present .
( b ) the right figure is an example of the finite @xmath4 case ( @xmath211 and @xmath212 ) where nonperturbative ( np ) poles indicated by blue `` @xmath221 '' are also present .
the green dotted line corresponds to @xmath222 mod @xmath4 with some @xmath223 . note that some of p poles and zeros of integrands coalesce and cancel out for integer @xmath4 .
, title="fig:",height=172 ] [ [ bullet - the - pole - structure - for - n-0-and - n_1le - n_2 ] ] @xmath78 the pole structure for @xmath224 and @xmath32 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + it is straightforward to generalize the contour prescription to the case of wilson loops . in this case
, however , we need to discuss two cases @xmath32 and @xmath225 separately .
we start with the former that is simpler than the latter .
we are interested in the summand in ( [ abjsfunction1 ] ) . again by multiplying the factor @xmath201 $ ] , this becomes the integrand with the replacement @xmath202 . for the p poles
the relevant factor is the same as in the partition function , @xmath213 $ ] .
this yields the p poles for generic ( non - integral ) @xmath4 at s_j = , -m-2 , -m-1;0 , 1 , 2 , .
[ ppoles2 ] a similar remark on the integer @xmath4 case applies to this case , and the gap between @xmath226 and @xmath114 repeats itself periodically modulo @xmath4 .
this implies that the p poles for an integer @xmath4 appear at s_j = 0 , 1 , , k - m-1 k. [ ppoles2kint ] for the np poles the relevant factors are @xmath227 and @xmath228 where @xmath229 runs from @xmath230 to @xmath2 .
the np poles are thus at @xmath231 note that the np poles are simply shifted by @xmath232 as compared to those for the partition function .
thus the pole structure differs from that of the partition function only for the integration variable @xmath126 .
as mentioned in the end of section [ wilsonloopresults ] , the integral representation for the sum ( [ abjsfunction1 ] ) is only well - defined for @xmath133 .
we thus restrict @xmath38 in this range . in fig .
[ contourfig2 ] shown are both p and np poles as well as the contour @xmath86 both for @xmath233 and @xmath234 .
similar to the partition function , as we decrease @xmath4 , np poles on the positive real axis move to the left . for @xmath235 , in particular , when @xmath4 becomes smaller than @xmath236 ( for @xmath175 ) , the np pole closest to the p pole at the origin crosses the imaginary axis . for wilson loops to be continuous as a function of @xmath4 , similar to the partition function , the contour @xmath86 has to be placed between @xmath237 and @xmath238 for @xmath175 .
note that the bound @xmath239 ensures that @xmath240 is left to @xmath241 .
for @xmath233 the contour is the same as that in the partition function .
the pole structure and the integration contour @xmath86 in this case are shown in figure [ contourfig2 ] . for wilson loops with @xmath32 : shown is the case , @xmath211 , @xmath212 and @xmath42 . ( a ) the left figure is for @xmath207 with @xmath233 .
the contour is placed between @xmath242 and @xmath243 .
( b ) the right figure is for @xmath207 with @xmath234 and the contour is placed between @xmath244 and @xmath245 .
the green dotted line corresponds to @xmath246 mod @xmath4 with some @xmath223 .
, title="fig:",height=172 ] for wilson loops with @xmath32 : shown is the case , @xmath211 , @xmath212 and @xmath42 . ( a )
the left figure is for @xmath207 with @xmath233 .
the contour is placed between @xmath242 and @xmath243 .
( b ) the right figure is for @xmath207 with @xmath234 and the contour is placed between @xmath244 and @xmath245 .
the green dotted line corresponds to @xmath246 mod @xmath4 with some @xmath223 . , title="fig:",height=172 ] [ [ bullet - the - pole - structure - for - n-0-and - n_1ge - n_2 ] ] @xmath78 the pole structure for @xmath224 and @xmath77 : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in this case we are interested in the summands ( [ s1 ] ) and ( [ s2 ] ) .
again by multiplying the factor @xmath247 $ ] , these summands become the integrands with the replacement @xmath248 .
we first discuss the pole structure of ( [ s1 ] ) .
this time the relevant factor for the p poles is different from the previous cases , @xmath249 $ ] where @xmath113 runs from @xmath114 to @xmath118 .
this yields the p poles for generic @xmath4 at s_a = ,
-m-2 , -m-1;-1-c ; 0 , , -2-c+n ;- c+n , [ ppoles3 ] note that there is an additional pole at @xmath102 in the gap between @xmath250 and @xmath114 and a hole at @xmath251 . in the case of integer @xmath4 , the gap ,
the additional point @xmath102 and the hole @xmath252 repeat themselves modulo @xmath4 and the p poles appear at @xmath253 to be more precise , if @xmath113 is small enough and the hole at @xmath252 falls into a gap , the hole is absent .
for the np poles the relevant factors are @xmath254 and @xmath255 .
the np poles are thus at @xmath256 the choice of integration contour @xmath257 is more involved than the previous cases : ( 1 ) in addition to the p poles at @xmath258 for a large @xmath4 ( i.e. , in the perturbative regime ) , we need to include , for a given @xmath113 , the p pole at @xmath259 .. ] this means that we place the integration contour _ for a given _
@xmath113 to the left of the p pole at @xmath102 .
( 2 ) as in the previous cases , we require the continuity with respect to @xmath4 .
the np pole at @xmath260 invades into the negative real axis , as @xmath4 is decreased , whereas the np pole at @xmath261 moves to the right .
hence the contour has to be placed to the left of @xmath262 and the right of @xmath263 for @xmath264 and to the left of @xmath102 and the right of @xmath265 for @xmath266 . and @xmath267 , the np pole at @xmath261 is always left to the np pole at @xmath260 .
similarly , the p pole at @xmath102 is always right to the np poles at @xmath261 since @xmath239 . ]
the pole structure and the contour @xmath257 are illustrated in figure [ contourfig3 ] .
note that the @xmath257 depends on @xmath113 and may thus be denoted as @xmath95\}$ ] ( @xmath268 . $ ] for wilson loops with @xmath77 : the figure is for the integrals ( [ integralii1 ] ) associated with the sum ( [ s1 ] ) .
the green dotted line corresponds to @xmath269 mod @xmath4 with some @xmath270 .
shown is the case @xmath211 , @xmath212 , @xmath42 and @xmath271.,height=172 ] next we turn to the pole structure of ( [ s2 ] ) . for the p pole
the relevant factor is @xmath272 $ ] where @xmath46 runs from @xmath230 to @xmath3 .
this yields for generic @xmath4 the p poles at s_a = , -m-2-n_ad , -m-1-n_ad ; -n_ad , 1-n_ad ,
note that the p poles are shifted by @xmath273 as compared to those in the @xmath274 case .
for an integer @xmath4 the gap between @xmath275 and @xmath273 repeats itself periodically modulo @xmath4 and the p poles appear at s_a = -n_ad , 1-n_ad ,
, k - m-1-n_ad k .
it could happen that if the winding @xmath38 is sufficiently large , @xmath276 becomes negative . for the np poles the relevant factors are @xmath277 and @xmath278 .
the np poles are thus at @xmath279 the choice of integration contour @xmath97 is similar to the @xmath32 case except that the contour for the variable @xmath280 is to the left of @xmath281 and the right of @xmath282 and picks up , in particular , the residues from the p poles at @xmath283 .
the pole structure and the contour @xmath97 are illustrated in figure [ contourfig4 ] . for wilson loops with @xmath77 : ( a ) the left figure is the pole structure and the contour for @xmath284 .
( b ) the right figure is for @xmath285 and the contour is shifted by @xmath286 as compared to that in ( a ) .
the green dotted line corresponds to @xmath287 mod @xmath4 with some @xmath270 .
the contours are the same .
shown is the case @xmath211 , @xmath212 and @xmath42.,title="fig:",height=172 ] for wilson loops with @xmath77 : ( a ) the left figure is the pole structure and the contour for @xmath284 .
( b ) the right figure is for @xmath285 and the contour is shifted by @xmath286 as compared to that in ( a ) .
the green dotted line corresponds to @xmath287 mod @xmath4 with some @xmath270 .
the contours are the same .
shown is the case @xmath211 , @xmath212 and @xmath42.,title="fig:",height=172 ] again this is a prescription that lacks a first principle derivation . in the case of @xmath32 ,
similar to the partition function , for a large @xmath4 , enclosing the contour with an infinite semi - circle to the right in the complex @xmath207 plane , the residue integral reduces to the sum ( [ abjsfunction1 ] ) , with the generalized @xmath153-function regularization ( [ genzeta ] ) implemented automatically , and the perturbative expansion in @xmath73 is correctly reproduced . in the case of @xmath77 , however , the way this prescription works is more subtle even for a large @xmath4 .
each of the two integral expressions ( [ integralii1 ] ) and ( [ integralii2 ] ) picks up _ extra _ perturbative contributions from the poles at @xmath288 . for @xmath41-bps wilson loops these extra contributions
cancel out in the sum of ( [ integralii1 ] ) and ( [ integralii2 ] ) , as shown in appendix [ cancellation ] , thereby reproducing the correct perturbative expansion in @xmath73.-bps wilson loops of the lens space and abj matrix models , via the analytic continuation , has been established and checked by direct perturbative calculations .
] as the nonperturbative test , we show in the next section that seiberg duality holds with our prescription , where it becomes clear that the inclusion of the p poles at @xmath288 is necessary . as noted in section [ wilsonloopresults ] ,
there are subtleties in taking the abjm limit @xmath289 , in particular , in the formula ( [ integralii1 ] ) for the case @xmath77 .
there are two points to be addressed ; ( 1 ) the agreement of the two formulae ( [ 1/6wilsoni1 ] ) and ( [ 1/6wilsoni2 ] ) , and ( 2 ) the @xmath63-bps wilson loop ( [ 1/2wilson ] ) in the abjm limit . to address the first point , notice that the integrands of ( [ integrali ] ) and ( [ integralii2 ] ) in the limit @xmath289 become identical .
however , as remarked before , the contours @xmath86 and @xmath97 are different .
now since the factors @xmath290 $ ] in ( [ integrali ] ) and @xmath291 $ ] in ( [ integralii2 ] ) are absent , there are no np poles on the real axis .
hence the difference due to the contours @xmath86 and @xmath97 only comes from the residues at the p poles @xmath292 in ( [ integralii2 ] ) .
therefore , in order for ( [ 1/6wilsoni1 ] ) and ( [ 1/6wilsoni2 ] ) to agree , these residues have to be canceled by ( [ integralii1 ] ) . to see it , we carefully take the @xmath289 limit of ( [ integralii1 ] ) : @xmath293}{\pi ds_a\over\sin(\pi s_a)}\right ] q^{n(2c - m)}{\left(q^{1-n}\right)_c\left(q^{1+n}\right)_{m-1-c}\over ( q)_c(q)_{m-1-c } } \prod_{a=1}^{n_2}{\left(q^{s_a+1}\right)_m\over 2\left(-q^{s_a+1}\right)_{m}}\nn\\ & \times \prod_{a=1}^{n_2}\left[{\left(-q^{s_a+1+c}\right)_1\left(q^{s_a+1+c - n}\right)_1 \over \left(q^{s_a+1+c}\right)_1\left(-q^{s_a+1+c - n}\right)_1}\prod_{b=1}^{a-1}{\left(q^{s_a - s_b}\right)_{1 } \over\left(-q^{s_a - s_b}\right)_{1 } } \prod_{b = a+1}^{n_2}{\left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a}\right)_{1}}\right]\ , \label{integralii1v2}\end{aligned}\ ] ] in particular , we focus on the factors near the p pole for a selected variable @xmath280 at @xmath294 as @xmath295 , _
s_d-1-c(q^1-n)_c(q^1+n)_m-1-c(q)_c(q)_m-1-c(q^s_d+1)_m(q^s_d+1+c)_1 & = & q^-nc(q^n - c)_m(q^n)_1(q^-c)_c(q)_m-1-c_s_d-1-c(q^s_d+1)_m(q^s_d+1+c)_1 + & = & q^-nc(q^n - c)_m(q^n)_1 [ limitresidue ] where we used @xmath296 and @xmath297 . note that in the @xmath295 limit these factors vanish away from the p pole @xmath294 , i.e. , if the limit @xmath298 is not taken in the first line .
therefore , in the abjm limit the only contribution comes from the residues at the p poles @xmath294 where @xmath299 .
an important remark is in order : when @xmath300 , the pole at @xmath294 is clearly a simple pole , since the factor @xmath301 that appear in the first line of is canceled by the same factor in the @xmath85-pochhammer symbol @xmath302 .
however , when @xmath266 , it is subtle , because there may not seem to be no apparent cancellation of these factors .
nevertheless , we treat the pole at @xmath294 as a simple pole .
as it turns out , the proper way to deal with this subtlety is to adopt an @xmath99-prescription for the parameter @xmath18 .
namely , @xmath18 is always kept off an integral value by the shift @xmath100 with @xmath101 .
the factor @xmath302 is always assumed to be @xmath303 with a non - integral index in our calculations and is defined by ( [ maux20sep12 ] ) for a non - integral @xmath304 . with this prescription , the factor @xmath301 is always canceled by the same factor in @xmath303 , making the pole at @xmath294 a simple pole .
however , there is one more subtlety in this prescription to be clarified .
namely , there appears a pole at @xmath305 even with this prescription when @xmath266 .
if this pole were included within the contour , the contribution would have been canceled . in other words
, it would have been the same as treating the pole at @xmath294 as a double pole .
thus our @xmath99-prescription involves a particular choice of the contour @xmath98 $ ] , i.e. , to place it between @xmath294 and @xmath103 so as to avoid the latter pole .
this is a very subtle point and so much of a detail but is absolutely necessary for getting sensible results . without loss of generality
, we can choose the index @xmath46 to be @xmath230 , since the expression is invariant under permutations of @xmath306 s .
this yields @xmath307}{\pi ds_a\over\sin(\pi s_a)}\right ] { \left(q^{s_a+1+c}\right)_1\left(q^{s_a+1+c - n}\right)_1 \over \left(-q^{s_a+1+c}\right)_1\left(-q^{s_a+1+c - n}\right)_1}\nn\\ & \quad\times \prod_{b=2}^{a-1 } { \left(q^{s_a - s_b}\right)_{1}\over\left(-q^{s_a - s_b}\right)_{1 } } \prod_{b = a+1}^{n } { \left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a}\right)_{1}}\ . \label{integralii1abjm}\end{aligned}\ ] ] as shown in appendix [ cancellation ] , this is exactly canceled out by the sum of residues in @xmath308 at the p poles @xmath309 .
hence the formula ( [ 1/6wilsoni2 ] ) in the abjm limit reduces to @xmath308 with the contour @xmath97 being replaced by the contour @xmath86 .
this proves the agreement of ( [ 1/6wilsoni1 ] ) and ( [ 1/6wilsoni2 ] ) in the abjm limit .
we also note that the formula ( [ integralii1abjm ] ) , when multiplied by @xmath310 , yields the @xmath63-bps wilson loop ( [ 1/2wilson ] ) in the abjm limit ( up to a normalization ) .
we now discuss more on the @xmath63-bps wilson loop in the next subsection .
the @xmath63-bps wilson loop is given by ( [ 1/2wilson ] ) that follows from the equality q^-n^2+ni^(2)(n_1,n_2;n)_k=(-1)^nq^n^2-ni(n_2,n_1;n)_-k , [ ii2relation ] where @xmath77 . in order to show this identity
, we recall that @xmath311 q^{-ns_d+n(d - m-2 ) } \prod_{\substack{a=1 \\
a\ne d}}^{n_2}{\left(q^{s_a - s_{d}-n}\right)_1\over \left(q^{s_a - s_{d}}\right)_1 } \label{integralii2v2}\\ & \quad\times\prod_{a=1}^{n_2}\left[{\left(q^{s_a+1+n\delta_{ad}}\right)_m \over\left(1+q^{n\delta_{ad}}\right)\left(-q^{s_a+1}\right)_{m } } \prod_{b=1}^{a-1}{\left(q^{s_a - s_b}\right)_{1 } \over\left(-q^{s_a - s_b+n\delta_{ad}}\right)_{1 } } \prod_{b = a+1}^{n_2}{\left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a - n\delta_{ad}}\right)_{1}}\right]\nn\end{aligned}\ ] ] and @xmath312 q^{ns_d - n(d-2)}\prod_{\substack{a=1 \\ a\ne d}}^{n_2 } { \left(q^{-s_a+s_d+n}\right)_1\over \left(q^{-s_a+s_d}\right)_1 } \label{integraliconjugate}\\ & \quad \times\prod_{a=1}^{n_2}\left[{(-1)^{m}\left(q^{-s_a-1};q^{-1}\right)_m \over\left(1+q^{-n\delta_{ad}}\right)\left(-q^{-s_a-1-n\delta_{ad}};q^{-1}\right)_{m } } \prod_{b=1}^{a-1}{\left(q^{-s_a+s_b}\right)_{1}\over\left(-q^{-s_a+s_b - n\delta_{ad}}\right)_{1 } } \prod_{b = a+1}^{n_2}{\left(q^{-s_b+s_a}\right)_{1}\over\left(-q^{-s_b+s_a+n\delta_{ad}}\right)_{1}}\right]\ , \nn \nn\end{aligned}\ ] ] where @xmath93 .
in fact , it is straightforward to check that these two are related , precisely as in ( [ ii2relation ] ) , by the change of variables , t_a= -s_a-1-m - n_ad , [ ttos ] for a given @xmath46 , where @xmath306 s are the variables in the latter ( [ integraliconjugate ] ) and @xmath313 s be identified with those in the former ( [ integralii2v2 ] ) .
the contour @xmath86 is placed in the intervals , @xmath314 for @xmath284 and @xmath315 . by the above change of variables ,
this becomes precisely the contour @xmath97 , where @xmath316 for @xmath284 and @xmath317 . in ( [ ttos ] )
compensates the orientation flip of the contour . ]
this proves that the @xmath63-bps wilson loop is given by @xmath318 up to the normalization .
there is a duality between two abj theories @xcite .
schematically , when @xmath319 , the following abj theories are equivalent : u(n_1)_ku(n_2)_-k= u(2n_1-n_2+k)_ku(n_1)_-k .
[ sd2 ] the partition functions of the two theories agree up to a phase @xcite .
it was further understood in @xcite how the perturbative and nonperturbative contributions to the partition function are exchanged under the duality map .
the wilson loops , in contrast , are not invariant under the duality .
the mapping rule for @xmath152-bps wilson loops in general representations in @xmath47 csm theories with a simple gauge group has been studied by kapustin and willett @xcite .
these wilson loops correspond to @xmath41-bps wilson loops in the abj theory .
our results are consistent with their rule and slightly generalize it to the case where the flavor group is gauged .
similar to the case of the partition function @xcite , our formulae for the wilson loops allow us to understand an important aspect of the duality , namely , how the perturbative and nonperturbative contributions are exchanged under the duality map .
in this section we provide a proof of the duality map by analyzing our expressions ( [ 1/6wilsoni1 ] ) , ( [ 1/6wilsoni2 ] ) and ( [ 1/2wilson ] ) for the wilson loops . to this end , let us recall our results for the duality map of the wilson loops , ( [ 1/6duality ] ) , ( [ flavorduality ] ) , and ( [ 1/2duality ] ) . [ [ bullet - the-1over-6-bps - wilson - loop - duality ] ] @xmath78 the @xmath41-bps wilson loop duality : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + @xmath320 [ [ bullet - the - flavor - wilson - loop - duality ] ] @xmath78 the flavor wilson loop duality : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + @xmath321 [ [ bullet - the-1over-2-bps - wilson - loop - duality ] ] @xmath78 the @xmath63-bps wilson loop duality : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + w_(n_1,n_2 ; n)_k = ( -1)^nw_(_2,n_1 ; n)_k , [ 1/2duality2 ] where we denoted the rank of the dual gauge group by @xmath151 .
these three relations are not independent , but one of them can be derived from the other two by using the relation between the @xmath63- and @xmath41-bps wilson loops w_(n_1,n_2 ; n)_k = w^i_16(n_1,n_2 ; n)_k-(-1)^nw^ii_16(n_1,n_2 ; n)_k .[1/21/6relation ] before going into a rigorous technical proof , we provide a heuristic yet very useful and intuitive way to understand how the wilson loops would be mapped . in refs .
@xcite , the brane realization of the abj(m ) theory was proposed .
the brane content of this configuration is given by the following : @xmath322_\theta \left[{4\atop 8}\right]_\theta \left[{5\atop 9}\right]_\theta \end{array}\label{mbxq12apr13}\end{aligned}\ ] ] here , @xmath323 is periodically identified , and @xmath324_\theta$ ] means the direction on the 3 - 7 plane with an angle @xmath325 with the @xmath326 axis , where @xmath327 . for @xmath328 theory with @xmath329 , there are @xmath2 d3-branes between an ns5-brane and a @xmath330 5-brane , and @xmath331 d3-branes between the @xmath330 and the ns5 .
this system realizes 3d supersymmetric field theory that lives in the 012 directions and flows in the ir to the abj scft .
seiberg duality corresponds to moving the ns5 and the @xmath330 branes past each other and , during the process , @xmath4 d3-branes are created by the hanany - witten effect @xcite while @xmath18 d3-branes are annihilated , in the end leaving d3-branes realizing the dual theory . for our purpose , it is convenient to consider the following m - theory lift of the configuration , in which wilson loops are geometrically realized @xcite .
assume that we have non - trivial wilson loop along e.g. @xmath332 , namely @xmath333 .
is not clear .
this is one of the reasons why the argument presented here is heuristic . ]
if we t - dualize the configuration along @xmath332 and further lift it to 11 dimensions , we obtain @xmath334_\theta \left[{3\atop 7}\right]_\theta \left[{4\atop 8}\right]_\theta \left[{5\atop 9}\right]_\theta \end{array } \end{split}\end{aligned}\ ] ] where `` a '' denotes the 11th direction .
note that the @xmath330 5-brane has lifted to an m5-brane ( denoted by m5@xmath335 ) that is tilted in four 2-planes with the same angle @xmath325 . in figure
[ fig : mlift1 ] , we schematically described this configuration . _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ) .
the left and right ends of the configuration are identified .
the position of the m2-branes in the @xmath332 direction corresponds to the wilson loop . among @xmath4 places in which fractional m2-branes can end , the occupied ones
are denoted by @xmath108 and the unoccupied ones by @xmath336 .
[ fig : mlift1],height=151 ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ because m5@xmath335-branes are tilted in the @xmath332-@xmath337 plane , there are only @xmath4 places in which `` fractional '' m2-branes can stretch between m5@xmath335 and m5 .
note that only one fractional m2-brane can exist in one place because of the s - rule @xcite .
@xmath338 fractional m2-branes are distributed among these @xmath4 places . on the other hand , @xmath2
`` entire '' m2-branes are going around the @xmath323 direction and they do not have to sit in these places but can be anywhere .
because of the wilson loop , different m2-branes are located at different positions along the @xmath332 direction .
let the @xmath332 coordinate of the @xmath2 m2-branes between m5 and m5@xmath335 be @xmath339 , @xmath340 , and that of the @xmath331 m2-branes ( both fractional and entire ) between m5@xmath335 and m5 be @xmath341 , @xmath342 ( see figure [ fig : mlift1 ] ) .
note that there is no direct relation between the @xmath343 here and the ones that appear in the abj matrix model . in the computation of wilson loops using localization @xcite , at saddle points , gauge fields ( which correspond to @xmath343 here )
vanish and wilson loops get contribution only from auxiliary fields ( which correspond to @xmath343 in the abj matrix model ) . ] furthermore , let the @xmath332 coordinate of the @xmath4 places in which fractional m2-branes can end be @xmath344 , @xmath345 .
if the radius of the @xmath332 direction is @xmath346 , we have @xmath347 and , for @xmath348 , @xmath349 as mentioned above , seiberg duality corresponds to moving m5 and m5@xmath335 past each other . in this process , @xmath18 fractional m2-branes get annihilated , and @xmath4 fractional m2-branes are created , leaving @xmath350 .
the resulting configuration is shown in figure [ fig : mlift2 ] .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ .
[ fig : mlift2],height=151 ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ let the @xmath332 coordinate of the @xmath351 m2-branes ( both fractional and entire ) between m5 and m5@xmath335 be @xmath352 , and that of the @xmath17 m2-branes between m5@xmath335 and m5 be @xmath353 ( see figure [ fig : mlift2 ] )
. in the original configuration in figure [ fig : mlift1 ] , among the @xmath4 spots @xmath354 at which fractional m2-branes can end , let the occupied ones be @xmath355 and the unoccupied ones be @xmath356 .
of course , @xmath357 .
clearly , @xmath358 in the dual theory , the position of the entire m2-branes are unchanged , while the occupied and unoccupied spots for fractional m2-branes are interchanged . therefore ,
@xmath359 now we want to use this picture to give a very heuristic explanation of seiberg duality .
consider the original configuration in figure [ fig : mlift1 ] .
the wilson loop in the fundamental representation simply measures the position of the m2-branes .
therefore , naively , we have '' . in reality , @xmath341 is a variable to be integrated over and is not localized at @xmath360 .
even if there is a sense in which they are localized at @xmath360 , we should sum over all possible ways to distribute @xmath18 fractional m2-branes over @xmath4 positions . ]
@xmath361 where in the second equation we used .
actually , it turns out that , in order to reproduce the explicit results obtained in the current paper , we must set @xmath362 by hand so that is replaced by @xmath363 in the dual theory , using and , we obtain @xmath364 where we introduced another ad hoc rule @xmath365 just as we did in
. also , in the second line of , we used , assuming that @xmath366 mod @xmath4 . therefore , comparing , and , , we `` derived '' the following duality relations : @xmath367 this means that the 1/2-bps wilson loop defined by @xmath368 is expected to have the following simple transformation rule : @xmath369 note that the above arguments are based on the identity and are valid only for @xmath370 mod @xmath4 .
we do not expect to get correct equations by setting @xmath124 in the above duality relations , as we commented in subsection [ wilsonloopresults ] . , but does not . ]
although the ad hoc rule @xmath371 , @xmath372 was crucial to reproduce the correct transformation rule for wilson loops , its physical meaning is unclear .
it is somewhat reminiscent of the fact that , in the abj matrix model at large @xmath373 @xcite , the eigenvalue distribution for @xmath45 is offset relative to the @xmath44 eigenvalue distribution on the complex eigenvalue plane , but further investigations are left for future research . since the arguments given in this subsection are meant to be only heuristic , we simply accept the rule as a working assumption and proceed . in passing , we note that , with the above ad - hoc rule , the 1/2-bps wilson loop can be understood simply as supertrace as follows : @xmath374 where @xmath375 , @xmath376 , and @xmath377 . the above heuristic method to guess the seiberg duality relation can be generalized to more general representations .
representation with a young diagram @xmath378 , the wilson loop is @xmath379 , where @xmath380 is the schur polynomial @xcite .
for example , @xmath381 , @xmath382 , @xmath383 .
] for example , in the original @xmath1 theory , consider the following 1/6-bps wilson loops : @xmath384 where @xmath385 denotes the wilson loop in the representations @xmath386 and @xmath387 for @xmath44 and @xmath45 , respectively , and `` @xmath108 '' means the trivial representation . also , in the last expression of each line , we used a schematic notation , whose meaning is defined by the immediately preceding expression . in the dual @xmath388 theory , we have @xmath389 the duality relation between @xmath390 and @xmath391
is readily found to be @xmath392 the combinations that have simple transformation rule are @xmath393 which transform as @xmath394 these are precisely the 1/2-bps wilson loops derived in @xcite . just as in , we can write these in terms of supertrace as follows : @xmath395 note that the right hand side is nothing but supertrace in the respective representations .
more generally , the 1/2-bps wilson loop for general representation @xmath386 is given by ] where @xmath396 are the @xmath343 that appear in matrix model ( see footnote [ ftnt : munu ] ) .
our ad hoc rule is to replace this with @xmath397 where @xmath343 are the positions of the m2-branes . ]
@xmath398 where @xmath399 is a polynomial in @xmath400 ( @xmath401 ) obtained from the schur polynomial .
each term in the polynomial contains a product of @xmath402 @xmath403 s .
the results from the previous section , more specifically and , say that @xmath404 under duality .
therefore , the transformation law for the general 1/2-bps wilson loop is @xmath405 none of the above is a derivation of seiberg duality for 1/6-bps wilson loops but is merely a motivation for it .
however , the fact that it predicts a simple duality law for 1/2-bps wilson loops is evidence that the 1/6-bps duality relation is also correct . since the mapping ( [ flavorduality2 ] ) for the flavor wilson loop and ( [ 1/2duality2 ] ) for the @xmath63-bps wilson loop are simple as compared to the mapping ( [ 1/6duality2 ] ) for @xmath41-bps wilson loops , it is the best strategy to prove ( [ flavorduality2 ] ) and ( [ 1/2duality2 ] ) and then infer ( [ 1/6duality2 ] ) from them .
in due course , we will also see manifestly the exchange of perturbative and nonperturbative contributions under the duality .
[ [ bullet - the - flavor - wilson - loop - duality-1 ] ] @xmath78 the flavor wilson loop duality : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we first prove the flavor wilson loop duality : @xmath406 which amounts to the equality = q^n^22-ni(n_1,_2;n)_-ki(n_1,_2;0)_-k .
[ flavortobeproved ] the basic idea for the proof is to show that ( 1 ) the integrands in the numerators are identical , i.e. , they share exactly the same zeros and poles and the same asymptotics up to the normalization , and ( 2 ) the contours are equivalent .
the explicit forms of @xmath143 and @xmath407 are given by @xmath408 q^{-ns_l+n(l-2)}\prod_{\substack{i=1 \\
i\ne l}}^{n_1 } { \left(q^{s_i - s_{l}-n}\right)_1\over \left(q^{s_i - s_{l}}\right)_1 } \label{integralidual}\\ & \quad\times\prod_{i=1}^{n_1}\left[{(-1)^{m}\left(q^{s_i+1}\right)_m \over\left(1+q^{n\delta_{il}}\right)\left(-q^{s_i+1+n\delta_{il}}\right)_{m}}\prod_{j=1}^{i-1}{\left(q^{s_i - s_j}\right)_{1}\over\left(-q^{s_i - s_j+n\delta_{il}}\right)_{1 } } \prod_{j = i+1}^{n_1}{\left(q^{s_j - s_i}\right)_{1}\over\left(-q^{s_j - s_i - n\delta_{il}}\right)_{1}}\right ] \nn\end{aligned}\ ] ] @xmath409 q^{ns_l - n(l+m-3)}\prod_{\substack{i=1 \\
i\ne l}}^{n_1 } { \left(q^{-s_i+s_l+n}\right)_1\over \left(q^{-s_i+s_l}\right)_1 } \label{integraliconjugatedual}\\ & \quad\times\prod_{i=1}^{n_1}\left[{\left(q^{s_i+1}\right)_{k - m } \over\left(1+q^{n\delta_{il}}\right)\left(-q^{s_i+1+n\delta_{il}}\right)_{k - m } } \prod_{j=1}^{i-1}{\left(q^{-s_i+s_j}\right)_{1}\over\left(-q^{-s_i+s_j - n\delta_{il}}\right)_{1 } } \prod_{j = i+1}^{n_1}{\left(q^{-s_j+s_i}\right)_{1}\over\left(-q^{-s_j+s_i+n\delta_{il}}\right)_{1}}\right ] , \nn \nn\end{aligned}\ ] ] where @xmath83 . at first glance
the zeros and poles of ( [ integralidual ] ) and ( [ integraliconjugatedual ] ) differ in the @xmath18-dependence , since @xmath18 is replaced by @xmath410 in the latter . however , introducing the dual variables in the latter _
i =- s_i+k2 - 1-n_il , [ dualmap ] we can easily see that they actually agree .
as simple as it may look , we stress that this is a very important map and can be regarded as the duality transformation , as we now justify . in terms of the original variables
@xmath223 , the poles of the integrand in ( [ integraliconjugatedual ] ) appear at s_i & = & 0 , 1 , , m-1k , + * np:*s_i & = & -k2+m - n_il , , k2 - 1-n_ilk , + * ( np):*s_i & = & s_j + k2-n_ilk . in terms of the dual variables @xmath411 , these poles are mapped to _
i & = & k2-m - n_il , , k2 - 2-n_il , k2 - 1-n_ilk , + * p:*_i & = & 0 , 1 , , k - m-1 k , + * ( np):*_i & = & _ j + k2-n_ilk .
these are precisely the same as the poles of the integrand in ( [ integralidual ] ) and thus the two integrands share exactly the same poles .
we emphasize that the p and np poles are exchanged by the duality transformation ( [ dualmap ] ) . in terms of the original variables ,
the contour @xmath412 is placed in the intervals , @xmath413 for @xmath414 and @xmath415 . in terms of the dual variables ,
this becomes the intervals , @xmath416 for @xmath414 and @xmath417 .
hence , the contours @xmath86 and @xmath412 are equivalent . in ( [ dualmap ] ) compensates the orientation flip of the contour . ]
meanwhile , the zeros appear only from the factors @xmath418 and @xmath419 and do not depend explicitly on @xmath18 .
it is easy to check that the zeros are invariant under the duality transformation ( [ dualmap ] ) .
indeed , the factors that depend on the differences @xmath420 in ( [ integralidual ] ) and @xmath421 in ( [ integraliconjugatedual ] ) only differ from each other by the factor @xmath422 multiplying the latter .
it remains to find the asymptotics of the integrands .
the asymptotics to be compared with are those at @xmath423 in ( [ integralidual ] ) and @xmath424 in ( [ integraliconjugatedual ] ) , and for the factors that depend on the differences of the variables we only need to care about the factor @xmath422 . collecting various factors together and taking into account the orientations of the contours , it is straightforward to find that the latter asymptotics is @xmath425 times the former .
since the factor @xmath426 is canceled by the same factor coming from the normalization , this completes the proof of the equality ( [ flavortobeproved ] ) .
[ [ bullet - the-1over-2-bps - wilson - loop - duality-1 ] ] @xmath78 the @xmath63-bps wilson loop duality : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we next prove the duality for the @xmath63-bps wilson loop : w_(n_1,n_2 ; n)_k = ( -1)^nw_(_2,n_1 ; n)_k [ 1/2duality3 ] which from ( [ 1/2wilson ] ) amounts to the equality = -q^-n^22+ni^(1)(_2,n_1;n)_ki^(2)(_2,n_1;0)_k .
[ 1/2bpstobeproved ] the explicit forms of @xmath427 and @xmath428 are given by @xmath429}{\pi ds_i\over\sin(\pi s_i)}\right ] q^{n}{(q^{1-n})_c(q^{1+n})_{m-1-c}\over ( q)_c(q)_{m-1-c } } \prod_{i=1}^{n_1}{\left(q^{-s_i-1};q^{-1}\right)_m\over 2\left(-q^{-s_i-1};q^{-1}\right)_{m}}\nn\\ & \qquad\times \prod_{i=1}^{n_1}\left[{\left(-q^{-s_i-1-c}\right)_1\left(q^{-s_i-1-c - n}\right)_1 \over \left(q^{-s_i-1-c}\right)_1\left(-q^{-s_i-1-c - n}\right)_1}\prod_{j=1}^{i-1}{\left(q^{-s_i+s_j}\right)_{1 } \over\left(-q^{-s_i+s_j}\right)_{1 } } \prod_{j = i+1}^{n_1}{\left(q^{-s_j+s_i}\right)_{1}\over\left(-q^{-s_j+s_i}\right)_{1}}\right]\ , \label{integralii1v3}\end{aligned}\ ] ] @xmath430}{\pi ds_i\over\sin(\pi s_i)}\right ] q^{n(2c+m)}{(q^{1-n})_c(q^{1+n})_{k - m-1-c}\over ( q)_c(q)_{k - m-1-c } } \prod_{i=1}^{n_1}{\left(q^{s_i+1}\right)_{k - m}\over 2\left(-q^{s_i+1}\right)_{k - m}}\nn\\ & \times \quad\prod_{i=1}^{n_1}\left[{\left(-q^{s_i+1+c}\right)_1\left(q^{s_i+1+c - n}\right)_1 \over \left(q^{s_i+1+c}\right)_1\left(-q^{s_i+1+c - n}\right)_1}\prod_{j=1}^{i-1}{\left(q^{s_i - s_j}\right)_{1 } \over\left(-q^{s_i - s_j}\right)_{1 } } \prod_{j = i+1}^{n_1}{\left(q^{s_j - s_i}\right)_{1}\over\left(-q^{s_j - s_i}\right)_{1}}\right]\ .
\label{integralii1v4}\end{aligned}\ ] ] similar to the flavor wilson loop duality , introducing the duality transformation in ( [ integralii1v3 ] ) _ i =- s_i - k2 - 1 , [ dualmap2 ] it becomes evident that the two integrands share the same zeros and poles . in terms of the original variables
@xmath223 , the poles of the integrand in ( [ integralii1v3 ] ) appear modulo @xmath4 at s_i & = & -1-c;0 , , -2-c+n ;- c+n , , k - m-1 , + * np:*s_i & = & k2-m , , k2 - 2-c ;- c, ,k2 - 1;+n-1-c , + * ( np):*s_i & = & s_j + k2 . in terms of the dual variables @xmath411 , these poles are mapped , after shifting by @xmath431 , to _ i & = & -k2+m , , k2 - 2-;-, ,k2 - 1;+n-1- , + * p:*_i & = & -1-;0 , , -2-+n;-+n , , m-1 , + * ( np):*_i & = & _ j + k2 , where @xmath432 .
indeed , these are exactly the poles of the integrand in ( [ integralii1v4 ] ) .
we again stress that the p and np poles are exchanged by the duality transformation ( [ dualmap2 ] ) .
the contour @xmath257 is placed in the interval , @xmath433 , that is mapped to @xmath434.more precisely speaking , these ranges are for @xmath264 as discussed in section [ intrep ] , but for @xmath266 the intervals are not conditional , @xmath435 and @xmath436 . ] hence the contour in terms of the variables @xmath411 is equivalent to @xmath437 . to the left of @xmath102 and the right of @xmath265 for @xmath266 as for the zeros , similar to the flavor wilson loop case
, they appear only from the factors that depend on the differences @xmath421 in ( [ integralii1v3 ] ) and @xmath420 in ( [ integralii1v4 ] ) .
in fact , these factors are exactly the same in ( [ integralii1v3 ] ) and ( [ integralii1v4 ] ) . in order to examine the asymptotics , we first note that , only @xmath271 gives a nonvanishing result in ( [ sindepfactor ] ) . the left hand side yields @xmath438 that equals the right hand side , @xmath439 , with a minus sign .
more generally , we find @xmath440 ] q^n(2+m)(q^1-n)_(q^1+n)_k - m-1-(q)_(q)_k - m-1- & = & q^n(n-1)+n(m - c)_0(q^)_1(q^n-)_k - m ( q^n)_1(q^-)_k - m + & = & -q^n(n-1)(q^1-n)_c(q^1+n)_m-1-c(q)_c(q)_m-1-c .
[ sindepfactor ] the asymptotics at @xmath424 in ( [ integralii1v3 ] ) and @xmath423 in ( [ integralii1v4 ] ) then differ only by the factor @xmath441 and the factor @xmath442 is canceled by the same factor from the normalization .
this completes the proof of ( [ 1/2bpstobeproved ] ) .
[ [ bullet - the-1over-6-bps - wilson - loop - duality-1 ] ] @xmath78 the @xmath41-bps wilson loop duality : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + having proved the two of the duality maps ( [ flavorduality2 ] ) and ( [ 1/2duality2 ] ) , the definition of the @xmath63-bps wilson loop ( [ 1/21/6relation ] ) implies the duality map ( [ 1/6duality2 ] ) for the @xmath41-bps wilson loops : an easiest way to see it is to add @xmath443 to the both sides of ( [ 1/6duality2 ] ) , w_(n_1,n_2;n)&=&-(w^i_16(_2,n_1 ; n)_k-(-1)^nw^i_16(n_1,n_2;n)_k ) + & & -2(-1)^n+1(w^ii_16(_2,n_1 ; n)_k - w^i_16(n_1,n_2;n)_k ) .using the flavor wilson loop duality ( [ flavorduality2 ] ) and the relation ( [ 1/21/6relation ] ) , this yields the @xmath63-bps wilson loop duality , w_(n_1,n_2;n)=(-1)^nw_(_2,n_1;n ) .
this thereby proves the @xmath41-bps wilson loop duality ( [ 1/6duality2 ] ) .
it is illustrative to work out the simplest case , the duality between the @xmath445 and @xmath446 abj theories , since the integral is one - dimensional and the integration can be explicitly carried out .
apart from its simplicity , the @xmath445 may also be relevant to the study of vasiliev s higher spin theory with @xmath447 symmetry in the t hooft limit , @xmath448 with @xmath449 fixed@xcite .
here we perform the integrals analytically and explicitly check seiberg duality for some small values of @xmath17 , @xmath4 and @xmath38 . in this connection , in the simplest case of @xmath450 , a check against the direct integral is provided for arbitrary @xmath4 and @xmath38 in appendix [ u1u1check ] .
we first consider the duality of flavor wilson loops that are @xmath34-bps wilson loops on the `` flavor group '' @xmath447 .
the @xmath34-bps flavor wilson loops are given by @xmath451 where the dual gauge group @xmath452 and the integral expression takes the form @xmath453 the function @xmath454 appearing in the denominators is essentially the partition function and has the following property under seiberg duality @xcite : @xmath455 meanwhile , the flavor wilson loop duality is given by w^ _
( 1,n;n)_k = w^_(,1;n)_k that implies @xmath456 this generalizes the relation for the partition function , and we are going to check this relation explicitly for some small values of @xmath17 , @xmath4 and @xmath38 .
we first calculate the odd @xmath4 case , since it is simpler than the even @xmath4 case .
the integrand flips the sign under the shift of integration variable , @xmath457 , in this case .
thanks to this property , the integrals can be evaluated by considering the closed contour depicted in fig.[contour1 ] : let us denote the original contour by @xmath86 and the closed contour by @xmath458 that consists of @xmath459 where @xmath97 is parallel to @xmath86 and shifted by @xmath4 .
it is easy to see that there are no contributions from the contours @xmath257 and @xmath460 , and the integral along @xmath97 is precisely the same as that along the original contour @xmath86 because the integrand only flips the sign under the shift @xmath461 .
it is then clear that the integral along the closed contour @xmath458 is twice that along @xmath86 : @xmath462 thus the integrals can be evaluated by residue calculations .
is shifted to the left from the vertical line running from @xmath463 to @xmath464 , exploiting the absence of poles on the strip between @xmath465 and @xmath466 . , width=377 ] it is straightforward to carry out the calculation and we find that @xmath467\ , \nn\end{aligned}\ ] ] where the first term is the contribution from the p poles at @xmath468 , while the second term is from the np poles at @xmath469 .
the even @xmath4 case requires more considerations .
in contrast to the odd @xmath4 case , the integrand is periodic under the shift @xmath457 .
in addition , some of p and np poles merge into double poles , since @xmath470 is an integer .
we can , however , apply a similar trick as that used in @xcite . for an illustration of this trick ,
let us consider a generic integral of the form @xmath471 with @xmath472 being periodic under the shift @xmath473 .
the trick is instead to consider the integrand @xmath474 with @xmath475 being an arbitrary constant .
this integrand shifts as @xmath476 when @xmath477 is shifted by @xmath4 .
thus the integral @xmath478 along the closed contour @xmath479 yields @xmath480 provided that there are no contributions from @xmath257 and @xmath460 . this way
the even @xmath4 case can also be evaluated by residue calculations .
note that the result does not depend on the choice of an arbitrary constant @xmath475 . in the following we
only show the case when @xmath481 .
the other case , @xmath482 , however , can be easily derived in a similar manner .
the poles encircled by the contour appear at @xmath483 the residue evaluation then yields @xmath484\ , \nn\end{aligned}\ ] ] where the first line is the contribution from simple poles and the second line from double poles .
these simple poles are a subgroup of p poles in , while the double poles are composed of p and np poles , i.e. , those p poles in coalescing np poles in . 0
the pole structure is classified as follows . @xmath485
[ even k case is under construction ... ] on the dual side , the integral @xmath486 can be calculated similarly as @xmath487\ .\nn\end{aligned}\ ] ] in the even @xmath4 case we only show result for the case when @xmath488 : @xmath489\ .\nn\end{aligned}\ ] ] we have now collected necessary data to explicitly check the flavor wilson loop duality .
let us check explicitly for the duality @xmath490 with winding @xmath491 .
the results are @xmath492 both in the original and dual theories , the first line is the perturbative contribution , while the second line is the nonperturbative contribution .
we can explicitly see that the perturbative and nonperturbative contributions are exchanged under seiberg duality .
next we consider the even @xmath4 case , @xmath493 with winding @xmath42 .
for the original theory we have @xmath494 \label{i(12)1 } \\ & = \frac{i-1}{16}\left[-2+\frac{2 - 2 i}{\pi } \right ] , \label{i(12)}\end{aligned}\ ] ] where the first term in the first line comes from the double pole at @xmath495 and the rest are from the simple p poles at @xmath496 .
note that the @xmath475-dependence is canceled out , as it should .
for the dual theory we have @xmath497 \label{i(14)1}\\ & = \frac{i-1}{16 } \left [ -2 + \frac{2 - 2 i}{\pi } \right]\ , \label{i(14)}\end{aligned}\ ] ] where the first term in the first line comes from the double pole at @xmath495 and the rest are from the simple np poles at @xmath498 .
the @xmath499-dependence is canceled out similar to the previous case .
however , we observe that the pole - by - pole maps agree if we identify @xmath500 even though the constants @xmath475 and @xmath499 do not seem to carry any physical meaning . as indicated ,
the p and np poles are exchanged in the original and dual wilson loops .
these examples provide evidence for the flavor wilson loop duality
. 0 first , on the original theory side , @xmath501 has the poles at @xmath502 second , on the dual theory side , @xmath503 has the following pole structure .
@xmath504 if we change the integration variable on the dual side as @xmath505 , we precisely obtain the same pole structure as the original side . in order to complete the proof of seiberg duality we need to analyze the asymptotic behavior .
let us consider the asymptotic region @xmath506 on the original theory side .
this corresponds to the limit @xmath507 on the dual side .
the integrand asymptotes to @xmath508 combining this with the partition function ratio , we find that seiberg duality holds ; @xmath509 .
we next turn to the @xmath35-bps wilson loops with winding @xmath38 @xmath510 where the integral expression takes the form @xmath511}{\pi ds\over\sin(\pi s ) } { \left(q^{s+1}\right)_{n-1}\over 2\left(-q^{s+1}\right)_{n-1}}\nn\\ & \qquad\times{\left(-q^{s+1+c}\right)_1\left(q^{s+1+c+n}\right)_1 \over \left(q^{s+1+c}\right)_1\left(-q^{s+1+c+n}\right)_1}\ . \label{i1u1}\end{aligned}\ ] ] the normalization factor has the relations @xmath512 and obeys the duality relation @xmath513 meanwhile , the @xmath63-bps wilson loop duality is given by @xmath514 that , together with , implies i^(1)(n,1,n)_-k =- i^-kq^-n^2 + 2ni^(1)(,1;n)_k .
[ 1/2tobechecked ] this is the relation we are going to check explicitly for some small value of @xmath17 , @xmath4 and @xmath38 .
similar to the previous case , the integrand for odd @xmath4 is anti - periodic under the shift @xmath473 , whereas it is periodic for even @xmath4 . in the latter case ,
some of p and np poles merge into double poles .
we can thus apply the same technique as that used in the previous case to this case . in the odd @xmath4 case ,
the poles encircled by the contour appear at @xmath515 for @xmath516 and a given @xmath113 and @xmath517 for @xmath518 and a given @xmath519 . as discussed , the change of the integration variable , @xmath520 , precisely exchanges p and np poles in the two theories .
it is straightforward to carry out the residue integrals and we find @xmath521\nn\end{aligned}\ ] ] where @xmath522 for @xmath523 with an abuse of notation .
the first two lines are the contributions from p poles and the last line is those from np poles . in the even @xmath4 case
we only show the case when @xmath524 .
the other case can be calculated in a similar manner .
the poles encircled by the contour appear at @xmath525 similar to the previous case , we find that @xmath526\end{aligned}\ ] ] where @xmath475 is an arbitrary constant . as an explicit check of
, we consider the example , the duality @xmath490 with winding @xmath491 . for the original theory we have @xmath527 \label{w21}\end{aligned}\ ] ]
where @xmath528 . in the first line
is the contributions from p poles for @xmath271 at @xmath529 , the first term in the second line from np poles for @xmath271 at @xmath530 , and the last term from the p pole for @xmath531 at @xmath532 . for the dual theory we have @xmath533\ .
\nn\end{aligned}\ ] ] the first term is the contributions from p poles for @xmath531 at @xmath532 , the second term from the np pole for @xmath271 at @xmath530 , and the second line from np poles for @xmath531 at @xmath534 .
it is clear that in and the p and np poles are interchanged under the duality as expected .
thus this provides evidence for the @xmath35-bps wilson loop duality . by deduction , this result , combined with the flavor wilson loop duality between and , also constitutes evidence for the @xmath41-bps wilson loop duality . 0
finally we consider the @xmath535 bps wilson loop which ends the @xmath536 gauge group .
@xmath537 @xmath538 \nonumber \\ & \qquad + \frac{1}{2}\sum_{c=0}^{n-1}(-1)^c q^{n(2c - n+1 ) } \frac{q^{-nc } ( q^{n - c } ) _ { n-1 } } { ( 1+q^n ) ( -q^{-c})_{n-1}}\end{aligned}\ ] ] where the first sum comes from the perturbative poles at @xmath539 and the second one comes from the non - perturbative poles at @xmath540 .
the third one is the contribution from the perturbative pole at @xmath541 which is canceled by the same contribution from @xmath111 .
in the current paper , we discussed the wilson loops of the abj theory and studied their properties , generalizing the techniques developed in @xcite for partition function . in more detail ,
the objects of our interest were the circular @xmath37- and @xmath36-bps wilson loops with winding number @xmath38 in the @xmath1 abj theory on @xmath23 . by the localization technique
, the wilson loop can be represented as an ordinary integral with @xmath2 variables @xmath543 and @xmath3 variables @xmath341 , corresponding to the eigenvalues of @xmath44 and @xmath45 adjoint matrices . rather than directly evaluating this abj matrix integral
, we followed @xcite and started instead with the wilson loop in the lens space matrix integral , which is related to the abj one by the analytic continuation @xmath29 . because the lens space wilson loop can be computed exactly , by continuing it back to the abj theory by setting @xmath29 , we arrived at an infinite sum expression for the abj wilson loop
. actually , this infinite sum is only formal and does not converge .
however , by means of a sommerfeld - watson transformation , we turned it into a convergent integral of @xmath30 variables and successfully obtained the `` mirror '' description of the abj wilson loop , generalizing that for partition function .
the final expressions are given in for the @xmath37-bps wilson loop with @xmath77 , in for the @xmath37-bps wilson loop with @xmath32 , and in for the @xmath36-bps wilson loop .
the abj theory is conjectured to possess a seiberg - like duality given in . based on the integral expressions for the abj wilson loops
, we showed that the wilson loops have non - trivial transformation rule , given by .
this result is consistent with the result by kapustin and willett @xcite and slightly generalize it to the case where the flavor group is gauged .
we also presented a heuristic explanation of the seiberg - like duality based on the brane construction of the abj theory , followed by a rigorous proof based on the integral representation of wilson loops .
the brane picture is heuristic but quite powerful and can be used to predict the duality rule for wilson loops with general representations .
we also presented another derivation of the duality in appendix [ kapustinwillett ] .
our method to start from the lens space theory and analytically continue it to abj theory involves subtleties associated with a sommerfeld - watson transformation to rewrite a divergent sum in terms of a well - defined contour integral .
in particular , this rewriting has possible ambiguities in the choice of integration contours including the @xmath99-prescription for the parameter @xmath18 .
it is necessary to keep @xmath18 slightly away from an integral value by the shift @xmath100 with @xmath101 in the course of calculations . in sync with this shift the contour @xmath98 $ ]
has to be placed between @xmath541 and @xmath103 so as to avoid the pole at @xmath544 .
although the choice we made is well - motivated by the continuity in @xmath4 and necessary to obtain sensible results and seiberg duality provides strong evidence in support for it , a direct derivation is certainly desirable .
the approaches taken in @xcite presumably provide promising directions for that purpose .
actually , however , this weakness of our approach can be turned around and regarded as its strength .
the infinite sum we encounter in the intermediate stage can be understood as giving a perturbative expansion of a gauge theory quantify which , by itself , is incomplete and divergent . rewriting it in terms of a finite contour integral can be thought of as supplementing it with non - perturbative corrections to make it well - defined and complete ; more precisely , summing over p poles corresponds to summing up perturbative expansion and including np poles corresponds to adding non - perturbative corrections .
we emphasize that it is very rare that we can carry out this non - perturbative completion in non - trivial field theories and abj wilson loops provide explicit and highly non - trivial examples for it . in @xcite ,
the partition function of abjm theory was evaluated using the fermi gas approach in detail and a cancellation mechanism was found between non - perturbative contributions .
namely , for certain values of @xmath4 , the contribution from worldsheet instantons diverges but , when that happens , the contribution from membrane instantons also diverges and they cancel each other to produce a finite result .
this was generalized to 1/2-bps wilson loops in abjm theory in @xcite and to abj theory in @xcite .
this phenomenon is reminiscent of what is happening in our formulation , in which partition function and wilson loops are expressed as contour integrals .
the integrals can be evaluated by summing over the residue of p and np poles , which are generically simple poles .
as we change @xmath4 continuously , at some integral values of @xmath4 , two such simple poles can collide and become a double pole . for this to happen
, the residue of each simple pole must diverge but their sum must remain finite . where @xmath166 . ]
it is reasonable to conjecture that this cancellation of residues is closely related to the cancellation mechanism of @xcite .
we leave this fascinating possibility for future study .
we observed that the @xmath37-bps wilson loop diverges for @xmath545 .
actually , because the lens space wilson loop is invariant under @xmath134 , we can define the analytically continued @xmath37-bps abj wilson loops to have this periodicity in @xmath38 as @xmath546 .
then the @xmath37-bps abj wilson loop is divergent only for @xmath138 mod @xmath4 , which can be checked in explicit expressions such as . on the other hand ,
the @xmath36-bps wilson loop is finite for all @xmath38 . in the type iia bulk
dual , in @xmath547 , the @xmath548-bps wilson loop corresponds to a fundamental string extending along @xmath549 inside @xmath550 and sitting at a point inside @xmath551 @xcite .
there is no problem having @xmath38 such fundamental strings , which must correspond to the @xmath36-bps wilson loop with arbitrary winding @xmath38 . on the other hand , for generic @xmath38 ,
the @xmath37-bps wilson loop has been argued to correspond to smearing the above fundamental string over @xmath552 @xcite , which seems a bit unnatural for an object as fundamental as a wilson loop .
however , particularly for @xmath138 mod @xmath4 , there is a @xmath37-bps configuration in which a d2-brane is along @xmath553 and carries fundamental string charge dissolved in worldvolume flux @xcite .
so , it is tempting to conjecture that , for @xmath554 mod @xmath4 , there is some different configuration dual to the @xmath37-bps wilson loop which becomes the d2-brane configuration at @xmath138 . the divergence is presumably related to this phase transition
. it would be interesting to actually find such a brane configuration . in @xcite
, it was conjectured that the @xmath328 abj theory in the fixed @xmath2 , large @xmath555 limit is dual to the @xmath22 supersymmetric , parity - violating version of the vasiliev higher spin theory ( where we assumed @xmath556 ) .
being based on @xmath2-dimensional integral , our formulation is particularly suited for studying this limit .
so , it is very interesting to use our results to evaluate wilson loops in the higher spin limit and compare them with predictions from the vasiliev side .
it is also interesting to see if our approach can be applied to more general csm theories with less supersymmetry , such as the necklace quiver @xcite .
we would like to thank nadav drukker , masazumi honda , sanefumi moriyama , kazutoshi ohta and kazumi okuyama for useful discussions .
ms thank the ipht , cea - saclay for hospitality where part of this work was done .
sh would like to thank yitp , nagoya university and kias for their hospitality where part of this work was done .
the work of ms was supported in part by grant - in - aid for young scientists ( b ) 24740159 from the japan society for the promotion of science ( jsps ) .
the results in the main text are given in terms of @xmath85-pochhammer symbols .
in this appendix we provide the definitions and some useful formulae and properties of related quantities .
roughly , a @xmath85-analog is a generalization of a quantity to include a new parameter @xmath85 , such that it reduces to the original version in the @xmath557 limit . in this appendix
, we will summarize definitions of various @xmath85-analogs and their properties relevant for the main text .
[ [ bullet - boldsymbolq - number ] ] @xmath108 @xmath558-number : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for @xmath559 , the @xmath85-number of @xmath560 is defined by @xmath561_q:={1-q^z\over 1-q},\qquad \end{aligned}\ ] ] [ [ bullet - boldsymbolq - pochhammer - symbol ] ] @xmath108 @xmath558-pochhammer symbol : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for @xmath562 , @xmath563 , the @xmath85-pochhammer symbol @xmath564 is defined by @xmath565 for @xmath559 , @xmath566 is defined by the last expression : @xmath567 this in particular means @xmath568 for @xmath569 , @xmath570 note that the @xmath571 limit of the @xmath85-pochhammer symbol is not the usual pochhammer symbol but only up to factors of @xmath572 : @xmath573 we often omit the base @xmath85 and simply write @xmath574 as @xmath575 . to denote the usual pochhammer symbol . ] some useful relations involving @xmath85-pochhammer symbols are @xmath576 where @xmath577 and @xmath578 is the @xmath85-gamma function defined below . for @xmath348 , we have the following formulae which `` reverse '' the order of the product in the @xmath85-pochhammer symbol : @xmath579 if @xmath580 with @xmath581 , the correction to this is of order @xmath582 : @xmath583 here we assumed that @xmath584 and @xmath585 , [ [ bullet - boldsymbolq - factorials ] ] @xmath108 @xmath558-factorials : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for @xmath563 , the @xmath85-factorial is given by @xmath586_q!:= [ 1]_q[2]_q\cdots[n]_q={(q)_n\over ( 1-q)^n},\qquad [ 0]_q!=1 , \qquad [ n+1]_q!=[n]_q[n-1]_q!~.\end{aligned}\ ] ] [ [ bullet - boldsymbolq - gamma - function ] ] @xmath108 @xmath558-gamma function : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for @xmath559 , the @xmath85-gamma function @xmath578 is defined by @xmath587 the @xmath85-gamma function satisfies the following relations : @xmath588_q\gamma_q(z),\\ \gamma_q(1)&=\gamma_q(2)=1,\qquad \gamma_q(n)=[n-1]_q !
\quad ( n\ge 1).\end{aligned}\ ] ] the behavior of @xmath589 near non - positive integers is @xmath590_q!~,\quad \ ] ] where @xmath563 , and @xmath166 . as @xmath571
, this reduces to the formula for the ordinary @xmath591 , @xmath592 [ [ bullet - boldsymbolq - barnes - boldsymbolg - function ] ] @xmath108 @xmath558-barnes @xmath593 function : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for @xmath559 , the @xmath85-barnes @xmath594 function is defined by @xcite @xmath595.\end{aligned}\ ] ] some of its properties are @xmath596_q ! = ( 1-q)^{-\half(n-1)(n-2)}\prod_{j=1}^{n-2}(q)_j = \prod_{k=1}^{n-2}[k]_q^{n - k-1},\\ \prod_{1\le a < b\le n}(q^a - q^b)= q^{{1\over 6}n(n^2 - 1)}(1-q)^{\half n(n-1)}g_2(n+1;q ) .
\label{ehwt6nov12}\end{gathered}\ ] ] the behavior of @xmath597 near non - positive integers is @xmath598 where @xmath569 , and @xmath166 . as @xmath571
, this reduces to the formula for the ordinary @xmath599 , @xmath600 we note that the vandermonde determinant can be expressed by the @xmath85-barnes @xmath594-function @xmath601 that follows from = q^n(n+1)_l=1^n(1-q^l)=q^n(n+1)(1-q)^n_q(n+1 ) .
some details of the calculations in the main text are given in this appendix .
in particular , we provide relevant details in the calculation of the wilson loops in the lens space matrix model and those of their analytic continuation to the wilson loops in the abj(m ) matrix model .
the analytic continuation presented here is a streamlined version of that given for the partition function in @xcite .
we provide computational details of the derivation of ( [ lensspacewilsonloop ] ) .
we first recall the definition of the ( un - normalized ) wilson loop in the lens space matrix model : @xmath602 where we have defined the expectation value of @xmath156 by : = _ lens _ i=1^n_1d_i2_a=1^n_2d_a2 _
sh()^2_sh()^2_ch(,)^2 o e^-12g_s(_i=1^n_1_i^2+_a=1^n_2_a^2 ) . [ vevapp ] the integrals ( [ wilsonlensapp ] ) are actually gaussian and can be performed exactly . to see it
, it is convenient to shift the eigenvalues as _ i_i - i2 , _ a_a+i2 .
these yield @xmath603 where @xmath604 .
the vandermonde determinant @xmath605 takes the following form and can be expanded as @xmath606 where @xmath607 is the permutation group of length @xmath17 and @xmath608 is the signature of an element @xmath609 .
because each term in is an exponential whose exponent is linear in @xmath610 , the integrals are all gaussian . to proceed , we define ( j):=_i=1^n_1d_i2_a=1^n_2d_a2 _
sh()^2_sh()^2_ch(,)^2 e^n_j e^-12g_s(_i=1^n_1_i^2+_a=1^n_2_a^2 ) , so that @xmath611 . by using the expansion of the vandermonde determinants ,
this becomes @xmath612}\ .\nn\end{aligned}\ ] ] it is then straightforward to perform the gaussian integrals and find that @xmath613 where in the second equality we used @xmath614 and @xmath615 notice that the sum over the permutation @xmath616 in ( [ wj0 ] ) is the determinant of the vandermonde matrix ( x_a)^b:= \ { cl ( -q^(a)+n_aj)^b & ( a=1,,j , , n_1 ) + ( q^-(a))^b & ( a = n_1 + 1 , , n ) .
, where we introduced @xmath617 .
thus ( [ wj0 ] ) yields @xmath618 we now rewrite the sum over the permutation @xmath619 as the sum over the ways of partitioning @xmath17 numbers @xmath620 into two groups of ordered numbers @xmath621 where @xmath622 and @xmath623 . in rewriting ,
we start with the sequence of numbers @xmath624 and then reorder it into \{d_1,,d_a_1 , c_1,d_a_1 + 1 , , d_a_2,c_2,d_a_2 + 1,,d_a_n_1,c_n_1,d_a_n_1 + 1 , , d_n_2 } that is just a way of expressing @xmath620 .
this obviously yields the sign @xmath625 .
we further reorder it into the partition \{c_1 , , c_n_1,d_1 , , d_n_2 } .
we can find the sign picked up by this reordering as follows .
we first move @xmath257 farthest to the left .
this gives the sign @xmath626 .
we next move @xmath97 to the next to and the right of @xmath257 .
this gives the sign @xmath627 . in repeating this process
, the move of @xmath191 picks up the sign @xmath628 .
thus the sign picked up in the end of all the moves is ( -1)^(-1)^_i=1^n_1c_i - n_1(n_1 + 1)=(-1)^+_a=1^n_1(a)-n_1(n_1 + 1 ) .
this is exactly the same sign factor as in ( [ wj1 ] ) and thus canceled out .
hence we obtain @xmath629 where @xmath630 is the partition that we have just discussed .
the factor @xmath631 arises for a fixed @xmath229 since there are so many numbers of the sequence that yield a given partition .
note that , as anticipated from the definition of the wilson loop , this is @xmath632-independent and thus yields @xmath633 this agrees with the partition function in @xcite ( up to the difference in normalizations ) when the winding @xmath124 .
as we have done so for the partition function , using ( [ qbarnesvande ] ) , we further rewrite the wilson loop ( [ wilson16ii ] ) as a product of the @xmath85-barnes @xmath594-function and a `` generalization of multiple @xmath85-hypergeometric function '' : @xmath634 where the special function @xmath635 is defined by @xmath636 we now define the normalized wilson loop by @xmath637 that takes the following simpler form @xmath638 in the rest of this appendix we massage ( [ sndef ] ) into a more convenient form : notice first that once the set of @xmath191 s is selected out of the numbers @xmath620 , @xmath176 s simply fill in the rest of the numbers .
taking this fact into account , we can rewrite the special function ( [ sndef ] ) as @xmath639 in going from ( [ snrw1 ] ) to ( [ snrw2 ] ) , we passed the expression @xmath640 note that we can extend the range of sums from @xmath641 to the semi - infinite one @xmath642 , since , by using @xmath643 , the factor @xmath644 for @xmath645 vanishes when @xmath646 . by repeatedly applying the formula @xmath647 , we can simplify ( [ snrw2 ] ) to ( [ lensspacesfunction ] ) @xmath648\ .\nn\end{aligned}\ ] ] we note that the summation over @xmath191 s is originally ordered , @xmath649 .
however , the summand in ( [ snsimple ] ) is invariant under permutation of @xmath191 s , as the marked index @xmath229 is summed over , we have replaced @xmath649 by the unordered sum @xmath650 .
having derived the lens space result , in particular , ( [ snsimple ] ) , here we provide the details of the analytic continuation , @xmath29 , to obtain the abj result : s(n_1,n_2;n)_ks^abj(n_1,n_2;n)_k .
what we are going to do is simply to replace @xmath3 in ( [ snsimple ] ) by @xmath165 and take the @xmath166 limit .
the basic formula to use is @xmath651 where we used ( [ maux20sep12 ] ) in the first equality .
the nontrivial part is the denominator in the factor _ i=1^n_1(-q^1+n_il)_c_i-1(-q^1-n_il)_n - c_i(q)_c_i-1(q)_n - c_i [ nontrivialfactor ] in the second line of ( [ snsimple ] ) .
however , we need to treat the two cases @xmath32 and @xmath77 separately , and the latter turns out to be more involved than the former . in this case it is straightforward to apply the formula ( [ analcontformula ] )
the denominator of ( [ nontrivialfactor ] ) yields ( q)_c_i-1(q)_n - c_i ( q)_c_i-1(q)_-m - c_i+ = -(-1)^m+c_i-1q^(m+c_i)(m+c_i-1)q(q^c_i)_m , where @xmath161 and @xmath652 and we used @xmath653 .
meanwhile , we can rewrite the numerator of ( [ nontrivialfactor ] ) after the analytic continuation as @xmath654 putting these together yields @xmath655 we thus find the analytic continuation ( [ n1lessn2analcont ] ) @xmath656 with the special function ( [ abjsfunction1 ] ) for the abj theory @xmath657\ .\nn\end{aligned}\ ] ] as mentioned , this case is more involved than the previous case .
the major difference stems from the fact that the factor @xmath658 in the denominator of ( [ nontrivialfactor ] ) , when analytically continued to @xmath659 , has the index @xmath660 that is not always negative in contrast to the previous case .
this index becomes negative when @xmath661 .
when the index is negative , the factor @xmath662 is singular and of order @xmath663 as in ( [ analcontformula ] ) .
this means that the factor ( [ nontrivialfactor ] ) , when analytically continued , vanishes with some power of @xmath99 . for the purpose of the analytic continuation
, we are only concerned with the leading vanishing term . in the previous case the leading vanishing term was of order @xmath664 as in ( [ nontrivialfactorcntd ] ) .
to extract the leading vanishing term of the factor ( [ nontrivialfactor ] ) , we need to find in which case the factor @xmath665 is least singular . now
the summand in ( [ analcontformula ] ) is identically zero whenever any of @xmath191 s coincide . in other words , it is only nonzero , when all of @xmath191 s are different from each other .
thus we can focus on the case where none of @xmath191 s are equal . clearly , the factor @xmath665 is least singular when @xmath18 of @xmath191 s , being all different , take values in @xmath666 . since the order does not matter , without loss of generality , we can choose @xmath667 by taking into account the combinatorial factor @xmath668 .
this also implies that the leading vanishing term is of order @xmath669 .
having understood this point , the remaining task is to ( 1 ) plug @xmath667 into ( [ snsimple ] ) and ( 2 ) apply the formula ( [ analcontformula ] ) to the factors @xmath662 with @xmath670 , while taking care of the sum over @xmath229 by splitting the sum @xmath671 into @xmath672 .
we first deal with the factors _ i=1^n_1 ( _ j=1^i-1(q^c_i - c_j)_1(-q^c_i - c_j+n_il)_1 _ j = i+1^n_1(q^c_j - c_i)_1(-q^c_j - c_i - n_il)_1 ) [ facn1gen2 ] in the second line of ( [ snsimple ] ) . the factors with all the indices less than or equal to @xmath18 yield @xmath673 the factors with one of the indices less than or equal to @xmath18 and the other one greater than @xmath18 yield & ( _ i = m+1^n_1_j=1^m(q^c_i - j)_1(-q^c_i - j+n_il)_1 ) ( _ i=1^m_j = m+1^n_1(q^c_j - i)_1(-q^c_j - i - n_il)_1 ) + & = \ { ll _ a=1^n_2(q^d_a)^2_m(-q^d_a+c)_1(-q^d_a)^2_m(-q^d_a+c - n)_1 & ( lm ) + _ a=1^n_2(q^d_a)^2_m(-q^d_a+n_ad)_m(-q^d_a)_m & ( l > m ) . where we relabeled @xmath674 with @xmath675 and @xmath676 for @xmath677 and @xmath678 for @xmath679
the factors with all the indices greater than @xmath18 are trivially of a similar form as ( [ facn1gen2 ] ) .
we next examine the factor in the first line of ( [ snsimple ] ) _ kl1-q^c_k - c_l - n1-q^c_k - c_l & = & \ { ll _ ^m1-q^k - l - n1-q^k - l_k = m+1^n_11-q^c_k - l - n1-q^c_k - l & ( lm ) + _ k=1^m1-q^k - c_l - n1-q^k - c_l_^n_11-q^c_k - c_l - n1-q^c_k - c_l & ( l > m ) .
+ & = & \ { ll q^-n(m-1-c)(q^1-n)_c(q^1+n)_m-1-c(q)_c(q)_m-1-c_a=1^n_21-q^d_a+c - n1-q^d_a+c & ( lm ) + q^-nm(q^d_d+n)_m(q^d_d)_m_^n_21-q^d_a - d_d - n1-q^d_a - d_d & ( l > m ) . . finally , we look into the factor ( [ nontrivialfactor ] ) .
the factors with @xmath680 are exactly the inverse of ( [ n1gen2factor1 ] ) and canceled out , whereas the analytic continuation of the rest of the factors yields @xmath681 putting all factors together , we find the analytic continuation ( [ n1gen2sn ] ) s(n_1,-n_2+;n)_k~(q)^n_2(s^abj_(1)(n_1,n_2;n)_k+s^abj_(2)(n_1,n_2;n)_k ) , where the special functions for the abj theory are given by @xmath682\nn\end{aligned}\ ] ] and @xmath683\ .\nn\end{aligned}\ ] ]
in this appendix , we present the expressions for wilson loops in general representations
. we will be very brief in explaining how to derive these results , because it is similar to the one for partition function ( ref .
@xcite ) and for wilson loops with winding number @xmath38 ( appendix [ details ] ) .
let us start with wilson loops in the @xmath28 lens space matrix model . in , we considered representations with winding number @xmath38 , but here we would like to consider general representations . for the @xmath44 representation with young diagram @xmath378 , the wilson loop can be computed by inserting @xmath684 in the matrix integral , where @xmath685 is the schur polynomial for @xmath378 @xcite . because each term in the polynomial @xmath684 has the form @xmath686 with @xmath687 , all we have to compute in principle is the matrix integral with @xmath688 inserted .
if the wilson line carries a non - trivial representation for @xmath45 , we must insert a similar factor also for @xmath341 .
therefore , given @xmath689 , the object of our interest here is the following matrix integral : @xmath690 where @xmath691 , @xmath692 , with @xmath693 . by symmetry ,
it is clear that this is invariant under @xmath694 and @xmath695 .
if @xmath696 , this reduces to partition function while , for the case with a single winding number , e.g. , @xmath697 , @xmath698 , this reduces to the wilson loop studied in the main text , , up to a factor : @xmath699 by carrying out the gauss integration and doing manipulations similar to , we arrive at the following simple combinatorial expression : @xmath700 s_{\{m\},\{n\}}(n_1,n_2 ) & : = \sum_{(\cn_1,\cn_2 ) } { 1\over n_1!}\sum_{\sigma\in s_{n_1 } } { 1\over n_2!}\sum_{\tau\in s_{n_2 } } q^{-\sum_{j=1}^{n_1}m_{\sigma(j)}c_j-\sum_{a=1}^{n_2}n_{\sigma(a)}d_a } \notag\\ & \qquad\times \prod_{c_j < c_k}{q^{c_j+m_{\sigma(j)}}-q^{c_k+m_{\sigma(k ) } } \over q^{c_j}-q^{c_k } } \prod_{d_a < d_b}{q^{d_a+n_{\tau(a)}}-q^{d_b+n_{\tau(b ) } } \over q^{d_a}-q^{d_b } } \notag\\ & \qquad\times \prod_{c_j < d_a}{q^{c_j+m_{\sigma(j)}}+q^{d_a+n_{\tau(a ) } } \over q^{c_j}-q^{d_a } } \prod_{d_a < c_j}{q^{d_a+n_{\tau(a)}}+q^{c_j+m_{\sigma(j ) } } \over q^{d_a}-q^{c_j}},\label{fsgt24may13}\end{aligned}\ ] ] where @xmath701 and @xmath702 , @xmath703 .
the symbol @xmath704 denotes the partition of the numbers @xmath705 into two groups @xmath706 and @xmath707 where @xmath191 s and @xmath176 s are ordered as @xmath708 and @xmath709 . to proceed ,
let us focus on the case with @xmath710 , namely on the @xmath44 wilson loop henceforth . in this case , we can rewrite the product in @xmath711 in favor of the @xmath85-pochhammer symbol , just as we did in . after using various formulas for the @xmath85-pochhammer symbol in appendix [ qanalogs ] , the final result can be written as @xmath712 in the above , we treated @xmath3 as a continuous variable . to obtain an expression for integral @xmath3 ,
let us shift @xmath713 with @xmath714 where it is understood that @xmath99 will be taken to zero at the end of computation . by extracting powers of @xmath99 from the @xmath85-pochhammer symbols in the first line of
, we find , to leading order in small @xmath99 expansion , @xmath715 we will drop subleading terms in small @xmath99 expansion henceforth . for @xmath716 , of course , the factor @xmath664 in front of must be canceled by the @xmath85-pochhammer symbol @xmath717 , because the original expression was finite to begin with .
this @xmath85-pochhammer symbol can be rewritten as @xmath717 @xmath718 @xmath719 @xmath720 $ ] , where `` @xmath721 '' means up to subleading terms in powers of @xmath99 .
indeed , for @xmath722 , this always contains the factor @xmath723 and , collecting contributions from @xmath724 , we see that this completely cancels the @xmath664 .
actually , this also means that , if @xmath725 , we have less powers of @xmath99 in the denominator and , as a result , the summand vanishes as @xmath166 .
therefore , we are free to remove the upper bound in the @xmath726-sum , as we have already done in .
in , summing over permutations @xmath727 acting on @xmath728 is the same as summing over permutations acting on @xmath726 ( it is easy to show that , if we set @xmath729 , summing over @xmath730 is the same as summing over @xmath731 , with @xmath732 replaced by @xmath733 ) .
therefore , we can relax the ordering constraint on @xmath726 and forget about the summation over permutations : @xmath734 note that the sum over @xmath724 is completely unconstrained , because the summand in vanishes if @xmath735 with @xmath736 .
we can easily see that , upon setting @xmath737 , this reduces to the @xmath711 function for the partition function @xcite .
we will refer to this as the unordered formula .
the formula is referred to as the ordered formula .
the abj wilson loop is obtained by setting @xmath739 with @xmath740 in the above expressions .
more precisely , according @xcite , the formula for analytic continuation is @xmath741.\label{vn11jun14}\end{aligned}\ ] ] first , consider the case with @xmath742 .
using the unordered expression , we straightforwardly obtain @xmath743 we have the @xmath664 as a prefactor , which remains uncanceled by the @xmath85-pochhammer .
it is not difficult to show that , when @xmath697 , this reduces to the formula , up to normalization ; namely , @xmath744 .
the integral representation is @xmath745 q^{-\sum_j ( s_j+1 ) m_j } \\
\times \prod_{j=1}^{n_1 } { ( q^{s_j+1})_m \over ( -q^{s_j+1+m_j})_m } \prod_{j < k}^{n_1 } { ( q^{s_k - s_j+m_k - m_j})_1 ( q^{s_k - s_j})_1 \over ( -q^{s_k - s_j - m_j})_1 ( -q^{s_k - s_j+m_k})_1},\end{gathered}\ ] ] where @xmath746 and @xmath747 . using the formula , the full expression including the prefactors in is @xmath748 the case with @xmath749 is more nontrivial as is the case for partition function @xcite . in this case , there are some powers of @xmath99 coming from @xmath750 in the summand , where we set @xmath751 .
one choice to get the most singular contribution is @xmath752 for this particular choice , by extracting powers of @xmath99 from the @xmath85-pochhammer symbol , we can show that can be rewritten as @xmath753 the choice is only one possibility and there are more ; there are @xmath754 ways to choose @xmath755 special @xmath726 s out of @xmath2 .
furthermore , there are @xmath756 ways to permute those @xmath755 special @xmath726 s .
so , in total , we have @xmath757 ways .
we should sum over all these choices . or
equivalently , as we have seen in going between and , we can fix the order of @xmath726 s to be @xmath758 , consider only , and sum over the permultations of @xmath728 .
so , can be expressed as @xmath759 if we want to relax the ordering constraint and let @xmath760 s run over all positive integers , then the summation will be over @xmath761 , meaning that two permutations @xmath762 are identified if @xmath763 and @xmath764 are permutations of each other .
one can show that , when @xmath697 , eq .
reduces to plus , up to normalization .
namely , @xmath765 $ ] .
the integral representation is @xmath766 q^{-\sum_{j=1}^{n_2'}(s_j+1+m')m_{\sigma(m'+j ) } } \notag\\ & \qquad\times { \prod_{j=1}^{n_2'}(-q^{s_j+1+m_{\sigma(m'+j)}})_{m ' } } \prod_{j=1}^{m ' } \prod_{k=1}^{n_2 ' } { ( q^{m'+s_k+1-j+m_{\sigma(m'+k)}-m_{\sigma(j)}})_1\over ( -q^{m'+s_k+1-j - m_{\sigma(j)}})_1(-q^{m'+s_k+1-j+m_{\sigma(m'+k)}})_1 } \notag\\ & \qquad\times \prod_{1\le j <
k\le n_2'}{(q^{s_k - s_j+m_{\sigma(m'+k)}-m_{\sigma(m'+j)}})_1 ( q^{s_k - s_j})_1 \over ( -q^{s_k - s_j - m_{\sigma(m'+j)}})_1(-q^{s_k - s_j+m_{\sigma(m'+k)}})_1}.\end{aligned}\ ] ] the full expression including the prefactors in is @xmath767
in this appendix we show that , in our expression ( [ 1/6wilsoni2 ] ) for @xmath41-bps wilson loops in the @xmath77 case , the contributions from the p poles at @xmath769 are absent .
namely , these contributions are canceled between those from @xmath142 and @xmath144 .
this is necessary , in particular , for the @xmath41-bps wilson loops ( [ 1/6wilsoni2 ] ) to correctly reproduce the perturbative expansion in @xmath73 and also fills the gap of a proof that the two expressions ( [ 1/6wilsoni1 ] ) and ( [ 1/6wilsoni2 ] ) agree in the abjm limit @xmath15 .
we first recall the expressions of our interest : @xmath770}{\pi ds_a\over\sin(\pi s_a)}\right ] q^{n(2c - m)}{\left(q^{1-n}\right)_c\left(q^{1+n}\right)_{m-1-c}\over ( q)_c(q)_{m-1-c } } \prod_{a=1}^{n_2}{\left(q^{s_a+1}\right)_m\over 2\left(-q^{s_a+1}\right)_{m}}\nn\\ & \times \prod_{a=1}^{n_2}\left[{\left(-q^{s_a+1+c}\right)_1\left(q^{s_a+1+c - n}\right)_1 \over \left(q^{s_a+1+c}\right)_1\left(-q^{s_a+1+c - n}\right)_1}\prod_{b=1}^{a-1}{\left(q^{s_a - s_b}\right)_{1 } \over\left(-q^{s_a - s_b}\right)_{1 } } \prod_{b = a+1}^{n_2}{\left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a}\right)_{1}}\right]\ , \label{integralii1app}\end{aligned}\ ] ] and @xmath92 q^{-ns_d+n(d - m-2 ) } \prod_{\substack{a=1 \\
a\ne d}}^{n_2}{\left(q^{s_a - s_{d}-n}\right)_1\over \left(q^{s_a - s_{d}}\right)_1 } \label{integralii2app}\\ & \times\prod_{a=1}^{n_2}\left[{\left(q^{s_a+1+n\delta_{ad}}\right)_m \over\left(1+q^{n\delta_{ad}}\right)\left(-q^{s_a+1}\right)_{m } } \prod_{b=1}^{a-1}{\left(q^{s_a - s_b}\right)_{1 } \over\left(-q^{s_a - s_b+n\delta_{ad}}\right)_{1 } } \prod_{b = a+1}^{n_2}{\left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a - n\delta_{ad}}\right)_{1}}\right]\ .\nn\end{aligned}\ ] ] in section [ intrep ] we discussed the pole structures of the integrands and the integration contours @xmath257 and @xmath97 in detail .
the contour @xmath257 for a given @xmath113 is placed to the left of @xmath262 and the right of @xmath771 , applies . ] whereas the contour @xmath97 is placed , for @xmath284 , to the left of @xmath772 and the right of @xmath773 and , for @xmath774 , to the left of @xmath281 and the right of @xmath775 . in particular , there are residues from the p poles at @xmath294 , ( @xmath299 ) , for ( [ integralii1app ] ) and @xmath776 for ( [ integralii2app ] ) . recall ( [ limitresidue ] ) for generic ( including non - integral ) @xmath18 _ s_d-1-c(q^1-n)_c(q^1+n)_m-1-c(q)_c(q)_m-1-c(q^s_d+1)_m(q^s_d+1+c)_1 & = & q^-nc(q^n - c)_m(q^n)_1(q^-c)_c(q)_m-1-c_s_d-1-c(q^s_d+1)_m(q^s_d+1+c)_1 + & = & q^-nc(q^n - c)_m(q^n)_1 [ limitresidue2 ] where we used @xmath296 , @xmath297 and ( [ maux20sep12 ] ) . as remarked below in section [ abjmlimit ] , implicit in this calculation is the @xmath99-prescription that always enables us to regard the p pole at @xmath294 as a simple pole .
namely , the integer @xmath18 is shifted to a non - integral value @xmath304 and the contour @xmath98 $ ] is placed between @xmath294 and @xmath103 so as to avoid the latter pole . this yields and @xmath144 by the same symbols . ]
@xmath777&={(-1)^c\over n_2 ! } \prod_{\substack{a=1 \\
a\ne d}}^{n_2}\left[{-1\over 2\pi i}\int_{c_1\![c]}{\pi ds_a\over\sin(\pi s_a)}\right ] { q^{n(c - m)}(q^{n - c})_m\over ( -q^{-c})_m(-q^{n})_1 } \nn\\ & \times \prod_{\substack{a=1 \\ a\ne d}}^{n_2}\biggl [ { \left(q^{s_a+1}\right)_m\left(q^{s_a+1+c}\right)_1\left(q^{s_a+1+c - n}\right)_1 \over 2\left(-q^{s_a+1}\right)_{m}\left(-q^{s_a+1+c}\right)_1\left(-q^{s_a+1+c - n}\right)_1 } \label{i1residue}\\ & \times \prod_{\substack{b=1 \\
b\ne d}}^{a-1}{\left(q^{s_a - s_b}\right)_{1 } \over\left(-q^{s_a - s_b}\right)_{1 } } \prod_{\substack{b = a+1\\ b\ne d}}^{n_2}{\left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a}\right)_{1}}\biggr]\ .\nn\end{aligned}\ ] ] similarly , it is straightforward to find that @xmath778&={(-1)^{c+1}\over n_2 ! } \prod_{\substack{a=1 \\
a\ne d}}^{n_2}\left[{-1\over 2\pi i}\int_{c_2}{\pi ds_a\over\sin(\pi s_a)}\right]{q^{n(c - m)}\left(q^{n - c}\right)_m \over\left(-q^n\right)_1\left(-q^{-c}\right)_{m } } \nn\\ & \times \prod_{\substack{a=1 \\
a\ne d}}^{n_2}\biggl[{\left(q^{s_a+1}\right)_m\left(q^{s_a+1+c}\right)_{1}\left(q^{s_a+1+c - n}\right)_{1 } \over 2\left(-q^{s_a+1}\right)_{m}\left(-q^{s_a+1+c}\right)_{1}\left(-q^{s_a+1+c - n}\right)_{1}}\label{i2residue}\\ & \times \prod_{\substack{b=1 \\
b\ne d}}^{a-1}{\left(q^{s_a - s_b}\right)_{1 } \over\left(-q^{s_a - s_b}\right)_{1 } } \prod_{\substack{b = a+1\\ b\ne d}}^{n_2}{\left(q^{s_b - s_a}\right)_{1}\over\left(-q^{s_b - s_a}\right)_{1}}\biggr]\ .\nn\end{aligned}\ ] ] hence the sum of the two residues @xmath779 $ ] exactly cancels out . as @xmath113 runs from @xmath114 to @xmath118 ,
this implies the cancellation of the residues in ( [ 1/6wilsoni2 ] ) at the p poles @xmath776 , @xmath780 .
more precisely , the cancellation requires the equivalence of the contours @xmath257 and @xmath97 . in this regard , note that the integrand of the residue ( [ i1residue ] ) , in particular , does not have a pole at @xmath102 and the contour @xmath257 can thus be shifted , past @xmath102 , to the left of @xmath781 so that @xmath257 becomes identical to @xmath97 .
we would like to address the subtlety remarked below ( [ s1 ] ) and ( [ s2 ] ) concerning the range , in particular , of the sum over @xmath113 in ( [ s1 ] ) and ( [ integralii1 ] ) . in its original form , the sum over @xmath113 ran from @xmath114 to @xmath115 .
when @xmath38 is less than @xmath18 , the sum simply terminates at @xmath782 , as the factors ( [ limitresidue2 ] ) vanish when @xmath121 .
when @xmath38 is greater than @xmath18 , however , in order to replace the upper limit @xmath115 by @xmath118 , we need to show that the contribution from @xmath783 to @xmath118 is absent in ( [ 1/6wilsoni2 ] ) and ( [ formalabj1/6wilsonloop ] ) .
note that when @xmath266 in the first line of ( [ limitresidue2 ] ) the factor @xmath784 in the denominator diverges .
thus , for @xmath266 , ( [ limitresidue2 ] ) might appear to vanish .
however , there are nonvanishing contributions coming from the poles at @xmath294 as the second line of ( [ limitresidue2 ] ) indicates .
this implies that the cancellation we have shown above is all we need to ensure the absence of the contribution from @xmath783 to @xmath118 in ( [ 1/6wilsoni2 ] ) and ( [ formalabj1/6wilsonloop ] ) .
the cancellation of the residues at @xmath785 also justifies the extension of the lower limits of the sum over @xmath176 s in ( [ s1 ] ) and ( [ s2 ] ) .
as a simplest check of our prescriptions that lack first principle derivations , we compare the result from the integral representation with that from the direct calculation in the case of the @xmath786 abjm theory .
the direct integral of yields @xmath787 where we used the fourier transform of @xmath788 @xmath789 as commented in section [ wilsonloopresults ] , the restriction on winding number , @xmath790 , is necessary for the convergence of integrals .
hence the direct calculation gives the normalized @xmath34-bps wilson loop @xmath791 on the other hand , as calculated in section [ u1unexample ] , the result from our integral representation yields , since @xmath792 , the sum @xmath793 for an integer @xmath38 .
meanwhile , this sum equals @xmath794 for a non - integral @xmath38 , to find the sum @xmath795 for even @xmath4 , we differentiate @xmath796 w.r.t .
@xmath38 and then send @xmath38 to an integral value . ]
@xmath797 thus the normalized @xmath34-bps wilson loop indeed agrees with providing evidence for our prescriptions .
in this appendix we provide an alternative derivation of the duality transformations of abj wilson loops by following kapustin and willett @xcite . in their approach seiberg duality
is understood as an isomorphism of the algebras that bps wilson loops generate . before discussing the wilson loops in the abj theory
, we first give a brief sketch of their construction of wilson loop algebras and derivation of duality transformations for generic three - dimensional supersymmetric gauge theories that admit localization . in order to construct the wilson loop algebras , we first define @xmath798 and @xmath799 where we denoted symmetric polynomials by the corresponding young diagrams .
each young diagram corresponds to a specific representation of the wilson loop .
the variables @xmath800 will be identified with the integration variables of the matrix models derived from localization .
the @xmath801 s and @xmath802 s generate the ring of symmetric polynomials .
notice , however , that @xmath803 by definition and thus the @xmath801 s and @xmath802 s are not independent .
next , to find the duality transformations for bps wilson loops , we construct an algebra for a given theory from the ring of symmetric polynomials that _ quantum _ wilson loops generate .
as it turns out , there is an isomorphism between the algebras for the original and dual theories that can be regarded as the duality transformations .
here we only outline the derivation of the duality transformations : 1 . using the matrix model obtained by localization , the invariance under the shift of an integration variable yields the following identity : @xmath804 where @xmath805 is an integration variable and @xmath806 means vev of the matrix model .
the polynomial @xmath807 is at most of the @xmath18-th order in @xmath805 , where @xmath18 is a constant determined by the rank and level of the gauge theory .
this is a _
constraint on the bps wilson loops that the wilson loop algebra is endowed with . 2 . from the polynomial @xmath807 we construct the following quantities
@xmath808 and @xmath809 : @xmath810 and @xmath811 in the algebra that kapustin and willett identify with a certain quotient of the ring , the classical constraint @xmath812 is deformed to @xmath813 that the elements , @xmath801 s and @xmath814 s , of the algebra obey in quantum theory .
the quantum constraint is crucial , and it is important that the left hand side of the constraint is truncated to a finite polynomial in @xmath815 in contrast to the classical constraint @xmath812 .
3 . owing to the truncation in and
thus in , there exists an isomorphism under the following transformations : @xmath816 that is , the quantum constraint is invariant under these transformations .
the first transformation can be identified with the map of the ranks of the two gauge groups in a dual pair and the second transformations with the maps of the bps wilson loops .
thus the duality transformations of the wilson loops can be extracted order by order in @xmath815 from . 0 @xmath817 at @xmath818 , for instance
, we find for the fundamental representation @xmath819 where the tilde indicates that the wilson loop is that in the dual theory .
we now apply the above method to abj wilson loops .
the original theory is the @xmath820 abj theory and the dual theory is the @xmath821 theory with the dual gauge group @xmath822 . in implementing the above procedure
it is useful to regard the original theory as the @xmath823 chern - simons matter theory and the @xmath44 part as flavor .
let us first recall that the @xmath34-bps wilson loop on the gauge group @xmath45 with winding @xmath38 is given by the eigenvalue integrals @xmath824 where we omitted the normalization factor as it is not relevant in the following discussion . for later convenience ,
we introduce the following notations : @xmath825 note that in only @xmath543 s are integrated and @xmath341 s are regarded as parameters of the theory . in the roles of
@xmath543 s and @xmath341 s are interchanged . in this notation the @xmath34-bps wilson loop on the @xmath45
is expressed as @xmath826 in the expression we can regard the @xmath820 abj theory as the @xmath45 chern - simons matter theory at level @xmath827 with @xmath828 hypermultiplets without fayet - iliopoulos terms . from this viewpoint
the flavor group is @xmath829 instead of @xmath44 . in , @xmath830 can be thought of as mass parameters for the hypermultiplets .
the seiberg dual is then the @xmath831 chern - simons matter theory at level @xmath4 with @xmath828 hypermultiplets of masses @xmath832 without fayet - iliopoulos terms @xcite .
these parameters are to be integrated in the end . in step 1 , to find for the abj theory , we consider the following quantity : @xmath833 integrals are invariant under shifts of integration variables . for our purpose , we consider , in particular , the shift @xmath834 where @xmath475 is a selected index but can be chosen arbitrarily .
note that the inserted factor @xmath835 exactly cancels the poles that appear in the integrand and allows us to make the above shift keeping the range of integration intact . under this shift
becomes @xmath836 since this is identical to , we obtain the identity : @xmath837 this equation must hold for any representation of bps wilson loops and then yields the operator identity @xmath838 where we introduced @xmath839 . for later convenience , we rewrite this as @xmath840 note that this polynomial is at most of order @xmath841 .
in step 3 , seiberg duality of the wilson loops is extracted from the transformation . at @xmath818
we obtain @xmath844 where the subscripts @xmath845 and @xmath846 indicate that the wilson loops are on the first and second gauge group , respectively . in this example , the bps wilson loop of the original @xmath823 theory in the fundamental representation maps to minus that of the dual @xmath847 theory in the fundamental representation shifted by a singlet that depends on the masses @xmath543 s in the dual theory .
case it might look naively that there is an additional identity operator coming from the @xmath848 term that contributes to the duality map . however , the @xmath34-bps wilson loop with winding @xmath38 is only well - defined for @xmath239 and thus the @xmath42 fundamental wilson loop is ill - defined for @xmath849 .
this method is therefore not applicable to the abj theory with @xmath849 .
for the same reason the @xmath850 terms in are meaningless because the @xmath34-bps wilson loops with winding @xmath851 are involved at this order . ] in order to find the duality maps for the abj theory , we need to integrate over the mass parameters @xmath830 .
the map then becomes @xmath852 in our notation this reads @xmath853 this indeed agrees with .
we can run the same procedure for the flavor wilson loops . since the flavor group is unchanged under seiberg duality , we simply obtain the following map : @xmath854 which reads @xmath855 this agrees with .
finally , the maps and imply that @xmath856 this agrees with .
we have thus succeeded to reproduce seiberg duality for the abj wilson loops with the winding @xmath42 . from the isomorphism of the wilson loop algebras
, we can also extract the duality transformations for the wilson loops in higher dimensional representations in the abj theory . here
we consider the wilson loops that involve two boxes in young diagrams corresponding to the @xmath857 terms in : @xmath858 integrating over the mass parameters @xmath543 , we find the duality transformation of the antisymmetric wilson loop : @xmath859 where @xmath860 is a fundamental wilson loop with winding @xmath491 and decomposed into @xmath861 using this relation , yields @xmath862 this agrees with the result from the heuristic derivation in the brane picture in section [ branepicturesection ] .
for the flavor wilson loops , we simply have @xmath863
lastly , we consider the wilson loop that is on both gauge groups @xmath864 the map of this wilson loop is obtained from the duality transformation for the fundamental wilson loops : @xmath865 where integration over @xmath341 s is implied . combining , and , we find the map of the @xmath35-bps wilson loop @xmath866 this again agrees with the result in section [ branepicturesection ] .
although it becomes increasingly cumbersome , it is straightforward to generalize this procedure to more general higher dimensional representations .
there is a subtlety for @xmath867 since contains the @xmath848 contribution .
then when we consider the low representation wilson loop , we need include the @xmath848 terms .
first we consider the @xmath849 case . in this case
the wilson loop duality becomes @xmath868 in the electric side , we regard @xmath543 as mass parameters . under the duality mapping we considered that @xmath543 are mapped @xmath869 .
but , is this correct ?
for example if @xmath870 is correct , seiberg duality is changed : @xmath871 n. seiberg , `` electric - magnetic duality in supersymmetric nonabelian gauge theories , '' nucl .
b * 435 * , 129 ( 1995 ) [ hep - th/9411149 ] .
e. witten , `` topological quantum field theory , '' commun .
* 117 * , 353 ( 1988 ) .
v. pestun , `` localization of gauge theory on a four - sphere and supersymmetric wilson loops , '' commun .
phys . *
313 * , 71 ( 2012 ) [ arxiv:0712.2824 [ hep - th ] ] .
m. marino , `` lectures on localization and matrix models in supersymmetric chern - simons - matter theories , '' j. phys . a a * 44 * , 463001 ( 2011 ) [ arxiv:1104.0783 [ hep - th ] ] .
o. aharony , o. bergman and d. l. jafferis , `` fractional m2-branes , '' jhep * 0811 * , 043 ( 2008 ) [ arxiv:0807.4924 [ hep - th ] ] .
o. aharony , o. bergman , d. l. jafferis and j. maldacena , `` n=6 superconformal chern - simons - matter theories , m2-branes and their gravity duals , '' jhep * 0810 * , 091 ( 2008 ) [ arxiv:0806.1218 [ hep - th ] ] . c. -m .
chang , s. minwalla , t. sharma and x. yin , `` abj triality : from higher spin fields to strings , '' arxiv:1207.4485 [ hep - th ] .
a. kapustin , b. willett and i. yaakov , `` exact results for wilson loops in superconformal chern - simons theories with matter , '' jhep * 1003 * , 089 ( 2010 ) [ arxiv:0909.4559 [ hep - th ] ] .
n. drukker , m. marino and p. putrov , `` from weak to strong coupling in abjm theory , '' commun .
* 306 * , 511 ( 2011 ) [ arxiv:1007.3837 [ hep - th ] ] .
c. p. herzog , i. r. klebanov , s. s. pufu and t. tesileanu , `` multi - matrix models and tri - sasaki einstein spaces , '' phys .
d * 83 * , 046001 ( 2011 ) [ arxiv:1011.5487 [ hep - th ] ] .
h. fuji , s. hirano and s. moriyama , `` summing up all genus free energy of abjm matrix model , '' jhep * 1108 * , 001 ( 2011 ) [ arxiv:1106.4631 [ hep - th ] ] .
m. marino and p. putrov , `` abjm theory as a fermi gas , '' j. stat .
mech . * 1203 * , p03001 ( 2012 ) [ arxiv:1110.4066 [ hep - th ] ] .
y. hatsuda , s. moriyama and k. okuyama , `` instanton effects in abjm theory from fermi gas approach , '' jhep * 1301 * , 158 ( 2013 ) [ arxiv:1211.1251 [ hep - th ] ] .
y. hatsuda , s. moriyama and k. okuyama , `` instanton bound states in abjm theory , '' jhep * 1305 * , 054 ( 2013 ) [ arxiv:1301.5184 [ hep - th ] ] .
s. matsumoto and s. moriyama , `` abj fractional brane from abjm wilson loop , '' jhep * 1403 * , 079 ( 2014 ) [ arxiv:1310.8051 [ hep - th ] ] .
m. honda and k. okuyama , `` exact results on abj theory and the refined topological string , '' arxiv:1405.3653 [ hep - th ] . h. awata , s. hirano and m. shigemori , `` the partition function of abj theory , '' arxiv:1212.2966 [ hep - th ] . f. benini , c. closset and s. cremonesi , `` chiral flavors and m2-branes at toric cy4 singularities , '' jhep * 1002 * , 036 ( 2010 ) [ arxiv:0911.4127 [ hep - th ] ] .
a. kapustin , b. willett and i. yaakov , `` nonperturbative tests of three - dimensional dualities , '' jhep * 1010 * , 013 ( 2010 ) [ arxiv:1003.5694 [ hep - th ] ] .
m. marino , `` chern - simons theory , matrix integrals , and perturbative three manifold invariants , '' commun . math .
phys . * 253 * , 25 ( 2004 ) [ arxiv : hep - th/0207096 ] .
m. aganagic , a. klemm , m. marino and c. vafa , `` matrix model as a mirror of chern - simons theory , '' jhep * 0402 * , 010 ( 2004 ) [ hep - th/0211098 ]
. m. honda , `` direct derivation of `` mirror '' abj partition function , '' jhep * 1312 * , 046 ( 2013 ) [ arxiv:1310.3126 [ hep - th ] ] .
s. hirano , m. honda , k. okuyama and m. shigemori , `` abj theory in the higher spin limit , '' work in progress .
a. giveon and d. kutasov , `` seiberg duality in chern - simons theory , '' nucl .
b * 812 * , 1 ( 2009 ) [ arxiv:0808.0360 [ hep - th ] ] .
a. hanany , d. vegh and a. zaffaroni , `` brane tilings and m2 branes , '' jhep * 0903 * , 012 ( 2009 ) [ arxiv:0809.1440 [ hep - th ] ] .
a. amariti , d. forcella , l. girardello and a. mariotti , `` 3d seiberg - like dualities and m2 branes , '' jhep * 1005 * , 025 ( 2010 ) [ arxiv:0903.3222 [ hep - th ] ] .
f. benini , c. closset and s. cremonesi , `` comments on 3d seiberg - like dualities , '' jhep * 1110 * , 075 ( 2011 ) [ arxiv:1108.5373 [ hep - th ] ] .
a. amariti , c. klare and m. siani , `` the large n limit of toric chern - simons matter theories and their duals , '' jhep * 1210 * , 019 ( 2012 ) [ arxiv:1111.1723 [ hep - th ] ] .
c. closset , `` seiberg duality for chern - simons quivers and d - brane mutations , '' jhep * 1203 * , 056 ( 2012 ) [ arxiv:1201.2432 [ hep - th ] ] . h. c. kim , j. kim da , s. kim and k. lee , `` vortices and 3 dimensional dualities , '' arxiv:1204.3895 [ hep - th ] .
d. xie , `` three dimensional seiberg - like duality and tropical cluster algebra , '' arxiv:1311.0889 [ hep - th ] .
s. dwivedi and p. ramadevi , `` is toric duality a seiberg - like duality in ( 2 + 1)-d ?
, '' jhep * 1407 * , 084 ( 2014 ) [ arxiv:1401.2767 [ hep - th ] ] . 39 a. amariti and d. forcella , `` spin(7 ) duality for @xmath872 = 1 cs - matter theories , '' jhep * 1407 * , 082 ( 2014 ) [ arxiv:1404.4052 [ hep - th ] ] .
a. amariti and c. klare , `` chern - simons and rg flows : contact with dualities , '' arxiv:1405.2312 [ hep - th ] .
a. giveon and d. kutasov , `` brane dynamics and gauge theory , '' rev .
phys . * 71 * , 983 ( 1999 ) [ hep - th/9802067 ] .
a. kapustin , b. willett and i. yaakov , `` tests of seiberg - like duality in three dimensions , '' arxiv:1012.4021 [ hep - th ]
. e. witten , `` quantum field theory and the jones polynomial , '' commun .
phys . * 121 * , 351 ( 1989 ) .
n. drukker , j. plefka and d. young , `` wilson loops in 3-dimensional n=6 supersymmetric chern - simons theory and their string theory duals , '' jhep * 0811 * , 019 ( 2008 ) [ arxiv:0809.2787 [ hep - th ] ] .
b. chen and j. -b .
wu , `` supersymmetric wilson loops in n=6 super chern - simons - matter theory , '' nucl . phys .
b * 825 * , 38 ( 2010 ) [ arxiv:0809.2863 [ hep - th ] ] . s. -j .
rey , t. suyama and s. yamaguchi , `` wilson loops in superconformal chern - simons theory and fundamental strings in anti - de sitter supergravity dual , '' jhep * 0903 * , 127 ( 2009 ) [ arxiv:0809.3786 [ hep - th ] ] .
n. drukker and d. trancanelli , `` a supermatrix model for n=6 super chern - simons - matter theory , '' jhep * 1002 * , 058 ( 2010 ) [ arxiv:0912.3006 [ hep - th ] ] .
m. marino , p. putrov , `` exact results in abjm theory from topological strings , '' jhep * 1006 * , 011 ( 2010 ) .
[ arxiv:0912.3074 [ hep - th ] ] .
n. drukker , m. marino and p. putrov , `` nonperturbative aspects of abjm theory , '' jhep * 1111 * , 141 ( 2011 ) [ arxiv:1103.4844 [ hep - th ] ] .
a. klemm , m. marino , m. schiereck and m. soroush , `` abjm wilson loops in the fermi gas approach , '' arxiv:1207.0611 [ hep - th ] . y. hatsuda , m. honda , s. moriyama and k. okuyama , `` abjm wilson loops in arbitrary representations , '' jhep * 1310 * , 168 ( 2013 ) [ arxiv:1306.4297 [ hep - th ] ] . y. hatsuda , m. marino , s. moriyama and k. okuyama , `` non - perturbative effects and the refined topological string , '' arxiv:1306.1734 [ hep - th ] .
a. kapustin and b. willett , `` wilson loops in supersymmetric chern - simons - matter theories and duality , '' arxiv:1302.2164 [ hep - th ] .
d. berenstein and d. trancanelli , `` three - dimensional n=6 scft s and their membrane dynamics , '' phys . rev .
d * 78 * , 106009 ( 2008 ) [ arxiv:0808.2503 [ hep - th ] ] . c. closset , t. t. dumitrescu , g. festuccia , z. komargodski and n. seiberg , `` comments on chern - simons contact terms in three dimensions , '' jhep * 1209 * , 091 ( 2012 ) [ arxiv:1206.5218 [ hep - th ] ] .
t. kitao , k. ohta and n. ohta , `` three - dimensional gauge dynamics from brane configurations with ( p , q ) - five - brane , '' nucl .
b * 539 * , 79 ( 1999 ) [ hep - th/9808111 ] .
k. ohta and y. yoshida , `` non - abelian localization for supersymmetric yang - mills - chern - simons theories on seifert manifold , '' phys .
d * 86 * , 105018 ( 2012 ) [ arxiv:1205.0046 [ hep - th ] ] .
m. honda and s. moriyama , `` instanton effects in orbifold abjm theory , '' arxiv:1404.0676 [ hep - th ] . m. nishizawa , `` on a q - analogue of the multiple gamma functions , '' lett . math
. phys . * 37 * , 201 ( 1996 ) [ arxiv : q - alg/9505086 ] .
|
we study supersymmetric wilson loops in the @xmath0 supersymmetric @xmath1 chern - simons - matter ( csm ) theory , the abj theory , at finite @xmath2 , @xmath3 and @xmath4 .
this generalizes our previous study on the abj partition function . first computing the wilson loops in the @xmath5 lens space matrix model exactly
, we perform an analytic continuation , @xmath3 to @xmath6 , to obtain the wilson loops in the abj theory that is given in terms of a formal series and only valid in perturbation theory . via a sommerfeld - watson type transform
, we provide a nonperturbative completion that renders the formal series well - defined at all couplings .
this is given by @xmath7-dimensional integrals that generalize the `` mirror description '' of the partition function of the abjm theory .
using our results , we find the maps between the wilson loops in the original and seiberg dual theories and prove the duality . in our approach
we can explicitly see how the perturbative and nonperturbative contributions to the wilson loops are exchanged under the duality .
the duality maps are further supported by a heuristic yet very useful argument based on the brane configuration as well as an alternative derivation based on that of kapustin and willett .
@xmath8_school of physics and center for theoretical physics _
+ _ university of the witwatersrand _ + _ wits 2050 , johannesburg ,
south africa
_ + @xmath9_department of physics _ + _ nagoya university , nagoya 464 - 8602 , japan _ + @xmath10_yukawa institute for theoretical physics _ + _ kyoto university , kyoto 606 - 8502 , japan _
+ @xmath11_hakubi center , kyoto university , kyoto 606 - 8501 , japan _
|
the magnetotransport phenomena @xcite in multiple - barrier heterostructures in which the magnetic is applied parallel to the electric field , was first reported by mendez , esaki and wang @xcite .
the experimental data acquired in the presence of a magnetic field exhibited abnormal features such as a shift of the onset voltage @xcite , an increase of the peak - to - valley current ratio , as well as the presence of additional shoulders in the valley region of the i - v characteristics @xcite .
however , these observations have not been fully accounted for on a theoretical basis .
this is partly due to the fact that a 3-dimensional treatment of the electron motion is required for rigorously analyzing the impact the quantization in the transverse plane has upon the longitudinal electron motion .
indeed relatively few works analyzing heterostructure devices with a magnetic field parallel to the electric field have been reported until now @xcite .
when the magnetic field and electric field are perpendicular to the interface plane of the composing heterostructure materials , the density of state ( dos ) of the electrons in the plane perpendicular to the superlattice direction assumes discrete values corresponding to the landau levels @xcite .
when the magnetic field is varied , the dos of the electrons changes and this strongly impacts the electron motion .
the resulting transmission probability and current - voltage characteristics of the heterostructure studied become then quite different from the case with zero magnetic field .
the additional shoulders observed in the valley region of the i - v characteristics in the presence of a magnetic field @xcite have been attributed to the longitudinal optical ( lo ) phonons which is one of the dominant scattering processes of electrons in the iii - v semiconductor heterostructure .
we present here a model of resonant - tunneling in the presence of of a magnetic field , which includes the electron - lo phonon interaction @xcite , so as to account for the transition of the incident electron to different landau states .
all possible transitions , including intra - landau state scattering will be considered in our model and numerical analysis .
the effect of the magnetic field is to quantize the energy levels of the electron into landau levels .
this is due to the fact that the projection of the helicoidal trajectory of the electron on the perpendicular plane corresponds to an harmonic oscillator . to account for the effect of this quantization on the longitudinal motion of electrons in the superlattice direction
, we will develop a 3-dimensional transport model .
we will then study the abnormal magnetotransport features with the aid of a 3-dimensional quantum simulator which numerically accounts for the contribution of the electron states in the perpendicular plane @xcite .
the heterostructure device considered consists of an undoped al@xmath0ga@xmath1as / gaas / al@xmath0ga@xmath1as , double barrier structure sandwiched by heavily doped @xmath2-gaas left and right contact layers .
our analysis only considers tunneling from electrons in the @xmath3 valley .
the @xmath3 conduction - band edge in the absence of an applied voltage for the device studied is shown in the inset of fig .
spin splitting in the magnetic field is ignored for simplicity .
finally in this paper we are interested in the region of operation for which we have @xmath4 , and our numerical calculations will be consequently performed at the low temperature of @xmath5 and for magnetic fields up to 22 tesla . in section ii , the 3d magnetotransport model used for electrons interacting with longitudinal optical phonons via polor scattering is introduced .
the electron state is expanded in the generalized wannier - landau basis .
expression for calculating the device transmission coefficient and current - voltage characteristics in the presence of phonon scattering are derived . in section iii ,
the numerical results obtained for the main abnormal magnetotransport features using the above theoretical model are presented and analyzed . finally in the last section , a brief summary of our results is given .
the hamiltonian of an electron system interacting with longitudinal optical phonons under a magnetic field in the double barrier structure is written as @xmath6 where the free and interaction terms are given by @xmath7 where @xmath8 ( @xmath9 ) are respectively the destruction and creation operators for electrons ( phonons ) and @xmath10 is the interaction weight for the phonon wave vector @xmath11 . here
@xmath12 stands for the longitudinal crystal wavevector @xmath13 and landau level @xmath14 and @xmath15 in the perpendicular plane , which results from the choice of landau gauge @xcite for the vector potential @xmath16 and @xmath11 is the mode of longitudinal optical phonon .
the lo phonons will be represented by an einstein model with the constant phonon frequency @xmath17 .
the schrdinger equation to be solved is @xmath18 we expand the electron wave function in terms of generalized wannier functions in the superlattice direction , @xmath19-direction @xcite and landau states in the @xmath20 plane , @xmath21 to solve the schrdinger equation , we calculate the matrix elements of the hamiltonian in the generalized wannier - landau basis . for the free part of hamiltonian
, we find @xmath22 where @xmath23 with @xmath24\delta_{n , n^{\prime}}\end{aligned}\ ] ] and @xmath25 , the cyclotron frequency at the lattice site @xmath26
. here , @xmath27 are the conduction band edge and applied voltage at the lattice site @xmath26 , respectively .
the band model selected consist of a tight - binding band structure in the superlattice direction and a parabolic band structure in the perpendicular plane .
note that the transversal mass in the hamiltonian matrix varies in the longitudinal direction .
it is convenient to rewrite the hamiltonian matrix element as @xmath28 in which the position ( @xmath26 ) dependent transversal component is now absorbed in @xmath29 @xmath30 which depends on the perpendicular states l , incorporates the 3-dimensional effects in the longitudinal motion of the electrons .
after some algebra , the matrix element of the interaction hamiltonian which determines the inter landau state transitions leads to @xmath31 where @xmath17 is the phonon frequency , @xmath32 is the transversal part of @xmath11 , and @xmath33 is the phase of the lo phonon of mode @xmath11 ( for a more detailed formalism , see ref . [ 9 ] ) .
the coefficients @xmath34 is given by @xmath35 which corresponds to the emission and absorption of phonons , respectively .
here , @xmath36 is the vacuum dielectric constant , and @xmath37 is the equilibrium number of phonons .
@xmath38 is given as @xmath39 with phonon energy @xmath40 mev and @xmath41 for gaas .
at liquid helium temperature , the number of optical phonons @xmath37 is very small , so only the emission of phonons is possible . in eqn.(10 ) , we introduced the coefficients @xmath42 } \bar{f}_{l^{\prime }
l } ( q^2_{\perp } l^2/2),\end{aligned}\ ] ] with @xmath43 where @xmath44 and @xmath45 are the maximum and the minimum of @xmath14 and @xmath46 , respectively , @xmath47 , and @xmath48 is an associated laguerre polynomial @xcite . using the obtained matrix elements , eqn.(5 ) and ( 10 ) , in the above subsection
, schrdinger equation can be reduced to the following @xmath49-dimensional envelope equation for @xmath50 , @xmath51 , \label{env}\end{aligned}\ ] ] where @xmath52 , is the cyclotron frequency at the left contact .
let us label the quantum state of an incident electron injected at the left contact with the energy @xmath53 using the quantum numbers @xmath54 .
an electron in the incident ballistic state @xmath54 which is scattered by a phonon will transit to other accessible states @xmath55 with total energy @xmath56 following the emission ( - ) or absorption ( + ) of a phonon .
therefore the solution of the envelope eqn.([env ] ) is chosen to be of the following form @xmath57 where @xmath17 is the phonon frequency @xmath58 , @xmath59 is the coherent solution for the envelope equation and where @xmath60 are the scattered waves due to absorption or emission of a phonon , respectively . after substituting this expression into the envelope equation and taking an ensemble average @xcite
, we arrive at @xmath61 where @xmath62 is the self energy due to absorption and emission of phonons .
the self energy is verified to be given by @xmath63 ^ 2,\ ] ] and @xmath64 is the longitudinal green s function in the generalized wannier basis evaluated using the relation @xmath65 for an incident electron state @xmath66 the current entering on the left hand side and leaving the quantum region on the right hand side are respectively given by @xmath67 here the phonon - assisted current component @xmath68 is given by @xmath69 ^ 2 .\ ] ] in the above equations , @xmath70 and @xmath71 are the electron velocity at the left and right contact , and @xmath72 is the lattice - site index of the right contact .
the transmission probability from the left to right contact is then defined as @xmath73 } { j_{i0}(e_{0x},l_0 ) } \ ] ] before calculating the current , the electron chemical potential @xmath74 at the left contact corresponding to the doping density @xmath26 in presence of the magnetic field @xmath75 , should be determined . as is shown in fig .
[ efbc ] , the chemical potential is found to exhibit an oscillatory behavior in @xmath76 as the magnetic field varies .
the chemical potential @xmath74 in the left contact is obtained from the carrier density through the following integration relation , @xmath77 where @xmath78 is a degeneracy factor in the plane perpendicular to the superlattice direction and where the fermi - dirac function @xmath79 is given by @xmath80 + 1}.\ ] ] here we have found it convenient to introduce the effective chemical potential @xmath81 for the landau sub - band @xmath82 which is related to the actual electron chemical potential @xmath74 according to @xmath83 @xmath81 can be interpreted as the effective chemical potential of the electrons in the landau state @xmath82 in the presence of a magnetic field .
thus , the total current density is @xmath84,\ ] ] where the second term in the right - hand side is the backward contribution to the total current . at very low temperatures such as 4.2 k ,
the current for a occupied sub - band @xmath82 reduces to @xmath85 . \label{cur}\ ] ] therefore , the area from 0 to @xmath81 of the transmission probability determines the contribution of each occupied sub - band to the total current .
as mentioned above the heterostructure device considered consists of an undoped al@xmath0ga@xmath1as / gaas / al@xmath0ga@xmath1as , double barrier structure ( see inset of fig . [ efbc ] ) sandwiched by heavily doped @xmath2-gaas left and right contact layers .
the barriers and well are both @xmath86 wide .
the donor density in the contact regions is @xmath87 .
the donors are assumed to be completely ionized due to the heavy doping density .
the value of the electron effective mass used for gaas and al@xmath0ga@xmath1as is @xmath88 and @xmath89 , respectively .
the numerical calculation starts with the calculation of the matrix element @xmath90 which is required to evaluate the self energy @xmath91 associated with the inter landau state transitions for all possible values of @xmath66 .
this self energy is then used to solve the envelope equation , and obtain the transmission coefficients .
the current density is finally obtained by integrating the transmission coefficient obtained over the incident energy .
we confined our numerical analysis to a single - sequential phonon scattering event since this is a quite reasonable approximation for the small heterostructure device under investigation . in order to calculate the currents , it is required to calculate the chemical potential @xmath92 as a function of the applied magnetic field @xmath75 . using eqn.([mub ] ) the chemical potential @xmath74 calculated for several temperatures
is plotted in fig .
[ efbc ] versus @xmath75 .
the chemical potential @xmath74 is found to be a periodic function in @xmath76 .
this figure clearly indicates that in the region of strong magnetic fields ( @xmath93 telsa ) using the approximation of a constant fermi energy @xmath94 would introduce a rather large error in the chemical potential .
this would in turn adversely affect the accuracy of the device current calculated as is verified in fig .
however at the temperature of @xmath95 , the oscillating behavior in @xmath92 disappears completely due to thermal population of all the landau states near the fermi energy . in fig .
[ teb ] , the transmission probability in a flat band ( no applied voltage ) is plotted for two magnetic fields of @xmath96 and @xmath97 telsa at @xmath5 . at this low temperature ,
the emission of phonons dominates since the number of phonons in the system is negligible .
the peaks beyond the main peak centered at 94 mev , are phonon - assisted resonant tunneling peaks which satisfy the following energy conservation relation @xmath98 where @xmath66 is the incident electron state and @xmath99 is the scattered state after the emission of one phonon and with @xmath100 .
the energy of the phonon is chosen to be @xmath101 mev for gaas contact layers for simplicity . a single - sequential phonon - emission has been assumed .
this was reported to be a valid approximation in a small size system like the double barrier structure considered @xcite the peak indicated by @xmath102 in fig .
[ teb ] arises when an incident electron moves in the structure with the emission of a single phonon and the landau level remains conserved while the peaks @xmath103 and @xmath104 indicate that phonon - assisted resonant - tunneling induced a transition to higher landau levels . as the magnetic field
is increased , the spacing @xmath105 between @xmath106 neighboring peaks increases in proportion to the applied external magnetic field where as the spacing between the main peak and @xmath102 peak remains constant as it is set by the phonon energy @xmath107 . for an incident state with landau levels @xmath108 ,
transitions of the type @xmath109 arise ( with @xmath26 an integer verifying @xmath110 ) , and additional peaks appears between the main and @xmath102 peaks .
the matrix element @xmath90 between landau states in eqn.(13 ) determines the strength of the inter landau state transitions in the transmission coefficient .
for example , at @xmath96 telsa its argument @xmath111 assume values in the range @xmath112 depending on the phonon mode . after the matrix @xmath90 is summed over all possible phonon modes @xmath32 , the inter landau state transition probability is obtained .
the analysis of the various scattering - assisted resonant - tunneling processes induced by lo phonon emission , indicates that the transmission probabilities for scattering events involving the transition between two different landau states ( @xmath113 ) is of the same order as the transmission probability for intra landau state ( @xmath114 ) scattering events .
[ teb ] clearly shows this property .
[ ivb1 ] , [ ibs ] , and [ ivb2 ] shows the electron current density calculated under various conditions .
the current density is obtained from the integration over the incident electron energy of the transmission coefficient @xmath115 multiplied by the fermi - dirac occupation function which is almost a step function at the low temperature considered .
a broadening of the landau levels up to @xmath104 mev was considered , but there were no noticeable difference in our results for the calculated currents for @xmath116 telsa since we have @xmath117 mev .
we observe in fig .
[ ivb1 ] and [ ivb2 ] , the presence of small additional shoulders of almost equal width in the valley region of the current - voltage curve .
these shoulders originate from that fact that the phonon - assisted resonant - tunneling peaks in the transmission coefficient are spaced by @xmath118 above the main quasi - resonant energy peak @xmath119 . before discussing the impact of phonon scattering ,
let us first briefly describe the main effects of the magnetic field upon the current of the resonant tunneling diode ( rtd ) in the absence of phonon scattering . as can be seen in fig .
[ ivb1 ] and [ ivb2 ] the magnetic field is found to induce a shift of the onset voltage of the rtd i - v characteristic by the amount @xmath120+\hbar w_{c0}(b)/2 $ ] , and to enhance the value of peak current of the rtd .
[ ibs ] reveals another interesting feature regarding the dependence of the rtd current - voltage characteristics upon the magnetic field . in this figure
the diode current is measured at @xmath121 volt which is in the current peak region of the rtd i - v curve .
note that phonon scattering was not included in this calculation as coherent ( ballistic ) tunneling gives the dominant contribution to the peak current .
the rtd current measured at that fixed diode voltage is seen to exhibits an oscillatory behavior versus @xmath76 which increases in amplitude as the magnetic field is increased . the current maxima observed in fig .
[ ibs ] for the current versus @xmath75 are found to occur for values @xmath122 of the magnetic field satisfying the following relation @xmath123 where the quasi - resonant level @xmath124 is varying with the diode voltage .
this leads to the interesting result that its periodicity versus @xmath76 which is approximately @xmath125^{-1}$ ] , depends on the quasi - resonant energy level @xmath126 ; for a smaller value of @xmath127 , one obtains an increased oscillatory behavior in the current - magnetic field curve . as can be seen fig .
[ ivb1 ] and [ ivb2 ] , the effect of phonon scattering on the rtd current density is to shift the current - voltage characteristic downward compared to the one obtained in the absence of scattering .
another effect is to increase the current and to introduce additional shoulders in the valley region of the current - voltage characteristic . the downward shift of onset voltage results from the negative value of the real part of the self energy induced by the scattering of the electron by the phonons . for magnetic fields of @xmath96 and @xmath97 telsa ,
the downward shifts obtained in the current - voltage characteristic are found to be @xmath128 and @xmath129 mv , respectively .
the small shoulders at @xmath130 telsa in the current valley region of fig .
[ ivb2 ] become larger than for the case of @xmath96 telsa .
thus in stronger magnetic fields , these additional shoulders in the valley region become higher and their intervals becomes wider , which agrees with the reported experiments @xcite to account for the origin of the shoulder at @xmath131 volt , let us consider the inset of fig .
[ ivb2 ] . at @xmath130 telsa , only the two lowest landau levels @xmath132 and @xmath133 contribute to the diode current .
the effective chemical potentials defined in eqn.(25 ) for the @xmath132 and @xmath133 states are respectively @xmath134 and @xmath135 mev .
they are indicated by two arrows in the inset of fig .
[ ivb2 ] . at @xmath131 volt ,
both the peaks @xmath102 and @xmath136 of transmission coefficients for the @xmath132 and @xmath133 incident states are respectively included into the integration range of the current eqn.([cur ] ) , explaining the formation of the shoulders . currently , it is difficult to compare our theoretical results in a quantitative way to the experiments reported in ref . 2 and 3 .
the reasons are as follows .
first , the authors of ref . 2 concluded that additional experiments are needed to confirm the weak minima of their preliminary data in the valley current region for strong magnetic fields .
second , our test device does nt have a spacer layer in the emitter region , while the device structure of ref .
3 have a spacer layer where a discrete triangular well state is formed .
the shoulders therein corresponding to @xmath137 beyond the main peak was reported to be as high as almost 2/5 of the main peak for extremely strong magnetic field , which were clarified to originate from the assistance of lo phonons to the electron current .
our results together with ref . 2 and 3 lead to the conclusion that the spacer layer tends to enhance the strength of the shoulders beyond the main peak .
but in order to explain such high shoulders , other contributions due to scatterings of electrons by interface roughness , acoustic phonon , etc . , might also be needed .
in our paper we have investigated the transport of electrons interacting with lo phonons in a double - barrier heterostructure with a magnetic field applied parallel to the electric field .
the effect of the magnetic field is to quantize the energy of the electron in the perpendicular plane to the magnetic field . to account for the effects of a magnetic field applied in parallel with the superlattice direction ,
a 3-dimensional transport model was developed which accounts for the variation of the transverse mass and the associated variation of the cyclotron frequency across the heterostructure . in order to determine the inter landau state transitions in the transmission probability , all the matrices @xmath90 for allowed phonon modes
have been calculated numerically .
it was established that the inter landau state transitions due to lo phonon scattering are as probable as intra landau state scattering . for a fixed diode voltage in the peak region of the rtd current - voltage curve ,
the rtd current versus the magnetic field was found to be a periodic function of @xmath76 .
the current maxima have an @xmath76 period which is approximately given by @xmath138 .
the dependence of this @xmath76 period on the quasi - resonant energy level @xmath126 of our test double barrier structure indicates that magnetotransport can be used for the spectroscopic analysis of rtds for diode voltages in the peak current region .
a theory and associated numerical analysis using scattering matrix elements was presented to describe the impact of electron - lo phonon interaction in the i - v curve in the presence of both parallel electric and magnetic fields .
the results obtained were found to be in qualitative agreement with the experimental results reported by ref . 2 and 3 .
however for further understanding and improved quantitative fit of the weak shoulders in the valley region of the double barrier iv , more theoretical work on magnetotransport including scattering processes such as interface roughness and acoustic phonon scattering , etc . , is required in addition to lo phonon scattering studied in this paper .
+ this work was supported in part by the korea science and engineering foundation ( kosef ) .
the first author is grateful to the ee department of the ohio - state university for hosting him during the course of this research .
the authors would like to thank the anonymous reviewers for their suggestions .
b. k. ridley , _ quantum processes in semiconductors _
( oxford university press , oxford , 1999 ) .
e. e. mendez , l. esaki , and w. i. wang , , 2893 ( 1986 ) .
m. l. leadbeater , e. s. alves , l. eaves , m. henini , o. h. hughes , a. celeste , j. c. portal , g. hill , and m. a. pate , , 3438 ( 1989 ) .
l. canali , m. lazzarino , l. sorba , and f. beltram , phys .
* 76 * , 3618 ( 1996 )
. k. s. chan , f. w. sheard , g. a. toombs , and l. eaves , 1447 ( 1997 ) . l. d. landau and e. m. lifshitz , _ quantum mechanics : non relativistic theory _
( pergamon press , oxford , 1965 ) .
h. asahi , m. tewordt , r. t. syme , m. j. kelly , v. j. law , d. r. mace , j. e. f. frost , d. a. ritchie , g. a. c. jones , and m. pepper , , 803 ( 1991 ) . for certain devices with p - type si / sige double barrier structures which has quite different band structures from our case
, the shoulders beyond a main peak can be induced due to non - parabolic , light and heavy hole bands even in the absence of other scattering events .
see , for example , a. zaslavsky , d. a. grtzmacher , s. y. lin , t. p. smith iii , r. a. kiehl , and t. o. sedgwick , , 16 036 ( 1993 ) ; d .- y .
lin , c .- w .
chen , and g. y. wu , , 4599 ( 1997 ) .
n. s. wingreen , k. w. jacobsen , and j. w. wilkins , phys .
* 61 * , 1396 ( 1988 ) .
p. roblin and w .- r .
liou , , 2146 ( 1993 ) ; p. sotirelis and p. roblin , , 13381 ( 1995 ) . w. kohn and j. r. onffroy , , 2485 ( 1973 ) ; j. g. gay and j. r. smith , , 4906 ( 1975 ) .
d. pfannkuche and r. r. gerhardts , , 12 606 ( 1992 ) .
|
the transport in double - barrier heterostructures of electrons interacting with longitudinal optical phonons in the presence of parallel electric and magnetic fields is analyzed theoretically with the aid of a 3-dimensional quantum transport simulator .
inter landau state transitions induced by lo - phonon scattering - assisted resonant - tunneling is shown to be an important process with a probability comparable to that of intra landau state scattering .
analysis of the current - voltage characteristics reveals also that the current peak is a periodic function of the inverse of the magnetic field , with a period dependent on the quasi - resonant energy level .
|
it has been known for quite a long time that superconductivity ( s ) and ferromagnetism ( f ) are two antagonistic orderings and that they can hardly coexist in a same compound .
although they tend to avoid each other , their mutual interaction may be studied when they are spatially separated , which is realized in superconductor / ferromagnet ( s / f ) hybrid structures ( as a review , see @xcite ) . indeed , in such systems , superconductivity and ferromagnetism
can influence each other through the so called proximity effect. the main peculiarity of the proximity effect in s / f structures is the damping oscillatory behavior of the superconducting correlations in the f layers ( while they monotonically decay in normal layers of superconductor / normal metal structures ) . in the dirty limit and for large exchange field , the characteristic lengths of the decay and oscillations are the same .
they are given by @xmath0 , where @xmath1 is the diffusion coefficient in the ferromagnet and @xmath2 is the exchange field acting on the electrons spins .
this unusual proximity effect leads to several striking phenomena , such as the non - monotonic dependence of the critical temperature and current of s / f multilayers on the f layers thickness and the realization of the so - called @xmath3 junction in s / f / s trilayers @xcite . although the existing theory provides rather good qualitative description of the observed _ _ _ _ effects , there is still no complete quantitative agreement with experiments .
this indicates that besides the exchange field , some additional pair - breaking mechanisms are present in the f layers .
indeed , spin - flip process is inherent to the ferromagnetic layers ( because of magnetic impurities , spin - wave or non stoichiometric lattices ) and may have dramatic consequences on superconductivity ( contrary to non magnetic impurities that have very little impact ) .
such a pair - breaking mechanism also arises in usually used weak ferromagnetic alloys , because they are close to ferromagnetism disappearance and then quite favorable to large magnetic disorder .
this can be inferred for instance from the very strong decrease of the critical current of s / f / s junctions as a function of the thickness of the ferromagnetic layer in experimental studies ryazanov2001,sellier , ryazanov1 . in such experiments ,
the ferromagnetic alloys used were cu@xmath4ni@xmath5 with @xmath6 , limit range of concentration for ferromagnetic properties .
in addition , the pair destruction due to spin - orbit interaction must be taken into account as well .
though the spin - flip and spin - orbit interactions in the ferromagnetic material were introduced in several recent theoretical studies on proximity effect in s / f hybrids , only few simple limiting cases were considered ( we ll give appropriate references below ) .
thus the problem of quantitative description of these effects in s / f multilayered systems is still unsolved . in this paper
we present the results of detailed theoretical study of the influence of spin - flip and spin - orbit scattering mechanisms on the critical temperature ( @xmath7 ) and critical current ( @xmath8 ) of s / f multilayered systems .
we obtain analytical and numerical solutions of the problem which provide the basis not only for qualitative understanding of experimental results but also to fit the data quantitatively .
in particular , we show that spin - orbit and spin - flip scattering mechanisms influence differently the properties of s / f structures : spin - orbit mechanism can destroy the @xmath7 and @xmath8 oscillations while spin - flip scattering can only modify them .
we also report on a striking non - intuitive behavior that the critical temperature and current can acquire with the variation of the s / f interface transparency__. _ _
a very convenient set of equations for an inhomogeneous superconductor was elaborated by eilenberger @xcite .
however , the eilenberger equations can be replaced by the much simpler usadel equations @xcite when the electron scattering free path in s / f systems is shorter than the superconducting length , which is often the case .
these equations are nonlinear but can be simplified when the temperature is close to the critical temperature @xmath9 or at any temperature in the f layer when the transparency is low .
we consider a s / f multilayered system where all physical quantities depend only on the coordinate @xmath10 perpendicular to the layers .
the natural choice of the spin - quantization axis is along the direction of the exchange field . in the general case , magnetic and spin - orbit scatterings mix up the up and down spin states .
therefore , two anomalous green functions , namely @xmath11 and @xmath12 are needed to describe this situation .
the linearized usadel equations may be written for @xmath13 as @xcite @xmath14 in the s layers and @xmath15 in the f layers , where @xmath16 is the superconducting order parameter , @xmath2 is the exchange field , @xmath17 is the diffusion coefficient in the s(f ) layer and @xmath18 are the matsubara frequencies , @xmath19 .
the parameter @xmath20 is the spin - orbit scattering time .
the magnetic scattering times are @xmath21 and @xmath22 .
the rate @xmath23 is proportional to the square of the exchange interaction potential ( the notations are the same as in @xcite ) .
note that the microscopical green functions theory of superconductors with magnetic impurities and spin - orbit scattering was proposed by abrikosov and gorkov @xcite .
besides , the usually used ferromagnets present a strong uniaxial anisotropy . in that case
, the perpendicular fluctuations of the exchange field are suppressed , that is @xmath24 .
therefore , henceforth , @xmath25 will be noted as the magnetic scattering time @xmath26 .
the spin - flip scattering is now simply incorporated replacing @xmath27 by @xmath28 in the standard usadel equation , and @xmath26 is the magnetic scattering time ( see for example @xcite ) . note that the usadel equations in the f layers are not coupled anymore when @xmath29 , and only one equation is needed .
the common feature of all the s / f bilayered and multilayered heterostructures is the non monotonic evolution of the critical temperature @xmath7 with the thickness of the ferromagnetic layer .
this behavior was first predicted by buzdin and kuprianov _ _ _ _ @xcite and _ _ _ _ radovic _ _ et al . _ _
@xcite , and was since then intensively studied both theoretically and experimentally ( as a review , see @xcite ) .
the presence of magnetic scattering can result in an additional decrease of the transition temperature ( on the contrary , nonmagnetic impurities do not affect the transition temperature ) . note that the theoretical description of @xmath7 for ferromagnetic layers with spin - orbit scattering was performed by demler _
_ @xcite and oh _ et al .
_ @xcite . in the present section ,
we therefore mainly focus on the influence of spin - flip process and neglect spin - orbit scattering .
we report here on the influence of the spin - flip scattering on the non monotonic dependence of @xmath7 when the thickness of the superconducting layer is supposed to be small , _ i. e. _ @xmath31 . in that case
, an analytical solution may be obtained .
it should also be underlined that the question of the spin - flip role was first addressed by tagirov @xcite in the discussion of @xmath7 of sf multilayers and by falko _ et al .
_ @xcite in the study of the resistance of a diffusive f / n junction .
we consider a s / f multilayered system with f layers of thickness @xmath33 and parallel magnetization directions and s layers of thickness @xmath34 , see fig .
1 ( this system is also equivalent to a s / f bilayer of thicknesses @xmath35 and @xmath36 respectively ) .
further , we assume that the sf interfaces are not magnetically - active , i.e. there is no rotation of the quasipartice spin at the interfaces as considered in refs .
rainer , fogel1,fogel2,eschrig . under these conditions
, long - range spin - triplet superconductivity does not appear @xcite .
discussion of the role of spin - flip scattering in a ferromagnet in combination with magnetically - active sf interfaces and/or noncollinear magnetizations requires separate study .
the anomalous green s function @xmath37 varies a little in the s layer and may be approximated by a simple expansion up to the second order @xmath38 ( see for example @xcite and @xcite ) @xmath39where @xmath40 is the value of the anomalous green s function at the center of the s layer . moreover , in that case , the spatial variation of the pair potential @xmath41 can be neglected @xmath42 .
it follows from eq .
( [ usadel_lin_s ] ) that @xmath43 where @xmath44 is the complex pair - breaking parameter . the usadel equation in the s layer is completed by the self - consistency equation@xmath45 where @xmath9 is the bare transition temperature of the superconducting layer in the absence of the proximity effect .
hence , this self - consistent equation gives for @xmath30@xmath46 if the temperature variation is small ( @xmath47 ) , ( temperature critique ) becomes @xmath48 the boundary conditions for the linearized usadel equation are lukichev @xmath49@xmath50 with @xmath51 the conductivity of the f(s ) layer , @xmath52 and @xmath53 is the interface transparency , related to the s / f resistance per unit area @xmath54.they lead to the determination of the pair - breaking parameter @xmath55 @xmath56 next , the resolution of the usadel equation in the f layers ( usadel_lin_f ) with symmetry consideration gives rise to the expression of the anomalous green s function in the f layer @xmath57 , \label{f0phase}\ ] ] in the 0 phase and @xmath58 , \ ] ] in the @xmath59 phase .
if @xmath60 , @xmath2 , we may neglect the matsubara frequencies in the expression for @xmath61 that becomes @xmath62 with @xmath63 and @xmath64 .
_ _ _ _ the two parameters @xmath65 and @xmath66 are respectively the decay characteristic length and the oscillations period , and may be written as@xmath67 for @xmath68 , eqs .
( [ decay ] ) and ( [ period ] ) reduce to @xmath69 . as expected , it is found that the decay length and oscillations period are the same in absence of spin - flip scattering .
the pair breaking parameter may be determined and it reads@xmath70 in the 0 phase and @xmath71 in the @xmath59 phase , where @xmath72 , @xmath73 , @xmath74 and @xmath75 . first , if the interface is supposed to be transparent , analytical expressions of the variation of the temperature may be found if @xmath76 : @xmath77 where two dimensionless parameters have been introduced , namely @xmath78 and @xmath79 , so that @xmath80 . in the general case however a numerical analysis has to be performed .
the ratio of the characteristic lengths _ _ _ _ @xmath81 clearly shows that the magnetic scattering decreases the decay length and increases the oscillation period . if @xmath82 , @xmath65 can become much smaller than @xmath66 .
besides , the decrease of @xmath65 makes the observation of the oscillations more difficult . the evolution of the critical temperature without and with spin - flip scattering is given in fig .
the phase that really occurs is the one with higher critical temperature .
it is seen that @xmath65 decreases in presence of spin - flip while @xmath66 increases .
the influence of the interface transparency parameter @xmath84 on @xmath7 of sf bilayers was studied before by aarts _
@xcite , fominov _ et al .
_ @xcite and tagirov tagirov . here
we extend this discussion taking into account spin - flip scattering in the f - layer and considering broader range of interfacetransparencies and demonstrate that new effects take place in this situation .
the intriguing evolution of @xmath85 with the interface transparency parameter @xmath86 must be underlined ( see fig .
3 ) . if @xmath87 , there is no magnetic scattering and it could intuitively be believed that , the higher the barrier , the less is the influence of the proximity effect on the s layer and therefore , @xmath88 .
however , it can be seen from fig .
3 that the critical temperature is a decreasing function of the interface transparency parameter @xmath86 for a small thickness of the f layer .
this counter - intuitive behavior can be qualitatively understood for a s / f bilayer .
the probability for a cooper pair to leave the s layer is smaller for a low transparent interface @xmath89 .
nevertheless , the probability for this pair to come back again in the s layer is much higher for a transparent interface . indeed ,
when the f layer is thin enough , the reflection of the cooper pair at the other interface of the f layer allows the pair to cross again the first interface , which is easier when @xmath86 is small .
consequently , the staying time in the f layer increases with the increase of the barrier , and when this time becomes higher than the coherence time of the cooper pair , the pair is destroyed , leading to a weakened superconductivity . therefore , the critical temperature decreases with the barrier in that case . on the other hand , if @xmath90 increases , the cooper pair is hardly reflected by the external interface of the f layer whatever the value of @xmath86 is and the critical temperature is expected to increase with the barrier .
besides , an analytical approach using formula ( [ tau ] ) leads to the determination of the critical temperature when @xmath91 . in that case , the pair breaking parameter in the 0 phase becomes @xmath92 two behaviors emerge whether @xmath93 or not . indeed , if @xmath94 , the critical temperature evolution is described by @xmath95 which is an increasing function of @xmath96 . on the contrary , if @xmath97 , the critical temperature decreases when @xmath96 increases , following @xmath98 the critical temperature evolution with @xmath96 presented in fig .
3 is therefore understood . in presence of small magnetic scattering ,
the critical temperature evolution remains qualitatively the same .
as shown in fig .
3 , it appears that superconductivity may be less destroyed when there is spin - flip scattering . in that case , the pair breaking parameter becomes @xmath99 the critical temperature is described by @xmath100 when @xmath101 , and magnetic scattering leads to a slight enhancement of the transition temperature . on the contrary , if @xmath102 @xmath103 and spin - flip implies the decrease of the transition temperature .
let us now consider briefly the general case , with spin - orbit and/or perpendicular spin - flip scattering .
an additional parameter has to be introduced , namely @xmath104 and the parameter @xmath105 now becomes @xmath106 in that case , expressions ( [ tau-0 ] ) and([tau pi ] ) are modified . in the 0 phase ,
@xmath107 is replaced by @xmath108 where @xmath109 becomes @xmath110 and @xmath111 is @xmath112 in the @xmath59 phase , @xmath113 is replaced by @xmath114 therefore , the influence of perpendicular spin - flip scattering and spin - orbit scattering is quite similar to the influence of parallel spin - flip processes , in the sense that it also implies the decrease of the decaying length and the increase of the oscillations period .
however , a special situation arises when @xmath115 .
then , the oscillations of the cooper pair wave function are completely destroyed .
similar conclusion for spin - orbit mechanism was obtained in @xcite .
in fact , the influence of the perpendicular magnetic scattering is analogous to the spin - orbit scattering ( see fig .
4 ) . probably
the role of perpendicular spin - flip or spin - orbit scattering is important for the understanding of experimental results where no oscillation of the critical temperature was detected . besides
, note that the critical temperature oscillations can not disappear when there is only parallel spin - flip .
( cgs ) , @xmath116 and @xmath117.,width=268 ]
a josephson junction may be realized with s / f / s sandwiches in which the weak link between the two superconductors is ensured by the ferromagnetic layer .
the supercurrent @xmath119 flowing across the structure can be expressed as @xmath120 , where @xmath8 is the critical current and @xmath121 stands for the phase difference between the two superconducting layers .
a standard junction has at equilibrium @xmath122 and @xmath123 , and therefore , no current exists .
it may appear however that @xmath8 becomes negative , which implies that the equilibrium phase difference is @xmath124 and the ground state undergoes a @xmath59 phase shift .
this so called @xmath59 junction was first predicted for s / f / s structures in the clean limit @xcite , and later in the more realistic case of the diffusive limit kuprianov .
the critical current @xmath8 in s / f / s junctions is controlled by the exchange field in the ferromagnet , the interface transparency parameter @xmath125 , the spin - flip and spin - orbit scattering rates . in this section
we will discuss the influence of these parameters on @xmath8 , paying the main attention to the role of spin - orbit and magnetic scatterings . besides , note that the spin - orbit effect on the critical current has already been studied for s / f bilayers separated by an insulator with thin f layers @xcite and for the fsfsf structure with noncollinear magnetizations of f - layers @xcite .
we will consider a symmetric s / f / s structure with the f layer having thickness @xmath33 and assume that the dirty limit conditions are fulfilled for s and f materials .
we will assume that the rigid boundary condition @xmath126 is fulfilled , when the suppression of superconductivity in s can be neglected and the linearized usadel equations ( [ usadel_lin_f ] ) can be used .
the case of transparent interfaces , @xmath127 , will be considered at the end . to derive the general expression for @xmath8 we should supplement ( usadel_lin_f ) by the boundary conditions @xcite at s / f interfaces @xmath128@xmath129 sign(\omega ) , \label{bc_sfs}\]]@xmath130 the solution of the boundary problem ( [ usadel_lin_f ] ) , ( [ bc_sfs ] )
has the form @xmath131 + b_{\omega+}\sinh\left [ k_{+}x\right ] \right\ } + ( 1-\eta_{\omega})\left\ { a_{\omega-}\cosh\left [ k_{-}x\right ] + b_{\omega -}\sinh\left [ k_{-}x\right ] \right\ } ) , \label{solp}\]]@xmath132 + b_{\omega+}\sinh\left [ k_{+}x\right ] \right\ } -(1+\eta_{\omega})\left\ { a_{\omega-}\cosh\left [ k_{-}x\right ] + b_{\omega -}\sinh\left [ k_{-}x\right ] \right\ } ) , \label{solm}\ ] ] where the coefficients @xmath133 and are given @xmath134 in the appendix , while @xmath135 the parameter @xmath136 is defined by @xmath137 , @xmath138 and @xmath139 .
note that in the general case perpendicular spin - flip has to be added in @xmath136 and @xmath105 .
the general expression for the supercurrent is @xmath140 , \label{curr}\ ] ] where @xmath141 . substituting ( [ solp ] ) and ( [ solm ] ) into the above expression and taking into account the symmetry relations ( given in the appendix ) , we get @xmath142 . for large spin - orbit scattering , i. e. @xmath143
, the critical current @xmath8 equals to @xmath144 where @xmath145 + 2g_{s}\gamma_{b}\xi_{n}k_{\pm } \cosh\left [ k_{\pm}d_{f}\right ] } .\ ] ] at larger exchange field @xmath146 , the critical current becomes @xmath147 } { ( ( \gamma_{b}\xi_{n}k)^{2}+g_{s}^{2})\sinh\left [ kd_{f}\right ] + 2g_{s}\gamma_{b}\xi_{n}k\cosh\left [ kd_{f}\right ] } \right\ } , \label{ic}\ ] ] where @xmath148 and @xmath80 , with @xmath149@xmath150 in appropriate limits , expressions ( [ ics]),([ic ] ) transform into the previously obtained results @xcite , @xcite , @xcite,@xcite,@xcite .
this approach is valid if @xmath151 or @xmath152 as follows from ( [ ic ] ) - ( [ q ] ) , in the case of strong spin - orbit scattering @xmath153 the critical current decays monotonically with the increase of @xmath33 with two decay lengths @xmath154 defined by eq.([etibe ] ) .
it is seen from ( [ etibe ] ) that in the limit @xmath155 , the parameter @xmath156 as a result , the contribution to the critical current in ( [ ics ] ) comes only from @xmath157 component with the length scale @xmath158 which describes the case of an s / n / s junction in which spin - orbit scattering does not influence @xmath159 with the increase of @xmath2 , the contribution to @xmath8 from the faster decaying @xmath160 component @xmath161 also increases and the difference between @xmath162 and @xmath163 decreases@xmath164 finally , when @xmath165 , both scales coincide , @xmath166 , and the components @xmath167 provide equal contributions to @xmath159 for relatively weak spin - orbit scattering , @xmath168 , the dependence @xmath169 follows the damped oscillation law , when two length scales @xmath170 can be introduced describing respectively the decay and the oscillation period of @xmath169 . in this section , we will concentrate on the case @xmath168 .
the scales @xmath65 , @xmath66 are related to @xmath171 and will be discussed in detail below in different limits . in the limit of small thickness @xmath173 and large interface transparency parameter @xmath174 we can neglect the terms @xmath175 and @xmath176 in the denominator of eq.([ic ] ) and with the accuracy of better than @xmath177 get @xmath178where @xmath179 is the normal junction resistance , @xmath180
@xmath181 and@xmath182if additionally @xmath183 h}\]]then @xmath184for @xmath185 and @xmath186 , the sum in ( [ ic_small_da ] ) can be calculated by transforming from summation into integration over @xmath18 resulting in _ _ _ _ @xmath187 , \label{ic_small_da0}\]]from ( [ ic_small_da0 ] ) it follows that the transformation from @xmath188 to @xmath3 junction may occur when the thickness @xmath33 exceeds some critical value @xmath189 @xmath190for @xmath191 this result agrees with buzdin2003 .
it is interesting to note that _ _
_ _ in this range of parameters _
_ , _ _ the condition ( [ cond_pi_sd ] ) of the transition to @xmath3 state does not depend on the properties of the interfaces . as is also seen from eq .
( [ cond_pi_sd ] ) , both spin - orbit and spin - flip scattering increase the thickness @xmath189 corresponding to the first @xmath188 to @xmath3 transition . if @xmath192 , then the term @xmath193 in ( [ ic_small_d ] ) is not small . as a result ,
the transition to a @xmath3 state should depend on the properties of interfaces and occurs at @xmath2 larger than the critical value @xmath194 following from eq.([ic_small_da0 ] ) .
therefore , in the limit of small @xmath195 and large @xmath196 the transition from @xmath188 to @xmath3 junction exists only if the exchange energy @xmath197 sufficiently exceeds @xmath198 .
the smaller the interface transparency parameter @xmath84 , the larger should be the exchange energy @xmath2 .
the possibility of the 0-@xmath3 transition in junctions with small thickness @xmath33 in the case of large @xmath84 is related to the multiple scattering at the boundaries . as a result
, the electrons spend more time in the f layer and its influence on superconductivity is enhanced .
this is the manifestation of the same mechanism that leads in s / f bilayers to the critical temperature decrease with the increase of the interface transparency parameter ( see section ii b ) . in the limit of large @xmath199 for the critical current from ( ic ) - ( [ q ] ) , we have @xmath200 where the coefficients @xmath201 and @xmath202 are defined by@xmath203 if additionally @xmath204 then both @xmath205 and @xmath206 may be considered as independent on matsubara frequencies , since the sum in ( [ ic ] ) converges at @xmath207 . in this case @xmath208
\exp\left\ { -\frac{d_{f}}{\xi_{f1}}\right\ } , \label{id1}\ ] ] where @xmath209@xmath210 the two characteristic length scales are given by the following expressions @xmath211@xmath212 which generalize eqs .
( [ decay ] ) and ( [ period ] ) for the case of the presence of the spin - orbit scattering in a ferromagnet .
one can see that with the increase of both scattering rates @xmath213 and @xmath214 the decay length @xmath65 decreases , while the oscillation period @xmath66 increases .
for a weak exchange field @xmath215 and sufficiently high temperatures @xmath216 , with @xmath217 , only the first term with @xmath218 in ( [ ic ] ) is important and we have @xmath219 where @xmath220 and @xmath221 and @xmath222 become @xmath223@xmath224 the parameters @xmath225 , @xmath226 , @xmath227 , @xmath228 are obtained replacing @xmath27 by @xmath229 in expressions ( [ p ] ) , ( [ q ] ) , ( [ uv ] ) .
the two characteristic lengths may be written in that case as @xmath230@xmath231 for @xmath232 , eqs.([ksi12 ] ) , ( [ ksi22 ] ) reduce to a simple expression @xmath233^{1/2}$ ] @xcite which describes the temperature variations of both length scales .
more precisely , in the considered limits , @xmath234 , while @xmath235 .
one can see that the scattering rates @xmath214,@xmath213 make the @xmath236 variation of @xmath237 weaker . from ( [ id1 ] ) , ( [ id2 ] )
, one can derive the condition for @xmath238 and obtain @xmath239 when the transitions between @xmath240 and @xmath59-states occur .
the position of the first zero , @xmath241 depends both on the material parameters of the ferromagnetic layer and the properties of the interfaces and superconducting electrodes , while the distance between the zeros is the function only of @xmath66 and therefore depends only on the transport parameters of the ferromagnetic material . from the structure of coefficients @xmath201 and @xmath202 ( see eq.([uv ] ) and ( zeros ) ) , it follows that in the limit of small @xmath125 ( @xmath242 at s / f interfaces ) and _ _ _ _ @xmath243 _ _ _ _
@xmath198@xmath244 in particular , it follows from ( [ zersim ] ) that for small spin - orbit and spin - flip scatterings @xmath245 , @xmath246 and the well known result @xmath247 for the first critical thickness for 0 to @xmath59 state is reproduced .
this @xmath248 value approximately satisfies the condition of validity of the large @xmath33 approximation considered in this section .
it follows from ( [ zersim ] ) that the critical thickness @xmath248increases with @xmath213 and @xmath214 ( see also numerical results below ) .
an increase of @xmath84 results in the suppression of the magnitude of @xmath249 @xmath250 near the s / f interfaces , which leads to the decrease of @xmath251 formally it is due to the fact that with an increase of @xmath84 , the coefficient @xmath206 in ( [ uv ] ) and , hence , @xmath252 may change its sign resulting in the existence of the solution @xmath253 of ( [ zeros ] ) for @xmath254 this solution corresponds to the thickness range in which simple large @xmath33approximation ( [ ic_ldg ] ) is no valid anymore .
this fact is in the full agreement with our consideration performed for the limit of small @xmath33 .
namely , it follows from ( [ cond_pi_sd ] ) that at large @xmath84 and @xmath2 we have@xmath255 note that in actual experimental situations , when the approximation @xmath199 is not fulfilled , simple expressions ( [ id1 ] ) , ( id2 ) , ( [ ksi11 ] ) , ( [ ksi12 ] ) , ( [ ksi21]),([ksi22 ] ) are not valid and to analyze the data it is convenient to introduce the effective decay length @xmath256@xmath257 where @xmath258 are the thicknesses at which the first @xmath259 and second @xmath259 maxima of @xmath169 occur .
below we will focus on the influence of @xmath260 @xmath261 and@xmath262 on the critical current .
consider first the simplest case of vanishing @xmath264 , @xmath265 . in this limit
, we immediately deduce from ( [ ic])-([q ] ) that @xmath266 + 2g_{s}\gamma _ { b}\xi _ { n}k\cosh \left [ kd_{f}\right ] } \right\ } , \label{ict0}\]]where the coefficients @xmath205 and @xmath206 become @xmath267@xmath268the dependence of the critical current as a function of the f layer thickness calculated from ( [ ict0 ] ) for @xmath269 and @xmath270 for different values of the interface transparency parameter @xmath84 is presented in fig .
5 . note that for small @xmath84 eq.([ict0 ] ) is formally not applicable , since is was derived under the condition ( cond_gamma_b ) of sufficiently large @xmath84 .
therefore , for @xmath2710 , @xmath2721 and @xmath2725 we have added for comparison the corresponding curves calculated numerically by direct solution of the usadel equations for arbitrary @xmath84 , which show that eq.([ict0 ] ) provides reasonable approximation even in the small @xmath84 range .
it is clearly seen from fig .
5 that with an increase of @xmath84 , the position of the first zero @xmath248 is shifted into the region @xmath273 as discussed above , the increase of @xmath84 results in the suppression of @xmath249 in the f layer , such that @xmath274 changes sign in the f layer center at smaller @xmath33 . the results plotted in fig . 5 make it possible to estimate the upper limit for the @xmath275 product of the sfs junction in a @xmath59-state .
this upper limit can be achieved in the case of highly transparent interfaces ( @xmath276=0 ) and in the absence of spin - flip and spin - orbit scattering . as follows from fig .
5 , @xmath277 that provides @xmath278 in the case of nb electrodes ( @xmath279=9k ) . on the thickness dependence of the critical current in s / f / s junction for @xmath280 , @xmath281 and @xmath282 .
open circles : the results of exact numerical calculations for @xmath2830 , @xmath2831 and @xmath2835 ( from top to bottom).,width=268 ] at large f layer thickness , @xmath169 is determined by eqs.([id1 ] ) , ( [ id2 ] ) .
if spin - flip and spin - orbit scattering are negligible , the ratio @xmath284 follows from ( ksi12 ) , ( [ ksi22 ] ) and depends only on @xmath2 and @xmath285 for typical ferromagnets @xmath286 and @xmath65 approximately equals to @xmath287 however , if the spin - flip scattering and spin - orbit rates become relatively large @xmath288 the situation may change drastically .
consider first the influence of spin - flip scattering on i@xmath289 if spin - flip scattering is not negligible , then the ratio of the characteristic lengths in the decaying solution ( [ id1 ] ) , ( [ id2 ] ) becomes @xmath290 and for strong spin - flip scattering @xmath291 the decay length @xmath65 may become much smaller than the oscillation period @xmath66 .
this results in the much stronger decrease of @xmath8 versus @xmath33 in s / f / s junctions . the evolution of @xmath8 for different values of @xmath213calculated from ( [ ic ] ) for @xmath269 @xmath270 and @xmath292 is given in fig .
6 . on the thickness dependence of the critical current in s / f / s junction for @xmath293 @xmath281 , @xmath294 , and @xmath282.,width=268 ] one can see that with increasing @xmath105 , the critical thickness @xmath248of the first @xmath295 crossover shifts to larger values of @xmath33 .
if spin - orbit scattering is not negligible , then the ratio of the characteristic lengths in the decaying solution ( [ id1 ] ) , ( [ id2 ] ) becomes @xmath296 eq.([r_so ] ) shows that the difference between the decaying @xmath65 and the oscillating length @xmath66 increases with @xmath214 even faster than for the case of spin - flip scattering .
moreover , the transition to monotonically decaying solution takes place at @xmath297 and @xmath298 the dependence of critical current on @xmath33 calculated from ( [ ic ] ) for @xmath269 @xmath270 and @xmath292 is shown in fig .
one can see that the oscillation period increases strongly with increasing @xmath136 and diverges when @xmath299 . at the same time , @xmath248shifts to larger values of @xmath33 .
, @xmath281 , @xmath294 , and @xmath282.,width=268 ]
let us now consider a s / f bilayer with a transparent interface .
the complete nonlinear usadel equation in the f layer has to be employed . for the easy axis of the ferromagnet and using the usual parametrization of the normal and anomalous functions @xmath300 and @xmath301 , it may be written in the absence of spin - orbit scattering as@xmath302note that eq.([usadel_nl_f ] )
transforms into ( [ usadel_lin_f ] ) in the limit of small @xmath303 for s / f / s junctions , this equation may be used and an analytical solution found if @xmath199 . in that case , the decay of the cooper pairs wave function in first approximation occurs independently near each interface @xcite .
it can therefore be treated separately enough to consider the behavior of the anomalous green s function near each s / f interface , assuming that the f layer thickness is infinite .
for one interface ( @xmath304 , a first integral of ( [ usadel_nl_f ] ) leads to @xmath305 } , \label{firsti}\ ] ] where the boundary conditions @xmath306 has been used . further integration in ( [ firsti ] )
gives @xcite @xmath307 } -\cos\frac{\theta_{f}}{2}}{\sqrt{\left [ 1-\varepsilon^{2}\sin^{2}\frac{\theta_{f}}{2})\right ] } + \cos\frac{\theta_{f}}{2}}=g_{0}\exp\left\ { -2q\frac{(d_{f}/2+x)}{\xi_{f}}\right\ } \label{g0}\ ] ] where @xmath308 and @xmath309 the integration constant @xmath310 in ( [ g0 ] ) should be determined from the boundary conditions at s / f interfaces .
for simplicity we will assume that the rigid boundary conditions are valid at @xmath311 then @xmath312 and from ( [ g0 ] ) , ( [ teta_in ] ) we finally get @xmath313 ^{2}},\ ] ] _ _ _ _ and @xmath314 ^{2}}=\frac{\left\vert \delta\right\vert ^{2}}{\left ( \omega+\omega\right ) ^{2}},\ ] ] where @xmath315 next , @xmath316 may be determined .
the anomalous green s function at the center of the f layer in the s / f / s junction may be taken as the superposition of the two decaying functions @xcite taking into account the phase difference in each superconducting electrode , @xmath317 as a result , the current - phase relation is sinusoidal and the critical current becomes @xmath318 ^{2}}\exp ( -q\frac{d_{f}}{\xi _ { f}})\right ) .
\label{ic2}\ ] ] the critical current is proportional to the small factor @xmath319 .
the terms neglected in our approach are much smaller and are of the order of @xmath320 .
therefore , they give a tiny second harmonic term in the current - phase relation . it should be underlined that this expression coincides with the one previously obtained in @xcite in the limit @xmath321 ( no magnetic scattering ) and for @xmath322 .
in addition , note that in the limit of vanishing magnetic scattering , the temperature dependence of the critical thickness @xmath323 of f layer when @xmath238 may appear only through the variation of@xmath324 with @xmath18 ( where @xmath325 in this case ) . as the characteristic range of the variation of the matsubara frequencies in the sum in ( [ ic2 ] ) is @xmath326 , then in the limit @xmath327 , the dependence of @xmath323 on @xmath236 will be weak .
however , when the spin scattering is not weak , i.e. @xmath328 another mechanism of the temperature dependence of the @xmath323 emerges due to the temperature dependent term @xmath329 in the usadel equation ( usadel_nl_f ) , or , in another words , due to the complex @xmath18 dependent function @xmath330 for a strong ferromagnet , @xmath331 the latter mechanism may become stronger than the thermal one related to @xmath332 it is not difficult to take into account both mechanisms in the numerical calculation of the sum ( [ ic2 ] ) . as an illustration , we present in fig . 8 the theoretical fit of the experimental data for nbcu@xmath333ni@xmath334nb junctions by sellier _
@xcite , making use of eq.([ic2 ] ) valid in the limit of small interface resistances ( @xmath335 ) .
the f layer thickness used for fit is @xmath336 nm while the experimental value presented in @xcite is @xmath337 nm . the difference may be explained by the uncertainty in f layers determination that is around @xmath338 nm , and may even increase due to the presence of a magnetically dead layer near s / f interface . taking this in mind
, the theoretical description of the critical current temperature dependence can be considered as rather satisfactory . besides , houzet _ et al .
_ @xcite have performed numerical calculations and got also good fit of another experimental curve of sellier2 ( @xmath339 nm ) for the set of parameters which are in the same range that we have used . ni@xmath334nb junction .
the fitting parameters are : @xmath340 and @xmath341.,width=268 ] as another application of the formalism , we present in fig .
9 the theoretical fit of the experimental data of ryazanov _ et al . _
ryazanexp for nbcu@xmath342ni@xmath343nb junctions with @xmath344 nm.good agreement is achieved assuming @xmath345 @xmath346 , @xmath347 and @xmath348 nm . as in the previous fit ,
small difference in @xmath33 may be explained by the uncertainty in f layers determination .
ni@xmath343nb junction .
the fitting parameters are : @xmath349 and @xmath350.,width=268 ] note that the rather complicated expression for the critical current ( ic2 ) simplifies near @xmath9 and may be written as ( for @xmath322 ) @xmath351 where @xmath352 is defined by @xmath353 .
the damping oscillatory behavior of the critical current is therefore retrieved , and the simplicity of the previous expression makes it useful for theoretical description of the evolution of the critical current with the thickness of the f layer .
we have made detailed theoretical investigation of the influence spin - dependent scattering in the ferromagnet on the critical temperature of s / f multilayered systems and the critical current of s / f / s josephson junctions .
more precisely , we have demonstrated that spin - flip and spin - orbit scatterings both lead to the decrease of the decaying length and the increase of the oscillations period . besides
, spin - orbit scattering may be more harmful for superconductivity than spin - flip scattering .
indeed , the oscillations of @xmath9 and @xmath8 can be destroyed by spin - orbit scattering , while spin - flip scattering can only modify them .
this allows to distinguish these pair - breaking mechanisms , which both should be taken into account for theoretical fits of experimental data .
moreover , the simultaneous introduction of the spin - flip and the spin - orbit scattering leads to important new prediction that the perpendicular spin flip can compensate spin - orbit and vice versa .
then the perpendicular magnetic scattering may restore the oscillations of @xmath9 and @xmath8 which would be otherwise absent if spin - orbit scattering is strong enough .
we have also studied the influence of the interface transparency on @xmath9 and @xmath8 in s / f and s / f / s structures , respectively . the nonmonotonous behavior of @xmath9 with the interface transparency parameter @xmath84 was predicted in the case of thin f - layers due to multiple scattering at the interfaces .
we have shown that the same mechanism is responsible for the 0-@xmath3 transition in s / f / s junctions with small thickness @xmath33 in the case of large @xmath84 . for larger @xmath33 detailed analytical and numerical study of the influence of @xmath84 parameter is presented .
it was predicted in ref.@xcite that in the case of noncollinear magnetic ordering the long - ranged triplet component of the superconducting condensate could be generated .
the spin - orbit and the perpendicular magnetic scattering are rather harmful for this long - ranged triplet component .
our analysis shows that the ferromagnetic alloys like cu@xmath4ni@xmath5 , where following our estimates the parameter @xmath354 may exceed unity , are not suitable candidates for the experimental search for the triplet component . finally , note that it may be interesting to study the josephson junctions with the ferromagnetic layer substituted by a paramagnetic one . applying an external magnetic field @xmath355
, it is possible to produce the polarization of the magnetic atoms and then generate an internal field @xmath356 , where @xmath357 is the paramagnetic susceptibility .
the variation the external field allows to change the relative contribution of the scattering mechanisms and the exchange field and then to modify the properties of the junction in a controllable way .
in particular , it could provoke the transition from 0 to @xmath3 state .
the authors thank m. daumens , m. houzet , m. kulic , c. meyers , and v.v .
ryazanov for useful discussions .
we also thank h. sellier and v.v .
ryazanov for communication of their experimental data .
this work was supported in part by egide programme 10197rc , esf pi - shift programme , rfbr grant n 06 - 02 - 90865 and nanoned programme under project tcs.7029 .
the coefficients @xmath133 and @xmath134 that appear in ( solp ) and ( [ solm ] ) may be written as the symmetry relations following from ( [ solp ] ) -([etibe ] ) are at @xmath362 @xmath363 and if @xmath146 : @xmath364 these relations allow us to simplify the calculations for the determination of the critical current .
|
we investigate superconductor / ferromagnet ( s / f ) hybrid structures in the dirty limit , described by the usadel equations .
more precisely , the oscillations of the critical temperature and critical current with the thickness of the ferromagnetic layers are studied .
we show that spin - flip and spin - orbit scattering lead to the decrease of the decay length and the increase of the oscillations period .
the critical current decay is more sensitive to these pair - breaking mechanisms than that of the critical temperature .
these two scattering mechanisms should be taken into account to get a better agreement between experimental results and theoretical description .
we also study the influence of the interface transparency on the properties of s / f structures .
|
the cranked shell model ( csm ) approach @xcite accounts well for the overall systematics of bandcrossing phenomena in rapidly rotating medium and heavy mass nuclei . in this model
the protons and neutrons move independently in a fixed rotating potential . as a consequence of this assumption ,
the alignment of a pair of neutrons with the rotational axis occurs at the same rotational frequency for different proton configurations .
however , recently substantial deviations from this simple picture have been reported using the csm approach in the high - spin study of nuclei near @xmath0 @xcite and for nuclei with aligning neutrons and protons occupying the same intruder subshell @xcite . these deviations , concerning the crossing between the g - band and the s - band in the even - neutron systems referred to as the ( ab)-crossing , include : * the crossing in the odd - proton nuclei
is considerably delayed as compared to the neighboring even - even nuclei , * in the case of @xmath0 even - even nuclei , the ( ab)-crossing is substantially delayed as compared to the neighboring @xmath1 .
[ f : crossex ] summarizes the experimental evidence .
the observed delay of the neutron alignment has been attributed to an increase of the deformation induced by the odd high-@xmath2 proton .
this mechanism has first been pointed out in ref .
@xcite as the possible origin of the delayed @xmath3 neutron backbend if an @xmath4 proton is present .
it has been substantiated by systematic calculations of the equilibrium deformations for bands with an odd - proton in the @xmath5 , @xmath4 and @xmath3 orbitals ( cf . e. g. @xcite ) . though being in the right order of magnitude , the calculations tend to underestimate the experimental shifts .
the p - n residual interaction has been invoked to be responsible for the remaining discrepancy @xcite . in a study of a qq- type p - n interaction acting in the particle - hole channel ref .
@xcite finds only very small shifts . in the pesent paper
we further investigate the role of the residual interaction .
at variance with ref .
@xcite , we use a short range force and calculate the exact solutions for the model system of interacting protons and neutrons in a @xmath2 shell , exposed to a deformed potential . the strong short - range proton - neutron ( p - n ) attraction probes the relative orientation of the high-@xmath2 orbitals , because of their strongly anisotropic , torus like density distributions .
it favors the anti parallel ( @xmath6 ) and the parallel ( @xmath7 ) coupling of a p - n pair , the energies being comparable .
since there are many low-@xmath8 pairs available , they tend to form a correlated state ( p - n pairing ) .
processes in which the high-@xmath2 particles change their relative orientation should be good means to study these aspects of the p - n interaction . in this manuscript
, we investigate a special kind of such reorientation , the alignment of a pair of neutrons with the axis of rotation . by means of a shell model calculation
it will be demonstrated that the rotational frequency at which the neutron alignment occurs changes when additional protons are present and that the frequency shift are sensitive to the p - n interaction and the correlations it generates .
it will also be shown that the p - n interaction causes a delay of the first double - bandcrossing in @xmath0 model systems studied as compared to cases with @xmath9 which may be related to the delayed bandcrossing observed in @xmath0 nuclei @xcite .
in our model the protons and neutrons in a @xmath2-shell interact via a delta force and move in the deformed rotating potential generated by the nucleons outside the @xmath2-shell .
the model hamiltonian is the same as used in refs .
@xcite , where one may look for the details , and is given by @xmath10 - g \sum_{i < j } \delta(|\hat r_i-\hat r_j|).\ ] ] the first term is the deformed quadrupole field where @xmath11 is related to the deformation parameter @xmath12 by @xmath13 in units of the coupling strength @xmath14 of the delta force .
the second , cranking term describes the uniform rotation about the x - axis with the frequency @xmath15 and the third term is the interaction .
the eigenfunctions of the hamiltonian are found by numerical diagonalization .
they can be classified by the isospin and the signature ( symmetry with respect to @xmath16 ) which are conserved by the model hamiltonian .
we would like to point that the present model predictions should only be considered as qualitative indications of the physics expected from more realistic studies .
the major restriction of our model is that we consider only few particles occupying the intruder subshell with rest of the nucleons giving rise to a deformed potential .
the assumption that the nucleons outside the intruder shell do not take part in the correlations generated by the short range residual interaction seems to be the most serious restriction of the model . since we choose a realistic value of the deformation for a given nucleus and estimate the number of particles occupying the high-@xmath2 orbital by looking at the realistic nilsson diagrams , the estimates of the frequency where the rotational alignment occurs should be qualitatively right . as a consequence of the restriction to a single @xmath2-shell , the total angular momentum is underestimated .
this is of little importance , because we study the _ frequency _ of the alignment processes , which will not be changed by the missing angular momentum from the nucleons outside the @xmath2-shell .
[ f : crossdef11 ] shows the expectation value of the @xmath17 component of the angular momentum calculated by assuming a deformation @xmath18 of a well deformed nucleus .
the situation with few protons and neutrons occupying the @xmath5 shell is studied .
the alignment of a pair of neutrons with the axis of rotation ( x ) , caused by the coriolis force , shows up as the step like rise of the angular momentum from small values to about 10 .
it corresponds to the crossing of g- band with the neutron s - band , where the inflection point defines the crossing frequency .
it is seen that the presence of the odd proton delays the alignment of the neutron pair . as demonstrated by fig .
[ f : crossex ] both in the @xmath5 and in the @xmath3 shells a similar delay of the crossing frequency is seen for the combination ( @xmath19 ) . the estimated number of protons and neutrons in the high-@xmath2 shell , denoted by @xmath20 and @xmath21 , must be compared with @xmath22 and @xmath23 of the calculations , respectively .
[ f : crossdef9 ] demonstrates that for near symmetric filling of the @xmath24 shell ( @xmath25 ) and ( @xmath26 ) the crossing frequency in the odd - a nuclei lies below the first crossing in the even - even nuclei ( @xmath27 ) , which corresponds to the simultaneous alignment of protons and neutrons @xcite .
this inversion of the order of the bandcrossings seems to be seen in the br- and kr - isotopes , as illustrated by fig .
[ f : crossex ] .
the figure also shows that the inverse ordering for near symmetric filling as compared to asymmetric filling can be attributed to a delay of the crossing in the even even @xmath0 nuclei as compared to the ones with @xmath9 @xcite , which is stronger than the delay in the odd-@xmath22 system .
the features discussed in the previous section reflect a systematic tendency that is illustrated in fig . [
f : f72jx ] and [ f : f72momi ] , showing @xmath28 and @xmath29 for the @xmath30 shell .
( we have chosen @xmath30 shell since the dimensions of the matrices to be diagonalized are lower and a systematic study is possible . )
the alignments show up as peaks in @xmath31 .
we are interested in the alignment of a pair of @xmath30 neutrons , which is displayed in lower panels of the figs . for the pure neutron @xmath32 and @xmath33 cases .
if there were no p - n correlations , the frequency at which these alignments appear would not change with the number of protons .
however , this is not seen .
instead , the critical frequency for the neutron alignment first grows with the number of added protons .
the shift reaches its maximum when @xmath34 and then decreases when @xmath22 becomes larger than @xmath23 .
of course , the role of @xmath23 and @xmath22 could be exchanged .
qualitatively , one may ascribe the progressive delay of the crossing to p - n correlations that disfavor the simultaneous alignment of protons and neutrons ( e. g. the presence of p - n pairs with a low @xmath8 ) , generated by the attractive p - n interaction .
the more protons are added the stronger the correlations become , until @xmath34 .
as discussed in @xcite , the character of the p - n interaction between the aligned protons and neutrons changes from attractive to repulsive when @xmath22 exceeds @xmath23 , because the character of the aligned configuration changes from particle - like to hole - like .
this shifts the p - n correlations to the highly excited states , whereas the yrast states are not much modified by the p - n interaction .
the shift of the crossing frequency is a consequence of the @xmath35 part of the @xmath36-interaction . the dashed dotted line in fig .
[ f : f72jx ] shows a calculation where the @xmath37 part of the @xmath36-interaction has been removed .
the function @xmath28 almost coincides with the one calculated with the full force .
the question if the p - n correlations that cause the shifts belong to the particle - hole channel or the particle - particle channel ( or to both ) can not be decided on the basis of the shell model calculations . for this , comparisons with mean field calculations are necessary , which will be addressed in a forthcoming paper .
we state here only the existence of strong p - n correlations of the @xmath35 type between nucleons in the same @xmath2-shell , which cause a delay of the first bandcrossing .
a delay of the onset of the bandcrossing for @xmath38 has recently been observed in the kr isotopes @xcite ( cf .
[ f : crossex ] ) .
it is suggested in the paper that the delay may be a consequence of @xmath37 pair correlations .
as already mentioned , the dashed dotted line in fig .
[ f : f72jx ] shows a calculation where we have switched off the @xmath37 part of the interaction .
it is seen that the results are almost the same as for the full force , i.e. the @xmath37 correlations do not influence the alignment significantly in the present model analysis and it is interesting to see whether this conclusion will hold in a realistic study . comparing in figs .
[ f : f72jx ] and [ f : f72momi ] the alignment of the odd - proton in the @xmath39 system with the one in the @xmath40 one notices also the the alignment of an odd - proton is hindered by the presence of neutrons in the same @xmath2-shell . in the previous work
@xcite it was mentioned that strong p - n correlations can alter the bandcrossing substantially and the double - bandcrossing can occur below the individual proton and neutron crossings .
however , most of the analysis was performed in the spherical approximation and the dominant component of the p - n interaction which changes the bandcrossing was not fully explored .
it is shown in the present work using the proper deformed basis that it is @xmath35 component of the n - p interaction which is responsible for the particle number dependence of the bandcrossing .
this is in accordance with a recent mean field study @xcite of the @xmath2-shell model system , which attributes the delay of the crossing frequency in the even - even @xmath0 system to the @xmath35 monopole pair field .
it is clear from the present simple model study that the alignment of high-@xmath2 nucleons with the rotational axis is sensitive to the p - n interaction .
the alignment of one kind of particles is delayed when the other kind of particles is present in the same @xmath2-shell .
the effect reaches its maximum for @xmath0 and is a consequence of @xmath35 correlations between protons and neutrons . for @xmath34 the crossing frequency in the even system is higher than for asymmetric filling .
the experimentally observed delay of the first backbend in @xmath34 even - even @xmath41kr can be attributed to these correlations .
the experimentally observed delay of the first backbend in even - neutron nuclei by the presence of an odd - proton in the same @xmath2-shell could be a combination of these correlations and the change of deformation induced by the odd - nucleon . by measuring the deformation and/or calculating it one may try to disentangle the two effects and probe the p - n correlations .
finally , we would like to mention that while the results of the present model analysis seem to point into the right direction they should not be compared with the experimental data in a quantitative form .
it would be quite interesting to study the effects in more realistic shell model configuration spaces .
j. heese , n. martin , c.j .
gross , w. fieber , k.p .
lieb , a. kuhnert , k.h .
maier and x. sun , phys . rev .
* c41 * , 1553 ( 1990 ) n. martin , c.j .
gross , j. heese and k.p .
lieb , j. phys .
* g15 * , l123 ( 1989 ) g. n. sylvan , j.e .
purcell , j. doring , j.w .
holcomb , g.d .
johns , t.d .
johnson , m. a. riley , p.c .
womble , v. a. wood and s.l .
tabor , phys .
rev . * c48 * , 2252 ( 1993 ) g. de angelis et al . ,
* b415 * , 217 ( 1997 ) j. heese , d.j .
blumenthal , a.a .
chisti , p. chowdhury , b. crowell , p.j .
ennis , c.j .
lister and ch .
winter , phys . rev .
* c43 * , r921 ( 1991 ) d. f. winchell , m.s .
kaplan , j.x .
saladin , h. takai , j.j .
kolata and j. dudek , phys . rev . * c40 * , 2673 ( 1989 ) c. gross et al .
. phys . * a501 * , 367 ( 1989 ) g. d. johns , j. doring , j.w .
holcomb , t.d .
johnson , m. a. riley , g.n .
sylvan , p.c .
womble , v.a .
wood and s.l .
tabor , phys .
* c50 * , 2786 ( 1994 ) v.p .
janzen et al . , phys .
lett . * 70 * , 1065 ( 1993 ) .
e. s. paul , d.b .
fossan , k. hauschild , i.m .
hibbert , h. schnare , j.m .
sears , i. thorsland , r. wadsworth , a.n .
wilson and j.n .
wilson , phys . rev . *
c51 * , r2857 ( 1995 ) f. liden et al .
, nucl . phys . *
a550 * , 365 ( 1992 ) r. a. bark , r. bengtsson and h. carlsson , phys . lett . *
b339 * , 11 ( 1994 ) and work quoted therein r. wyss , a. johnson , proc .
int . conf . on high spin physics , pittsburgh 1990 , world scientific , singapore , 1991 w. satula , r. wyss , and f.dnau , nucl
a565 * , 573 ( 1993 ) j.a .
sheikh , n. rowley , m.a .
nagarajan and h.g .
price , phys .
* 64 * , 376(1990 ) s. frauendorf , j.a .
sheikh and n. rowley , phys . rev .
* c50 * , 196 ( 1994 ) k. kaneko , jing - je zhang , phys .
c 57 , 1732 ( 1998 )
|
the effects of the residual proton - neutron interactions on bandcrossing features are studied by means of shell model calculations for nucleons in a high - j intruder orbital .
the presence of an odd - nucleon shifts the frequency of the alignment of two nucleons of the other kind along the axis of rotation .
it is shown that the anomalous delayed crossing observed in nuclei with aligning neutrons and protons occupying the same intruder subshell can be partly attributed to these residual interactions .
|
in this paper i show how to formulate an asymptotically free theory in such a way that it includes perturbative and non - perturbative phenomena simultaneously .
the idea is the following .
contrary to an infrared free theory , in an asymptotically free theory the divergences prevent us from writing even perturbation expansions in a unique , well - defined way .
we can , however , make use of the fact that divergences in the theory enter only the green s functions and the vertices . on the other hand , knowing the green s functions and the vertices we can express all the other amplitudes through them perturbatively in a unique way .
thus , we have to formulate equations for green s functions and vertices in a form which does not contain any divergences .
if this is done , the solutions of these equations will contain both perturbative and non - perturbative phenomena . to avoid technical complications , in the first section of the paper we will derive these equations in an abelian theory
in which usual perturbation theory is also well defined . in the second section
we generalize the equations for a non - abelian theory and discuss the connection between perturbative and non - perturbative effects .
the equations have an integro - differential structure in which the asymptotic behaviour of the green s functions is defined by the boundary conditions .
the main conclusion in this section is that in the region of large momenta the green s functions of quarks and gluons contain additional parameters compared to normal perturbation theory which can be associated with different types of `` condensates '' .
these non - perturbative parameters can be defined by solving the equations and finding non - singular solutions in the infrared region . a priori it is not clear what type of additional conditions have to be imposed on this system of equations in order to fix a non - singular solution .
it can be , for example , the conservation of the axial current or of other currents which is formally satisfied from the point of view of the lagrangian but is not ensured because of the divergences . in the third section
we will show that these equations make it possible , for the first time , to analyse the problem of spontaneous symmetry breaking in an asymptotically free theory and to find approximately the value of the critical coupling at which symmetry breaking occurs .
also , they allow us to answer the fundamental question for asymptotically free theories , namely : how the bound states the hadrons have to be treated in these theories and how these bound states influence the equations for the green s functions .
the analysis leads to the following conclusion .
the conditions for axial current conservation of flavour non - singlet currents ( in the limit of zero bare quark masses ) require that eight goldstone bosons ( the pseudo - scalar octet ) have to be regarded as elementary objects with couplings defined by ward identities .
this is so in spite of the fact that the couplings of these states to fermions decrease at large fermion virtualities .
the same analysis provides a new possibility for the solution of the @xmath0 problem . in this solution
the flavour singlet pseudoscalar @xmath1 is a normal bound state of @xmath2 without a point - like structure .
the mass of this bound state is different from zero and can be calculated in the limit of massless quarks . for massive quarks
the pseudoscalar octet becomes massive .
their masses , however , are not calculable in terms of bare quark masses because of logarithmic divergences and have to be regarded as unknown parameters which in the real case of confined quarks are defined by the self - consistence of the solution of the infrared problem .
these states have an essential influence on the equations for the green s functions , which , as it will be shown in the next paper , can be used constructively in solving the confinement problem if the effective coupling in the infrared region is not too large . in this case
the integro - differential equations can be reduced to a system of non - linear differential equations and the theory looks like a theory of particle propagation in self - consistent fields defined by the green s functions themselves ( as it is the case in landau s fermi liquid theory ) .
the self - consistent fields are fields of gluons and @xmath3-mesons .
in qed we have two green s functions : that of the photon @xmath4 and of the electron @xmath5 , and a vertex function @xmath6
. the photon green s function is defined by the vacuum polarization operator @xmath7 which can be expressed symbolically by a sum of feynman diagrams @xmath8 in order to obtain series not containing divergences we will try to consider the derivatives of @xmath7 as functions of momenta . due to current conservation , we have @xmath9 where @xmath10 contains only logarithmic divergences . differentiating ( [ 1.2 ] ) twice we will have @xmath11 we here use @xmath12 the second term in ( [ 1.3 ] ) does not contain divergences , the first one does . in order to obtain a finite expression , consider @xmath13 because of this , the quantity @xmath14 does not contain divergences . in feynman gauge
@xmath15 is equal @xmath16 any diagram contains only the product @xmath17 and therefore @xmath18 in first order this means @xmath19 the integrals on the right - hand - side are convergent .
this fact has to hold in any order .
hence , we can include in ( [ 1.8 ] ) the exact electron green s functions and the exact vertex functions and add all the corresponding diagrams which were not included . as a result
, we can write @xmath20 where @xmath21 denotes the sum over the permutations of the indices similarly to ( [ 1.8 ] ) ; a quantity @xmath22 corresponds to each photon line .
( [ 1.9 ] ) is an equation for @xmath23 ( in the form of series in @xmath23 ) provided that @xmath24 and @xmath6 are known . in order to obtain equations for electron green s functions and vertices
we have to remember that these functions can change rapidly even if @xmath23 is small because they can have infrared singularities of the type @xmath25 or ultraviolet singularities of the type @xmath26 .
it is proven to be possible to arrange the differentiation in such a way that there will be an expansion only in @xmath23 . to write an equation for the fermion green s function not containing any divergences
, we have to differentiate twice the self - energy of the fermion or its inverse green s function .
let us consider @xmath27 ; the diagrammatic expression for @xmath28 will be the following .
the simplest diagram is @xmath29 diagrams of the next order are @xmath30 it can be easily shown that @xmath31 using this equality , we have @xmath32 @xmath33 @xmath34 @xmath35 restricting ourselves to @xmath36 , we can write @xmath37 @xmath38 and , consequently , @xmath39 the term @xmath40 has a contribution to @xmath41 not containing overlapping divergences and is of the form @xmath42 & & + \,\ , { \setbox0=\hbox{\pandocstartinclude{fig22.pstex_t.tex}\pandocendinclude{input}{21}{28 } } \picraise=-0.5\ht0 \advance\picraise by 0.5\dp0 \advance\picraise by 3pt \hbox{\raise\picraise \box0 } } \,\ , + \,\ , { \setbox0=\hbox{\pandocstartinclude{fig23.pstex_t.tex}\pandocendinclude{input}{21}{28 } } \picraise=-0.5\ht0 \advance\picraise by 0.5\dp0 \advance\picraise by 3pt \hbox{\raise\picraise \box0 } } \label{1.12}\end{aligned}\ ] ] for the calculation of higher order diagrams it is convenient to adopt the following principle .
beginning from the first point of interaction we shall relate the external momentum to the photon line : @xmath43 the next photon interaction can be expressed as @xmath44 now we relate the external momentum to the positron line .
if the positron is emitting a photon , we keep sending the external momentum along the positron line up to its annihilation . as a result
, we get a tree structure @xmath45 in which the initial photon line can end only at the final electron .
it can be shown that every line in the diagram will be passed only once if the photons included in the fermion self energy are not taken into account .
this means that in this approach exact electron green s functions have to be used with bare vertices since the idea is basically the exclusion of overlapping divergences . taking the second derivative of one of the photon propagators and restricting ourselves to the contribution
@xmath46 we obtain both on the left - hand side and the right - hand side photon emission amplitudes with zero momentum . due to gauge invariance ,
however , a zero momentum photon can not change the state of the system ( it is emitted from the external lines ) .
the photon emission amplitude equals @xmath47 [ within this convention the bare vertex is @xmath48 .
the final contribution of the differentiation has the structure ( [ 1.11 ] ) .
the above statement , which essentially means that the emission of a zero momentum photon is determined by the total charge , can be proven by the ward identity .
thus , we have @xmath49 the remaining diagrams contain first derivatives of photon and positron lines and second derivatives of positron lines .
all these diagrams can be expressed in terms of the exact @xmath50 , @xmath51 and @xmath52 functions .
they do not contain divergences except those graphs which correspond to the photon self energy . the first term in ( [ 1.16 ] )
has the structure @xmath53 where @xmath54 is a quantity close to the compton scattering amplitude in the sense that they would be equal if we differentiated all fermion propagators including those inside the electron green s function .
( the factor @xmath55 is due to the fact that the compton amplitude contains the sum of diagrams with momenta @xmath56 and @xmath57 ) .
the compton scattering amplitude @xmath58 satisfies the ward identity @xmath59 which has the simple diagrammatical meaning @xmath60 the second term on the right - hand side of ( [ 1.17 ] ) represents the contribution of photons which , if we carried out the mentioned differentiation , would correspond to photons inside the green s function and which are not present in @xmath61 .
hence , @xmath62 all the diagrams for @xmath63 are of the form @xmath64 and contain the photon lines with the exact green s functions ( [ 1.6 ] ) which equal @xmath65 . by differentiating the photon lines we have calculated the contribution of @xmath36 .
so , there remains @xmath66 where we have introduced @xmath67 in order to avoid a contribution from @xmath68 .
this expression contains logarithmic corrections coming from the ultraviolet region ( @xmath69 ) . replacing @xmath70 by @xmath71 and performing an integration by parts , we get @xmath72 the first term in ( [ 1.20 ] )
can be explicitly calculated while the second one does not contain any logarithms because of the presence of the difference @xmath73 .
we can rewrite ( [ 1.20 ] ) in the form @xmath74 consequently , @xmath63 can be expressed as @xmath75 where @xmath76 is defined by [ the second line of ] ( [ 1.20 ] ) and @xmath40 contains first order derivatives of the photon and positron lines and second order derivatives of the positron lines , as it has been explained above .
the first term contains all singularities of the types @xmath77 and @xmath78 .
to obtain the equation for the vertex function , let us express @xmath79 as a set of diagrams containing exact green s functions : @xmath80 we shall relate the external momenta to the lines in the diagrams in the same way as we did when we derived the equation for the green s function . calculating the second order derivative in @xmath81
we obtain @xmath82 where @xmath83 is the contribution to the amplitude of the process @xmath84 which corresponds to the emission of photons with momenta @xmath85 from the external legs : @xmath86 inserting ( [ 2.3 ] ) into ( [ 2.2 ] ) and replacing @xmath67 by @xmath87 we get @xmath88 here we have introduced the notations @xmath89 the correction terms @xmath90 are defined as a set of diagrams with exact green s functions and vertices .
they contain first order derivatives of both the photon lines and the positron lines , second order derivatives of the positron lines and , in the same way as in the case of the green s function , corrections due to the replacement of @xmath67 by @xmath87 .
the same equation is valid for the interaction amplitude of fermions with the external field provided this interaction does not depend on the relative momentum @xmath81 of the fermions .
the equations for the interaction with external fields are essential . indeed ,
if these equations have solutions that decrease at large virtualities of the fermions i.e. solutions which do not require driving terms this means the existence of bound states . for the sake of convenience
we rewrite the equation ( [ 2.4 ] ) in the form @xmath91 which will be understood as the equation for the bound state with a spin which is defined by the invariant structure of the matrix @xmath92 .
it is important to note that the accuracy of the equations for the vertex @xmath79 and for the bound states differs from the accuracy of the equation for the green s function . the functions @xmath79 and @xmath93 depend , among others , on the ratios @xmath94 , @xmath95 .
if these parameters become large , then @xmath50 and @xmath93 contain , in general , the so - called sudakov logarithms which are not included in the equations ( [ 2.4 ] ) and ( [ 2.6 ] ) .
hence , the equations are valid only if @xmath96
qcd is the theory of interacting quarks and gluons .
the description of quarks is more or less the same as in qed .
gluons , however , are very different .
even the fact that being a spin @xmath97 particle the gluon has to have a multi - component wave function is not seen explicitly . in the usual approach
it is described by a green s function @xmath98 which in momentum space ( in feynman gauge ) always can be written in the form @xmath99 it contains only one unknown function .
all spin properties of the gluon are included in the momentum dependence of its interaction . in order to introduce multi - component green s functions
, we have to formulate the theory in a form in which the interaction is momentum independent : we have to replace the usual description of the gluons and their interactions by a duffin - kemmer type formulation . in the framework of this description
the gluon lagrangian is @xmath100\}f_{\mu\nu } + \frac{1}{2}f_{\mu\nu}f_{\mu\nu},\ ] ] [ where the trace over colour indices is implied ] .
the potential @xmath101 and the field strength @xmath102 are independent quantities .
the interaction is momentum independent and equals @xmath103f_{\mu\nu}$ ] . in this formulation
we have three independent green s functions @xmath104 in order to fix the gauge in a covariant way , it is , of course , necessary to introduce ghosts by adding a gauge - fixing term @xmath105 the three green s functions can be combined into one by introducing the ten - component state [ implying @xmath106 @xmath107 we use the parameter @xmath108 , which has the dimension of mass , to convert the lower component of @xmath109 into the same dimension as the upper component .
the equations for the fields corresponding to the lagrangian ( [ 3.2 ] ) , ( [ 3.4 ] ) are @xmath110 @xmath111 - f_{\mu\nu } = 0 .\ ] ] in terms of the state @xmath109 we have @xmath112 @xmath113 here @xmath114 are duffin - kemmer matrices satisfying the commutation relation @xmath115 they are connecting the [ four ] upper @xmath116 and [ six ] lower @xmath117 components of the field @xmath109 in ( [ 3.5 ] ) , @xmath118 [ their non - zero components , the @xmath119 matrices @xmath120 have a simple representation @xmath121 the quantities @xmath122 are projection operators onto the upper and lower components of @xmath109 .
if we want to preserve the dimension of the @xmath123 component of the free green s function , @xmath124 has to satisfy the equation @xmath125
d_0 = \frac{1}{m^2}.\ ] ] the solution of this equation is @xmath126 .\ ] ] in feynman gauge @xmath127 and @xmath128 , in landau gauge @xmath129 .
the three terms in ( [ 3.10 ] ) correspond to the three independent green s functions ( [ 3.3 ] ) .
the vertex for the interaction of three gluons with momenta @xmath130 , colours @xmath131 and duffin - kemmer indices @xmath132 has the structure @xmath133 @xmath134 @xmath135 the coupling constant remains in our notation @xmath136 .
let us consider in this approach the properties of the exact green s function @xmath137 @xmath138 where @xmath139 is defined diagrammatically : @xmath140 it contains four matrix elements @xmath141 , @xmath142 , @xmath143 , @xmath144 .
it can be also represented in the form @xmath145 the factor @xmath146 in the third term is necessary to preserve the current conservation ; in the first order the ghost contributes only to @xmath147 .
we added the term @xmath148 in ( [ 3.12 ] ) to be able to fix the gauge for the exact green s function . instead of ( [ 3.12 ] )
we can write @xmath149 the green s function then will be equal @xmath150 where @xmath151 .\ ] ] in feynman gauge @xmath152 in order to understand the meaning of the denominator in ( [ 3.16 ] ) , let us consider the renormalization properties of simple diagrams , for example @xmath153 .
@xmath154 where @xmath155 is the effective bare coupling ; [ we expect , see ( [ 3.27 ] ) below ] @xmath156 if this is true , we have in feynman gauge @xmath157 however , @xmath158 [ in the manuscript , @xmath144 ] has to be proportional to @xmath160 .
this means that we have to expect @xmath161 i.e. @xmath162 this is our definition of @xmath163 .
it has all known properties of the renormalized coupling @xmath164 . as a result
, @xmath52 can be written in the form @xmath165 in this approach @xmath166 is not a quantity coming from the normalization : it is the bare coupling .
the theory has to be defined as the limit @xmath167 . in this context
the expression ( [ 3.23 ] ) has a very interesting property . in the limit @xmath168 the green s function @xmath52 contains only @xmath169 and @xmath170 . according to ( [ 3.22 ] ) , in this limit @xmath171 .
because of this , when we regard the equations for @xmath169 and @xmath170 in the way we did for the fermionic green s function in qed , @xmath166 has to disappear .
consequently , we have equations for @xmath169 and @xmath170 with @xmath172 being arbitrary .
the equation for @xmath173 will not help since , due to the equality @xmath171 it has to be an identity .
we will see that an equation for @xmath164 appears when we will consider the correction of the order of @xmath174 . before formulating the equation for the green s function ,
let us see what the ward identity looks like in this formulation .
the bare vertex is @xmath175 [ with @xmath176 the colour matrix ( the structure constant ) ] .
consider the relation between @xmath177 and the green s function : @xmath178 \nonumber \\ & = & m[d_0^{-1}(k_2 ) - d_0^{-1}(k_1 ) ] + m\,\gamma_+\zeta_0[(\hat{k}_2 ^ 2-k_2 ^ 2 ) - ( \hat{k}_1 ^ 2-k_1 ^ 2 ) ] .\end{aligned}\ ] ] hence , at @xmath179 [ hereafter , @xmath180 @xmath181,\ ] ] which is , of course , the usual complication due to the slavnov - taylor ward identity .
but in this formalism the additional term is equal @xmath182 inserting this vertex into an arbitrary diagram we obtain [ for the full vertex ] the relation @xmath183 .\ ] ] according to ( [ 3.15 ] ) , ( [ 3.18 ] ) , in feynman gauge we can write [ @xmath184 @xmath185.\ ] ] the quantities @xmath186 , @xmath187 must be calculated from the equations for @xmath169 and @xmath170 .
but if the equations are formulated in terms of the exact green s functions , the dependence on @xmath188 enters only through these green s functions @xmath52 which , according to ( [ 3.16 ] ) , ( [ 3.17 ] ) , include @xmath188 only through the quantity @xmath189 defined in ( [ 3.17 ] ) . in the limit @xmath190 , however , @xmath189 does not depend on @xmath188 .
consequently , @xmath186 and @xmath187 are equal to zero and we have the following simple relation for the vertex at zero momentum : @xmath191 here @xmath192 is defined by ( [ 3.27 ] ) ; it does not contain gauge - fixing terms : @xmath193 let us first consider the equations for @xmath169 and @xmath170 in the same way as we did for fermions .
as in the case of qed , we will use the feynman gauge in order to simplify the one - gluon contribution to the equation for the green s function . @xmath194
each line here contains the exact propagator ( [ 3.23 ] ) . differentiating the propagator along the upper line
, we obtain the contribution corresponding to the second derivative of one line @xmath195 where @xmath196 and we introduced the notation @xmath197 the contribution of this derivative to the equation will have the form @xmath198 here @xmath199 is the gluon - gluon scattering amplitude at zero momentum of one of the gluons . the second term @xmath200 is defined diagrammatically : @xmath201 taking the first derivative of two different lines , we get the contribution @xmath202 as a result we have @xmath203 we have to remember that on the right - hand side we have to take only the matrix elements @xmath204 , @xmath205 and @xmath206 . writing @xmath207 and using the ward identity ( [ 3.28 ] ) , we obtain an equation of the same structure as for fermions : @xmath208 @xmath209 if we now make the same trick as before , namely replace @xmath210 by @xmath211 and redefine @xmath212 , we will have in the limit @xmath167 @xmath213 \!\!\ ! & = & \!\!\ !
3\frac{\alpha(p)}{\pi\,k^2 } \,[\gamma_- , \partial_{\mu}\tilde{d}^{-1}(z_1\hat{k } + z_1 ^ 2 z_2^{-1}\gamma_+ + z_2\hat{k}^2\gamma_-)\partial_{\mu } \tilde{d}^{-1 } ] + [ \gamma_-,l ' ] , \\
\label{3.40 } \partial^2 z_2^{-1}\gamma_- \!\!\!&=&\!\!\ !
3\frac{\alpha(p)}{\pi\,k^2}\,\gamma_- \partial_{\mu}\tilde{d}^{-1 } \left ( z_1\hat{k } + z_1 ^
2 z_2^{-1}\gamma_+ + z_2\hat{k}^2\gamma_- \right)\partial_{\mu}\tilde{d}^{-1}\gamma_- + \gamma_-l'\gamma_- .\end{aligned}\ ] ] [ [ equation - for - quark - greens - function . ] ] [ equation for quark green s function . ]
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the equation for the fermionic green s function in qcd will differ from ( [ 1.21 ] ) only by the factor @xmath214 ( @xmath215 are colour matrices ) .
the reason is the following .
the derivation of the equation ( [ 1.21 ] ) was based on the relation between the fermionic green s function and the amplitude of zero momentum photon emission by a fermion @xmath216 in the usual formulation this relation is not correct in qcd .
the simple relation ( [ 3.28 ] ) for the amplitude of the zero momentum gluon emission by a gluon implies , however , that the amplitude of a zero momentum gluon emission by a quark has to be equal @xmath217 together with ( [ 3.23 ] ) it leads to the equation ( [ 1.21 ] ) .
[ [ equation - for - gamma . ] ] [ equation for @xmath218 . ] + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the equation for the colourless vertices remains also the same .
the equation for a three - gluon vertex , however , will be essentially different . in the same way as for the fermionic case , but taking into account the non - commutativity of the gluon coupling , we can show that @xmath219 satisfies the following equation : @xmath220 @xmath221 the right - hand side corresponds to all possible gluon emissions from the external lines : @xmath222 @xmath223 higher order terms can be written in the same diagrammatic way .
the fourth term in ( [ 3.43 ] ) and in the corresponding diagrammatic expression equals zero at @xmath224 since @xmath225 at @xmath224 .
the equation ( [ 3.43 ] ) is , indeed , quite different from the equation for a colourless vertex .
the main difference comes from the fact that on the right - hand side it contains derivatives of @xmath218 not only over @xmath81 but also over @xmath226 .
hence , in order to find @xmath218 , we have to write three different equations for second derivatives over three different momenta .
a similar equation is valid for the quark - gluon vertex . knowing the green s function and the vertices , all the other amplitudes for the interactions and quarks and gluons can be written in the usual perturbative way .
these amplitudes have no divergences and contain inside the gluon green s function the unknown function @xmath163 .
to formulate the theory in an unambiguous way , without any references to the cutoff and the regularization , we have to find the equation for @xmath163 as we did in qed , and learn , how to write the higher terms more elegantly and constructively .
i postpone the investigation of this problem to another paper .
what i want to discuss now is the main difference between an asymptotically free theory and an infrared free theory .
the equations for green s functions of quarks and gluons are proven to be second order integro - differential equations . to solve them ,
we need boundary conditions . in an infrared theory
the boundary conditions are known : they are the conditions for the existence of free elementary particles at small momenta . in an asymptotically free theory
the interaction is small at large momenta , and we expect to have here a perturbative solution .
however , in the region of large momenta all the equations have two types of solutions : the perturbative solution and solutions which decrease as some power of the momenta and therefore contain dimensional parameters reflecting the density of different condensates .
these parameters have to be defined either by the introduction of additional conditions of the type of conservation laws or by the self - consistency of the solutions in the small momentum region ( or both ) .
the next section of this paper will be devoted , essentially , to the discussion of this problem .
in the present section we will consider the equation for the fermion green s function in qcd @xmath227 where @xmath228 this equation was discussed extensively [ 1 ] in connection with the problem of quark confinement .
here we shall use the equation for the discussion of the spontaneous breaking of chiral symmetry in asymptotically free theories . the asymptotic freedom is reflected in this equation by the fact that @xmath229 decreases when @xmath230 .
we will assume that in the limit @xmath231 @xmath229 approaches a finite value .
@xmath232 at @xmath230 the solution of the equation ( [ 4.1 ] ) has the form @xmath233 , \ ] ] @xmath234 if @xmath235 , the quantities @xmath236 , @xmath108 , @xmath237 , @xmath238 are arbitrary parameters . in the case when @xmath229 is defined by perturbation theory ,
@xmath239 are the following functions : @xmath240 the anomalous dimensions @xmath241 , @xmath242 and @xmath243 can easily be found from the equation ( [ 4.1 ] ) . generally speaking ,
the solution depends on four parameters . in the limit @xmath230
the chiral invariant solution can be written as @xmath244 the general solution ( [ 4.2 ] ) corresponds to massive quarks . in the solution which corresponds to spontaneously broken chiral symmetry , @xmath245 . in this case
the mass term decreases when @xmath246 ; the term @xmath237 is responsible for the violation of the symmetry . multiplying ( [ 4.1 ] ) by @xmath247
we obtain the equation @xmath248 where @xmath249 clearly , it has a structure which corresponds to the equation for particle propagation in the self - consistent field @xmath250 .
it is easily seen that ( [ 4.1 ] ) is equivalent to the equation for @xmath101 of the form @xmath251 the matrix @xmath28 is defined by two invariant functions and can be written as can be treated as a dimensionless quantity since the equation ( [ 4.5 ] ) is homogeneous ] ] @xmath252\equiv \rho \ , e^{-\hat{n}\frac{\phi}{2}}\ ] ] where @xmath253 here @xmath101 is @xmath254 inserting ( [ 4.8 ] ) and ( [ 4.9 ] ) into ( [ 4.5 ] ) , we obtain a set of non - linear equations for @xmath255 and @xmath92 .
we can linearize the equation for @xmath255 by writing @xmath256 for a constant @xmath257 , or @xmath258 for @xmath257 which is a function of @xmath81 . here
@xmath259 as a result , we get for @xmath260 and @xmath92 the following set of equations : @xmath261u = \frac{\dot{\beta}}{\beta}(\dot{u}-u)\ ] ] @xmath262 for a constant @xmath257 the conservation law @xmath263 is fulfilled ; @xmath264 u^2 .\ ] ] the term @xmath265 is of the order of @xmath266 and it can almost always be neglected .
the asymptotic behaviour ( [ 4.2 ] ) of the green s function in the limit @xmath267 ( @xmath268 ) corresponds to @xmath269 the chiral - invariant solution corresponds to @xmath270 .
close to the value @xmath271 ( i.e. @xmath272 ) , at large @xmath273 we have @xmath274 hence , ( [ 4.13 ] ) is an oscillator equation with damping if @xmath273 is growing and with acceleration if @xmath273 is decreasing : @xmath275 this means that the chiral invariant solution @xmath271 is unstable in an asymptotically free theory .
let us consider the equation for @xmath92 at negative @xmath273 values in detail . due to ( [ 4.8 ] ) , @xmath276 is in this case purely imaginary and ( [ 4.13 ] ) describes the motion of the particle as a function of the `` time '' @xmath277 in a periodic field ; the damping ( or the acceleration ) is defined by ( [ 4.15 ] ) .
@xmath278 at @xmath279 ( @xmath280 ) the particle is situated at one of the minima of the potential ; it accelerates as @xmath277 decreases .
the acceleration rate is defined by the parameters of the solution ( [ 4.2 ] ) in the limit @xmath281 .
when @xmath81 is decreasing we have , generally speaking , two possible behaviours for the solution .
if @xmath282 remains positive all the time , then the solution goes to infinity and @xmath51 develops a singularity at some euclidean @xmath81 . if instead the potential @xmath282 changes sign , then the solution may approach again a minimum . in the latter case it is easy to see that @xmath28 has a singularity at @xmath283 .
there is only one possibility to avoid having a singularity in the euclidean region including @xmath283 .
we have to choose the parameters which determine the acceleration at @xmath284 so that the particle approaches the maximum of the potential when @xmath285 ( @xmath286 ) . in order to find these exceptional solutions without singularities at @xmath287
it is natural to start solving the equation at @xmath283 by fixing the behaviour of the solution at @xmath286 .
the solution of ( [ 4.12 ] ) , ( [ 4.13 ] ) corresponding to a maximum ( e.g. @xmath288 ) at @xmath231 is @xmath289 in the case of such a solution the constant @xmath290 in ( [ 4.14 ] ) equals zero , and @xmath291 } \>.\ ] ] inserting ( [ 4.18 ] ) into ( [ 4.13 ] ) we obtain one non - linear equation .
it can be analysed easily for arbitrary @xmath273 values .
the solution we are interested in contains essentially one parameter @xmath292 which can be related to the renormalized fermion mass .
the parameter @xmath293 is irrelevant ; it defines the normalization of the green s function at @xmath283 and can be chosen as @xmath97 .
the solution of the equation leads to the unambiguous determination of the asymptotic parameters of the green s function ( [ 4.2 ] ) by @xmath292 and by the parameter @xmath294 of the strong interaction is the momentum scale where the coupling in figure 1 becomes of order unity or , more precisely , when it reaches the critical value ( see below ) ] ] which enter @xmath295 . as it was mentioned before , to the spontaneous breaking of chiral symmetry corresponds an asymptotic behaviour of @xmath28 in which @xmath296 .
the existence of such a solution requires a connection between @xmath292 and @xmath294 .
the renormalized fermion mass as well as the other `` condensate '' parameters are then determined by the strong interaction parameter @xmath294 .
let us consider this solution in detail .
starting from the point @xmath288 at @xmath297 , the solution will , obviously , reach the minimum of the potential either monotonically ( if the damping is strong enough ) or in an oscillating way . in our case the damping depends on the value of @xmath257 . if @xmath136 is small , @xmath257 is close to unity , the damping is strong and the solution has a monotonic behaviour . by decreasing @xmath257
the solution may become an oscillating one . for obtaining the value of @xmath257
at which the oscillation begins there is no need to solve the equation for all values of @xmath81
. it will be sufficient to investigate the solution in the region where @xmath298 becomes close to @xmath3 ; here @xmath299 and @xmath300 satisfies the equation ( [ 4.16 ] ) .
the solution can be written in the form @xmath301 it oscillates if @xmath302 oscillations in @xmath298 mean that the mass term in ( [ 4.8 ] ) , @xmath303 starts to oscillate . by solving the equation for bound states we can check that at @xmath304 bound states appear with wave functions behaving like ( [ 4.19 ] ) which coincides with the behaviour of the solution of the dirac equation in the field of a point - like static charge @xmath305 when @xmath306 .
the simplest example for such bound states are goldstone states the wave functions of which , as we shall see in the next section , are proportional to @xmath307 .
we have found the oscillations and the critical value @xmath308 using the assumption that @xmath309 . in reality @xmath136
depends on @xmath81 ( see fig . 1 ) and the situation is somewhat more complicated .
it reminds the case of the equation for a critical charge of finite radius . in the region of small @xmath81 values where @xmath87 is close to a constant @xmath310 we can consider @xmath298 as an independent variable and @xmath273 as a function of @xmath298 .
it can be shown that for @xmath67 satisfying the condition ( [ 4.20 ] ) there are two regions @xmath311 and @xmath312 in the @xmath273 , @xmath298 plane ( fig .
3 ) @xmath313 where the solution @xmath314 is monotonic , and there is a region @xmath315 where the solution oscillates .
the value of @xmath316 is determined by the equality @xmath317 it exists if @xmath318 . considering @xmath257 as a function of @xmath273 in ( [ 4.22 ] ) and taking into account that @xmath319 at @xmath281
, we obtain a region bound by the dashed curve and the vertical axis where the solution oscillates .
there are no oscillations if @xmath320 ( @xmath321 ) .
due to ( [ 4.2 ] ) and ( [ 4.3 ] ) , we can write in the region @xmath322 @xmath323 if @xmath324 , the solution which equals @xmath325 at @xmath231 , transforms monotonically into ( [ 4.23 ] ) with @xmath326 .
if @xmath318 , @xmath327 and @xmath328 start to oscillate as functions of @xmath292 , and @xmath329 can turn into zero at a certain @xmath292 value .
this means that we have a solution corresponding to broken chiral symmetry .
if @xmath245 , there exist also a large number of solutions depending on the parameters @xmath237 , @xmath238 . with the same sign of @xmath237 we can have different solutions corresponding to the curves i and ii which in the limit @xmath231 reach @xmath288 and @xmath330 , respectively
the solutions @xmath331 correspond to smaller values @xmath237
, @xmath292 .
let us consider the solutions of types i and ii in detail for complex @xmath81 this is justified since @xmath28 has to satisfy the requirements of analyticity and unitarity .
we will show that both solutions have singularities at real positive @xmath273 values .
the solutions are chosen in such a way that they are regular as @xmath285 and have no singularities at @xmath332 . due to the analyticity of the equations ,
the behaviour of the solution for @xmath333 can be found by solving the same equations ( [ 4.13 ] ) , ( [ 4.18 ] ) with the same boundary conditions at @xmath231 .
if @xmath257 is fixed , the equations ( [ 4.12 ] ) ( [ 4.14 ] ) can be rewritten in a simpler form .
denote @xmath334 then @xmath335 @xmath336 with the boundary condition @xmath337 at @xmath338 for a i - type solution .
[ equation ( [ 4.25 ] ) solves ( [ 4.14 ] ) for @xmath339 .
equation ( [ 4.24 ] ) follows from ( [ 4.25 ] ) and @xmath340 which is equivalent to eq .
( [ 4.12 ] ) for @xmath341 . ]
the phase diagram corresponding to the equation ( [ 4.24 ] ) is shown in fig .
4 where the solid line represents the solution of the equation @xmath342 the function @xmath343 is shown by the dashed line in fig . 4 .
@xmath344 the @xmath92-dependence of @xmath226 at @xmath345 is different for @xmath346 and @xmath347 . in both cases
@xmath92 approaches infinity at finite @xmath277 values .
we have at @xmath346 , @xmath348 , @xmath349 and at @xmath347 , @xmath348 , @xmath350 the equation ( [ 4.25 ] ) enables us to find @xmath277 as a function of @xmath92 .
@xmath351 the integral in ( [ 4.29 ] ) converges in both cases ( @xmath346 and @xmath347 ) ; because of this , @xmath92 goes to infinity at a finite @xmath352 ( @xmath353 ) .
near @xmath354 at @xmath346 ( @xmath355 ) we have @xmath356 and , consequently , @xmath357 if @xmath347 ( @xmath358 ) , @xmath359 and thus @xmath360 it can be shown that the relation between the position of the singularity at @xmath354 of the green s function and the quantity @xmath292 can be written as @xmath361.\ ] ] the formulae ( [ 4.31 ] ) and ( [ 4.33 ] ) define the behaviour of @xmath362 near the singularity for a solution of type i. in order to obtain the behaviour of @xmath28 near the singularity for a solution of type ii it is sufficient to notice that at @xmath231 such a solution has the form @xmath363 the replacement of @xmath92 by @xmath364 changes only the sign of @xmath28 . replacing @xmath365 by @xmath366 and solving ( [ 4.24 ] ) and ( [ 4.25 ] ) for @xmath367
we will find @xmath92 to be negative , and near the singularity @xmath92 will go to @xmath368 .
this means the following .
if the first term @xmath369 of the expression @xmath370 equals zero for the solution i at @xmath354 , then @xmath371 is zero for the solution ii at @xmath372 .
if @xmath373 , both solutions become complex : @xmath374 & + & \frac{i\pi}{\beta } \quad\mbox{,}\quad \beta>\frac{1}{2 } \nonumber\\ \phi = -2 \ln \left[\,c_1 \left(\frac{q}{m^ * } - 1 \right)\right ] & + & 2i\pi \quad\mbox{,}\quad \beta<\frac{1}{2 } .\end{aligned}\ ] ] the trajectories of @xmath375 at @xmath376 in the complex plane @xmath92 are shown in fig.5 for the solutions i , ii [ for @xmath377 .
@xmath378 the solution of the type i has a remarkable feature : @xmath379 for any @xmath380 values .
hence , taking @xmath381 , the imaginary part of @xmath307 in ( [ 4.8 ] ) @xmath382 turns out to be positive . at the same time
@xmath383 for the solution of the type ii is an oscillating function ; this can lead to a contradiction with the unitarity condition for the green s function . for complex @xmath81 values
@xmath362 has no singularities .
if we move in the complex plane along an arbitrary ray [ from the origin to infinity ] , the trajectory of @xmath375 does nt approach infinity ( curves i and ii ) . the main result of this section is the following . in the framework of the equation for the fermion green s function ( [ 4.1 ] ) there exist solutions corresponding to broken symmetry provided @xmath87 has an asymptotically free behaviour and @xmath384 .
these solutions behave at @xmath267 as @xmath385.\ ] ] the expression ( [ 4.39 ] ) has a mass term decreasing at infinity .
if a fermion green s function corresponds to symmetry breaking , it is natural to expect the existence of goldstone - type bound states .
this expectation is connected with the belief that if @xmath245 , the axial current has to be conserved .
this is , however , not necessarily true because due to divergences in the theory a leakage of the current is possible into the region of the ultraviolet cutoff .
a typical example for this phenomenon is the anomaly .
but even in a non - anomalous case it is not obvious whether the current conservation is implied by the equation for the green s function or it is imposed as a condition on the solution of the equation . in order to clarify this , let us consider the equation for the bound state @xmath92 supposing that it is a pseudoscalar .
@xmath386 this equation has to have a solution decreasing for large @xmath273 at @xmath387 .
it is easy to see that indeed there exists such a solution . at @xmath388
the equation for @xmath92 is an equation for the variation of a function which satisfies the equation for @xmath28 .
if this variation is taken in the form @xmath389 , @xmath92 obviously satisfies the equation and decreases at @xmath281 as @xmath390 this , however , does not mean that we have particles with @xmath387 .
indeed , the equation ( [ 5.1 ] ) has a solution decreasing at @xmath284 for any @xmath391 values .
the reason for this is that the equation is highly degenerate .
it has a solution of the form @xmath392 where @xmath393 and @xmath394 are any combinations of dirac matrices with coefficients not depending on @xmath81 .
this can be checked by substituting directly ( [ 5.3 ] ) into ( [ 5.1 ] ) and using the equation for @xmath28 .
the reason for this degeneracy is the invariance of the equation ( [ 5.1 ] ) under lorentz transformations , under all discrete symmetries of the dirac equation and , for @xmath309 , even scale invariance and translational invariance in momentum space . if , for example , @xmath395 , then , due to the ward identity @xmath396 @xmath397 is the divergence of the axial current .
hence , if the equation ( [ 5.1 ] ) is satisfied for @xmath398 , it has to have the solution ( [ 5.4 ] ) . in order to show that ( [ 5.1 ] )
has a decreasing solution at @xmath399 for any @xmath226 value , let us notice that in euclidean space this equation has the structure of the schrdinger equation @xmath400 in four dimensions at @xmath401 , with a potential depending on @xmath81 , spin variables and the external vector @xmath226 .
for such a potential the total four - dimensional angular momentum is not conserved , only its projection @xmath402 onto @xmath403 .
an equation of this type has always non - singular solutions with incoming waves of given @xmath402 .
e.g. , considering an incoming wave with @xmath404 for the pseudoscalar @xmath92 , we will have a solution @xmath405 at @xmath230 . or
, taking an incoming wave of the form @xmath406 , we will have @xmath407 suppose we found the solution @xmath408 .
then , due to the fact that for @xmath409 the solution ( [ 5.4 ] ) behaves as @xmath410 we will find that @xmath411 is decreasing with @xmath284 at any @xmath226 . by stating that the equation for bound states has a solution at any value of @xmath226
we do not mean that there are no bound states ; it means only that the mass of the bound state has to be calculated independently .
the most natural way to do this is to calculate the self - energy of the state @xmath412 and to solve the equation @xmath413 alternatively , one can calculate the forward compton scattering of the bound state on a fermion , @xmath414 and after that integrate over the distribution of fermions in the vacuum .
but this way we will never get a massless goldstone state .
the reason for this is the almost obvious fact that ( [ 5.9 ] ) and ( [ 5.11 ] ) contradict the condition of axial current conservation .
let us consider the condition for current conservation in detail .
if we introduce @xmath415 as a set of diagrams @xmath416 with a massive fermionic green s function , it will not satisfy the ward identity .
however , the ward identity will be satisfied by the sum of @xmath415 and of the goldstone contribution : @xmath417 namely [ @xmath418 , @xmath419 here @xmath136 is the yukawa coupling of a goldstone boson to a fermion [ and @xmath420 is related to @xmath136 via the fermion loop diagram ] @xmath421 knowing @xmath422 , we can define the yukawa coupling @xmath136 by ( [ 5.14 ] ) , ( [ 5.15 ] ) .
let us apply the operator @xmath423 to ( [ 5.14 ] ) .
we find that @xmath424 satisfies the equation ( [ 5.1 ] ) since @xmath425 and the right - hand - side of ( [ 5.14 ] ) satisfy the same equation .
but the existence of ( [ 5.13 ] ) implies that the mass of the goldstone boson has to be zero . in order to clarify the situation with compton scattering ,
let us consider the ward identity for the amplitude @xmath426 where the dashed line with the momentum @xmath427 corresponds to the axial current , the wavy line with momentum @xmath56 corresponds to the goldstone state , and the other two lines to the fermions .
the ward identity for this amplitude is @xmath428 to fulfil ( [ 5.17 ] ) , @xmath429 has to be equal @xmath430 here @xmath431 is defined by ( [ 5.13 ] ) . using ( [ 5.13 ] ) and
( [ 5.17 ] ) we obtain @xmath432 at large @xmath433 , @xmath434 values we will have @xmath435 this means that similarly to the case of asymptotically non - free theories the goldstone - fermion scattering amplitude does not depend on the momentum of the goldstone boson ; it decreases only as a function of fermion virtuality . under these circumstances
it is obvious that even in an asymptotically free theory the goldstone boson has kind of a point - like structure .
amplitudes for the interactions of many goldstones with fermions can be found in an analogous way and have the same property [ the property of being independent of the goldstone momenta ] .
the self - energy of the goldstone state is now different from ( [ 5.9 ] ) .
it contains two terms @xmath436 and equals @xmath391 at small @xmath391 .
as we already have said , in this approach the ward identity becomes the definition of the yukawa coupling @xmath136 ( the wave function of the goldstone boson ) through @xmath398 which has a clear diagrammatic meaning .
the equation contains the amplitude @xmath420 of the transition between the goldstone and the axial current .
the expression ( [ 5.15 ] ) for @xmath437 is highly symbolic , because it contains overlapping divergences . in order to write a sensible expression
we can use the same procedure as in section [ i ] where we calculated the polarization operator of the photon . applying the ward identity , we can write instead of ( [ 5.15 ] ) @xmath438 differentiating ( [ 5.22 ] ) over @xmath226 we will get for @xmath388 @xmath439 [ this diagrammatic expression contains exact fermion propagators and bare vertices . ]
if we want to get rid of the @xmath440-s we have to commute @xmath440 with the green s functions and the interaction vertices along one of the fermionic lines . doing this , we obtain @xmath441 due to the conservation of the vector current , the first two terms in ( [ 5.24 ] ) are zero at @xmath388 .
the first one is just the photon polarization operator at @xmath388 , the second one is the amplitude for the decay of a zero momentum scalar into two zero - momentum photons .
the last term looks like the zero - momentum pseudoscalar - photon scattering amplitude which also has to be zero .
this , however , is not true , because it does not contain all the necessary diagrams . observing that this term does not contain overlapping divergences we can write @xmath442 in the zeroth order in @xmath443 , @xmath444
up to now we discussed the goldstone states in a relatively abstract way without fixing the concrete asymptotically free theory . in real qcd
we have quarks with different flavours and there is a difference between flavour singlet and flavour non - singlet states . in order to clarify the picture it will be useful to describe the goldstone state in a different way .
the previous discussion shows that the goldstone states in asymptotically free and non - free theories are rather similar .
therefore it is natural to try to introduce the goldstone boson in the usual way as a point - like state and to see how this state will interact . in the usual discussion of a goldstone particle
we suppose that there is a point - like pseudoscalar interaction between this particle and a fermion with the pseudovector coupling @xmath445 this interaction induces the radiative correction to the propagator @xmath446 of the pseudoscalar @xmath447 where the wavy line describes the bare pseudoscalar green s function @xmath124 .
if the fermion is massless and the axial current is conserved , this pseudoscalar will not interact ; its self - energy @xmath448 is equal to zero .
if , due to symmetry breaking , the fermion becomes massive , it starts to interact and acquires a self - energy different from zero . by using the diagrammatic definition of @xmath449
[ see ( [ 5.22 ] ) ] , we will find @xmath450 where @xmath420 is the same amplitude for the goldstone - current transition as what was discussed in the previous section .
hence , @xmath451 in the limit @xmath452 we will have @xmath453 the pseudovector interaction between goldstone and fermion is @xmath454 with a pseudovector coupling @xmath455 defined by the fermion mass .
the limiting procedure @xmath452 can be understood if we accept that the interaction responsible for symmetry breaking changes the fermion vacuum fluctuations not only at finite momenta but also near the ultraviolet cutoff .
this change in the fermion vacuum fluctuations is responsible for the leakage of the axial current in [ from ] the region of finite momenta ; it can produce the driving term for the goldstone state , recovering the current conservation .
in general , the pseudovector coupling has a disadvantage compared to the pseudoscalar coupling which we have discussed before : it looks unrenormalizable .
but in the case of flavour non - singlet states it can always be replaced by the pseudoscalar coupling with the help of the trivial relation [ @xmath418 @xmath456 which leads to the ward identity ( [ 5.14 ] ) for pseudoscalar coupling .
if we include the emission and the absorption of goldstone bosons inside the diagram , then in the process of this replacement a point - like amplitude appears which corresponds to the quark interaction with many goldstone bosons .
nevertheless i believe it is possible to prove that , due to the decrease of these amplitude as functions of quark virtuality , the theory is renormalizable . due to the anomaly , the case of a flavour singlet current is very different even on the level of the goldstone green s function . in this case the corresponding polarization operator @xmath139 will contain not only the quark loop which we have discussed but also a gluonic contribution @xmath457 the triangle diagram @xmath458 present in ( [ 6.8 ] ) was calculated many years ago by adler , bell and jackiw [ 3 ] @xmath459 with this expression for @xmath458 , @xmath460 still has the form ( [ 6.4 ] ) but it will be quadratically divergent and @xmath461 where @xmath462 is the ultraviolet cutoff .
this means that goldstone particles exist in the anomalous case but they are decoupled from all physical states . at the same time the ward identity still makes sense because the product @xmath424 does not depend on @xmath420 .
nevertheless , the concrete form of the ward identity will change .
the reason for this is again the adler - bell - jackiw anomaly [ 3 ] . for the triangle diagram @xmath463 the replacement of the pseudovector coupling by the pseudoscalar coupling gives an incorrect result : instead of the correct expression @xmath464 which was obtained in [ 3 ]
, we get just the first term .
we can try to write the ward identity using ( [ 6.12 ] ) , but this turns out not to be necessary .
the reason is that in this approach the axial current @xmath465 between gluonic states , defined symbolically by the relation @xmath466 ( where the term @xmath467 is defined diagrammatically and @xmath468 ) is just the transverse part of @xmath467 : @xmath469 the axial current between quark states @xmath470 can be written as @xmath471 here @xmath472 is the same set of diagrams as in the non - singlet case and @xmath473 is the `` axial current of gluons '' @xmath474 in the same way the goldstone - quark interaction can also be divided into two parts . in the first part we can replace the pseudovector coupling by the pseudoscalar coupling .
the second part is the longitudinal part of @xmath473 .
consequently , @xmath475 and the ward identity can be written in the form @xmath476 this expression enables us to answer the question , what happens with particles like @xmath1 which would be goldstone states if we did not take into account that they can decay into two gluons .
let us consider the contribution of the massive pseudoscalar flavour singlet particle @xmath1 to the ward identity ( [ 6.19 ] ) .
the divergence @xmath477 has a pole corresponding to @xmath1 : @xmath478 the yukawa coupling of the goldstone state has also a pole : @xmath479 the right - hand side of ( [ 6.19 ] ) has , however , no poles .
this condition can be satisfied if @xmath480 the same ward identity ( [ 6.19 ] ) [ substituting ( [ 6.20 ] ) , ( [ 6.21 ] ) and using ( [ 6.22 ] ) ] gives at @xmath387 @xmath481 from which equation it follows that @xmath482 for @xmath1 is equal ] @xmath483 this means that @xmath1 acquired a mass due to the transition into a goldstone boson which itself is decoupled .
it is interesting to notice that at relatively small @xmath482 when the comparison between ward identities for different values of @xmath391 ( @xmath484 , @xmath387 ) makes sense , ( [ 6.24 ] ) gives us the same result as the equation ( [ 5.9 ] ) ( without the additional point - like term which is present in the @xmath3-meson case ( [ 5.21 ] ) ) . indeed , in ( [ 5.9
] ) we can write @xmath485 the sum of the first two terms is equal to @xmath391 ( as in ( [ 5.21 ] ) ) and therefore @xmath486 in agreement with ( [ 6.24 ] ) ) and ( [ 6.25 ] ) have different signs ; see the previous footnote . ] ] .
we see that the subtraction term in the @xmath3-meson self - energy , reflecting the quasi - local structure of the pion , disappears in the case of @xmath1 . in this sense @xmath1
is a normal bound state without a point - like structure .
the approach we presented here for the resolution of the @xmath0-problem is technically very close to the approach developed by veneziano [ 4 ] but the underlying physics differs essentially . in veneziano s
approach big long - range fluctuations are responsible for the @xmath1 mass . in our approach
@xmath1 is a normal @xmath2 bound state with no local structure which would be responsible for its small mass if it were a goldstone state .
this local structure is destroyed by the decay on hard gluons .
we have discussed in detail an asymptotically free theory with massless fermions .
we came to the conclusion that in order to obtain the correct spectrum of goldstone particles , axial current conservation has to be imposed on the theory .
qcd , however , contains massive quarks and the same spectrum of massive quasi - goldstone particles ( the pseudo - scalar octet ) as the theory with massless quarks .
the problem is , how to impose the condition of axial current conservation on a theory which obviously does not conserve the axial current .
strictly speaking ,
i do nt know how to do this . for our real world
, however , there is a natural possibility to solve the problem . in the real world qcd
is part of the standard model describing strong , electromagnetic and weak interactions . in the standard model all particles
are supposed to be intrinsically massless and their masses appear as the result of symmetry breaking due to some kind of higgs mechanism with or without elementary higgs particles .
the possibility of such a mechanism is guaranteed by the conservation of the left - handed @xmath487current @xmath488 . for the matrix elements @xmath489 of this current between any two quarks with momenta @xmath490 , @xmath491
, we have the ward identity [ @xmath418 @xmath492 in the case of massive fermions @xmath493 contains the contribution of three goldstone bosons responsible for the masses of @xmath494 and @xmath495 : @xmath496.\ ] ] for large @xmath273-values @xmath28 contains a massive term @xmath497 .
hence , at large @xmath273 we have @xmath498 .\ ] ] at small @xmath273 of the order of the qcd scale , @xmath499 is defined by the total quark mass @xmath307 which for the light quark is much larger than @xmath329 . using the relations ( [ 7.1 ] ) ( [ 7.3 ] ) we can calculate the masses of the pseudoscalar octet in the same way as we did for @xmath1 .
suppose there is a bound state of light quarks which is a pseudoscalar particle ( the @xmath3-meson ) with a finite mass @xmath402 .
the pole corresponding to this particle will contribute to both terms in ( [ 7.2 ] ) . with this pole contribution ,
@xmath500 and @xmath499 can be written in the form [ with @xmath501 standing for `` non - pion '' or `` non - pole '' ] @xmath502 @xmath503 as in the case of @xmath1 , the condition for the pole cancellation gives us the value of @xmath482 : @xmath504 at @xmath387 we have @xmath505.\ ] ] [ here @xmath506 is the light - quark mass matrix , @xmath507 , with @xmath508 . ]
the yukawa coupling @xmath509 of the bound state has to decrease at large @xmath273 .
therefore we have to identify @xmath510 with @xmath511 .
then @xmath512 where @xmath513 is the isotopically invariant part of the quark mass term behaving at @xmath409 as @xmath514 .
finally , @xmath515 is defined by weak interactions with goldstone states , and @xmath516 is determined by pion exchange .
the conclusion of these considerations is that in the case of massive quarks the conservation of the left - handed @xmath487 current can play the same role for the calculation of the coupling and the masses of the flavour non - singlet pseudoscalar particles as in the massless case the conservation of the axial current .
the result is that the masses become different from zero while the yukawa coupling remains the same at least for small @xmath329 values .
the expression ( [ 7.6 ] ) for @xmath482 can be also obtained without discussing the current conservation , just by calculating @xmath517 as it is defined by ( [ 5.21 ] ) for the quark mass @xmath518 , keeping only the term linear in @xmath329 .
it is in agreement with the idea that @xmath329 affects the mass of the goldstone state but not its wave function , which always has been the common belief .
it is not trivial that , if we calculate @xmath482 by using ( [ 7.6 ] ) or ( [ 5.21 ] ) , we get a logarithmic divergence . for the @xmath3-meson mass
the main divergent part equals @xmath519 ( here @xmath520 is of the order of @xmath294 ) . in the standard model
the behaviour of @xmath329 and @xmath521 can be calculated in a fantastically wide region of @xmath273 : from 1 gev up to the scale where one of the couplings of the standard model ( the @xmath0-coupling @xmath522 , the yukawa coupling @xmath523 or the coupling of the self - interaction of the higgs particles , @xmath524 ) becomes of the order of unity . if the @xmath0-coupling @xmath522 is the first to become large , which seems natural , this scale is @xmath525 gev .
at @xmath526 the behaviour of @xmath329 and @xmath521 is unknown .
because of this , @xmath527 is not calculable in principle and it has to be considered as an arbitrary parameter .
if we , however , assume that at @xmath273 larger than @xmath462 , @xmath528 and @xmath529 vanish , we will be able to calculate @xmath527 and , surprisingly , this calculation gives a reasonable value for @xmath527 with @xmath525 gev [ 2 ] .
the expression ( [ 7.9 ] ) for @xmath530 is also in agreement with the naive expression @xmath531 the important difference is that now @xmath532 is determined not only by strong but also by weak interactions : it goes to infinity if the weak interaction is removed .
from the previous discussion it is obvious that the small - mass pion contribution has to be included in the equation for the light - quark green s function .
fortunately this is very easy to do .
having in mind that now the diagrams contain not only the gluon contribution but also the emission and the absorption of pions we will find that as before , the main contribution to @xmath63 comes from the simplest diagram 0= = -0.50 by 0.50 by 3pt with the coupling @xmath533 at zero momentum instead of the gluon coupling @xmath534 .
it leads to the following equation for the green s function : @xmath535 the equation for bound states ( [ 2.6 ] ) has also to be changed .
the correction comes from the diagrams @xmath536 instead of ( [ 2.6 ] ) we will have @xmath537.\end{aligned}\ ] ] here @xmath524 is the emission amplitude of the zero momentum pion in the transition of the bound state to the @xmath2-pair .
this amplitude has to be defined by the axial current conservation .
there is another important quantity in this equation , namely : @xmath538 . in section [ v ]
we have written an explicit expression for @xmath539 including only the gauge field contribution and ignoring the pion contribution .
now we include the pion contribution in the equation for the green s function and , to be self - consistent , we have to do the same for @xmath539 .
i was not able to carry out this in any order in the yukawa coupling , but in the first order in @xmath266 the equation ( [ 5.25 ] ) is correct if one adds the diagrams @xmath540 as we see , the gluonic correction of the order of @xmath443 and the pionic correction of the order of @xmath266 have the same diagrammatic structure . in order to estimate the value of these corrections ,
let us take just the contributions of zero momentum gluons and pions to them .
it can be shown that the contribution of zero momentum gluons cancels in the last two diagrams of ( [ 5.25 ] ) .
zero momentum pion contribution comes only from the first diagram of ( [ 8.3 ] ) .
transferring the differentiation from the fermionic line to the pionic line in this diagram we obtain the contribution of zero momentum pions in the form @xmath541 the expression for @xmath539 which includes the zero - momentum pion contribution is @xmath542 it gives us an understanding of the scale of possible pion contributions .
it is interesting to note , that ( [ 8.5 ] ) is not an expression in terms of the green s functions but an equation for @xmath538 . in the next paper
we will show that the pion contribution changes essentially the structure of the equation for the green s function .
the new equation has a solution corresponding to the confined quark . at the same time the symmetry breaking solution will not necessarily survive ( at least if @xmath67 is large ) .
i would like to thank yu .
dokshitzer , c. ewerz , b. metsch , j. nyiri , h. petry , d. schtte and a. vainshtein for many helpful discussions .
i am very grateful to all the members of the institute for theoretical nuclear physics for their kind hospitality and the warm and inspiring atmosphere i experienced during my stay in bonn .
i express my gratitude to the humboldt foundation for making this stay possible .
1 . v.n .
gribov , lund preprint lu - tp 91 - 7 ( 1991 ) 2 .
gribov , bonn preprint tk-95 - 35 ( 1995 ) , hep - ph/9512352 3 .
s. adler , phys .
* 177 * ( 1969 ) , 2426 ; j.s .
bell , r. jackiw , nuov . cimen . *
60a * ( 1969 ) , 47 4 . g. veneziano , nucl . phys . * b159 * ( 1979 ) , 357 ; phys . lett . * 95b * ( 1980 ) , 90
|
the formulation of qcd which contains no divergences and no renormalization procedure is presented .
it contains both perturbative and non - perturbative phenomena .
it is shown that , due to its asymptotically free nature , the theory is not defined uniquely .
the chiral symmetry breaking and the nature of the octet of pseudo - scalar particles as quasi - goldstone states are analysed in the theory with massless and massive quarks .
the @xmath0 problem is discussed .
-1 cm .0 cm .0 cm 9 in 6.5 in = 2ex `
bonn tk 97 - 08 * qcd at large and short distances + ( annotated version ) * vladimir gribov _ landau institute for theoretical physics , moscow + and + research institute for particle and nuclear physics , budapest + and + institut fr theoretische kernphysik der universitt bonn _
july 28 , 1997
|
the phenomenon that a class of externally slow driven systems evolves naturally into a state of no single characteristic size or time without fine tuning of any parameter is known as self - organized criticality ( soc)@xcite . in soc , the system evolves into a non - equilibrium steady state characterized by long range spatio - temporal correlations and power law scaling behavior as observed in equilibrium critical phenomena@xcite .
the phenomenon is observed in many natural physical and chemical processes@xcite .
sandpile , a prototypical model , was introduced by bak , tang and wiesenfeld ( btw)@xcite for studying soc .
the system here is driven by adding sand grains , one at a time , randomly to the sites of a regular lattice . in btw
, the model evolves following certain deterministic rules for distributing grains of a sand column among all the nearest neighbors equally if the height of a column attains a predefined critical height .
this intermittent burst of toppling activity is called avalanche . soon after the introduction of btw sandpile model , a stochastic version of the model namely manna s stochastic model ( msm)@xcite was proposed . in msm
, sand grains flow in two randomly selected directions out of four possible directions on a square lattice after toppling .
it was initially believed that btw and msm belong to the same universality class@xcite .
however , calculating an extended set of exponents , introduced by christensen and olami@xcite , ben - hur and biham claimed for the first time that msm belongs to a new universality class@xcite .
different universality class of msm has been confirmed later by performing multifractal@xcite and moment@xcite analysis of the avalanche distribution functions .
it is found that the distribution functions of avalanche properties obey finite size scaling ( fss ) in msm@xcite whereas in btw , some of them obey multi - scaling@xcite .
external bias fields were found to have non - trivial effect on the critical properties of various lattice statistical models in equilibrium critical phenomena .
for example , in directed self - avoiding walks ( dsaw)@xcite , spiral self - avoiding walks ( ssaw)@xcite , spiral lattice animal ( sla)@xcite , directed percolation ( dp)@xcite , spiral percolation ( sp)@xcite , directed spiral percolation ( dsp)@xcite , etc , the external bias has changed the critical behavior of these models .
the effect of external bias on soc is studied so far applying external global directional bias only .
there are two such models , directed sandpile model ( dsm ) introduced by dhar and ramaswamy@xcite and directed fixed energy sandpile ( dfes ) model introduced by karmakar and manna@xcite . in these models ,
sand grains flow in a globally preferred direction after toppling .
application of global bias on the sandpile model introduces anisotropy in the systems and leads to new universality classes for dsm@xcite and dfes@xcite . in contrary , the lattice statistical models like sla and sp not only remain isotropic under a local rotational bias but also exhibit nontrivial critical behavior@xcite . it is then intriguing to study the effect of a rotational constraint on the critical properties of sandpile models at out of equilibrium situation . in this paper , a new two state `` quasi - deterministic '' rotational sandpile model ( rsm ) is introduced imposing rotational constraint on the flow of sand grains . in this model ,
a site topples if it exceeds a predefined critical height and two sand grains flow , one in the forward direction , the direction from which the last sand grain was received and the other in a specific rotational direction , say clockwise with respect to the forward direction . since the direction of last sand grain received varies site to site , the rotational constraint is local in nature . during avalanche ,
the direction of the last grain received may change if the toppling sequence is changed .
this introduces certain `` randomness '' or `` internal stochasticity '' in the model .
rsm includes features like mass conservation , open boundary , local deterministic rule for grain distribution along with toppling imbalance , certain stochasticity .
the model is studied here numerically in two dimensions ( @xmath0 ) and the critical avalanche properties at the non - equilibrium steady state are characterized .
an extended set of critical exponents are calculated , nature of the scaling functions are determined , and time auto correlations between the toppling waves are analyzed .
interestingly , it is found that some of the critical exponents are similar to that of btw but different from that of msm whereas the scaling functions follow usual fss as in msm .
consequently rsm belongs to a new universality class .
a physical realization of rsm could be a sandpile on a disk rotating slowly about an axis perpendicular to the plane of the disk and passing through the center of the disk .
rsm is defined here on a square lattice of size @xmath1 in two dimensions ( @xmath0 ) .
a positive integer @xmath2 , the height of the sand column , is assigned to each lattice site .
initially , all @xmath2s are set to zero . sand grains are added , one at a time , to randomly chosen lattice sites and the variable @xmath2 is incremented to @xmath3 if a sand grain is added to the @xmath4th site .
a site is called active or critical when the height of a site becomes greater than or equal to a predefined threshold value @xmath5 , as in msm .
the active site will burst into a toppling activity .
toppling of the first active site initiates an avalanche .
this site can be called the origin of the avalanche .
the flow of sand grains is now demonstrated with the help of fig.[model ] . on the very first toppling of the active site , the central black circle ,
two sand grains are given away to two randomly selected nearest neighbors out of four nearest neighbors on a square lattice .
note that this is the only externally imposed stochastic step in the model .
the grey circles ( sites @xmath6 and @xmath7 ) represent the recipient sites . as soon as a site receives a sand grain , the direction @xmath8 from which the grain was received is assigned to it along with the increment of its height @xmath2 .
the value of @xmath8 can change from @xmath9 to @xmath7 as there are four possible directions on a square lattice .
the solid arrows in fig.[model ] indicates different possible directions . as the avalanche propagates , the direction @xmath8 and height @xmath2 are updated on receiving a sand grain and the information from which direction the last sand grain was received is only kept .
now the next active sites with @xmath10 in the avalanche will topple following a deterministic rule .
two sand grains from a critical site will flow , one in the forward direction , the direction from which the last grain was received , and the other in a clockwise rotational direction with respect to the forward direction .
note that , distribution of sand grains from an active site depends on the receiving direction of the last sand grain before toppling which varies site to site .
thus , the rotational constraint represents a spatially local bias here in rsm in contrast to the directed bias where the directions of sand flow are globally fixed .
if the grey sites ( @xmath6 and @xmath7 ) in fig.[model ] are the new active sites after receiving sand grains from the central site , the flow of sands after toppling of the grey sites will be in the forward and along a rotational direction as indicated by dotted arrows .
the toppling rules for the @xmath4th active site can be stated as @xmath11 where @xmath8 is the direction from which the last sand grain was received by the @xmath4th site .
if the index @xmath12 becomes greater than @xmath7 it is taken to be @xmath9 .
it is important to note that the number of sand grains outgoing from a site after toppling is not necessarily equal to the number of sand grains incoming to the same site after toppling of its nearest neighbors once each .
therefore , there is toppling imbalance in rsm as in msm@xcite whereas there is complete toppling balance in btw .
note that , toppling imbalance in rsm is due to the imposition of rotational constraint whereas the same appears in msm due to stochasticity in grain distribution . toppling of an active site may cause a series of intermittent bursts which constitutes an avalanche . during an avalanche
no sand grain is added .
propagation of an avalanche stops if all sites of the lattice become under critical .
the avalanche dynamics is studied with open boundary condition .
the number of sand grains remains conserved in the model .
the steady state corresponds to the constant average height of the sandpile when the current of incoming flux of sand grains is equal to that of the outgoing flux .
it should be noted here that the final state in an avalanche depends on the sequence of toppling of the critical sites due to the rotational rules considered here . it is demonstrated in fig.[nonab ] considering two nearest neighbors at the critical state ( @xmath13 ) at the same time step .
the numbers represent the height of the sand column at that site .
the short arrows associated with the numbers represent the direction from which the last sand grain was received .
it can be seen that reversing the order of toppling , two different final states are obtained .
it has been verified for a larger lattice that the interchange of toppling sequence leads to different final configurations starting from the same initial state .
consequently , the model is non - abelian .
note that toppling of one site may change the present state ( direction from which the last sand grain was received ) of the nearest neighbors .
consequently , the sand grains will flow in different directions than it was expected before toppling of its neighbors .
apart from the initial stochastic step , the imposition of the rotational rule then also introduces some randomness in the model during time evolution .
this can be considered as `` internal stochasticity '' in the model in contrast to the externally imposed stochasticity in msm .
this inherent randomness makes the model not only non - abelian but also `` quasi - deterministic '' . due to the stochastic dynamical rules , msm had already been found non - abelian@xcite .
it is also important to notice that the local correlation in the rotational toppling rule then can not propagate throughout the avalanche as in btw because of the `` internal stochasticity '' in the model .
rsm is thus a new two state , `` quasi - deterministic '' , non - abelian sandpile model under local external rotational bias . despite few attempts , the critical behavior of non - abelian sandpile models are less understood .
the non - abelian zhang model @xcite belongs to the same universality class as that of abelian btw in the limit of small quantum of energy added to the system@xcite .
directed slope model is non - abelian@xcite but it shows the same critical behavior as that of the abelian directed model@xcite .
a crossover behavior of critical exponents from a generalized zhang model to that of btw model depending upon a non universal parameter @xmath14 , the probability of sand flow in a given direction , is shown by biham _
et al_@xcite . on one hand ,
rsm has features of mass conservation , open boundary , local deterministicity in gain distribution to the neighborhood after toppling as that of btw , on the other hand , it has features like toppling imbalance , certain stochasticity and non - abelianity as that of msm .
it is then interesting to characterize the critical properties of rsm which has microscopic features of both btw and msm .
below , the avalanche properties at the steady state are characterized in three different ways : @xmath15 calculating an extended set of critical exponents , @xmath16 performing moment analysis of the probability distribution functions and @xmath17 coarsening an avalanche into a series of toppling waves .
the non - equilibrium steady state is defined by the constant average height of the sandpile at which the current of influx of sand grain to the system is equal to the current of out flux of the same at the open boundary . in order to identify the steady state ,
the average height @xmath18 has been measured generating a large number of avalanches . the average height @xmath19 is plotted against the number of avalanches upto @xmath20 in fig.[sat ] for the system size @xmath21 .
it can be seen that a constant average height @xmath19 is achieved and it remains constant over a large number of avalanches . for smaller lattice
sizes the steady states are reached by smaller number of avalanches . a slight variation of the average height with the system size is observed .
the values of @xmath19 against the system size @xmath22 is shown in the inset . in order to characterize the physical properties of the avalanches occurred at the non - equilibrium steady state ,
simulations have been performed on the square lattice of sizes @xmath23 to @xmath21 in multiple of @xmath6 .
first @xmath20 avalanches were skipped to achieve the steady state .
extensive data collection have been made for each lattice size for averaging , ranging from @xmath24 avalanches for @xmath23 down to @xmath25 avalanches for @xmath21 in ten configurations . in each configuration , the initial @xmath26 avalanches are neglected again on the steady state before collecting data .
it should be mentioned here that due to the rotational constraint the lifetime of an avalanche in rsm is much higher in comparison to that in other models .
generation of large number of avalanche then requires huge computer time in rsm .
a comparison of the morphology of avalanche clusters in the steady state of different sandpile models is made here .
typical large avalanche clusters of btw , msm and rsm obtained in their respective steady states are shown in fig.[avch ] ( color online ) .
the avalanche clusters are generated on a square lattice of size @xmath27 dropping sand grains one at a time at the center of the lattice .
the clusters shown here have @xmath28 maximum number of toppling in each and it is represented by the red color .
different colors correspond to different number of toppling of sites in an avalanche as : red for @xmath28 , green for @xmath29-@xmath30 , blue for @xmath31-@xmath32 , orange for @xmath33-@xmath34 , magenta for @xmath35-@xmath36 , and grey for @xmath7-@xmath9 toppling numbers .
white spaces inside the avalanche correspond to the sites that did not topple at all during the avalanche .
there are few things to notice .
first , rsm avalanche cluster is different from that of both btw as well as msm .
btw avalanche cluster consists of concentric zones of lower and lower number of toppling around a single maximal toppling zone@xcite whereas avalanche cluster of msm is random in nature@xcite .
rsm avalanche cluster neither fully consists of concentric zones as in btw nor it is totally random .
it is important to note that the local correlation in rotational constraint does not lead to long range correlation generating btw like correlated structure .
second , there are more than one red zones ( maximum number of toppling zones ) in the avalanche cluster of rsm in fig.[avch](@xmath37 ) as in the avalanche cluster of msm ( fig.[avch](@xmath38 ) ) whereas there is only one red zone at the center of the avalanche cluster of btw ( fig.[avch](@xmath39 ) ) .
it is then possible to have several maximal toppling zones in an avalanche in rsm whereas in btw always one maximal toppling zone appears@xcite .
rsm avalanche looks like superimposition of several btw like structures around different maximal toppling zones .
occurrence of several maximal toppling zones is also a common feature in msm avalanche@xcite .
third , though the sand grains are added at the central site of the lattice the maximal toppling zones appear at arbitrarily different places in rsm as well as in msm . in btw ,
the central zone correspond to the maximal toppling zone@xcite .
fourth , btw avalanche clusters are compact without holes or no toppling regions inside an avalanche ( see fig.[avch](@xmath39 ) ) . on the other hand , there are several holes appear in msm as seen in fig.[avch](@xmath38 ) .
rsm avalanche cluster is almost compact with few holes here and there .
the appearance of holes in msm is due to the stochastic rule of sand distribution and the same in rsm is due the rotational rule of sand distribution .
note that holes of a single site also appear in the avalanche clusters of rsm .
apparently it seems generation of single sited hole is forbidden by the rotational rule .
single sited holes only could appear at the termination point of two different branching of toppling of a previous site .
an avalanche can be considered as a branching process since toppling of a site can make more than one neighbors critical@xcite .
the rsm avalanche cluster therefore has properties of both the deterministic btw and stochastic msm .
it is then expected that the critical avalanche properties in the steady state will show a mixed behavior .
note that , the avalanche clusters are isotropic in space in rsm as in both btw and msm .
however , the avalanche clusters are anisotropic in the cases of dsm and dfes where two correlation lengths are required to characterize their spatial extensions . in order to characterize different physical properties of the avalanches occurred at the steady state , different quantities like toppling size @xmath40 of an avalanche , avalanche area @xmath39 , lifetime @xmath41 and spatial extension @xmath42
are measured .
toppling size @xmath40 is defined as the total number of toppling occurred in an avalanche .
avalanche area @xmath39 is equal to the number of distinct sites toppled in an avalanche .
lifetime @xmath41 of an avalanche is taken as the number of parallel updates to make all the sites under critical .
spatial extension @xmath42 of an avalanche is given by @xmath43 where @xmath44 , @xmath45 is the position vector of the distinct sites toppled .
the related critical exponents are estimated determining the probability distributions of all these properties ( @xmath46 and @xmath42 ) .
the probability distribution function of an avalanche related quantity @xmath47 at the steady state of a given system size @xmath22 is expected to obey power law behavior given by @xmath48 where @xmath49 is the corresponding critical exponent and @xmath47 stands for @xmath46 and @xmath42 .
@xmath50 is the finite size scaling function and @xmath51 is called a capacity dimension . in the @xmath52 limit ,
the scaling function @xmath53 become a constant and the power law behavior given in eq.[pd ] can be approximated as @xmath54 .
the corresponding exponents @xmath49 can be estimated from the slope of the best fitted straight line through the data points in log - log scale .
data are collected in bins of interval of @xmath55s , @xmath56s , @xmath57s and so on .
finally the data are normalized by the bin widths . in fig.[pds ] , the probability distribution @xmath58 of the toppling size @xmath40 is plotted for different system sizes , @xmath23 to @xmath21 in multiple of @xmath6 .
it can be seen that the toppling size distribution has the same power law behavior for different system size @xmath22 with a cutoff that increases with @xmath22 . since there is no typical toppling sizes of an avalanche at the non - equilibrium steady state , rsm then exhibits self - organized criticality .
the solid line represents the best fitted straight line between the data points of @xmath21 with a slope of @xmath59 .
the error is due to the least square fitting taking into account of statistical error of each data point .
the slopes obtained from the best fitted part of data for other values of @xmath22 remain within this error bar . in order to extract the critical exponents related to other avalanche properties the same procedure
has been followed .
however , in fig.[pdatl ] the probability distribution @xmath60 is plotted against @xmath47 only for @xmath21 where @xmath47 corresponds to area @xmath39 ( circles ) , lifetime @xmath41 ( triangles ) and spatial extension @xmath42 ( squares ) .
it can be seen that the probability distributions @xmath60 follow reasonable power law behavior for each property @xmath47 .
the solid lines represent the best fitted straight line through the data points .
the values of the associated critical exponents @xmath49 are obtained from the slopes of the best fitted straight lines as @xmath61 , @xmath62 and @xmath63 .
the error bars quoted here are the least square fit errors taking into account of statistical error of each data point .
a comparison of the values of the exponents obtained here is made with that of btw and msm in table [ table1 ] .
the values of the exponents for btw and msm are taken from ref.@xcite .
interestingly , the toppling size exponent @xmath64 and the lifetime exponent @xmath65 are different whereas @xmath66 and @xmath67 are almost the same as that of btw
. the disagreement of the lifetime and toppling size distribution exponents with the corresponding btw exponents can be accounted by the fact that in rsm the avalanche waves generally have a spiraling nature around several maximal toppling zones within the avalanche cluster and as a consequence it will take longer time and large number of toppling for an avalanche to die away than that in btw where a single maximal toppling zone occurs and the toppling wave propagates outwardly . on the other hand , in comparison to msm , most of the exponents are found different .
thus , from the point of view of power law correlations , some of the avalanche properties are similar to that of btw but different from msm .
note that @xmath68 , conjectured by majumder and dhar@xcite , is satisfied in case of msm but it is not valid for btw .
it can be seen that the conjecture is just outside the error bar here in case of rsm .
the expected value of @xmath64 in rsm from the conjecture is @xmath69 close to the obtained value @xmath70 .
since the avalanche properties are related to each other , conditional expectation values are defined as introduced by christensen _
et al_@xcite .
the conditional expectation value of an avalanche property @xmath47 when another property is exactly equal to @xmath71 is defined as @xmath72 where @xmath73 is the probability to find a property @xmath47 when the other property is exactly equal to @xmath71 for a given system size @xmath22 . in the steady state ,
the expectation values scale with its argument as @xmath74 where @xmath75 is a critical exponent .
four expectation values @xmath76 , @xmath77 , @xmath78 , and @xmath79 are calculated on a square lattice of size @xmath21 and their scaling behavior are determined . in fig.[ealt ] , the expectation values are plotted against their arguments in order to evaluate the exponents @xmath75 . from the slope of the best fitted straight lines , the values of the exponents are estimated as : @xmath80 , @xmath81 , @xmath82 and @xmath83 .
the values of @xmath75 in dsm@xcite are smaller in comparison to the values obtained here . in table
[ table1 ] , the values of @xmath75 are compared with that of btw and msm .
there are few things to notice .
first , @xmath84 is found greater than one and a relevant exponent .
this is expected because in this model , a site topples many times in an avalanche due to rotational constraint .
second , the exponent @xmath85 is found @xmath86 since the avalanche clusters are almost compact with a few holes here and there .
third , the value of the dynamical exponent @xmath87 is the highest in rsm and it is lowest in btw . because , due to the rotational constraint the sand grains rotates around several maximal zones inside the avalanche and take longer time to complete an avalanche .
fourth , according to the scaling function form given in eq.[gamma ] , the exponents should satisfy the scaling relation @xmath88 .
it can be seen that the scaling relation @xmath89 is satisfied within error bars .
fifth , the values of @xmath84 , @xmath90 and @xmath87 are found different from that of btw as well as msm except @xmath85 .
finally , a set of scaling relations between the probability distribution exponents @xmath49 and the exponents @xmath75 describing the conditional expectation values of the avalanche properties can be obtained from the following identity @xmath91 which would be satisfied by any set of three stochastic variables @xmath47 , @xmath71 and @xmath92 . using this identity and the relation @xmath93
, the following scaling relation can be obtained @xmath94 the above scaling relation is satisfied within error bars for @xmath95 .
thus , the extended set of exponents obtained here in rsm from both power law analysis and conditional probabilities are consistent with the scaling relations .
note that the values of the critical exponents obtained here should remain invariant under the reversal of rotational symmetry . though some of the exponents are close to that of btw
, rsm belongs to a new universality class because the extended set of exponents are not identical .
note that , dsp and defs already belongs to different universality classes because of their anisotropic character .
directed models show a continuous phase transition from an absorbed phase to an active phase@xcite .
now the avalanche properties are analyzed to understand the nature of the scaling functions , fss or multi - scaling , following the method of moment analysis@xcite .
the probability distribution @xmath96 of an avalanche property @xmath47 in a finite system of size @xmath22 is expected to obey a scaling function form as given in eq.[pd ] , @xmath97 where @xmath51 is called a capacity dimension as already mentioned .
the finite system size @xmath22 causes a cutoff of the probability distributions at @xmath98 .
the @xmath99-moments of a property @xmath47 is defined as @xmath100 where @xmath101 .
thus , if the probability distributions obey fss then the moment exponent @xmath102 should have a constant gap between two successive values , @xmath103 . on the other hand ,
if they obey multi - scaling , @xmath102 should have a continuous dependence on @xmath99 .
for a given @xmath99 , the value of @xmath104 has been obtained from the slope of the plot of @xmath105 versus @xmath106 changing lattice size from @xmath23 to @xmath107 in multiple of @xmath6 .
the least square fit error for each @xmath104 value is found @xmath108 . a sequence of exponents @xmath102 , @xmath109 , is obtained for @xmath110 values of @xmath99 between @xmath111 and @xmath7 . in fig.[sq1 ] , the exponent related to the average toppling size @xmath112 is plotted against the moment @xmath99 .
an important point to note here is that the value of @xmath113 is @xmath86 in all three models .
the average toppling size varies with the system size as @xmath114 where @xmath40 is equivalent to the number of steps by a random walker required to reach the lattice boundary@xcite .
thus , the value of @xmath115 represents the diffusive character of the model , one of the characterizing features of the sandpile models .
the diffusive behavior of rsm is also consistent with the fact that spiral random walks are diffusive@xcite . in order to compare the results of rsm with that of btw and msm ,
a sequence of exponents @xmath102 are also obtained for these two models and plotted in the same figure .
it can be seen that the variation of @xmath112 with @xmath99 in fig.[sq1 ] is not identical with either btw or msm .
further analysis of this sequence of exponents is then needed in order to understand the nature of the scaling function .
the slopes @xmath116 are then estimated using finite difference method .
if the probability distributions obey multi - scaling , the rate of change of @xmath102 with @xmath99 should not be a constant . in fig.[mdsat ]
, @xmath116s are plotted against the moment @xmath99 for @xmath109 and compared with that of btw and msm . in fig.[mdsat ] , the solid line represents the data of rsm , the dashed line represents the data of msm and the dotted line represents the data of btw .
the rate of change of @xmath102 with respect to @xmath99 for all three properties @xmath109 in rsm are different from that of both btw and msm and remain unchanged in higher moments as in the case of msm . in btw ,
the derivatives corresponding to toppling size @xmath40 and lifetime @xmath41 do not saturate with the moment @xmath99 . a comparison of variation of @xmath116 with @xmath99 in btw and rsm with respect to msm can be made . in order to have a comparative study a quantity @xmath117
is defined as @xmath118 where @xmath119 and @xmath120 stands for the models btw or rsm .
@xmath117 is plotted against @xmath99 in the inset of corresponding plots .
the dotted line corresponds to @xmath121 and the solid line corresponds to @xmath122 .
it can be seen that the value of @xmath123 remains constant with respect to msm whereas in case of btw , data corresponding to toppling size and lifetime increases slowly with respect to msm .
thus , the scaling functions of the avalanche properties in rsm follow fss as in msm rather than multi - scaling as in btw . the values of the capacity dimensions @xmath51 can be calculated taking the large @xmath99 limit of @xmath124 .
the values obtained are : @xmath125 , @xmath126 and @xmath127 .
the errors are due to the finite difference method adopted for differentiation of the @xmath128 sequence .
the value of @xmath129 is trivially equal to @xmath9 because @xmath130 . since @xmath131 for a given system size @xmath22
, it can be shown that @xmath132 .
taking @xmath133 , one should have @xmath134 . for rsm ,
the values of @xmath51 and @xmath135 are found close .
since @xmath136 and @xmath137 do not saturate in btw , the corresponding capacity dimensions are not possible to estimate .
however , the capacity dimension @xmath138 in all three models are found @xmath86 as it is expected .
the values of @xmath139 and @xmath140 in msm ( as in ref.@xcite ) and rsm differ slightly .
note that @xmath141 is slightly higher than @xmath115 because @xmath142 remains in the nonlinear regime in rsm .
knowing the values of capacity dimensions , it is now possible to check the scaling function form given in eq.[pd ] studying the distribution functions for different system sizes @xmath22 .
the scaling function form is checked by plotting a scaled distribution @xmath143 against the scaled variable @xmath144 for @xmath145 in fig.[fss ] following chessa _
et al_@xcite .
for both the properties , toppling size @xmath40 and lifetime @xmath41 , a reasonable collapse of data are observed for @xmath146 and @xmath107 in support of the assumed scaling function form in eq.[pd ] . in the inset of fig.[fss ] , data collapse for toppling size @xmath40 is also shown in log - normal scale .
the fss forms assumed here in rsm for the avalanche properties are then correct . in spite of the fact that rsm has locally deterministic rule for grain distribution , it is conservative , its avalanche cluster morphology is almost compact , it is diffusive , and some of the critical exponents are similar to that of btw , it is interesting to note that the scaling functions in rsm do not follow multi - scaling as in btw . a sandpile model with microscopic as well as macroscopic characteristics of a deterministic model like
btw follows fss which is characteristics of a stochastic model like msm is a new result .
this has happened due to the nature of the rotational constraint which incorporates `` internal stochasticity '' by changing the state of the critical sites in a time step during time evolution of the system .
note that , due to the presence of rotational constraint on the sand flow the toppling balance of btw is also broken as already mentioned .
recently , it is demonstrated by karmakar _
et al_@xcite that the scaling functions obey fss rather than multi - scaling if toppling imbalance is introduced in the btw sandpile model .
existence of fss in rsm is possibly due to toppling imbalance as well as `` internal stochasticity '' in the model .
fss of avalanche properties can be confirmed by studying time autocorrelation between toppling waves .
the time evolution of toppling dynamics is studied here coarsening the avalanches into a series of toppling waves@xcite .
toppling waves are defined as the number of toppling during the propagation of an avalanche starting from a critical site @xmath147 without toppling @xmath147 more than once .
each toppling of @xmath147 creates a new toppling wave .
thus , the total number of toppling @xmath40 in an avalanche can be considered as @xmath148 where @xmath149 is the number of toppling in the @xmath150th wave and @xmath120 is the number of toppling waves during the avalanche .
it is then possible to generate a wave time series @xmath151 .
following menech and stella@xcite , for a given lattice size @xmath22 , a time autocorrelation function is defined as @xmath152 where @xmath153 and @xmath154 represents the time average .
it has already been demonstrated by menech and stella@xcite and karmakar _
et al_@xcite that @xmath155 is long range for btw whereas it remains negative initially and then becomes zero in case of msm .
thus , the waves in btw have correlation over a longer period of time whereas they are uncorrelated in case of msm .
this is also argued by menech and stella@xcite that this observation is consistent with the fact that the toppling sizs follow multi - scaling in btw and fss in msm .
the time autocorrelation function @xmath155 for the toppling waves has also been calculated here in rsm for a system size @xmath156 taking time average over @xmath20 toppling waves .
@xmath155 is plotted against @xmath41 for rsm in fig.[corr ] . in order to compare the data of rsm with that of btw , and msm , @xmath155s of these models
are also calculated and plotted in the same figure .
it can be seen that there is long range correlation for btw and anti - correlation for msm as expected .
the toppling waves are also uncorrelated here in rsm as in the case of msm .
@xmath155 is found negative initially and then becomes zero .
it is then consistent with the fact that the toppling size distribution follow fss rather than multi - scaling .
the origin of negative autocorrelation in msm is the stochasticity . in rms ,
the local deterministic toppling rule picks up certain randomness during the evolution and as a consequence the toppling wave shows negative auto correlation .
though the rotational constrain has local correlation , the toppling waves become uncorrelated because of the `` hidden stochasticity '' .
it should be mentioned here that in the cases of msm and rsm , the sites involved in a toppling wave may topple more than once unlike in the case of btw . moreover , the toppling numbers as well as the final configurations of msm and rsm strongly depend on the sequence of toppling . as a consequence , eq.[toppl ]
may not satisfy always in rsm and msm .
however , on an average the collection of toppling waves can be considered as a representation of an avalanche in these models .
a new two state `` quasi - deterministic '' sandpile model , rsm , is defined imposing rotational constraint on the sand flow in order to study the effect of local external bias on self - organized critical systems .
the model has microscopic properties like mass conservation , open boundary , local deterministicity in sand grain distribution on toppling as that of btw . at the same time
, the rotational bias introduces toppling imbalance and certain stochasticity at the microscopic level as in msm .
the non - equilibrium steady state of rsm is characterized by power law distribution of avalanche properties .
the avalanche cluster morphology in the steady state exhibits characteristics of both btw as well as msm .
rsm is found to be diffusive in character like other sandpile models .
calculating an extended set of critical exponents it is found that some of the exponents are close to that of btw but different from msm .
the values of the exponents satisfy the scaling relations among them within error bars .
rsm then belongs to a new universality class .
the scaling function forms are determined .
it is found that the scaling functions obey usual fss as in the case of msm rather than multi - scaling as in the case of btw .
this has been confirmed by negative time autocorrelation of toppling waves constituting an avalanche .
a sandpile model having certain microscopic as well as macroscopic features of btw follows fss as that of msm is a new result .
the appearance of fss in rsm may be due to local toppling imbalance and `` internal stochasticity '' caused by the imposed rotational constraint on the model .
* acknowledgment : * the authors thank s. s. manna , d. dhar and m. a. munoz for critical comments and several suggestions on the manuscript .
this work is financially supported by board of research in nuclear sciences , department of atomic energy , india , grant no.2005/37/5/brns .
for example : coagulation ( h. takayasu , phys .
* 63 * , 2563 ( 1989 ) ) , earthquake ( k. chen , p. bak and s. p. obukhov , phys .
a * 43 * , 625 ( 1990 ) ) , river network ( h. takayasu and h. inaoka , phys . rev
. lett . * 68 * , 966 ( 1992 ) ) , forest fire ( b. drossel , s. clar and f. schwabl , phys . rev . lett .
* 71 * , 3739 ( 1993 ) ) , biological evolution ( p. bak and k. snappen , phys .
lett . * 71 * , 4083 ( 1993 ) ) , dissolution of random solids ( s. sinha , v. kishore and s. b. santra , europhys . lett . * 71 * , 632 ( 2005 ) ) , etc .
.[table1 ] comparison of critical exponents obtained in different sandpile models .
the values of the critical exponents for btw and msm are taken from ref.@xcite .
some of the exponents are close to that of btw and most of them are different from that of msm . [ cols="^,^,^,^ " , ]
|
rotational constraint representing a local external bias generally has non - trivial effect on the critical behavior of lattice statistical models in equilibrium critical phenomena . in order to study the effect of rotational bias in a out of equilibrium situation like self - organized criticality
, a new two state `` quasi - deterministic '' rotational sandpile model is developed here imposing rotational constraint on the flow of sand grains .
an extended set of new critical exponents are found to characterize the avalanche properties at the non - equilibrium steady state of the model .
the probability distribution functions are found to obey usual finite size scaling supported by negative time autocorrelation between the toppling waves .
the model exhibits characteristics of both deterministic and stochastic sandpile models .
|
the nature of the low temperature phase of finite dimensional spin glasses is still a subject of controversy @xcite .
recently bray , moore , bokil and drossel @xcite have questioned many of the numerical results obtained with monte carlo methods in the three dimensional edwards - anderson ( ea ) model @xcite . inspired by the study of the migdal - kadanoff approximation ( mka ) of the ea model , they argued that the numerical results , that were obtained at temperatures @xmath6 , could be strongly affected by finite size effects and that one should go to sizes larger than the crossover length @xmath7 in order to see the right ( droplet ) behavior .
they found ( in the framework of the mka ) that the crossover length is @xmath8 for @xmath9 and that it decreases for lower temperatures : @xmath10 when @xmath11 @xcite ( see also the comment @xcite ) .
this is maybe the main motivation that pushed us to study the ea model in the very low temperature region : verify whether the behavior already found at @xmath12 persists at @xmath13 .
in fact at these temperatures we can simulate ( using off - equilibrium techniques ) a system of size larger than @xmath7 ( in this paper , we will present data for sizes @xmath14 and @xmath15 ) .
the numerical data showed in this paper have been measured in the off - equilibrium dynamical regime .
this way of probing the system properties , apart from being much more similar to the experimental procedure , does not present the thermalization problems of a simulation performed at the equilibrium , that would be insurmountable obstacles at so low temperatures .
the efficiency of this way of measuring has been largely tested in the recent past @xcite .
moreover the off - equilibrium dynamics of spin glasses has received in the last years a great attention both from the experimental @xcite and from the analytical @xcite points of view .
taking the measurements in the off - equilibrium regime we are able to confront the droplet model ( dm ) @xcite and the mean field like theory ( mf)@xcite on two grounds : the off - equilibrium regime itself and the equilibrium one , that can be obtained in the limit of very large times .
we can take this limit quite safely thanks to the very large time reached in our simulations .
a preliminary analysis based on the data at temperatures @xmath16 and @xmath4 was reported in reference @xcite . in this paper
we present an extended analysis based on nine different temperatures obtaining a precise temperature dependence of the dynamical critical exponent in order to have an accurate confront with recent experiments .
we have simulated the gaussian ising spin glass on a three - dimensional cubic lattice of volume @xmath17 with periodic boundary conditions .
the hamiltonian of the system is @xmath18 we denote by @xmath19 the sum over nearest neighbor pairs .
@xmath20 are gaussian variables with zero mean and unit variance .
we focus our attention on the study of the point - point overlap correlation function computed at distance @xmath21 and time @xmath22 @xmath23 where @xmath24 and @xmath25 are two real replicas ( systems which evolve with the same disorder ) and the index @xmath26 runs over all the points of the lattice . as usual we denote by @xmath27 the average over the disorder and , in this context , @xmath28 is the average over the dynamical process until time @xmath22 ( for a given realization of the disorder ) . the two replicas ( @xmath24 and @xmath25 ) evolve with different random numbers . the simulation has been performed in a similar way to the experimental procedure : the system is prepared in a high temperature configuration ( actually the initial configurations were chosen at random , i.e.@xmath29 ) and suddenly it is quenched below the ( estimated ) critical temperature , @xmath30 @xcite .
immediately we start taking the measurements , which obviously depend on time .
the equilibrium behavior is recovered in the large time limit .
we have used as dynamical process the standard metropolis method .
we have simulated 4 samples ( 8 systems ) of an @xmath15 lattice , measuring the correlation function at times @xmath31 ( with @xmath32 ) and temperatures @xmath33 , @xmath34 , @xmath35 , @xmath36 , @xmath37 , @xmath38 , and @xmath39 .
in addition , we have simulated 4096 samples of an @xmath14 lattice measuring at times @xmath40 ( with @xmath41 ) and at three temperatures : @xmath16 , @xmath37 and @xmath39 . for the study of the fluctuation - dissipation relation we have used @xmath15 systems and we have simulated them for more than @xmath42 monte carlo steps .
all the simulations have been performed with the help of the parallel computer ape100 @xcite .
first , we analyze the correlation functions computed with the @xmath15 runs .
the study of the numerical data suggest us the following ansatz for the time and spatial dependences of the correlation function @xcite @xmath43\ , \label{main}\ ] ] where @xmath44 is the dynamical correlation length .
the numerical data clearly show that the dynamical correlation length depends on the time following a power law @xmath45 where @xmath0 is the dynamical critical exponent .
the exponents @xmath46 , @xmath47 and @xmath0 and the amplitude @xmath48 could , in principle , depend on the temperature .
however we obtain ( see below ) that @xmath46 , @xmath47 and @xmath48 are almost temperature independent , while @xmath49 is inversely proportional to @xmath5 . .
parameters of the point - point correlation function [ cols="^,^,^,^,^",options="header " , ] [ table : param ] in table [ table : param ] we report the results of our fits ( always done using the cern routine minuit @xcite ) .
we remark that , for a given temperature , we have fitted our numerical data to the ansatz of eq.([main ] ) in two steps . in the first step we fix the distance in the correlation function and we perform the following three parameters fit in the variable @xmath22 @xmath50
we have found that @xmath51 is independent of @xmath21 .
the second step has been to extract from @xmath52 and @xmath53 the exponents @xmath46 and @xmath47 and the amplitude @xmath48 using the formul : @xmath54 and @xmath55 .
we report our final values of @xmath0 , @xmath47 , @xmath48 and @xmath46 in table [ table : param ] .
the resulting values for @xmath49 ( see table [ table : param ] ) can be fitted to a power law ( using all the temperatures of table [ table : param ] ) obtaining @xmath56 from the previous fit we can guess a simpler law for the dynamical critical exponent @xmath57 , obtaining three dimensional spin glass ( see references @xcite ) .
moreover this law was guessed for the gaussian model using numerical data taken at temperatures @xmath16 and @xmath4 in reference @xcite . ]
@xmath58 this kind of behavior suggests that the low temperature dynamics in spin glasses is dominated mainly by activated processes with free energy barriers diverging logarithmically with the size of the system .
we can finally write down the dependence of the dynamical correlation length on the time as well as on the temperature : @xmath59 where we have assumed that the temperature of the phase transition is @xmath30 @xcite .
the agreement of the previous formula with the experiments is very good .
we recall that in experiments @xcite it was found the following dependence for the dynamical correlation length @xmath60 where @xmath61 is the experimental critical temperature ( the authors of this result do not quote the error in the exponent )
. a further check of eq.([main ] ) would be the collapse of the data ( measured at different times and different temperatures ) when plotting @xmath62 versus @xmath63 . to this purpose
we use the data from 4 samples of the @xmath64 runs , together with those measured on 4096 samples of @xmath65 runs .
we remark that , in the @xmath65 runs , the volume is near 19 times less than the @xmath15 runs but we have computed 1000 times more samples and so we expect the errors to be smaller . in fig .
[ plot_corr ] we plot the correlation function for two low temperatures ( @xmath3 and @xmath39 ) using as variables @xmath66 and @xmath67 ( we have taken @xmath68 , see table [ table : param ] ) . in the plots we use the data from both runs ( @xmath14 and @xmath15 ) and they superimpose perfectly . in the insets we present the same data in a log - linear scale in order to let the reader evaluate better the collapse .
it is clear that the scaling is impressive even at the lowest temperature @xmath4 .
we can also state that the finite size effects are negligible for the lattice sizes used .
scaling arguments tell us that the more general scaling function for the correlation function is ( for large @xmath21 and @xmath22 ) @xmath69 where the scaling function @xmath70 is smooth .
moreover , in the scaling regime , @xmath70 should not depend neither on temperature nor on the lattice size .
note that in our ansatz ( [ main ] ) we have chosen an exponential function for the scaling function : @xmath71 , and we show that it fits very well the data .
however to check that our estimates of @xmath46 and @xmath44 are correct we do not need to know @xmath70 .
we can simply plot @xmath67 versus @xmath66 ( as it was done in fig .
[ plot_corr ] ) and check how well the data collapse . in order to check the temperature independence of @xmath70 we show in fig .
[ tutte_t ] the scaling function for three different temperatures ( @xmath72 and @xmath35 ) , together with the exponential function @xmath73 $ ] ( see table [ table : param ] ) obtained through the fitting procedure .
it is clear that the scaling function is really temperature - independent and it can be very well approximated by the exponential function as we have chosen in our ansatz .
another interesting issue is the extrapolation of the correlation function to infinite time . in this limit
we can compare again our numerical results with the predictions of the droplet model and with that of the rsb theory . in the former
the extrapolated correlation function tends to the value for large distances , whereas the rsb prediction is a pure power law going asymptotically to zero @xcite .
our ansatz , which describes perfectly the numerical data , supports the rsb prediction even for the lowest temperatures .
nevertheless , we have tried to fit our data with a functional dependence compatible with the droplet model , that is , @xmath74 , where @xmath75 . if @xmath76 , then the previous formula is exactly our ansatz ( and it implies a breaking of the replica symmetry ) , while if @xmath77 then it would support a droplet picture .
fitting the data to the previous formula , @xmath78 , we have found that at every temperature and even at @xmath4 ( our lowest temperature ) , the best value for @xmath79 is always compatible with zero . at very low temperatures ,
i.e. @xmath4 , the edwards - anderson order parameter is so close to one ( @xmath80 ) than we can safely distinguish between the two competing theories .
in fact in the droplet like formula we have that @xmath78 is almost constant , but for all our purposes it can be considered as a constant . ]
and so we simply should fit the data into the scaling formula of eq.([scaling ] ) without the factor @xmath81 in order to check the correctness of the droplet model . in fig .
[ con_senza ] we present the data rescaled with the formula suggested by the droplet model ( left plot ) and with that implied by rsb ( right plot ) .
it is clear that the rsb prediction fits much better the numerical data .
note that the data error is sufficiently small to affirm safely that the data in the left plot have no collapse at all .
a scaling plot like the left one has been recently presented by komori et al . in @xcite
( see also reference @xcite ) .
we believe that the rather poor collapse of their data ( see fig . 5 in @xcite ) is due to the fact that they neglect the factor @xmath81 in the scaling formula .
a much better collapse would be obtained by plotting @xmath82 versus @xmath66 ( see @xcite ) .
now , we present the results of the analysis based on the generalization of the fluctuation - dissipation theorem ( fdt ) in the out of equilibrium regime @xcite . in this section we will focus on the scaling properties of the aging region and the violation of fluctuation - dissipation at very low temperatures .
a preliminary analysis was done in @xcite studying the violation of fdt at temperature , @xmath16 .
here we have simulated different lower temperatures and so , as byproduct , we can study the scaling properties of the violation of fdt .
an analogous analysis , with many temperatures , was done in @xcite but on the four - dimensional ea model . for the sake of conciseness
, we do not repeat all the formalism and we address the interested reader to one of the previous publications on the subject @xcite . here
we simply recall the main formul that we use . as usual
we define the integrated response to a very small external field as @xmath83 where @xmath84 and @xmath85 .
the autocorrelation function is defined as @xmath86 relating these two functions , in the large times limit , via @xmath87 we have that , at the equilibrium , the fluctuation - dissipation theorem ( fdt ) holds and @xmath88 , while in the aging regime the function @xmath89 can be linked to the equilibrium overlap distribution through @xmath90 @xcite .
models that , in the frozen phase , do not show any breaking of the replica symmetry , have , at the equilibrium level , a static @xmath91 , which dynamically corresponds to the absence of response in the aging regime .
this means that , plotting @xmath92 versus @xmath93 , we obtain a horizontal line in the range @xmath94 @xcite ( in the quasi - equilibrium regime , @xmath95 , and it always holds @xmath96 independently of the model ) in fig .
[ fdt ] we show the results for different temperatures in the usual plot @xmath92 versus @xmath93 .
note that in this plot the fdt line is @xmath97 and so it is different for different temperatures .
it is quite clear that , even for very large times , the curves are far from been horizontal when they leave the fdt line .
this result gives more evidence in favor of a replica symmetry breaking in the very low temperature phase of the 3d ea model @xcite .
we present the data for different temperatures on a single plot in order to make more evident the fact that the numerical data seem to stay on the same curve once the system enters into the aging regime , i.e.when the points leave the fdt line .
this kind of behavior has been observed in the four - dimensional ea model @xcite and it is reminiscent of the mean - field solution .
indeed in the sk model , using the parisi - toulouse ( pat ) hypothesis @xcite , it can be shown @xcite that @xmath98 the formula can be easily generalized assuming a generic power law behavior in the aging regime : @xmath99 ( the mean - field value for the exponent is @xmath100 ) .
we use this generalization to fit the data and we obtain very good results .
the best fit parameters have been estimated from the collapse of the data reported in fig .
[ fdt ] and they are @xmath101 and @xmath102 ( to be compared with the mean - field values @xmath103 and @xmath100 , and those obtained for the 4d ea model @xmath104 and @xmath105 @xcite ) . in order to show the validity of the fitting formula , we present in fig .
[ scaling_fdt ] the collapse of the scaled data using the variables @xmath106 and @xmath107 , where @xmath108 .
it is easy to see that , if the previous scaling holds , the data should stay on two power laws : @xmath109 and @xmath110 in the quasi - equilibrium and aging regime , respectively .
the two power laws are reported in fig .
[ scaling_fdt ] .
even if we may expect a breakdown of the assumed scaling for large values of the scaling variable @xmath21 ( i.e. the scaled data are no longer described by a power law ) , we note however that for a quite large range the collapse is very good and very well approximated by a power law .
moreover , we remark that the collapse has been obtained adjusting only one parameter .
we have studied the off equilibrium dynamics of the three dimensional gaussian spin glass in the very low temperature phase . in particular
we have studied the scaling properties of the dynamical overlap correlation functions and the scaling properties of the violation of the fluctuation - dissipation .
we have tried to fit our correlation functions to the functional form predicted by the droplet model but the fits were poor . moreover
a correlation length diverging following a power law with the time implies , as was noted by rieger @xcite , barriers diverging not as @xmath111 ( as predicted by the droplet model with the lower bound @xmath112 ) but as @xmath113 .
this latter results implies @xmath114 , hence violating the droplet lower bound .
it is interesting to note that the experimental data could be fitted to the droplet formula assuming that @xmath115 @xcite .
however while both , the results of numerical simulations and the experiments are in very good agreement with a power law fit for @xmath116 , the numerical fit assuming a droplet formula for @xmath116 @xcite disagrees with the experimental fit assuming the same hypothesis @xcite . as it has been noted above , our final result for the dynamical correlation length is in a very good agreement with the experimental result .
we remark that the same scenario ( power law dependence of @xmath116 and linear dependence of @xmath117 with temperature ) also emerges in four and six dimensions . in the latter case
it was found that @xmath118 @xcite ( @xmath119 at the transition is the value predicted by mean field ) while in the former one @xmath120 @xcite .
moreover in these two dimensions the overlap correlation function constrained to zero overlap follows a pure power law as in three dimensions . if we send to infinity the time in our ansatz for the overlap - overlap correlation function we obtain a pure power decay @xmath121 with @xmath122 , with a small dependence of @xmath46 on the temperature for the whole spin glass phase .
we recall again that the droplet prediction is @xmath123 in contradiction with our numerical correlation functions ( this fact was already noted in @xcite ) . instead
, the pure power behavior is supported by the gaussian approximation using the mean field solution @xcite .
one could argue that the simulated temperatures are not low enough and the times and sizes not large enough in order to see the `` true '' ( droplet ) behavior of the ea model .
however , as we stressed in the introduction , @xmath15 is large enough for temperatures as low as @xmath4 and @xmath3 .
moreover our large times extrapolations are very safe thanks to the measurements have been taken over 6 time decades .
we have shown numerical results that contradict the droplet predictions in a wide range of temperatures ( @xmath124 ) .
in particular we point out that our results ( both for correlation functions and for violation of fdt ) at a very low temperature , @xmath4 , support a mean field picture .
finally , we remark that using the ( pat ) mean field scaling relations for the @xmath125 @xcite we have obtained a very good scaling plot of the violation of fluctuation - dissipation ( like in four dimensions @xcite ) .
this provides us another strong evidence calling for a low temperature phase being well described by mean field @xcite .
jjrl is partially supported by cicyt aen97 - 1693 .
jjrl wishes to thank l.a .
frnandez and a. muoz sudupe for interesting suggestions .
moreover we wish to thank a. j. bray for pointing out an arithmetic error in equation ( [ final ] ) .
w. l. mcmillan , j. phys .
c * 17 * , 3179 ( 1984 ) .
a. j. bray , m. a. moore , in _ heidelberg colloquium on glassy dynamics _ , lecture notes in physics * 275 * , j. l. van hemmen and i. morgenstern ( ed .
springer , berlin ) .
d. s. fisher , d. a. huse , phys .
lett . * 56 * , 1601 ( 1986 ) and phys . rev .
b * 38 * 386 ( 1988 ) .
|
we present a high statistic systematic study of the overlap correlation function well below the critical temperature in the three dimensional gaussian spin glass .
the off - equilibrium correlation function has been studied confirming the power law behavior for the dynamical correlation length . in particular
we have computed the dynamical critical exponent @xmath0 in a wide range of temperatures , @xmath1 , obtaining a dependence @xmath2 in a very good agreement with recent experiments .
moreover , we report a study of the violation of the fluctuation - dissipation theorem for very low temperatures @xmath3 and @xmath4 .
all our numerical results avoid a droplet model interpretation even when @xmath5 is as low as @xmath4 .
|
molecule of the dna has unusual elastic properties owing to its helicoidal symmetry and the sequence of base - pairs . from the physical view - point , i.e. neglecting its genetic information , the latter looks random .
the vibrational dynamics of the dna has an important bearing upon biological phenomenae in cell , @xcite , see also@xcite .
a specific feature of the dynamics is localized motions in the duplex which spread only over several base - pairs .
their existence has been confirmed by the experimental research using the raman scattering , @xcite,@xcite , @xcite , @xcite , the far - infrared absorption , @xcite,@xcite , and the brillouin scattering @xcite . in paper
@xcite the vibrational modes have been studied using the submillimeter - wave absorption spectroscopy in the range of @xmath0 thz .
thus , woolard et al , @xcite , have found multiple dielectric resonances in the long - wavelength portion of the submillimeter - wave regime , i.e. @xmath1 , which they ascribe to phonon modes of the dna .
these results provide valuable methods of bio - diagnostics , @xcite . theoretical study of phonon modes of the dna had revealed elastic excitations of the duplex which may correspond to the approximate helicoidal symmetry of a molecule of the dna , and its random elastic structure , @xcite , @xcite .
it is important that phonon modes of the dna are believed to be strongly attenuated owing to an interaction with ambient medium .
the effect can be mitigated , to some extent , by preparing samples of appropriate character .
thus , films formed by molecules of the dna appears to be less prone to the attenuation as regards phonon modes .
in contrast , it is especially strong in liquid solutions of the dna .
it should be noted that the estimates are based on the navier - stokes hydrodynamics in _
sub_-ghz range , even more so , the classical stokes formula for the viscous drag at small reynolds number . but
the above arguments are wrong in the ghz range .
in fact , van zandt , @xcite , showed that within the framework of the maxwell hydrodynamics , there could exist underdamped phonon modes .
thus , whether the phonon modes are damped or not so damped , remains to be seen . in this paper
we follow the analysis performed by chia c.shih and s.georghiou who have put forward powerful arguments in favour of the existence of underdamped vibrational modes of a molecule of the dna , besides the overdamped ones @xcite .
the authors visualize a molecule of the dna as a duplex of two strands formed by sugar - phosphate backbones framed by base - pairs located inside the strands , and assume that the bases are shielded by sugar
phosphates from the bombardment by molecules of solvent , the base - pairs being influenced by the ambient medium indirectly , through their interaction with the backbone .
the motion of the bases has the shape of librations inside the cages formed by the sugars of the backbone .
consequently , the dynamics of the backbone is damped owing to the strong attenuation caused by the medium , whereas that of the base - pairs turns out to be underdamped .
their angular frequencies are insensitive to the viscosity and lie in the low range of the raman spectrum .
thus , the backbone modulates the motion of base - pairs in accord with the environment medium .
we aim at finding effects of resonance attenuation of hypersound propagating in a sample of molecules of the dna . to that end
one may employ the interaction between solvent and molecules of the dna to generate phonon modes of the dna by pumping ghz - excitations in the liquid .
the strong viscous interactions of the molecules and the solvent may result in a dragging that could promote torsional phonon modes generating the interstrand ones , and result in the additional absorption of hypersound .
thus , the study of solutions of the dna could provide important information as to the nature of the liquid state , precisely through taking into account effects of dissipation in solvent in the ghz range .
it is worth noting that hypersound acoustics has made considerable progress in recent years and has become a powerful method in experimental research , @xcite , @xcite . in the context of the dna
it may turn out to be a valuable means for studying hydrodynamic phenomenae .
in this paper we are concerned with the phonon modes of the dna . to that end
we employ the theoretical model worked out in our earlier paper @xcite , in which we follow the guidelines cast by h.kappellmann and w.biem , @xcite . for the convenience of the reader we recall the main points of paper @xcite . in considering the dynamics of the dna
one has to take into account : 1 . the dna having the two strands ; + 2 .
the base - pairs being linked by the hydrogen bonds ; + 3 . the helical symmetry .
we shall utilize a one - dimensional lattice model of the dna which accommodates these requirements , in agreement with paper @xcite .
we follow the scheme introduced by el hasan and calladine , @xcite .
aiming at a qualitative description of the dna dynamics we use a simplified set of variables , and describe the relative position of the bases of a base - pair by means of the vector @xmath2 ; which is equal to zero when the base - pair is at equilibrium , see figs.[fig : pairs ] , [ fig : model ] .
the relative positions of different base - pairs are described by torsional angles @xmath3 of the sugar - phosphate backbone .
they correspond to deviations from the standard equilibrium twist of the double helix , so that a twist of the dna molecule , which does not involve inter - strand motion or mutual displacements of the bases inside the pairs , is determined by the torsional angles @xmath3 of rotation of the base - pairs about the axis of the double - helix .
thus , at equilibrium , a base - pair can be transformed into the next one by rotation through the pitch angle @xmath4 and simultaneous translation through a distance of neighbouring base - pairs .
this picture of the conformation of the dna is in full agreement with that of paper @xcite .
we suppose that the size of dna molecule is small enough that it can be visualized as a straight double helix , that is not larger than the persistence length .
hence the number of base - pairs , @xmath5 , approximately .
consequently , the twist energy of the molecule is given by the equation @xmath6\ ] ] in which @xmath7 is the moment of inertia of the @xmath8th base - pair , and @xmath9 are the twist coefficients , which may change from one base - pair to another .
interstrand motions should correspond to the relative motion , or libration in terms of paper @xcite , of the bases inside the base - pairs , therefore the kinetic energy due to this degree of freedom may be cast in the form @xmath10 where @xmath11 is the effective mass of the @xmath8th base - pair . for each base - pair
we have the reference frame in which z - axis corresponds to the axis of the double helix , y - axis to the long axis of the base - pair , x - axis perpendicular to z- and y- axes ( see fig . 1 of paper @xcite ) . at equilibrium the change in position of neighbouring base - pairs is determined only by the twist angle @xmath4 of the double helix .
we shall assume @xmath12 as for the b - form of dna . to determine the energy due to the inter - strand displacements we need to find the strain taking into account the helical structure of our system .
aiming at a simplified picture of inter - strand motion , we confine ourselves to the torsional degrees of freedom of the double lattice and assume the vectors @xmath13 being parallel to x - y plane , or two - dimensional .
consider the displacements @xmath14 determined within the frames of the two consecutive base - pairs , n , n+1 . since we must compare the two vectors in the same frame
, we shall rotate the vector @xmath15 to the frame of the n - th base pair , @xmath16 here @xmath17 is the inverse matrix of the rotation of the n - th frame to the ( n+1)-one given by the equation @xmath18 \label{rot}\ ] ] the matrix @xmath19 is 2 by 2 since the vectors @xmath13 are effectively two - dimensional
. then the strain caused by the displacements of the base - pairs is determined by the difference @xmath20 it is important that the angle @xmath21 is given by the twist angle , @xmath4 , describing the double helix , in conjunction with the torsional angles @xmath3 , so that @xmath22 therefore , the energy due to the _ inter - strand _
stress reads @xmath23 ^ 2 \right \}\ ] ] where @xmath24 are elastic constants determining the librational motion of base - pairs .
it is important that they may change from one base - pair to another .
the equilibrium position of the double helix is the twisted one determined by @xmath4 , with all @xmath3 being equal to zero . combining the formulae given above
we may write down the total energy of the dna molecule in the form @xmath25 \nonumber \\ & + & \sum_n \left\ { \frac{m_n}{2}\ , \dot{\vec y_n}^2 + \displaystyle{\frac{k_n}{2a^2 } } \ , \left [ r^{-1}(\omega + \varphi_{n+1 } - \varphi _ n)\ , \vec y_{n+1 } - \vec y_n \right]^2 + \frac{\epsilon_n}{2}\ , \vec y_n^2 \right \ } \label{main}\end{aligned}\ ] ] in which @xmath24 and @xmath26 are the torsional elastic constants and the inter - pairs distance , correspondingly . in summations given above n is the number of a site corresponding to the n - th base - pair , and @xmath27 , @xmath28 being the number of pairs in the segment of the dna under consideration .
the last term , @xmath29 accommodates the energy of the inter - strand _ separation _ due to the _ slides of the bases inside the base - pairs_. thus , we have the model of a molecule of the dna which may be visualized as @xmath30dimensional one , in the sense that it is one dimensional , from the formal point of view , and at the same time _ accomodates librational motions of base - pairs _ into `` two transversal dimensions '' outside the axis of the duplex . this is due to @xmath13 being directed outside the axis of a molecule , while scalar angles @xmath3 describe the twist of the sugar - phosphate backbone .
our next step is to split the above @xmath31dimensional lattice into three interacting linear chains . to that end
we shall cast our variables in a complex form .
let us introduce complex quantities @xmath32 in accord with the equations @xmath33 then we may cast the expression for the energy , @xmath34 , in the form @xmath35 for the kinetic energy , and @xmath36 for the potential one . on using the substitution @xmath37
we may cast the above equations in the form @xmath38 and @xmath39 we cosider only small deviations from the equilibrium , and assume that difference @xmath40 is small .
therefore , we have the equation @xmath41 in what follows we shall neglect terms of the second order in @xmath42 . within the above approximation the lagrangian equations of motion have the form @xmath43 @xmath44 and @xmath45 being the real and imaginary parts of @xmath46 .
thus , we split the equations of motion into three parts , and cast the system in a form of three interacting one - dimensional lattices of variables @xmath47 .
this mathematical device is crucial for the subsequent analysis of the dynamics .
it is worth noting that , in real life , a molecule of the dna is not a perfect duplex , so that constants @xmath48 are , generally , random sequences . indeed , owing to different sets of base - pairs , the conformational parameters of a molecule of the dna , such as
angles determining positions of base - pairs , suffer considerable , upto tens of per cent , deviations from the regular positions , @xcite .
further , their being of the two types , adenine thymine and guanine cytosine , results in constants @xmath48 having different values for different pairs .
therefore , equations ( [ eqn : phi ] ) ( [ eqn : b ] ) describe the motion of irregular interacting lattices
. one can appreciate this circumstance according to ones lights .
for one thing , it results in formidable difficulties as far as the standard analytical treatment is concerned , for another , it is conducive to a phenomenon peculiar to molecules of the dna in real life , that is the formation of localized excitations . in this paper
we shall prefer the second option .
we feel that one can understand the phenomenon of localized excitations within the framework of the theory for dynamic excitations in random media worked out , years ago , by i.m.lifshitz and his colleagues , @xcite @xcite ( see also the excellent review @xcite ) . according to the lifshitz theory the dynamics of elastic excitations in random media , for example lattices of real crystals or amorphous systems , has two important properties .
firstly , there exists forbidden bands in the spectrum of elastic excitations , such as phonons . secondly , there are excitations localized in bounded regions of the medium , their characteristic frequencies being located at the edges of the forbidden zones . the key observation due to i.m .
lifshitz was that defects of media could serve centers for the localization of elastic excitations , see @xcite @xcite . in his first paper on the subject he considered defects caused by an atom of isotope that has a mass different from the regular one , and showed that the defect could generate localized vibrations at its side .
later , the approach was further extended to defects of a more general kind and higher dimensions , see @xcite , for example plains of defects , and also boundaries of crystals .
the study of spectra for disordered , or random , lattices has followed mainly two paths .
a number of papers employed analytical methods , which generally could provide only qualitative description of the spectra , see @xcite .
it was p.dean , @xcite , who developed a powerful computer technic for analysis situs of the frequency distribution of disordered lattices .
the results by p.dean comply with the lifshitz theory , and provide a powerful insight into a specific structure of spectra .
phonon modes may result in the excessive absorption of hypersound in films and solutions of the dna .
in fact , the passage of an acoustic wave may promote the transfer of molecules from the equilibrium state to a state in which phonon modes of molecules of the dna are excited .
the time delays in this process and its reversal should lead to a relaxational dissipation of acoustic energy , and an absorption of hypersound at certain resonance frequencies corresponding to frequencies of the phonon modes . equally important , hypersound irradiation of molecules of the dna could make for generatung phonon modes .
both the experimental data , @xcite @xcite , @xcite , and the theoretical arguments , @xcite , @xcite , @xcite , indicate that the frequencies of phonons of the dna are in ghz range , and therefore one may expect resonance interaction between hypersound waves and phonons of the dna .
the availability of hypersound transducers , see @xcite , @xcite , suggests that there are technical means so as to employ high resolution acoustical spectroscopy for studying phonon modes of the dna in ghz range .
the alleged attenuation requires a careful choice of right samples of the dna for studying phonon modes . by now films of the dna
are generally employed to that end .
perhaps , this could make for diminishing the attenuation effects mentioned above .
in fact , solutions of the dna are not the best proposition owing to the attenuation effects being large in this case .
but , any way , it should be very interesting to use liquid crystallin phases of the dna , as was done for inelastic x - ray scattering , see paper @xcite and references therein .
the attenuation of phonon modes is closely related to the problem of hydration of the dna .
it is alleged that there are two relaxational processes in a hydrated molecule of the dna due to the primary and the second hydration shell with relaxational times @xmath49 and @xmath50 sec .
the residence time for a molecule of water at grooves of a molecule of the dna is estimated as @xmath51 sec , @xcite .
phonon dynamics having characteristic times @xmath52 sec , there is a need for using the generalized hydrodynamics in ghz range , @xcite , to estimate , for example , the size of dissipation forces acting on a vibrating molecule .
presently , this problem is hard to solve , and we have got only the information on the dispersion and attenuation of sound waves in this range , required for the theoretical treatment of the mandelstam brillouin light scattering .
but , curious enough , one may expect that the strong attenuation of some phonon modes could bring about the generation of others because of the attenuation due to the hydration shell mentioned above .
the latter primarily involve only external regions of the duplex , or according to paper @xcite does not directly affect the librations of base - pairs .
hence , we may suggest that an acoustic wave could drag the molecule , promote its rotational motion , and generate librations of base - pairs owing to the internal dynamics of the duplex .
we may infer from equations ( [ eqn : phi ] ) ( [ eqn : b ] ) that our model allows for the process .
the problem is to write down a reasonable force of molecular - liquid interaction . for the convenience of numerical simulation
it is worthwhile to choose appropriate scales for length , mass , and time , which agree with the conformational structure of the dna .
we shall take the following quantities as the dna units * @xmath53 as unit of mass , by the order of magnitude close to the mass of a base - pair ; + * @xmath54 as unit of length , close to the distance between neighbouring base - pairs ; + * @xmath55 as unit of time ; corresponding to the upper edge of phonon frequencies of the dna .
it is important that the sound velocities , @xmath56 , of a molecule of the dna are of order @xmath57 , according to various experimental and theoretical estimates , @xcite , @xcite . in the units indicated above we have therefore @xmath58 .
the moment of inertia , @xmath7 , of a base - pair is equal by orders of magnitude @xmath59 , where @xmath19 is the radius of a molecule of the dna , that is @xmath60 .
therefore , we have @xmath61 in the dna units introduced above . to obtain numerical estimates for the constants
@xmath62 we shall require that the values of sound velocities , @xmath56 , and frequencies , @xmath63 , of librations inside base - pairs , be of the orders of magnitude @xmath64 and @xmath65 , respectively .
since @xmath66 and @xmath67 , we have @xmath68 and @xmath69 , respectively . for the libration frequency
we have @xmath70 , if we assume that it be of order @xmath71 . for the viscosity coefficient of water we have , accordingly , @xmath72 .
summarizing , we have the following characteristic quantities expressed in the dna units @xmath73 as was mentioned above , the hydrodynamical theory , presently , does not provide reliable theoretical instruments for studying the interaction between a molecule and solvent , and one should turn to a rule of thumb to find a solution .
the interaction could be small enough , and therefore proportional to the velocity of a molecule .
confining us to the picture given by paper @xcite , we may suggest that the viscous force imposed on the @xmath8-th cite of the molecule should have the form @xmath74 where @xmath75 is a dissipative constant .
the size of @xmath75 is difficult to assess .
estimates based on the conventional picture of liquid motion , that is the navier
stokes one , are apparently far from reality because of small characteristic ( `` nano '' ) times , and the specific structure of water close to the molecule ( see @xcite ) , so that there is a problem as to whether it is reasonable to employ the usual hydrodynamic viscosity coefficient in this situation .
if we assume that it is possible , the dimension analysis gives for the moment of viscousity at site n @xmath76 or @xmath77 in the dna units introduced above @xmath78 so that we have @xmath79 .
in choosing the drag due to the action of a sound wave on the molecule we have to take into account that the wavelength of a hypersound wave in ghz region is by several orders of magnitude larger than the molecular segment .
the latter being of a few hundred , the force depends only on time .
we assume also that the torsion of molecule due to the force is small , and may take it in the form @xmath80 the main point is to determine the nature of the force and to assess its size .
if we confine ourselves to the _ macroscopical _ navier stokes hydrodynamics , there are two different pictures surfacing .
first , there could be an effect similar to that of the rayleigh disc , @xcite .
to see the point let us neglect the dissipation and attribute the force to the streamlines of flow round a molecule .
there is the turning
moment @xmath81 experienced by an oblate ellipsoid , or disk , in a vibrating fluid , @xcite . @xmath82
the dimension analysis shows that there is the general formula for the turning
moment @xmath83 where @xmath84 is the characteristic length of a body , @xmath85 velocity and @xmath86 the density of the fluid . in the case we are considering
, we may suggest that the rotation angle is small so as to employ @xmath42 instead of its sinus . since the coefficient at @xmath87 is the product @xmath88
, we obtain the expression @xmath89 now we are in a position to assess the value of @xmath90 as regards the energy pumped in a sample .
let us recall that the density @xmath91 of the energy current is given by the equation @xmath92 , c being the velocity of sound in liquid .
we are interested only in rough estimates .
therefore we assume @xmath93 . for the energy current @xmath94 we obtain @xmath95 , or @xmath96 in the dna units . in the equation given above @xmath19
is the radius of the molecule , that is @xmath97 , or @xmath98 in the dna units .
consequently , we have @xmath99 .
compared with the values of @xmath100 , it is too small to produce any appreciable effect .
in fact , even for @xmath101 we obtain only @xmath102 , which is also too small .
therefore , the effect of the rayleigh disk could not have any bearing on phonon modes of the dna .
but it is worth noting that the above estimate is due to the use of the laws of _ macroscopical hydrodynamics _ , whereas there are no definite conclusions as to their validity on _ microscopical scale_. the second option is provided by the viscous interaction of the solvent with the molecule .
we are looking for an analogue of the familiar stokes force " @xmath103 .
using the dimensional analysis we obtain the expression @xmath104 for @xmath105 it provides the estimate @xmath102 , or @xmath106 for @xmath107 .
the above estimates apparently preclude any opportunity for the observation of phonon modes of the dna by means of hypersound pumping .
similar arguments , @xcite , are generally put forward as regards phonon modes studied by the sub - millimeter absorption spectroscopy .
but nonetheless the sub thz - phonon modes are observed , see @xcite and the references therein .
we feel that this discrepancy between theory and experiment is likely to be due to the use of conventional hydrodynamics in the region where it does not work properly .
in particular , it results in the values of the coefficients @xmath90 and @xmath75 in equations ( [ eqn : phi])([eqn : b ] ) that precludes the existence of phonon modes .
therefore , in what follows we consider consequencies of @xmath90 and @xmath75 being outside the range prescribed by the navier stokes hydrodynamics , and choose appropriate values that permit the existence of phonon modes .
the equations of motion are obtained by inserting the forces ( [ eqn : viscforce ] ) , ( [ eqn : drag ] ) in equation([eqn : phi ] ) .
the equations are nonlinear and require numerical simulation for their studying . for the numerical simulation of the above equations we have used the algorithms of verlet , leapfrog , and the explicit and implicit adams algorithms .
we have systematically made comparisons between results provided by different algorithms so as to escape possible mistakes and numerical artifacts .
we have taken the integration step @xmath108 , or @xmath109 for verification , the unit of time being the period of external force generated by sound pumping .. we are going to simulate the generation of phonon modes by pumping hypersound . according to the equations of motion it is allowed owing to the term given by equation ( [ eqn : drag ] ) .
thus , we shall generate firstly a @xmath42-mode , and the latter shall promote the @xmath44- and the @xmath45-ones , according to equations ( [ eqn : a ] ) and ( [ eqn : b ] ) . as far as these modes are concerned , we have a parametric excitation through the interaction terms in equations ( [ eqn : a]),([eqn : b ] ) , and , as is shown below , there may emerge a parametric resonance at a certain frequency @xmath110 of the excitation pulse . here
it should be noted that the usual theory of parametric resonance for the harmonic oscillator , @xcite , can not be directly employed because the equations ( [ eqn : a ] ) ( [ eqn : b ] ) of motion for the interacting @xmath111 chains are nonlinear , and we need to find the resonance frequencies by `` trial and error '' , lifshitz s theory providing at this point general recommendations .
it is important that there exists the threshold @xmath112 that depends on the dissipative coonstant @xmath75 , the resonance taking place only for amplitudes of acoustic field which are large enough , @xmath113 .
if the frequency of the acoustic wave is very close to the resonance one , @xmath112 depends linearly on the dissipative constant @xmath75 , see fig.[fig : dissipation ] , and for small @xmath75 , @xmath114 is also small , so that there is no threshold value at @xmath115 and @xmath116 .
if the excitation frequency appreciably deviates from the resonance one , there is a threshold for @xmath112 at @xmath115 , see fig.[fig : threshold ] .
thus , the numerical simulation of equations ( [ eqn : phi ] ) ( [ eqn : b ] ) is in agreement with the general theory of parametric resonance .
localization of a phonon mode excited by the resonant acoustic wave is illustrated in fig.[fig : localization ] for zero - dissipation case @xmath115 , in fig.[fig : loc01 ] for the dissipative constant @xmath117 , and in fig.[fig : loc02 ] for @xmath118 .
our results thus in agreement with van zandt and saxena , @xcite , who predicted the phonons in the submillimiter range corresponding to localized excitations spread over several base pairs .
we feel that the acoustic spectroscopy could be helpful in analyzing the conformational structure of the dna through the analysis of its phonon modes .
the model we have constructed qualitatively agree with the experimental data by providing the existence of the phonon modes and the reasonable orders of magnitude for their frequencies .
it implies a phonon localization due to the random structure of the duplex , and shows that it would be worthwhile to study the action of hypersound on molecules of the dna .
it is important that absorption peaks for hypersound could be expected at frequencies corresponding to the parametric resonance of phonon modes of the dna under hypersound pumping . in this respect
the parametric resonance could be a valuable means for studying the vibrational modes of the dna .
there still remains a problem of accommodating the dissipative effects .
the standard navier stokes model promotes a size of dissipation that exclude the existence of phonon modes , but they are observed in real life , @xcite . the circumstance challenges theoreticians to working out an adeqate framework for the hydrodynamics at molecular scale .
the part played by the dissipation , perhaps different from the navier
stokes one , could be crucial , and torsional modes of the backbone could make for understanding its nature on the nano - scale .
hypersound acoustic spectroscopy could be instrumental in this respect . to that end liquid samples
may turn out to be more interesting , solutions of the dna providing a probe into the physics of liquid state .
the observation of the parametric resonance we have indicated could serve a means for understanding the hydrodynamics on molecular scale .
99 w.n.mei , m.kohli , , e.m.prohofsky , and l.l.van zandt , biopolymers * 20 * , 833 ( 1981 ) .
jing ju , thesis , stevens institute of technology ( 2001 ) .
h.urabe , y.tominaga , and k.kubatta , j.chem.phys . * 78 * , 5937 ( 1983 ) .
y.tominaga , m.shida , k.kubota , h.urabe , y.nishimura , and m.tsuboi , j.chem.phys . * 83 * , 5972 ( 1985 ) .
t.weidlich , s.m.lindsay , r.qi , a.rupprecht , w.l.peticolas , and g.a.thomas , j.biomol.struc.dynam .
* 8 * , 139 ( 1990 ) .
m. krisch,1 a.mermet , h.grimm , v.t.forsyth , and a.rupprecht , phys.rev .
* e73 * , 061909 ( 2006 ) .
c.liu , g.s.edwards , s.morgan , and e.silberman , phys.rev . * a 40 * , 7394 ( 1989 ) .
w.k.scroll , v.v.prabhu , e.w.prohofsky , and l.l.van zandt , biopolymers * 28 * , 1189 ( 1989 ) .
j.w.powell , g.s.edwards , l.genzel , f.kremer , a.wittlin , w.kubasek,and a.peticolas , phys.rev . * a 35 * , 3929 ( 1987 ) .
s.m.lindsay and j.powell , structure and dynamics : nucleic acids and proteins , eds . e.clementi and r.sarma , adenine ( 1983 ) .
d.l.woolard , t.r.globus , b.l.gelmont , m.bykhovskaia , a.c.samuels , d.cookmeyer , j.l.hesler , t.w.crowe , j.o.jensen , j.l.jensen , and w.r.loerop , phys.rev . * e65 * , 051903 ( 2002 ) d.l.woolard , t.koscica , d.l.rhodes , h.l.cui , r.a.pastore , j.o.jensen , j.l.jensen , w.r.loerop , r.h.jacobsen , d.mittleman , and m.c.nuss , j.apll.toxicology , * 17*(4 ) , 243 ( 1997 ) .
r.k.adair , biophys.j . , * 82 * , 1147 ( 2002 ) .
b.f.putnam , l.l .
van zandt , e.w.prohofsky , and w.n.mei , biophysical journal * 35 * , 271 ( 1981 ) .
v.k.saxena and l.l.van zandt , biophysical journal , * 67 * , 2448 ( 1994 ) .
l.l.van zandt , phys.rev.lett . *
57 * , 2085 ( 1986 ) . l. l. van zandt and v. k. saxena , in structure & functions , volume 1 : nucleic acids edited by r. h. sarma and m. h. sarma adenine , new york , 1992. chia c.shih and s.georghiou , j.biomol.structure.dynamics , * 17 * , 921 ( 2000 ) .
a.huynh , b.perrin , n.d.lanzilotti-kimura , b.jusserand , a.fainstein , and a.lemtre , arxiv : 0803.2090v1 [ cond-mat.other ] 14mar 2008 ; a. huynh , n. d. lanzillotti - kimura , b. jusserand , b. perrin , a. fainstein , m.f .
pascual - winter , e. peronne , and a. lemaitre , phys .
. lett . * 97 * , 115502 ( 2006 ) .
b.j.linde , molecular and quantum acoustics , * 27 * , 169 ( 2006 ) .
v.l.golo , jetp * 128 * , 428 ( 2005 ) , the three - wave interaction between inter strand modes of the dna . h.capellmann and w.biem , z.phys . *
209 * , ( 1968 ) .
m.e.el hassan and c.r.calladine , j.mol.biol .
* 251 * , 648 ( 1995 ) .
hunter , j.mol.biol .
, * 230 * , 1025 ( 1993 ) .
ja.i.frenkel , kinetic theory of liquids , moscow , nauka ( 1945 ) .
v.k.saxena and l.l.van zandt , phys.rev . * a 42 * , 4993 ( 1990 ) .
lifshitz , s.a.gredeskul , and l.a.pastur , introduction into the theory of random media , ?
nauka ? , ( 1982 ) . i.m.lifshitz and l.n .
rosenzweig , izv.ac.nauk ussr ser.fyz .
* 12 * , 667 ( 1948 ) .
i.m.lifshitz , journ . of phys .
( ussr)7 , 211 , 249 ( 1943 ) .
i.m.lifshitz , zhetf , 17 , 1017 , 1076 ( 1947 ) .
i.m.lifshitz , nuovo cimento suppl . , 10 , 3 , 716 ( 1956 ) .
maradudin , e.w .
montroll , and g.h .
weiss , theory of lattice dynamics in the harmonic approximation , academic press , new york and london ( 1963 ) .
ziman , models of disorder , ch.8 , cambridge university press , cambridge ( 1979 ) .
p.dean , phys.rev .
* 44 * , 127168 ( 1972 ) .
j.w.rayleigh , the theory of sound , vol . 1 , ch.3 , new york , dover publications ( 1926 ) .
w.knig , ann.phys . , lpz . , * 42 * , 549 ( 1891 ) ; * 43 * , 43 ( 1891 ) .
f.merzel , f. fontaine - vive , m. r. johnson , and g. j. kearley , phys.rev .
* e76 * , 031917 ( 2007 ) .
m.krisch , a.mermet , h.grimm , v.t.forsyth , and a.rupprecht , phys.rev .
* e73 * , ( 2006 ) .
samir kumal pal , liang zhao , and ahmed h.zewail , pnas * 100 * , 8113 ( 2003 ) . phys.rev.lett . ,
* 88 * , 156101 ( 2002 ) .
|
we find a parametric resonance in the ghz range of the dna dynamics , generated by pumping hypersound .
there are localized phonon modes caused by the random structure of elastic modulii due to the sequence of base pairs .
|
the results presented here are based on a data sample collected with the belle detector @xcite at the kekb asymmetric - energy @xmath3 collider @xcite .
the experiment is located at the high energy accelerator research organization ( kek ) in tsukuba , japan .
electrons and positrons are being collided at the center - of - mass ( cm ) energy of about 10.58 gev , corresponding to the mass of the @xmath4 resonance .
the main objective of the belle experiment is a study of the @xmath5 violation in the @xmath6 meson system produced from @xmath4 , however a large data sample collected with the belle detector and its excellent performance also make possible to perfrom searches for new hadronic states as well as studies of their properties .
there are several possible mechanisms of the particle production at @xmath6 factories : production in the @xmath6 meson decays , fragmentation of quarks in @xmath3 annihilation or two - photon collisions .
most of the newly discovered states , the so - called @xmath0 mesons , almost certainly contain a @xmath7 quark pair among their constituent particles but do not fit into any of the unassigned charmonium levels predicted within the potential models . as a result , they are considered to be possible candidates for new , exotic types of particles such as multiquark states that include either molecular states ( two loosely bound charm mesons ( @xmath8)(@xmath9 ) ) @xcite , or tetraquarks ( tightly bound four - quark states , [ @xmath8][@xmath9 ] ) @xcite , charmonium hybrids ( @xmath7-gluon with excited gluonic degrees of freedom ) @xcite , or hadro - charmonium ( compact charmonium states , @xmath12 , @xmath13 , @xmath14 , `` coated '' by excited light - hadron matter ) @xcite .
more conservative models @xcite suggest to reconsider the effect of the numerous open charm thresholds which are located in many cases near the new states on the parameters of the conventional charmonium levels . in this paper we review review recent results from belle collaboration on the @xmath15 , @xmath16 and @xmath17 charmoniumlike meson states as well as on their possible counterparts in @xmath1 and @xmath2 systems , summarized also in table [ tab_newstates ] .
[ cols= " < , < , < , < , < , < , < " , ] nonzero electric charge makes these three states by definition exotic and excludes the possibility to interpret them as a charmonium or @xmath7-gluon hybrid mesons .
they are prime candidates for a multiquark mesons .
it is therefore important that the belle results get confirmed by other experiments .
_ vs. _ @xmath18 dalitz plot distribution for candidate @xmath19 events .
( bottom ) the @xmath20 dalitz plot projection for events from the @xmath21 gev@xmath22 ( second vertical band shown in top plot ) .
, title="fig:",scaledwidth=49.0% ] _ vs. _ @xmath18 dalitz plot distribution for candidate @xmath19 events .
( bottom ) the @xmath20 dalitz plot projection for events from the @xmath21 gev@xmath22 ( second vertical band shown in top plot ) .
, title="fig:",scaledwidth=49.0% ]
isr has proven to be powerful tool to search for @xmath23 states at @xmath6-factories , since it allows to scan a broad energy range of @xmath24 below the initial @xmath3 cm energy , while the high luminosity compensates for the suppression due to the hard - photon emission .
first state , discovered with the isr technique by babar , was @xmath26 in the @xmath27 process @xcite . using
the same method belle confirmed the @xmath26 state and in addition found another resonant structure , called @xmath25 @xcite .
a fit using two interfering breit - wigner amplitudes to the @xmath28 invariant mass distribution describes the data better than a fit assuming one resonance , especially for the lower - mass side of the @xmath29 gev/@xmath30 enhancement ( see fig . [
figy1mm](a ) ) .
obtained masses and widths of the @xmath25 and @xmath26 are given in table [ tab_newstates ] .
the babar collaboration subsequently reported a similar structure in the cross section for the isr @xmath3 annihilation process resulting in the @xmath33 final state @xcite .
the mass and width of the state denoted as @xmath31 are both significantly higher than the values found for the @xmath26 .
belle performed a similar study on a larger data sample ( 673 fb@xmath34 for belle compared to 272 fb@xmath34 for babar ) and observed that the structure is formed from two narrower peaks @xcite ; one , in agreement with the babar study , is observed near 4.36 @xmath35 and another , called @xmath32 , near 4.66 @xmath35 .
fit to the @xmath33 distribution with a coherent sum of two breit - wigner amplitudes is shown in fig .
[ figy1mm](b ) . fitted masses and widths of the two states
are given in table [ tab_newstates ] .
no sign was found either of @xmath26 ( @xmath25 ) decay to @xmath33 , or of @xmath31 ( @xmath32 ) decay to @xmath36 .
the nature of @xmath16 states still reamins unclear .
there is only one unassigned @xmath23 charmonium level in this mass region , the @xmath37 level , which might accommodate the @xmath32 , however there are no other available charmonium level in the spectrum for all other peaks discussed above . in addition , these states , if considered to be conventional charmonium states , should decay mainly to @xmath38 , however measurements of cross sections for exclusive open - charm final states in this energy range , perfromed by belle @xcite , babar @xcite and cleo - c @xcite , shown no evidence for peaking near the masses of the @xmath16 states .
the one exception is @xmath40 , for which belle has reported a near - threshold enhancement , called the @xmath39 ( see fig . [ figy2 mm ] ) . the fitted mass and width
are given in table [ tab_newstates ] and are consistent within errors with the mass and width of the @xmath32 state .
combinations from @xmath3 annihilation with the isr photon .
the curves show the result of the fit ( solid line ) , threshold function ( dashed line ) and combinatorial background contribution ( dashed - dotted line ) .
, scaledwidth=49.0% ]
an interesting question is whether or not there exist any @xmath41 counterpart states in the @xmath1 and @xmath2 systems , predicted by many of the models proposed to explain the charmoniumlike @xmath0 states .
some recent results , discussed below , indicate that this may be the case . in the @xmath1 system
one possible candidate is @xmath42 , a @xmath23 state , first observed by babar in the isr process @xmath43
@xcite and later also confirmed by bes @xcite . at belle , we also measured the @xmath44 and @xmath45 cross sections for the isr processes @xmath46 and @xmath47 with cm energy ranging from 1.5 to 3.5 gev @xcite .
figure [ figyssbar ] shows observed @xmath44 and @xmath45 cross section distributions .
the former is fitted with two coherent breit - wigner functions and the latter with one breit - wigner function interfering with a non - resonant continuum function .
the mass and width of the high mass peak , corresponding to @xmath42 , are found to be @xmath48 mev and @xmath49 mev , which are consistent with the previous measurements .
first measurements of mass and width are reported for the low mass peak in the @xmath44 cross section distribution , which corresponds to the @xmath50 .
they are found to be @xmath51 mev and @xmath52 mev .
the widths of the @xmath50 and @xmath42 are found to be quite similar and both are at the 200 mev level .
this may suggest that the @xmath42 is an excited @xmath23 @xmath1 state .
since the @xmath53 is thought to have a large @xmath1 component , @xmath54 can be viewed as an open - flavor decay as opposed to @xmath55 , which is a hadronic transition .
studies of @xmath42 in other decay modes are needed to distinguish , whether @xmath42 is a conventional @xmath1 state or an @xmath56-quark counterpart of the @xmath26 .
cross section distribution with superimposed fit result with two coherent breit - wigner functions .
( b ) the @xmath57 cross section distribution with superimposed fit result with one resonance and a coherent non - resonant term .
, title="fig:",scaledwidth=49.0% ] cross section distribution with superimposed fit result with two coherent breit - wigner functions .
( b ) the @xmath57 cross section distribution with superimposed fit result with one resonance and a coherent non - resonant term . , title="fig:",scaledwidth=49.0% ] belle and babar measured the partial decay widths of the order of few kev for @xmath59 as well as @xmath60 @xcite and found that are similar to those from dipion transitions from the @xmath61 to the @xmath62 and @xmath63 @xcite
. the same measurement was also perfromed by belle on a data sample at the cm energy of 10.87 gev , corresponding to the @xmath64 , and found huge signals for @xmath65 , @xmath60 and @xmath66 ( see fig .
[ fig5strans ] ) . assuming the observed signal events are due solely to the @xmath64 resonance , than the corresponding partial widths
are found to be in the range ( 0.520.85 ) mev , more than two orders of magnitude larger than the corresponding partial widths for @xmath4 , @xmath61 and @xmath62 decays to @xmath65 .
a possible explanation is a @xmath2 counterpart to the @xmath26 , denoted as @xmath67 @xcite , which may overlap with the @xmath64 .
alternative explanations include a nonperturbative approach for the calculation of the decay widths of dipion transitions of heavy quarkonia @xcite , the presence of final state interactions @xcite , or the existence of a tetraquark intermediate state @xcite .
mass difference distributions for events with @xmath68 ( a ) and @xmath69 ( b ) .
vertical dashed lines indicate the expected @xmath70 values for the @xmath71 transitions .
, title="fig:",scaledwidth=49.5% ] mass difference distributions for events with @xmath68 ( a ) and @xmath69 ( b ) .
vertical dashed lines indicate the expected @xmath70 values for the @xmath71 transitions .
, title="fig:",scaledwidth=49.5% ] in order to distinguish these hypotheses belle performed a measurement of the energy dependence of the cross sections for @xmath72 ( @xmath73 ) at energies around 10.87 gev @xcite .
peaks are found in all three channels at around 10.899 gev .
the preliminary values for the peak mass and width , obtained by performing a fit with a common breit - wigner function to the measured @xmath74 cross section distribution ( shown in fig .
[ figyb ] ) , are found to be @xmath75 mev and @xmath76 mev .
a fit using the pdg resonance parameters for the @xmath64 and @xmath77 @xcite , shown in fig .
[ figyb](b ) , fails to describe the observed @xmath74 cross section .
( @xmath73 ) processes .
the results of the fits which are shown as smoothed curves with ( a ) free common mean and width and ( b ) with fixed mean and width to the pdg @xmath64 and @xmath77 parameters . , title="fig:",scaledwidth=49.5% ] ( @xmath73 ) processes
. the results of the fits which are shown as smoothed curves with ( a ) free common mean and width and ( b ) with fixed mean and width to the pdg @xmath64 and @xmath77 parameters .
, title="fig:",scaledwidth=49.5% ]
the belle experiment at the kekb collider provides an excellent environment for charmonium spectroscopy . as a result
, many new particles , summarized in table [ tab_newstates ] , have been discovered during the ten year operation of the belle detector .
we have reported on a new belle @xmath78 mass peak in @xmath79 .
the measured @xmath80 mass and width are in agreement with the babar s mass and width values for the @xmath81 resonance seen in @xmath82 decays .
the problem with the interpretation of this state as a conventional charmonium is its large partial decay width to @xmath83 .
the mass of the @xmath84 produced in @xmath85 and @xmath86 are found to be consistent within around 0.9 mev .
the mass of the @xmath84 decaying into @xmath87 final state is consistent with that from @xmath88 within uncertainties .
another feature of @xmath84 is that the non - resonant @xmath89 contribution in @xmath90 decay dominates over the resonant @xmath91 contribution , which is not the case for conventional charmonium states .
there are too many @xmath23 @xmath16 states and too few unassigned charmonium levels between 4.0 gev and 4.7 gev .
in addition measurements of cross sections for exclusive open - charm final states in this energy range show no evidence for peaking near the masses of the @xmath16 states , except for @xmath92 , which has a threshold peak near the @xmath32 peak mass . 99 a. abashian _ et al . _ [ belle collaboration ] , nucl .
instrum .
a * 479 * , 117 ( 2002 ) .
s. kurokawa and e. kitunani , nucl .
instrum .
a * 499 * , 1 ( 2003 ) and other papers in the volume .
m. b. voloshin and l. b. okun , jetp lett .
* 23 * , 333 ( 1976 ) .
m. bander , g. l. shaw and p. thomas , phys .
lett . * 36 * , 695 ( 1977 ) .
a. de rujula , h. georgi and s.l .
glashow , phys .
lett . * 38 * , 317 ( 1977 ) .
n. a. trnqvist , z. phys .
c * 61 * , 525 ( 1994 ) .
a. v. manohar and m.b .
wise , nucl .
b * 339 * , 17 ( 1993 ) .
n. a. trnqvist , phys .
b * 590 * , 209 ( 2004 ) .
f. e. close and p.r .
page , phys .
b * 578 * , 119 ( 2003 ) .
wong , phys .
c * 69 * , 055202 ( 2004 ) .
e. braaten and m. kusunoki , phys .
d * 69 * , 114012 ( 2004 ) .
e. s. swanson , phys .
b * 588 * , 189 ( 2004 ) .
t .- w . chiu and t .- h .
hsieh , phys .
b * 646*,95 ( 2007 ) .
i. bigi _ et al .
d * 72 * , 114016 ( 2005 ) .
l. maiani _ et al .
d * 72 * , 031502 ( 2005 ) .
l. maiani , f. piccinini , a. d. polosa and v. riquer , phys .
d * 71 * , 014028 ( 2005 ) [ arxiv : hep - ph/0412098 ] .
t. barnes , s. godfrey , phys . rev .
d * 69 * , 054008 ( 2004 ) .
eichten , k. lane , c. quigg , phys .
d * 69 * , 094019 ( 2004 ) . c. meng and k .-
chao , phys .
d * 75 * , 114002 ( 2007 ) .
j. l. rosner , phys .
d * 74 * 076006 ( 2006 ) .
s. l. olsen [ belle collaboration ] , nucl .
a * 827 * , 53c ( 2009 ) [ arxiv:0901.2371 [ hep - ex ] ] .
k. abe _ et al .
_ , phys .
lett . * 98 * , 082001 ( 2007 ) [ arxiv : hep - ex/0507019 ] . k. abe _ et al .
_ [ belle collaboration ] , phys .
lett . * 94 * , 182002 ( 2005 ) [ arxiv : hep - ex/0408126 ] .
s. uehara _ et al . _ [ belle collaboration ] , phys .
lett . *
96 * , 082003 ( 2006 ) [ arxiv : hep - ex/0512035 ] .
p. pakhlov _ et al .
_ [ belle collaboration ] , phys .
lett . *
100 * , 202001 ( 2008 ) [ arxiv:0708.3812 [ hep - ex ] ] .
b. aubert _ et al .
_ [ babar collaboration ] , phys .
* 101 * , 082001 ( 2008 ) [ arxiv:0711.2047 [ hep - ex ] ] .
i. adachi _ et al . _ [ belle collaboration ] , arxiv:0810.0358 [ hep - ex ] .
s. uehara _
et al . _ [ babar collaboration ] , in preparation .
d. e. acosta _ et al .
_ [ cdf ii collaboration ] , phys . rev .
lett . * 93 * , 072001 ( 2004 ) [ arxiv : hep - ex/0312021 ] . v.
m. abazov _ et al . _
[ d0 collaboration ] , phys .
lett . * 93 * , 162002 ( 2004 ) [ arxiv : hep - ex/0405004 ] .
b. aubert _ et al .
_ [ babar collaboration ] , phys .
d * 71 * , 071103 ( 2005 ) [ arxiv : hep - ex/0406022 ]
. i. adachi _ et al . _
[ belle collaboration ] , arxiv:0809.1224 [ hep - ex ] .
t. aaltonen _ et al . _
[ cdf collaboration ] , arxiv:0906.5218 [ hep - ex ] . c. amsler _ et al . _
[ particle data group ] , phys .
b * 667 * , 1 ( 2008 ) .
g. gokhroo _ et al . _
( belle collaboration ) , phys .
lett . * 97 * , 162002 ( 2006 ) .
b. aubert _ et al . _ [ babar collaboration ] , phys .
d * 74 * , 091103 ( 2006 ) [ arxiv : hep - ex/0610018 ] .
m. ablikim _ et al . _
[ bes collaboration ] , phys .
* 100 * , 102003 ( 2008 ) [ arxiv:0712.1143 [ hep - ex ] ] .
c. p. shen _ et al .
_ [ belle collaboration ] , phys .
d * 80 * , 031101 ( 2009 ) [ arxiv:0808.0006 [ hep - ex ] ] .
b. aubert _ et al . _
[ babar collaboration ] , phys .
lett . * 96 * , 232001 ( 2006 ) [ arxiv : hep - ex/0604031 ] .
k. f. chen _ et al .
_ [ belle collaboration ] , phys .
lett . *
100 * , 112001 ( 2008 ) [ arxiv:0710.2577 [ hep - ex ] ] .
w. s. hou , phys .
d * 74 * , 017504 ( 2006 ) [ arxiv : hep - ph/0606016 ] .
a. simonov , jetp lett .
* 87 * , 121 ( 2008 ) [ arxiv:0712.2197 [ hep - ph ] ] . c. meng and k. t. chao , phys .
d * 78 * , 034022 ( 2008 ) [ arxiv:0805.0143 [ hep - ph ] ] .
m. karliner and h. j. lipkin , arxiv:0802.0649 [ hep - ph ] .
i. adachi _ et al . _
[ belle collaboration ] , arxiv:0808.2445 [ hep - ex ] .
|
we report on the most recent experimental progress on @xmath0 charmoniumlike meson states and their possible counterparts in the @xmath1 and @xmath2 systems .
|
at the end of 2002 japanese collaboration leps at spring-8 @xcite and the bubble chamber experiment diana at itep , moscow @xcite , announced evidence of a strange baryon , called @xmath4 , whose quantum numbers can not be constructed from 3 quarks .
this truly exotic state : ( @xmath5 ) has been anticipated already by the founders of the quark model , however it was believed to be rather heavy ( 1800 1900 mev ) and wide .
the excitement created by the findings of leps and diana was due to the low mass of the putative @xmath4 , of the order of 1530 mev , and a very small width .
such properties , however , are natural in chiral models where the antistrange quark is excited in form of a chiral field rather than as a constituent quark .
early predictions of the @xmath4 mass in different versions of the chiral models were very close to the experimental numbers of leps and diana @xcite , and moreover the width , as estimated by diakonov , petrov and polyakov in 1997 , was very small @xcite .
the discovery by leps and diana triggered both experimental and theoretical activity in a somewhat extinct field of hadron spectroscopy .
present experimental situation is , however , rather confusing .
many collaborations announced the existence of @xmath4 in different experimental setups , but the comparable number of searches ended with null result .
furthermore , dedicated high statistics runs with clas detector at cebaf did not show any signal of @xmath4 contradicting previous findings of the same collaboration @xcite . on the other hand leps has confirmed its original result @xcite .
also high statistics analysis by diana turned out to be positive @xcite .
we refer the reader to recent reviews of the experimental situation @xcite . for the purpose of the present study
we assume that @xmath4 exists with a mass equal 1540 mev and total width @xmath6 mev .
one of the immediate consequences of a possible existence of @xmath4 is the existence of the whole su(3 ) multiplet : @xmath7 ( see fig.[fig : multiplets ] ) .
indeed , na49 experiment at cern @xcite announced discovery of another exotic state , namely @xmath8 .
unfortunately the searches of @xmath9 by other groups have not confirmed the results of na49 .
apart from truly exotic states that can not be constructed from three quarks , antidecuplet contains cryptoexotic states that are primarily built from 5 quarks , however their quantum numbers can be constructed from three quarks as well .
these are nucleon - like states ( @xmath10 ) and @xmath11like states ( @xmath12 ) that are the subject of the present paper .
the interpretation of these states is not well understood : one may try to associate them with some known resonances , as it was done in the case of @xmath10 in the original paper of diakonov , petrov and polyakov for example @xcite , or one may postulate the existence of new , yet undiscovered resonances with nucleon or @xmath13 quantum numbers . in this paper
we follow the latter approach trying to predict the range of masses and widths for cryptoexotic @xmath12 and also @xmath9 states .
weight diagrams for octet and antidecuplet . states that can mix lie on dashed lines . ] in the chiral quark - soliton model ( @xmath14qsm ) the spin - parity quantum numbers of the antidecuplet members are unambiguously predicted to be @xmath15 @xcite , so that @xmath10 and @xmath16 are predicted to be @xmath17 resonances .
one of the striking properties of @xmath10 is that it can be excited by an electromagnetic probe from the neutron target much stronger than from the proton one @xcite .
the photoexcitation of charged isocomponent of @xmath10 is possible only due to su@xmath18 violation ; therefore its suppression by a factor @xmath19 in the amplitude is expected .
the existence of a new nucleon resonance with the mass near @xmath20 mev was suggested in refs .
the authors of ref .
@xcite used the gell - mann okubo mass relations in the presence of mixing , in order to predict the mass of this new nucleon resonance .
as an input for the gell - mann okubo mass formula the authors of ref .
@xcite used the mass of the @xmath21 baryon reported by the na49 collaboration @xcite . in ref .
@xcite , in order to constrain the mass of this possible new narrow @xmath22 , the modified pwa of @xmath23 scattering data was employed .
it was found that the easiest way to accommodate a narrow @xmath22 is to set its mass around @xmath24 mev and quantum numbers to @xmath17 ( @xmath25 ) . in the same paper
the width of the possible @xmath22 was analyzed in the framework of @xmath14qsm .
it was found that the width of new n@xmath26 is in the range of tens of mev with very small @xmath27 partial width of @xmath28 mev @xcite .
one should stress that the decay to @xmath27 is not suppressed in the su@xmath18 limit and it can be made small only if the symmetry violation is taken into account .
it follows that the preferred decay channels are @xmath29 , @xmath30 and @xmath31 @xcite .
the search of the new nucleon has been performed in the @xmath32 photoproduction on the neutron at graal @xcite .
the narrow peak in the quasi - free neutron cross section and in the @xmath33 invariant mass spectrum has been observed . the original observation of refs .
@xcite has been recently confirmed by two other groups : cbelsa / taps @xcite and lns - sendai @xcite .
all three experiments found an enhancement in the quasi - free cross - section on the neutron .
in addition , the graal and cbelsa / taps groups have observed a narrow peak in the @xmath33 invariant mass spectrum at @xmath34 mev .
the corresponding data can be explained by existence of a new narrow resonance @xcite .
one should , however , stress that the part of above mentioned experimental results may have a different interpretation that does not require to postulate a new narrow nucleon resonance @xcite . further evidence for a new narrow nucleon resonance came from the analysis of the @xmath13 beam asymmetry in @xmath32 photoproduction on the proton @xcite . in these papers the narrow structure in the @xmath13 beam asymmetry around invariant mass @xmath35 mev has been observed . that structure can be interpreted as the contribution of a narrow nucleon resonance with the mass 1685 mev , total width @xmath36 mev and the photocoupling to the proton much smaller than to the neutron ; the properties that are expected for the nonstrange partner of @xmath37@xcite .
the properties of possible new narrow nucleon resonance that crystalized out recent experiments on @xmath32 photoproduction are summarized in ref . @xcite . throughout this paper
we shall assume that the new @xmath38 nucleon resonance exists with total width below 25 mev .
one of the striking and to some extent counterintuitive properties of @xmath4 is its small width .
in particular , diana @xcite that has doubled the statistics of their formation experiment @xmath39 as compared to the original result reported in ref.@xcite , claims @xmath40 and the width @xmath41 ( with possible systematic uncertainties ) .
the only other available formation experiment with the secondary kaon beam at belle sets an upper limit @xmath42 ( at a 90% confidence level ) @xcite which is beyond the above value .
also the reanalysis of the old @xmath43 scattering data @xcite shows that there is room for the exotic resonance with a width below 1 mev .
the small width implies that the coupling @xmath44 is at least an order of magnitude smaller than @xmath45 .
the small value of @xmath44 appears naturally in a relativistic field - theoretic approach to baryons , allowing for a consistent account for multi - quark components in baryons ; in particular in refs .
@xcite an upper bound @xmath46 has been obtained without any parameter fixing .
recent calculation of the @xmath37 width in @xmath14qsm @xcite also gave small width of @xmath47 mev .
the width below 1 mev also comes out from the parameter - free qcd sum rules analysis @xcite . in any case
the small value of @xmath44 is related to the small value of @xmath48 reduced matrix element that is responsible for the direct decay @xmath49 .
indeed@xmath50 where @xmath51 is the mixing angle between the nucleon - like states in octet and antidecuplet , and @xmath52 denotes the transition reduced matrix element @xmath53 ( see tab .
[ tablematel ] for definitions ) . for ordinary baryons
we expect terms like @xmath54 ( @xmath55 being of the order of @xmath56 ) to be small in comparison with the leading term . for antidecuplet both terms
@xmath48 and @xmath57 are small and comparable in magnitude
. they may add or cancel depending on the decay channel ( for @xmath4 we have only two equal decay channels @xmath58 ) violating completely the su@xmath59 relations between the decay couplings . in this way one
is able to explain the suppression of @xmath27 decay channel in @xmath10 decays mentioned above , and also the existence of @xmath60 transition that is forbidden in the su@xmath59 limit .
we see therefore that in the case of antidecuplet mixing is an important ingredient primarily to understand the decays , but also the masses @xcite .
unfortunately , at least at the first sight , there is large arbitrariness as far as mixing angles and transition matrix elements are concerned .
here we shall try to constrain them from the existing data .
if the data are not available we shall make estimates based on @xmath14qsm .
we shall consider mixing of the ground state octet ( @xmath61 ) , the octet of the @xmath62 roper resonance ( @xmath63 ) and antidecuplet . in the su@xmath59 limit
all states in the ground state octet and in antidecuplet are separately degenerate in mass . when @xmath56 corrections are switched on the masses split and take values given by gell - mann okubo ( gmo ) mass formulae . at the same time
the wave functions become mixtures of the original representation ( 8 or @xmath7 ) and all other allowed su@xmath59 representations that appear in the tensor product @xmath64 or @xmath65 .
this introduces @xmath66 mixing for nucleon - like and @xmath13-like states characterized by angle @xmath51 ( note that due to the accidental equality of the pertinent su(3 ) clebsch - gordan coefficients mixing angles between @xmath67- and @xmath13-like states are equal in the leading order in @xmath68 ) .
we assume that ground state octet gmo states correspond to the physical states , which is true with 0.5% accuracy @xcite .
we assume next that antidecuplet gmo states undergo further mixing with the roper octet . for roper octet gmo mass formulae work with much worse accuracy of approximately 3% @xcite , so there is a need for additional mixing .
again both for @xmath67- and @xmath13-like states the mixing angle @xmath69 is the same .
this is depicted in fig.[fig : angles ] . throughout this paper
we take into account only these two mixings which we generally believe to be small .
-like states .
full circles denote physical states , open circles in the case of roper octet and @xmath7 correspond to the gmo states that undergo further mixing with angle @xmath69 .
grey circles correspond to particles not considered in the present paper . ]
our analysis is based on the following assumptions concerning antidecuplet .
we assume @xmath70 mev and @xmath71 mev . we follow analysis of refs .
@xcite assuming that @xmath10 is a new resonance with mass @xmath72 mev and total width @xmath73 mev .
we also assume hierarchy of the branching ratios that is described in more detail in sect .
[ sect : where ] . with these assumptions
we are able to provide limits on the mixing angles both for the nucleon - like and @xmath13-like states .
we find a small region in the space of mixing angles where the required properties of @xmath10 are reproduced .
for these allowed angles we calculate masses of @xmath12 and @xmath9 and their decay patterns .
the paper is organized as follows . in the next section
we discuss general formulae for quantum mixing and define mixing angles and wave functions used throughout this paper . in sect.[sect : decays ] we calculate the decay constants in the presence of mixing .
we define reduced matrix elements and discuss their hermiticity properties .
section [ sect : where ] contains numerical results of our analysis . finally in sect.[sect : conclusions ] we briefly summarize our results and present conclusions .
before we discuss three state mixing , let us recall the formulae for two state mixing , that can be found for example in ref.@xcite . consider perturbation hamiltonian ( @xmath74 @xmath75 ) where @xmath76:@xmath77{cc}m_{1 } & -v\\ -v & m_{2}\end{array } \right ] .
\label{hprim}\ ] ] here we have chosen `` @xmath78 '' sign in front of @xmath79 in order to be in agreement with the sign dictated by the @xmath14qsm .
let us consider for the moment @xmath80 mixing . in that case
hamiltonian ( [ hprim ] ) represents mixing between nucleon states or @xmath13 states ( with different entries for each case ) .
exotic states @xmath4 and @xmath9 remain unmixed .
it is important to note here that the su(3 ) clebsch - gordan coefficients are identical for @xmath67 and @xmath13 states and therefore @xmath79 is the same .
introducing@xmath81 we get the following mass eigenvalues@xmath82 and eigenvectors@xmath83 the mixing angle @xmath51 @xmath84{c}\left\vert { } \right . 1^{\text{phys}}\left
. { } \right\rangle \\ \left\vert { } \right .
2^{\text{phys}}\left .
{ } \right\rangle \end{array } \right ] = \left [ \begin{array } [ c]{rc}\cos\alpha & \sin\alpha\\ -\sin\alpha & \cos\alpha \end{array } \right ] \left [ \begin{array } [ c]{c}\left\vert 1\right\rangle \\
\left\vert 2\right\rangle \end{array } \right ] \label{rotation}\ ] ] is conveniently defined through the second equation ( [ eigv2])@xmath85 the last approximation consists in approximating @xmath86 by the difference of the mean multiplet masses in the spirit of the first order perturbation theory in @xmath68
. then eq.([eigv2 ] ) implies@xmath87 let us consider three states belonging to ordinary octet
@xmath61 , `` roper '' octet @xmath63 and antidecuplet @xmath7 , with unperturbed masses satisfying@xmath88 the vector of nucleon - like ( or @xmath13-like ) states@xmath84{l}\left\vert 8_{1}\right\rangle \\
\left\vert 8_{2}\right\rangle \\
\left\vert \overline{10}\right\rangle \end{array } \right ] \label{vector}\ ] ] is a subject of mixing by the orthogonal matrix@xmath89{ccc}\cos\theta & \sin\theta & 0\\ -\sin\theta & \cos\theta & 0\\ 0 & 0 & 1 \end{array } \right ] \left [ \begin{array } [ c]{ccc}1 & 0 & 0\\ 0 & \cos\phi & \sin\phi\\ 0 & -\sin\phi & \cos\phi \end{array } \right ] \left [ \begin{array } [ c]{ccc}\cos\alpha & 0 & \sin\alpha\\ 0 & 1 & 0\\ -\sin\alpha & 0 & \cos\alpha \end{array } \right ] .
\label{omatrix}\ ] ] throughout this paper we neglect roper - ground octet mixing and set @xmath90 . angle @xmath91 as in sect.[sect : two ] describes @xmath92 mixing , angle @xmath69 is responsible for the mixing of roper states with already mixed @xmath7 , as depicted schematically in fig.[fig : angles ] .
since we anticipate that the mixing will be not large , the physical states can be labeled by the su(3 ) representations so that:@xmath93 the sequential mixing ( [ omatrix ] ) allows to calculate physical masses by iterative application of ( [ physmass ] ) .
after the first mixing of antidecuplet with the ground state octet , antidecuplet states satisfy gell - man
okubo mass formulae ( from now on particle names stand for their masses):@xmath94 these states ( @xmath10 and @xmath12 then mix further with the roper octet.@xmath95 these formulae are valid both for nucleon- and sigma - like states . since physical masses of the nucleon - like states are known we can calculate bare @xmath10 from ( [ barmass])@xmath96 since we know @xmath97 ( which is not mixed ) and @xmath10 ( from ( [ barmass1 ] ) ) we can calculate @xmath98 and then @xmath99 ( because @xmath9 is not mixed as well ) and finally @xmath100 from the last equation in ( [ barmass ] ) since both @xmath101 and @xmath102 ( from ( [ go ] ) ) are known:@xmath103
to calculate the decay width of baryon @xmath104 ( where @xmath105 stands for the pseudoscalar meson ) we shall use following @xcite
the generalized goldberger - treiman relation employed first by witten , adkins and nappi in ref.@xcite : @xmath106 here @xmath107 is the mass of the decaying baryon , @xmath108 the mass of the decay product , @xmath109 meson momentum given by:@xmath110 the decay constant @xmath111 stands for the matrix element of the tensor decay operator @xmath112:@xmath113 explicit form of the decay operator is known for example in @xmath14qsm @xcite .
here we shall simply assume , that @xmath112 transforms as a @xmath105 component of the octet and as spin @xmath114 .
this , together with the assumption that @xmath112 satisfies hermiticity condition will allow us to express the relevant matrix elements between the unmixed states with the help of su@xmath115(3 ) isoscalar factors and the reduced matrix elements .
below we list all matrix elements needed in the present analysis .
decays to octet ( ground state or roper ) are given by:@xmath116{cc}8 & 8\\ \varphi & b_{1}\end{array } \right .
\left\vert \begin{array } [ c]{c}\overline{10}\\ b_{2}\end{array } \right ] g_{\overline{10}}\qquad\left ( \text{or } g_{\overline{10}}^{r}\right ) , \nonumber\\ \left\langle 8,b_{2}\right\vert o_{\varphi}^{(8)}\left\vert 10,b_{1}\right\rangle & = \sqrt{2}\left [ \begin{array } [ c]{cc}8 & 8\\ \varphi & b_{1}\end{array } \right . \left\vert \begin{array } [ c]{c}10\\ b_{2}\end{array } \right ] g_{10}\qquad\left ( \text{or } g_{\overline{10}}^{r}\right ) .
\label{to8}\ ] ] diagonal matrix elements are defined as follows:@xmath117{cc}8 & \overline{10}\\ \varphi & b_{1}\end{array } \right .
\left\vert \begin{array } [ c]{c}\overline{10}\\ b_{2}\end{array } \right ] h_{\overline{10}},\nonumber\\ \left\langle 8,b_{2}\right\vert o_{\varphi}^{(8)}\left\vert 8,b_{1}\right\rangle & = 2\left [ \begin{array } [ c]{cc}8 & 8\\ \varphi & b_{1}\end{array } \right .
\left\vert \begin{array } [ c]{c}8_{1}\\ b_{2}\end{array } \right ] a+\sqrt{20}\left [ \begin{array } [ c]{cc}8 & 8\\ \varphi & b_{1}\end{array } \right .
\left\vert \begin{array } [ c]{c}8_{1}\\ b_{2}\end{array } \right ] b , \label{diag}\ ] ] with @xmath118 when one of the octets is roper .
note that transitions @xmath119 are equal to @xmath7 @xmath120 of ( [ to8 ] ) by the hermiticity requirement . for transitions @xmath121 and @xmath122 more care
is needed since decuplet has spin 3/2 ; we shall comment upon this later .
matrix elements used in this paper are displayed in table [ tablematel ] [ c]r|ccc@xmath123 & @xmath124 & @xmath125 & @xmath126 + @xmath127 & @xmath48 & @xmath52 & @xmath78 + @xmath128 & @xmath129 & @xmath130 & @xmath131 + @xmath132 & @xmath129 & @xmath133 & @xmath134 + @xmath135 & @xmath136 & @xmath78 & @xmath137 + @xmath138 & @xmath129 & @xmath52 & @xmath139 + @xmath140 & @xmath141 & @xmath142 & @xmath143 + @xmath144 & @xmath129 & @xmath78 & @xmath145 + @xmath146 & @xmath136 & @xmath78 & @xmath145 + @xmath147 & @xmath141 & @xmath148 & @xmath149 + @xmath150 & @xmath151 & @xmath78 & @xmath152 + @xmath153 & @xmath154 & @xmath78 & @xmath78 + @xmath155 & @xmath156 & @xmath157 & @xmath78 + in the case of pion - nucleon coupling and roper decays we may safely neglect small mixing corrections proportional to the sinuses of the mixing angles . with notation of table [ tablematel ]
we have:@xmath158 furthermore roper decay constants read:@xmath159 since for the roper octet @xmath160 @xcite , which makes @xmath161 very small ( note that the tiny decay width of the roper to @xmath162 @xcite results both from the smallness of the coupling constant and of the phase space ) , we will assume in the following that@xmath163 which corresponds to@xmath164 for antidecuplet decays we keep mixing terms , since they are comparable in magnitude to the primary decay constants . for @xmath4 we get : @xmath165 since @xmath166 can be directly read off from @xmath4 decay width , it is convenient to express the remaining decay constants through @xmath167 and @xmath168 , @xmath169 and @xmath170 .
this leads to the following decay constants for @xmath10:@xmath171 approximate equalities in eq.([gn10bar ] ) correspond to the small mixing angle limit ( @xmath172angle@xmath173 ) .
apart from physical decay constants mentioned above , @xmath174 depends additionally on @xmath52 that canceled in the expression for @xmath175 .
this is _ a priori _ , apart from the mixing angles , new free parameter that can not be constrained from the data . in what follows
we shall estimate @xmath52 using model calculations within the framework of @xmath14qsm .
decay constant to @xmath176 reads@xmath177 in the su@xmath115(3 ) limit antidecuplet states can not decay to decuplet .
however , in the presence of mixing such decays are possible:@xmath178 where we have introduced new decay constant @xmath179 describing roper decay to @xmath180 . here a remark concerning factor 2 in front of @xmath181 in eq.([g10bar10 ] ) is in order .
factor 2 implies@xmath182 this relation follows from the fact , that the decay operator transforms as an octet in su@xmath115(3 ) but has also spin 1 , since the decays considered here occur in @xmath183-wave . while calculating the width we average over initial spin and isospin and sum over the final state . in the case of @xmath184 decay
initial spin and isospin are @xmath185 and averaging of amplitude @xmath186 gives@xmath187 where @xmath188 is the reduced amplitude . when calculating transition @xmath189 which appears due to the mixing @xmath190 we have , due to hermiticity , the same amplitude @xmath188 , however averaging is different:@xmath191 hence@xmath192 which is effectively ( up to the overall phase ) the same as ( [ gdeln ] ) .
decay constant for @xmath12 to nucleon reads as follows:@xmath193 decays to @xmath13 take the following form:@xmath194 decay to @xmath195 is given by@xmath196 finally for the decays to decuplet we have:@xmath197
in order to get some numerical insight into eqs.([gn10bar ] ) let us recall the results of @xmath14qsm@xmath198 in notation of ref.@xcite .
taking fits to decays without mixing we get@xmath199 in the present work we completely eliminate @xmath48 through eq.([gthnk ] ) , however @xmath52 reappears in other decay constants .
therefore in the following we shall choose @xmath200 , however we also checked other values , namely @xmath201 and @xmath202 .
note that even if we know some decay constant such as @xmath168 from experiment , we still have freedom in choosing the relative phase with which it enters eqs.([gthnk])([g10bar10 ] ) .
some of these phases can be absorbed to the sign of the mixing angles , the other ones have to be chosen arbitrarily .
for example taking @xmath203 we can absorb the relative phase into the sign of @xmath51 ( see eq.([gthnk ] ) ) .
therefore the sign of @xmath204 can not be absorbed into redefinition of @xmath51 .
choosing negative @xmath52 we have followed the sign dictated by the @xmath14qsm , therefore we have to check sensitivity of our predictions for @xmath205 .
we find that in this case the results are not compatible with experimentally acceptable pattern of decay widths .
similarly the phase of @xmath170 can be absorbed into the sign of angle @xmath69 , therefore the relative phase between @xmath206 and @xmath179 is not fixed ( see eq.([g10bar10 ] ) ) . in what follows
we assume that both @xmath181 and @xmath179 are positive and show that for negative phase we again get results that ar not compatible with experiment . throughout this paper
we assume the following values for the parameters entering decay constants:@xmath207 and the following decay widths for @xmath184 and roper @xcite:@xmath208 decays into : @xmath209 long dash ( red ) , @xmath162 short dash ( blue ) , @xmath176 dash - dot ( dark green ) , @xmath180 dash - dot - dot ( purple ) and total width
solid ( black ) as functions of angle @xmath51 for fixed @xmath69 : @xmath210 upper plots , @xmath211 lower plots .
for all plots @xmath200 . , title="fig:",width=226 ]
decays into : @xmath209 long dash ( red ) , @xmath162 short dash ( blue ) , @xmath176 dash - dot ( dark green ) , @xmath180 dash - dot - dot ( purple ) and total width
solid ( black ) as functions of angle @xmath51 for fixed @xmath69 : @xmath210 upper plots , @xmath211 lower plots .
for all plots @xmath200 . ,
title="fig:",width=226 ] + decays into : @xmath209 long dash ( red ) , @xmath162 short dash ( blue ) , @xmath176 dash - dot ( dark green ) , @xmath180 dash - dot - dot ( purple ) and total width
solid ( black ) as functions of angle @xmath51 for fixed @xmath69 : @xmath210 upper plots , @xmath211 lower plots .
for all plots @xmath200 . , title="fig:",width=226 ]
decays into : @xmath209 long dash ( red ) , @xmath162 short dash ( blue ) , @xmath176 dash - dot ( dark green ) , @xmath180 dash - dot - dot ( purple ) and total width
solid ( black ) as functions of angle @xmath51 for fixed @xmath69 : @xmath210 upper plots , @xmath211 lower plots . for all plots @xmath200 .
, title="fig:",width=226 ] + but as functions of angle @xmath69 for fixed @xmath51 : @xmath212 upper plots , @xmath213 lower plots.,title="fig:",width=226 ] but as functions of angle @xmath69 for fixed @xmath51 : @xmath212 upper plots , @xmath213 lower plots.,title="fig:",width=226 ] + but as functions of angle @xmath69 for fixed @xmath51 : @xmath212 upper plots , @xmath213 lower plots.,title="fig:",width=226 ] but as functions of angle @xmath69 for fixed @xmath51 : @xmath212 upper plots , @xmath213 lower plots.,title="fig:",width=226 ] + in figs .
[ fig : hm7al ] and [ fig : hm7fi ] we plot partial decay widths and branching ratios for @xmath10 decays as functions of mixing angles @xmath51 and @xmath69 .
total width has been calculated by adding all partial widths plus 10% for unaccounted three body decays .
first of all let us observe that without mixing the dominant decay mode @xmath27 has branching ratio of 66% , next is @xmath29 with branching ratio 20% and finally @xmath31 4%
. decays to decuplet are forbidden .
this decay pattern contradicts experiment ( provided we want to interpret @xmath38 which decays predominantly to @xmath29 , as a cryptoexotic member of antidecuplet ) .
however , the branching ratios and decay widths change rather rapidly when mixing is included .
we can see from figs .
[ fig : hm7al ] and [ fig : hm7fi ] that for positive @xmath51 and negative @xmath69 there exist regions where the decay to @xmath29 is dominant and the total width is relatively large . in order to find the preferable region in @xmath214 plane
we impose the following requirements @xcite:@xmath215 and plot pertinent contours in fig.[fig : contour ] .
the allowed region defined in eq.([conditions ] ) should lie between the two continuous ( black ) ellipses corresponding to the allowed range of the total width , inside the outer dashed blue ellipse corresponding to br@xmath216 , and between two solid ( red ) lines where @xmath217mev .
we see that there is a rather narrow strip of allowed angles concentrated in the vicinity of the point @xmath218 and @xmath219 . in fig .
[ fig : contour ] we also plot the dotted ( orange ) line inside the allowed region given by the equation @xmath220 ) .
we also plot an orange dotted line inside the allowed region of mixing angles along which we later plot partial decays widths and branching ratios . ] in the following figures we plot various quantities along the line ( [ or_lin ] ) , _ i.e. _ for mixing angles inside the allowed region , indicating the limits on angle @xmath51 by vertical thin lines . first in fig .
[ fig : orange_line ] we show partial decay widths and branching ratios of @xmath10 .
we see that indeed the decay to @xmath27 is strongly suppressed , however another channel , namely the decay to @xmath30 , starts to dominate for larger mixing angles . in any case
decays to decuplet are large ( remember they are forbidden if there is no mixing ) and this prediction provides a stringent test of our model .
decays into : @xmath209 long dash ( red ) , @xmath162 short dash ( blue ) , @xmath176 dash - dot ( dark green ) , @xmath180 dash - dot - dot ( purple ) and total width
solid ( black ) plotted along the line ( [ or_lin ] ) .
for all plots @xmath200 . ,
title="fig:",width=226 ] decays into : @xmath209 long dash ( red ) , @xmath162 short dash ( blue ) , @xmath176 dash - dot ( dark green ) , @xmath180 dash - dot - dot ( purple ) and total width
solid ( black ) plotted along the line ( [ or_lin ] ) .
for all plots @xmath200 . , title="fig:",width=226 ] .
solid lines correspond to the physical masses , whereas dashed lines to the gmo masses before mixing with roper octet .
note that for @xmath4 and @xmath9 gmo states and physical states coincide . ]
next we come to the predictions for @xmath12 and also for @xmath9 . in fig .
[ fig : sigmaxi ] we plot masses of antidecuplet states as functions of @xmath69 .
solid lines correspond to the physical masses .
since we take @xmath4 and @xmath10 as input they do not depend on mixing angle .
dashed lines correspond to gmo states before mixing .
we see that@xmath221 within the allowed limits of fig .
[ fig : contour ] ( not only at the endpoints of the line ( [ or_lin ] ) ) . finally in fig .
[ fig : gsig_orange_line ] we plot partial widths for the decays of @xmath12 .
we see two dominating decay modes : @xmath222 and @xmath223 . in the right panel
we magnify the scale to distinguish between the remaining decays . in fig .
[ fig : brsig_orange_line ] we plot the pertinent branching ratios . in table
[ tablewidths ] we give the range of widths within the allowed limits of fig .
[ fig : contour ] ( not only along the line ( [ or_lin ] ) ) .
[ c]ccc & @xmath224}$ ] & @xmath225}$ ] + @xmath43 & 5.50 & 15.27 + @xmath223 & 1.70 & 4.40 + @xmath226 & 0.09 & 2.57 + @xmath227 & 0.01 & 1.74 + @xmath228 & 0.05 & 1.95 + @xmath229 & 0.22 & 0.50 + @xmath230 & 9.7 & 26.9 + decays into : nk
double dash dot ( brown ) , @xmath231 dash dot ( dark green ) , @xmath232 long dash ( red ) , @xmath184k double dash dot ( purple ) , @xmath233 short dash ( blue ) , @xmath234 dash - dot - dot ( pink ) and total width solid ( black ) plotted along the line ( [ or_lin ] ) .
for all plots @xmath235 . on right panel
we present the enlargement of the left plot in order to distinguish different decay modes that are below 1.2 mev , title="fig:",width=226 ] decays into : nk
double dash dot ( brown ) , @xmath231 dash dot ( dark green ) , @xmath232 long dash ( red ) , @xmath184k double dash dot ( purple ) , @xmath233 short dash ( blue ) , @xmath234 dash - dot - dot ( pink ) and total width
solid ( black ) plotted along the line ( [ or_lin ] ) .
for all plots @xmath235 . on right panel
we present the enlargement of the left plot in order to distinguish different decay modes that are below 1.2 mev , title="fig:",width=226 ] .
all lines as in fig .
[ fig : gsig_orange_line ] .
, title="fig:",width=226 ] .
all lines as in fig . [
fig : gsig_orange_line ] .
, title="fig:",width=226 ] decays into : @xmath209 long dash ( red ) , and total width
solid ( black ) plotted along the line ( [ or_lin ] ) . for all plots @xmath200.,width=226 ] finally we also plot in fig.[fig : gxi_orange_line ] partial decay widths for the two allowed decays modes of @xmath9 .
in the present paper we have examined mixing scenario in which the crypts - exotic antidecuplet states mix with the respective states in the ground state octet and in the roper octet .
this scenario is motivated by recent experimental results on new narrow nucleon resonance @xmath38 .
its small decay width @xmath236 mev and the fact that @xmath237 ( neutron - like state ) undergoes photoexcitation in @xmath32 meson production and @xmath238 does not , is easily explained if these states are interpreted as members of antidecuplet .
however @xmath38 seems to have very small coupling to the @xmath23 channel and comparatively large one to @xmath29 channel in contradiction to pure su@xmath115(3 ) predictions for these decays .
since roper resonance has very small partial width to @xmath29 one can adjust its mixing angle with @xmath38 to suppress @xmath27 coupling not affecting the @xmath32n one .
we have shown that there exists a small , but stable against variation of the unknown couplings , region in the space of mixing angles where such mechanism is effectively working .
our main result concerning the strange members of antidecuplet is based on the observation that @xmath13-like states have the same mixing angles as the nucleonic states and on the fact that @xmath239 does not mix neither with roper octet nor with the ground state octet .
this allows us to constrain masses of these states and their decay patterns .
before we discuss our results in more detail we want to stress that the mass limits given in eqs.([sigmarange],[xirange ] ) and the decay widths summarized in tab .
[ tablewidths ] should be considered as the qualitative ones .
variations of roper decay constants within the experimental limits and roper @xmath240 parameter would enlarge limits derived in this paper .
also our discussion concerning @xmath52 parameter was only qualitative .
we predict that @xmath12 has a mass around 1815 mev within the limits of eq.([sigmarange ] ) .
there are no known @xmath13 resonances in this energy range @xcite .
total width of @xmath12 does not exceed 30 mev but is also constrained from below being larger that 10 mev .
note that we have calculated total widths by adding all partial widths and 10% for unaccounted three body decays .
most prominent decay channels are @xmath43 and @xmath223 with branching ratios approximately 60% and 20% respectively . due to the mixing su@xmath115(3 ) forbidden decays to decuplet are possible , but small , at the level of 5 to 9% .
the simplest way to detect @xmath12 is its formation in @xmath241 scattering with the kaon beam of @xmath242 gev .
the corresponding resonance cross section is about 5 millibarn .
note , however , that the non - resonant @xmath241 cross - section in this energy region is tens of millibarn .
that requires high statistics experiment and detailed partial wave analysis .
additionally , the small width of predicted @xmath12 demands high energy resolution of the kaon beam .
@xmath12 can be produced in the photoproduction experiments , e.g. in @xmath243 , note however that according to simple estimate based on the u - spin the corresponding production cross section is about three times smaller than the analogous @xmath244 cross section , i.e. very small .
one can reveal the small signal of the @xmath245 processes using its interference with much stronger amplitude of @xmath246 @xcite .
details of possible ways to see @xmath12 in various processes we shall give elsewhere @xcite . from ( [ xirange ] ) we see that @xmath247 is larger than 1900 mev .
the latter estimate is in disagreement with the result of na49 @xcite .
there is one known three star resonance in the energy range ( [ xirange ] ) , namely @xmath248 of unknown spin and parity @xcite .
however it has isospin @xmath249 and therefore it can not belong to @xmath7 . indeed in ref.@xcite @xmath248 has been attributed to the same octet as @xmath250 .
our results are based on , to some extent arbitrary , assignment of the relative signs of two reduced matrix elements defining decay constants : @xmath204 and @xmath179 .
we have fixed @xmath52 to be negative ( @xmath251 ) in agreement with @xmath14qsm and @xmath179 to be positive ( @xmath252 ) . in fig.[fig : wrong ] we show contour plots corresponding to ( [ conditions ] ) with these phases inverted .
we see that there is no common stable intersection region in the space of mixing angles for @xmath253 or / and @xmath254 .
in fact there is small allowed spot for @xmath200 and @xmath255 , but it is unstable for small variations of these values or conditions ( [ conditions ] ) . finally , in the same figure we present contour plot for large width of @xmath4 where also no allowed region of mixing angles exists .
to conclude : we have proposed a scenario in which roper octet can mix with putative antidecuplet of exotic baryons and predicted the properties of its strange members @xmath12 and @xmath9 .
we hope that these estimates will be helpful in eventual experimental searches . for three unconstraint phases taking values different than in the main analysis , and for the standard phases but for large decay width of @xmath4.,title="fig : " ] for three unconstraint phases taking values different than in the main analysis , and for the standard phases but for large decay width of @xmath4.,title="fig : " ] + for three unconstraint phases taking values different than in the main analysis , and for the standard phases but for large decay width of @xmath4.,title="fig : " ] for three unconstraint phases taking values different than in the main analysis , and for the standard phases but for large decay width of @xmath4.,title="fig : " ] + the authors are grateful to m. amarian , d. diakonov , v. kuznetsov and v. petrov for discussions .
the paper was partially supported by the polish - german cooperation agreement between polish academy of sciences ( pan ) and deutsche forschungsgemeinschaft ( dfg ) , and sfb tr16 ( bochum - bonn - giessen ) of dfg .
l.c biedenharn and y. dothan , _ monopolar harmonics in su(3)@xmath256 as eigenstates of the skyrme - witten model for baryons _ , e. gotsman and g. tauber ( eds . ) , _ from su(3 ) to gravity _ , cambridge university press 1985 , p. 15 - 34 .
b. mckinnon _ et al . _
[ clas collaboration ] , phys .
lett . * 96 * , 212001 ( 2006 ) [ arxiv : hep - ex/0603028],r . de vita _ et al . _ [ clas collaboration ] , phys .
d * 74 * , 032001 ( 2006 ) [ arxiv : hep - ex/0606062 ] .
k. s. choi , s. i. nam , a. hosaka and h. c. kim , phys .
b * 636 * , 253 ( 2006 ) [ arxiv : hep - ph/0512136];s .
i. nam , k. s. choi , a. hosaka and h. c. kim , prog .
theor . phys .
suppl . * 168 * , 97 ( 2007 ) [ arxiv:0704.3101 [ hep - ph]];k .
s. choi , s. i. nam , a. hosaka and h. c. kim , arxiv:0710.2185 [ hep - ph]k .
s. choi , s. i. nam , a. hosaka and h. c. kim , arxiv:0707.3854 [ hep - ph ] .
r. a. arndt , i. i. strakovsky and r. l. workman , phys .
c * 68 * , 042201 ( 2003 ) [ erratum - ibid .
c * 69 * , 019901 ( 2004 ) ] [ arxiv : nucl - th/0308012 ] ; + j. haidenbauer and g. krein , phys . rev .
c * 68 * ( 2003 ) 052201 [ arxiv : hep - ph/0309243 ] ; + r. a. arndt , i. i. strakovsky and r. l. workman , nucl .
phys . a * 754 * ( 2005 ) 261 [ arxiv : nucl - th/0311030 ] ; + w. r. gibbs , phys .
c * 70 * ( 2004 ) 045208 [ arxiv : nucl - th/0405024 ] .
m. amarian , d. diakonov and m. v. polyakov , phys .
d * 78 * ( 2008 ) 074003 [ arxiv : hep - ph/0612150 ] ; + an review of interference effects in particle physics is given by y. azimov , arxiv:0904.1376 [ hep - ph ] .
|
we propose a scenario in which roper octet can mix with a putative antidecuplet of exotic baryons and predict the properties of its strange members .
we show that @xmath0 and @xmath1 .
we also estimate total widths : @xmath2 and @xmath3 and branching ratios for different decay modes .
|
ultra - massive white dwarfs , generally defined as having masses exceeding @xmath3 , remained relatively rare until deep spectroscopic surveys uncovered these intrinsically faint objects . the hot hydrogen - rich ( da ) white dwarf gd 50 @xcite , the carbon - rich white dwarf g35 - 26 @xcite , and two da white dwarfs ( pg 1658 + 441 and pg 0136 + 251 ) from the palomar - green survey @xcite were rare examples of this phenomenon .
interestingly , spectroscopic follow - up of extreme ultraviolet ( euv ) surveys of the local ( @xmath4 pc ) , hot white dwarf population managed to identify many new ultra - massive white dwarfs on the basis of a _ large , spectroscopically determined surface gravity _ ( @xmath5 ) .
based on the _ extreme ultraviolet explorer _
( _ euve _ ) and rosat _ wide field camera _ ( _ wfc _ ) surveys , @xcite , @xcite , @xcite , and @xcite added a dozen new objects to the population .
more recently , a re - analysis of the palomar - green ( pg ) sample of da white dwarfs @xcite and a study of the white dwarf mass distribution in the sloan digital sky survey ( sdss ; * ? ? ?
* ) seem to generate similar yields of ultra - massive white dwarfs . are these objects products of single star evolution or double degenerate mergers ?
@xcite reviews theoretical arguments in favour of an upper mass limit for white dwarf stars larger than the canonical upper limit of @xmath3 , and possibly as large as @xmath6 . a limit of @xmath3 is generally assumed because carbon ignition in high mass cores would lead to thermonuclear runaway or core collapse .
however , the effect of mass loss may alter the scenario and lead to the formation of a massive oxygen - neon - magnesium white dwarf @xcite . @xcite and @xcite successfully evolved 9 and 10 @xmath7 stars , which , following an off - centre carbon ignition in partial electron degenerate conditions , generated oxygen - neon core white dwarfs with carbon - oxygen shells , and total masses of @xmath8 and @xmath9 , respectively .
on the other hand , the merger scenario proposed for the origin of type ia supernovae @xcite also generates ultra - massive white dwarfs .
however , the merger process itself and the fate of these objects are uncertain @xcite .
we re - examine the evidence for the existence of ultra - massive white dwarfs .
mass measurements are based on gravitational redshift measurements ( 2 ) , radius ( parallax ) measurements ( 3 ) , orbital parameters ( 4 ) , and surface gravity measurements ( 5 ) . the gravitational redshift measurements ( @xmath10 ) , radius measurements , and surface gravity measurements ( @xmath11 ) are converted into mass measurements by adopting mass - radius relations for a variety of model structures .
we adopted the models of @xcite with a c / o core and zero metallicity , a helium mantle ( @xmath12 ) , and with a _ thick _ ( @xmath13 ) hydrogen layer to describe hydrogen - rich da white dwarfs or without a hydrogen layer to describe non - da white dwarfs . in 5 , we also present evidence that deep surveys of high - proper motion white dwarfs are now extending the ultra - massive population toward lower temperatures , and it is estimated that @xmath14% of white dwarfs have masses in excess of @xmath3 .
we summarize and conclude in 6 .
the massive white dwarf and member of the pleiades lb 1497 @xcite remained a unique candidate for having a mass in excess of @xmath15 and possibly in excess of @xmath3 among stars with gravitational redshift measurements @xcite .
however , applying the same techniques to white dwarfs in common proper motion binaries @xcite proposed new ultra - massive white dwarf candidates ( table [ tbl-1 ] ) .
we demonstrate that this is not the case and that the ultra - massive white dwarf sequence presented by @xcite and based on gravitational redshift measurements does not exist .
we also re - evaluate the mass of lb 1497 and show that its case remains compelling .
.the cpmb sample@xmath16 [ cols="^,^,^",options="header " , ] + @xmath16 using the mass - radius relations for a co core @xcite . for masses above @xmath17 , mass - radius relations of @xcite were used .
+ @xmath18 using the mass - radius relations for a one core @xcite . for masses above @xmath19 ,
mass - radius relations of @xcite were used .
+ @xmath20 based on the model fit of kawka & vennes ( 2006 ) assuming @xmath21 .
@xmath22 based on the distance modulus of the da companion lb 9802 @xcite .
+ references : ( 1 ) @xcite ; ( 2 ) @xcite ; ( 3 ) @xcite ; ( 4 ) @xcite .
s. v. is grateful for the hospitality and support of the astronomical institute at ondrejov observatory .
acknowledges support from the centre for theoretical astrophysics ( lc06014 ) .
we are grateful to jay farihi for pointing out recent developments in the study of gd 362 .
99 althaus , l. g. , garca - berro , e. , isern , j. , & crsico , a. h. 2005 , a&a , 441 , 689 barstow , m. a. , bond , h. e. , burleigh , m. r. , & holberg , j. b. 2001 , mnras , 322 , 891 barstow , m. a. , holberg , j. b. , & koester , d. 1994 , mnras , 270 , 516 benvenuto , o. g. , & althaus , l. g. 1999 , mnras , 303 , 30 bergeron , p. , kidder , k. m. , holberg , j. b. , liebert , j. , wesemael , f. , & saffer , r. a. 1991 , apj , 372 , 267 bergeron , p. , leggett , s. k. , & ruiz , m. t. 2001 , apjs , 133 , 413 bergeron , p. , liebert , j. , & fulbright , m. s. 1995 , apj , 444 , 810 berghfer , t. w. , vennes , s. , & dupuis , j. 2000 , apj , 538 , 854 bessell , m. s. , & brett , j. m. 1988 , pasp , 100 , 1134 burleigh , m.r . , & barstow , m.a . 1999 ,
a&a , 341 , 795 burleigh , m.r . , & barstow , m.a .
2000 , a&a , 359 , 977 cataln , s. , isern , j. , garcia - berro , e. , & ribas , i. 2008a , arxiv e - prints , 804 , arxiv:0804.3034 cataln , s. , ribas , i. , isern , j. , & garcia - berro , e. 2008b , a&a , 477 , 901 chiba , m. , & beers , t. c. 2000 , aj , 119 , 2843 cutri , r.m .
, et al . 2006 , explanatory supplement to the 2mass all sky data release and extended mission products ( pasadena : caltech ) dahn , c. c. 1999 , 11th european workshop on white dwarfs , astronomical society of the pacific conference series , 169 , 24 dahn , c. c. , bergeron , p. , liebert , j. , harris , h. c. , canzian , b. , leggett , s. k. , & boudreault , s. 2004 , apj , 605 , 400 dotter , a. , chaboyer , b. , jevremovic , d. , kostov , v. , baron , e. , & ferguson , j. w. 2008 , arxiv e - prints , 804 , arxiv:0804.4473 dupuis , j. , chayer , p. , vennes , s. , allard , n.f . , & hbrard , g. 2003 , apj , 598 , 486 eggen , o. j. ; greenstein , j. l. 1965 , apj , 141 , 83 finley , d. s. , koester , d. , & basri , g. 1997 , apj , 488 , 375 garcia - berro , e. , ritossa , c. , & iben , i. j. 1997 , apj , 485 , 765 gianninas , a. , bergeron , p. , & fontaine , g. 2005 , apj , 631 , 1100 gianninas , a. , dufour , p. , & bergeron , p. 2004 , apjl , 617 , l57 gould , a. , & chanam , j. 2004 , apjs , 150 , 455 hamada , t. & salpeter , e. e. 1961 , apj , 134 , 683 handler , g. , et al .
2002 , mnras , 333 , 262 harrington , r. s. , & dahn , c. c. 1980 , aj , 85 , 454 hubeny , i. , hummer , d.g .
, & lanz , t. 1994 , a&a , 282 , 151 hummer , d.g . & mihalas , d. 1988 , apj , 331 , 794 inglis , d.r . , & teller , e. 1939 , apj , 90 , 439 johnson , d. r. h. , & soderblom , d. r. 1987 , aj , 93 , 864 kawka , a. , & vennes , s. 2006 , apj , 643 , 402 kawka , a. , vennes , s. , schmidt , g. d. , wickramasinghe , d. t. , & koch , r. 2007 , apj , 654 , 499 kepler , s. o. , kleinman , s. j. , nitta , a. , koester , d. , castanheira , b. g. , giovannini , o. , costa , a. f. m. , & althaus , l. 2007 , mnras , 375 , 1315 kilic , m. , von hippel , t. , leggett , s. k. , & winget , d. e. 2005 , apjl , 632 , l115 kirkpatrick , j. d. , & mccarthy , d. w. , jr .
1994 , aj , 107 , 333 landsman , w. , simon , t. , & bergeron , p. 1993 , pasp , 105 , 841 landsman , w. , simon , t. , & bergeron , p. 1995 , proceedings of the 9th european workshop on white dwarfs , lecture notes in physics , 443 , springer - verlag , berlin , 191 liebert , j. , bergeron , p. , & holberg , j. b. 2005 , apjs , 156 , 47 luyten , w. j. 1976 , a catalogue of 1849 stars with proper motions greater than 0"5 annually ( minneapolis : university of minnesota ) marsh , m. c. , et al .
1997 , mnras , 286 , 369 maxted , p.
f. l. , & marsh , t. r. 1999 , mnras , 307 , 122 moos , h.w .
, et al . 2000 , apj , 538 , l1 munari , u. , sordo , r. , castelli , f. , & zwitter , t. 2005 , a&a , 442 , 1127 nomoto , k. 1984 , apj , 277 , 791 parthasarathy , m. , branch , d. , jeffery , d. j. , & baron , e. 2007 , new astronomy review , 51 , 524 redfield , s. , & linsky , j. l. 2008 , apj , 673 , 283 reid , i. n. 1996 , aj , 111 , 2000 ritossa , c. , garcia - berro , e. , & iben , i. j. 1996 , apj , 460 , 489 schmidt , g. d. , bergeron , p. , liebert , j. , & saffer , r. a. 1992 , apj , 394 , 603 segretain , l. , chabrier , g. , & mochkovitch , r. 1997 , apj , 481 , 355 silvestri , n. m. , hawley , s. l. , & oswalt , t. d. 2005 , aj , 129 , 2428 silvestri , n.m . , oswalt , t.d . , wood , m.a . , smith , j.a . ,
reid , i.n . , sion , e.m .
2001 , aj , 121 , 503 thejll , p. , shipman , h. l. , macdonald , j. , & macfarland , w. m. 1990 , apj , 361 , 197 uytterhoeven , k. , willems , b. , lefever , k. , aerts , c. , telting , j. h. , & kolb , u. 2004 , a&a , 427 , 581 van altena , w. f. , lee , j. t. , & hoffleit , d. 1995 , vizier online data catalog , 1174 , 0 vennes , s. 1999 , apj , 525 , 995 vennes , s. , berghoefer , t. , & christian , d. 1997a , apjl , 491 , l85 vennes , s. , christian , d. j. , & thorstensen , j. r. 1998 , apj , 502 , 763 vennes , s. , kawka , a. , & dupuis , j. 2008 , in preparation vennes , s. , schmidt , g. d. , ferrario , l. , christian , d. j. , wickramasinghe , d. t. , & kawka , a. 2003 , apj , 593 , 1040 vennes , s. , thejll , p. a. , galvan , r. g. , & dupuis , j. 1997b , apj , 480 , 714 vennes , s. , thejll , p. a. , wickramasinghe , d. t. , & bessell , m. s. 1996 , apj , 467 , 782 wegner , g. , & reid , i. n. 1991 , apj , 375 , 674 wegner , g. , reid , i. n. , & mcmahan , r. k. , jr .
1991 , apj , 376 , 186 weidemann , v. 2000 , a&a , 363 , 647 yungelson , l. r. , livio , m. , tutukov , a. v. , & saffer , r. a. 1994 , apj , 420 , 336 zacharias , n. , monet , d. g. , levine , s. e. , urban , s. e. , gaume , r. , & wycoff , g. l. 2005 , vizier online data catalog , 1297 , 0 zuckerman , b. , koester , d. , melis , c. , hansen , b. m. , & jura , m. 2007 , apj , 671 , 872 zuckerman , b. , koester , d. , reid , i. n. , & hnsch , m. 2003 , apj , 596 , 477
|
we re - examine the evidence for the existence of ultra - massive ( @xmath0 ) white dwarfs based on gravitational redshift of white dwarfs in common proper motion binaries or in clusters , on parallax measurements , on orbital solutions , and , finally , on the analysis of hydrogen line profiles .
we conclude that the best evidence is largely based on the analysis of balmer line profiles although the companion to the a8v star hr 8210 is a compelling case made initially using the large binary mass function and confirmed by an analysis of the lyman line spectrum .
the confirmation and identification of high - mass white dwarfs , more particularly non - da white dwarfs , using parallax measurements may prove critical in establishing the population fraction of these objects and in constraining the high - end of empirical initial - mass to final - mass relations .
the existence of a substantial population of ultra - massive white dwarfs supports the concept of a steeper initial - mass to final - mass relations linking @xmath1 progenitors with @xmath2 white dwarfs .
[ firstpage ] stars : evolution stars : fundamental parameters white dwarfs
|
thanks to the ubiquity of powerful , general - purpose computers , large - scale numerical calculations have become an important part of every field of science and engineering , enabling quantitative predictions , analysis , and design of ever more complex systems .
there are a wide variety of different approaches to such calculations , and there is no single `` best '' method for all circumstances not only are some methods better suited to particular situations than to others , but there are also often severe trade - offs between generality / simplicity and theoretical efficiency . even in relatively mature areas like
computational classical electromagnetism ( em ) , a variety of techniques spanning a broad range of sophistication and generality remain in widespread use ( and new variations are continually developed ) @xcite .
semi - analytical approaches also remain important , especially perturbative techniques to decompose problems containing widely differing length scales ( the most challenging situation for brute - force numerics ) .
nevertheless , many commonalities and guiding principles can be identified that seem to apply to a range of numerical techniques .
until a few years ago , casimir forces and other em fluctuation - induced interactions occupied an unusual position in this tableau .
realistic , general numerical methods to solve for casimir forces were simply unavailable ; solutions were limited to special high - symmetry geometries ( and often to special materials like perfect metals ) that are amenable to analytical and semi - analytical approaches .
this is not to say that there were not , _ in principle _ , decades - old theoretical frameworks capable of describing fluctuations for arbitrary geometries and materials , but practical techniques for _ evaluating _ these theoretical descriptions on a computer have only been demonstrated in the last few years @xcite . in almost all cases ,
these approaches turn out to be closely related to computational methods from _ classical _ em , which is fortunate because it means that casimir computations can exploit decades of progress in computational classical em once the relationship between the problems becomes clear .
the long delay in developing numerical methods for casimir interactions , from the time the phenomenon was first proposed in 1948 @xcite , can be explained by three factors .
first , accurate measurements of casimir forces were first reported only in 1997 @xcite and experimental interest in complex casimir geometries and materials has only recently experienced dramatic growth due to the progress in fabricating nanoscale mechanical devices .
second , even the simplest numerical prediction of a single force requires the equivalent of a large number of classical em simulations , a barrier to casual numerical experimentation .
third , there have historically been many equivalent theoretical formulations of casimir forces , but some formulations are much more amenable to computational solution than others , and these formulations are often couched in a language that is opaque to researchers from classical computational em .
this purpose of this review is to survey the available and proposed numerical techniques for evaluating casimir forces , energies , torques , and related interactions , emphasizing their relationships to standard classical - em methods .
our goal is not to identify a `` best '' method , but rather to illuminate the strengths and weaknesses of each approach , highlighting the conclusions that can be gleaned from the classical experience .
we will review an intellectual framework in which to evaluate different numerical techniques , comparing them along several axes for which quasi - independent choices of approach can be made .
we will also emphasize a few key departures of casimir problems from ordinary classical em , such as the necessity of imaginary- or complex - frequency solutions of maxwell s equations and the need for wide - bandwidth analyses , that impact the adaptation of off - the - shelf computational methods .
numerical methods from distinct groups or research papers often differ in several ways simultaneously , complicating the task of directly comparing or even describing them . in order to organize one s understanding of numerical approaches , it is useful to break them down along three axes of comparison , representing ( roughly ) independent choices in the design of a method : * what * problem * does the method solveeven within a single area such as classical em , there are several conceptually different _ questions _ that one can ask and several _ ways of asking _ them that lead to different categories of methods . *
what * basis * is used to express the unknownshow the _ infinite _ number of unknowns in the exact partial differential equation ( pde ) or integral equation are reduced to a _
finite _ number of unknowns for solution on a computer .
* what * solution technique * is used to determine these unknownseven with the same equations and the same unknowns , there are vast differences among the types of direct , sparse , and iterative methods that can be used to attack the problem , and the efficient application of a particular solution technique to a particular problem is sometimes a research task unto itself . in this section ,
we briefly summarize the available problems , basis choices , and solution techniques for casimir problems . in subsequent sections ,
we then discuss in more detail the specific approaches that have currently been demonstrated or proposed . in classical em
, there are several types of problems that are typically posed ( * ? ? ?
* appendix d ) , such as computing source - free time - harmonic eigensolutions @xmath0 and eigenfrequencies @xmath1 , computing time - harmonic fields resulting from a time - harmonic current source @xmath2 , or computing the time - dependent fields created by an arbitrary time - dependent source @xmath3 starting at @xmath4 . although these are all closely mathematically related , and in some sense the solution of one problem can give solutions to the other problems , they lead to very different types of numerical simulations . in a similar way ,
despite the fact that different formulations of casimir - interaction problems are ultimately mathematically equivalent ( although the equivalencies are often far from obvious)and are usually answering the same conceptual question , such as what is the force or interaction energy for some geometry each one leads most naturally to distinct classes of computational methods . here , we exclude formulations such as proximity - force ( `` parallel - plate '' ) approximations @xcite , pairwise summation of casimir polder forces ( valid in the dilute - gas limit ) @xcite , and ray optics @xcite , that are useful in special cases but represent uncontrolled approximations if they are applied to arbitrary geometries .
although at some point the distinctions are blurred by the mathematical equivalencies , we can crudely categorize the approaches as : * computing the eigenfrequencies @xmath5 and summing the zero - point energy @xmath6 @xcite .
see sec .
[ sec : eigenmode - summation ] . * integrating the mean energy density or force density ( stress tensor ) , by evaluating field correlation functions @xmath7 and @xmath8 in terms of the classical em green s functions at @xmath1 via the fluctuation
dissipation theorem @xcite .
see sec .
[ sec : fluctuation - dissipation ] . * evaluating a path - integral expression for the interaction energy ( or its derivative ) , constrained by the boundary conditions usually , portions of the path integrals are performed analytically to express the problem in terms of classical scattering matrices or green s functions at each @xmath1 @xcite .
see sec .
[ sec : path - integrals ] . in each case , the result must be summed / integrated over all frequencies @xmath1 to obtain the physical result ( corresponding to thermodynamic / quantum fluctuations at all frequencies ) .
the relationship of the problem to _ causal _ green s functions ( fields appear _ after _ currents ) means that the integrand is analytic for @xmath9 @xcite . as a consequence , there is a choice of _ contours _ of @xmath1 integration in the upper - half complex plane , which is surprisingly important it turns out that the integrands are wildly oscillatory on the real-@xmath1 axis and require accurate integration over a huge bandwidth , whereas the integrands are much better - behaved along the imaginary-@xmath1 axis ( `` wick - rotated '' or `` matsubara '' frequencies ) .
this means that casimir calculations almost always involve classical em problems evaluated at _ complex or imaginary frequencies _ , as is discussed further in sec .
[ sec : complex - frequency ] .
the nonzero - temperature case , where the integral over imaginary frequencies becomes a sum ( numerically equivalent to a trapezoidal - rule approximation ) , is discussed in sec .
[ sec : finite - temperature ] .
there is also another way to categorize the problem to be solved : whether one is solving a partial differential equation ( * pde * ) or an * integral equation*. in a pde , one has _ volumetric unknowns _ :
fields or other functions at every point in space , related to one another _ locally _ by derivatives and so on . in an integral equation
, one typically has _
surface unknowns _ : the fields or currents on the _ boundaries _ between piecewise - homogeneous regions , related to one another _ non - locally _ by the green s functions of the homogeneous regions ( typically known analytically ) @xcite ( described further in sec .
[ sub : bem - stress ] ) .
the key point is to take advantage of the common situation in which one has piecewise - constant materials , yielding a _ surface integral equation_. ( there are also _ volume integral equations _ for inhomogeneous media @xcite , as well as hybrid integral / pde approaches @xcite , but these are less common . )
there are other hybrid approaches such as _ eigenmode expansion _
@xcite , also called _ rigorous coupled - wave analysis _ ( rcwa ) @xcite or a _ cross - section _
method @xcite : a structure is broken up along one direction into piecewise - constant cross - sections , and the unknown fields at the _ interfaces _ between cross - sections are propagated in the uniform sections via the _ eigenmodes _ of those cross - sections ( computed analytically or numerically by solving the pde in the cross - section ) .
eigenmode expansion is most advantageous for geometries in which the cross - section is constant over substantial regions , just as integral - equation methods are most advantageous to exploit large homogeneous regions .
casimir problems , for the most part , reduce to solving classical em linear pdes or integral equations where the unknowns reside in an infinite - dimensional vector space of functions . to _
discretize _ the problem approximately into a finite number @xmath10 of unknowns , these unknown functions must be expanded in some finite _ basis _ ( that converges to the exact solution as @xmath11 ) .
there are three typical types of basis : * * finite differences * @xcite ( * fd * ) : approximate a function @xmath12 by its values on some uniform grid with spacing @xmath13 , approximate derivatives by some difference expression [ e.g. second - order center differences @xmath14 and integrals by summations ( e.g. a trapezoidal rule ) . *
* finite - element methods * @xcite ( * fem * ) : divide space into geometric _ elements _ ( e.g. triangles / tetrahedra ) , and expand an unknown @xmath12 in a simple _ localized _ basis expansion for each element ( typically , low - degree polynomials ) with some continuity constraints .
( fd methods are viewable as special cases of fems for uniform grids . ) for an integral - equation approach , where the unknowns are functions on surfaces , the same idea is typically called a * boundary - element method * ( * bem * ) @xcite . * * spectral * methods @xcite : expand functions in a non - localized complete basis , truncated to a finite number of terms .
most commonly , fourier series or related expansions are used ( cosine series , fourier
bessel series , spherical or spheroidal harmonics , chebyshev polynomials , _ etc .
finite differences have the advantage of simplicity of implementation and analysis , and the disadvantages of uniform spatial resolution and relatively low - order convergence ( errors typically @xmath15 @xcite or even @xmath16 in the presence of discontinuous materials unless special techniques are used @xcite ) .
fems can have nonuniform spatial resolution to resolve disparate feature sizes in the same problem , at a price of much greater complexity of implementation and solution techniques , and can have high - order convergence at the price of using complicated curved elements and high - order basis functions .
spectral methods can have very high - order or possibly exponential ( `` spectral '' ) convergence rates @xcite that can even suit them to analytical solution hence , spectral methods were the dominant technique before the computer era and are typically the first class of methods that appear in any field , such as in mie s classic solution of wave scattering from a sphere @xcite .
however , exponential convergence is usually obtained only if all discontinuities and singularities are taken explicitly into account in the basis @xcite . with discontinuous materials
, this is typically only practical for very smooth , high - symmetry geometries like spheres , cylinders , and so on ; the use of a generic fourier / spectral basis for arbitrary geometries reduces to a brute - force method that is sometimes very convenient @xcite , but may have unremarkable convergence rates @xcite .
bems require the most complicated implementation techniques , because any nontrivial change to the green s functions of the homogeneous regions ( e.g. a change in dimensionality , boundary conditions , or material types ) involves tricky changes to the singular - integration methods required to assemble the matrix @xcite and to the fast - solver methods mentioned in sec .
[ sub : solution - techniques ] .
given fem / bem or spectral basis functions @xmath17 and a linear equation @xmath18 for an unknown function @xmath19 in terms of a linear differential / integral operator @xmath20 , there are two common ways @xcite to obtain a finite set of @xmath10 equations to determine the @xmath10 unknown coefficients @xmath21 in @xmath22 .
one is a * collocation * method : require that @xmath23 be satisfied at _ n _ collocation points @xmath24 .
the other is a * galerkin * method : require that @xmath25 be satisfied for @xmath26 , where @xmath27 is some inner product on the function space .
both approaches result in an @xmath28 matrix equation of the form @xmath29 .
a galerkin method has the useful property that if @xmath20 is hermitian and/or definite then the matrix @xmath30 has the same properties .
the specific situation of vector - valued unknowns in em creates additional considerations for the basis functions . in order to obtain center - difference approximations for all the field components ,
fd methods for em typically use a staggered * yee grid *
@xcite , in which each component of the em fields is offset onto its own @xmath31-shifted grid . in fems for em , in order to maintain the appropriate continuity conditions for curl or divergence operators , one uses special classes of _ vector - valued _ basis functions such as ndlec elements @xcite . in bems for em ,
vector - valued * rwg * ( rao , wilton , and glisson ) basis functions @xcite ( or generalizations thereof @xcite ) are used in order to enforce a physical continuity condition on surface currents ( to preclude accumulation of charge at element edges ) ; see also fig .
[ fig : crossed - capsules ] in sec .
[ sub : bem - stress ] . a spectral integral - equation method for em with cylindrical or spherical scatterers
is sometimes called a * multipole - expansion * method @xcite , since the obvious spectral basis is equivalent to expanding the scattered fields in terms of multipole moments . given a particular problem and basis choice ,
one at the end obtains some @xmath28 set of linear equations @xmath32 to solve ( or possibly eigenequations @xmath33 ) . where one evaluates a log determinant or a trace of an inverse , since this is done using either eigenvalues or the same matrix factorizations that are used to solve @xmath32 .
] note also that a _
single _ casimir - force calculation requires the solution of _ many _ such equations , at the very least for an integral over frequencies ( see sec .
[ sec : complex - frequency ] ) .
there are essentially three ways to solve such a set of equations : * * dense - direct * solvers : solve @xmath32 using direct matrix - factorization methods ( e.g. gaussian elimination ) , are necessarily iterative , but modern dense - eigensolver techniques employ direct factorizations as steps of the process @xcite . ] requiring @xmath34 storage and @xmath35 time @xcite . *
* sparse - direct * solvers @xcite : if @xmath36 is _ sparse _
( mostly zero entries ) , use similar direct matrix - factorization methods , but cleverly re - arranged in an attempt to preserve the sparsity .
time and storage depend strongly on the _ sparsity pattern _ of @xmath36 ( the pattern of nonzero entries ) .
* * iterative methods * @xcite : repeatedly improve a guess for the solution @xmath37 ( usually starting with a random or zero guess ) , only referencing @xmath36 via repeated matrix vector multiplies .
time depends strongly on the properties of @xmath36 and the iterative technique , but typically requires only @xmath38 storage .
exploits any fast way [ ideally @xmath38 or @xmath39 ) ] to multiply @xmath36 by any arbitrary vector .
if the number @xmath10 of degrees of freedom is small , i.e. if the basis converges rapidly for a given geometry , dense - direct methods are simple , quick , and headache - free ( and have a standard state - of - the - art implementation in the free lapack library @xcite ) .
for example , @xmath40 problems can be solved in under a second on any modern computer with a few megabytes of memory .
up to @xmath41 is reasonably feasible , but @xmath42 requires almost 100 gb of memory and days of computation time without a large parallel computer .
this makes dense - direct solvers the method of choice in simple geometries with a rapidly converging spectral basis , or with bem integral - equation methods for basic shapes that can be accurately described by a few thousand triangular panels , but they rapidly become impractical for larger problems involving many and/or complex objects ( or for moderate - size pde problems even in two dimensions ) . in pde methods with a localized ( fd or fem ) basis ,
the matrices @xmath36 have a special property : they are _ sparse _
( mostly zero ) .
the locality of the operators in a typical pde means that each grid point or element directly interacts only with a bounded number of neighbors , in which case @xmath36 has only @xmath38 nonzero entries and can be stored with @xmath38 memory .
the process of solving @xmath32 , e.g. computing the lu factorization @xmath43 by gaussian elimination @xcite , unfortunately , ordinarily destroys this sparsity : the resulting @xmath44 and @xmath45 triangular matrices are generally not sparse .
however , the pattern of nonzero entries that arises from a pde is not random , and it turns out that clever re - orderings of the rows and columns during factorization can partially preserve sparsity for typical patterns ; this insight leads to _ sparse - direct _ solvers @xcite , available via many free - software packages implementing different sparsity - preserving heuristics and other variations @xcite .
the sparsity pattern of @xmath36 depends on the dimensionality of the problem , which determines the number of neighbors a given element interacts with . for meshes / grids having nearest - neighbor interactions
, a sparse - direct solver typically requires @xmath38 time and storage in 1d ( where the matrices are band - diagaonal ) , @xmath46 time with @xmath47 storage in 2d , and @xmath34 time with @xmath48 storage in 3d @xcite .
the practical upshot is that sparse - direct methods work well for 1d and 2d pdes , but can grow to be impractical in 3d . for bem and spectral methods
, the interactions are not localized and the matrices are not sparse , so sparse - direct methods are not directly applicable ( but see below for an indirect technique ) . for the largest - scale problems , or for problems lacking a sparse @xmath36 ,
the remaining possibility is an _ iterative _ method . in these methods ,
one need only supply a fast way to multiply @xmath36 by an arbitrary vector @xmath49 , and the trick is to use this @xmath50 operation on a clever sequence of vectors in such a way as to make an arbitrary initial guess @xmath51 converge as rapidly as possible to the solution @xmath52 , ideally using only @xmath38 storage .
many such techniques have been developed @xcite .
the most favorable situation for @xmath53 occurs when @xmath36 is hermitian positive - definite , in which case an ideal krylov method called the _ conjugate - gradient _ method can be applied , with excellent guaranteed convergence properties @xcite , and fortunately this is precisely the case that usually arises for the imaginary - frequency casimir methods below .
there are two wrinkles that require special attention , however .
first , one must have a fast way to compute @xmath50 .
if @xmath36 is sparse ( as for pde and fd methods ) , then only @xmath38 nonzero entries of @xmath36 need be stored ( as above ) and @xmath50 can be computed in @xmath38 operations . in a spectral method , @xmath36 is generally dense , but for spectral pde methods there are often fast @xmath47 techniques to compute @xmath50 using only @xmath38 storage ( @xmath36 is stored implicitly ) , based on fast fourier transform ( fft ) algorithms @xcite . in a bem , where @xmath36 is again dense , a variety of sophisticated methods that require only @xmath47 computation time and @xmath38 storage to compute @xmath50 ( again storing @xmath36 implicitly ) have been developed @xcite , beginning with the pioneering fast - multipole method ( fmm ) @xcite .
these fast bems exploit the localized basis and the decaying , convolutional nature of the green s function to approximate long - range interactions ( to any desired accuracy ) .
fmms can be viewed as an approximate factorizations into sparse matrices , at which point sparse - direct methods are also applicable @xcite .
a second wrinkle is that the convergence rates of iterative methods depend on the condition number of @xmath36 ( the ratio of largest and smallest singular values ) @xcite , and condition numbers generally worsen as the ratio of the largest and smallest lengthscales in the problem increases . to combat this ,
users of iterative methods employ * preconditioning * techniques : instead of solving @xmath53 , one solves @xmath54 or similar , where the preconditioning matrix @xmath55 is some crude _ approximate inverse _ for @xmath36 ( but much simpler to compute than @xmath56 ! ) such that the condition number of @xmath57 is reduced @xcite .
the difficulty with this approach is that good preconditioners tend to be highly problem - dependent , although a variety of useful approaches such as incomplete factorization and coarse - grid / multigrid approximations have been identified @xcite .
the upshot is that , while the largest - scale solvers almost invariably use iterative techniques , for any given class of physical problems it sometimes takes significant research before the iterative approach becomes well - optimized .
perhaps the simplest way to express the casimir energy , at zero temperature , is as a sum of zero - point energies of all oscillating em modes in the system : @xmath58 where @xmath1 is the frequency of the mode ( @xmath59 ) @xcite .
that is , when the electromagnetic field is quantized into photons with energy @xmath60 , it turns out that the vacuum state in the absence of photons is not empty , but rather has the energy equivalent of `` half a photon '' in each mode .
the computational strategy is then straightforward , in principle : compute the em eigenfrequencies @xmath1 in the problem by some numerical method ( many techniques are available for computing eigenfrequencies @xcite ) and sum them to obtain @xmath45 .
forces are then given by the derivative of @xmath45 with respect to changes in the geometry , which could be approximated by finite differences or differentiated analytically with a hellman
feynman technique @xcite ( more generally , derivatives of any computed quantity can be computed efficiently by an adjoint method @xcite ) .
of course , @xmath45 in eq .
( [ eq : u - sum ] ) has the disadvantage of being formally infinite , but this is actually a minor problem in practice : as soon as one discretizes the problem into a finite number of degrees of freedom ( e.g. , a finite number of grid points ) , the number of eigenfrequencies becomes finite ( with the upper bound representing a nyquist - like frequency of the grid ) .
this is the numerical analogue @xcite of analytical regularization techniques that are applied to truncate the same sum in analytical computations @xcite .
( these regularizations do not affect energy _ differences _ or forces for rigid - body motions . )
matters are also somewhat subtle for dissipative or open systems @xcite . but
the most serious problem is that , even in the lossless case , this sum is badly behaved : even when one differentiates with separation @xmath61 to obtain a finite force @xmath62 , the summand is wildly oscillatory and includes substantial contributions from essentially _ every _ frequency , which mostly cancel to leave a tiny result @xcite .
numerically , therefore , one must ostensibly compute _ all _ of the modes , to high precision , which requires @xmath35 time and @xmath34 storage ( for a dense - direct eigensolver @xcite ) given @xmath10 degrees of freedom .
this is possible in simple 1d problems @xcite , but is impractical as a general approach .
because of the mathematical equivalence of the different approaches to the casimir problem , the mode - summation method is sometimes useful as a starting point to derive alternative formulations , but the end result is invariably quite different in spirit from computing the eigenfrequencies one by one and summing them . for example , if one has a function @xmath63 whose roots are the eigenfrequencies , then one can equivalently write @xmath45 , via the residue theorem of complex analysis , as @xmath64}{d\omega}d\omega$ ] , where @xmath65 is any closed contour in the complex-@xmath1 plane that encloses the roots @xcite . however , finding functions whose roots are the eigenfrequencies naturally points towards green s functions ( to relate different boundary conditions ) , and the contour choices typically involve wick rotation as in sec .
[ sec : complex - frequency ] , so this approach leads directly to imaginary - frequency scattering - matrix techniques as in sec . [
sec : path - integrals ] @xcite . a similar contour integral arises from a zeta - function regularization of ( [ eq : u - sum ] ) @xcite .
in order to better understand the frequency integration / summation in casimir problems , it is illustrative to examine the analytical formula for the simple case of two perfect - metal plates in vacuum separated by a distance @xmath61 , in which case it can be derived in a variety of ways that the attractive force @xmath66 is given by @xcite:@xmath67\\ & = { \operatorname{re}}\left[\int_{0}^{\infty}f(\omega)d\omega\right]={\operatorname{im}}\left[\int_{0}^{\infty}f(i\xi)d\xi\right]=\frac{\hbar c}{240a^{4}},\end{split } \label{eq : lifshitz - metal}\ ] ] where @xmath68 is the contribution of each frequency @xmath1 to the force and @xmath69 is related to the plate - parallel momentum of the contributing modes / fluctuations . in this special case , the entire integral
can be performed analytically , but for parallel plates of some finite permittivity @xmath70 the generalization ( the _ lifshitz formula _
@xcite ) must be integrated numerically . in practice , however , the formula and its generalizations are never integrated in the form at left
instead , one uses the technique of contour integration from complex analysis to _ wick rotate _ the integral to imaginary frequencies @xmath71 , integrating over @xmath72 .
( in fact , the formula is typically derived _ starting _ in imaginary frequencies , via a matsubara approach @xcite . ) in this section , we review why a trick of this sort is both _ possible _ and _ essential _ in numerical computations for all of the methods described below .
wick rotation is always _
possible _ as a consequence of causality .
it turns out that the frequency contributions @xmath68 for arbitrary materials and geometries , for all of the different formulations of the casimir force below , are ultimately expressed in terms of classical em _ green s functions _ at @xmath1 : the em fields in response to time - harmonic currents @xmath2 . as a consequence of the causality of maxwell s equations and physical materials
em fields always arise _ after _ the source currents , not before
it mathematically follows that the green s functions must be analytic functions ( no poles or other singularities ) when @xmath73 ( the upper - half complex plane ) @xcite .
poles in the green s function correspond to eigenfrequencies or resonances of the source - free maxwell s equations , and must lie at @xmath74 for any physical system with dissipative materials ( with the poles approaching @xmath75 in the idealized lossless limit ) .
[ one can easily see explicitly that this is true for the @xmath68 above : the poles result from a vanishing denominator in the @xmath69 integrand , which only occurs for purely real @xmath1 corresponding to the real - frequency modes trapped between two perfect - metal plates . ] as an elementary consequence of complex analysis , this analyticity means that the @xmath76 can be arbitrarily deformed to any contour in the upper - half complex-@xmath1 plane without changing the integration result .
wick rotation is _ essential _ for computation because the frequency contributions @xmath68 to the force ( or interaction energy or other related quantities ) are extremely ill - behaved near to the real-@xmath1 axis : they are wildly oscillatory and slowly decaying . for example , the magnitude and phase of the function @xmath68 are plotted in the complex @xmath1 plane in fig .
[ fig : lifshitz - omega ] , where the @xmath69 integral was evaluated numerically with a high - order clenshaw curtis quadrature scheme @xcite .
contributions @xmath68 to the casimir force , from each fluctuation / mode frequency @xmath1 , for two perfect - metal plates with separation @xmath61 , in the complex-@xmath1 plane .
_ left _ : magnitude @xmath77 .
_ right : _
phase @xmath78 .
[ the magnitude is truncated at @xmath79 , as it diverges towards the real-@xmath1 axis , and some numerical artifacts ( rapid oscillations ) are visible near the real-@xmath1 axis in the phase due to difficulty in evaluating @xmath68 . ]
the key point is that @xmath68 is badly behaved ( oscillatory and non - decaying ) along contours parallel to the real-@xmath1 axis , whereas @xmath68 is nicely behaved ( non - oscillatory and exponentially decaying ) along contours parallel to the imaginary-@xmath1 axis . ]
merely evaluating @xmath68 along the real-@xmath1 axis is difficult because of singularities ( which ultimately reduce the integral to a summation over eigenfrequency - contributions as in sec .
[ sec : eigenmode - summation ] ) ; in physical materials with dissipation , the real-@xmath1 axis is non - singular but is still badly behaved because of poles ( lossy modes ) located just below the axis . along any contour parallel to the real-@xmath1 axis , the integrand is oscillatory ( as can be seen from the phase plot ) and non - decaying ( as can be seen from the magnitude plot ) : formally , just as with the infinite summation over eigenmodes in sec .
[ sec : eigenmode - summation ] , one must integrate over an infinite bandwidth , regularized in some way ( e.g. by the nyquist frequency placing an upper bound on @xmath1 for a finite grid ) , where the oscillations almost entirely cancel to leave a tiny remainder ( the force ) .
( any physical materials must cease to polarize as @xmath80 where the susceptibility vanishes @xcite , which will make the force contributions eventually vanish as @xmath80 even in 1d , but a very wide - bandwidth oscillatory integral is still required . )
this is a disaster for any numerical method even when one is only integrating an analytical expression such as the lifshitz formula , mere roundoff errors are a severe difficulty for real @xmath1 . along the imaginary-@xmath1 axis , on the other hand ( or any sufficiently vertical contour ) , @xmath68 is exponentially decaying and mostly non - oscillatory an ideal situation for numerical integration .
therefore , in order for classical em solvers to be used for casimir problems , they need to be adapted to solve maxwell s equations at complex or imaginary @xmath1 .
although this sounds strange at first , the frequency - domain problem actually becomes numerically _ easier _ in every way at imaginary @xmath1 ; this is discussed in more detail in sec .
[ sub : imaginary - green ] .
in fact , one can even identify an exact mathematical equivalence between a particular complex-@xmath1 contour and a _ real_-frequency system where an _ artificial dissipation _ has been introduced , as discussed in sec .
[ sub : fdtd ] below using this trick , one can actually use classical em solvers with no modification at all , as long as they handle dissipative media . in any case
, one needs an integral over frequencies to compute a physically meaningful quantity , which means that solvers and material models , not to mention any physical intuition used for guidance , must be valid for more than just a narrow real-@xmath1 bandwidth ( unlike most problems in classical em ) .
numerically , it should be pointed out that the @xmath81 integrand is smooth and exponentially decaying , and so the @xmath72 integral can be approximated to high accuracy by an exponentially convergent quadrature ( numerical integration ) scheme using evaluations at relatively few points @xmath72 .
for example , one can use gauss laguerre quadrature @xcite , gaussian quadrature with an appropriate change of variables @xcite , or clenshaw curtis quadrature with an appropriate change of variables @xcite .
another , equivalent , viewpoint on casimir interactions is that they arise from geometry - dependent fluctuations of the electromagnetic fields @xmath82 and @xmath83 , which on average have some nonzero energy density and exert a force .
if we can compute these average fields , we can integrate the resulting energy density , stress tensors , and so on , to obtain the energy , force , or other quantities of interest .
the good news is that there is a simple expression for those fluctuations in terms of the * fluctuation dissipation * theorem of statistical physics : the correlation function of the fields is related to the corresponding _ classical _ green s function @xcite .
ultimately , this means that any standard classical em technique to compute green s functions ( fields from currents ) can be applied to compute casimir forces , with the caveat that the techniques must be slightly modified to work at imaginary or complex frequencies as described below .
the temperature-@xmath84 correlation function for the fluctuating electric field at a given frequency @xmath1 is given by @xcite : @xmath85\coth(\hbar\omega/2k_{\mbox{b}}t),\label{eq : ecorr - real}\ ] ] where @xmath86 is the classical dyadic `` photon '' green s function , proportional from a dipole current @xmath87 is @xmath88 .
] to the relationship between an electric - dipole current in the @xmath89 direction at @xmath90 to the electric field at @xmath37 , and solves @xmath91\vec{g}_{k}^{e}(\omega,\vec{x},\vec{x}')=\delta^{3}(\vec{x}-\vec{x}')\hat{e}_{k},\label{eq : green - real}\ ] ] where @xmath70 is the electric permittivity tensor , @xmath92 is the magnetic permeability tensor , and @xmath93 is a unit vector in direction @xmath89 .
similarly , the magnetic - field correlation function is @xmath94\coth(\hbar\omega/2k_{\mbox{b}}t).\label{eq : hcorr - real}\ ] ] the magnetic green s function @xmath95 can be defined in two essentially equivalent ways .
the first is as derivatives @xmath96 of the electric green s function @xmath97 , where @xmath98 and @xmath99 denote derivatives with respect to @xmath37 and @xmath90 ( @xmath99 acting to the left ) , respectively @xcite . the second way to define
@xmath95 is proportional to the magnetic field in response to a magnetic - dipole current , analogous to ( [ eq : green - real]):@xmath100\vec{g}_{k}^{h}(\omega,\vec{x},\vec{x}')=\delta^{3}(\vec{x}-\vec{x}')\hat{e}_{k},\label{eq : green - real - h}\ ] ] which can be more convenient for numerical calculation @xcite .
these two definitions are related @xcite by @xmath101 ( with @xmath102 being the @xmath103 identity matrix ) , into ( [ eq : green - real - h ] ) , with @xmath104 denoting @xmath105 and @xmath92 or @xmath106 denoting @xmath107 or @xmath108 , respectively .
in particular , @xmath109(\frac{1}{\omega^{2}}\frac{1}{\mu}\nabla\times\mathbf{g}^{e}\times\nabla'\frac{1}{\mu'}-\frac{1}{\omega^{2}\mu'}\delta)$ ] yields @xmath110\times\nabla'\frac{1}{\mu'}-\nabla\times\frac{1}{\omega^{2}\mu'\varepsilon}\nabla\times\delta+\delta$ ] , which via ( [ eq : green - real ] ) gives @xmath111 as desired , where in the last step we have used the fact that @xmath112 [ since @xmath113 is antisymmetric under transposition and @xmath114 . ] where the second ( diagonal ) term has no effect on energy differences or forces below and is therefore irrelevant .
now , these equations are rather nasty along the real-@xmath1 axis : not only will there be poles in @xmath115 just below the axis corresponding to lossy modes , but in the limit where the dissipative losses vanish ( @xmath70 and @xmath92 become real ) , the combination of the poles approaching the real axis with the @xmath116 in the correlation function results in a delta function at each pole$ ] , viewed as a distribution , yields @xmath117 @xcite . ] and integrals of the correlation functions turn into sums over modes as in sec .
[ sec : eigenmode - summation ] .
however , the saving grace , as pointed out in sec .
[ sec : complex - frequency ] , is that green s functions are _ causal _ , allowing us to transform any integral over all real fluctuation frequencies into an integral over _ imaginary _ fluctuation frequencies @xmath71 .
the @xmath118 factor has poles that alter this picture , but we will eliminate those for now by considering only the @xmath119 case where @xmath120 , returning to nonzero temperatures in sec . [
sec : finite - temperature ] . in particular , to compute the casimir energy @xmath45
, we merely integrate the classical energy density in the em field @xcite over all positions and all fluctuation frequencies , wick - rotated to an integral over imaginary frequencies , resulting in the expression : @xmath121d^{3}\vec{x},\label{eq : u - mean - eh}\ ] ] where we have simplified to the case of isotropic materials ( scalar @xmath70 and @xmath92 ) . at thermodynamic equilibrium , this expression remains valid even for arbitrary dissipative / dispersive media thanks to a direct equivalence with a path - integral expression @xcite , which is not obvious from the classical viewpoint in which the energy density is usually only derived in the approximation of negligible absorption @xcite .
( thanks to the relationship between the green s function and the local density of states @xcite , there is also a direct equivalence between this energy integral and eigenmode summation @xcite . ) in the common case where @xmath92 has negligible frequency dependence ( magnetic responses are usually negligible at the short wavelengths where casimir forces are important , so that @xmath122 ) , we can use the identity , and from the identity after ( [ eq : green - real - h ] ) we know that @xmath123 .
however , because @xmath113 is self - adjoint @xcite , we can integrate by parts to move @xmath113 from the first argument / index of @xmath124 to the second , obtaining @xmath125 from the first term under the integral .
( here , we employ the fact that @xmath124 is real - symmetric at imaginary @xmath71 , from sec . [
sub : imaginary - green ] , to apply ( [ eq : green - imag ] ) to the second index / argument instead of the first . )
this cancels the other delta from @xmath126 and leaves @xmath127 , giving @xmath128 as desired . ]
that @xmath129 for fields at any given frequency @xcite to simplify this expression to @xcite : @xmath130 here , the zero - temperature imaginary - frequency mean - square electric field is given by : @xmath131 where @xmath132 denotes the trace @xmath133 and the @xmath116 has disappeared compared to ( [ eq : ecorr - real ] ) because @xmath134 ) is real and the @xmath116 cancels the @xmath135 in @xmath136 .
equation ( [ eq : ee - imag ] ) may at first strike one as odd , because one is evaluating the green s function ( field ) at @xmath37 from a source at @xmath37 , which is formally infinite .
this is yet another instance of the formal infinities that appear in casimir problems , similar to the infinite sum over modes in sec .
[ sec : eigenmode - summation ] . in practice
, this is not a problem either analytically or numerically .
analytically , one typically regularizes the problem by subtracting off the vacuum green s function ( equivalent to only looking at the _ portion _ of the fields at @xmath37 which are reflected off of inhomogeneities in @xmath70 or @xmath92 ) @xcite . numerically , in an fd or fem method with a finite grid , the green s function is everywhere finite ( the grid is its own regularization ) @xcite . in a bem ,
the green s function is explicitly written as a sum of the vacuum field and scattered fields , so the former can again be subtracted analytically @xcite . as in sec .
[ sec : eigenmode - summation ] , these regularizations do not affect physically observable quantities such as forces or energy differences , assuming rigid - body motion .
this imaginary - frequency green s function is actually a remarkably nice object .
wick - rotating eq .
( [ eq : green - real ] ) , it satisfies : @xmath137\vec{g}_{k}^{e}(i\xi,\vec{x},\vec{x}')=\delta^{3}(\vec{x}-\vec{x}')\hat{e}_{k}.\label{eq : green - imag}\ ] ] because of causality , it turns out that @xmath70 and @xmath92 are strictly real - symmetric and positive - definite ( in the absence of gain ) along the imaginary - frequency axis , even for dissipative / dispersive materials @xcite .
furthermore , the operator @xmath138 is real - symmetric positive - semidefinite for a positive - definite real - symmetric @xmath92 @xcite .
thus , the entire bracketed operator @xmath139 $ ] in eq .
( [ eq : green - imag ] ) is _ real - symmetric positive - definite _ for @xmath140 , which lends itself to some of the best numerical solution techniques ( cholesky decomposition @xcite , tridiagonal qr @xcite , conjugate gradients @xcite , and rayleigh - quotient methods @xcite ) .
( this definiteness is also another way of seeing the lack of poles or oscillations for @xmath71 . )
it follows that the integral operator whose kernel is @xmath124 , i.e. the inverse of the @xmath139 $ ] operator in eq .
( [ eq : green - imag ] ) , is also real - symmetric positive - definite , which is equally useful for integral - equation methods . in vacuum ,
the 3d real-@xmath1 green s function @xmath141 @xcite is wick - rotated to @xmath142 , an _ exponentially decaying , non - oscillatory _ function .
this is yet another way of understanding why , for @xmath71 , there are no interference effects and hence no `` modes '' ( poles in @xmath115 ) , and integrands tend to be non - oscillatory and exponentially decaying ( as @xmath143 , @xmath115 becomes exponentially short - ranged and does not `` see '' the interacting objects , cutting off the force contributions ) .
( it also means , unfortunately , that a lot of the most interesting phenomena in classical em , which stem from interference effects and resonances , may have very limited consequences for casimir interactions . )
one other property we should mention is that the operator becomes semidefinite for @xmath144 , with a nullspace encompassing any static field distribution ( @xmath145 for any scalar @xmath146 ) .
this corresponds to the well - known singularity of maxwell s equations at zero frequency @xcite , where the electric and magnetic fields decouple @xcite .
since we eventually integrate over @xmath72 , the measure - zero contribution from @xmath144 does not actually matter , and one can use a quadrature scheme that avoids evaluating @xmath144 . however , in the nonzero - temperature case of sec . [
sec : finite - temperature ] one obtains a sum over discrete-@xmath72 contributions , in which case the zero - frequency term is explicitly present . in this case , @xmath144 can be interpreted if necessary as the limit @xmath147 ( which can be obtained accurately in several ways , e.g. by richardson extrapolation @xcite , although some solvers need special care to be accurate at low frequency @xcite ) ; note , however , that there has been some controversy about the zero - frequency contribution in the unphysical limit of perfect / dissipationless metals @xcite . in practice , one often wants to know the casimir force ( or torque ) on an object rather than the energy density . in this case , instead of integrating an electromagnetic energy density over the volume , one can integrate an electromagnetic _ stress tensor _ over a surface enclosing the object in question , schematically depicted in fig .
[ fig : stress - schematic ] @xcite . schematic depiction of two objects whose casimir interaction is desired .
one computational method involves integrating a mean stress tensor around some closed surface ( dashed red line ) surrounding an object , yielding the force on that object . ]
the mean stress tensor for the casimir force is @xcite : @xmath148\\ + \mu(\vec{x},\omega)\left[\langle h_{j}(\vec{x})h_{k}(\vec{x})\rangle-\frac{\delta_{jk}}{2}\sum_{\ell}\langle h_{\ell}(\vec{x})^{2}\rangle_{\omega}\right].\label{eq : stress}\end{gathered}\ ] ] as above , the field correlation functions are expressed in terms of the classical green s function , and the integral of the contributions over all @xmath1 is wick - rotated to imaginary frequencies @xmath71 : @xmath149 with ( zero - temperature ) correlation functions @xmath150 @xmath151 corresponding to the fields on the stress - integration surface in response to currents placed on that surface . to compute a casimir torque around an origin @xmath152 ,
one instead uses @xmath153 @xcite .
the derivation of this stress tensor ( [ eq : stress ] ) is not as straightforward as it might at first appear .
if the stress - integration surface lies entirely in vacuum @xmath154 and @xmath122 , then one can interpret ( [ eq : stress ] ) as merely the ordinary em stress tensor from the microscopic maxwell equations @xcite , albeit integrated over fluctuations .
if the stress - integration surface lies in a dispersive / dissipative medium such as a fluid , however , then the classical em stress tensor is well known to be problematic @xcite and ( [ eq : stress ] ) may seem superficially incorrect .
however , it turns out that these problems disappear in the context of thermodynamic equilibrium , where a more careful free - energy derivation of the casimir force from fluctuations indeed results in equation ( [ eq : stress ] ) @xcite , is not neglected , then there is an additional @xmath155 term in ( [ eq : stress ] ) resulting from fluctuations in the density @xmath156 @xcite . ] which has also proved consistent with experiments @xcite .
note also that , while ( [ eq : stress ] ) assumes the special case of isotropic media at @xmath37 , it can still be used to evaluate the force on objects made of anisotropic materials , as long as the stress - integration surface lies in an isotropic medium ( e.g. vacuum or most fluids ) .
this formulation is especially important for methods that use an iterative solver for the green s functions as discussed below , because it only requires solving for the response to currents on the stress - integration surface , rather than currents at every point in space to integrate the energy density , greatly reducing the number of right - hand sides to be solved for the linear equation ( [ eq : green - imag ] ) @xcite ; additional reductions in the number of right - hand sides are described in sec .
[ sub : spatial - integration - basis ] . in order to determine the casimir energy or force ,
one evaluates the green s function by solving eq .
( [ eq : green - imag ] ) and then integrates the appropriate energy / force density over volume / surface and over the imaginary frequency @xmath72 .
the central numerical problem is then the determination of the green s function by solving a set of linear equations corresponding to ( [ eq : green - imag ] ) , and probably the simplest approach is based on a finite - difference ( fd ) basis : space is divided into a uniform grid with some resolution @xmath13 , derivatives are turned into differences , and one solves ( [ eq : green - imag ] ) by some method for every desired right - hand side . in classical em ( typically finding the field from a given current at real @xmath1 ) , this is known as a finite - difference frequency - domain ( fdfd ) method , and has been widely used for many years @xcite .
for example , in one dimension for @xmath157-directed currents / fields with @xmath158 , ( [ eq : green - imag ] ) becomes @xmath159 .
if we approximate @xmath160 , then the corresponding finite - difference equation , with a standard center - difference approximation for @xmath161 @xcite , is @xmath162 replacing the @xmath163 with a discrete equivalent at @xmath164 .
equation ( [ eq : green - imag-1d - fd ] ) is a tridiagonal system of equations for the unknowns @xmath165 .
more generally , of course , one has derivatives in the @xmath49 and @xmath157 directions and three unknown @xmath166 ( or @xmath82 ) components to determine at each grid point . as mentioned in sec .
[ sub : choices - of - basis ] , it turns out that accurate center - difference approximations for the @xmath167 operator in three dimensions are better suited to a `` staggered '' grid , called a _ yee _
grid @xcite , in which different field components are discretized at points slightly offset in space : e.g. , @xmath168\delta x , n_{y}\delta y , n_{z}\delta z)$ ] , @xmath169\delta y , n_{z}\delta z)$ ] , , and @xmath170\delta z)$ ] for the @xmath82 field components .
note that any arbitrary frequency dependence of @xmath70 is trivial to include , because in frequency domain one is solving each @xmath72 separately , and a perfect electric conductor is simply the @xmath171 limit .
one must , of course , somehow truncate the computational domain to a finite region of space in order to obtain a finite number @xmath10 of degrees of freedom .
there are many reasonable ways to do this because casimir interactions are rapidly decaying in space ( force @xmath172 or faster with distance @xmath61 in @xmath173 dimensions , at least for zero temperature ) .
one could simply terminate the domain with dirichlet or periodic boundary conditions , for example , and if the cell boundaries are far enough away from the objects of interest then the boundary effects will be negligible ( quite different from classical em problems at real @xmath1 ! ) @xcite . in classical em ,
one commonly uses the more sophisticated approach of a * perfectly matched layer * ( pml ) , an artificial reflectionless absorbing material placed adjacent to the boundaries of the computational domain to eliminate outgoing waves @xcite .
mathematically , a pml in a direction @xmath52 is equivalent to a complex `` coordinate stretching '' @xmath174 for an artificial pml `` conductivity '' @xmath175 @xcite , where the @xmath176 factor is introduced to give an equal attenuation rate at all frequencies .
however , at imaginary frequencies @xmath71 , a pml therefore results simply in a _ real _
coordinate stretching @xmath177 : in a pde with decaying , non - oscillatory solutions ( such as the imaginary-@xmath1 maxwell equations ) , it is well known that a reasonable approach to truncating infinite domains is to perform a ( real ) coordinate transformation that compresses space far away where the solution is small @xcite .
a convenience of maxwell s equations is that any coordinate transformation ( real or complex ) can be converted merely into a change of @xmath70 and @xmath92 @xcite , so any pml can be expressed simply as a change of materials while keeping the same pde and discretization @xcite .
such a center - difference scheme is nominally second - order accurate , with discretization errors that vanish as @xmath178 @xcite .
one can also construct higher - order difference approximations ( based on more grid points per difference ) . as a practical matter , however , the accuracy is limited instead by the treatment of material interfaces where @xmath70 changes discontinuously . if no special allowance is made for these interfaces , the method still converges , but its convergence rate is reduced by the discontinuity to @xmath179 @xcite ( unless one has @xmath82 polarization completely parallel to all interfaces so that there is no field discontinuity ) .
there are various schemes to restore second - order ( or higher ) accuracy by employing specialized fd equations at the interfaces @xcite , but an especially simple scheme involves unmodified fd equations with modified materials : it turns out that , if the discontinuous @xmath70 is smoothed in a particular way ( to avoid introducing first - order errors by the smoothing itself ) , then second - order accuracy can be restored @xcite .
discontinuities are dealt with in this way , however , one may still fail to obtain second - order accuracy if the geometry contains sharp corners , which limit the accuracy to @xmath180 for some @xmath181 @xcite .
this is an instance of _ darboux s principle _ : the convergence rate of a numerical method is generally limited by the strongest singularity in the solution that has not been explicitly compensated for @xcite . ] given the fd equations , one must must then choose a solution technique to solve the resulting linear equations @xmath32 , where @xmath37 is the green s function ( or field ) , @xmath182 is the delta - function ( or current ) right - hand side , and @xmath36 is the discretized @xmath183 operator .
note that @xmath36 is a real - symmetric positive - definite matrix at imaginary frequencies , as discussed in sec .
[ sub : imaginary - green ] . because @xmath36 is sparse [ only @xmath38 nonzero entries ] , one can utilize a sparse - direct cholesky factorization @xmath184 ( @xmath185 is upper - triangular ) @xcite ( for which many software packages are available @xcite ) . given this factorization ,
any right - hand side can be solved quickly by backsubstitution , so one can quickly sum the energy density over all grid points ( essentially computing the trace of @xmath56 ) to find the casimir energy , or alternatively sum the stress tensor over a stress - integration surface to find the force . precisely such a sparse - direct fd method for the casimir energy was suggested by pasquali and maggs @xcite , albeit derived by a path - integral @xmath186 expression that is mathematically equivalent to summing the energy density ( [ eq : u - mean - e ] ) @xcite .
the alternative is an iterative technique , and in this case @xmath36 s hermitian definiteness means that an ideal krylov method , the conjugate - gradient method @xcite can be employed @xcite .
the conjugate - gradient method requires @xmath38 storage and time per iteration , and in the absence of preconditioning requires a number of iterations in @xmath173 dimensions proportional to the diameter @xmath187 of the grid for each right - hand side @xcite .
the stress - tensor approach reduces the number of right - hand sides to be solved compared to energy - density integration : one only needs to evaluate the green s function for sources on a stress - integration surface , which has @xmath188 points in @xmath173 dimensions .
this gives a total time complexity of @xmath189 for an unpreconditioned iterative method ; an ideal multigrid preconditioner can in principle reduce the number of iterations to @xmath190 @xcite ( when @xmath10 is increased by improving spatial resolution ) , yielding an @xmath191 time complexity .
substantial further improvements are obtained by realizing that one does not , in fact , need to sum over every point on the stress - integration surface , instead switching to a different spatial integration scheme described in sec .
[ sub : spatial - integration - basis ] . in some sense ,
a volume discretization such as an fd method is too general : in most physical situations , the medium is piecewise - constant , and one might want to take advantage of this fact . in particular , for the basic problem of finding the field in response to a current source at a given frequency , one can instead use a surface integral - equation approach : the unknowns are _ surface currents _ on the interfaces between homogeneous materials , and one solves for the surface currents so that the total field ( source + surface currents ) satisfies the appropriate boundary conditions at the interfaces @xcite .
for example , in the case of a perfect electric conductor , the surface - current unknowns can be the physical electric currents @xmath192 at the interface , and the boundary condition is that of vanishing tangential @xmath82 field . in the case of permeable media @xmath70 and @xmath92 , the physical ( bound )
currents are volumetric within the medium [ e.g. , the electric bound current is @xmath193 , not surface currents @xcite .
however , it turns out that one can introduce _ fictitious _ surface electric and magnetic currents at all interfaces to provide enough degrees of freedom to satisfy the boundary condition of continuous tangential @xmath82 and @xmath83 , and thus to fully solve maxwell s equations .
the application of this _ _ equivalence principle _ _ to obtain surface integral equations for bem is known as the pmchw approach ( poggio , miller , chang , harrington , and wu ) @xcite . in either case , one has surface ( electric and/or magnetic ) currents @xmath194 , plus an external current source @xmath192 [ e.g. , the right - hand side of ( [ eq : green - real ] ) ] , so one can express the @xmath82 or @xmath83 field at any point @xmath37 as a convolution of @xmath195 with the _ analytically known _
green s function @xmath196 of the corresponding _ homogeneous _ medium at @xmath37 . in bem
, one expresses @xmath194 , in turn , as a sum of _ localized _ basis functions @xmath197 associated with some discrete mesh approximation of the surface .
for example , fig .
[ fig : crossed - capsules ] depicts a standard triangular - type mesh of two objects , where there is a localized basis function @xmath197 ( inset ) associated with each _ edge _ of this mesh such that @xmath197 is nonzero only on the adjacent two triangles @xcite ; this is the rwg basis mentioned in sec .
[ sub : choices - of - basis ] .
example triangular mesh of the surfaces of two objects for a bem solver @xcite . associated with each edge
@xmath89 is an `` rwg '' basis function @xmath197 @xcite , schematically represented in the inset , which vanishes outside the adjacent two triangles . ]
abstractly , the resulting equations for the fields could then be written in the following form : @xmath198 where @xmath199 denotes convolution with the ( dyadic ) analytical green s function of the homogeneous medium at @xmath37 , and @xmath200 are the unknown coefficients of each basis function .
contribute to this sum . ]
( more generally , @xmath201 could be replaced by any arbitrary incident field , regardless of how it is created . ) in a galerkin method ( see sec . [
sub : choices - of - basis ] ) , one obtains @xmath10 equations for the @xmath10 unknowns @xmath200 by taking the inner product of both sides of this equation ( substituted into the appropriate boundary condition ) with the same basis functions @xmath202 ( since they work just as well as a basis for the tangential field as for the tangential surface currents ) .
this ultimately results in a set of linear equations @xmath203 , where the matrix @xmath36 multiplying the unknown coefficients @xmath200 is given by @xmath204 [ for the case of a perfect conductor with a vanishing tangential @xmath82 , the right - hand - side @xmath205 is given by @xmath206 .
] one then solves the linear system for the unknown coefficients @xmath200 , and hence for the unknown surface currents @xmath194 .
implementing this technique is nontrivial because the @xmath207 integrands ( [ eq : bem - matrix ] ) are singular for @xmath208 or for @xmath209 adjacent to @xmath89 , necessitating specialized quadrature techniques for a given form of @xmath210 @xcite , but substantial guidance from the past decades of literature on the subject is available . given these currents , one can then evaluate the electric or magnetic field at any point @xmath37 , not just on the surface , by evaluating eq .
( [ eq : bem - abstract ] ) at that point .
in particular , one can evaluate the field correlation functions via the fluctuation - dissipation theorem ( [ eq : ecorr - real ] ) : @xmath211 is given in terms of the electric field in the @xmath209 direction at @xmath37 from a delta - function current @xmath192 in the @xmath89 direction at @xmath37 .
of course , as noted previously , this is infinite because the @xmath212 term ( the field from the delta function ) blows up at @xmath37 , but in the casimir case one is only interested in the _ change _ of the correlation functions due to the geometry
so , one can use the standard trick @xcite of subtracting the vacuum contribution @xmath201 and only computing the surface - current contribution @xmath213 to the field at @xmath37 . in this way
, one can compute the stress tensor , the energy density , and so on , as desired . as explained in sec .
[ sec : complex - frequency ] , the integral of contributions over all frequencies is best performed at imaginary frequencies , so all of the above must use @xmath71 . this only has the effect of wick - rotating the homogeneous - medium dyadic green s function @xmath210 to the @xmath214 imaginary - frequency green s function .
this makes the problem _ easier _ , in principle .
first , the exponential decay cuts off long - range interactions , making fast - solver techniques ( see sec .
[ sub : solution - techniques ] ) potentially even more effective .
second , the matrix @xmath36 is now real - symmetric and positive - definite , which allows the use of more efficient linear solvers as noted previously .
fortunately , the @xmath215 singularity of @xmath210 is the same at real and imaginary frequencies , allowing existing techniques for the integration of ( [ eq : bem - matrix ] ) to be leveraged . at first glance
, this approach seems most straightforwardly applicable to the stress - tensor technique , as suggested in ref .
@xcite : one uses the bem solution to evaluate the mean stress tensor @xmath216 on any integration surface around a body , integrating via some quadrature technique to obtain the force .
if one uses a dense - direct solver ( when @xmath10 is not too big ) , the cholesky factorization of @xmath36 can be computed once for a given @xmath72 and then many right - hand sides can be solved quickly via backsubstitution @xcite in order to integrate @xmath216 over the stress - integration surface .
precisely such a dense - direct bem stress - tensor method was recently demonstrated to compute casimir forces in two and three dimensions @xcite . as described in sec .
[ sub : solution - techniques ] , fast - solver techniques can be applied to multiply @xmath36 by a vector in @xmath47 time with @xmath38 storage ; given a good preconditioner , this implies that an iterative method such as conjugate - gradient ( applicable since @xmath36 is real - symmetric positive - definite ) could find @xmath216 at a single @xmath37 and @xmath72 in @xmath47 time .
the remaining question is the number of points required for the surface integral of @xmath216 , which depends on _ why _ one is increasing @xmath10 : either to increase accuracy for a fixed geometry or to increase the complexity of the geometry for a fixed accuracy . in the former case
, the smoothness of @xmath216 in @xmath37 means that exponentially convergent quadrature techniques are applicable , which converge much faster than the ( polynomial ) bem basis for the surface currents , so that ultimately the number of stress - quadrature points by a sum @xmath217 over quadrature points @xmath218 with weights @xmath219 .
there are many techniques for the selection of these points and weights , and in general one can obtain an error that decreases exponentially fast with the number of points for analytic integrands @xcite .
multidimensional quadrature , sometimes called _ cubature _ , should be used to integrate the stress tensor over a 2d surface , and numerous schemes have been developed for low - dimensional cubature @xcite ( including methods that adaptively place more quadrature points where they are most needed @xcite ) . for spherical integration surfaces ( or surfaces that can be smoothly mapped to spheres ) , specialized methods are available @xcite .
] should be independent of @xmath10 and the overall complexity becomes @xmath47 . in the latter case , for a fixed accuracy and increasingly complex geometry ( or smaller feature sizes ) , it appears likely that the number of stress - quadrature points will increase with @xmath10 , but detailed studies of this scaling are not yet available .
it turns out that this bem approach is closely related to the bem path - integral approach described in sec .
[ sub : bem - path ] . both approaches
end up solving linear equations with exactly the same matrix @xmath36 of eq .
( [ eq : bem - matrix ] ) , with the same degrees of freedom . the path - integral approach shows , however , that this same matrix can be applied to compute the casimir interaction energy as well as the force , with comparable computational cost for dense solvers .
moreover , as explained below , expressing the force in terms of the derivative of the path - integral energy results in a trace expression that is conceptually equivalent to integrating a stress tensor over the surface of an object , where the number of `` quadrature points '' is now exactly equal to @xmath10 .
an unanswered question , at this point , is whether a fast solver can be more efficiently ( or more easily ) exploited in the stress - tensor approach or in the path - integral approach .
there are of course , many other frequency - domain techniques from classical em that could potentially be used to solve for the green s function and hence the energy / force density . for example , one could use spectral integral - equation methods , such as multipole expansions for spheres and cylinders @xcite , to compute responses to currents , although the advantages of this approach compared to the spectral path - integral approach in sec .
[ sub : spectral - path - integral ] are unclear .
one can also solve the pde formulation of the green s function ( [ eq : green - imag ] ) using a finite - element ( fem ) approach with some general mesh ; in principle , existing fem techniques from classical em @xcite are straightforwardly applicable .
one subtlety that arises in fem methods with a nonuniform resolution is the regularization , however @xcite . in principle , as mentioned above , one needs to subtract the vacuum green s function contribution from the field correlation functions in order to get a physical result [ since the vacuum green s function @xmath220 diverges as @xmath221 , although the divergence is cut off by the the finite mesh resolution ] . with a uniform mesh ,
this vacuum contribution is the same everywhere in the mesh and hence automatically integrates to zero in the force ( when the stress tensor is integrated over a closed surface or the energy is differentiated ) . for a nonuniform mesh
, however , the vacuum contribution varies at different points in space with different resolution , so some `` manual '' regularization seems to be required ( e.g. , subtracting a calculation with the same mesh but removing the objects ) .
these possibilities currently remain to be explored for casimir physics .
casimir effects are fundamentally broad - bandwidth , integrating contributions of fluctuations at all frequencies ( real or imaginary ) , although the imaginary - frequency response is dominated by a limited range of imaginary frequencies . in classical em , when a broad - bandwidth response is desired , such as a transmission or reflection spectrum from some structure , there is a well - known alternative to computing the contributions at each frequency separately instead
, one can simulate the same problem in _ time _
, fourier - transforming the response to a short pulse excitation in order to obtain the broad - bandwidth response in a single _ time - domain _ simulation @xcite .
the same ideas are applicable to the casimir problem , with a few twists , yielding a practical method @xcite that allows casimir calculations to exploit off - the - shelf time - domain solvers implementing the standard _ finite - difference time - domain _ ( fdtd ) method @xcite .
there are two key components of this approach @xcite : first , converting the frequency integral to a time integral and , second , finding a time - domain equivalent of the complex - frequency idea from sec .
[ sec : complex - frequency ] .
as reviewed above , the mean fluctuations in the fields , such as @xmath222 , can be expressed in terms of the fields at @xmath37 from a frequency-@xmath1 current at @xmath37 .
if , instead of a frequency-@xmath1 current , one uses a current with @xmath223 time dependence , it follows by linearity of ( [ eq : green - real ] ) that the fourier transform of the resulting fields must yield exactly the same @xmath222 .
roughly , the procedure could be expressed as follows : first , we compute some function @xmath224 of the time - domain fields from a sequence of simulations with @xmath223 sources , e.g. where @xmath224 is the result of spatially integrating the fields making up the mean stress tensor @xmath225 [ noting that each point @xmath37 involves several separate @xmath223-response simulations ] .
second , we fourier transform @xmath224 to obtain @xmath226 .
third , we obtain the force ( or energy , etcetera ) by integrating @xmath227 with appropriate frequency - weighting factor @xmath228 ( which may come from the frequency dependence of @xmath70 in @xmath216 , a jacobian factor from below , etcetera ) . at this point , however , it is clear that the fourier transform of @xmath229 was entirely unnecessary : because of the unitarity of the fourier transform ( the plancherel theorem ) , @xmath230 .
that is , we can compute the force ( or energy .
etcetera ) by starting with @xmath223 sources and simply integrating the response @xmath224 in time ( accumulated as the simulation progresses ) multiplied by some ( precomputed , geometry - independent ) kernel @xmath231 ( which depends on temperature if the @xmath118 factor is included for @xmath232 ) .
the details of this process , for the case of the stress tensor , are described in refs .
@xcite .
although it turns out to be possible to carry out this time - integration process as - is , we again find that a transformation into the complex - frequency plane is desirable for practical computation ( here , to reduce the required simulation time ) @xcite .
transforming the _ frequency _ in a _
time_-domain method , however , requires an indirect approach .
the central observation is that , in the equation ( [ eq : green - real ] ) for the electric - field green s function @xmath124 , the frequency only appears explicitly in the @xmath233 term , together with @xmath70 .
so , any transformation of @xmath1 can equivalently be viewed as a transformation of @xmath70 .
in particular suppose that we wish to make some transformation @xmath234 to obtain an @xmath1 in the upper - half complex plane , where @xmath72 is a real parameter ( e.g. @xmath71 for a wick rotation ) .
equivalently , we can view this as a calculation at a _ real _ frequency @xmath72 for a transformed _ complex material _ : @xmath235 where the transformed material is @xcite @xmath236 for example , a wick rotation @xmath237 is equivalent to operating at a real frequency @xmath72 with a material @xmath238 .
however , at this point we run into a problem : multiplying @xmath70 by @xmath239 yields exponentially growing solutions at negative frequencies @xcite , and this will inevitably lead to exponential blowup in a time - domain simulation ( which can not avoid exciting negative frequencies , if only from roundoff noise ) . in order to obtain a useful time - domain simulation
, we must choose a contour @xmath240 that yields a _ causal _ , _ dissipative _ material @xmath241 , and one such choice is @xmath242 for any constant @xmath243 @xcite .
this yields @xmath244 , where the @xmath245 term behaves exactly like an artificial _ conductivity _ added everywhere in space . in the frequency - domain picture , we would say from sec .
[ sec : complex - frequency ] that this @xmath240 contour will improve the computation by moving away from the real-@xmath1 axis , transforming the frequency integrand into something exponentially decaying and less oscillatory . in the time - domain picture
, the @xmath246 term adds a _ dissipation _ everywhere in space that causes @xmath224 to _ decay exponentially in time _ , allowing us to truncate the simulation after a short time .
as long as we include the appropriate jacobian factor @xmath247 in our frequency integral , absorbing it into @xmath231 , we will obtain the _ same result _ in a much shorter time .
the computational details of this transformation are described in refs .
more generally , this equivalence between the casimir force and a relatively narrow - bandwidth real - frequency response of a dissipative system potentially opens other avenues for the understanding of casimir physics @xcite .
the end result is a computational method for the casimir force in which one takes an off - the - shelf time - domain solver ( real time / frequency ) , adds an artificial conductivity @xmath246 everywhere , and then accumulates the response @xmath224 to short pulses multiplied by a precomputed ( geometry independent ) kernel @xmath231 .
the most common time - domain simulation technique in classical em is the fdtd method @xcite .
essentially , fdtd works by taking the same spatial yee discretization as in the fdfd method above , and then also discretizing time with some time step @xmath248 .
the fields are then marched through time in steps of @xmath248 , where each time step requires @xmath38 work for @xmath10 spatial grid points . because the complex-@xmath1 contour is implemented entirely as a choice of materials @xmath241 , existing fdtd software can be used without modification to compute casimir forces , and one can exploit powerful existing software implementing parallel calculations , various dimensionalities and symmetries , general dispersive and anisotropic materials , pml absorbing boundaries , and techniques for accurate handling of discontinuous materials .
one such fdtd package is available as free / open - source software from our group @xcite , and we have included built - in facilities to compute casimir forces @xcite .
finally , we should mention a few techniques that accelerate the convergence and reduce the computational cost of the finite - difference approaches .
these techniques are not _ necessary _ for convergence , but they are simple to implement and provide significant efficiency benefits .
the simplest technique is extrapolation in @xmath13 : since the convergence rate of the error with the spatial resolution @xmath13 is generally known _ a priori _ , one can fit the results computed at two or more resolutions in order to extrapolate to @xmath249 .
the generalization of this approach is known as _ richardson extrapolation _
@xcite , and it can essential increases the convergence order cheaply , e.g. , improving @xmath179 to @xmath178 @xcite .
second , suppose one is computing the force between two objects @xmath36 and @xmath250 surrounded by a homogeneous medium .
if one of the objects , say @xmath250 , is removed , then ( in principle ) there should be no net remaining force on @xmath36 .
however , because of discretization asymmetry , a computation with @xmath36 alone will sometimes still give a small net force , which converges to zero as @xmath249 .
if this `` error '' force is subtracted from the @xmath36@xmath250 force calculation , it turns out that the net error is reduced .
more generally , the error is greatly reduced if one computes the @xmath36@xmath250 force and then subtracts the `` error '' forces for @xmath36 alone and for @xmath250 alone , tripling the number of computations but greatly reducing the resolution that is required for an accurate result @xcite .
third , when integrating the stress tensor @xmath251 over @xmath37 to obtain the net force ( [ eq : stress - integral ] ) , the most straightforward technique in fd is to simply sum over all the grid points on the integration surface recall that each point @xmath37 requires a linear solve ( a different right - hand side ) in frequency domain , or alternatively a separate time - domain simulation ( a separate current pulse ) .
this is wasteful , however , because @xmath251 is conceptually smoothly varying in space
if one could evaluate it at arbitrary points @xmath37 ( as is possible in the bem approach ) , an exponentially convergent quadrature scheme could be exploited to obtain an accurate integral with just a few @xmath37 s .
this is not directly possible in an fd method , but one can employ a related approach . if the integration surface is a box aligned with the grid , one can expand the fields on each side of the box in a cosine series ( a discrete cosine transform , or dct , since space is discrete)this generally converges rapidly , so only a small number terms from each side are required for an accurate integration .
but instead of putting in point sources , obtaining the responses , and expanding the response in a cosine series , it is equivalent ( by linearity ) to put in cosine sources directly instead of point sources .
[ mathematically , we are exploiting the fact that a delta function can be expanded in any orthonormal basis @xmath252 over the surface , such as a cosine series , via : @xmath253 . substituting this into the right - hand side of ( [ eq : green - imag ] ) ,
each @xmath252 acts like a current source and @xmath254 scales the result , which is eventually integrated over @xmath90 . ]
the details of this process and its convergence rate are described in ref .
@xcite , but the consequence is that many fewer linear systems ( fewer right - hand sides ) need be solved ( either in frequency or time domain ) than if one solved for the stress tensor at each point individually .
another formulation of casimir interactions is to use a derivation based on path integrals .
although the path - integral derivation itself is a bit unusual from the perspective of classical em , and there are several slightly different variations on this idea in the literature , the end result is straightforward : casimir energies and forces are expressed in terms of log determinants and traces of classical scattering matrices @xcite , or similarly the interaction matrices ( [ eq : bem - matrix ] ) that arise in bem formulations @xcite . here ,
we omit the details of the derivations and focus mainly on the common case of piecewise - homogeneous materials , emphasizing the relationship of the resulting method to surface - integral equations from classical em via the approach in ref . @xcite .
path integrals relate the casimir interaction energy @xmath45 of a given configuration to a functional integral over all possible vector - potential fields @xmath255 . assuming piecewise - homogeneous materials , the constraint that the fields in this path integral must satisfy the appropriate boundary conditions
can be expressed in terms of auxiliary fields @xmath192 at the interfaces ( a sort of lagrange multiplier ) @xcite . , resulting in an expression equivalent to the sum over energy density in sec .
[ sec : fluctuation - dissipation ] @xcite and which in an fd discretization reduces in the same way to repeated solution of the greens - function diagonal at every point in space @xcite . ] at this point , the original functional integral over @xmath255 can be performed analytically , resulting in an energy expression involving a functional integral @xmath256 over only the auxiliary fields @xmath192 at each imaginary frequency @xmath72 , of the form ( at zero temperature ) : @xmath257 @xmath258 here , @xmath259 denotes @xmath260 when the objects are at infinite separation ( non - interacting ) , regularizing @xmath45 to just the ( finite ) interaction energy ( see also the chapter by s. j. rahi et al . in this volume for additional discussion of path integrals and casimir interactions . ) in the case of perfect electric conductors in vacuum , @xmath192 can be interpreted as a surface current on each conductor ( enforcing the vanishing tangential @xmath82 field ) , and @xmath261 is the vacuum green s function in the medium outside the conductors @xcite . for permeable media ( finite @xmath70 and @xmath92 ) ,
it turns out that a formulation closely related to the standard pmchw integral - equation model ( see sec . [
sub : bem - stress ] ) can be obtained : @xmath192 represents fictitious surface electric and magnetic currents on each interface ( derived from the continuity of the tangential @xmath82 and @xmath83 fields ) , with @xmath261 again being a homogeneous green s function ( with one @xmath260 factor for each contiguous homogeneous region ) @xcite . alternatively , because there is a direct correspondence between surface currents and the outgoing / scattered fields from a given interface , `` currents '' @xmath192 can be replaced by scattered fields , again related at different points @xmath37 and @xmath90 by the green s function of the homogeneous medium ; this is typically derived directly from a t - matrix formalism @xcite . here
, we will focus on the surface - current viewpoint , which is more common in the classical - em integral - equation community .
the path integral ( [ eq : path - integral ] ) is somewhat exotic in classical em , but it quickly reduces to a manageable expression once an approximate ( finite ) basis @xmath197 is chosen for the currents @xmath192 . expanding in this basis ,
@xmath262 and the functional integral @xmath263 is replaced by an ordinary integral over the basis coefficients @xmath264 .
equation ( [ eq : path - integral ] ) is then a gaussian integral that can be performed analytically to obtain @xmath265 for a proportionality constant @xmath266 @xcite , where @xmath267 is essentially the same as the bem matrix ( [ eq : bem - matrix ] ) , albeit here in an arbitrary basis . in the @xmath186 of ( [ eq : path - integral - energy ] )
, proportionality constants and exponents cancel , leaving : @xmath268d\xi.\label{eq : path - integral - energy - matrix}\ ] ] just as in sec .
[ eq : bem - matrix ] , the use of a real - symmetric positive - definite homogeneous green s function @xmath261 at imaginary frequencies means that @xmath269 is also real - symmetric and positive - definite , ensuring positive real eigenvalues and hence a real @xmath186 .
several further simplifications are possible , even before choosing a particular basis . for example
, let @xmath270 be the position of some object for which the force @xmath271 is desired .
the components @xmath272 of the force ( in direction @xmath273 ) can then be expressed directly as a trace @xcite : @xmath274 equivalently , this trace is the sum of eigenvalues @xmath275 of the generalized eigenproblem @xmath276 ; again , these @xmath275 are real because @xmath36 is real - symmetric positive - definite and @xmath277 is real - symmetric .
( if dense - direct solvers are used , computing @xmath278 via cholesky factorization is much more efficient than computing eigenvalues , however @xcite . )
the matrix @xmath36 can be further block - decomposed in the usual case where one is computing the interactions among two or more disjoint objects ( with disjoint surface currents @xmath192 ) .
for example , suppose that one has two objects @xmath279 and @xmath280 , in which case one can write @xmath281 where @xmath282 and @xmath283 couple currents on each object to other currents on the same object , and @xmath284 and @xmath285 couple currents on object 1 to object 2 and vice versa . in the limit of infinite separation for @xmath286 , one obtains @xmath287 while @xmath282 and @xmath283 are unchanged , and one can simplify the @xmath186 in ( [ eq : path - integral - energy - matrix ] ) to @xmath288=\log\det\left[i - a_{22}^{-1}a_{12}^{\mathrm{t}}a_{11}^{-1}a_{21}\right].\label{eq : logdet - block}\ ] ] computationally , only @xmath284 depends on the relative positions of the objects , and this simplification immediately allows several computations to be re - used if the energy or force is computed for multiple relative positions .
before we continue , it should be noted that there also exists a fundamentally different approach for evaluating a path - integral casimir formulation . instead of reducing the problem to surface / scattering unknowns and analytically integrating @xmath260 to obtain a matrix @xmath186 expression , it is possible to retain the original path - integral expression , in terms of a functional integral over vector potentials @xmath255 in the volume , and perform this functional integral numerically via monte - carlo methods @xcite .
this reduces to a monte - carlo integration of an action over all possible closed - loop paths ( `` worldlines '' ) , discretized into some number of points per path .
because this technique is so different from typical classical em computations , it is difficult to directly compare with the other approaches in this review .
evaluating its computational requirements involves a statistical analysis of the scaling of the necessary number of paths and number of points per path with the desired accuracy and the complexity of the geometry @xcite , which is not currently available .
a difficulty with this technique is that it has currently only been formulated for scalar fields with dirichlet boundary conditions , not for the true casimir force of vector electromagnetism .
one choice of basis functions @xmath197 for the path - integral expressions above is a spectral basis , and ( mirroring the history of integral equations in classical em ) this was the first approach applied in the casimir problem . with cylindrical objects , for example ,
the natural spectral basis is a fourier - series @xmath289 in the angular direction @xmath146 . for planar surfaces the natural choice is a fourier transform , for spheres
it is spherical harmonics @xmath290 ( or their vector equivalents @xcite ) , and for spheroids there are spheroidal harmonics @xcite .
equivalently , instead of thinking of surface currents expanded in a fourier - like basis , one can think of the scattered fields from each object expanded in the corresponding fourier - like basis ( e.g. plane , cylindrical , or spherical waves ) , in which case @xmath36 relates the incoming to outgoing / scattered waves for each object ; this has been called a `` scattering - matrix '' or `` t - matrix '' method and is the source of many pioneering results for casimir interactions of non - planar geometries @xcite .
even for nonspherical / spheroidal objects , one can expand the scattered waves in vector spherical harmonics @xcite , and a variety of numerical techniques have been developed to relate a spherical - harmonic basis to the boundary conditions on nonspherical surfaces @xcite .
these spectral scattering methods have their roots in many classical techniques for em scattering problems
@xcite ( see also the chapters by s. j. rahi et al . and a. lambrecht et al . in this volume for additional discussions of scattering techniques and casimir interactions . )
here , we will use the surface - current viewpoint rather than the equivalent scattered - wave viewpoint .
many simplifications occur in the interaction matrix @xmath36 of ( [ eq : bem - matrix - block ] ) for geometries with highly symmetrical objects and a corresponding spectral basis @xcite .
consider , for example , the case of spherical objects , with surface currents expressed in a vector spherical - harmonic basis ( spherical harmonics for two polarizations @xcite ) . in the interaction matrix @xmath291 , the convolution @xmath292 of a green s function @xmath166 with @xmath197 is known analytically : it is just the outgoing spherical wave produced by a spherical - harmonic current .
if @xmath202 is another spherical - harmonic current on the same sphere , then the orthogonality of the spherical harmonics means that the @xmath37 integral of @xmath293 against the spherical wave is zero unless @xmath208 .
thus , the self - interaction blocks @xmath282 and @xmath283 of ( [ eq : bem - matrix - block ] ) , with an appropriate normalization , are simply identity matrices .
the @xmath284 entries are given by the coupling of a spherical wave from @xmath197 on sphere 2 with a spherical - harmonic basis function @xmath202 on sphere 1 , but again this integral can be expressed analytically , albeit as an infinite series : the spherical wave from sphere 2 can be re - expressed in the basis of spherical waves centered on sphere 1 via known translation identities of spherical waves , and as a result @xmath284 takes the form of a `` translation matrix '' @xcite .
furthermore , if there are only two spheres in the problem , then their spherical harmonics can be expressed with respect to a common @xmath157 axis passing through the centers of the spheres , and a @xmath290 on sphere 1 will only couple with a @xmath294 on sphere 2 if @xmath295 , greatly reducing the number of nonzero matrix elements .
related identities are available for coupling cylindrical waves around different origins , expanding spherical / cylindrical waves in terms of planewaves for coupling to planar surfaces , and
so on @xcite .
as was noted in sec .
[ sub : choices - of - basis ] , such a spectral basis can converge exponentially fast if there are no singularities ( e.g. corners ) that were not accounted for analytically , and the method can even lend itself to analytical study . especially for cylinders and spheres ,
the method is simple to implement and allows rapid exploration of many configurations ; the corresponding classical `` multipole methods '' are common in classical em for cases where such shapes are of particular interest @xcite . on the other hand ,
as the objects become less and less similar to the `` natural '' shape for a given basis ( e.g. less spherical for spherical harmonics ) , especially objects with corners or cusps , the spectral basis converges more slowly @xcite . even for the interaction between two spheres or a sphere and a plate , as the two surfaces approach one another the multipole expansion will converge more slowly @xcite
conceptually , a spherical - harmonic basis has uniform angular resolution all over the sphere , whereas for two near - touching surfaces one would rather have more resolution in the regions where the surfaces are close ( e.g. by using a nonuniform bem mesh )
. this exponential convergence of a spectral ( spherical harmonic @xcite ) casimir calculation is depicted in fig .
[ fig : sphere - conv ] for the case of the casimir interaction energy @xmath45 between two gold spheres of radius @xmath296 m , for various surface - to - surface separations @xmath61 .
fractional error @xmath297 in the casimir interaction energy @xmath45 between two gold spheres of radius @xmath296 m , for various surface - surface separations @xmath61 , as a function of the maximum spherical - harmonic order @xmath298 of the spectral path - integral ( scattering - matrix / multipole ) method .
the error converges exponentially with @xmath298 , but the exponential rate slows as @xmath299 shrinks .
( calculations thanks to a. rodriguez . ) ] the error @xmath297 decreases exponentially with the maximum spherical - harmonic order @xmath298 [ corresponding to @xmath300 degrees of freedom for two spheres ] , but the exponential rate slows as @xmath299 decreases .
( on the other hand , for small @xmath299 a perturbative expansion or extrapolation may become applicable @xcite . ) in a bem ,
one meshes the interfaces , say into triangles , and uses a set of localized basis functions @xmath197 as discussed in sec .
[ sub : bem - stress ] . in this case , the interaction matrix @xmath36 that arises in the path - integral formulation is _ _ exactly the same as the interaction matrix that arises in classical bem methods ( albeit at an imaginary frequency ) , and is the same as the matrix @xmath36 that arises in a bem stress - tensor approach as described in sec . [ sub : bem - stress ] .
the main difference , compared to the stress - tensor approach , lies in how one _ uses _ the matrix @xmath36 : instead of solving a sequence of linear equations to find the mean stress tensor @xmath216 at various points on a surface around an object , one computes @xmath301 or @xmath302 $ ] to obtain the energy ( [ eq : path - integral - energy - matrix ] ) or force ( [ eq : path - integral - force - matrix ] ) .
we have demonstrated this approach for several three - dimensional geometries , such as the crossed capsules of fig .
[ fig : crossed - capsules ] @xcite .
if one is using dense - matrix techniques , the advantage of this approach over the stress - tensor technique seems clear @xcite : it avoids the complication of picking a stress - integration surface and an appropriate surface - integration technique , and allows the size of the linear system to be easily reduced via blocking as in eq .
( [ eq : logdet - block ] ) .
the situation is less clear as one moves to larger and larger problems , in which dense - matrix solvers become impractical and one requires an iterative method . in that case , computing @xmath302 $ ] straightforwardly requires @xmath10 linear systems to be solved ; if each linear system can be solved in @xmath47 time with a fast solver ( as discussed in secs .
[ sub : solution - techniques ] and [ sub : bem - stress ] ) , then the overall complexity is @xmath303 [ with @xmath38 storage ] , whereas it is possible that the stress - tensor surface integral may require fewer than @xmath10 solves . on the other hand , there may be more efficient ways to compute the trace ( or @xmath186 ) via low - rank approximations : for example , if the trace ( or @xmath186 ) is dominated by a small number of extremal eigenvalues , then these eigenvalues can be computed by an iterative method @xcite with the equivalent of @xmath304 linear solves .
the real - symmetric property of @xmath36 , as usual , means that the most favorable iterative methods can be employed , such as a lanczos or rayleigh - quotient method @xcite .
another possibility might be sparse - direct solvers via a fast - multipole decomposition @xcite . the most efficient use of a fast @xmath47 bem solver in casimir problems , whether by stress - tensor or path - integral methods , remains an open question ( and
the answer may well be problem - dependent ) . in the bem approach with localized basis functions ,
the @xmath302 $ ] expression for the force corresponds to a sum of a diagonal components for each surface element , and in the exact limit of infinite resolution ( infinitesimal elements ) this becomes an integral over the object surfaces .
expressing the force as a surface integral of a quantity related to greens - function diagonals is obviously reminiscent of the stress - tensor integration from sec .
[ sub : stress - tensor ] , and it turns out that one can prove an exact equivalence using only vector calculus @xcite .
( at least one previous author has already shown the algebraic equivalence of the stress tensor and the derivative of the path - integral energy for forces between periodic plates @xcite . )
it is possible , and sometimes very useful , to employ a hybrid of the bem and spectral techniques in the previous two sections .
one can discretize a surface using boundary elements , and use this discretization to solve for the scattering matrix @xmath305 of each object in a spectral basis such as spherical waves .
that is , for any given incident spherical wave , the outgoing field can be computed with bem via ( [ eq : bem - abstract ] ) and then decomposed into outgoing spherical waves to obtain one row / column of @xmath305 at a time ; alternatively , the multipole decomposition of the outgoing wave can be computed directly from the multipole moments of the excited surface currents @xmath194 @xcite .
this approach appears to be especially attractive when one has complicated objects , for which a localized bem basis works well to express the boundary conditions , but the interactions are only to be computed at relatively large separations where the casimir interaction is dominated by a few low - order multipole moments .
one can perform the bem computation once per object and re - use the resulting scattering matrix many times via the analytical translation matrices , allowing one to efficiently compute interactions for many rearrangements of the same objects and/or for `` dilute '' media consisting of many copies of the same objects @xcite .
( essentially , this could be viewed as a form of low - rank approximation of the bem matrix , capturing the essential details relevant to moderate - range casimir interactions in a much smaller matrix . )
such a hybrid approach is less attractive for closer separations , however , in which the increasing number of relevant multipole moments will eventually lead to an impractically large matrix to be computed .
consider the case of the interaction between two corrugated surfaces depicted in fig [ fig : rcwa ] , separated in the @xmath157 direction .
schematic problem for which eigenmode - expansion is well suited : the interaction between two corrugated surfaces , with period @xmath306 .
the casimir problem reduces to computing the reflection matrices @xmath307 and @xmath308 for each individual surface , in a planewave basis .
eigenmode expansion works by expanding the field in each cross - section ( dashed lines ) in the basis of eigenmodes of a @xmath157-invariant structure with that cross - section , and then matching boundary conditions whenever the cross - section changes . ] from the scattering - matrix viewpoint , it is natural to consider scattering off of each object by planewaves . in this case , the self - interaction matrices @xmath309 and @xmath310 can be re - expressed in terms of reflection matrices @xmath307 and @xmath308 for each surface , relating the amplitudes of incident waves at some plane ( dashed line ) above each surface to the reflected ( specular and non - specular ) planewave amplitudes .
the matrices @xmath284 and @xmath311 are replaced by a diagonal matrix @xmath312 that relates the planewave amplitudes at the planes for objects 1 and 2 , separated by a distance @xmath61at real frequencies , this would be a phase factor , but at imaginary frequencies it is an exponential decay as discussed below .
this results in the following expression for the casimir interaction energy : @xmath313d\xi.\label{eq : casimir - energy - r}\ ] ] alternatively , instead of viewing it as a special case of the t - matrix / scattering - matrix idea @xcite , the same expression can be derived starting from an eigenmode - summation approach @xcite .
the problem then reduces to computing the scattering of an incident planewave off of a corrugated surface , with the scattered field decomposed into outgoing planewaves . for this problem
, one could use any of the tools of computational em ( such as bem , fd , and so on ) , but there is a notable method that is often well - suited to the case of periodic surfaces , especially periodic surfaces with piecewise - constant cross - sections ( as in object 2 of fig .
[ fig : rcwa ] ) .
this method is called _ eigenmode expansion _
@xcite or _ rigorous coupled - wave analysis _ ( rcwa ) @xcite , or alternatively a _ cross - section _ method @xcite .
rcwa has a long history because it is closely tied to semi - analytical methods to study waveguides with slowly / weakly varying cross - sections @xcite .
an analogous method was recently applied to casimir problems @xcite . in rcwa
, one computes reflection and scattering matrices at a given frequency @xmath1 along some direction @xmath157 by expanding the fields at each @xmath157 in the basis of the eigenmodes of the cross - section at that @xmath157 ( waves with @xmath157 dependence @xmath314 at a given @xmath1 , where @xmath315 is called the _ propagation constant _ of the mode ) . along regions of uniform cross - section
, the @xmath157 dependence @xmath314 of each mode is known analytically and no computation is required ( the mode amplitudes are multiplied by a diagonal propagation matrix @xmath316 ) .
regions of continuously varying cross - section are approximated by breaking them up into a finite number of constant cross - section layers ( as in object 1 of fig .
[ fig : rcwa ] ) . at any @xmath157 where the cross - section changes , a change of basis is performed by matching boundary conditions (
the @xmath317 components of the fields must be continuous ) , yielding a _ transfer matrix _ at that interface .
all these transfer and propagation matrices can then be combined to compute scattering / reflection matrices for an entire structure .
the main difference here from classical rcwa computations is that the modes are computed at imaginary frequencies @xmath72 . as in sec .
[ sub : imaginary - green ] , this actually simplifies the problem . at an imaginary frequency @xmath71 ,
the modes of a given cross - section @xmath318 and @xmath319 with @xmath157 dependence @xmath320 ( @xmath321 ) satisfy the eigenequation ( for isotropic materials ) @xcite:@xmath322 where the @xmath317 subscript indicates a two - component vector with @xmath317 ( transverse ) components .
the operators on both the left- and right - hand sides are real - symmetric , while the operator on the left - hand side is positive - definite , and as a result the eigenvalues @xmath323 are purely real .
this means that the propagation constants @xmath315 are purely imaginary ( all of the imaginary - frequency modes are evanescent in @xmath157 ) , and the analogues of incoming / outgoing waves are those that are exponentially decaying towards / away from the surface .
moreover , the numerical problem of solving for these eigenmodes in a given cross - section reduces to a positive - definite generalized eigenvalue problem ( a definite matrix pencil @xcite ) , to which the most desirable numerical solvers apply @xcite ( unlike the classical real-@xmath1 problem in which there are both propagating and evanescent modes because the problem is indefinite @xcite ) . for homogeneous cross - sections
( as in the space between the two objects ) , the solutions are simply planewaves of the form @xmath324 , where for vacuum @xmath325 . for sufficiently simple cross - sections , especially in two - dimensional or axisymmetric geometries ,
it is possible to solve for the modes analytically and hence obtain the scattering matrices , and this is how the technique was first applied to the casimir problem @xcite . for more general geometries ,
one can solve for the modes numerically by a variety of techniques , such as by a transfer - matrix method in two dimensions @xcite or by a planewave expansion ( in the @xmath317 cross - section ) in three dimensions @xcite .
of course , one truncates to a finite number of modes via some cutoff @xmath326 ( which follows automatically from discretizing the cross - section in a finite grid , for example ) , and convergence is obtained in the limit as this cutoff increases . given a basis of eigenmodes with some cutoff
, the process of constructing the scattering / reflection matrices is thoroughly discussed elsewhere @xcite , so we do not review it here .
the strength of rcwa is that regions of uniform cross - section are handled with at most a 2d discretization of the cross - section , independent of the thickness of the region , so very thick or very thin layers can be solved efficiently .
the main limitation of rcwa methods is that the transfer matrices ( and the resulting reflection matrices @xmath307 and @xmath308 ) are dense @xmath28 matrices , where @xmath10 is the number of modes required for convergence . if @xmath10 is large , as in complicated three - dimensional structures , the problem can quickly become impractical because of the @xmath34 storage and @xmath35 computation requirements .
the most favorable case is that of periodic structures with relatively simple unit cells , in which case the problem can be reduced to that of computing the modes of each periodic unit cell ( with bloch - periodic boundary conditions ) as discussed below , and rcwa can then be quite practical even in three dimensions .
non - periodic structures , such as compact objects , can be handled by perfectly matched layer ( pml ) absorbing boundaries @xcite , albeit at greater computational cost because of the increased cross - section size .
in this section , we briefly discuss the issue of periodicity and other symmetries , which can be exploited to greatly reduce the computational effort in casimir calculations just as for classical em calculations . if a structure is periodic in the @xmath52 direction with period @xmath306 , as in fig .
[ fig : rcwa ] , the problem simplifies considerably because one can reduce the computation to a single unit cell of thickness @xmath306 .
in particular , one imposes bloch - periodic boundary conditions the fields at @xmath327 equal the fields at @xmath328 multiplied by a phase factor @xmath329and computes the casimir energy or force for each bloch wavevector @xmath330 separately , then integrates the result over @xmath330 via @xmath331 .
this can be derived in a variety of ways , for example by applying bloch s theorem @xcite to decompose the eigenmodes into bloch - wave solutions for each @xmath330 , or by expanding the delta functions of the fluctuation
dissipation approach in a fourier series @xcite .
more generally , for any periodic unit cell , one can perform the casimir energy / force computation for the unit cell with bloch periodic boundaries and then integrate the bloch wavevector @xmath332 over the irreducible brillouin zone ( multiplied by the volume ratio of the brillouin zone and the irreducible brillouin zone ) .
the specific case of continuous translational symmetry , say in the @xmath52 direction , corresponds @xmath333 and one must integrate over all @xmath330 ( the brillouin zone is infinite ) .
certain additional simplifications apply in the case of a perfect - metal structure with continuous translational symmetry , in which case the fields decompose into two polarizations and the @xmath89 integration can be performed implicitly @xcite .
rotational symmetry can be handled similarly : the fields can be decomposed into fields with @xmath289 angular dependence , and the total force or energy is the sum over all integers @xmath334 of the contributions for each @xmath334 @xcite .
more generally , the casimir contributions can be decomposed into a sum of contributions from irreducible representations of the symmetry group of the structure ( e.g. all eigenmodes can be classified into these representations @xcite ) ; translational and rotational symmetries are merely special cases . as another example , in a structure with a mirror symmetry one could sum even- and odd - symmetry contributions ( in fact , this is the underlying reason for the te / tm polarization decomposition in two dimensions @xcite ) .
in the preceding sections , we discussed only the computation of casimir interactions at zero temperature @xmath119 .
however , the modification of any imaginary - frequency expression for a casimir interaction from @xmath335 to @xmath232 is almost trivial : one simply performs a sum instead of an integral . if the @xmath335 interaction ( energy , force , etc . )
is expressed as an integral @xmath336 of some contributions @xmath337 at each imaginary frequency @xmath72 , then the @xmath232 interaction is well known to be simply @xcite : @xmath338 where @xmath339 is boltzmann s constant and @xmath340 indicates a sum with weight @xmath341 for the @xmath342 term .
the frequencies @xmath343 are known as matsubara frequencies , and the corresponding ( imaginary ) matsubara wavelengths are @xmath344 where @xmath345 .
the conversion of the @xmath335 integral into a summation can be derived in a variety of ways , most directly by considering thermodynamics in the matsubara formalism @xcite .
physically , this arises from the @xmath346 bose
einstein distribution factor that appears in the fluctuation dissipation expressions ( [ eq : ecorr - real ] ) for nonzero temperatures . when the contour integration is performed over @xmath1 , the @xmath118 introduces poles at @xmath347 that convert the integral into a sum via the residue theorem ( with the @xmath342 residue having half weight because it lies on the real-@xmath1 axis ) @xcite . as explained in sec .
[ sub : imaginary - green ] , some care must be applied in evaluating the @xmath342 term because of the well known singularity of maxwell s equations at @xmath348 ( where the @xmath82 and @xmath83 fields decouple ) , and one may need to take the limit @xmath147 ( although there is some controversy in the unphysical case of perfect metals @xcite ) mathematically , the sum of eq .
( [ eq : nonzero - t ] ) is exactly the same as a trapezoidal - rule approximation for the @xmath335 integral , with equally spaced abscissas @xmath349 @xcite .
thanks to the @xmath350 convergence of the trapezoidal rule @xcite , this means that the @xmath232 result is quite close to the @xmath335 result unless @xmath337 varies rapidly on the scale of @xmath351 .
in particular , suppose that @xmath337 varies on a scale @xmath352 , corresponding to some lengthscale @xmath61 in the problem ( typically from a surface surface separation ) .
in that case , assuming @xmath337 has nonzero slope has zero slope at @xmath353 , then the trapezoidal rule differs from the integral by @xmath354 or less , depending upon which derivative is nonzero at @xmath353 @xcite .
] at @xmath353 ( typical for interactions between realistic metal surfaces ) , then the nonzero-@xmath84 correction should be of order @xmath355 . at room temperature ( @xmath356 k ) , @xmath357 m , and the temperature corrections to casimir interactions are typically negligible for submicron separations @xcite .
on the other hand , it is possible that careful material and geometry choices may lead to larger temperature effects @xcite .
there is also the possibility of interesting effects in nonequilibrium situations ( objects at different temperatures ) @xcite , but such situations are beyond the scope of this review .
the area of numerical casimir computations remains rich with opportunities .
relatively few geometry and material combinations have as yet been explored , and thus many newly answerable questions remain regarding the ways in which casimir phenomena can be modified by exploiting the degrees of freedom available in modern nanofabrication . in the regime of computational techniques , while several effective methods have already been proposed and demonstrated , the parallels with computational electromagnetism lead us to anticipate ongoing improvements and developments for some time to come .
the same parallels also caution against any absolute `` rankings '' of the different approaches , as different numerical techniques have always exhibited unique strengths and weaknesses in both theory and practice . and
because computer time is typically much less expensive than programmer time , there is something to be said for methods that may be theoretically suboptimal but are easy to implement ( or are available off - the - shelf ) for very general geometries and materials . nor is the value of analytical and semi - analytical techniques diminished , but rather these approaches are freed from the tedium of hand computation to focus on more fundamental questions .
this work was supported in part by the army research office through the isn under contract w911nf-07-d-0004 , by the mit ferry fund , and by the defense advanced research projects agency ( darpa ) under contract n66001 - 09 - 1 - 2070-dod .
we are especially grateful to our students , a. w. rodriguez , a. p. mccauley , and h. reid for their creativity and energy in pursuing casimir simulations .
we are also grateful to our colleagues f. capasso , d. dalvit , t. emig , r. jaffe , j. d. joannopoulos , m. kardar , m. levin , m. lonar , j. munday , s. j. rahi , and j. white , for their many suggestions over the years .
rodriguez , a. , ibanescu , m. , iannuzzi , d. , capasso , f. , joannopoulos , j.d . ,
johnson , s.g . : computation and visualization of casimir forces in arbitrary geometries : non - monotonic lateral - wall forces and failure of proximity force approximations .
99*(8 ) , 080,401 ( 2007 ) rodriguez , a. , ibanescu , m. , iannuzzi , d. , joannopoulos , j.d .
, johnson , s.g . : virtual photons in imaginary time : computing casimir forces in arbitrary geometries via standard numerical electromagnetism .
phys . rev .
a * 76*(3 ) , 032,106 ( 2007 ) sedmik , r. , vasiljevich , i. , tajmar , m. : detailed parametric study of casimir forces in the casimir polder approximation for nontrivial 3d geometries .
j. computer - aided materials design * 14*(1 ) , 119132 ( 2007 ) zaheer , s. , rodriguez , a.w . ,
johnson , s.g . ,
jaffe , r.l . : optical - approximation analysis of sidewall - spacing effects on the force between two squares with parallel sidewalls .
phys . rev .
a * 76*(6 ) , 063,816 ( 2007 ) schaubert , d.h . , wilton , d.r . , glisson , a.w .
: a tetrahedral modeling method for electromagnetic scattering by arbitrarily shaped inhomogeneous dielectric bodies .
ieee trans .
antennas propagat .
* 32 * , 7785 ( 1984 ) moharam , m.g . ,
grann , e.b . ,
pommet , d.a . ,
gaylord , t.k . : formulation for stable and efficient implementation of the rigorous coupled - wave analysis of binary gratings .
a * 12 * , 10681076 ( 1995 ) moharam , m.g . ,
pommet , d.a . ,
grann , e.b .
, gaylord , t.k .
: stable implementation of the rigorous coupled - wave analysis for surface - relief gratings : enhanced transmittance matrix approach .
a * 12 * , 10771086 ( 1995 ) barrett , r. , berry , m. , chan , t.f . ,
demmel , j. , donato , j. , dongarra , j. , eijkhout , v. , pozo , r. , romine , c. , van der vorst , h. : templates for the solution of linear systems : building blocks for iterative methods , 2nd edn .
siam , philadelphia ( 1994 ) anderson , e. , bai , z. , bischof , c. , blackford , s. , demmel , j. , dongarra , j. , du croz , j. , greenbaum , a. , hammarling , s. , mckenney , a. , sorensen , d. : lapack users guide , 3rd edn .
siam , philadelphia ( 1999 ) rodriguez , a.w . ,
munday , j. , davlit , d. , capasso , f. , joannopoulos , j.d . ,
johnson , s.g . : stable suspension and dispersion - induced transition from repulsive casimir forces between fluid - separated eccentric cylinders .
lett . * 101*(19 ) , 190,404 ( 2008 ) zhao , l. , cangellaris , a.c . : a general approach for the development of unsplit - field time - domain implementations of perfectly matched layers for fdtd grid truncation .
ieee microwave and guided wave lett .
* 6*(5 ) , 209211 ( 1996 ) farjadpour , a. , roundy , d. , rodriguez , a. , ibanescu , m. , bermel , p. , joannopoulos , j.d . , johnson , s.g . ,
burr , g. : improving accuracy by subpixel smoothing in fdtd .
optics letters * 31 * , 29722974 ( 2006 ) kottke , c. , farjadpour , a. , johnson , s.g . :
perturbation theory for anisotropic dielectric interfaces , and application to sub - pixel smoothing of discretized numerical methods .
e * 77 * , 036,611 ( 2008 ) oskooi , a.f . ,
roundy , d. , ibanescu , m. , bermel , p. , joannopoulos , j.d . ,
johnson , s.g .
: meep : a flexible free - software package for electromagnetic simulations by the fdtd method .
phys . comm .
* 181 * , 687702 ( 2010 ) mccauley , a.p . ,
zhao , r. , reid , m.t.h . ,
rodriguez , a.w .
, zhao , j. , rosa , f.s.s . , joannopoulos , j.d . ,
dalvit , d.a.r . ,
soukoulis , c.m . ,
johnson , s.g .
: microstructure effects for casimir forces in chiral metamaterials .
arxiv:1006.5489 ( 2010 ) johnson , s.g . ,
bienstman , p. , skorobogatiy , m. , ibanescu , m. , lidorikis , e. , joannopoulos , j.d . : adiabatic theorem and continuous coupled - mode theory for efficient taper transitions in photonic crystals .
e * 66 * , 066,608 ( 2002 )
|
we review several different approaches for computing casimir forces and related fluctuation - induced interactions between bodies of arbitrary shapes and materials . the relationships between this problem and well known computational techniques from classical electromagnetism are emphasized .
we also review the basic principles of standard computational methods , categorizing them according to three criteria choice of problem , basis , and solution technique that can be used to classify proposals for the casimir problem as well . in this way
, mature classical methods can be exploited to model casimir physics , with a few important modifications .
|
there has been much interest in the past few years in ultracold systems of dipoles @xcite .
anisotropic long - range interactions can arise due to the interactions between the magnetic moments of atoms or due to the interactions between heteronuclear diatomic molecules ( in this case , each molecule is treated as an electric dipole ) .
experimentally , dipole - dipole interactions play a prominent role in the dynamics of ultracold atomic clouds consisting of cr @xcite , dy @xcite , and er @xcite . in these systems , the effects of the dipole - dipole interactions have been observed in the degenerate regime via expansion cloud imaging @xcite .
additionally , polar molecules such as krb @xcite , rbcs @xcite and lics @xcite have been created and the effects of the anisotropic interactions have been observed in krb clouds with phase space densities just above degeneracy @xcite . it remains an open question whether the effects from the anisotropic and long - range dipole - dipole interactions can be observed in thermodynamic observables at high temperature , i.e. , at temperatures above the degeneracy temperature .
for isotropic short - range interactions , the high - temperature behavior of two - component fermi gases is well described by the virial equation of state @xcite . at unitarity ,
i.e. , for infinitely large interspecies @xmath0-wave scattering length , the virial equation of state has been validated through careful measurements @xcite .
this paper derives the virial equation of state for dipolar gases and investigates the interplay between the short - range and long - range interactions . specifically , we focus on the temperature regime above quantum degeneracy and show that signatures of the dipole - dipole interactions can be enhanced if one of the generalized scattering lengths in a low partial wave channel is tuned .
we consider a single - component gas of aligned dipolar bosons or fermions under spherically symmetric harmonic confinement and calculate the second - order virial coefficient for weak dipole - dipole interactions .
if the dipole - dipole scattering properties are calculated in the first - order born approximation and if the dipole length @xmath1 is small compared to the harmonic oscillator length @xmath2 , then the second - order virial coefficient @xmath3 , which accounts for the two - body interactions , is negative at large @xmath4 for both identical bosons and identical fermions and depends quadratically ( and not linearly ) on @xmath1 .
this implies that the leading order effect of the two - body interactions for aligned dipolar gases under spherically symmetric harmonic confinement is repulsive at high temperature .
moreover , because @xmath3 depends quadratically on @xmath1 , the effects of the dipole - dipole interactions will be hard to extract from thermodynamic observables .
it is well known that the generalized scattering lengths , which characterize the scattering properties of the two dipole system , can be modified through the application of an external magnetic or electric field @xcite . for fermions ,
we show that by reducing one of the generalized scattering lengths , @xmath3 depends linearly ( and not quadratically ) on @xmath1 .
this implies that the effects of the dipole - dipole interactions can be enhanced by tuning one of the generalized scattering lengths to near zero .
the remainder of this paper is organized as follows .
section [ sec_system ] introduces the virial equation of state for a single - component quantum gas , the two - body hamiltonian for aligned dipoles , and the different models used to treat the interaction between the dipoles .
section [ sec_dipscatt ] studies the low - energy dipole - dipole scattering properties and characterizes the behavior of the generalized scattering lengths .
special attention is paid to low - energy shape resonances .
our main results are presented in secs .
[ sec_results ] and [ sec_thermo ] .
section [ sec_results ] determines the virial coefficient @xmath3 for dipolar bose and fermi gases under spherically symmetric confinement within various approximations .
the resulting virial coefficients are used in sec .
[ sec_thermo ] to determine thermodynamic observables such as the total energy and free energy
. finally , sec .
[ sec_conclusion ] concludes .
for a single component gas , where the chemical potential @xmath5 is held fixed , the thermodynamic potential @xmath6 in the grand canonical ensemble reads @xcite @xmath7 where @xmath8 is the hamiltonian of the @xmath9-body system , @xmath10 is @xmath11 , @xmath12 is boltzmann s constant , @xmath4 is the temperature , and tr is the trace operator . in the high - temperature limit , the chemical potential becomes large and negative , such that the fugacity @xmath13 , @xmath14 , is a small parameter .
taylor expanding eq . in the high - temperature limit
, i.e. , for @xmath15 , yields the thermodynamic potential in terms of the virial coefficients @xmath16 , @xmath17 the @xmath16 s are determined by the canonical partition functions @xmath18 , @xmath19 we evaluate the trace by inserting a complete set of eigenstates with eigenenergies @xmath20 , yielding @xmath21 where the summation index @xmath22 collectively denotes the complete set of quantum numbers allowed by symmetry . to isolate the effect of the two - body interactions , we express the grand canonical potential @xmath6 in terms of the grand canonical potential @xmath23 of the noninteracting system plus corrections @xmath24 @xcite , @xmath25 we are interested in accounting for the leading - order effect of the two - body interactions
correspondingly , we truncate eq . at @xmath26 . collecting terms of the same order in the fugacity
, the second - order virial coefficient reads @xmath27 where @xmath28 characterizes the noninteracting two - body system and @xmath3 , @xmath29 contains the leading - order effect of the two - body interactions . in eq .
( [ eq_db2 ] ) , @xmath30 and @xmath31 denote the canonical partition functions of the interacting and noninteracting two - body systems , respectively . for spherically symmetric harmonic confinement with angular trapping frequency @xmath32 , as considered throughout this paper
, we find @xcite @xmath33 where @xmath34 in eq . , the plus sign is for identical bosons while the minus sign is for identical fermions .
we define the dimensionless quantity @xmath35 as the ratio of the harmonic oscillator energy @xmath36 , @xmath37 ( i.e. , the characteristic energy of the noninteracting system ) , and the thermal energy scale @xmath38 , @xmath39 . to determine the second - order virial coefficient @xmath3 that accounts for the interactions between the dipoles [ see eq .
( [ eq_db2 ] ) ] , we need to determine @xmath30 , i.e. , we need to determine the eigenenergies @xmath40 of the interacting two - body system [ see eq .
( [ eq_qne ] ) ] .
since the two - body interaction depends only on the distance vector @xmath41 , the center of mass and relative motion separate , implying @xmath42 .
the center of mass contribution has been determined in the literature @xcite and we thus focus in the following on the relative contribution .
the schrdinger equation for the relative distance vector @xmath41 reads @xmath43 \psi({\bf r } ) = e_2^{\rm rel}\psi({\bf r}),\end{aligned}\ ] ] where @xmath44 denotes the reduced mass and @xmath45 the two - body interaction potential ( see below for details ) . in what follows ,
we assume that the dipoles are aligned by an external field .
this implies that @xmath46 is azimuthally symmetric and that the projection quantum number @xmath47 is a good quantum number .
the relative orbital angular momentum @xmath48 , in contrast , is not a good quantum number .
specifically , the interaction potential couples angular momentum channels with the same parity . for identical bosons , the even relative angular momenta @xmath48 are coupled . for identical fermions , in contrast ,
the odd relative angular momenta @xmath48 are coupled .
we solve eq .
using two different representations of the interaction potential . in the first approach , @xmath46 is given by @xcite @xmath49 where @xmath50 denotes the angle between the distance vector @xmath41 and the axis along which the dipoles are aligned .
the model potential @xmath51 is characterized by two length scales , the hard wall radius ( or short - range length ) @xmath52 and the dipole length ( or long - range length ) @xmath1 , @xmath53 , where @xmath54 denotes the dipole moment . in the second approach ,
we use the regularized zero - range pseudopotential @xmath55 @xcite , @xmath56 equation contains an infinite number of terms .
the @xmath57 are defined by how they act on a wavefunction , @xmath58_{r\to0 } , \end{aligned}\ ] ] where the @xmath59 denote the spherical harmonics and where @xmath60 is the relative scattering momentum . in the absence of the confinement , @xmath61 is given by the relative scattering energy @xmath62 , @xmath63 . in the presence of the harmonic trap ,
@xmath61 is determined , in a self - consistent manner , by the eigenenergies @xmath64 of eq .
( [ eq_se ] ) @xcite .
since @xmath47 is a good quantum number , each @xmath47-channel is treated by a @xmath47-specific pseudopotential ; for details on the @xmath65 case , see ref .
the generalized scattering lengths @xmath66 depend on the dipole moment @xmath54 .
to use @xmath55 , the @xmath66 have to be provided as input .
the next section discusses how these input parameters are determined .
this section determines the generalized free - space scattering lengths @xmath66 for the shape - dependent model potential @xmath51 , eq . .
to determine the @xmath66 , we consider the relative free - space schrdinger equation [ eq . with @xmath67 and propagate the logarithmic derivative matrix using johnson s algorithm @xcite with an adaptive step size . at large @xmath68 ,
the logarithmic derivative matrix is matched to the asymptotic solution and the generalized scattering lengths are determined from the partial wave phase shifts @xmath69 , @xmath70/k$ ] @xcite . before presenting our numerical results
, we briefly review the expected low - energy properties of the @xmath66 . away from resonances , the generalized scattering lengths @xmath66 [ @xmath71 are , in general , well approximated by applying the first - order born approximation to the dipole - dipole potential @xmath72 , @xmath73 @xcite . the born approximation yields @xmath74}{(2l-1)(2l+3)l(l+1)}\end{aligned}\ ] ] and @xmath75 } } { ( 2l-1)l(l-1)\sqrt{(2l+1)(2l-3)}}.\end{aligned}\ ] ] all other @xmath66 with @xmath76 are zero . for @xmath65 , eqs . and
agree with eqs .
( 9 ) and ( 10 ) of ref .
it is important to note that the scattering length @xmath77 can not , in general , be determined reliably within the first - order born approximation @xcite . in the following ,
we focus on the scattering properties of two identical fermionic dipoles , implying that @xmath48 and @xmath78 are odd , with @xmath65 . , ( b ) @xmath79 , and ( c ) @xmath80 as a function of the relative scattering energy @xmath62 for @xmath81 .
the dotted lines indicate the zero - energy born approximation values of @xmath66 . in ( a )
, the vertical dashed line indicates the location of the shape resonance .
note the different vertical scales of panel ( a ) and of panels ( b ) and ( c ) .
, width=264 ] solid lines in fig .
[ fig_scl ] show the numerically obtained generalized scattering lengths @xmath82 with @xmath83 for @xmath51 with @xmath81 as a function of the scattering energy @xmath62 .
the scattering energy has been scaled by the dipole energy @xmath84 , @xmath85 . for comparison ,
dotted horizontal lines show the generalized scattering lengths predicted by the born approximation .
the numerically obtained scattering lengths @xmath79 and @xmath80 [ see figs .
[ fig_scl](b ) and [ fig_scl](c ) ] are well approximated by the born approximation estimates over a wide range of scattering energies .
deviations occur in narrow energy windows around @xmath86 and @xmath87 .
the deviations between the exact values and the corresponding born approximation values are larger for @xmath88 [ see fig .
[ fig_scl](a ) ] than for @xmath89 and @xmath80 .
the deviations from the born approximation values can be attributed to a quasi - bound state that is , approximately , tied to the @xmath90 channel [ see dashed line in fig .
[ fig_scl](a ) ] . to understand the emergence of the quasi - bound state
, we treat eq . using @xmath51 in the adiabatic representation @xcite .
figure [ fig_potcurves ] for two aligned fermionic dipoles .
though the angular momentum @xmath48 is not a good quantum number , the curves can be approximately labeled by @xmath48 .
solid , dash - dotted , and dash - dot - dotted lines are for @xmath91 , and @xmath87 , respectively .
the hard wall is indicated by a hatched region extending from zero to @xmath92 .
the horizontal dashed line indicates the energy of the quasi - bound state .
, width=264 ] shows the lowest three adiabatic potential curves for two aligned fermionic dipoles with @xmath92 as a function of the distance coordinate @xmath68 .
though the angular momentum @xmath48 is not a good quantum number , each curve can be labeled approximately by @xmath48 .
solid , dash - dotted , and dash - dot - dotted lines are for @xmath93 and @xmath87 , respectively .
the @xmath94 potential curve exhibits a barrier of height around @xmath95 , indicating that the true bound state which is to a good approximation described by the lowest potential curve @xcite turns into a quasi - bound state with positive energy after crossing the zero - energy threshold .
as a function of the relative scattering energy @xmath62 for various values of @xmath96 .
( a ) solid , dotted , dash - dotted , and dash - dot - dotted lines are for @xmath97 and @xmath98 , respectively .
( b ) solid , dotted , and dash - dotted lines are for @xmath99 [ see also fig . [ fig_scl](a ) ] , @xmath100 and @xmath101 , respectively . , width=264 ] figure [ fig_a110 ] shows the energy dependence of @xmath88 for various ratios between the long - range and short - range length scales in the vicinity of a shape resonance . for @xmath96 values , for which no quasi- or shallowly - bound states exist , the born approximation is valid for @xmath102 [ see the solid and dotted lines in fig .
[ fig_a110](a ) ] . as @xmath96 increases
, @xmath88 develops a `` dip '' around @xmath103 [ see the dash - dotted line in fig .
[ fig_a110](a ) ] . at yet larger @xmath96
, the dip turns into two resonance - like features , i.e. , @xmath88 diverges at energies above and below the barrier of the @xmath94 adiabatic potential curve [ see the dash - dot - dotted and solid lines in figs .
[ fig_a110](a ) and [ fig_a110](b ) , respectively ] .
we associate the resonance features with the appearance of a quasi - bound state that `` lives '' to the left of the barrier of the adiabatic potential curve ( see fig . [ fig_potcurves ] ) . as @xmath96 increases further , the energy of the quasi - bound state moves to smaller ( positive ) energies [ see the solid line in fig .
[ fig_a110](b ) ] . for @xmath104 ,
@xmath88 is nearly constant for @xmath105 [ see the dotted line in fig .
[ fig_a110](b ) ] . for larger @xmath96
, the quasi - bound state turns into a true bound state and @xmath88 `` returns '' , in the low energy regime , to its born approximation value [ see dash - dotted line in fig . [
fig_a110](b ) ] . in the next section , we make two assumptions .
( i ) we assume that one of the generalized scattering lengths is tunable while the other generalized scattering lengths are well described by their born approximation values .
( ii ) we assume that all generalized scattering lengths are , to a good approximation , independent of energy . to fulfill these assumptions approximately
, the external field should be tuned to a value near resonance .
experiments on fermionic alkali atoms @xcite have demonstrated that the regime where only one of the scattering lengths differs notably from its born approximation value can be reached .
the requirement to sit near but not on resonance allows one to take advantage of the fact that the resonance feature of the tunable scattering length is much broader than that of the other scattering lengths ; this implies that the energy - dependence of the scattering lengths approximated by their born approximation values is weak .
lastly , as shown by the dotted line in fig .
[ fig_a110](b ) for @xmath88 , the near - resonance regime also allows one to realize a regime where the tunable scattering length is roughly independent of energy over a wide energy regime .
this section discusses the virial coefficient @xmath3 that accounts for the two - body interactions for two identical bosons and two identical fermions as a function of the temperature . to determine @xmath3 , we take advantage of the fact that the relative eigenenergies @xmath64 for the trapped two - dipole system interacting through @xmath55 can , provided the generalized scattering lengths @xmath66 are known , be obtained by solving a transcendental equation @xcite . as indicated in our discussion at the end of sec .
[ sec_dipscatt ] , we solve the transcendental equation assuming that the generalized scattering lengths are energy - independent .
we have checked that the use of an energy - dependent pseudo - potential leaves our main conclusions unaltered . combining @xmath64 and @xmath106 , the canonical partition function @xmath30 [ see eq . ] and correspondingly the virial coefficient @xmath3 can be calculated .
since the relative two - body energy spectrum is , in practice , only known up to a maximum energy @xmath107 , @xmath3 is only accurate up to a maximum temperature or down to a minimum dimensionless inverse temperature @xmath35 . to distinguish the second - order virial coefficient for two identical bosons and two identical fermions , we use the superscripts @xmath108 and @xmath109 , respectively .
we start our discussion by considering two identical bosons .
the solid lines in figs .
[ fig_b20all](a ) and [ fig_b20all](b ) as a function of @xmath35 for @xmath110 for ( a ) and ( b ) identical dipolar bosons , and ( c ) and ( d ) identical dipolar fermions . for comparison , the right vertical axis
is scaled by the magnitude of the noninteracting virial coefficient at @xmath111 .
( a ) all generalized scattering lengths for @xmath112 are given by their born approximation values and @xmath113 .
( b ) all generalized scattering lengths for @xmath112 are given by their born approximation values and @xmath114 ; the dashed line shows eq . .
( c ) all generalized scattering lengths are given by their born approximation values .
( d ) all generalized scattering lengths are given by their born approximation values except @xmath115 ; the dashed line shows eq . .
, width=264 ] show the virial coefficient @xmath116 for a relatively small dipole length @xmath1 , @xmath110 [ @xmath2 denotes the oscillator length , @xmath117 . to determine the relative eigenenergies @xmath64 ,
the generalized scattering lengths @xmath66 with @xmath76 are approximated by the born approximation values [ see eqs .
( [ eq_all ] ) and ( [ eq_allp ] ) ] . as discussed in sec .
[ sec_dipscatt ] , the @xmath0-wave scattering length @xmath77 can not be determined within the born approximation .
figure [ fig_b20all](a ) assumes that @xmath118 .
correspondingly , the resulting virial coefficient is that for a `` purely '' dipolar system .
figure [ fig_b20all](b ) assumes that @xmath119 .
thus , the virial coefficient shown in fig .
[ fig_b20all](b ) reflects the combined effect of the long - range dipole - dipole interaction and the short - range @xmath0-wave interaction .
for the @xmath35 values shown , the solid lines are converged to better than 1% and include all states with relative energy less than @xmath120 , that is , @xmath121 . for the born approximation to be applicable ,
two conditions have to be met .
first , the virial coefficient @xmath116 has to be converged for an energy cutoff @xmath107 that is smaller than @xmath84 .
this condition comes from the fact that the born approximation is only valid for @xmath122 . in figs .
[ fig_b20all](a ) and [ fig_b20all](b ) , we have @xmath123 and thus @xmath107 and @xmath124 .
second , the system has to be away from shape resonances that are associated with quasi - bound states whose energy lies below @xmath125 . the second - order virial coefficient @xmath116 for the purely dipolar bose gas under spherically symmetric harmonic confinement [ fig .
[ fig_b20all](a ) ] is negative for all @xmath35 shown , indicating that the leading - order effect of the dipole - dipole interaction is repulsive for aligned bose gases with @xmath126 small compared to one .
figure [ fig_b20all](a ) shows that the effect of the two - body interaction is very small .
specifically , the contribution @xmath116 is about three orders of magnitude smaller than the noninteracting contribution @xmath28 at @xmath111 [ see the right axis of fig .
[ fig_b20all](a ) ] , suggesting that the effects of the dipole - dipole interactions would be hard to extract from thermodynamic observables for a purely dipolar gas with small @xmath126 .
the virial coefficient for @xmath119 [ fig .
[ fig_b20all](b ) ] is about three orders of magnitude larger than that for @xmath118 [ fig .
[ fig_b20all](a ) ] .
moreover , the sign is reversed . since the dipole length @xmath1 is the same in figs .
[ fig_b20all](a ) and [ fig_b20all](b ) , it is clear that the increase of @xmath116 is due to the change in @xmath77 .
the fact that @xmath116 is positive for @xmath119 makes sense intuitively since a positive @xmath116 implies an effective attraction as would be naively expected for a negative @xmath77 .
we now consider two identical fermions .
the solid lines in figs .
[ fig_b20all](c ) and [ fig_b20all](d ) show @xmath127 for two identical dipolar fermions for @xmath110 . in fig .
[ fig_b20all](c ) , the energies @xmath64 are obtained by approximating all generalized scattering lengths by the born approximation values .
thus , the virial coefficient shown in fig .
[ fig_b20all](c ) is that for a purely dipolar gas under spherically symmetric external confinement with small @xmath126 .
@xmath127 is negative at high temperature , changes sign at @xmath128 , and is positive for smaller temperatures ( larger @xmath35 ) . in fig .
[ fig_b20all](d ) , in contrast , the energies @xmath64 are obtained by approximating all but one of the generalized scattering lengths by their born approximation value .
specifically , we choose @xmath115 , i.e. , we assume that this generalized scattering length is tuned away from its born approximation value through the application of an external field
. figure [ fig_b20all](d ) shows that @xmath127 is negative for all temperatures and that @xmath129 is between one and two orders of magnitude larger than the absolute value of the virial coefficient for the purely dipolar gas [ see fig .
[ fig_b20all](c ) ] . since the dipole length @xmath1 is the same in figs .
[ fig_b20all](c ) and [ fig_b20all](d ) , it is clear that the increase of @xmath130 is due to the change in @xmath88 .
figure [ fig_scl ] shows that the scenario considered here can be realized , at least approximately , for the model potential @xmath51 . for the parameters considered in fig .
[ fig_scl ] , the @xmath66 with @xmath131 are , except for a tiny energy window , well approximated by the born approximation value for @xmath132 up to @xmath133 . moreover , the discussion around fig .
[ fig_a110 ] shows that @xmath88 can , at least approximately , be tuned to a value different from the born approximation value while all other @xmath66 are quite well described by their born approximation value .
we now develop a perturbative approach that ( i ) shows analytically that the second - order virial coefficients @xmath116 and @xmath127 for the purely dipolar gases do not contain terms linear in @xmath1 , thereby explaining the small values of @xmath116 and @xmath127 in figs .
[ fig_b20all](a ) and [ fig_b20all](c ) , and ( ii ) yields analytic expressions in the case where one or more of the generalized scattering lengths are assumed to be tunable [ see dashed lines in figs .
[ fig_b20all](b ) and [ fig_b20all](d ) ] .
the idea is to calculate the relative energies @xmath64 perturbatively and to then use the perturbative energy expressions to calculate @xmath3 analytically to linear order in @xmath126 .
we assume that @xmath126 is small and employ first - order degenerate perturbation theory .
specifically , we treat the noninteracting harmonic oscillator hamiltonian in the relative degrees of freedom as the unperturbed hamiltonian and the regularized pseudopotential @xmath55 as a perturbation .
we label the unperturbed states @xmath134 by the principal quantum number @xmath9 , the orbital angular momentum quantum number @xmath48 and the projection quantum number @xmath47 , and find compact analytical expressions for the matrix elements @xmath135 , @xmath136 diagonalizing the perturbation matrix spanned by the matrix elements @xmath137 for each noninteracting energy manifold , i.e. , for the manifolds with unperturbed energies @xmath138 , @xmath139 , and so on , we find the first - order energy shifts @xmath140 , @xmath141 . using these energy shifts in eqs . and , the second - order virial coefficient in the weakly - interacting regime can be written as @xmath142 so far , the discussion is general , i.e. , we have not yet specified whether we are dealing with bosons or fermions .
correspondingly , eq . ( [ eq_db2_perturbative ] ) writes @xmath3 without the @xmath108/@xmath109 superscript .
inspection of eq .
( [ eq_pert_matrix_element ] ) shows that the matrix elements @xmath143 depend on @xmath47 only through the generalized scattering lengths . using this and rewriting the sum of eigenvalues for each energy manifold in terms of the trace of the corresponding perturbation matrix ,
we find @xmath144 where we have replaced the scattering momentum @xmath60 by the unperturbed energies , that is , we used @xmath145 . in eq .
( [ eq_db2_perturbative2 ] ) , the sum over @xmath48 is restricted to even or odd @xmath48 for bosons or fermions , respectively .
we now apply eq .
( [ eq_db2_perturbative2 ] ) to the purely dipolar bose and fermi gases . using the born approximation values
[ see eq .
( [ eq_all ] ) ] , we find @xmath146 this shows that @xmath147 and @xmath148 do not , in the small @xmath126 regime , contain terms that are linear in the dipole length @xmath1 , in agreement with the fact that the virial coefficients shown in figs .
[ fig_b20all](a ) and [ fig_b20all](c ) are very small . to determine the leading - order dependence of @xmath3 on @xmath1 for purely dipolar gases , our treatment would have to be modified in two places .
first , the generalized scattering lengths would have to be known to second order in @xmath1 and , second , the perturbation treatment would have to be pushed to second order .
these generalizations are not pursued in this work .
next , we focus on the scenario where one or more of the generalized scattering lengths are tunable [ see figs .
[ fig_b20all](b ) and [ fig_b20all](d ) ] . for identical bosons , we assume that the @xmath66 with @xmath76 are given by the born approximation values and that @xmath77 takes a small but finite value ( @xmath149 much smaller than one ) . under these assumptions ,
the perturbative treatment yields @xmath150 equation ( [ eq_b2b ] ) breaks down in the @xmath151 limit but is expected to be valid in the @xmath35 regime considered in this paper . the dashed line in fig .
[ fig_b20all](b ) shows eq .
( [ eq_b2b ] ) .
the agreement with the solid line is excellent , indicating that the virial coefficient @xmath147 is , to a very good approximation , determined by @xmath77 alone and independent of @xmath1 .
this implies that the dipolar bose gas with tunable @xmath77 behaves , to leading order , like a bose gas with isotropic @xmath0-wave interactions .
the reason is that the leading order dipolar contributions to @xmath147 average to zero in the regime where the born approximation is applicable [ see eq . ( [ eq_average ] ) ] . for identical fermions ,
we assume that the scattering lengths with @xmath90 are tunable and that all other generalized scattering lengths are approximated by the born approximation values . using the perturbative energies ( i.e. , assuming @xmath152 ) and evaluating the sums in eq .
, we find @xmath153 where @xmath154 is the hypergeometric function .
equation breaks down in the @xmath155 limit but is expected to be valid in the @xmath35 regime considered in this paper . the dashed line in fig .
[ fig_b20all](d ) shows eq .
( [ eq_b2f ] ) with @xmath156 and @xmath157 given by their born approximation value .
the agreement with the solid line is excellent .
our analysis shows that the virial coefficient @xmath158 depends linearly on @xmath1 if one of the @xmath66 is tuned away from its born approximation value .
perturbative expressions analogous to eqs . and can be derived if other low partial - wave scattering lengths are assumed to be tunable .
this section examines the thermodynamics of aligned dipolar gases under spherically symmetric harmonic confinement using the virial equation of state up to second order in the fugacity . figures [ fig_energy ] and [ fig_freeenergy ] for @xmath159 identical fermions under spherically symmetric confinement as a function of the temperature @xmath4 .
the solid line shows @xmath160 for the non - interacting fermi gas . dashed and dotted lines show @xmath160 for the interacting fermi gas with @xmath161 obtained using the virial equation of state to second order in the fugacity .
the dashed line shows results for the purely dipolar fermi gas .
the dotted line shows results for @xmath156 while other @xmath66 are approximated by their born - approximation value .
, width=264 ] show the total energy per particle @xmath160 and the helmholtz free energy per particle @xmath162 , respectively , as a function of the temperature @xmath4 for @xmath159 identical fermions with @xmath110 .
both quantities are scaled by the fermi energy @xmath163 , which in the semiclassical approximation reads @xmath164 . for @xmath159 identical fermions under external spherically symmetric confinement as a function of the temperature @xmath4 . the parameters and line styles are the same as in fig .
[ fig_energy ] .
the dashed and solid lines are indistinguishable on the scale shown . ,
width=264 ] the dashed lines show @xmath160 and @xmath162 for the purely dipolar fermi gas with the @xmath66 approximated by their born approximation value .
the dotted lines , in contrast , are for the case where @xmath88 is assumed to be tunable ( here , @xmath156 ) and where all other @xmath66 are approximated by their born approximation value . for comparison ,
the solid lines show @xmath160 and @xmath162 for the non - interacting fermi gas . on the scale of figs .
[ fig_energy ] and [ fig_freeenergy ] , the dashed lines are nearly indistinguishable from the solid lines ; for temperatures above @xmath165 , which is roughly where we expect that the virial equation of state yields reliable results , the deviations are less than 0.01% .
this implies that the effects of the dipolar interactions will be hard to observe for weakly - interacting purely dipolar gases .
in essence , the dipolar effects get averaged over , making the weakly - interacting purely dipolar gas behave approximately like a non - interacting gas . if the scattering length @xmath88 is tuned to zero ( see dotted lines in figs .
[ fig_energy ] and [ fig_freeenergy ] ) , the dependence of the thermodynamic observables on the dipolar interactions is enhanced .
the difference between the thermodynamic observables for the interacting and non - interacting systems is of the order of 3% for @xmath166 .
this suggests that dedicated cold atom experiments on erbium or dysprosium might be able to detect the effects of the dipolar interactions in thermodynamic measurements at temperatures around the fermi temperature . as discussed above
, the tunability of one of the scattering lengths is possible if the system under study exhibits non - overlapping resonances . in practice ,
a given scattering length may change a fair amount in the relevant energy range or overlapping resonances may lead to deviations in more than one scattering length .
while an exhaustive study of all possible non - universal parameter combinations is beyond the scope of this work , our key point , namely , the fact that tuning one of the scattering lengths away from its born approximation value ( and possibly to zero ) can enhance the signatures of the dipole - dipole interactions of weakly - interacting gases in thermodynamic observables , should apply also in more complicated situations .
this paper considered the high - temperature virial equation of state of weakly - interacting single - component dipolar quantum gases under spherically symmetric harmonic confinement . for
purely bosonic and fermionic dipoles , we found that the lowest - order correction to the ideal gas is repulsive in the high - temperature limit .
compared to the non - interacting gases , the modifications of the thermodynamic observables are very small . using a perturbative approach , we showed that the corrections are small since the second order virial coefficient depends quadratically on the dipole length @xmath1 ; terms linear in @xmath1 average to zero within the born approximation .
we argued that the effects of the dipole - dipole interactions can be enhanced if one or more generalized scattering lengths can be tuned away from their born approximation value .
for the fermionic gas , we showed explicitly that the generalized scattering length @xmath88 can be tuned in the vicinity of a low - energy shape resonance , leading to a second order virial coefficient that depends linearly on the dipole length .
correspondingly , the effects of the dipolar interaction are , compared to the purely dipolar case , greatly enhanced . throughout
, we considered single - component bose and fermi gases .
if we considered two - component dipolar gases , then the thermodynamic properties would depend on the interspecies and intraspecies virial coefficients .
for a two - component fermi gas , e.g. , the intraspecies virial coefficients would be given by @xmath167 while the interspecies virial coefficient would be given by a combination of the even and odd @xmath48 virial coefficients , i.e. , of @xmath168 and @xmath167 .
while the thermodynamic observables would differ from those considered in this paper , our perturbative analysis of the virial coefficients carries over unchanged .
the present work presents a first step toward understanding the thermodynamic properties of dipolar systems .
while the present work provides some insights into weakly - interacting dipolar gases at finite temperature , many questions remain unanswered . what happens if the dipolar interactions are of intermediate strength or strong ?
do these systems exhibit universal thermodynamic properties ?
what are the effects of the third- and higher - body virial coefficients ?
what are the effects of anisotropic confinement geometries ?
|
we consider dilute gases of dipolar bosons or fermions in the high - temperature limit in a spherically symmetric harmonic trapping potential . we examine the system using a virial expansion up to second order in the fugacity .
using the born approximation and assuming purely dipolar interactions , we find that the second - order virial coefficient for both bosons and fermions depends quadratically on the dipole length and is negative at high temperature , indicating that to lowest order in the dipole - dipole interactions the dipolar single - component quantum gases are repulsive . if the @xmath0-wave scattering length for the bosonic system is tunable and its absolute value is made small , then the @xmath0-wave interactions dominate and the dipolar gas behaves like a weakly - interacting bose gas with isotropic @xmath0-wave interactions .
if the generalized scattering lengths for the fermionic system are tunable , then the dipole length can enter linearly in the virial equation of state , enhancing the dipole - dipole effects in the thermodynamic observables .
|
the balitsky , fadin , kuraev and lipatov ( bfkl ) @xcite equation resums all leading logarithms of bjorken @xmath2 for hadronic cross sections and structure functions . the solution of the bfkl equation grows like a power of center of mass energy @xmath3 , therefore violating the unitarity bound at very high energies @xcite .
this is one of the major problems of small @xmath2 physics , since the froissart bound @xcite states that the total cross section should not raise faster than @xmath4 at asymptotically high energies .
there is a general belief that the unitarity problem could be cured by resumming all multiple bfkl pomeron exchanges in the total cross section , or , equivalently , in the structure function . as was argued by mueller in @xcite multiple pomeron exchanges
become important at the values of rapidity of the order of @xmath5 with @xmath6 the strong coupling constant , which is assumed to be small , and @xmath7 .
this result could be obtained if one notes that one pomeron contribution to the total cross section of , for instance , onium onium scattering , is parametrically of the order of @xmath8 $ ] and the contribution of the double pomeron exchange is @xmath9 $ ] . since multiple pomeron exchanges become important when the single and double pomeron exchange contributions become comparable , we recover eq .
( [ yu ] ) by just equating the two expressions .
after completion of the calculation of the next to leading order corrections to the kernel of the bfkl equation ( nlo bfkl ) by fadin and lipatov @xcite , and , independently , by camici and ciafaloni @xcite , it was shown in @xcite that due to the running coupling effects these corrections become important at the rapidities of the order of @xmath10 one can see that @xmath11 for parametrically small @xmath6 .
that implies that the center of mass energy at which the multiple pomeron exchanges become important is much smaller , and , therefore , is easier to achieve , than the energy at which nlo corrections start playing an important role .
that is multiple pomeron exchanges are probably more relevant than nlo bfkl to the description of current experiments @xcite .
also multiple pomeron exchanges are much more likely to unitarize the total hadronic cross section . in this paper
we are going to propose a solution to the problem of resummation of multiple pomeron exchanges and show how the bfkl pomeron unitarizes . in the previous paper on the subject @xcite we proposed an equation which resums all multiple bfkl pomeron @xcite exchanges in the nuclear structure function . in deriving the equation we used the techniques of mueller s dipole model @xcite . as was proven some time ago @xcite the dipole model can provide us with the equation which is exactly equivalent to the bfkl equation @xcite . in @xcite
it was shown that the dipole model methods , when applied to hadronic or nuclear structure functions , yield us with the equation which in the traditional feynman diagram language resums the so called pomeron `` fan '' diagrams in the leading logarithmic approximation , an example of which is depicted in fig .
[ fan ] for the case of deep inelastic scattering . to review the results of @xcite
let us consider a deep inelastic scattering ( dis ) of a virtual photon on a hadron or a nucleus .
an incoming photon splits into a quark
antiquark pair and then the @xmath12 pair rescatters on the target hadron . in the rest frame of the hadron
all qcd evolution should be included in the wave function of the incoming photon . that way the incoming photon develops a cascade of gluons , which then scatter on the hadron at rest . in @xcite
this gluon cascade was taken in the leading longitudinal logarithmic ( @xmath13 ) approximation and in the large @xmath14 limit .
this is exactly the type of cascade described by mueller s dipole model @xcite . in the large @xmath14 limit
the incoming gluon develops a system of color dipoles and each of them independently rescatters on the hadron ( nucleus ) @xcite .
the forward amplitude of the process is shown in fig .
the double lines in fig .
[ dis ] correspond to gluons in large @xmath14 approximation being represented as consisting of a quark and an antiquark of different colors .
the color dipoles are formed by a quark from one gluon and an antiquark from another gluon . in fig .
[ dis ] each dipole , that was developed through the qcd evolution , later interacts with the nucleus by a series of glauber type multiple rescatterings on the nucleons .
the assumption about the type of interaction of the dipoles with the hadron or nucleus is not important for the evolution .
it could also be just two gluon exchanges .
the important assumption is that each dipole interacts with the target independently of the other dipoles , which is done in the spirit of the large @xmath14 limit . for the case of a large nucleus independent dipole interactions with the target nucleus
are enhanced by some factors of the atomic number @xmath15 of the nucleus compared to the case when several dipoles interact with the same nucleon @xcite .
this allowed us to assume that the dipoles interact with the nucleus independently .
the summation of these contribution corresponds to summation of the pomeron `` fan '' diagrams of fig .
however , in general there is another class of diagrams , which we will call pomeron loop diagrams . in a pomeron loop diagram
a pomeron first splits into two pomerons , just like in the fan diagram , but then the two pomerons merge back into one pomeron . in other words the graphs containing pomerons not only splitting , but also merging together will be referred to as pomeron loop diagrams . in a large nucleus these pomeron loop diagrams could be considered small , since they would be suppressed by powers of @xmath15 .
this approximation strictly speaking is not valid for the calculation of a hadronic wave function . at very high energies
it also becomes not very well justified even in the nuclear case for the following reason .
the contribution of each additional pomeron in a fan diagram on a nucleus is parametrically of the order of @xmath16 , which is large at the rapidities of the order of @xmath17 given by eq .
( [ yu ] ) .
the contribution of an extra pomeron in a pomeron loop diagram is of the order of @xmath18 , which is not enhanced by powers of @xmath15 and is , therefore , suppressed in the nuclear case .
however , one can easily see that at extremely high energies , when @xmath19 becomes greater than or of the order of one , pomeron loop diagrams become important even for a nucleus , still being smaller than the fan diagrams .
resummation of pomeron loop diagrams in a dipole wave function seems to be a very hard technical problem , possibly involving nlo bfkl or , even , next - to - next - to - leading order bfkl kernel calculations @xcite .
nevertheless , it is the author s belief that these pomeron loop diagrams are not going to significantly change the high energy behavior of the structure functions .
still in the rigorous sense the results of this paper should be considered as resumming the fan diagrams only , which is strictly justified only for a large nucleus . in @xcite , rewriting the @xmath0 structure function of the target as a convolution of the wave function of the @xmath20 fluctuations of a virtual photon with the propagator of the quark
antiquark pair through the nucleus , we obtained @xmath21 where the photon s wave function @xmath22 consists of transverse @xmath23\ ] ] and longitudinal @xmath24 contributions , with @xmath25 , and we assumed that quarks are massless and have only one flavor and unit electric charge .
as always @xmath26 is the virtuality of the photon and @xmath2 is the bjorken @xmath2-variable .
@xmath27 is the fraction of the photons light cone momentum carried by the quark , @xmath28 is the transverse separation of the quark
antiquark pair , @xmath29 is the impact parameter of the original pair and @xmath30 is the rapidity variable .
@xmath31 is the propagator of the quark
antiquark pair through the nucleus , or , more correctly , the forward scattering amplitude of the @xmath20 pair with the nucleus ( hadron ) .
it obeys the equation @xcite @xmath32 + \frac{\alpha c_f}{\pi^2 } \int_0^y d y \
, \exp \left [ - \frac{4 \alpha c_f}{\pi } \ln \left ( \frac{x_{01}}{\rho } \right ) ( y - y ) \right]\end{aligned}\ ] ] @xmath33 , \end{aligned}\ ] ] with @xmath34 the propagator of a dipole of size @xmath35 at the impact parameter @xmath29 through the target nucleus or hadron , which was taken to be of glauber form in @xcite : @xmath36 - 1.\ ] ] here @xmath37 is the transverse area of the hadron or nucleus , @xmath15 is the atomic number of the nucleus , and @xmath38 is the gluon distribution of the nucleons in the nucleus , which was taken at the two gluon ( lowest in @xmath6 ) level @xcite .
( [ gla ] ) is written for a cylindrical hadron or nucleus .
in the large @xmath14 limit @xmath39 . in eq .
( [ eqn ] ) @xmath40 is an ultraviolet regulator @xcite .
( [ eqn ] ) resums all multiple hard pomeron exchanges in the @xmath0 structure function . the linear in @xmath41 term on the right hand side of eq .
( [ eqn ] ) corresponds to the usual bfkl evolution @xcite , whereas the quadratic term in @xmath41 is in a certain way responsible for triple pomeron vertices .
this quadratic term on the right hand side of eq .
( [ eqn ] ) introduces some damping effects , which at the end unitarize the growth of the bfkl pomeron .
we note that an equation similar to ( [ eqn ] ) was suggested by gribov , levin and ryskin ( glr ) in momentum space representation in @xcite and was subsequently studied by bartels and levin in @xcite .
however , usually by the glr equation one implies an equation summing all fan diagrams in the double leading logarithmic approximation .
( [ eqn ] ) resums all the fan diagram in the leading longitudinal logarithmic approximation . in the double logarithmic limit ( large @xmath26 ) eq .
( [ eqn ] ) reduces to the glr equation .
we also note that an equation very similar to eq .
( [ eqn ] ) has been proposed by balitsky in @xcite to be used in describing the wavefunctions of each of the quarkonia in onium
onium scattering , similarly to @xcite .
a renormalization group techniques based on mclerran venugopalan @xcite model have been developed by jalilian - marian , kovner , leonidov , and weigert @xcite leading to an equation which is also supposed to resum multiple hard pomeron exchanges @xcite .
unfortunately the resulting equation @xcite differs from our eq .
( [ eqn ] ) .
an equation summing all multiple pomeron diagrams in the double logarithmic limit has been proposed by ayala , gay ducati and levin @xcite , which is also different from the appropriate limit of our eq .
( [ eqn ] ) . in this paper
we are going to analyze eq .
( [ eqn ] ) . in sect .
ii we will find a solution of eq .
( [ eqn ] ) in the form of a series in eq .
( [ psol ] ) which will be constructed by solving eq .
( [ eqn ] ) perturbatively in the momentum space .
the details of transforming eq .
( [ eqn ] ) into momentum space will be given in the appendix a with the result given by eq .
( [ bsp ] ) .
the terms of the series will correspond to multiple pomeron exchange contributions .
partial sums of the series will be plotted in fig .
[ uni ] , which shows how multiple pomeron exchanges unitarize the amplitude .
we will argue that the kinematic region for which the series is convergent corresponds to the case when the dis cross sections and structure functions have not reached saturation . in a certain sense
we will define the saturation region as the kinematic regime where the series ceases to converge . by resumming the terms of the series corresponding to exchanges of large numbers of pomerons
we construct a function which we can analytically continue outside the region of convergence of the series of eq .
( [ psol ] ) . that would provide us with the ansatz for the solution of eq .
( [ bsp ] ) outside of the saturation region [ see eq .
( [ asym ] ) ] . we will prove that this ansatz actually is the only possible energy independent solution of eq .
( [ bsp ] ) . to transform our results into coordinate space on would have to separate the momentum integration in two regions , one of them being above and the other one being below the saturation scale .
since we will find only the asymptotic solution inside the saturation region our results may not be very precise in determining the coordinate space amplitude . a numerical solution of eq . ( [ eqn ] )
would probably allow one to achieve better precision . nevertheless for asymptotically high energies we can transform the amplitude into coordinate space obtaining eq .
( [ nsa ] ) .
thus we will show that multiple pomeron exchanges unitarize the forward amplitude , making it independent of energy at high energies , which is shown in fig .
[ nsat ] . in sect .
iii we will analyze the high energy asymptotics of the @xmath0 structure function . we will show that , contrary to some expectations , @xmath0 structure function does not become independent of energy at high energies .
it becomes linearly proportional to @xmath1 , still satisfying the froissart bound [ see eq .
( [ f2a ] ) ] . in sect .
iv we will discuss how in the framework of the dipole model one can describe diffractive scattering by including the multiple pomeron evolution in rapidity of eq .
( [ eqn ] ) in the quasi
elastic structure functions considered in @xcite .
we will derive an expression for the quasi elastic structure function @xmath42 ( [ f2diff ] ) and discuss its asymptotics ( [ f2da ] ) .
finally , we will conclude in sect .
v by summarizing our results and discussing the possible implications of the developed techniques on different issues of small-@xmath2 physics .
we start by considering a dis on a large cylindrical nucleus .
this assumption about the size and geometry of the nucleus is being done for simplicity of calculations and our results could be easily generalized to a large nucleus of any shape .
if the nucleus is sufficiently large we can assume that the forward amplitude of the scattering of the original quark
antiquark pair on the nucleus , @xmath43 , is a slowly varying function of the impact parameter of the virtual photon @xmath29 .
this is equivalent to assuming that the transverse extent of the dipole wave function is much smaller than the size of the nucleus , and that the impact parameter between the virtual photon and the nucleus is smaller than the radius of the nucleus @xmath44 .
we note that in order to construct @xmath43 in @xcite we integrated over the impact parameters @xmath45 s of the nucleons with which each of the pomerons in fig .
[ fan ] interacted .
that way @xmath29 does not have a meaning of the impact parameter of a pomeron exchange anymore and @xmath29 dependence in @xmath43 becomes purely `` geometrical '' . in a more formal language
we assume that @xmath46 , where @xmath47 is the typical transverse size of the dipole wave function , and @xmath48 ( @xmath49 is zero at the center of the nucleus ) .
if these conditions are satisfied we may neglect @xmath50 and @xmath51 compared to @xmath29 on the right hand side of eq .
( [ eqn ] ) , which means that the changes in @xmath52 when @xmath29 is varied by @xmath53 are negligibly small . thus suppressing the @xmath29 dependence
we write @xmath54 instead of @xmath43 in eq.([eqn ] ) . then differentiating eq .
( [ eqn ] ) with respect to @xmath55 we can rewrite it as @xmath56 n({\bf x}_{02 } , y)\end{aligned}\ ] ] @xmath57 with @xmath58 as the initial condition , where we also suppressed ( but did not neglect ) the impact parameter dependence of @xmath59 .
now one can see that the evolution equation ( [ 1diff ] ) itself in the large nucleus approximation does not impose any impact parameter @xmath29 dependence .
the latter is only included in the initial conditions given by @xmath60 . since @xmath59 is a slowly varying function of @xmath29 ( constant for a cylindrical nucleus ) our assumption about the slow dependence of @xmath41 on @xmath29 is self consistent .
physically this implies that before the collision , when the incoming virtual photon develops its dipole evolved wave function , it does not have any information about the nucleus .
the information about the thickness of the nuclear medium as a function of the impact parameter comes through @xmath59 when the dipoles interact with the nucleus .
let us rewrite @xmath61 we note that the inverse of this transformation is @xmath62 since @xmath63 does not depend on the direction of @xmath64 , @xmath65 does not depend on the direction of the relative transverse momentum of the @xmath20 pair @xmath66 . that allowed us to simplify the integrations in eqs .
( [ trfm ] ) and ( [ invtrfm ] ) .
now eq . ( [ 1diff ] ) becomes @xmath67 where @xmath68 is the eigenvalue of the bfkl kernel @xcite with @xmath69 . in eq .
( [ bsp ] ) the function @xmath70 is taken as a differential operator with @xmath71 acting on @xmath65 .
we also put @xmath72 instead of @xmath73 everywhere in the spirit of the large @xmath14 approximation .
the details of obtaining eq .
( [ bsp ] ) from eq .
( [ 1diff ] ) are presented in the appendix a. eq .
( [ bsp ] ) shows explicitly that the first , linear , term on its right hand side gives the usual bfkl equation , whereas the quadratic term corresponds to the triple pomeron vertex , which unitarizes the growth of the bfkl pomeron .
one can easily prove that eq .
( [ bsp ] ) has a unique solution satisfying given initial condition at @xmath74 .
the strategy of the proof is conventional : assume that there are two different solutions of eq .
( [ bsp ] ) satisfying given initial conditions .
then one can easily derive an equation for the function given by the difference of these solutions with the initial condition stating that the function is equal to zero when @xmath74 .
after a little analysis of the resulting equation with that initial condition one could easily see that the difference between the two solutions of the original eq .
( [ bsp ] ) is zero .
that way one proves uniqueness of the solution of eq .
( [ bsp ] ) .
we want to find a solution of eq .
( [ bsp ] ) satisfying the initial condition given by the bfkl pomeron contribution at relatively `` small '' rapidity @xmath75 . unfortunately finding an exact analytical solution of eq .
( [ bsp ] ) seems to be a very difficult task , partly because eq .
( [ bsp ] ) is non linear , partly because of a complicated structure of the bfkl kernel @xmath70 .
instead we are going to construct a series resulting from the perturbative solution of eq .
( [ bsp ] ) , which , in the region where it is convergent , provides us with the _ exact _ solution of eq .
( [ bsp ] ) .
as we will show below the region of convergence of that series corresponds to the regime when the dis cross sections and structure functions are _ not _ saturated . in the saturation regime the series diverges , but allows us to construct an asymptotic solution by analytical continuation .
let us start by assuming that at @xmath75 the function @xmath76 is very small , @xmath77 .
this allows us to neglect the quadratic term in eq .
( [ bsp ] ) , since it would be of higher order in @xmath76 .
( [ bsp ] ) becomes @xmath78 which corresponds to the usual bfkl equation @xcite .
the solution of eq .
( [ bfkl ] ) is given by @xmath79 \
, c ( k),\ ] ] with @xmath80 being some unknown function of @xmath81 , to be specified by initial conditions .
rewrite @xmath80 as a mellin transformation @xmath82 where @xmath83 is some scale characterizing the nucleus or hadron . here
the integration in @xmath84 runs parallel to the imaginary axis to the right of all the singularities of @xmath85 . substituting eq .
( [ cmel ] ) into eq .
( [ solb ] ) we obtain @xmath86 \left ( \frac{k}{\lambda } \right)^\lambda c_\lambda .\ ] ] assuming that the transverse momentum of the @xmath20 pair @xmath81 is not very large , such that @xmath87 and that @xmath88 is a slowly varying function of @xmath84 we come to the conclusion that the integral in eq .
( [ nmel ] ) is dominated by the saddle point in the exponential in the vicinity of @xmath89 in @xmath90 .
expanding @xmath91 , with @xmath92 the riemann zeta function , we can perform a saddle point approximation in the integral of eq .
( [ nmel ] ) .
the saddle point is given by @xmath93 where we have defined @xmath94 which is small .
the result of the integration yields @xmath95 } { \sqrt{14 \alpha n_c \zeta ( 3 ) y } } \exp \left ( - \frac{\pi}{14 \alpha n_c \zeta ( 3 ) y } \ln^2 \frac{k}{\lambda } \right ) \equiv p_1 ( k , y),\ ] ] a factor usually associated with the zero momentum transfer single hard pomeron exchange , which we denoted by @xmath96 .
@xmath97 is determined from initial conditions . if the initial conditions are given by the two gluon exchange approximation , then @xmath98 . that way , parametrically , when @xmath75 , the amplitude @xmath99 , and our assumption of the smallness of @xmath76 at @xmath75 is justified .
( [ n1 ] ) provides us with the usual single bfkl pomeron solution .
now we are going to find corrections to this expression resulting from eq .
( [ bsp ] ) , which will form a series @xmath100 with @xmath101 being of the order of @xmath102 .
the first correction is @xmath103 . rewriting @xmath104 , substituting it back into eq .
( [ bsp ] ) , neglecting non linear terms in @xmath103 and employing eq .
( [ bfkl ] ) we obtain @xmath105 where @xmath106 is given by eq .
( [ nmel ] ) .
one can easily see that the solution of eq .
( [ n2eq ] ) with the initial condition that it disappears at small @xmath55 is @xmath107 [ { \tilde n}_1 ( k , y)^2 ] .\ ] ] plugging @xmath106 of eq .
( [ nmel ] ) into eq .
( [ n21 ] ) and integrating over @xmath108 yields @xmath109 } \end{aligned}\ ] ] @xmath110 \right ) - \exp \left ( 2 \ , \frac{\alpha n_c}{\pi } \ , y \ , \chi ( - \lambda_1 - \lambda_2 ) \right ) \right ] .\end{aligned}\ ] ]
( [ n2eq ] ) corresponds to putting one triple pomeron vertex ( given by quadratic term in @xmath111 ) into the dis diagram , so that @xmath103 consists of one pomeron splitting into two .
the integration over @xmath108 in eq .
( [ n21 ] ) corresponds to integration over all possible rapidity positions of the triple pomeron vertex between the incoming quark
antiquark pair and the target .
thus the first exponential term in the square brackets in eq .
( [ n22 ] ) corresponds to the case when the pomeron splitting occurs very close ( in rapidity ) to the @xmath20 pair , so that most of the rapidity interval is covered by two pomeron contribution . the second exponential term in eq .
( [ n22 ] ) corresponds to the situation when the splitting happens very close to the target , so that most of the interaction corresponds to the single pomeron exchange .
therefore the first exponential term in eq .
( [ n22 ] ) corresponds to two pomeron contribution and the second term corresponds to the one pomeron contribution .
similar conclusions have been reached by navelet and peschanski in @xcite when analyzing two pomeron contribution to onium
onium scattering in dipole model @xcite .
as we will see below the two pomeron contribution gives a term proportional to @xmath112 $ ] , while the one pomeron exchange is proportional to @xmath113 $ ] .
that way the two pomeron term dominates in @xmath103 and the single pomeron term could be safely neglected .
one should also remember that @xmath103 is proportional to the square of @xmath85 , so that the single pomeron exchange term from @xmath103 represents just an order @xmath114 ( i.e. , next - to - next - to - leading order ) correction to @xmath106 . neglecting this contribution we rewrite eq .
( [ n22 ] ) as @xmath109 } \end{aligned}\ ] ] @xmath115 \right ) .\end{aligned}\ ] ] here again we can perform the integration over @xmath116 and @xmath117 in the saddle point approximation .
the saddle points of @xmath116 and @xmath117 integrations would have been independent if it were not for the @xmath118 in the denominator .
the problem is that @xmath70 has a pole at @xmath119 , and since the saddle points of @xmath116 and @xmath117 integrations are located in the vicinity of @xmath120 [ see eq .
( [ sp ] ) ] this could influence the exact location of the saddle points of the integral .
however , as one can see analyzing the saddle points of eq .
( [ n23 ] ) , if @xmath121 the effect of @xmath118 could be safely neglected .
this imposes an additional constraint on the possible values of transverse momentum @xmath81 .
remember that in order to perform the saddle point integration in eq .
( [ nmel ] ) we had to assume that @xmath122 .
these two conditions specify the range of @xmath81 in which our analysis is applicable .
the condition @xmath123 is not very hard to satisfy , since , after all , the transverse momentum @xmath81 should be large enough for perturbative qcd to be applicable , i.e. , @xmath124 .
we will return to this question below . performing the saddle point integration in eq .
( [ n23 ] ) in the spirit of the above discussion one obtains @xmath125 with @xmath126 defined in eq .
( [ a ] ) . in arriving at eq .
( [ p2 ] ) we had to also include the non singular terms in @xmath127 in the denominator of eq .
( [ n23 ] ) .
this was done for the reasons which will become apparent below , when we will explore the convergence of the series we will obtain .
we can continue construction of our perturbative solution by finding the third order correction @xmath128 .
writing @xmath129 we obtain the following equation for @xmath128 from eq .
( [ bsp ] ) @xmath130 the solution to eq .
( [ n3eq ] ) is given by @xmath131 [ \ , 2 \ , { \tilde n}_1 ( k , y ) \ , { \tilde n}_2 ( k , y ) ] , \ ] ] similarly to eq .
( [ n21 ] ) . substituting @xmath132 from eq .
( [ nmel ] ) and @xmath133 from eq .
( [ n23 ] ) , performing the integration over @xmath108 and neglecting the two pomeron contribution which arises in the same way as one pomeron contribution appeared in eq .
( [ n22 ] ) we reduce eq .
( [ n31 ] ) to the following form @xmath134 } \end{aligned}\ ] ] @xmath135 } \
, \exp \left ( 2 \ , \frac{\alpha n_c}{\pi } \ , y \ , [ \chi ( - \lambda_1 ) + \chi ( - \lambda_2 ) + \chi ( - \lambda_3 ) ] \right ) .\end{aligned}\ ] ] after saddle point integrations eq .
( [ n33 ] ) yields @xmath136 now we are ready to write down the general expression for the perturbation series of ( [ ps ] ) . an obvious generalization of eqs .
( [ n21 ] ) and ( [ n31 ] ) allows us to write @xmath137 \
, \left ( \ , \sum_{m=1}^{n-1 } \ , { \tilde n}_m ( k , y ) \ , { \tilde n}_{n - m } ( k , y ) \right).\ ] ] as one could see from above the term of order @xmath138 in the series is proportional to @xmath139 .
each pomeron splitting through the integration over @xmath108 introduces a denominator of the type present in eqs .
( [ n23 ] ) and ( [ n33 ] ) .
the series becomes @xmath140 where the coefficients of the series should be determined from the relationship @xmath141 with @xmath142 \}^{-1 } , \hspace*{1 cm } \mbox{even } \ , \
\\ - \frac{1}{2 } \ , \ { n \ , 2 \ , \ln 2 - \psi ( 1 ) + \psi ( n/2 ) \}^{-1 } , \hspace*{1 cm } \mbox{odd } \ , \ , n .
\end{array } \right.\ ] ] eqs .
( [ psol ] ) , ( [ fn ] ) and ( [ an ] ) provide us with the exact solution of eq .
( [ bsp ] ) in the kinematic region where the series of eq .
( [ psol ] ) is convergent . since each term @xmath143 is given by eq .
( [ neq ] ) the infinite sum of these terms ( [ psol ] ) satisfies eq .
( [ bsp ] ) giving the exact solution to it in the region where it is finite . to obtain eq .
( [ an ] ) we had to assume that @xmath144 ( as @xmath145 ) and expand the @xmath146 functions in the denominators to the first non - singular terms in @xmath147 .
( [ psol ] ) represents an expansion of the amplitude in terms of multiple hard pomeron exchanges , the @xmath138th term in the series corresponding to the @xmath138-pomeron exchange contribution .
( [ fn ] ) and ( [ an ] ) yield us with the prescription of calculation of an arbitrary term in the series for @xmath148 .
one could explore the effects of the higher order corrections to the bfkl pomeron arising from eq .
( [ bsp ] ) by plotting the partial sums of the series of eq .
( [ psol ] ) . defining @xmath149 , we plot @xmath150 as a function of @xmath96 in fig .
[ uni ] for @xmath151 .
we fixed @xmath127 to be a constant equal to @xmath152 for simplicity .
the single bfkl pomeron contribution corresponds to the first term in the series , @xmath153 , and is given by the straight line in fig .
[ uni ] , because we plotted @xmath96 as an argument on the horizontal axis .
each partial sum @xmath150 gives us some approximation to the exact result for @xmath111 , and then starting from some value of @xmath96 starts diverging .
inclusion of higher orders gives better approximations .
that way the thick line formed by the overlap of the curves corresponding to different partial sums represents the true value of @xmath111 , the solution of eq .
( [ bsp ] ) to which the series is converging .
one can see that the exact solution shown by that thick line lies below the single bfkl pomeron contribution ( straight line ) , deviating from it more and more as the energy ( and , therefore , @xmath96 ) increases .
thus the multiple pomeron contributions start unitarizing the bfkl pomeron .
an analogous conclusion was reached by salam in @xcite through numerical study of the multiple pomeron exchanges in the onium
onium scattering in the framework of the dipole model @xcite . to find the radius of convergence and the high energy asymptotic behavior of the series in eq .
( [ psol ] ) let us study the terms corresponding to large values of @xmath138 .
these terms describe the multiple pomeron exchange contributions .
we would like to explore very large @xmath138 , such that @xmath154 .
then the coefficients @xmath155 determining the recursion relations in ( [ fn ] ) for the coefficients of the series become @xmath156 thus , redefining @xmath157 we obtain the following recursion relations instead of eq .
( [ fn ] ) @xmath158 for very large @xmath138 eq .
( [ recur ] ) could be satisfied up to order @xmath159 corrections with the ansatz @xmath160 where @xmath161 could be some arbitrary constant .
here we have to point out that since eq .
( [ recur ] ) is written for very large @xmath138 we can not use @xmath162 as a `` normalization '' condition anymore and that way we have lost the information which would have allowed us to fix the value of @xmath161 . in principle @xmath161 should depend on @xmath147 , but this function is a much slower varying function of @xmath81 and @xmath55 than @xmath163 , so we assume @xmath161 to be constant in our analysis .
if we fix @xmath164 then numerical estimates show that @xmath165 , which allows one to hope that for asymptotics of eq .
( [ fn ] ) @xmath161 is not very large and does not introduce a large correction to our analysis . using the ansatz of eq .
( [ ans ] ) together with eq .
( [ resc ] ) in eq .
( [ psol ] ) we conclude that the constructed perturbation series is convergent as long as @xmath166 outside the region specified by the condition of eq .
( [ conv ] ) the series in eq .
( [ psol ] ) is not convergent anymore .
multiple pomeron exchange contributions corresponding to large values of @xmath138 become much larger and , consequently , more important than the one- or two - pomeron exchanges .
defining the saturation momentum scale @xmath167 by the following condition @xmath168 we obtain ( for not very large @xmath167 ) @xcite @xmath169 } { \sqrt{14 \alpha n_c \zeta ( 3 ) y}}.\ ] ] using this quantity we can say that for the transverse momenta of the quark
antiquark pair greater than the saturation momentum , @xmath170 , the series converges . for the transverse momenta below the saturation scale , @xmath171 , that is , inside the saturation region , the series diverges ceasing to be a reliable approximation to the solution of eq .
( [ bsp ] ) .
one might wonder whether the condition @xmath172 , which we needed to perform the saddle point integrations above , would prevent us from exploring the high energy asymptotics with the series of eq .
( [ psol ] ) by putting an upper limit on the possible values of @xmath55 for fixed @xmath81 .
really , this condition provides us with the limit on the range of rapidities for the perturbative solution we found @xmath173 however we also have to note that the series ( [ psol ] ) is convergent as long as @xmath174 , i.e. , outside of the saturation region .
this yields us with another constraint on @xmath55 @xmath175 now one can see that the two conditions ( [ ysp ] ) and ( [ ysat ] ) are roughly of the same order of magnitude .
thus the condition @xmath176 does not impose any additional constraint on the kinematic region of the applicability of the series in eq .
( [ psol ] ) since the series diverges in the region where the condition is violated anyway .
now we want to find some approximation to the solution of eq .
( [ bsp ] ) inside the saturation region .
the saturation region is specified by the condition that @xmath177 and could be achieved either by decreasing the transverse momentum or by increasing the center of mass energy of the system .
first we note that one may find a correction to eq .
( [ ans ] ) . rewriting it as @xmath178 where @xmath179 for large @xmath138 and @xmath180 for small @xmath138 , i.e. , it satisfies eq .
( [ fn ] ) with the large @xmath138 asymptotics given by leading in @xmath138 behavior of eq .
( [ ans ] ) . substituting eq .
( [ cor1 ] ) into eq .
( [ ans ] ) , expanding it to the lowest order in @xmath181 and neglecting @xmath182 for small values of @xmath138 one obtains @xmath183 using eqs .
( [ cor2 ] ) , ( [ ans ] ) and ( [ resc ] ) in ( [ psol ] ) we could sum up leading in @xmath138 terms of the series ( [ psol ] ) obtaining some expressions which could be analytically continued outside the region of convergence of the series of eq .
( [ psol ] ) . this way we might obtain an expression for the solution of eq . ( [ bsp ] ) inside the saturation region , that is for large @xmath96 .
while the leading in @xmath138 part of the series given by eq .
( [ ans ] ) gives a function which goes to a constant for large @xmath96 , the subleading terms given by eq .
( [ cor2 ] ) yield us with @xmath184 analytically continuing @xmath185 in the region of large @xmath96 we obtain for asymptotically large energies @xmath186 .
of course we do not know whether the higher order corrections in @xmath159 would not give us some other function of @xmath96 which would be large for large @xmath96 possibly interfering with our @xmath187 ansatz .
therefore @xmath187 should be considered only as a a possible high energy asymptotics of @xmath76 and the above argument should not be taken as a proof of this statement . to justify that the above ansatz of @xmath188 is the solution of eq .
( [ bsp ] ) at asymptotically high energies we first note that @xmath189 with @xmath190 some large momentum scale ( @xmath191 ) satisfies the equation @xmath192 really , rewriting @xmath193 where the contour @xmath194 is a circle of radius less than one centered around @xmath195 , one can see that since @xmath90 also has a pole at @xmath195 then @xmath196 we can assume that in the saturation region @xmath197 ceases to depend on @xmath55
. then eq .
( [ bsp ] ) reduces to eq .
( [ hen ] ) . to find the most general solution of eq .
( [ hen ] ) we represent @xmath76 as a series @xmath198 and substitute it into eq .
( [ hen ] ) in order to find the unknown coefficients @xmath199 . using the technique outlined above for @xmath200
we can show that @xmath201 = \left\ { \begin{array}{c } \frac{2}{m + 1 } \left ( \frac{k}{m } \right)^n \ , \ln^{m+1 } \frac{m}{k } , \hspace*{1 cm } \mbox{even } \ , \ , n , \\ \\ 2 \ , \chi ( -n ) \left ( \frac{k}{m } \right)^n \ , \ln^m \frac{m}{k } , \hspace*{1 cm } \mbox{odd } \ , \ , n. \end{array } \right.\ ] ] with the help of eq .
( [ chiac ] ) one can show that in order for the series of eq .
( [ hens ] ) to satisfy eq .
( [ hen ] ) all of its coefficients must be zero with the exception of @xmath202 .
thus we proved that the most general energy independent solution of eq .
( [ bsp ] ) is given by eq .
( [ log ] ) . at asymptotically high energies
there is only one large momentum scale in the scattering of the quark
antiquark pair on a hadron or nucleus the saturation scale @xmath167 .
this scale is also suggested by the ansatz of eq .
( [ sans ] ) .
therefore we should put @xmath203 and use eq .
( [ log ] ) as a solution of eq .
( [ bsp ] ) in the region when @xmath204 .
the derivative of this new solution with respect to rapidity is of the order of @xmath205 and could be neglected compared to any of the terms on the right hand side of eq .
( [ bsp ] ) , each of them being of the order of @xmath206 .
this could be done since @xmath207 .
therefore we conclude that in the saturation region @xmath208 now that we know the solution for @xmath209 outside of the saturation region given by eqs .
( [ psol ] ) , ( [ fn ] ) and ( [ an ] ) and the high energy asymptotics inside the saturation region given by eq .
( [ asym ] ) , one might want to construct the solution in coordinate space ( @xmath210 ) using the fourier transformation of eq .
( [ trfm ] ) . to do
that one would have to integrate over all values of the transverse momentum @xmath81 from @xmath211 to @xmath212 . if @xmath213 then the transverse momentum , which in eq .
( [ trfm ] ) is effectively cut off by the bessel function and , therefore , varies approximately from @xmath211 to @xmath214 , will go through a range of values both above and below @xmath167 . for the first case one would have to use the perturbation series which we constructed as @xmath76 , while for the second case one might use the asymptotic value of @xmath76 given by eq .
( [ asym ] ) , although a perturbative expansion around it would be more accurate .
the result would be a complicated combination of special function and we are not going to list it here . if @xmath215 then the transverse momentum in eq .
( [ trfm ] ) simply gets cut off by @xmath216 and always stays within the saturation region .
thus one could just perform a fourier transformation of formula ( [ asym ] ) using eq .
( [ trfm ] ) .
the result is @xmath217 this corresponds to the blackness of the total cross section of the quark
antiquark pair on the target nucleus , since it is given by @xmath218 for @xmath219 and for a cylindrical nucleus of radius @xmath44 as described above .
that way the total cross section is independent of energy at asymptotically high energies and is completely unitary .
it reaches its `` geometrical '' limit ( [ geo ] ) which could be predicted even from quantum mechanics . that way we have shown that @xmath220 given by the solution of eq .
( [ bsp ] ) behaves like a single bfkl pomeron exchange contribution at moderately high energies ( @xmath75 ) , which follows from the perturbation series we have constructed , and as energy increases to very high quantities ( @xmath221 ) , saturates to a constant independent of energy . a qualitative sketch of @xmath220 as a function of @xmath222 ( one pomeron contribution )
is shown in fig .
for a large cylindrical nucleus the @xmath0 structure function is given by @xmath223 which follows from eq .
( [ f2 ] ) . we can rewrite eq . ( [ f21 ] ) in momentum space using eq .
( [ trfm ] ) @xmath224 where @xmath225 employing eqs .
( [ ft ] ) and ( [ fl ] ) we obtain @xmath226 \ , \frac{2}{3 a^2 } \ , _
2f_1 \left ( 2 , 3 , 2.5 , - \frac{k^2}{4 a^2 } \right ) + \right.\end{aligned}\ ] ] @xmath227 \right\},\end{aligned}\ ] ] where @xmath228 is a hypergeometric function .
( [ f22 ] ) shows that in order to write down an expression for the @xmath0 structure function one has to know @xmath197 in the areas above and below saturation .
the situation is similar to the fourier transformation of the previous section .
the wave function @xmath229 given by eq .
( [ hyp ] ) becomes very small for transverse momenta @xmath230 , effectively providing an upper cutoff on the @xmath81 integration .
thus if @xmath231 the integral over @xmath81 includes @xmath232 and @xmath170 and to calculate it we have to use the solution for @xmath76 both inside and outside of the saturation region
. we can again propose the approximation outlined in sect .
ii for calculation of @xmath0 , which consists of using perturbation series of eq .
( [ psol ] ) outside saturation region and approximating the solution inside that region by eq .
( [ asym ] ) .
the result would also include a complicated list of special functions which we are not going to list here .
we have to admit that numerical solution of eq .
( [ eqn ] ) could give a more precise results for @xmath0 .
here we can just point out that at moderately high energies the behavior of the @xmath0 structure function will be dominated by single bfkl pomeron exchange , providing us with the following formula which could be obtained by substituting the first term in the series of eq .
( [ psol ] ) into eq .
( [ f22 ] ) @xmath233 } { \sqrt{14 \alpha n_c \zeta ( 3 ) y } } , \hspace*{1 cm } y \sim \frac{1}{\alpha},\ ] ] where we have neglected the transverse momentum diffusion term in the exponent of eq .
( [ n1 ] ) and @xmath234 .
( [ f2b ] ) is valid only for moderately high energies , when the saturation momentum @xmath167 is small and we can neglect the @xmath81 integration in eq .
( [ f22 ] ) below that scale .
when @xmath235 or the energy gets very high to sufficiently increase @xmath167 the integral over @xmath81 in eq .
( [ f22 ] ) is limited to @xmath236 , so that we can use eq .
( [ asym ] ) for @xmath197 .
the result yields @xmath237 as energy becomes very large , corresponding to @xmath238 , @xmath239 becomes approximately equal to @xmath240 .
therefore @xmath241 we conclude that @xmath242 at asymptotically high energies
. this conclusion may seem to be a little unusual , so we will reproduce the same result using more conventional coordinate space language . as was shown in the previous section
the forward amplitude of the @xmath243 pair scattering on a nucleus is @xmath244 , when @xmath245 .
when @xmath246 one could argue that the amplitude is roughly proportional to some positive power of @xmath47 , at least @xmath247 , as given by one pomeron exchange .
thus , for @xmath235 we can neglect the part of the integral with @xmath248 in eq .
( [ f21 ] ) , since the wave function @xmath249 in eq .
( [ f21 ] ) behaves like @xmath250 and is not singular enough to make that portion of the integral significant . that way , expanding the modified bessel functions in @xmath249 of eq .
( [ ft ] ) and integrating over @xmath27 we obtain @xmath251 the upper cutoff on the @xmath252 integration in eq .
( [ f23 ] ) is provided by the fact that the modified bessel functions fall off exponentially at the large values of the argument . integrating over @xmath252 in eq .
( [ f23 ] ) we arrive at eq .
( [ f2a ] ) . that way we have shown that the @xmath0 structure function given by the solution of eq .
( [ neq ] ) is proportional to the single bfkl pomeron exchange contribution ( [ f2b ] ) at moderately high energies with @xmath253 , and , as energy increases the @xmath0 structure function unitarizes , becoming linearly proportional to @xmath1 ( [ f2a ] ) .
we stress again that even though the cross section of the quark
antiquark pair of a fixed transverse size @xmath47 scattering on a nucleus saturates to a constant at large @xmath55 , the integral over @xmath47 makes @xmath0 depend on @xmath55 , such that @xmath254 at very large @xmath55 .
in @xcite a diffractive deep inelastic scattering process was considered , an amplitude of which is shown in fig .
[ ampl ] . a virtual photon scatters on a nucleus
elastically , leaving it intact .
two quarks in the final state in fig . [ ampl ] may become a pair of jets or form a vector meson .
these type of processes were called quasi elastic virtual photoproduction in @xcite . the interactions between the quark
antiquark pair and the nucleus were taken in the quasi
classical approximation , similar to @xcite .
each nucleon in @xcite would interact with the @xmath243 pair by a two gluon exchange .
strictly speaking this approximation corresponds to not very high energies , when the logarithms of @xmath3 resulting from qcd evolution are not large enough to become important .
that is why the approximation is called quasi classical . in @xcite
it provided us with a simple glauber type expression for the diffractive , or , more correctly , quasi
elastic structure function @xmath42 [ cf .
( 20 ) in @xcite ] @xmath255 + 4 q^2 z^2 ( 1 - z)^2 k_0 ^ 2 ( x_{01 } a ) \right\}\end{aligned}\ ] ] @xmath256 where we again assumed that the quarks are massless and have only one flavor .
@xmath34 is the glauber propagator of the quark
antiquark pair , which for the case of a cylindrical nucleus is given by eq .
( [ gla ] ) .
one can obtain eq .
( [ f2diff0 ] ) from eq .
( [ f2 ] ) by first noting that the quasi
classical limit corresponds to putting @xmath74 ( no evolution in rapidity ) in it and @xmath257 . recalling that , as was argued in @xcite , one can rewrite the total and elastic cross sections for dis on a nucleus in terms of the @xmath258-matrix at a given impact parameter of the collision , @xmath259 , yielding @xmath260 \end{aligned}\ ] ] @xmath261 ^ 2 \end{aligned}\ ] ] and associating @xmath0 with @xmath262 and @xmath42 with @xmath263
, we can easily derive eq .
( [ f2diff0 ] ) from eq .
( [ f2 ] ) @xcite .
now let us imagine our quasi elastic process at very high energies
. then , strictly speaking , one has to include the evolution in rapidity in the structure function @xmath42 .
we still do not want to have particles being produced in the central rapidity region . one could model the interaction by a single pomeron exchange between the quark
antiquark pair and the nucleus , similarly to what was done previously in @xcite and references mentioned there for elastic scattering .
here we are going to propose inclusion of multiple pomeron exchanges in @xmath42 .
since the above formalism allows us to sum up pomeron fan diagrams in the leading logarithmic approximation , we can just insert this evolution in the amplitude of the quasi elastic process .
the diagram of the process is depicted in fig .
[ diffr ] . the diffractive rapidity gap between the nucleus in the final state and the remains of the @xmath20 pair results not only from a single pomeron exchange , but also from multiple pomeron exchanges .
still nothing is produced in the central rapidity region .
the corresponding expression for the diffractive structure function @xmath42 is easy to derive .
we know the expression for the total @xmath0 structure function in eq .
( [ f2 ] ) , which includes all the discussed multiple pomeron exchanges .
one can easily argue along the lines outlined in deriving eq .
( [ f2diff0 ] ) , that for the case of multiple pomeron exchanges we again have a relationship between the total and elastic cross sections similar to the one given by eqs .
( [ stot ] ) and ( [ sel ] ) . that way we conclude that to obtain a formula for @xmath42 one has to substitute @xmath264 in eq .
( [ f2diff0 ] ) with @xmath265 .
the result is @xmath255 + 4 q^2 z^2 ( 1 - z)^2 k_0 ^ 2 ( x_{01 } a ) \right\}\end{aligned}\ ] ] @xmath266 where @xmath265 is , in the large @xmath14 limit , a solution of eq .
( [ eqn ] ) .
( [ f2diff ] ) is our master formula for the diffractive ( quasi elastic ) structure function .
when @xmath74 ( no evolution ) it reduces to eq .
( [ f2diff0 ] ) . using the results of sect .
ii we can derive the asymptotics of the diffractive structure function . at moderately high energies the one pomeron exchange contribution dominates the diffractive amplitude of fig .
[ diffr ] . again ,
similar to the previous section we assume that since at @xmath75 the saturation scale @xmath167 is small one can neglect the @xmath81 integration below @xmath167 . performing a fourier transformation ( [ trfm ] ) of the first term in the series of eq .
( [ psol ] ) and substituting the result into eq .
( [ f2diff ] ) we obtain for a cylindrical nucleus @xmath267 } { 14 \alpha n_c \zeta ( 3 ) y } \ ,
y , \hspace*{1 cm } y \sim \frac{1}{\alpha},\ ] ] where we had to cut off the @xmath27 integral in eq .
( [ f2diff ] ) by @xmath268 and we again neglected the transverse momentum diffusion term in eq .
( [ n1 ] ) . the diffractive structure function of eq .
( [ f2db ] ) behaves like a square of a single pomeron exchange contribution , similar to what was found for elastic parton scattering in @xcite .
even though @xmath269 in eq .
( [ f2db ] ) is growing with energy faster than @xmath270 in eq .
( [ f2b ] ) , unitarity is not violated , since both expressions are valid in the region of relatively small @xmath167 corresponding to @xmath271 . thus at moderate energies @xmath272 , as @xmath273 .
we also note that formula ( [ f2db ] ) was derived assuming that the photon s virtuality @xmath26 is not very large .
therefore , the fact that @xmath42 in eq .
( [ f2db ] ) has no powers of @xmath274 in the numerator , whereas the corresponding @xmath0 given by eq .
( [ f2b ] ) is proportional to @xmath274 , does _ not _ imply that @xmath42 is not a leading twist effect , violating the theoretical predictions @xcite and experimental observations @xcite .
for the case of large @xmath274 one should perform a different saddle point approximation for the terms in the perturbative series given by eqs .
( [ nmel ] ) , ( [ n22 ] ) , etc . , [ see @xcite ] , getting a different series solving eq .
( [ bsp ] ) outside of the saturation region .
this should result in concluding that @xmath42 given by the first term in that series is not suppressed by any powers of @xmath274 compared to @xmath0 .
however , doing that is beyond the scope of this paper . at very high energies , when @xmath275 is large enough so that @xmath276 , the quasi elastic structure function @xmath42 saturates .
similarly to sect .
iii we can derive the high energy asymptotic behavior of eq .
( [ f2diff ] ) .
putting @xmath277 for @xmath278 and imposing the cutoffs on the @xmath252 integration in eq .
( [ f2diff ] ) analogously to what was done in deriving eq .
( [ f23 ] ) we deduce @xmath279 that way the @xmath42 also becomes linearly proportional to the logarithm of center of mass energy energy at very high energies , similarly to the total structure function @xmath0 .
comparing eq .
( [ f2da ] ) with the high energy behavior of the @xmath0 structure function given by eq .
( [ f2a ] ) we conclude that @xmath42 is still smaller than @xmath0 and unitarity is again not violated .
one can see that at this asymptotically high energy @xmath280 , which corresponds to the prediction for the black cross sections resulting from quantum mechanics .
in this paper we have found a solution of the evolution equation ( [ eqn ] ) which resums all pomeron fan diagrams in the leading logarithmic limit for large @xmath14 , that was proposed in @xcite .
the solution consists of two parts . outside of the saturation region
we have constructed a perturbative series in momentum space given by eqs .
( [ psol ] ) , ( [ fn ] ) and ( [ an ] ) .
we have demonstrated that the series tends to unitarize the single bfkl pomeron exchange contribution ( see fig .
[ uni ] ) . in deriving formula ( [ bsp ] ) , which led to eqs .
( [ psol ] ) , ( [ fn ] ) and ( [ an ] ) , we had to assume that dis cross sections are slowly varying functions of the impact parameter , which corresponds to scattering on a large nucleus . only in the large nucleus case we could argue that the fan diagrams dominate the total dis cross sections . in order to perform the saddle point approximation in eqs .
( [ nmel ] ) , ( [ n22 ] ) , ( [ n33 ] ) and for the series ( [ psol ] ) to be convergent we had to assume that @xmath281 and @xmath282 , restricting the possible values of @xmath274 to a certain not very broad range .
nevertheless the abovementioned assumptions did not change the qualitative behavior of the obtained result and the conclusions could be easily extended to large @xmath274 by performing a different saddle point approximation . inside the saturation region , for @xmath235
, we found an approximate solution given by eq .
( [ asym ] ) corresponding to the saturation of the total scattering cross section of a quark antiquark pair of a given transverse size on a nucleus to a constant independent of energy ( [ nsa ] ) . that way we have shown that the coordinate space forward amplitude of the @xmath20 scattering on a nucleus or hadron grows like a bfkl pomeron exchange contribution at moderately high energies ( [ n1 ] ) , and , as energy gets very large it saturates to a constant ( [ nsa ] ) . a qualitative plot of @xmath210 is presented in fig .
[ nsat ] .
we have to admit that in order to construct a solution of eq .
( [ eqn ] ) in coordinate space , @xmath210 , and the corresponding structure function @xmath0 one has to have a better knowledge of the momentum space solution inside the saturation region
. a numerical solution of eq . ( [ eqn ] ) would probably be very helpful in determining exact values of @xmath210 and @xmath0 at intermediately large rapidities .
we have shown that the @xmath0 structure function given by the solution of eq .
( [ eqn ] ) grows as a power of energy at moderately high energies , corresponding to rapidities of the order of @xmath271 .
this behavior , presented in eq .
( [ f2b ] ) , reproduces the usual single bfkl pomeron exchange contribution to the structure function . as energy grows very high , corresponding to the rapidities of the order of @xmath283 \ln ( 1 / \alpha^2 ) $ ] , the @xmath0 structure functions unitarizes , becoming linearly proportional to @xmath1 , which satisfies froissart bound
this is shown in eq .
( [ f2a ] ) .
thus we have shown that the bfkl pomeron on the nucleus is unitarized by summation of pomeron fan diagrams .
we also note that we derived an expression for the quasi elastic ( diffractive ) structure function @xmath42 presented in eq .
( [ f2diff ] ) , which also includes all multiple pomeron exchanges .
we have shown that at moderately high energies the obtained expression behaves like a square of the bfkl pomeron contribution [ eq .
( [ f2db ] ) ] and at very high energies it saturates and becomes a linear function of @xmath1 [ eq .
( [ f2da ] ) ] , being equal to a half of the structure function @xmath0 , as expected for black cross sections . in @xcite
it was argued that the dipole model provides us with the techniques that allow one to resum pomeron fan diagrams at any subleading logarithmic order in the large @xmath14 limit .
that is , if one calculates the next - to - leading order kernel in the dipole model , the resulting equation for the generating functional @xmath284 [ see @xcite ] would provide us with the equation resumming all nlo pomeron fan diagrams .
the resulting equation would be , probably , more complicated than our eq .
( [ eqn ] ) .
it would have a much more sophisticated kernel @xcite and would have cubic terms in @xmath41 on the right hand side .
however , the analysis presented in this paper allows one to hope that this resummation will cure some of the problems of the recently calculated nlo bfkl kernel @xcite .
resummation of one loop running coupling corrections , which are a part of the nlo bfkl kernel , brings a factor of @xmath285 in the one pomeron exchange amplitude , with the constant @xmath286 for three flavors @xcite .
since it introduces an even faster growth with energy than the leading order bfkl pomeron , this factor seems to complicate the problem of unitarization of the pomeron .
however , the inclusion if this factor of @xmath287 in our definition of @xmath96 in eq .
( [ n1 ] ) would not change the perturbation series of eq .
( [ psol ] ) and the quark
antiquark scattering amplitude would still be unitarized by multiple pomeron exchanges .
of course a careful inclusion of higher order corrections would lead to much more complicated results .
nevertheless we may hope that inclusion of one loop running coupling corrections would just modify the one pomeron exchange contribution and the multiple pomeron exchanges would still unitarize the total dis cross section .
however , a rigorous proof of this statement is beyond our goals here .
finally we point out that one can use the developed technique to construct a gluon distribution function @xmath288 of the nucleus including multiple hard pomeron exchanges similarly to the way we have constructed the @xmath0 structure function . to do that one has to consider scattering of the `` current '' @xmath289 on the nucleus , similarly to @xcite .
then , in the large @xmath14 limit , it should be possible to derive an equation for the generating functional of the dipole wave function of the current @xmath290 , which , at the end , would provide us with an expression for the gluon distribution function @xmath291 .
after this paper was finished the author has learned about another effort to solve eq .
( [ eqn ] ) which was carried out in @xcite .
i wish to thank ulrich heinz , edmond iancu and cern theory division for their hospitality and support during the final stages of this work .
i would like to thank gregory korchemsky , andrei leonidov , genya levin , larry mclerran , alfred mueller , raju venugopalan and samuel wallon for many helpful and encouraging discussions .
this work is supported in part by doe grant de - fg02 - 87er40328 .
here we are going to show how the fourier transformation of eqs .
( [ trfm ] ) and ( [ invtrfm ] ) reduces eq .
( [ 1diff ] ) to eq .
( [ bsp ] ) .
first we note that @xcite @xmath292 which would allow us to integrate independently over @xmath293 and @xmath294 in the kernel of eq .
( [ 1diff ] ) . using the jacobian of eq .
( [ jac ] ) in the second ( quadratic ) term on the right hand side of eq .
( [ 1diff ] ) , multiplying both sides of eq .
( [ 1diff ] ) by @xmath295 , integrating over @xmath252 from @xmath211 to @xmath212 and employing @xmath296 we obtain @xmath297 n ( { \bf x}_{02 } , y ) \end{aligned}\ ] ] @xmath298 the first ( linear in @xmath41 ) term on the right hand side of eq . ( [ 2diff ] ) can be rewritten employing eqs .
( [ jac ] ) , ( [ trfm ] ) and ( [ invtrfm ] ) as @xmath299 using the techniques outlined in eqs . ( 32 ) and ( 33 ) of @xcite the integral in the parenthesis of eq .
( [ 1te ] ) can be done , yielding @xmath300 substituting this back into eq .
( [ 1te ] ) we end up with @xmath301 expanding @xmath302 in a taylor series we rewrite eq .
( [ 1te2 ] ) as @xmath303.\end{aligned}\ ] ] performing the integration over @xmath252 in eq .
( [ 1te3 ] ) we obtain [ see eq .
( 33 ) in @xcite ] @xmath304.\end{aligned}\ ] ] after differentiating over @xmath305 and employing the fact that @xmath306 which is a direct consequence of eq .
( [ del ] ) , we can reduce eq .
( [ 1te4 ] ) to @xmath307 with @xmath308 defined by eq .
( [ chi ] ) above . rewriting @xmath309 in eq .
( [ 1te5 ] ) as @xmath310 and employing eq .
( [ del2 ] ) to perform the summation , which makes the integration over @xmath311 trivial , we obtain the following expression for the first term on the right hand side of eq .
( [ 2diff ] ) @xmath312 that way eq.([2diff ] ) reduces to eq .
( [ bsp ] ) , as desired .
j. jalilian - marian , a. kovner , a. leonidov , and h. weigert , nucl .
* b 504 * , 415 ( 1997 ) ; phys . rev . *
d 59 * 014014 ( 1999 ) ; phys . rev . * d59 * , 034007 ( 1999 ) ; j. jalilian - marian , a. kovner , and h. weigert , phys .
rev . * d 59 * 014015 ( 1999 ) .
|
we analyze the evolution equation describing all multiple hard pomeron exchanges in a hadronic or nuclear structure functions that was proposed earlier in @xcite .
we construct a perturbation series providing us with an exact solution to the equation outside of the saturation region .
the series demonstrates how at moderately high energies the corrections to the single bfkl pomeron exchange contribution which are due to the multiple pomeron exchanges start unitarizing total deep inelastic scattering cross section .
we show that as energy increases the scattering cross section of the quark
antiquark pair of a fixed transverse separation on a hadron or nucleus given by the solution of our equation inside of the saturation region unitarizes and becomes independent of energy .
the corresponding @xmath0 structure function also unitarizes and becomes linearly proportional to @xmath1 .
we also discuss possible applications of the developed technique to diffraction .
16.5 cm -1 cm 22.4 cm cern - th/99 - 166 + nuc - mn-99/8-t + tpi - minn-99/26 * unitarization of the bfkl pomeron on a nucleus * + yuri v. kovchegov + + _ theory division , cern + ch-1211 , geneva , switzerland _ + + and + + _ school of physics and astronomy , university of minnesota , + minneapolis , mn 55455 , usa _ + + +
|
identification of two tens of the long grb ( @xmath9 ) with the extragalactic objects located at cosmological distances ( @xmath10 ) raises a question about the source of energy release . for the spherical symmetry
the energy yield is turned to be as abnormally large as @xmath11 only in the band of thousands of @xmath12 .
such energy release is comparable with that of a supernova explosion ( sn ) , for which @xmath13 of the energy is escaped by neutrinos during the first several seconds . to overcome the energy catastrophe in the problem of grb a hypothesis
was suggested about a strong collimation of the gamma radiation in the narrow solid angle of @xmath14 .
this assumption gives a possibility to diminish the luminosity of grb , needed for the explanation of the observed gamma - ray flux , by a factor of @xmath15 which for @xmath14 gives the gain of @xmath16 .
such directivity can be explained by the radiation of a narrow jet of particles accelerated to relativistic energies .
however , it is not clear now how this jet is formed when two compact objects merge or when one of them explodes .
that is why there exists a large amount of models suggested for grb explanation , @xcite . in last years the model of strong anisotropic supernova explosion becomes very popular ( see , for example , @xcite ) , which demands extreme conditions for its realization .
our model suggests the mechanism of origin of a narrow beam of gamma - rays when a shock of an sn explosion in the binary system , containing a magnetized neutron star or a white dwarf , is interacting with its magnetosphere .
if the observer is placed in the plane of the orbit of such a binary system and at the moment of the origin of the magnetospheric tail the line of observation coincides with the tail direction , then a grb will be observed .
we give the estimations confirming that our model can explain the general properties of grb , both long ( @xmath17 ) cosmological , and short ( @xmath18 ) possibly local .
for the short grb we consider the binary system in which one companion , as previously , is a neutron star or a white dwarf and another is a star of cataclysmic flare variable .
as an example we discuss a binary stellar system with a distance between the companions of @xmath19 .
the system consists of a blue giant - supernova progenitor ( @xmath20 ) and a magnetized neutron star ( @xmath21 ) or a magnetized white dwarf ( @xmath22 ) . according to the review @xcite
the parameters of the shock are kinetic energy @xmath23 , velocity @xmath24 , density @xmath25 , width @xmath26 . under the impact of the shock on the dipole magnetic field of a neutron star or a white dwarf a magnetospheric tail forms , aligned along shock wave propagation .
the magnetospheric magnetic field is amplified and compressed by the cumulative effect , forming the narrow tail .
the dependencies of the magnitude of the magnetic field in the tail @xmath27 and the tail s diameter @xmath28 on the distance to the center of the compact star @xmath29 follow from the condition of freezing of the magnetic field into the shock plasma and the conservation of the magnetic field flux @xmath30 here @xmath31 is the alfven radius @xmath32 where the magnetic field pressure is equal to the pressure of the shock matter ; @xmath33 is the value of the magnetic field on the alfven radius @xmath34 the length of the tail @xmath35 is of the order of shock s width @xmath36 , @xmath37 thus , a stretched ( @xmath38 ) , magnetized ( @xmath39 ) , almost parallel magnetospheric tail forms with a stored magnetic energy @xmath40 , @xmath41 for neutron stars and for white dwarfs the values of @xmath42 are equal to @xmath43 respectively .
we see that for a not very compact pair ( @xmath44 ) a part of the energy caught from the total kinetic energy of the shock of the sn explosion is @xmath45 and @xmath46 for the white dwarfs and the neutron stars . , that is typical of a compact sn of ib / c type .
for the ia type @xmath47 may @xmath48 times larger . ] the energy caught increases as @xmath49 for a closer binary system .
for the narrow directivity of gamma radiation @xmath50 from the relativistic particles ( @xmath3 ) , when the magnetospheric tail is directed to the observer , it is natural to obtain the effective power needed for the cosmological grb .
the rate of grb of @xmath51 for a total number of @xmath52 galaxies of a visible universe gives the rate per one galaxy of the order of @xmath53 .
if we assume the rate of sn explosions in one young galaxy as @xmath54 , then we obtain that approximately @xmath55 sn explosions result in one grb .
this corresponds to the directivity of gamma radiation of @xmath56 .
the lifetime of the tail is defined by the annihilation of its magnetic field @xmath57 ( @xmath58 is the alfven velocity ) and is equal to the time of its forming @xmath59 .
it is @xmath60 .
the non - stationarity and the reconnection of the magnetic field in the tail ( note that the magnetic field is antiparallel at the opposite sides of the tail ) lead to the appearance of a strong electric field @xmath61 accelerating particles . their mean lorentz factor @xmath62 can be estimated from the equilibrium between the acceleration rate and synchrotron losses of fast particles in the magnetic field @xmath33 , @xmath63 the characteristic value of the lorentz factor of accelerated electrons ( and positrons ) is @xmath64 . in the magnetic field
@xmath65 particles of such energies radiate at the frequency @xmath66 , which corresponds to the photon energy @xmath67 , typical of a gamma burst .
in some cases the magnetospheric tail produced by the interaction of a sn shock with a magnetosphere of a magnetized companion can be torn due to a global reconnecton of its magnetic field lines ( tearing instability ) .
such kind of phenomenon is observed during intense chromospheric flares on the sun . as a result
, part of the tail can be torn off and obtain high kinetic energy .
it will move with a lorentz factor @xmath68 , which is almost equal to the factor @xmath62 , estimated early in ( 7 ) , @xmath7 .
thus , a narrow , magnetized , relativistic jet is moving towards the observer . in the frame
moving together with the jet , let the magnetic field strength be @xmath69 , the particle density be @xmath70 , the mean lorentz factor of particles be @xmath62 .
due to synchrotron loses the value of @xmath62 will decrease . inside the jet microreconnection processes will support the equipartision between the magnetic energy and the particle energy , @xmath71 . besides
, particles are frozen into the magnetic field , and their density is proportional to the magnetic field strength , @xmath72 . the index `` 0 '' is related to the initial values of the parameters .
it follows that @xmath73 during the adiabatic expansion of the jet s diameter with the conservation of the magnetic flux , @xmath74 , the quantities @xmath70 and @xmath69 will decrease in time .
because of the optical depth with respect to the thompson scattering is small @xmath75 for our parameters ( @xmath76 ) we can consider that the synchrotron radiation escapes freely .
so , the jet s cooling is associated with the synchrotron losses @xmath77 here @xmath78 is the induced electric field arising due to the change of the magnetic field , @xmath79 . from (
8) we obtain the equation describing the evolution of the mean particle energy @xmath80 = -\frac{2e^4 } { 3m^3 c^5}\left ( \frac{8\pi mc^2n_0}{b_0}\right ) ^2 \gamma^4.\ ] ] the second term in the left - hand side of ( 9 ) describes the betatron cooling of the particles with decreasing magnetic field @xmath81 @xmath82 is the plasma frequency and @xmath83 is the cyclotron frequency in the jet at the initial time @xmath84 . in our case for @xmath85 the value of @xmath86 , and we can write the solution of equation ( 9 ) as @xmath87 where the time @xmath88 ( @xmath89 is the classic electron radius ) . for not very small times in the frame of the jet , @xmath90 , we have @xmath91 the characteristic frequency @xmath92 , at which relativistic particles radiate synchrotron photons , will change in time @xmath93 for such a dependence @xmath94 the observed frequency @xmath95 and the time in the frame of observer @xmath96 are related as @xmath97 we see that the time of glowing at a given frequency in the observer frame is @xmath98 for the photon energy @xmath99 , which is typical of a grb , the time of glowing is @xmath100 .
this time just corresponds to the maximum of the distribution of long grb over their duration ( at the level 90% from the peak luminosity - @xmath101 ) .
the observations showed that it is long grb ( @xmath102 ) that are sometimes accompanied the afterglow in x - ray and/or optic bands . according to our model
the duration of the x - ray afterglow ( @xmath103 ) is several days , and that of the optic afterglow ( @xmath104 ) is several months .
these estimations do not contradict the observations of grb . from our model
we obtain also the estimations for the rate of the change of the afterglow intensity , @xmath105 .
the observations show ( see , for example @xcite , @xcite ) the dependence @xmath106 , which does not contradict our estimation .
probably , if the energy of the shock from a sn is not enough to tear off the magnetospheric tail of magnetized component , then there will not be any jet and consequently no afterglow will be observed .
this situation is more typical of a not energetic grb and they will be short ( @xmath107 ) .
|
the sn explosion in the closed binary can give the magnetospheric flare possessing the properties of grb .
the sn shock , flowing around the magnetosphere of a magnetized neutron star or a white dwarf , produces a narrow magnetic tail @xmath0 long , @xmath1 wide and a magnetic field of @xmath2 .
fast particles ( @xmath3 ) , generated in the tail by reconnection processes , radiate gamma rays of the @xmath4 - @xmath5 energies .
the duration of radiation @xmath6 corresponds to a short grb .
apart , the powerful shock can tear and accelerate part of the tail .
that is the relativistic ( @xmath7 ) , strongly magnetized jet , producing gamma radiation and also x - ray and optic afterglow .
that is a long ( @xmath8 ) grb .
the duration of the afterglow is inversely proportional to the photon energy and is several months for optic . , gamma rays bursts , supernovae , magnetic tail ( magnetosphere ) , synchrotron radiation 98.70.r , 97.60.b , 94.30.c
|
new advances in different areas require the development of tunable , wide - band , high - power sources of coherent electromagnetic radiation in gigahertz , terahertz and higher frequency ranges .
conventional electron vacuum devices have restricted possibility of frequency tuning ( usually it does not exceed 5 - 10% ) for the certain carrier frequency at certain e - beam energy .
volume free electron laser ( vfel)@xcite was proposed as a new type of free electron laser .
frequency tuning , possibility of use of wide electron beams ( several e - beams ) and reduction of threshold current density necessary for start of generation , provided by vfel , make it a basis for development of more compact , high - power and tunable radiation sources then conventional electron vacuum devices could let .
first lasing of vfel was reported at fel 2001 @xcite .
benefits given by vfel : \1 .
volume fel provides frequency tuning by rotation of diffraction grating ; = 16 cm \2 .
use of multi - wave diffraction reduces generation threshold and the size of generation zone . starting current j
depends on interaction length l as @xcite : @xmath0 , @xmath1 is the number of surplus waves appearing due to diffraction ( for example , in case of two - wave bragg diffraction @xmath2 , for three - wave diffraction @xmath3 and so on ) .
= 15 cm \3 .
wide electron beams and diffraction gratings of large volumes can be used in vfel .
two or three - dimensional diffraction gratings allow to distribute interaction over large volume and to overcome power restrictions in resonator .
volume distributed feedback provides mode discrimination in vfel resonator .
= 4 cm \4 .
vfel can simultaneously generate radiation at several frequencies that is extremely important , for example , for development of multifrequency radiation source .
vfel operates in wavelength range @xmath44 mm at electron beam energy up to 10kev .
such generating device is very interesting for practical use in submillimeter ( terahertz ) range , which is actively employing for medical examination .
vfel resonator is formed by two parallel diffraction gratings with different periods and two smooth sidewalls ( see figure in lanl2001 ) .
the interaction of the exciting diffraction grating with an electron beam arouses smith - purcell radiation .
the resonant diffraction grating provides distributed feedback of generated radiation with electron beam by bragg dynamical diffraction .
resonant grating can rotate to change orientation of grating grooves with respect to electron beam velocity that provides tuning of diffraction conditions .
= 16 cm the length of resonator is 100 mm , periods of the exciting diffraction grating and resonant diffraction grating are 0.67 mm and 3 mm , respectively .
the ribbon electron beam with profile 10 mm x 1 mm is emitted by thermal cathode , formed in pierce gun without beam compression and guided through resonator along dffraction gratings .
only electrons passing at the distance @xmath5 @xcite near the exciting grating surface interact with electromagnetic wave effectively ( i.e. less than 10% of total cross size of electron beam participates in lasing process ) . to provide the most effective interaction of electron beam with the exciting grating , it is placed in a beam and the value of `` grating current '' ( i.e. part of total electron beam current falling to the grating ) is monitored as a touchstone of interaction efficiency and indicator of beam position with respect to grating surface .
the following parameters can be monitored during experiment : beam voltage ( electron beam energy ) ; total electron beam current ( number of emitted electrons per second ) ; grating current ; angle between resonant grating grooves and electron beam velocity ( angle of grating rotation ) ; distance between gratings ( transverse size of resonator ) ; microwave power and frequency .
electrons are emitted in pulsed regime ( unipolar pulse with sinusoidal shape and pulse duration 10 ms ) in sequence of two or three voltage pulses .
voltage can vary from 0 to 10 kv ( hence , electron beam energy is varied from 0 to 10 kev ) .
different longitudional modes of resonator are excited while electron energy passes through the whole variation interval .
frequency of data reading is determined by the condition that change of electron energy between two consecutive measurements of each monitored parameter should not cause change of generation regime . estimated admissible change of energy is about @xmath41% of rating value .
data reading interval is @xmath6 s. a lot of system parameters can be varied during experiment , so complete investigation of the system requires multiple measurements .
experimental study of frequency tuning and change of radiation intensity at mechanical rotation of resonant diffraction grating ( i.e. at variation of the angular orientation of grooves of resonant diffraction grating with respect to electron beam velocity ) is presented in this paper .
general theoretical statements @xcite should be reminded for better understanding of the below experimental results . considering processes in resonator one should discern the difference between two cases : ( a ) presence of resonator sidewalls
can be neglected ; ( b ) presence of sidewalls of resonator can not be neglected . for unbounded waveguide ( case ( a ) )
two geometries of diffraction are distinguished : those of bragg and laue @xcite .
laue geometry of diffraction implies that both incident and diffracted waves passes to vacuum through the same boundary of resonator , while at bragg diffraction incident and diffracted waves passes through the different boundaries .
absolute instability can appear in bragg geometry and such system works as generator ( for backward bragg diffraction it converts into a well - known backward - wave tube ) .
for laue case convective instability provides only amplification regime .
= 4 cm generation process in resonator is described by maxwell equations , containing the space - periodic permittivity of diffraction grating @xmath7 , where @xmath8 is the reciprocal lattice vector of the diffraction grating @xcite .
if @xmath9 in the above equations is equal to zero , they describe smooth waveguide , which eigenfunctions @xmath10 are well known . if presence of waveguide sidewalls can not be neglected , ( @xmath11 , where @xmath12 , @xmath13 are the integer numbers , @xmath14 and @xmath15 are the transversal dimensions of resonator ) .
= 5 cm expanding maxwell equations over eigenfunctions of smooth waveguide @xmath16 one obtains equations equations describing generation in continuous effective grating with periodic permittivity @xmath17 @xcite . in single - mode approximation
the above problem comes to that of generation ( amplification ) in one - dimensional periodic medium with effective permittivity @xmath18 presence of diffraction grating in waveguide makes nondiagonal matrix elements @xmath19 , describing transition between modes , different from zero : @xmath20 presence of sidewalls imparts new feature to diffraction process in resonator , because even in laue - like case ( which corresponds to laue geometry of diffraction if neglecting the sidewalls ) the matrix element @xmath21 differs from zero . and , as a result , generation regime can be reached at laue - like geometry , in contrast with an unbounded waveguide , where laue geometry provides only amplification regime , while generation is possible at bragg diffraction .
general view of microwave signal is presented on fig.[microwave]a .
presence of several microwave peaks is the manifestation of longitudinal cavity modes .
frequency difference @xmath22 associated with the change @xmath23 of electron energy @xmath24 ( @xmath25 ) can be estimated as @xmath26 . according to measurements @xmath27 , @xmath28 for two next distinguished peaks ( fig.[microwave]b . ) , hence , frequency difference for them can be estimated as @xmath29 ghz .
= 16 cm now let us consider behavior of lasing efficiency with grating rotation .
rotation of diffraction grating changes components of reciprocal lattice vector . from ( [ chi_eff ] )
it is easy to see that @xmath30 decreases with @xmath31 growth .
change of @xmath21 yields change of generation conditions and , in particular , generation threshold .
reduction of @xmath30 results in decrease of generation efficiency ( fig.[voltage ] ) .
= 4 cm but , interaction efficiency can be increased if recollect that at certain value of current density the intensity is proportional to the term @xmath32@xcite where @xmath33 is the distance from electron to the grating , @xmath34 is the lorentz - factor , @xmath35 is the radiation wavelength , @xmath36 . as a result , energy of electron beam at certain current density should be increased ( to increase factor @xmath37 ) to overcome decrease of @xmath38 .
non - one - dimensional feedback , being used in vfel , provides tuning of radiation frequency by diffraction grating rotation .
measured frequencies at rotation of resonant diffraction grating for microwave peak , corresponding electron beam energy 2,9 kev are shown in fig.[frequency ] .
measured frequency change well accords theoretical predictions @xcite = 4 cm
vfel generator can be considered as a new type of tunable backward wave tube with variable period ( due to gratings rotation ) and vfel amplifier is a new type of tunable travelling wave tube .
frequency tuning , possibility of use of wide electron beams ( several e - beams ) and reduction of threshold current density necessary for start of generation , provided by vfel , make it a basis for development of more compact , high - power and tunable radiation sources then conventional electron vacuum devices could let .
|
use of non one - dimensional distributed feedback in vfel gives possibility of frequency tuning in wide range . in present work dependence of lasing process on the angle between resonant diffraction grating grooves and direction of electron beam velocity
is discussed .
volume free electron laser ( vfel ) , volume distributed feedback ( vdfb ) , diffraction grating , smith - purcell radiation , electron beam instability 41.60.c , 41.75.f , h , 42.79.d
|
it is believed by most astro - cosmologists as well as astro - particle physicists that the only solution to resolve the flatness and horizon problems in the universe is to think of exponential expansion called inflation @xcite right after birth of our universe . in recent years
several observational results to confirm the inflation were reported @xcite .
in particular , the most important evidence in the present days to provide data is the temperature fluctuation of the cosmic microwave background radiation ( cmb ) which shows us quantitatively that there was the era of the inflation in the early universe . using this data we can investigate the mechanism of the inflationary expansion in detail .
the cmb is the photon emitted in the recombination era of which the wavelength is very long so that cmb reaches us without much scattering afterwards .
therefore cmb carries to us the information of the early universe .
the most peculiar properties of cmb is that it is homogeneous and uniform with very high accuracy and furthermore the spectrum agrees very well with the black body radiation with the temperature 2.725 [ k ] while the temperature fluctuation @xmath7 is of the order of @xmath8 .
this temperature fluctuation is believed to come from the gravitational potential @xmath9 through sachs - wolfe effect @xmath10 @xcite . in the usual inflation models
the gravitational potential @xmath9 is believed to be created when the hypothetical scalar particle(s ) called inflaton(s ) fluctuates quantum mechanically during rolling down classically along with the potential .
there are many models proposed for creation mechanism , such as slow - roll inflation @xcite , chaotic inflation @xcite etc .
in our previous papers @xcite , we show that the temperature fluctuation of cmb can be explained as the vacuum fluctuation of a large number of matter fields . here
the matter fields are originated from the string excited states and/or the kaluza - klein modes on the compactified space . there
appear very many kinds of fields in the exponentially expanding universe , where the hubble parameter plays the role of temperature .
these fields fluctuate around the bottom of the mass potential , and generate the energy density of fluctuation .
finally the fluctuation creates the gravitational potential @xmath9 through the einstein equation .
although each fluctuation due to single field is very small , but enormous number of species of fields contribute and it turns out to be summed up to a huge fluctuation compatible to the present observational value . furthermore it is possible to determine the radii of the compactified spaces , as well as the string coupling constant by comparison with the observational values of cmb .
the very merit of our theory is in the fact that we do not need to put any assumption or fine - tuning on the form of the matter fields potential , but we simply consider free fields . however in our above described theory
there still are several problems to be resolved . in the first place the spectral index @xmath11 which indicates the distance dependence of the fluctuation or , in other words , the 2-point correlation function of the fields exceeds one , i.e. @xmath12 in our model , although the observed value of the planck is @xmath13 @xcite .
as was pointed out in our foregoing papers @xcite the reason why we obtained @xmath12 is in the fact that we assumed the background metric is exactly the de sitter space - time . in order to explain the experimental value of the spectral index , we may have to introduce a time dependent hubble parameter @xmath2 , something like @xmath14 .
so far we have not discussed the origin of the inflation , and simply assumed the de sitter space - time . in this paper
we point out the possibility that the inflation is caused by the vacuum energy of the many fields in the expanding universe .
it is the purpose of the present article that we present a self - consistent theory of the time evolution of the universe by incorporating the backreaction of the quantized matter fields .
the main idea of the present paper follows our pervious articles @xcite , except that we do not assume de sitter background and incorporate the backreaction of the matter fields . in the early universe we consider @xmath15 species of the free scalar fields .
the time evolution of the early universe is assumed to follow the classical einstein equation , @xmath16 on the right hand side the energy - momentum tensor of the species @xmath17 of the matter fields is regarded as the vacuum expectation value @xmath18 , in the metric @xmath19 .
when one expresses @xmath18 by a functional of arbitrary metric @xmath19 , @xmath20 $ ] , the einstein equation ( [ 1-einstein.eq ] ) describes the time evolution of @xmath19 caused by the backreaction from the matter fields .
here we take the vacuum of the bunch - davies type in the calculation of @xmath20 $ ] . as we will see @xmath20$]is proportional to @xmath21 . therefore in the early universe , where @xmath2 is large
, we can ignore the initial state dependence of the energy - momentum tensor and simply take @xmath20 $ ] .
we mainly use the eikonal approximation to make solutions of the equation of motion of the matter fields .
then we perform the 1-loop momentum integration to calculate the correlation functions . by so doing the eq .
( [ 1-einstein.eq ] ) becomes a second order differential equation for @xmath2 .
the present paper is organized as follows . in section
[ 2 ] , we review briefly the eikonal approximation .
then we investigate the solution of matter field of mass @xmath22 in two cases , that is , for the cases of @xmath23 and @xmath24 , and evaluate @xmath25 for each cases . in section [ 3 ] we assume that there exist both light and heavy particles in the universe and search for de sitter - like expanding solutions of eq .
( [ 1-einstein.eq ] ) , where @xmath2 is almost constant
. then we examine the solutions when the hubble parameter is very small compared to that of the de sitter - like expanding era . from this analysis
we will see that the hubble parameter becomes zero and the de sitter era ends for a certain region of the initial conditions .
section [ 4 ] is devoted to conclusion and outlook . in appendix
[ app - airy ] we describe in detail the calculation by using the airy function around boundary between uv and ir regimes in the eikonal approximation . in appendix
[ app - damping ] we show the calculation of the correlation functions using the solution in the ir regime in the eikonal approximation .
it is legitimate to assume that the background metric is homogeneous and uniform as are indicated from cmb data .
we thus start with investigating the one species of the scalar field described by the action @xmath26 in the friedmann - robertson - walker space - time @xmath27 for the later convenience to apply the dimensional regularization , the space - time dimension in eq .
( [ 2-action ] ) is taken to be @xmath28 with @xmath29 . in terms of the co - moving time
@xmath30 we write the differentiation as @xmath31 and then the fourier mode @xmath32 of the rescaled scalar field @xmath33 obeys the following equation of motion : @xmath34\right)\chi_k=0 .
\label{2-eom}\end{aligned}\ ] ] here the potential @xmath35 $ ] is defined by @xmath36 & \equiv \left(1-\frac{\epsilon}{2}\right)\frac{a^{\prime \prime}}{a } -\epsilon \left(\frac{1}{2}-\frac{\epsilon}{4}\right ) \left(\frac{a^{\prime}}{a}\right)^2-m^2a^2 \nonumber \\ & = \left(1-\frac{\epsilon}{2}\right)a^2\left(2h^2+\dot{h}\right ) -\epsilon \left(\frac{1}{2}-\frac{\epsilon}{4}\right)a^2h^2-m^2a^2 , \label{2-potential}\end{aligned}\ ] ] where @xmath2 is the hubble parameter @xmath37 the quantity we are interested in is the @xmath38-component of the energy - momentum tensor as a functional of @xmath0 ( or @xmath2 ) , @xmath39 & = \frac{\mu^{4-n}}{2}\left\langle\dot{\phi}^2 + \frac{1}{a^2}\left(\nabla \phi \right)^2+m^2\phi^2\right\rangle \nonumber \\ & = \frac{\mu^{\epsilon}}{2a^{4-\epsilon}}\bigg\ { \left\langle \left(\chi^{\prime}\right)^2\right\rangle -\left(1-\frac{\epsilon}{2}\right)ah \left\langle \chi^{\prime}\chi + \chi \chi^{\prime}\right\rangle + \left(1-\frac{\epsilon}{2}\right)^2a^2h^2 \left\langle \chi^2\right\rangle \nonumber \\ & \qquad \qquad \quad + \left\langle \left(\nabla \chi \right)^2\right\rangle + m^2a^2\left\langle \chi^2\right\rangle \bigg\}. \label{2-energy - momentum.tensor}\end{aligned}\ ] ] here a mass scale @xmath40 is introduced to keep the mass dimension of the energy - momentum tensor .
we solve the equation of motion ( [ 2-eom ] ) with recourse to the eikonal approximation by putting @xmath41 then by using the boundary condition @xmath42 and the assumption of the eikonal approximation @xmath43 ( [ 2-eom ] ) has the solution @xmath44}. \label{2-eikonal.solution}\end{aligned}\ ] ] here the wronskian is used to normalize the solution : @xmath45 the general solution for @xmath46 is given by @xmath47 in the oscillatory region , @xmath48 we then quantize the field @xmath49 canonically by using the commutation relation @xmath50 = i\delta^{3-\epsilon}(\vec{x}-\vec{y } )
. \label{2-commutation.relation}\end{aligned}\ ] ] the vacuum state is defined as usual @xmath51 so that the vacuum expectation value of an operator @xmath52 is given by @xmath53 . in the following
we consider the situation where @xmath54 and @xmath55 are much smaller than @xmath2 itself , @xmath56 which means that @xmath2 and @xmath54 vary much slower than the time evolution of the universe . in this subsection
we evaluate the energy density @xmath58 of a scalar field when hubble parameter is much larger than mass ; @xmath57 .
it is well known that if @xmath2 is constant @xmath58 is given by @xmath59 note that this result is obtained without using the eikonal approximation . when @xmath60 and @xmath61 , the potential ( [ 2-potential ] ) is always positive , i.e. @xmath36>0 . \label{2-potential.is.positive}\end{aligned}\ ] ] then the eq .
( [ 2-oscillatory.region ] ) is not always satisfied , and we should analyze the following three momentum regions to continue the solution @xmath62 in the ultra - violet limit @xmath63 ( equivalently @xmath64 ) into the infrared region : 1 .
@xmath65 + in this region the solution is oscillatory as @xmath66{k^2-v(\tau ) } } e^{-i\int_{\tau_k}^{\tau } d\tau^{\prime } \sqrt{k^2-v(\tau^{\prime})}}. \label{2-oscillation.solution}\end{aligned}\ ] ] here @xmath67 is determined from @xmath68 .
2 . @xmath69 + the solution is expressed in terms of the airy function @xmath70 3 .
@xmath71 + the damping and enhancing solutions given by @xmath72{v(\tau ) -k^2}}\left\{\frac{i}{2 } e^{-\int_{\tau_k}^{\tau}d\tau^{\prime}\sqrt{v(\tau^{\prime})-k^2 } } + e^{+\int_{\tau_k}^{\tau}d\tau^{\prime}\sqrt{v(\tau^{\prime})-k^2 } } \right\}. \label{2-damping.solution}\end{aligned}\ ] ] in the above ( ii ) and ( iii ) , the boundary @xmath73 is chosen to satisfy the validity condition of the eikonal approximation ( [ 2-eikonal.condition ] ) .
in fact , @xmath74 here @xmath75 is obtained from ( [ 2-comoving.time-diff ] ) and ( [ 2-potential ] ) , @xmath76 we next evaluate the vacuum expectation value of the energy density in each region . in the region ( i )
it is given by @xmath77 \nonumber \\ & \quad -\frac{1}{64\pi^2}\left(-m^4 + 2m^2h^2 - 6h^2\dot{h}-2h\ddot{h}+\dot{h}^2 \right)\log \frac{2h^2+\dot{h}-m^2}{\mu^2}+o(\epsilon ) .
\label{2-em.tensor - oscillation}\end{aligned}\ ] ] only in this region there appears an ultra - violet divergence which should be renormalized .
indeed in ( [ 2-em.tensor-oscillation ] ) , the @xmath78 term gives the divergence in dimensional regularization .
there are five terms in the coefficient of @xmath78 .
the @xmath79 and @xmath80 terms are renormalized into the cosmological constant term @xmath81 and the einstein - hilbert action @xmath82 , respectively .
the rest three terms are @xmath83 , @xmath84 , and @xmath85 and from the dimensional analysis their counter terms are given by @xmath86 and @xmath87 . in the case of frw metric ( [ 2-frw.metric ] ) , these two actions give the following additional terms to the euler - lagrange equation : @xmath88 therefore the three coefficients of the @xmath78 term can be renormalized by these two counter terms . in principle
we may introduce finite renormalization for the three terms @xmath79 , @xmath80 and @xmath89 .
however , since we have to make the same renormalization to the two cases @xmath57 and @xmath90 of the single theory , we relegate to argue this problem after the following subsection [ 2.2 ] for the case of a small hubble parameter @xmath90 .
then we make use of the minimal subtraction , and expand ( [ 2-em.tensor-oscillation ] ) for large @xmath2 assuming @xmath91 and @xmath61 : @xmath92 for the regions ( ii ) and ( iii ) the contributions to @xmath93 are described in the appendices [ app - airy ] and [ app - damping ] respectively .
we here show the combined result of ( ii ) and ( iii ) calculated numerically under the conditions @xmath57 , @xmath60 and @xmath61 : @xmath94 note that the coefficient of @xmath84 does not remain stationary when we vary the artificial parameter @xmath73 , which was introduced to separate two regions ( ii ) and ( iii ) .
however this coefficient is very small compared to the other terms and does not affect the physical result . in this subsection
we derive the energy density in the case of a small hubble parameter , @xmath90 . in the same manner as the previous subsection [ 2.1 ]
we calculate the vacuum expectation value of the energy density .
again we assume that the change of the hubble parameter is fairly slow , @xmath95 in this case the potential term ( [ 2-potential ] ) of the equation of motion ( [ 2-eom ] ) is always negative , @xmath36<0 .
\label{2-potential.is.negative}\end{aligned}\ ] ] then the solution of the eikonal approximation ( [ 2-eikonal.solution ] ) is expressed as @xmath96\right)}.\end{aligned}\ ] ] apparently any turning point does not exist so that the solution ( [ 2-eikonal.solution ] ) is valid in the whole region of @xmath97 .
thus we obtain the energy density as @xmath98 \nonumber \\ & \quad -\frac{1}{64\pi^2}\left(-m^4 + 2m^2h^2 - 6h^2\dot{h}-2h\ddot{h}+\dot{h}^2 \right)\log \frac{m^2 - 2h^2-\dot{h}}{\mu^2}+o(\epsilon ) .
\label{2-em.tensor - damping}\end{aligned}\ ] ] we then make the renormalization by using the minimal subtraction , and the expansion assuming @xmath99 and @xmath61 gives the result , @xmath100 hereby we next make use of the finite renormalization for the cosmological constant term @xmath81 and the einstein - hilbert action @xmath82 in such a way that @xmath79 and @xmath101 terms and @xmath80 and @xmath102 terms are set to zero .
the reason for so doing is because the energy density for small values of @xmath2 should be zero .
although we should add the same counter terms to ( [ 2-em.tensor-oscillation-approx ] ) , such terms are negligible when @xmath2 is much larger than @xmath22 .
furthermore we may make a finite renormalization for @xmath86 and @xmath87 which generates an additional term @xmath103 to ( [ 2-em.tensor-damping-approx ] ) .
so far we have obtained the energy density of a single scalar field for two cases . the explicit form for the case of a large hubble parameter
is given by @xmath104 which is the sum of ( [ 2-em.tensor-oscillation-approx ] ) and ( [ 2-em.tensor-airy+damping ] ) . for the case of a small hubble parameter
we have @xmath105 we next consider the einstein equation assuming that there are many fields with various masses . then the total energy density is written as @xmath106 where @xmath107 and @xmath108 stand for the numbers of the fields whose masses are lighter and heavier than the hubble parameter , respectively .
the ratio of @xmath107 to @xmath108 may vary as the hubble parameter evolves in time .
however , the coefficients of each terms in ( [ 3-energy.density-right ] ) and ( [ 3-energy.density-heavy ] ) are roughly the same , so that we may replace @xmath93 and @xmath109 with their average
. then ( [ 3-total.energy0 ] ) becomes @xmath110 where @xmath15 is the total number of the species .
then the @xmath38-component of the einstein equation reads @xmath111 rewriting this equation after plugging ( [ 3-energy.density-right ] ) and ( [ 3-energy.density-heavy ] ) , we obtain the equation of describing the time evolution of the hubble parameter @xmath2 : @xmath112 where @xmath22 is the typical mass of the fields heavier than the hubble parameter .
a comment is in order : the mass scale @xmath40 is originally arbitrary , but as a cut - off scale in the general relativity , field theories and/or string theory , it may be natural to choose planck scale @xmath113 or string scale @xmath114 . in this subsection
we analytically solve the above eq .
( [ 3-einstein.eq ] ) by using linear approximation around the de sitter solution .
it is clear that the eq .
( [ 3-einstein.eq ] ) has the solution , where @xmath2 is a constant , @xmath115 which is nothing but the de sitter solution .
we next consider a fluctuation @xmath116 around this solution @xmath117 , @xmath118 in the linear approximation the eq .
( [ 3-einstein.eq ] ) becomes @xmath119 where @xmath120 and @xmath121 are constants given by @xmath122 the general solution reads @xmath123 let us turn to investigate the behavior of the solution ( [ 3-einstein.eq-begin-solution ] ) by using observational values .
if we assume that the matter field potential is about the forth power of the gut scale , @xmath124}$ ] @xcite , we obtain @xmath125}. \label{3-hubble}\end{aligned}\ ] ] consequently , from ( [ 3-pure.de.sitter ] ) , the number of species @xmath15 is evaluated as @xmath126 let us , for example , choose @xmath40 and @xmath22 in ( [ 3-einstein.eq ] ) as @xmath127 } , \nonumber \\ & m=10^{16}\text { [ gev ] } , \label{3-scales}\end{aligned}\ ] ] where @xmath114 is the string scale . in the case of @xmath128 ,
in which we do not make an additional finite renormalization to the minimal subtraction , the solution ( [ 3-einstein.eq-begin-solution ] ) becomes @xmath129 if @xmath130 satisfies @xmath131 the duration of the inflation becomes @xmath132 i.e. the e - folding number exceeds @xmath133 . in fig .
[ fig : energy - backreaction_inflation1](a ) we depict the time evolution of the hubble parameter with the choices @xmath134 and @xmath135 .
the present approximation @xmath60 and @xmath61 is valid in the region @xmath136 in view of ( [ 3-pure.de.sitter ] ) this is nothing but the region @xmath137 however , if we introduce an additional finite renormalization in the form of @xmath138 we can relax the condition of @xmath130 by choosing appropriate @xmath139 .
for instance if we choose @xmath140 , the solution ( [ 3-einstein.eq-begin-solution ] ) reads @xmath141 @xmath2 is depicted in fig .
[ fig : energy - backreaction_inflation1](b ) for @xmath142 and @xmath143 , which indicates that @xmath2 is close to @xmath117 in the region ( [ 3-inflation.era0 ] ) . : the inflation.,title="fig:",width=302 ] : the inflation.,title="fig:",width=302 ] ( -350,0)@xmath144 ( -120,0)@xmath144 ( -480,120)@xmath145 ( -380,100)(a ) ( -360,100)@xmath128 ( -360,80)@xmath134 ( -360,60)@xmath135 ( -150,100)(b ) ( -130,100)@xmath140 ( -130,80)@xmath142 ( -130,60)@xmath143 [ fig : energy - backreaction_inflation1 ] in the previous subsection [ 3.1 ] , we have found the inflationary solution has a sufficiently large e - folding number , and after the exponential expansion , the hubble parameter starts to decrease . in this subsection we show that in the above solution the hubble parameter becomes zero for large @xmath146 , which indicates that the inflation will end automatically . as we have seen in the previous subsection the eq .
( [ 3-einstein.eq ] ) becomes invalid sometime after @xmath2 starts to decrease .
however the eq .
( [ 3-einstein.eq ] ) becomes valid again if @xmath2 is sufficiently small .
therefore if we assume that @xmath2 continues to decrease , we can use eq .
( [ 3-einstein.eq ] ) to describe how the inflation stops . in this region @xmath147
term on the right hand side of ( [ 3-einstein.eq ] ) is much smaller than @xmath148 term on the left hand side because they balanced before @xmath2 becomes small : @xmath149 thus the eq .
( [ 3-einstein.eq ] ) reads @xmath150 where the constants @xmath151 and @xmath152 are given by @xmath153 here @xmath15 and @xmath40 are given by ( [ 3-number.of.species ] ) and ( [ 3-scales ] ) .
the solution @xmath2 of ( [ 3-einstein.eq-end2 ] ) rapidly decreases as @xmath146 increases and then , with small oscillation , @xmath154 as @xmath155 for various choices of @xmath22 with @xmath156 .
for example when we impose an appropriate initial condition so as to satisfy @xmath157 and @xmath158 and perform the numerical calculation for two cases @xmath159 and @xmath160 [ gev ] .
the results are depicted in fig .
[ fig : energy - backreaction_inflation2 ] . here fig .
[ fig : energy - backreaction_inflation2](a ) and fig .
[ fig : energy - backreaction_inflation2](b ) shows the results for @xmath128 and @xmath140 respectively . with the initial condition @xmath161 and @xmath162.,title="fig:",width=302 ] with the initial condition @xmath161 and @xmath162.,title="fig:",width=302 ] ( -340,0)@xmath144 ( -110,0)@xmath144 ( -490,120)@xmath163 ( -370,140)(a ) ( -350,140)@xmath128 ( -320,120)@xmath159 [ gev ] ( -350,123 ) ( -320,100)@xmath164 [ gev ] ( -350,103 ) ( -140,140)(b ) ( -120,140)@xmath140 ( -90,120)@xmath159 [ gev ] ( -120,123 ) ( -90,100)@xmath164 [ gev ] ( -120,103 )
in the present paper we investigate the effect of the backreaction to the expansion of the space - time due to the matter fields existing in the universe . on arbitrary friedmann - robertson - walker space - time
, we quantize , as an example of the matter field , the scalar field and then we calculate the vacuum expectation value of the energy density .
since we are not able to obtain the exact solution of the scalar field , we make use of the eikonal approximation to construct the solution .
we then plug it in the energy density of the matter field . combining it with the friedmann equation
we obtain a self - consistent equation that describes the time evolution of the space - time .
thus we can determine the behavior of the early universe including the backreaction of the matter fields . as a result
we find the following two eras in the early universe : 1 .
we obtain the inflationary solution with the e - folding number @xmath1 if we make a fine tuning of few orders of magnitude for the initial condition
. this initial condition may be determined by a scenario of the birth of the universe in string theory and/or quantum gravity .
2 . after the era ( 1 )
the hubble parameter will continue to decrease and then with small oscillation @xmath2 goes to zero asymptotically .
then finally the inflation ends . in this manner
by imposing some reasonable conditions we can make a inflation scenario without inflaton .
the problems next to be discussed are why the temperature fluctuation in the observed cmb is of the order @xmath165 and very small non - gaussianity is realized ( cf .
@xcite , @xmath166 ( 95`% ` cl ) ) . in our previous papers @xcite , by using the de sitter background without incorporating the backreaction , we estimated the cmb temperature fluctuation and non - gaussianity @xmath167 using our theory of many scalar fields , and obtained that if @xmath165 then @xmath168 . the difference between the present and previous papers ia in the fact that in the present paper we incorporated the effect of backreaction , which may not affect largely to the fluctuation , because in both cases the fluctuation is produced during the exponentially expanding era . furthermore in the previous papers @xcite we had the problem that the spectral index @xmath11 exceeded 1 and contradicted the observed value .
however as is mentioned in section [ 1 ] in the present paper the condition for @xmath169 translated into @xmath170 and @xmath171 is satisfied in the whole range of the inflation era ( see fig [ fig : energy - backreaction_inflation1 - 2 ] ) . for the case of fig .
[ fig : energy - backreaction_inflation1](a).,title="fig:",width=302 ] ( 7,37)@xmath144 ( -270,120)@xmath172 ( -80,180)@xmath128 ( -80,160)@xmath134 ( -80,140)@xmath135 furthermore we consider a new mechanism to produce the temperature fluctuation in cmb that is different from our previous papers @xcite . in @xcite , we pointed out that the fluctuation of the energy density @xmath173 in the universe creates the gravitational potential @xmath9 through the einstein equation and turns to be the temperature fluctuation of cmb .
a new mechanism we would like to discuss here is that the fluctuation @xmath174 of the time @xmath175 at which the inflation ends generates the density perturbation , as in the ordinary scenario with the inflaton , @xmath176 .
@xmath174 can be roughly estimated as follows .
first we express @xmath175 in terms of the initial values @xmath177 and @xmath178
. we define @xmath175 as the time @xmath146 when the third term of the eq .
( [ 3-einstein.eq-begin-solution1 ] ) becomes of order @xmath179 : @xmath180 on the other hand from ( [ 3-einstein.eq-begin-solution1 ] ) we obtain @xmath181 by combining ( [ 4-inflation.ends ] ) and ( [ 4-initial.condition ] ) , we obtain @xmath182 then @xmath174 is given by @xmath183 on the other hand by taking the variation of the friedmann equation @xmath184 , we have @xmath185 from ( [ 4-fluctuation.of.time ] ) and ( [ 4-variation.of.friedmann ] ) we obtain the density perturbation as @xmath186 using the eq .
( [ 3-number.of.species ] ) , @xmath187 is evaluated as @xmath188 which is consistent with the observational value .
it would be interesting to investigate this new mechanism in detail .
the authors acknowledge y. sekino for discussion in the early stage of the present work .
one of the authors h.k is supported by the jsps grant in aid for scientific research no .
m.n is also supported by the jsps grant in aid for scientific research no .
in the region @xmath69 , there is no ultra - violet divergence , so that we can take @xmath190 .
the solution of the scalar field in this region is given by the eq .
( [ 2-airy.solution ] ) , @xmath191 where the lower end of the integration of @xmath192 is in fact the boundary @xmath73 obtained in eq .
( [ 2-airy.boundary ] ) , @xmath193 in order to carry out the @xmath192-integration of the correlation functions such as @xmath194 , let us evaluate @xmath67 and @xmath195 approximately under the conditions @xmath60 and @xmath61 .
firstly , @xmath67 is determined from the equation @xmath196 imposing the condition @xmath60 , we can approximate the scale factor @xmath0 to be the pure de sitter one , @xmath197 , and obtain @xmath198 thus @xmath67 is evaluated as @xmath199 in the same manner @xmath195 becomes @xmath200 then we would like to calculate the four kinds of correlation functions in eq .
( [ 2-energy-momentum.tensor ] ) in order .
firstly , @xmath194 is calculated as @xmath201 here ( [ app - tau_k.approx ] ) and ( [ app-b.approx ] ) are used to obtain the last line . after the change of variable @xmath202 , this integration turns to be @xmath203 here we define the upper end of the integration region as @xmath204 by assuming that @xmath205 and @xmath206 are small , this upper end can be approximated as @xmath207 then we perform the integrations including the airy functions numerically and expand them in terms of @xmath205 and @xmath206 . in the same way , we calculate the correlations @xmath208 , @xmath209 and @xmath210 as follows : @xmath211 @xmath212 @xmath213
similarly to the region ( ii ) , we do not have the ultra - violet divergence in the region @xmath71 , so that we take @xmath190 . as we stated in the subsection [ 2.1 ] , from ( [ 2-airy.boundary ] )
, we have the upper end of this region @xmath214 and the solution of the scalar field is given by eq .
( [ 2-damping.solution ] ) : @xmath215{v(\tau ) -k^2}}\left\ { \frac{i}{2}e^{-\int_{\tau_k}^{\tau}d\tau^{\prime } \sqrt{v(\tau^{\prime})-k^2 } } + e^{+\int_{\tau_k}^{\tau}d\tau^{\prime } \sqrt{v(\tau^{\prime})-k^2}}\right\}. \label{app - damping.solution}\end{aligned}\ ] ] here let us calculate @xmath216 integration of the exponent of the fourier mode @xmath217 @xmath218 under the condition @xmath60 .
in this situation we can approximate the scale factor as the pure de sitter solution @xmath219 , and we obtain @xmath220 now we evaluate the four kinds of correlation functions included in the energy density ( [ 2-energy-momentum.tensor ] ) .
firstly , as for the correlation @xmath194 , applying the result ( [ app-damping.exponent ] ) , we find @xmath221 by changing the integration variable as @xmath222 , and using the lower end of @xmath223 integration , @xmath224 we rewrite @xmath194 as @xmath225 furthermore we expand @xmath226 with respect to @xmath227 and @xmath206 : @xmath228 .
\label{app - correlation - mass}\end{aligned}\ ] ] here , in each terms with @xmath229 , we approximate the lower end @xmath230 of the @xmath223 integration as @xmath231 here we have used ( [ 2-potential ] ) and ( [ 2-potential - prime ] ) to obtain the second equality . in practice we calculate the @xmath223 integrations numerically . in the same manner
we evaluate the rest correlations @xmath208 , @xmath209 and @xmath210 . @xmath232 .
\label{app - correlation - nabla}\end{aligned}\ ] ] @xmath233 \nonumber \\ & = \frac{1}{4\pi^2}\int_0^{k_0}dk\ > k^2 \bigg[\frac{1}{4}\left\{\frac{v^{\prime}}{4}(v - k^2)^{-\frac{5}{4 } } + ( v - k^2)^{\frac{1}{4}}\right\}^2 \left(\textstyle \frac{\sqrt{v}+\sqrt{v - k^2}}{\sqrt{v}-\sqrt{v - k^2 } } \right)^{-\frac{\sqrt{v}}{ah}}e^{2\frac{\sqrt{v - k^2}}{ah } } \nonumber \\ & \qquad \qquad \qquad \qquad + \left\{\frac{v^{\prime}}{4}(v - k^2)^{-\frac{5}{4 } } -(v - k^2)^{\frac{1}{4}}\right\}^2 \left(\textstyle \frac{\sqrt{v}+\sqrt{v - k^2}}{\sqrt{v}-\sqrt{v - k^2 } } \right)^{\frac{\sqrt{v}}{ah}}e^{-2\frac{\sqrt{v - k^2}}{ah}}\bigg ] \nonumber \\ & = \frac{v^2}{4\pi^2 } \int_{z_0}^1dz\ > z^2\sqrt{1-z^2}\left[\frac{1}{4 } \left\{\frac{1}{4}\frac{v^{\prime}}{v^{\frac{3}{2}}}\frac{1}{z^3}+1 \right\}^2\left(\frac{1+z}{1-z}\right)^{-\frac{\sqrt{v}}{ah } } e^{2\frac{\sqrt{v}}{ah}z } + \left\{\frac{1}{4}\frac{v^{\prime}}{v^{\frac{3}{2}}}\frac{1}{z^3}-1 \right\}^2\left(\frac{1+z}{1-z}\right)^{\frac{\sqrt{v}}{ah } } e^{-2\frac{\sqrt{v}}{ah}z}\right ] \nonumber \\ & = \frac{1}{4\pi^2}\frac{(v^{\prime})^2}{v}\times \frac{1}{16 } \int_{z_0}^1dz\ >
\frac{\sqrt{1-z^2}}{z^4}\left\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\frac{\sqrt{v}}{ah } } e^{2\frac{\sqrt{v}}{ah}z } + \left(\frac{1+z}{1-z}\right)^{\frac{\sqrt{v}}{ah } } e^{-2\frac{\sqrt{v}}{ah}z}\right\ } \nonumber \\ & \quad + \frac{1}{4\pi^2}(v^{\prime})^2\sqrt{v}\times \frac{1}{2 } \int_{z_0}^1dz\ > \frac{\sqrt{1-z^2}}{z}\left\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\frac{\sqrt{v}}{ah } } e^{2\frac{\sqrt{v}}{ah}z } -\left(\frac{1+z}{1-z}\right)^{\frac{\sqrt{v}}{ah } } e^{-2\frac{\sqrt{v}}{ah}z}\right\ } \nonumber \\ & \quad + \frac{1}{4\pi^2}v^2 \int_{z_0}^1dz\ > z^2\sqrt{1-z^2}\left\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\frac{\sqrt{v}}{ah } } e^{2\frac{\sqrt{v}}{ah}z } + \left(\frac{1+z}{1-z}\right)^{\frac{\sqrt{v}}{ah } } e^{-2\frac{\sqrt{v}}{ah}z}\right\ } \nonumber \\ & = \frac{a^4h^4}{4\pi^2 } \left(4 + 6\frac{\dot{h}}{h^2}+\frac{\ddot{h}}{h^3}-2\frac{m^2}{h^2 } \right)^2 \left(2+\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^{-1 } \nonumber \\ & \quad \times \bigg [ \frac{1}{16}\int_{z_0}^1dz\ > \frac{\sqrt{1-z^2}}{z^4 } \bigg\{\textstyle \frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } + \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \quad -\frac{\sqrt{2}}{64}\left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right ) \int_{z_0}^1dz\ >
\frac{\sqrt{1-z^2}}{z^4}\textstyle \left(\log \frac{1+z}{1-z}-2z\right)\bigg\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } -\left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \quad + \frac{1}{16\times 32 } \left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^2 \nonumber \\ & \qquad \qquad \times \int_{z_0}^1dz\ >
\frac{\sqrt{1-z^2}}{z^4}\textstyle \bigg\{\frac{1}{4}\left((\sqrt{2}-8z)\log \frac{1+z}{1-z } + 2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z+8z^2\right ) \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } \nonumber \\ & \qquad \qquad \qquad \qquad \qquad \qquad \textstyle -\left((\sqrt{2}+8z)\log \frac{1+z}{1-z } -2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z-8z^2\right ) \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \quad + \cdots \bigg ] \nonumber \\ & \quad + \frac{a^4h^4}{4\pi^2 } \left(4 + 6\frac{\dot{h}}{h^2}+\frac{\ddot{h}}{h^3}-2\frac{m^2}{h^2 } \right ) \left(2+\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^{\frac{1}{2 } } \nonumber \\ & \qquad \times \bigg [ \frac{1}{2}\int_{z_0}^1dz\ > \frac{\sqrt{1-z^2}}{z } \bigg\{\textstyle \frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } -\left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad -\frac{\sqrt{2}}{8}\left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right ) \int_{z_0}^1dz\ >
\frac{\sqrt{1-z^2}}{z}\textstyle \left(\log \frac{1+z}{1-z}-2z\right)\bigg\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } + \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad + \frac{1}{64 } \left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^2 \nonumber \\ & \qquad \qquad \quad \times \int_{z_0}^1dz\ >
\frac{\sqrt{1-z^2}}{z}\textstyle \bigg\{\frac{1}{4}\left((\sqrt{2}-8z)\log \frac{1+z}{1-z } + 2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z+8z^2\right ) \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } \nonumber \\ & \qquad \qquad \qquad \qquad \qquad \qquad \quad \textstyle + \left((\sqrt{2}+8z)\log \frac{1+z}{1-z } -2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z-8z^2\right ) \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad + \cdots \bigg ] \nonumber \\ & \quad + \frac{a^4h^4}{4\pi^2 } \left(2+\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^2 \nonumber \\ & \qquad \times \bigg [ \int_{z_0}^1dz\ > z^2\sqrt{1-z^2 } \bigg\{\textstyle \frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } + \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad -\frac{\sqrt{2}}{4}\left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right ) \int_{z_0}^1dz\ >
z^2\sqrt{1-z^2}\textstyle \left(\log \frac{1+z}{1-z}-2z\right)\bigg\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } -\left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad + \frac{1}{32 } \left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^2 \nonumber \\ & \qquad \qquad \quad \times \int_{z_0}^1dz\ >
z^2\sqrt{1-z^2}\textstyle \bigg\{\frac{1}{4}\left((\sqrt{2}-8z)\log \frac{1+z}{1-z } + 2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z+8z^2\right ) \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } \nonumber \\ & \qquad \qquad \qquad \qquad \qquad \qquad \quad \textstyle -\left((\sqrt{2}+8z)\log \frac{1+z}{1-z } -2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z-8z^2\right ) \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad + \cdots \bigg ] .
\nonumber \\ \label{app - correlation - kinetic}\end{aligned}\ ] ] @xmath234 \nonumber \\ & = -\frac{1}{2\pi^2}\int_0^{k_0}dk\ > k^2 \bigg[\frac{1}{4}\left\{\frac{v^{\prime}}{4}(v - k^2)^{-\frac{3}{2}}+1 \right\ } \left(\textstyle \frac{\sqrt{v}+\sqrt{v - k^2}}{\sqrt{v}-\sqrt{v - k^2 } } \right)^{-\frac{\sqrt{v}}{ah}}e^{2\frac{\sqrt{v - k^2}}{ah } } \nonumber \\ & \qquad \qquad \qquad \qquad + \left\{\frac{v^{\prime}}{4}(v - k^2)^{-\frac{3}{2}}-1\right\ } \left(\textstyle \frac{\sqrt{v}+\sqrt{v - k^2}}{\sqrt{v}-\sqrt{v - k^2 } } \right)^{\frac{\sqrt{v}}{ah}}e^{-2\frac{\sqrt{v - k^2}}{ah}}\bigg ] \nonumber \\ & = -\frac{v^{\frac{3}{2}}}{2\pi^2 } \int_{z_0}^1dz\ > z\sqrt{1-z^2}\left[\frac{1}{4 } \left(\frac{1}{4}\frac{v^{\prime}}{v^{\frac{3}{2}}}\frac{1}{z^3}+1 \right)\left(\frac{1+z}{1-z}\right)^{-\frac{\sqrt{v}}{ah } } e^{2\frac{\sqrt{v}}{ah}z } + \left(\frac{1}{4}\frac{v^{\prime}}{v^{\frac{3}{2}}}\frac{1}{z^3}-1 \right)\left(\frac{1+z}{1-z}\right)^{\frac{\sqrt{v}}{ah } } e^{-2\frac{\sqrt{v}}{ah}z}\right ] \nonumber \\ & = -\frac{1}{4\pi^2}v^{\prime}\times \frac{1}{2 } \int_{z_0}^1dz\ > \frac{\sqrt{1-z^2}}{z^2}\left\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\frac{\sqrt{v}}{ah } } e^{2\frac{\sqrt{v}}{ah}z } + \left(\frac{1+z}{1-z}\right)^{\frac{\sqrt{v}}{ah } } e^{-2\frac{\sqrt{v}}{ah}z}\right\ } \nonumber \\ & \quad -\frac{1}{4\pi^2}v^{\frac{3}{2}}\times 2 \int_{z_0}^1dz\ > z\sqrt{1-z^2}\left\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\frac{\sqrt{v}}{ah } } e^{2\frac{\sqrt{v}}{ah}z } -\left(\frac{1+z}{1-z}\right)^{\frac{\sqrt{v}}{ah } } e^{-2\frac{\sqrt{v}}{ah}z}\right\ } \nonumber \\ & = -\frac{a^3h^3}{4\pi^2 } \left(4 + 6\frac{\dot{h}}{h^2}+\frac{\ddot{h}}{h^3}-2\frac{m^2}{h^2 } \right ) \nonumber \\ & \quad \times \bigg [ \frac{1}{2}\int_{z_0}^1dz\ > \frac{\sqrt{1-z^2}}{z^2 } \bigg\{\textstyle \frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } + \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \quad -\frac{\sqrt{2}}{8}\left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right ) \int_{z_0}^1dz\ >
\frac{\sqrt{1-z^2}}{z^2}\textstyle \left(\log \frac{1+z}{1-z}-2z\right)\bigg\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } -\left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \quad + \frac{1}{64 } \left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^2 \nonumber \\ & \qquad \qquad \times \int_{z_0}^1dz\ >
\frac{\sqrt{1-z^2}}{z^2}\textstyle \bigg\{\frac{1}{4}\left((\sqrt{2}-8z)\log \frac{1+z}{1-z } + 2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z+8z^2\right ) \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } \nonumber \\ & \qquad \qquad \qquad \qquad \qquad \qquad \textstyle -\left((\sqrt{2}+8z)\log \frac{1+z}{1-z } -2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z-8z^2\right ) \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \quad + \cdots \bigg ] \nonumber \\ & \quad -\frac{a^3h^3}{4\pi^2 } \left(2+\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^{\frac{3}{2 } } \nonumber \\ & \qquad \times \bigg [ 2\int_{z_0}^1dz\ > z\sqrt{1-z^2 } \bigg\{\textstyle \frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } -\left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad -\frac{\sqrt{2}}{2}\left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right ) \int_{z_0}^1dz\ >
z\sqrt{1-z^2}\textstyle \left(\log \frac{1+z}{1-z}-2z\right)\bigg\{\frac{1}{4 } \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } + \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad + \frac{1}{16 } \left(\frac{\dot{h}}{h^2}-\frac{m^2}{h^2}\right)^2 \nonumber \\ & \qquad \qquad \quad \times \int_{z_0}^1dz\ >
z\sqrt{1-z^2}\textstyle \bigg\{\frac{1}{4}\left((\sqrt{2}-8z)\log \frac{1+z}{1-z } + 2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z+8z^2\right ) \left(\frac{1+z}{1-z}\right)^{-\sqrt{2}}e^{2\sqrt{2}z } \nonumber \\ & \qquad \qquad \qquad \qquad \qquad \qquad \quad \textstyle + \left((\sqrt{2}+8z)\log \frac{1+z}{1-z } -2\left(\log \frac{1+z}{1-z}\right)^2 - 2\sqrt{2}z-8z^2\right ) \left(\frac{1+z}{1-z}\right)^{\sqrt{2}}e^{-2\sqrt{2}z}\bigg\ } \nonumber \\ & \qquad \qquad + \cdots \bigg ] .
\label{app - correlation - cross}\end{aligned}\ ] ]
|
it is shown that the inflationary era in early universe is realized due to the effect of backreaction of quantized matter fields .
in fact we start by quantizing a free scalar field in the friedmann - robertson - walker space - time , and the field is fluctuating quantum mechanically around the bottom of the mass potential .
we then obtain the vacuum expectation value of the energy density of the scalar field as a functional of the scale factor @xmath0 of the universe . by plugging it into the einstein equation
, a self - consistent equation is established in which the matter fields determine the time evolution of the universe .
we solve this equation by setting few conditions and find the following solution : the universe expands la de sitter with e - folding number @xmath1 and then it turns to shrink with a decreasing hubble parameter @xmath2 which rapidly goes to zero . * a new mechanism of realizing inflationary universe + * with recourse to backreaction of quantized free fields + inflation without inflaton + * * yoshinobu habara@xmath3 , hikaru kawai@xmath4 and masao ninomiya@xmath5 @xmath5okayama institute for quantum physics , 1 - 9 - 1 kyoyama , okayama 700 - 0015 , japan @xmath6department of physics , kyoto university , kyoto 606 - 8502 , japan [email protected] , + [email protected] , + [email protected] pacs numbers : 04.62.+v , 98.80.cq , 11.25.-w , 11.25.mj , 98.80.qc keywords : inflation , string theory , cosmic microwave background
|
40004000 = 1000 = ifundefinedreset@font@font footnotetext#1 tempcntc citex[#1]#2@fileswauxout tempcnta@tempcntb@neciteaciteforciteb:=#2citeo#1 citeotempcnta > tempcntbcitea citea , m tempcnta = tempcntbtempcnta # 1 o p # 1#1 # 1 # 1 # 1 # 1 # 1#2 # 1#2_#1f_#2 # 1#1 # 1#1 _ # 1@xmath0 # 1 _ 0
it is now generally believed that the type iia string compactified down to six dimensions on a k3 surface gives the same physics as a heterotic string compactified on a four - torus ( see , for example , @xcite ) .
it was suggested in @xcite that much the same effect could be seen in four dimensions for @xmath1 theories .
that is , a type ii string compactified on a threefold could be equivalent to a heterotic string compactified on an object , essentially a k3 surface times a two - torus , for a suitable manifold .
let us suppose that we have a type iia string compactified on a manifold , @xmath2 .
it was realized in @xcite that some of the evidence supporting the conjectured dual pairs of @xcite appeared to originate in the fact that @xmath2 could be written in the form of a k3 fibration .
that is @xmath2 can be written as a bundle where the generic fibre is a k3 surface and the base space is @xmath3 . over a finite number of points on the base @xmath3 the fibre degenerates and will not be a k3 surface .
it was realized in @xcite that this fibration structure fits in nicely with the six - dimensional duality above .
a general result concerning k3 fibrations was derived in @xcite .
that is , if there is a heterotic string dual to the type iia string compactified on @xmath2 , such that the weakly - coupled heterotic string can be understood in terms of @xmath2 near its large - radius limit , then @xmath2 _ must _ be a k3 fibration .
in particular , the divisor class of the k3 fibre ( or equivalently , the size of the base @xmath3 ) on the type iia side maps to the dilaton on the heterotic side . it is natural to ask whether the converse to this statement is true . that is ,
if @xmath2 is a k3 fibration then must there exist a heterotic dual ? certainly there is a field in the theory which has every right to be called a dilaton for the heterotic string .
this question is analogous to one familiar from the relationship between conformal field theories and non - linear s with a target space .
when does a conformal field theory have a target space interpretation ?
a necessary condition for this is that the moduli space should have a large - radius limit point , that is a point of _ maximal unipotency _ @xcite .
it seems reasonable to say that this is also a sufficient condition certainly no counter - example is known .
this condition of maximal unipotency appears to be on the same footing as the k3 fibration condition above and so again it seems very reasonable to assert that it is again a sufficient condition .
this raises an immediate question .
suppose @xmath2 is such that it admits more than one k3 fibration .
that is , i can fibre @xmath2 in two ( or more ) ways such that the resulting k3 fibres in each case are not homologous .
we now have two ( or more ) heterotic strings , each of which can equally be called the dual of the type iia theory .
since the k3 fibres are not homologous , the dilatons for the heterotic strings must correspond to different directions in the moduli space of vector moduli . thus the relationship between the heterotic strings will not be manifest from the perturbative analysis of the heterotic string using conformal field theory .
thus , manifolds with multiple k3 fibrations give examples of nontrivial heterotic dual pairs in four dimensions .
an example of dual heterotic strings in six dimensions was given recently by duff , minasian and witten @xcite .
each member of this pair gives an @xmath4 theory in six dimensions and so can be compactified on a two - torus to give @xmath1 in four dimensions .
the resulting theory is precisely one discussed in @xcite .
thus , if the conjectures of @xcite are to be correct then the @xmath2 given as the dual partner _ must _ admit at least two fibrations .
we will see that this is correct , aside from some minor subtleties , in section [ s : eg1 ] .
we will also see this double fibration structure essentially requires @xmath2 to be unique .
an important part of the analysis of @xcite was the necessity of the possible appearance of gauge groups which could not be understood perturbatively from the heterotic string .
such groups were discussed in the context of type iia strings in @xcite .
we will see in section [ s : enhg ] that the dual fibration picture discussed in section [ s : eg1 ] fits in perfectly with this picture .
we will also see that the type iia string predicts more than can be seen from simple analysis of either of the dual heterotic strings .
there is actually no limit to how many k3 fibrations some manifolds may admit . in section [ s : inf ] we discuss an extreme case of an infinite number . while preparing this paper we became aware of @xcite which has some overlap with this work
in this section we will discuss the example of @xcite in the context of a double fibration . whereas a type ii compactification is specified by the space on which it is compactified , a heterotic string requires , in addition , a vector bundle structure in which the gauge fields of the heterotic string live .
the heterotic strings in six dimensions studied in @xcite were compactified on a k3 surface with the following vector bundle .
let @xmath5 be a stable @xmath6-bundle on a k3 with @xmath7 .
embed this instanton in an @xmath8-bundle .
the @xmath9 bundle required is simply the sum of two such bundles .
since the structure group of this bundle commutes with an @xmath10 subgroup , we expect to see an @xmath11 gauge group in the resultant six - dimensional physics .
this bundle may be deformed however , that is , there are massless scalars in the six - dimensional theory .
giving expectation values to these scalars will deform the bundle and `` higgs '' part of the gauge group away . a generic point in the moduli space will have no gauge group whatsoever .
the analysis of @xcite indicated that for any such heterotic string one can find a dual partner .
the dual partner is physically the same as the original but the field identified as the dilaton is different a strongly coupled string maps to a weakly coupled partner and the other moduli determining the shape of the k3 surface and the moduli of the bundle undergo a nontrivial transformation . phrased differently , the complete moduli space of these heterotic strings , including the dilaton , admits a nontrivial @xmath12 isometry which should be modded out by in order to obtain the true moduli space of inequivalent theories in six dimensions .
now let us compactify this heterotic string on @xmath13 .
the resulting theory then has a @xmath14 gauge group two @xmath15 s from the metric from the isometries of the torus and two from reduction of the @xmath16-field .
this is one of the examples studied in @xcite .
one of the @xmath15 s comes from the @xmath1 supergravity multiplet in four dimensions and the other three come from vector multiplets .
the vector bundle @xmath5 described above lives in a moduli space with 448 real dimensions .
thus , together with the 80 real dimensions of the moduli space of strings on a k3 surface , our original theory in six dimensions ( without the dilaton ) lived in a moduli space of @xmath17 dimensions .
this gives the moduli space coming from the hypermultiplets in the four - dimensional theory .
this space has a quaternionic structure there are @xmath18 hypermultiplets . for this four dimensional theory to be dual to a type iia
string compactified on a manifold @xmath2 we therefore require @xmath19 and @xmath20 ( the dilaton of the type ii string gives one hypermultiplet ) .
in @xcite it was conjectured that @xmath2 is given by the blow - up of a hypersurface in a weighted projective space namely the hypersurface given by @xmath21 in the space @xmath22 . when this is blown - up , it is a k3 fibration as discussed in @xcite .
this then satisfies the condition from @xcite that it be dual to a heterotic string . from the result of @xcite
however , this is not enough we need it to be dual to two heterotic strings .
thus there must be two k3 fibrations of @xmath2 . at first sight
we are in trouble the blow - up of the hypersurface defined by ( [ eq : d24 ] ) does _ not _ admit two fibrations .
all is not lost however .
the manifold @xmath2 has 243 deformations of complex structure .
the equation ( [ eq : d24 ] ) only has 242 deformations .
therefore there is one deformation of @xmath2 which takes it outside the class of those embedded in @xmath22 . when we deform outside this class , the fibration structure of @xmath2 changes . onto the first three coordinates induces an elliptic fibration of @xmath2 onto @xmath23 .
the space @xmath23 is singular and should be blown - up to give the `` hirzebruch surface '' @xmath24 .
this elliptic fibration is easily seen to have a section , and hence is birationally equivalent to a weierstrass model @xmath25 where @xmath26 and @xmath27 .
( see @xcite and also @xcite . )
now we can deform @xmath24 to @xmath28 , and deform the above weierstrass model to @xmath25 where @xmath29 and @xmath30 .
this yields the additional deformation of complex structure not visible in @xmath22 . ]
the result is that , at a generic complex structure , @xmath2 is not only a k3 fibration but can also be written as an elliptic fibration with base space @xmath28 .
fix a point in the first @xmath3 factor .
then the fibration over the second @xmath3 defines a k3 surface as an elliptic fibration .
thus the first @xmath3 can be seen as the base for writing the whole threefold as a k3 fibration .
clearly we could have chosen instead to fix a point first in the second @xmath3 factor so @xmath2 admits two k3 fibrations .
we believe that the fact that there is only one k3 fibration for special values of the complex structure is probably not very significant from the point of view of string theory .
it appears to be tied up with the phenomenon observed by wilson @xcite that the khler cone for a manifold can `` jump '' as one moves about the moduli space of complex structures as rational curves suddenly jump into existence .
one may worry that this may have adverse effects on rational curve counts along the lines of @xcite but it turns out that any such rational curve will give a zero contribution when the analysis of @xcite is performed .
similarly we expect the dual heterotic string to be unaffected .
let us clarify the interpretation of our construction of the fibration in terms of the heterotic string .
let us again fix a generic point in the first @xmath3 and consider building a k3 surface as an elliptic fibration over the second @xmath3 .
this elliptic fibration has one section .
this section thus defines a rational curve embedded inside this k3 surface .
the size of this rational curve is given by the size of second @xmath3 .
now we fibre the threefold as a k3 fibration over the first @xmath3 .
the result is shown in figure [ fig : fib1 ] .
thus the size of one of the @xmath3 s in the base @xmath28 gives the size of the rational curve in the generic fibre and the other gives the size of the base of the k3 fibration . as explained in @xcite the size of the base gives the value of the dilaton for the heterotic string .
what is the size of the other @xmath3 in terms of the heterotic string ?
let @xmath31 and @xmath32 be our two heterotic strings . in six dimensions let their string couplings ( exponential of the dilaton ) be given by @xmath33 and @xmath34 . in @xcite
it was shown that @xmath35 now let us compactify further down to four dimensions to give a dual pair of heterotic strings with coupling constants @xmath36 and @xmath37 . to do this we compactify on a two - torus .
the four - dimensional coupling depends on the six - dimensional coupling and the area of this two torus .
this area differs according to which heterotic string measures it @xcite .
let us denote the area of the torus @xmath38 or @xmath39 depending on whether it is measured by @xmath31 or @xmath32 .
we then have @xmath40 thus the size of the other @xmath3 in the type iia picture is immediately apparent it is the size of the torus in the heterotic string .
note that the act of exchanging the fibrations with @xmath2 generically leads to a change in the complex structure of @xmath2 .
that is , there is a @xmath12 symmetry in the moduli space of complex structures _ together _ with khler forms but not on either alone .
translated into the heterotic string language this tells us that as we map from one heterotic string to its dual partner , there is a non - trivial map on the moduli space of hypermultiplets .
this is exactly what was found in @xcite .
for some values of the moduli of the hypermultiplets the exchange of the base @xmath3 s will be a symmetry of @xmath2 .
thus we have arrived at a manifest geometric realization of the duality between the two heterotic strings in terms of a symmetry of the space on which the equivalent type iia string is compactified .
note that some evidence of the symmetry in the context of type iib strings has previously been noted in @xcite .
it is interesting to note that we can now go some way to prove the conjecture made in @xcite that a type iia string compactified on @xmath2 is indeed dual to the heterotic string above .
this is because one may show that @xmath2 is the _ only _
manifold which is an elliptic fibration over @xmath28 which has the right hodge numbers . to
more precise , suppose the following statements hold . 1 .
the heterotic - heterotic duality in six dimensions proposed in @xcite is correct .
this heterotic string compactified on a torus down to four dimensions is dual to a type iia string compactified on some manifold @xmath41 .
3 . the weakly - coupled heterotic string can be understood in terms of @xmath41 near its large radius limit .
we can then state that _
@xmath41 is isomorphic to @xmath2_. finally let us prove this uniqueness .
this is a little technical and the reader may wish to skip to the next section .
suppose @xmath42 is any calabi - yau manifold with two k3 fibrations and @xmath43 .
then we claim that @xmath42 is of the same type as the @xmath2 constructed above . indeed , the two distinct maps @xmath44 induces an elliptic fibration @xmath45 .
if @xmath46 does not have a section , then there is an elliptic threefold @xmath47 , the jacobian of @xmath42 , which is fiber - wise isomorphic to @xmath46 except for possibly some isolated fibers , and such that @xmath48 has a section .
see @xcite for details .
then @xmath48 is birationally equivalent to a weierstrass model over @xmath28 as in the footnote above .
since @xmath49 and each @xmath50 can arise from only a finite number of different @xmath51 in the same deformation class of @xmath42 ( as follows from the theory of @xcite ) , we see that we obtain a dominant map from the moduli space of @xmath42 to the moduli space of @xmath2 .
however , again using the theory of @xcite , for general @xmath2 , there is no elliptic fibration @xmath52 without a section whose jacobian is @xmath2 .
thus @xmath53 must already have a section , so @xmath42 is deformation equivalent to @xmath2 .
it was discussed in @xcite that going between the dual heterotic strings exchanges the rles of the perturbative gauge group and the nonperturbative gauge group .
we will now see that this is also in accord with the double fibration structure .
the construction of the heterotic string in question began with a theory with gauge group @xmath11 .
thus was then removed by picking a generic value for the bundle moduli .
thus we know that the @xmath14 gauge group in our four - dimensional theory can be enhanced to at least @xmath54 by tuning the hypermultiplet moduli suitably .
when we are at such a point of enhanced symmetry , some vector moduli become massless and we can move into a new branch of the moduli space roughly along the lines of @xcite .
the new theory at a generic point in its moduli space will have rank@xmath55 more vector multiplets than the original and fewer hypermultiplets since we had to specialize a bundle with structure group @xmath9 to one of @xmath56 .
an @xmath6 bundle with @xmath57 has a moduli space of 84 real dimensions and so the number of hypermultiplets lost is @xmath58 .
thus if this new heterotic string is dual to a type iia string on a manifold @xmath59 , we have @xmath60 and @xmath61 . the appearance of enhanced gauge symmetries in the context of type ii string compactifications has been discussed in @xcite . as seen from this new heterotic string
, the @xmath11 gauge group appears by varying the vector moduli . for our purposes it will be the analysis of @xcite which will be of most use .
there it was claimed that the enhanced gauge groups which could be seen perturbatively by the heterotic string as the vector moduli varied must originate from singularities in the k3 fibre .
thus , the @xmath11 gauge group appears when the generic k3 fibre has two @xmath62 singularities in it @xcite .
so long as this extremal transition between @xmath2 and @xmath59 preserves the k3 fibration structure then we can see that the @xmath11 gauge group appears in the original heterotic string when we deform the _ complex structure _ of @xmath2 so as to acquire two @xmath62 singularities in the generic fibre .
we can do this explicitly as follows .
let us consider @xmath2 as an elliptic fibration over @xmath28 .
when viewed as a k3 fibration , the generic fibre will itself be an elliptic fibration .
let us denote the homogeneous coordinates of the base @xmath28 by @xmath63 $ ] and @xmath64 $ ] .
a fibre can then be written in weierstrass form as @xmath65 in order for the resulting threefold to be a manifold we require that @xmath66 is of bidegree @xmath67 in @xmath68 and @xmath69 and that @xmath70 be of bidegree @xmath71 @xcite .
as we move over the base @xmath28 , the discriminant of the elliptic , @xmath72 , will have zeros leading to singular fibres .
if the polynomials @xmath66 and @xmath70 are generic , the singular fibres will not be too bad and the resulting manifold will be smooth . for special choices of the polynomials however , the singular fibre will be sufficiently bad as to introduce singularities into the threefold .
this has been discussed in the context of string theory and @xmath73-branes in @xcite .
let us fix a generic point on one of the @xmath3 s ( i.e. , specify @xmath74 and @xmath75 ) in the base .
the elliptic fibration given in ( [ eq : efib ] ) will then give the generic k3 fibre as an elliptic fibration .
we now want to deform this as to get two @xmath62 singularities in the k3 surface . following the work of @xcite we can do this by choosing the following polynomials for @xmath66 and @xmath70 :
@xmath76 where @xmath46 is of bidegree @xmath77 and @xmath78 is of bidegree @xmath79 .
we now claim that the generic k3 fibre has an @xmath62 singularity at @xmath80 and another at @xmath81 .
we can now blow each k3 fibre up to resolve this singularity .
this process of deforming the complex structure to acquire a singularity and then blowing it up is shown for a generic k3 fibre in figure [ fig : k3trans ] .
we show complex dimensions as real . in the each case
the k3 surface is an elliptic fibration over the base @xmath3 .
the blowing up procedure introduces two sets of rational curves in the shape of the dynkin diagram for @xmath62 . in terms of the generic k3 fibre
, this process has simply changed one k3 surface into another . the effect globally on the threefold is not so trivial however .
we need to worry that the degenerate fibres of the k3 fibration may be sufficiently bad as to introduce singularities into the threefold .
it may be shown that this is not the case here .
our new smooth manifold obtained by this extremal transition is , of course , the obvious candidate for @xmath59 .
indeed one can show that it satisfies @xmath82 and @xmath83 .
we claim then that we have constructed the manifold @xmath59 which is dual to the heterotic string obtained from the @xmath10 heterotic string deformed by vector moduli to obtain a string with gauge group @xmath84 .
thus we have another example to add to those in @xcite of the geometric interpretation of a phase transition in the heterotic string in terms of the type iia string .
now let us ask what happens when we exchange the @xmath3 s in the base of the elliptic fibration as we try to go through this extremal transition .
the base of the k3 fibration is now what was the base of generic k3 fibre when it was written as an elliptic fibration .
thus , we have our @xmath62 trees sticking out over a finite number of points on this base as shown in the bottom of figure [ fig : k3trans ] .
that is , this contribution to the counting of the vector multiplets comes from having reducible bad fibres in the k3 fibration .
this is exactly the situation where we expect to have nonperturbative contributions to the gauge group of the heterotic string as explained in @xcite .
thus we see that the @xmath10 that was perturbative in the original heterotic string becomes nonperturbative in the dual heterotic string .
what s more , one can also see that the generic fibre of the second fibration will not contain any singularities and so the perturbative gauge group consists only of the @xmath14 which is always present .
it is easy to extend this analysis to the simply - laced gauge groups which are subgroups within @xmath11 and in each case we see the structure conjectured in @xcite appearing .
that is , the parts of the gauge group which are perceived as perturbative and nonperturbative are exchanged ( except for the @xmath14 ) between the dual heterotic strings .
up to this point we have managed to reproduce the effects seen in @xcite in terms of the geometry of the type iia string .
it turns out that we can say more however .
the same method as was used to introduce two @xmath62 singularities into the generic k3 fibre can actually be used to introduce two @xmath8 singularities . thus we appear to be claiming that the heterotic string can be deformed to a theory with a gauge group containing @xmath85 .
this is troublesome to interpret in terms of the usual picture since the centre of @xmath9 is trivial we need to deform our @xmath85 bundle so that it becomes trivial and yet maintain @xmath86 .
clearly this is impossible .
what we can do however in addition is to bend the base k3 around and possibly introduce singularities into the base , or the bundle ( perhaps reinterpreting it as a sheaf ) .
( one may also try to analyze this in terms of open strings along the lines of @xcite . )
such processes can introduce further gauge groups .
one should also suspect that this may introduce nonperturbative gauge groups however following witten s work @xcite .
we will now see that this is indeed the case .
something new happens when we degenerate to an @xmath85 singularity compared to this cases of @xmath11 and its subgroups .
that is , when we blow - up to resolve the generic fibre it turns out that some singular fibres are still sufficiently badly - behaved as to make the resulting threefold singular . if we want to obtain a smooth threefold we require further blow - ups within these bad fibres
. thus we necessarily have contributions to @xmath87 of the threefold from bad fibres and thus , following @xcite , we have a nonperturbative part to the gauge group . in the case at hand
one can show that each of six bad fibres for each @xmath8 need to be blown up once to smooth the threefold .
this suggests that we have a nonperturbative gauge group of rank 12 when we try to enhance the perturbative part of the gauge group to include @xmath85 .
it is nice to see that a nonperturbative contribution to the gauge group arises when one tries to enhance to @xmath85 but we have no definite geometric interpretation for this on the heterotic side yet in terms of singular bundles and/or singular k3 surfaces .
we should say that finding the heterotic string description of this model is necessarily going to be difficult since there is always some nonperturbative contribution to the gauge group for any point in its moduli space .
we have discussed above an example of a manifold which admits two k3 fibrations .
actually there is no limit to the number of k3 fibrations one may obtain and in this section we discuss an example with an infinite number . in a way this will turn out to be more trivial than the previous case .
when one has more than one k3 fibration , the relationship between the different fibrations can be of one of three types : 1 .
the fibrations have quite different generic fibres .
that is , the picard lattice for the generic fibres differ
the generic fibres have the same picard lattice but , for a generic complex structure , the act of exchanging the fibrations has a nontrivial action on the complex structure .
3 . the fibrations are completely diffeomorphic .
an example of the first case is @xmath59 from section [ s : enhg ] . after going through the @xmath11 transition ,
the picard lattice the generic k3 fibre will now contain the root lattice of @xmath11 whereas the generic fibre of the other fibre will not . in terms of the heterotic string case 1
corresponds to the perturbative part of the gauge group for one heterotic string being different to the perturbative part for the other . in the case of @xmath59 ,
one string has a large perturbative gauge group of rank 18 , whereas its dual partner only sees @xmath14 .
an example of the second case is @xmath2 .
both heterotic strings have a perturbative gauge group of @xmath14 but , as discussed in @xcite , the map between them is nontrivial on the hypermultiplet moduli space .
we will now discuss an example of the third type following @xcite .
consider the manifold , @xmath42 , obtained by blowing up the orbifold @xmath88 .
this orbifold has been discussed many times in the string literature , for example @xcite .
the first @xmath12 quotient may be thought of as building a k3 surface as @xmath89 .
the second quotient then acts on this k3 surface and the other @xmath13 to build a space .
let us analyze this intermediate k3 surface and in particular its picard lattice , @xmath90 ( see , for example @xcite for a discussion of the picard lattice in this context ) .
the two @xmath13 s used to build this k3 surface each contribute a rational curve class to the k3 surface as well as the 16 exceptional divisors coming from blowing up the orbifold .
thus , for a generic model , @xmath90 is a rank 18 lattice of signature @xmath91 .
note that this lattice is _ not _ self - dual .
now , thanks to the work of @xcite , it turns out that any primitive element of length 0 in @xmath90 defines the class of an elliptic curve in the k3 surface and that there is an elliptic fibration of the k3 surface whose generic fibre is this class .
thus , since there are an infinite number of such vectors , there are an infinite number of elliptic fibrations for such a k3 surface .
now divide out by the second @xmath12 .
there are two possibilities for how this @xmath12 acts on our elliptic fibration of the k3 surface .
it may either fix the base @xmath3 point - wise or it may not . in the case that the base @xmath3 is fixed , one may show that this may be used as the base of @xmath42 as a k3 fibration @xcite . loosely speaking
, this happens for `` half '' of the primitive elements of length zero in @xmath90 .
thus @xmath42 has an infinite number of k3 fibrations .
if our assertion in the introduction concerning the existence of a heterotic string for every k3 fibration is correct then we have an infinite number of heterotic strings any two of which are dual to each other !
actually this is not as bad as it sounds .
one can see that a subgroup ( of finite index ) of @xmath92 must act as a symmetry group on this manifold since it acted as such on the k3 surface above .
this can be used to identify all the different k3 fibrations we have generated .
thus all the k3 fibrations are diffeomorphic .
one should note that this does not mean that this example is trivial however .
one should think of this subgroup of @xmath92 as acting as a group of @xmath93-dualities on the heterotic string in the sense of @xcite .
the group acts nontrivially on the dilaton and so identifies heterotic strings that do not look isomorphic from their conformal field theories .
one should also note that it mixes contributions to @xmath87 from generic fibres and from bad fibres and so mixes perturbative gauge groups with nonperturbative gauge groups as in the earlier example .
the difference between this example and that in section [ s : eg1 ] is that this time the duality maps acts trivially on the space of hypermultiplets . in @xcite
a conjecture was made about manifolds with such infinite symmetries .
if it is correct then it will always be the case that a manifold with an infinite number of fibrations will lead to heterotic strings which are mainly equivalent in this manner .
dividing out by this group of dualities one can be left with only a finite number of heterotic strings which are equivalent in the sense of type 1 or type 2 above .
we would like to thank m. duff , s. kachru , v. kaplunovsky and d. morrison for useful conversations .
the work of the authors is supported by grants from the national science foundation .
|
a type iia string compactified on a calabi
yau manifold which admits a k3 fibration is believed to be equivalent to a heterotic string in four dimensions .
we study cases where a calabi yau manifold can have more than one such fibration leading to equivalences between perturbatively inequivalent heterotic strings .
this allows an analysis of an example in six dimensions due to duff , minasian and witten and enables us to go some way to prove a conjecture by kachru and vafa .
the interplay between gauge groups which arise perturbatively and nonperturbatively is seen clearly in this example . as an extreme case
we discuss a calabi
yau manifold which admits an infinite number of k3 fibrations leading to infinite set of equivalent heterotic strings .
epsf # 1date_#1 _ # 1thefnmarkfootnotetext1991 _ mathematics subject classification .
_ # 1 # 1thefnmarkfootnotetext _ key words and phrases .
_ # 1 ps . @firstpageps .
@empty oddhead evenheadoddhead # 1
|
after the discovery of higgs boson by the atlas and cms collaborations at the lhc @xcite , properties of higgs boson have been studied extensively .
higgs boson couplings to the standard model ( sm ) particles are important because these couplings may give some hints for physics beyond the sm .
tree - level neutral bosons coupling for @xmath0 vanish within the electroweak interactions . therefore , any detected signals of these tree - level couplings would indicate the existence of new physics . in a common scenario beyond the sm
, the corrections from higher dimensional operators can be calculated through an effective lagrangian .
being consistent with lorentz and gauge invariance , the general structure for @xmath0 vertex can be written as @xcite , @xmath10,\ ] ] where @xmath11 and @xmath12 represent momentum of photon and @xmath13 boson , respectively .
@xmath14 is the mass of @xmath13 boson ; @xmath15 ; @xmath1 , @xmath2 and @xmath16 are constants which can be written in form factors . here , @xmath1 vanishes according to the electromagnetic gauge invariance for @xmath17 . in
the photon - induced processes a quasi - real photon emitted from one of the proton beam @xcite can be described in the framework of equivalent photon approximation ( epa ) . for a process in a photon
induced @xmath4 collision can be different from pure deep inelastic scattering process by means of two experimental signatures emerge in the following way @xcite : first , forward detectors can detect the particles with large pseudorapidity .
if the proton emits a photon , it scatters with a large pseudorapidity and can not be detected from the central detectors .
however , potential forward detectors located at 220 m and 420 m away from the interaction point can detect the particles with large pseudorapidity providing some information on the scattered proton energy .
second , if the photon emitting intact protons exits the central detector without being detected , the energy deposit in the forward region decreases compared to the case in which the proton remnant is detected by the colorimeters . accordingly
, one of the forward regions of the central detector has a large energy deficiency .
the region devoid of particles defines forward rapidity gaps .
@xmath18 backgrounds from deep inelastic processes can be seperated by applying a selection cut on this quantity .
photon induced processes at the lhc have been studied as a probe of new physics beyond the sm @xcite .
extensive studies on determining the sensitivity to the @xmath0 vertex have been performed at @xmath19 @xcite , @xmath20 @xcite and @xmath18 @xcite collisions . in this work ,
we have studied anomalous @xmath0 vertex to obtain the limits on @xmath1 , @xmath2 and @xmath3 couplings in a model independent way through a @xmath4 collision via the process @xmath21 at the lhc . for all calculations ,
we use the computer package calchep @xcite by implementing the anomalous interaction vertices given in eq .
[ couplings ] .
we consider the acceptance regions of the atlas and cms forward detectors to tag the protons with some energy loss fraction @xmath22 .
atlas and cms experiments will have forward detectors with the acceptance of @xmath23 @xcite within the pseudorapidity ranges @xmath24 .
when the forward detectors are installed closer to the interaction points , higher @xmath25 value is obtained .
therefore , we define the parameter sets corresponding to these acceptances and pseudorapidity ranges .
the parameters can also be related to the transverse momentum cuts on the tag protons as , @xmath26 where @xmath27 and @xmath28 are the energy and mass of proton , respectively .
we define the parameter sets , namely pi , pii and piii within the pseudorapidity range @xmath24 for minimal transverse momentum cuts on the scattered protons with @xmath29 gev , @xmath30 gev and @xmath31 gev , respectively .
the decay width of @xmath32 with anomalous couplings can be calculated considering the effective vertex in eq . [ couplings ] as , @xmath33 \label{eq : decay}\end{aligned}\ ] ] where @xmath34 and @xmath35
, @xmath36 denotes mass of the higgs boson .
there is also contribution from one loop diagram including @xmath0 anomalous couplings .
however , this contribution is proportional to the power of six of the anomalous couplings . taking into account the experimental bounds on these couplings , which are smaller than @xmath37
, tree level contribution dominates . depending on the anomalous couplings @xmath1 , @xmath2 , @xmath38.,width=377 ] the decay width of @xmath32 depending on the anomalous couplings
is given in fig .
[ decay ] . in fig .
[ decay ] , we set two of these anomalous couplings equal to zero while varying the other coupling changes in the range @xmath39 $ ] .
the horizontal dashed line ( black ) denotes the one loop calculation of the decay width @xmath32 , which is @xmath40 mev @xcite for the higgs boson mass of 125 gev within the framework of the sm .
the limits on the anomalous couplings can be extracted from the intersection points of the experimental predictions and the calculated values of decay width @xmath41 .
it is seen from fig .
[ decay ] that the prediction from atlas @xcite experiment limits these anomalous couplings as @xmath42 and @xmath6 , @xmath43 .
the cms @xcite prediction leads to more strict bounds . due to more restriction on anomalous coupling @xmath1
we assume @xmath44 , while other couplings allowed to changed in the considered range .
the contributing tree level feynman diagrams for the single production of higgs boson through @xmath45 subprocess ( where @xmath46 ) are shown in fig .
the differential cross section for the subprocess @xmath45 is obtained as neglecting quark masses , @xmath47}[4a_\gamma^2m_z^4 + 4a_\gamma b_\gamma m_z^2(\hat{t}-m_h^2)\\\nonumber&&+(b_\gamma^2 + \tilde{b}_\gamma^2)[m_h^4+\hat{t}^2 + 2(\hat{s}-2m_h^2)(\hat{s } + \hat{t})]].\label{eq:2}\end{aligned}\ ] ] we find the total cross section of @xmath21 process by integrating differential cross section of @xmath48 subprocess over the parton distribution functions cteq6l @xcite and photon spectrum in epa @xcite , by using the calchep .
.,width=529 ] in figs .
[ fig : csb ] and [ fig : csbt ] the total cross sections of @xmath21 process as a function of @xmath2 and @xmath38 are illustrated for the parameter sets pi , pii , piii . with @xmath9 tev .
dot - dashed , dashed and solid lines correspond to parameter sets pi , pii and piii , respectively.,width=377 ] with @xmath9 tev .
dot - dashed , dashed and solid lines correspond to parameter sets pi , pii and piii , respectively.,width=377 ]
the @xmath49 test for the bounds of anomalous @xmath50 and @xmath51 couplings at 95 % c.l . can be written as @xmath52 where @xmath53 is the cross section including considered anomalous couplings ; @xmath54 is the statistical error and here @xmath55 is the number of sm events .
the number of events are given by @xmath56 where @xmath57 denotes the @xmath58-tagging efficiency and @xmath59 is the integrated luminosity .
we use kinematical cuts for transverse momentum of final state quarks to be @xmath60 gev and pseudorapidity to be @xmath61 . in tables
[ tab : tab1 ] , [ tab : tab2 ] and [ tab : tab3 ] , we have listed 95% c.l .
sensitivity limits on the couplings @xmath50 and @xmath51 for various integrated luminosities by varying one coupling at a time , for sets of minimal transverse momentum cuts pi , pii and piii , respectively .
.sensitivity ( 95% c.l . ) to anomalous @xmath0 couplings for parameter set pi in the @xmath4 collisions at the lhc with @xmath9 tev for various integrated luminosities .
[ tab : tab1 ] [ cols="^,^,^,^",options="header " , ] the sensitivities to the anomalous couplings in @xmath62 plane are plotted when two of the anomalous parameters are changed independently in figs .
[ fig : cont1 ] , [ fig : cont2 ] and [ fig : cont3 ] for the parameter sets pi , pii and piii , respectively . and @xmath38 for the process @xmath63 with @xmath9 tev for parameter set pi.,width=377 ] and @xmath38 for the process @xmath63 with @xmath9 tev for parameter set pii.,width=377 ] and @xmath38 for the process @xmath63 with @xmath9 tev for parameter set piii.,width=377 ]
we have analyzed the single production of higgs boson including anomalous @xmath0 vertices at the photon - induced @xmath4 collisions at the lhc .
this process gives opportunity to investigate anomalous @xmath0 couplings @xmath6 and @xmath64 at the order of @xmath65 .
we find the best limits on these anomalous couplings which can be compared to the bounds @xmath66 obtained from the associate production of @xmath67 @xcite at linear collider with @xmath68 gev .
99 g. aad _ et al . _ [ atlas collaboration ] , phys .
b * 716 * , 1 ( 2012 ) .
s. chatrchyan _ et al . _
[ cms collaboration ] , phys .
b * 716 * , 30 ( 2012 ) .
k. hagiwara , s. ishihara , j. kamoshita and b. a. kniehl , eur .
j. c * 14 * , 457 ( 2000 ) .
s. s. biswal , r. m. godbole , r. k. singh and d. choudhury , phys .
d * 73 * , 035001 ( 2006 ) [ erratum - ibid .
d * 74 * , 039904 ( 2006 ) ] . v. m. budnev , i. f. ginzburg , g. v. meledin and v. g. serbo , phys .
rept . * 15 * , 181 ( 1975 ) .
i. f. ginzburg , g. l. kotkin , v. g. serbo and v. i. telnov , nucl .
instrum . meth . *
205 * , 47 ( 1983 ) .
k. piotrzkowski , phys .
d * 63 * , 071502 ( 2001 ) . c. royon [ rp220 collaboration ] , arxiv:0706.1796 [ physics.ins-det ] .
m. g. albrow _ et al . _
[ fp420 r and d collaboration ] , jinst * 4 * , t10001 ( 2009 ) .
s. atag and i. t. cakir , phys .
d * 63 * , 033004 ( 2001 ) .
o. kepka and c. royon , phys .
d * 78 * , 073005 ( 2008 ) .
j. de favereau de jeneret , v. lemaitre , y. liu , s. ovyn , t. pierzchala , k. piotrzkowski , x. rouby and n. schul _ et al .
_ , arxiv:0908.2020 [ hep - ph ] .
e. chapon , c. royon and o. kepka , phys .
d * 81 * , 074003 ( 2010 ) .
m. g. albrow , t. d. coughlin and j. r. forshaw , prog .
phys . * 65 * , 149 ( 2010 ) .
i. sahin and a. a. billur , phys .
d * 83 * , 035011 ( 2011 ) .
i. sahin and b. sahin , phys .
d * 86 * , 115001 ( 2012 ) .
b. sahin and a. a. billur , phys .
d * 86 * , 074026 ( 2012 ) .
a. senol , phys .
d * 87 * , 073003 ( 2013 ) .
e. gabrielli , v. a. ilyin and b. mele , phys .
d * 56 * , 5945 ( 1997 ) [ erratum - ibid .
d * 58 * , 119902 ( 1998 ) ] .
u. cotti , j. l. diaz - cruz and j. j. toscano , phys . lett .
b * 404 * , 308 ( 1997 ) .
s. dutta , k. hagiwara and y. matsumoto , phys .
d * 78 * , 115016 ( 2008 ) .
s. d. rindani and p. sharma , phys .
d * 79 * , 075007 ( 2009 ) . s. d. rindani and p. sharma , phys .
b * 693 * , 134 ( 2010 ) .
a. gutierrez - rodriguez , j. montano and m. a. perez , j. phys .
g * 38 * , 095003 ( 2011 ) .
v. hankele , g. klamke , d. zeppenfeld and t. figy , phys .
d * 74 * , 095001 ( 2006 ) .
a. belyaev , n. d. christensen and a. pukhov , arxiv:1207.6082 [ hep - ph ] .
|
we have examined @xmath0 vertex to obtain the limits on anomalous @xmath1 , @xmath2 and @xmath3 couplings in a model independent way through the @xmath4 collisions via the process @xmath5 .
the sensitivities to the anomalous couplings can be obtained as @xmath6 , @xmath7 for the integrated luminosity of @xmath8 at the lhc with @xmath9 tev .
|
the long - range behavior of gapped systems in condensed matter physics is accessible via topology . for noninteracting fermionic systems
there is a classification of topological phases using ideas related to @xmath0-theory @xcite . over the past few years
the interacting case has been vigorously studied , for both fermionic systems and bosonic systems , with an emphasis on _ short - range entanglement _ ( sre ) ; a small sampling of papers is @xcite .
symmetry protected topological _ ( spt ) phases are captured by group cohomology @xcite , but other investigations ( e.g. @xcite ) reveal the existence of additional sre phases and raise the question of a complete classification . in this paper
we propose an invariant of bosonic and fermionic sre topological phases constructed from effective field theory .
computations and examples demonstrate that it effectively detects known sre phases .
our proposal applies to gapped systems which at low energy ( long time ) can be approximated by topological field theories . such a system must be sufficiently local that it can be formulated on arbitrary manifolds , and it must have a continuum limit which is a field theory , at least at low energy .
we do not investigate microscopic behavior at all in this paper , but rather simply assume the existence of a long - range topological theory .
short range entanglement , or the absence of topological order , is a microscopic assumption .
kitaev @xcite has been studying sre phases from first principles microscopically , and has suggested the macroscopic consequence that the long - range topological theory has a unique vacuum on any background manifold .
we go further and assume that the long - range topological field theory describing an sre phase is _ fully extended _ and _ invertible _ , concepts that we explain below . from mathematical investigations
it has been known for a long time that fully extended , invertible , _ topological _ field theories are equivalent to maps between spectra in the sense of algebraic topology .
this link with stable homotopy theory is the basis of our proposal . while our proposal in [ subsec:5.2 ] is specific and precise , we neither formulate nor prove a mathematical theorem which justifies it .
also , while we enumerate groups which should house invariants of sre phases , we do not argue either that the effective field theory is a complete invariant or that every possible effective field theory is realized by a microscopic system .
in place of proof the paper marshals evidence in two stages .
pre-[subsec:5.2 ] is a long conceptual march leading to the proposal .
post-[subsec:5.2 ] is a series of experimental checks , including the relationship to group cohomology , boundary terminations , and specific computations .
we include a long discussion in [ subsec:6.3 ] about detecting kitaev s
@xmath1 phase using invertible topological field theories .
there are additional possible effective field theories which are `` @xmath2 roots '' , perhaps an indication that not all possible effective theories are realized by microscopic systems .
another example , the `` 3d bosonic @xmath1 phase with half - quantized surface thermal hall effect '' @xcite , @xcite , is also treated in detail .
it may be useful to broadly characterize our proposal in field - theoretic language : whereas the group cohomology captures pure gauge theories , the additional sre phases contain couplings to gravity or are purely gravitational .
the spt phases have no purely gravitational component .
more fundamentally , the long - range field theory is envisioned as the low - energy behavior of the coupling to gravity of the original system .
( and , if there are global symmetries , we gauge them and so couple to gauge theory too . )
bordism as a tool to classify spt phases appears differently in the recent papers of kapustin @xcite , @xcite , @xcite . in unpublished work kitaev
@xcite develops a classification of sre phases based on microscopic considerations .
their results and approach differ from ours , and it will be very interesting to reconcile them
. it may be that there is more microscopic information in the physics which leads to different or additional input into the effective field theories .
in particular , there are a few ingredients in our proposal ( choice of tangential structure , choice of target spectrum ) which involve leaps of faith and can easily be adjusted if further microscopic implications are discovered .
we begin in [ sec:2 ] with an exposition of several formal points in field theory : extended field theory , invertible field theory , relative field theory , anomalies , global symmetries and gauging , topological field theory , unitarity . while many of these concepts are familiar to physicists , the mathematical language may be unfamiliar and we hope to provide some bridge here .
we touch on the cobordism hypothesis , which classifies fully extended topological theories , in [ subsec:2.5 ] ; the precise nature of this classification is illuminated in [ subsec:2.7 ] with a toy example in preparation for a later discussion of the `` kitaev @xmath1 phase '' in [ subsec:6.3 ] .
but the cobordism hypothesis is overkill for _ invertible _ theories . in
[ subsec:2.6 ] we describe the link between fully extended , invertible , topological field theories and stable homotopy theory in general terms . that involves particular _ madsen - tillmann spectra _ , which are unstable analogs of thom s bordism spectra ; we describe them in [ sec:4 ] .
there we also discuss the crucial theorem of galatius - madsen - tillmann - weiss @xcite which identifies these spectra as geometric realizations of bordism categories . that these _ unstable _ bordism spectra are appropriate to field theories is natural since field theories are dimension - specific .
the material in [ sec:2 ] and [ sec:4 ] is general background not particular to condensed matter systems . in [ sec:3 ] we give much of the general argument about effective field theories for sre topological phases .
we begin in [ subsec:3.1 ] with elementary thoughts indicating why topology may sufficiently describe the low energy behavior of gapped systems .
most of our assumptions are stated explicitly in [ subsec:3.3 ] .
there is still conceptual work to translate those assumptions into concrete mathematical statements .
specifically , once we know that sre gapped phases give rise to invertible topological theories , there are still parameters to choose : the tangential structure on manifolds representing space and the target category for the field theory .
the latter is discussed in [ subsec:5.1 ] separately for bosonic and fermionic theories ; the bosonic case is a bit surprising and we settle on a kludge for the target spectrum . for the tangential structures we assume without much justification that in bosonic theories the space manifolds are oriented and in fermionic theories they are spin . along the way we encounter a few tricky issues , for example the gauging of antilinear symmetries ( [ subsubsec:2.4.3 ] , [ subsubsec:5.1.3 ] ) and implementation of unitarity ( [ subsubsec:4.2.5 ] ) .
we state our proposal in [ subsec:5.2 ] .
we divide theories into bosonic and fermionic .
also , symmetries may be anomalous and we propose a classification of anomaly theories and anomalous gauged theories as well .
possible effective invertible topological theories for sre phases with fixed symmetry form an abelian group ; for spt phases they form a subgroup which we also delineate .
our first deduction in [ subsec:5.3 ] from the proposal is that the phases previously identified using group cohomology are included . in [ subsec:6.1 ] we show that for bosonic theories in @xmath3 space dimensions group cohomology provides a complete classification , which agrees with known results @xcite .
already for fermionic theories in @xmath3 the situation is more interesting , as described in [ subsec:6.2 ] : we detect the majorana chain @xcite in our classification . in [ subsec:6.3 ] we identify bosonic @xmath4 sre phases .
these were introduced by kitaev @xcite , @xcite and are related to 2-spacetime above is the dimension of space ; spacetime has dimension @xmath5 . ]
dimensional chiral conformal field theories whose chiral central charge is an integer divisible by 8 . these central charges
do not show up in the usual account of the associated 3-spacetime dimensional topological field theory ; there only the reduction mod 8 is used .
our explanation of how they fit in here , and so the role of the chiral central charge as a real number not taken mod 8 , is based on the easier examples discussed in [ subsec:2.7 ] . in [ subsec:6.4 ] we illustrate a constraint imposed by unitarizability . in
[ subsec:6.5 ] we compute that the abelian group of @xmath6 bosonic time - reversal symmetric effective sre field theories is isomorphic to @xmath7 .
one generator is accounted for by group cohomology , and we claim the other is the 3d bosonic @xmath1 phase with half - quantized surface thermal hall effect mentioned earlier .
finally , in [ sec:7 ] we give a general discussion of boundary conditions / terminations / excitations and use it to justify the aforementioned claim .
an appendix includes topological computations which are needed in the text .
the notion of a non - extended _ invertible _ field theory arose in joint work with greg moore @xcite .
fully extended invertible topological theories , and the relation to stable homotopy theory , has been a longstanding discussion topic with mike hopkins and constantin teleman , as have many other general ideas described in [ sec:2 ] . in particular , we used these ideas in @xcite to construct a topological field theory based on the verlinde ring .
the specific application to sre phases described here crystallized during the _ symmetry in topological phases _ workshop in princeton , and i thank the organizers for inviting me .
i had long conversations with alexei kitaev after a first draft of this paper was complete , and those inspired a significant modification of [ subsubsec:5.1.1 ] and [ subsec:6.3 ] .
i thank him for sharing his perspectives .
i also thank zheng - cheng gu , mike hopkins , anton kapustin , constantin teleman , ashvin vishwanath , kevin walker , oscal randal - williams , and xiao - gang wen for very helpful conversations and correspondence .
we begin with a formal viewpoint on the structure of a field theory , which is the lens through which we analyze the long - range effective topological theory in [ sec:3 ] . in the mathematics literature
this approach was abstracted in segal s axioms for two - dimensional conformal field theory @xcite and in atiyah s axioms for topological field theories @xcite .
the lectures @xcite treat general quantum field theories from this perspective .
our focus in this paper is on _ invertible _ field theories , which we define in [ subsec:2.2 ] .
other general topics we quickly review include extended field theories , relative field theories , anomalies , gauging symmetries , and unitarity .
we then focus on fully extended topological theories , for which the powerful _
cobordism hypothesis _
@xcite provides a classification result .
invertible topological theories can be analyzed using homotopy theory , and in this section we explain why that is true but defer a more precise description to [ sec:4 ] .
we conclude with a few toy examples which illuminate subtleties we will encounter in the condensed matter systems of [ sec:6 ] .
the subtleties discussed there may have broader interest .
there are many expositions of this material , in addition to the ones referenced earlier in this paragraph , and here we offer another .
a somewhat different point of view on topological field theories may be found in @xcite .
the reader may wish to use this section for reference and skip on first reading to later parts of the paper .
we remark that this formal viewpoint does not distinguish `` classical '' from `` quantum '' , and indeed we will give examples of both types .
another remark is that the field theories which arise in condensed matter physics are usually defined on spacetimes which are products of space and time , whereas the discussion in this section models theories defined on more general spacetimes .
we discuss the necessary modification in [ subsec:4.2 ] and account for it in the proposals of [ subsec:5.2 ] .
let @xmath8 be the _ spacetime _ dimension of a field theory @xmath9 , which we simply call the dimension of @xmath9 .
we write @xmath10 where @xmath11 is the _ space _ dimension .
is often called the dimension of the theory , whereas in quantum field theory and string theory literature it is @xmath8 which is the dimension .
we use the terms ` spacetime dimension ' and ` space dimension ' to avoid confusion . ]
the case @xmath12 ( @xmath13 ) is mechanics ; there is only time .
a quantum mechanical system assigns a complex vector space @xmath14 ( the ` quantum hilbert space ' ) to a point and the time evolution @xmath15 to a closed interval of length @xmath16 .
the group law @xmath17 is encoded by gluing intervals , as illustrated in figure [ fig:1 ] .
an @xmath8-dimensional euclidean field theory assigns a _ partition function _
@xmath18 to a compact @xmath8-dimensional manifold @xmath19 with no boundary .
there is a complex vector space @xmath20 for each compact @xmath21-manifold , thought of as a spatial slice , and now this ` quantum hilbert space ' may depend on @xmath22 . roughly speaking , the vector space @xmath23 attached to a small sphere
@xmath24 is the space of local operators , and to a closed manifold @xmath19 with @xmath25 small open balls removed we attach the _ correlation functions _ @xmath26 as illustrated in figure [ fig:2 ] . here
all boundary components are `` incoming '' , whereas each interval in figure [ fig:1 ] has an incoming boundary component and an outgoing boundary component , each a single point .
there are analogous quantum evolution operators in any dimension for manifolds with both incoming and outgoing components , and the group law of quantum mechanics has a generalization .
the mathematical expression of this formal structure is the assertion that @xmath27 is a homomorphism , or _ functor _ , between _
symmetric monoidal categories_. the _ bordism _ category @xmath28 consists of closed @xmath21-manifolds and bordisms between them .
a bordism @xmath29 is a compact @xmath8-manifold with boundary , the boundary has a continuous partition @xmath30 which divides it into incoming and outgoing components , and there are diffeomorphisms @xmath31 and @xmath32 ; see figure [ fig:3 ] in which one should view time as flowing from left to right . the target category has complex vector spaces as objects and linear maps as morphisms . ]
there are two kinds of composition .
the internal composition glues morphisms ( figure [ fig:4 ] ) and the external composition is disjoint union .
similarly @xmath33 has two composition laws : the internal composition is the usual composition of linear maps and the external composition is tensor product .
the homomorphism is required to preserve both composition laws .
this formulation is rather compact and one must unpack it to see the usual structures in field theory .
of morphisms @xmath29 and @xmath34 typically one does not have bare @xmath8-manifolds , but rather each @xmath8-manifold @xmath19 is endowed with a space @xmath35 of fields .
for example , in quantum mechanics ( figure [ fig:1 ] ) the 1-manifolds have a riemannian metric : the total length represents time .
higher dimensional field theories are often formulated on riemannian manifolds , though conformal field theories only require a conformal structure .
other possible fields include orientations , spin structures , scalar fields , spinor fields , etc .
there is a bordism category @xmath36 of manifolds equipped with a specified collection of fields , and a functor @xmath37 represents a field theory with @xmath38 as the set of ( background ) fields .
[ thm:8 ] to illustrate the notation , consider @xmath39 spacetime dimensional chern - simons theory with gauge group @xmath40 .
there is a classical and a quantum theory .
the fields @xmath35 in the classical theory on a 3-manifold @xmath19 consist of an orientation @xmath41 and a principal @xmath42-bundle with connection @xmath43 .
a closed 3-manifold @xmath19 appears in the bordism category @xmath44 as a morphism @xmath45 from the empty 2-manifold to itself .
we have @xmath46 and so @xmath47 is multiplication by a complex number , and if we make the fields explicit we denote it as @xmath48 .
it is given by the formula @xmath49 where implicitly , since @xmath50 is a 1-form on the total space of a @xmath42-bundle , we have used a global section to pull it down to the base @xmath19 .
the orientation is used to define integration of differential forms .
an oriented 2-manifold @xmath22 with @xmath42-connection has an attached chern - simons line @xmath51 , and a bordism with fields has a relative chern - simons invariant mapping between the chern - simons lines of the boundaries .
this theory is _
invertible _ in the sense described in [ subsec:2.2 ] . in the quantum theory
@xcite one integrates over the gauge field @xmath50 , and to get a well - defined 3-dimensional theory one needs in addition to the orientation @xmath41 a field @xmath52 which is a certain sort of `` framing '' called a _
@xmath53-structure_. the quantum invariant @xmath54 of a closed 3-manifold is an arbitrary complex number for example , it vanishes for some manifolds and the quantum vector spaces @xmath55 do not necessarily have dimension one .
so the quantum theory is not generally invertible .
[ thm:1 ] this description of a field theory has an important deficiency : it does not encode _ smooth _ dependence on parameters .
for example , the partition function on a closed @xmath8-manifold @xmath19 must depend smoothly on the background fields in @xmath35 .
( example : the smooth dependence of the partition function of 2-dimensional yang - mills theory on the area of a surface . )
formally , the definitions are enlarged to include fiber bundles @xmath56 of @xmath8-manifolds with fields , and these must map to smoothly varying linear maps of smooth vector bundles over @xmath57 . for topological theories , which are our main concern , instead of families
one usually postulates instead that the morphism sets in the categories @xmath36 and @xmath33 have a topology and all maps are _
continuous_. there are two obvious topologies on the set of linear maps @xmath58 between two finite dimensional vector spaces , equivalently on the set of @xmath59 matrices .
we can use the usual topology induced from the usual topology on the real numbers , or we can use the discrete topology . both are used in the classification scheme of [ subsec:5.2 ] .
[ thm:2 ] one should also allow smooth families of field theories @xmath9 . for topological field theories
it is more natural to study _ continuous _ families , i.e. , to form a _ topological space _ of topological field theories .
this is crucial for the classification of gapped topological phases : if two gapped systems are connected by a continuous path , we expect their effective topological field theories can also be joined by a continuous path .
finally , in this paper we will always consider _ fully extended _
field theories , usually topological .
an @xmath8-dimensional fully extended theory assigns invariants to manifolds of all dimensions @xmath60 .
these compact manifolds with fields , which are now allowed corners , are organized into an algebraic structure called a _
symmetric monoidal @xmath61-category _ , denoted @xmath62 .
the target for an extended field theory is a symmetric monoidal @xmath63-category @xmath64 , which typically has its `` @xmath65 loop space '' isomorphic to @xmath33 .
if so , then an extended field theory @xmath66 restricts on @xmath21- and @xmath8-manifolds to a usual field theory .
as already indicated , the invariants attached to manifolds of dimension @xmath67 tend to be categorical in nature .
a theory which extends in this way is fully local , and it is natural to make this strong locality hypothesis for the effective topological theory which comes from a gapped physical theory . see @xcite for a modern description of fully extended topological field theories .
` invertibility ' refers to the tensor product operation on vector spaces and linear maps .
first , @xmath68 is a `` unit element '' for tensor product in the sense that for any vector space @xmath69 we have an isomorphism @xmath70 which is naturally defined .
thus we call @xmath68 a _
tensor unit_. a complex vector space @xmath69 is _ invertible _ if there exists a vector space @xmath71 and an isomorphism @xmath72 . since @xmath73
, it follows immediately that if @xmath69 is invertible , then @xmath74 .
conversely , if @xmath69 is 1-dimensional then @xmath75 is isomorphic to @xmath68 .
thus the invertible vector spaces are precisely the _
lines_. invertibility of a linear map @xmath76 under tensor product is equivalent to the usual definition of invertibility , namely that there exist @xmath77 such that the compositions @xmath78 and @xmath79 are identity maps .
if @xmath80 is a line , then a linear map @xmath81 is multiplication by a complex number @xmath82 and it is invertible if and only if @xmath83 .
we denote the nonzero complex numbers as @xmath84 .
invertible complex vector spaces and invertible linear maps between them form a subcategory @xmath85 which by definition has the property that all morphisms are invertible .
such a category is called a _
groupoid_. an _ invertible field theory _
@xmath86 is one for which all vector spaces @xmath20 and linear maps @xmath87 are invertible.for a fully extended invertible field theory the value of @xmath9 on _ any _ manifold of dimension @xmath60 is invertible under the symmetric monoidal product of the target .
a theorem of the author and constantin teleman asserts that for oriented theories if the number @xmath88 is nonzero and the vector spaces @xmath89 are one - dimensional , then @xmath9 is invertible . ]
we can express that by saying that @xmath9 factors through a functor @xmath90 .
thus all quantum hilbert spaces are one - dimensional and all propagations are invertible .
the tensor product of invertible theories is invertible , so ( isomorphism classes of or deformation classes of ) invertible theories form an abelian group .
[ thm:3 ] here is a simple example with @xmath12 .
fix a smooth manifold @xmath91 of any dimension and a smooth complex line bundle @xmath92 with connection .
we define a 1-dimensional field theory whose set of fields @xmath35 on a 1-manifold @xmath19 consists of a pair @xmath93 of an orientation @xmath94 and a smooth map @xmath95 . then to @xmath96 a point with the positive orientation and @xmath97 we set @xmath98 to be the fiber of the line bundle @xmath92 at @xmath99 .
to @xmath100 $ ] with the usual orientation and a map @xmath101\to m$ ] we assign the parallel transport @xmath102 along the path @xmath103 .
the reader can easily work out the values of @xmath9 on other manifolds .
we can make a family of such field theories by varying the line bundle @xmath92 and its connection .
the path components of this family of field theories are parametrized by the topological equivalence classes of line bundles @xmath92 ( without connection . ) in [ sec:4 ] and [ sec:5 ] we will learn that the set of path components can be computed by stable homotopy theory.\cong h^2(m;{{\mathbb z}})$ ] . in all computations
@xmath104 $ ] denotes _ pointed _ homotopy classes of maps between the pointed spaces @xmath105 . ] [ thm:6 ] continuing with @xmath12 we now take the line bundle to be one of the fields , rather than being pulled back from an external manifold .
thus let @xmath35 consist of an orientation @xmath41 and a complex line bundle @xmath106 _ with connection_. the definition of the theory is similar to that in example [ thm:3 ] .
( the two theories are related : choose the universal line bundle @xmath107 in example [ thm:3 ] . )
we continue this example in [ subsec:2.7 ] .
[ thm:4 ] let @xmath108 and suppose @xmath38 includes just an orientation .
the line @xmath20 attached to any closed 1-manifold @xmath22 is the trivial line @xmath68 and the number attached to any closed 2-manifold @xmath19 is @xmath109 the exponential of the euler number with base some @xmath110 .
this is a connected\cong h^3(bso_2;{{\mathbb z}})=0 $ ] . ]
family of theories with parameter space\cong h^2(bso_2;{{{\mathbb c}}/{{\mathbb z}}})\cong { { { \mathbb c}}/{{\mathbb z}}}$ ] .
aficionados may relish the following .
if we compose with @xmath111 , then the theories with parameter @xmath110 and @xmath112 become isomorphic . here
@xmath113 is the brown - comenetz dual of the sphere spectrum ( [ subsubsec:5.1.1 ] ) .
note the numerical invariants of 2-manifolds only depend on @xmath114 . ]
if we drop the orientation , then there is another theory not connected to this family : we compute : @xmath115\cong h^3(bo_2;\widetilde{{{\mathbb z}}})\cong { { { { \mathbb z}}/2{{\mathbb z}}}}$ ] . here
@xmath116 is the nontrivial local system on @xmath117 . ]
the invariant of a closed surface @xmath19 is @xmath118 , where @xmath119 is the characteristic number associated to the square of the first stiefel - whitney class of the tangent bundle .
( it is nontrivial for the real projective plane , for example . )
classical lagrangian field theories are invertible field theories : the invariant of a closed @xmath8-manifold is the _ exponentiated _ action @xmath120 . here ` @xmath19 ' includes a choice of background fields .
[ thm:5 ] finite gauge theories provide a typical example of a classical topological theory in any dimension @xmath8 @xcite .
let @xmath121 be a finite group and fix a cocycle which represents a cohomology class @xmath122 .
the fields @xmath35 are an orientation and a principal @xmath121-bundle @xmath123 .
( as principal @xmath121-bundles have automorphisms deck transformations the fields in this case form a groupoid , or stack , rather than a space ; see @xcite for one mathematical treatment . )
let @xmath124 be the pairing of the characteristic class @xmath125 of @xmath126 in @xmath127 with the fundamental class of the orientation .
the invariant of the invertible field theory is @xmath128 .
note that no orientation is required if the cocycle vanishes .
this _ classical _ dijkgraaf - witten theory is invertible ; the _ quantum _ theory , obtained by a finite sum over bundles @xmath126 , is typically not invertible .
the case @xmath39 is a special case of chern - simons theory ( briefly described in example [ thm:8 ] with gauge group @xmath42 ) .
a field theory @xmath9 as described in [ subsec:2.1 ] might be termed _
suppose @xmath129 is an ( absolute ) @xmath130-dimensional field theory .
then we can have an @xmath8-dimensional theory @xmath9 which is defined _ relative _ to @xmath129 .
in fact , it is more precise and important to realize that we need only the truncation @xmath131 of @xmath129 which remembers the values on manifolds of dimension @xmath60 .
thus @xmath132 need only be defined on such manifolds in the first place . to a closed @xmath8-manifold @xmath19 ( with fields )
the theory @xmath129 assigns a vector space @xmath133 .
a relative theory @xmath9 then assigns either a linear map @xmath134 or a linear map @xmath135 in the first case we evaluate the map on @xmath136 to obtain a vector in @xmath133 ; in the second case @xmath87 is a covector , an element of the dual vector space .
there are similar statements for lower dimensional manifolds . in the first case
we write @xmath137 and in the second @xmath138 where @xmath139 is the trivial theory .
if @xmath129 is invertible , then @xmath9 is termed an _ anomalous _ field theory with _
anomaly _ theory @xmath129 .
we refer to @xcite for more explanations and for nontrivial examples .
here is an easy one , which illustrates the relationship with boundary conditions .
relative theories can be viewed as boundary conditions in any dimension , an idea we take up in [ sec:7 ] .
[ thm:7 ] let @xmath140 and suppose @xmath129 is a quantum mechanics theory ( figure [ fig:1 ] ) with hilbert space @xmath14 attached to a point
. then a relative theory @xmath141 is determined by evaluating on 0-manifolds , and it is enough to evaluate on a point with each orientation . since @xmath142 and @xmath143
we obtain a vector @xmath144 and a dual vector @xmath145 .
we can use the relative theory @xmath9 as a boundary condition for @xmath129 , as illustrated in figure [ fig:5 ] .
let @xmath121 be a lie group . for simplicity
we discuss a non - extended field theory .
a global symmetry group may have an action on fields , and it also may have an action on the category @xmath33 . in the simplest cases those actions are trivial , and then @xmath121 is a _ global symmetry _ if the functor lifts to a functor @xmath146 into the category of representations of @xmath121 .
more plainly , the group @xmath121 acts on the vector space @xmath20 attached to each @xmath21-manifold and the linear maps assigned to bordisms are @xmath121-invariant .
in this situation we might `` gauge the symmetry '' or , in less ambiguous terms , construct a _
@xmath121-equivariant extension _ of the theory .
this means that there is a new field which is a @xmath121-connection ; if @xmath121 is finite , then a @xmath121-connection is simply the underlying principal @xmath121-bundle .
whence the terminology : this is the gauge field .
there is a new set of fields @xmath147 which maps to the single field is most naturally a simplicial sheaf on the category of smooth manifolds @xcite . ]
@xmath148 of a @xmath121-connection , and the fiber over the trivial @xmath121-connection is the old set @xmath38 of fields .
the @xmath121-equivariant extension is a functor @xmath149 whose restriction to the trivial @xmath121-connection is the original theory .
this makes sense since the trivial @xmath121-connection has the group @xmath121 as its automorphism group .
[ thm:9 ] a typical example in field theory is a @xmath150-model into a riemannian manifold @xmath91 with a group @xmath121 of isometries .
the classical model makes sense in any spacetime dimension @xmath8 .
the fields @xmath35 on an @xmath8-manifold consist of a metric @xmath151 and a map @xmath95 . in the @xmath121-equivariant extension
an element of @xmath152 is a triple @xmath153 where @xmath154 is a connection on a principal @xmath121-bundle @xmath123 and now @xmath155 is a @xmath121-equivariant map @xmath156 .
the map to @xmath148 sends @xmath157 to @xmath154 .
if @xmath154 is the trivial @xmath121-connection then @xmath155 is equivalent to a map @xmath158 , and the deck transformations of the trivial bundle @xmath159 become the original @xmath121-action on the fields .
this example does _ not _ fit our simplified description , since the global symmetry group @xmath121 _ does _ act on the fields @xmath38 , but there is a modification which covers this situation . in the quantum @xmath150-model we integrate over the field @xmath103 , and as there is no @xmath121-action on the remaining fields our description applies as is
. there may be obstructions to constructing this @xmath121-equivariant extension : see @xcite for a recent discussion and examples .
one well - known example is the gauged wzw model @xcite . in good cases
there is only a single obstruction which can be interpreted as an _ anomaly_. in these cases the extension does not exist but rather there is an invertible @xmath130-spacetime dimensional theory @xmath129 and an extension @xmath160 which is a theory relative to @xmath129 in the sense of [ subsec:2.3 ] . [ thm:10 ] a standard example in @xmath161 is quantum chromodynamics .
this theory has a global @xmath162 symmetry ( for @xmath163 the number of flavors ) which is anomalous .
we account for such anomalies in our proposals ( [ subsec:5.2 ] ) . in quantum mechanics , due to wigner s theorem ,
the global symmetry group @xmath121 is equipped with a homomorphism is the group of @xmath164 roots of unity . ]
@xmath165 which tracks whether a given symmetry acts linearly or antilinearly
. because states are lines in a hilbert space , rather than vectors , the group @xmath121 acts projectively and there is an extension by the group @xmath42 of unit norm scalars . for theories tied to spacetime , as opposed to abstract theories , one can also track whether or not symmetries reverse the orientation of time by another homomorphism @xmath166 in many situations @xmath167 , but that need nt be so in general .
see @xcite for a general discussion .
we handle the extension by simply replacing @xmath121 with the extended symmetry group . in doing
so we must take care that the group of scalars acts by scalar multiplication on all vector spaces in the theory .
we discuss time - reversal in [ subsubsec:4.2.3 ] . here
we explain how to gauge antilinear symmetries .
consider a 1-spacetime dimensional theory @xmath9 , so a quantum mechanical model as in figure [ fig:1 ] .
let @xmath168 and suppose @xmath121 is a group of global symmetries as above
. for simplicity , assume @xmath121 is a discrete group . as explained in [ subsubsec:2.4.1 ] a @xmath121-equivariant extension @xmath160 is a theory on manifolds equipped with a principal @xmath121-bundle .
now @xmath14 is the value of @xmath160 on @xmath169 equipped with the trivial ( which means _ trivialized _ ) @xmath121-bundle . consider the 1-manifold @xmath100 $ ] with a ( necessarily trivial , but not trivialized ) @xmath121-bundle , and suppose there are trivializations over the endpoints @xmath170 .
let @xmath171 be the parallel transport .
then @xmath172 is the action of the global symmetry @xmath151 .
however , if @xmath173i.e . , if @xmath151 acts antilinearly then this does not fit into since the category @xmath33 has only _ linear _ maps .
the way out is that @xmath155 defines a 2-dimensional invertible anomaly theory @xmath129 , and @xmath160 is an anomalous theory with anomaly @xmath129 , as in .
the anomaly theory assigns @xmath174 , the complex conjugate to the trivial line of complex numbers , and then the relative theory gives a _ linear _ map @xmath175 ( a linear map @xmath176 is equivalent to an antilinear map @xmath177 . see ( *
* ( 2.7 ) ) for the analog of in an arbitrary relative theory . )
this discussion extends to higher dimensions . for extended field theories with values in a higher category @xmath64
we would need to explain how complex conjugation acts on @xmath64 . in this paper
we focus on invertible field theories , and we will implement this `` antilinearity anomaly '' using twisted cohomology ; see [ subsubsec:5.1.3 ] . the formal setup of topological field theory described here is based on the _ euclidean _ version of quantum field theory . for euclidean qfts unitarity is expressed by both a reality condition and a _ reflection - positivity _ condition .
( a standard reference is @xcite ; see @xcite for a heuristic explanation . )
the unitarity condition for a fully extended topological theory implements only the reality condition .
namely , assuming the fields @xmath38 include an orientation there is an involution of @xmath62 which reverses the orientation .
also , assuming that the target @xmath64 is based on complex numbers , then it has an involution of complex conjugation .
unitarity is the statement that @xmath178 is equivariant for these involutions .
a formal justification from the path integral stems from a basic fact : orientation reversal conjugates the euclidean action . in this paper
we indicate how to implement unitarity for invertible topological field theories , which are maps of spectra .
the involutions are quite explicit , and unitarity amounts to a twisted extension of the field theory to unoriented manifolds .
one subtlety , which also occurs in non - invertible theories , is that for spin theories there are two notions of unitarity the two different euclidean pin groups lead to two orientation - reversing involutions on spin manifolds . because of the gaps in our understanding of unitarity , related to positivity and the choice of pin group
, we do not implement unitarity fully in our proposal in [ subsec:5.2 ] .
we discuss unitarity further in [ subsubsec:4.2.4 ] , [ subsubsec:4.2.5 ] .
one can debate which theories deserve the moniker ` topological . ' ) is not topological if the gauge group is not discrete . ]
we will say that a fully extended theory is topological if the fields @xmath38 are topological , and the fields are topological if they satisfy homotopy invariance : if @xmath179 is a homotopy of local diffeomorphisms of @xmath8-manifolds , then the pullbacks by @xmath180 and @xmath181 on fields are equal .
thus orientations , spin structures , and @xmath121-bundles for discrete groups @xmath121 are all examples of topological fields .
metrics , conformal structures , and connections for positive dimensional lie groups are all examples of non - topological fields . on the other hand , _ flat _
@xmath121-connections are topological fields for any lie group @xmath121
. fully extended topological field theories are a topic of current interest in topology and other parts of mathematics .
the _ cobordism hypothesis _
, conjectured by baez - dolan @xcite and proved by hopkins - lurie in dimensions @xmath182 and in general by lurie @xcite , is a powerful result which determines the space of fully extended topological theories of a fixed type .
the ` type ' refers to both the fields @xmath38 and the target @xmath64 .
thus one speaks of `` oriented '' theories or `` framed '' theories , which tells about the topological fields in the theory .
the theorem very roughly states that a theory @xmath9 is determined by its value @xmath183 on the 0-manifold consisting of a single point .
one can intuitively think of this as the value on an @xmath8-dimensional ball , and the idea is that any @xmath8-manifold is glued together from balls , so that if the theory is fully local then its values can be reconstructed from those on a point .
furthermore , the value on a point is constrained to satisfy strong finiteness conditions .
we refer the reader to @xcite and the expository account @xcite .
[ thm:11 ] if @xmath12 and we let the field @xmath38 be an orientation , then a theory @xmath184 is determined by the vector space @xmath185 .
the finiteness condition is that @xmath185 is finite dimensional .
of course , in usual quantum mechanics the quantum hilbert space is typically infinite dimensional ; not so in this topological version .
the cobordism hypothesis tells not just about individual theories , but rather about the collection of theories with fixed @xmath38 and @xmath64 .
the first ( easy ) theorem is that this collection is an ordinary _ space _ rather than a more abstract category .
one should think of this space as parametrizing families of theories , as in example [ thm:3 ] and example [ thm:4 ] . yet in the homotopical setting for field theories , it is only the _ homotopy type _ of the space of theories which is well - defined ; see [ subsec:2.7 ] for more discussion .
the theorem in particular computes the set of path components of this space .
two theories lie in the same path component if and only if they can be continuously connected .
this matches well the notion of a _ topological phase _ , and
indeed the cobordism hypothesis is a powerful tool for distinguishing topological phases of gapped theories . in this paper
we focus on _ invertible _ theories , which describe sre phases , and the cobordism hypothesis reduces to a much easier statement , as we explain in [ subsec:2.6 ] .
we emphasize that the cobordism hypothesis for invertible and non - invertible theories determines the complete homotopy type of the space of theories , not just the set of path components .
we begin with an analogy .
let @xmath186 $ ] be the ring of polynomials in a variable @xmath187 with complex coefficients and let @xmath68 be the ring of complex numbers .
define the ring homomorphism @xmath188\to{{\mathbb c}}$ ] which sends a polynomial @xmath189 to its value @xmath190 at @xmath191 .
let @xmath192 $ ] denote the subset of polynomials with nonzero constant term .
note that @xmath57 is closed under multiplication : if @xmath193 , then @xmath194 .
extend the homomorphism @xmath9 to ratios of polynomials @xmath195 where @xmath196 .
this is for the simple reason that @xmath197 if @xmath196 , so @xmath198 makes sense .
we write @xmath199 $ ] for the ring of such ratios : we have inverted elements in @xmath57 .
this inversion construction is easy in this case since @xmath186 $ ] is a _ commutative _ ring ; it is trickier in the noncommutative case and in the categorical context to which we now turn .
suppose @xmath200 is an _ invertible _ topological field theory . by definition @xmath9
takes values in the subset @xmath201 consisting of invertibles : invertible objects , invertible 1-morphisms , and invertible morphisms at all levels .
now , as in the analogy , the fact that all values are invertible means that the theory factors through the symmetric monoidal @xmath63-category @xmath202 obtained by adjoining inverses of all morphisms : @xmath203^(.6)f \ar[d]&{\mathcal{c}}\\ |{\operatorname{bord}_{n}}({\mathcal{f}})|\ar[r]^(.6){\widetilde f } & { \mathcal{c}}^\times \ar[u]}\ ] ] the map @xmath160 encodes all information about the theory @xmath9 .
the domain and codomain of @xmath160 are each a higher category , in fact an @xmath204-category , in which all arrows are invertible . such categories are called _
@xmath204-groupoids_. the basic idea is that an @xmath205-groupoid is equivalent to a space .
this is easiest to see in the opposite direction : from a space @xmath57 we can extract an @xmath204-groupoid @xmath206 . putting aside the ` @xmath204 ' for a moment , we extract an ordinary groupoid called the _ fundamental groupoid _ @xmath207 .
its objects are the points of @xmath57 and a morphism @xmath208 between points @xmath209 is a continuous path @xmath210\to s$ ] from @xmath211 to @xmath212 up to homotopy .
this is a groupoid because paths are invertible : reverse time .
the higher groupoids track homotopies of paths , homotopies of homotopies , etc .
the conclusion is that @xmath160 may be considered as a continuous map of spaces .
this already brings us into the realm of topology .
but more is true .
the domain and codomain of @xmath160 are _ symmetric monoidal _
@xmath205-groupoids , which induces more structure on the corresponding spaces .
( recall that the monoidal product on the bordism category is disjoint union and on the category @xmath64 it is some sort of tensor product . )
namely , those spaces are _ infinite loop spaces_. so for each of the spaces @xmath57 there exist a sequence of pointed spaces @xmath213 such that the loop space of @xmath214 is for further discussion . ]
furthermore , the fact that @xmath9 preserves the symmetric monoidal structure a field theory takes compositions to compositions and disjoint unions to tensor products implies that @xmath160 is an infinite loop map .
such sequences of spaces are called _ spectra _ and an infinite loop map gives rise to a map of spectra .
the bottom line is that the space of invertible field theories ( with specified @xmath216 ) is a space of maps in homotopy theory .
we are interested in the abelian group of path components , which houses an invariant of gapped topological phases , and that is the group of homotopy classes of maps between spectra
. therefore , the computation of invertible field theories starts by recognizing the spectra in the domain and codomain , which in turn depend on the choice of @xmath38 and @xmath64 .
the domain spectra , obtained from bordism multicategories , will be discussed in [ sec:4 ] .
we remark that to study invertible field theories one does not need the cobordism hypothesis ; the power of the latter is for more general non - invertible topological field theories .
[ thm:12 ] the set of path components here has a natural abelian group structure .
in fact , a spectrum @xmath217 determines a collection @xmath218 of abelian groups , its homotopy groups .
there is additional information in the spectrum which binds these groups together , but a first heuristic is that a spectrum is some topological version of a @xmath219-graded abelian group .
[ thm:13 ] we continue with example [ thm:4 ] . in this case
@xmath108 and the field @xmath38 is an orientation . now
the target @xmath64 is a 2-category , and we need to determine the spectrum corresponding to the invertibles @xmath220 .
a typical choice is to take @xmath64 to be the 2-category of complex linear categories .
this conforms to the usual picture that a 2-dimensional field theory assigns a complex number to a closed 2-manifold , a complex vector space to a closed 1-manifold , and a complex linear category to a 0-manifold . in the invertible sub 2-groupoid @xmath221
all of these are `` 1-dimensional '' .
this means that the category is equivalent to @xmath33 , the category of vector spaces ; the vector spaces are isomorphic to @xmath68 ; and the numbers we encounter are nonzero , so elements of @xmath84 . when we make the corresponding spectrum @xmath217 , the fact that invertible categories are all equivalent implies . ] @xmath222 .
the fact that all invertible 1-morphisms are equivalent implies @xmath223 .
finally we come to @xmath224 , which captures the 2-morphisms @xmath84 . here
we get two different answers , depending on whether we consider @xmath84 to have the discrete topology or the continuous topology . in the discrete case
we have @xmath225 . for the ordinary topology we use @xmath226 to deduce @xmath227 .
higher homotopy groups of @xmath217 vanish .
hence each of @xmath228 and @xmath229 has only a single nonzero homotopy group .
such spectra are called _ eilenberg - maclane spectra _ and are basic building blocks .
the notation is @xmath230 the domain spectrum , to be explained in [ sec:4 ] , is denoted @xmath231 .
thus the two sets of path components are : @xmath232&\cong { { { \mathbb c}}^\times } , \\ [ \sigma ^2mtso_2,\sigma ^3h{{\mathbb z}}]&=0.\end{aligned}\ ] ] resuming example [ thm:4 ] we see that the first of the computations in distinguishes the euler theories , parametrized by the base @xmath110 of the exponential in .
the discreteness in @xmath228 means that theories for distinct @xmath125 can not be connected by a smooth path . in the usual topology they can , and this explains the second computation in : the space of theories is connected .
[ thm:42 ] for a general invertible bosonic theory in higher than 2-spacetime dimensions we need more nonzero homotopy groups in the target spectrum . this surprise is discussed in [ subsec:5.1 ] .
[ thm:43 ] the distinction between @xmath228 and @xmath229 is important and carries through to more elaborate target spectra .
the discrete target , based on @xmath84 with the discrete topology , is where we detect individual theories .
the other target , based on @xmath84 with its usual topology and manifested as @xmath219 shifted up one degree , is where we detect deformation classes of theories . in the next section we elaborate on the meaning of spaces of maps into targets such as @xmath229 .
we have alluded several times to the _ homotopical _ setting of the cobordism hypothesis and so too the computations of deformation classes of invertible field theories .
it often happens that the geometric interpretation of a homotopical computation is subtle .
there are many examples in topology , enumerative geometry , etc .
we illustrate in our present context with a simple example . in [ sec:6 ] we encounter a more sophisticated example of the same type in the classification of gapped topological phases .
in example [ thm:6 ] we discussed an invertible field theory in @xmath12 spacetime dimensions with fields @xmath233 a triple consisting of an orientation , complex line bundle , and connection .
then there is an obvious theory @xmath234 with values in the category of complex lines .
namely , to a point @xmath169 with positive orientation and a line bundle @xmath235 we attach the line @xmath80 .
( a line bundle over a point is a single line . )
if we reverse the orientation of @xmath169 , so consider @xmath236 , we take the dual line ; if there are several points we form the tensor product .
if @xmath237 $ ] is a line bundle with covariant derivative over the interval with its usual orientation , the field theory assigns to it the parallel transport @xmath238 from the fiber over the initial endpoint to the fiber at the terminal endpoint .
if @xmath239 is a line bundle with connection over an oriented circle , then @xmath240 assigns to it the holonomy , which is a number in @xmath84 .
this is a well - defined theory , and it is not `` topological '' according to our definition , since the connection @xmath241 is not a homotopy - invariant field .
we can instead take the set of topological fields @xmath242 consisting of an orientation and a complex line bundle but no connection .
now we ask to _ classify _ topological field theories @xmath243 as emphasized in example [ thm:13 ] it is important to specify which topology we use on linear isomorphisms in the category @xmath244 : the discrete or continuous topology . here
we use the continuous topology .
the computation of equivalence classes of theories is @xmath245 = [ s^0\wedge b{{{\mathbb c}}^\times } _ + , \sigma ^2{{\mathbb z } } ] = h^2(b{{{\mathbb c}}^\times } ; { { \mathbb z } } ) \cong { { \mathbb z}}.\ ] ] according to the discussions in [ subsec:2.5 ] we conclude that the space of theories is not connected , but rather there is an integer invariant which distinguishes deformation classes of theories . there is always a trivial theory
it sends every 0-manifold with fields to the trivial line @xmath68 and every closed 1-manifold with fields to the number @xmath246and it is in the deformation class of theories labeled by the integer 0 in .
so we are led to ask : [ thm:14 ] what theory is labeled by the integer @xmath25 in ?
can we construct a single example of such a theory ?
it is clear what to do on 0-manifolds .
for example , @xmath247 the @xmath248 tensor power of the line @xmath80 .
in other words , we observe that the truncation @xmath249 of the geometric theory above does not use the covariant derivative , and so we take @xmath250 on 0-manifolds to be that theory to the @xmath248 power
. what do we do on 1-manifolds ?
the theory @xmath240 uses the connection on @xmath251 $ ] to define a definite linear map parallel transport and the connection on @xmath252 to define a definite number the holonomy .
but the now the fields do not include a connection and we have no apparent way to determine these linear maps and numbers .
there are several ways out , and they illuminate what is is being computed in and in later computations .
similar remarks apply to all topological theories and to the cobordism hypothesis .
the first comment , already made in remark [ thm:1 ] , is that a field theory gives invariants for _ families _ of manifolds with fields parametrized by a smooth manifold @xmath57 , not just for single manifolds .
so , for example , we can consider a family of points @xmath169 parametrized by the 2-sphere @xmath253 endowed with a complex line bundle @xmath254 on the total space .
let @xmath11 be the degree of this line bundle .
a field theory @xmath250 returns another line bundle over @xmath57 , and according to @xmath250 of this family is the @xmath248 tensor power @xmath255 , which has degree @xmath256 .
the ratio @xmath257 of the degrees is the integer in , and it is detected by this 2-parameter family of points .
similarly , we can consider a 1-parameter family of circles @xmath258 parametrized by @xmath259 .
the fields are an orientation along the fibers of this map and a line bundle @xmath260 on the total space , say of degree @xmath11 .
a field theory @xmath250 returns a map @xmath261 and what the computation tells is that the winding number of this map around the origin in @xmath68 is @xmath256 . in other words ,
the computation determines homotopical information in any particular field theory in the deformation class , and that information may need to be measured in families .
this still does not construct a particular theory ; it only constrains any such . to construct a theory
we need to make some choices which in itself is not surprising but what may be surprising is that we can choose to change the domain @xmath262 or the codomain @xmath244 in order to construct a particular theory .
what s more , we may go outside the realm of _ topological _ field theories and use more general field theories .
for instance we can replace @xmath263 with @xmath264 .
there is an obvious map @xmath265 which forgets the connection @xmath241 , and the important point is that the fibers of this map are _
contractible_. that is , the space of covariant derivatives @xmath241 on a fixed line bundle @xmath106 is a contractible space .
( in fact , it is an infinite dimensional affine space . )
if we make this substitution , then we know how to construct a theory .
namely , we take @xmath266 where @xmath240 is defined after
. another choice would be to fix a smooth model of the classifying space for line bundles , say by fixing a complex hilbert space @xmath14 and taking the classifying space to be the infinite dimensional projective space @xmath267 .
furthermore , we fix a smooth line bundle @xmath268 with covariant derivative .
augment the topological fields by a contractible choice : a classifying map @xmath269 for each line bundle @xmath106 .
again there is a map from triples @xmath270 to @xmath271 and the fibers are contractible . with these choices and replacements
it is easy to construct a particular theory using parallel transport in the bundle @xmath272 , pulled back via the classifying map @xmath273 .
[ thm:28 ] a different possibility is to keep the domain fields as is but change the target category @xmath244 .
we can replace it by the _ 2-category _ of @xmath219-gerbes ; the automorphisms of any object comprise the category of @xmath219-torsors , and the automorphism group of any automorphism is @xmath219 .
this is the target of a _
2-dimensional _ field theory with fields .
this theory assigns to a line bundle @xmath106 over a closed oriented surface its degree , which is an integer .
the information on lower dimensional manifolds can be used to compute this degree locally ; see @xcite .
another insight can be gleaned from , which we unwrap : @xmath274 let double vertical bars around a higher category denote its _ geometric realization _ , which is a topological space .
the process of geometric realization inverts all arrows in the category , so factors through the single bar construction . in these terms
what we calculate in is the group of homotopy classes of maps of _ spaces _ @xmath275 whereas a particular field theory @xmath250 is a map of _ categories _ @xmath276 the functor ` geometric realization ' takes a map of categories to a map of spaces .
furthermore , grothendieck s _ homotopy hypothesis _
asserts that the functor of geometric realization is an _ equivalence _ between higher _ groupoids _ and spaces , so we expect to be able to invert @xmath277 since @xmath278 is a higher groupoid .
such an inversion would let us pass from a map of spaces to a field theory , but in practice we can not explicitly construct an inverse without making choices . in our example , @xmath279 is the groupoid of complex lines and its geometric realization as a space is @xmath280 .
we need to identify that geometric realization with an explicit model of @xmath281 to get back to one - dimensional vector spaces . in summary ,
the most relevant interpretation we can offer of the computation for this paper is our first : it computes homotopy classes of a class of theories obtained by augmenting the topological fields with geometric fields which constitute a contractible choice .
this resonates with the construction of quantum chern - simons theory @xcite , for example , in which a riemannian metric is introduced to evaluate the path integral .
[ thm:44 ] the complication here arises since we seek to interpret _ nontorsion _ elements of the group of deformation classes , computed in . _ torsion _ elements , such as in footnote [ foot:12 ] , lift to maps into @xmath282 shifted down a degree , and so are realized by definite topological theories .
in this section we begin to apply the generalities about invertible topological field theories to short - range entangled phases .
we start in [ subsec:3.1 ] with some general remarks about scaling , energy gaps , and effective theories . in [ subsec:3.3 ] we give general arguments about short - range entangled ( sre ) phases in gapped condensed matter systems . in the last subsection [ subsec:3.4 ] we bring in local symmetries , including time - reversal , and in particular define symmetry protected topological ( spt ) phases in terms of invertible topological field theories . in classical nonrelativistic physics
there are three basic _ dimensions _ : length ( @xmath80 ) , time ( @xmath283 ) , and mass ( @xmath91 ) .
( for mathematical discussions of dimensions and units , see @xcite , @xcite . )
the dimensions of other physical quantities can be expressed in terms of these .
for example , energy has dimension @xmath284 = \frac{ml^2}{t^2}.\ ] ] universal physical constants also have dimensions , for example planck s constant @xmath285 and the speed of light @xmath286 : @xmath287 = \frac{ml^2}{t},\qquad [ c]=\frac lt.\ ] ] the constant @xmath285 is present in any quantum system , the constant @xmath286 in any relativistic system , and both in a relativistic quantum system .
these constants allow us to convert between dimensions , often silently by assuming units in which @xmath288 and @xmath289 .
thus in a quantum system we have @xmath290=\frac 1t\qquad \textnormal{(assuming~$\hbar$)}.\ ] ] this dimensional analysis suggests that the _ low energy _ behavior of a quantum system is reflected in its _ long time behavior_. in a relativistic quantum system we have @xmath291 using @xmath286 , and so @xmath292=m=\frac 1t=\frac 1l\qquad \textnormal{(assuming~$\hbar , c$)},\ ] ] thereby relating low energy to low mass and long time to large distance .
consider , then , a riemannian manifold @xmath91 with riemannian metric @xmath151 .
let @xmath293 be the laplace operator on differential forms , which we take as the hamiltonian of a nonrelativistic quantum system .
, where @xmath294 ; instead , we set @xmath288 and @xmath295 . ]
the eigenvalues of @xmath293 are energies , so the low energy behavior involves the low lying eigenvalues .
equivalently , we can consider the long time evolution , which is @xmath296 for @xmath16 large .
nonzero eigenvalues lead to oscillations , which tend to cancel out if @xmath16 is large , and once more we are led to the low lying spectrum .
is more familiar and leads to the same conclusion .
] now we encounter a fundamental dichotomy . if 0 is an isolated point of the spectrum , then for @xmath16 large the kernel of @xmath293 dominates .
however , if 0 is not an isolated point of the spectrum , then the low lying continuous spectrum mixes inextricably with the kernel .
therefore , we isolate the kernel at long time only if there is a _
gap _ in the spectrum above 0 .
this always happens if @xmath91 is compact . in that case
the hodge and de rham theorems combine to prove that the kernel of the laplace operator @xmath293 on differential forms has _ topological _ significance : the dimension of the kernel of @xmath293 on @xmath297 is the @xmath298 betti number of @xmath91 .
a _ gapped quantum system _ is one in which 0 is an isolated point of the spectrum of the hamiltonian @xmath299 . by the dimensional analysis above , this _ energy gap _ is equivalent to a _ mass gap _ in a relativistic quantum system .
one is interested in the long time ( and large distance ) behavior of a system since that is what we can observe , and often that behavior is described by an _
effective _ system .
we use the general term ` long - range ' for either ` long time ' or ` large distance ' .
for example , quantum field theories at large distance are often well - approximated by another quantum field theory , and the approximating theory contains more than the vacuum state if there is no mass gap ; see @xcite for a discussion .
it is sometimes said that if there is a mass gap then the low energy theory is trivial , but there is a more nuanced truism : @xmath300 for this to make sense , we need to study the theory on arbitrary spacetimes , so couple to background gravity .
that can be done for many field theories .
for example , 4-dimensional yang - mills theory makes sense on arbitrary riemannian manifolds , and conjecturally there is a mass gap , so one should obtain a 4-dimensional topological field in the low energy limit .
( here @xmath161 is the spacetime dimension . )
[ thm:15 ] the @xmath301 supersymmetric yang - mills theory also has a conjectural mass gap , and presumably the low energy effective topological theory is more interesting in that case .
we turn to condensed matter systems , which we treat _ very _ heuristically .
we assume the quantum hilbert space @xmath14 is the tensor product over a set @xmath57 of sites of finite dimensional hilbert spaces @xmath302 and we may assume @xmath303 is a fixed hilbert space independent of the site @xmath304 ; the sites are initially arranged in a regular pattern , such as a lattice ; and the hamiltonian @xmath299 is _ local _ in that it is a sum @xmath305 where @xmath306 relative to a decomposition @xmath307 in which @xmath308 runs over sites in a small vicinity of @xmath304 and the operator @xmath283 is independent of @xmath304 .
the sites @xmath57 are located in a background _ space _
@xmath22 of dimension @xmath11 .
( we emphasize that @xmath11 is the dimension of space and @xmath5 the dimension of spacetime . )
this is a nonrelativistic system , and galilean boosts have been broken by the sites , which do nt move in time .
thus time is completely separate from the geometry of space .
initially @xmath22 typically lies in flat euclidean space , and coupling to gravity in this context means that the system can be studied on any curved @xmath11-dimensional manifold @xmath22 .
fix a compact @xmath22 and imagine that the finite set of sites @xmath57 becomes more and more dense . with no pretense of precision
we assume that the system is gapped in that 0 is the lowest eigenvalue of @xmath299 and the spectrum of @xmath299 has a fixed size gap above 0 which persists in the limit that @xmath57 becomes dense .
furthermore , we assume the kernel of @xmath299 stabilizes to a finite dimensional vector space @xmath20 .
in this situation we would like to assert : there is an effective @xmath11-space dimensional topological field theory @xmath9 which approximates the low energy / long time behavior of the system .
the vector spaces @xmath20 are part of that theory .
the deformation class of the theory @xmath9 is then a _ topological invariant _ of the original system , much the same way that the betti numbers are topological invariants of the laplace operator on a compact riemannian manifold .
[ thm:18 ] some parts of the gluing laws of a topological field theory are clearly going to hold .
for example , if @xmath309 is a disjoint union , and we envision a system as described after remark [ thm:15 ] , then the quantum hilbert space of @xmath22 is the tensor product of those on @xmath310 and @xmath311 and the hamiltonian decomposes accordingly .
therefore , so too does the kernel @xmath20 .
one expects a more subtle decomposition emerges from a continuum limit process .
[ thm:17 ] this procedure does not obviously give numerical invariants on all compact @xmath312-dimensional manifolds .
we expect invariants corresponding to time evolution , and since the field theory is defined for compact manifolds we must take circular time @xmath313 .
when we return to the initial time the manifold and its fields can undergo a symmetry , so the manifold we obtain is a fiber bundle @xmath314 .
we do expect invariants for these _ mapping cylinders_. this kind of impoverished @xmath312-dimensional field theory , perhaps with variations , goes under many names : a _ @xmath315-dimensional theory _
, a theory of _ @xmath299-type _ , .
many , in fact most , of the effective theories of @xmath299-type that we encounter do extend to full @xmath312-dimensional theories , and it may be that additional considerations in the microscopic theory can imply that property of the long - range topological theory .
even more is possible .
a 1-parameter family of @xmath312-manifolds parametrized by @xmath57 is the total space of a fiber bundle @xmath316 , and the theory gives a map @xmath317 .
its winding number is an integer invariant associated to @xmath91 .
it may happen that these integer invariants are defined for _ arbitrary _ closed @xmath318-manifolds , not just mapping cylinders .
this would impose a more severe constraint on the low energy effective theory , as we illustrate in [ subsec:6.3 ] .
geometrically , an @xmath299-type theory gives invariants for manifolds equipped with a rank @xmath11 bundle stably equivalent to the tangent bundle .
if one starts with a gapped quantum field theory , at first glance one can imagine its low energy behavior giving rise to a specific topological field theory , or at least a contractible space of theories depending on some mild choices in the approximation . for a condensed matter system
there seem to be more choices as one must , in addition to any cutoffs , take a continuum limit .
this leads to the expectation that we obtain a space of theories
again presumably contractible , but hopefully at least connected and makes it more plausible that one could encounter the phenomenon highlighted in [ subsec:2.7 ] .
indeed , we will in [ subsec:6.3 ] .
we make explicit the assumptions beyond which underlie our proposal in [ subsec:5.2 ] .
the most drastic of these is invertibility , which is an expression of short - range entanglement ; see ( [ srehyp ] ) below . 1 .
we assume that the low energy effective topological theory @xmath9 is _ fully extended _ , in the sense discussed after remark [ thm:2 ] .
this is a strong form of locality , and it seems reasonable since the continuum limit theory is obtained from local discrete systems , as in and . to define the extended theory
we must specify what sorts of ( higher categorical ) invariants @xmath9 computes on low dimensional manifolds .
in other words , we must specify the target @xmath64 for the field theory ; see .
since we restrict to invertible theories , we need only specify a spectrum .
that is one of the key choices to be made ; see [ subsec:5.1 ] for a full discussion .
we assume that @xmath9 is _ unitary _ , since microscopic condensed matter systems are typically unitary .
3 . one important consideration which affects the choice of @xmath64 is the _ topology _ on the space of theories , as already mentioned and illustrated in remark [ thm:2 ] , example [ thm:4 ] , and example [ thm:13 ] .
to the extent that the theory @xmath9 has numerical invariants of closed @xmath312-manifolds , they lie in @xmath68 ; the invariants of closed @xmath11-manifolds are @xmath68-vector spaces . in both cases
we use the usual topology on @xmath68 to allow the theory @xmath9 to deform by continuously varying the numerical invariants , the linear maps between vector spaces , etc .
two microscopic gapped systems related by a continuous deformation are considered to define the same topological phase , and they should give rise to effective topological field theories which are deformation equivalent .
on the other hand , we also consider anomaly theories @xmath129 , which can arise when gauging an anomalous global symmetry .
( see [ subsec:2.3 ] and the text before example [ thm:10 ] . ) in that case we do not allow continuous deformations of the anomaly theory : anomalous theories relative to distinct anomaly theories should not be viewed as the same topological phase .
thus , for the classification of anomalies we take the _ discrete _ topology on @xmath68 .
another important choice is of the _ background fields _
@xmath38 in the low energy effective topological theory .
we expect only topological fields .
our choice here is based on limited experience , and is more or less a guess .
naively one may think a lattice system on a manifold induces a framing , but we hope there is rotational invariance which allows us to formulate the theory on more general manifolds . as orientations are typically used to define local hamiltonians , we require invariance only under rotations connected to the identity . therefore ,
for purely bosonic systems we choose @xmath38 to include an orientation , and if the theory includes fermions then we augment this to a spin structure . in the fermionic case one can think that the coupling of such a system to background gravity involves spacetime spinor fields , whence the necessity of a spin structure .
there is an additional field a background principal bundle if we gauge a global symmetry , as we elaborate below in [ subsec:3.4 ] . 5 .
as already discussed in remark [ thm:17 ] we take @xmath9 to be a theory defined in @xmath11 space dimensions which defines complex numbers only on special closed @xmath312-manifolds . in remark [ thm:17 ]
we explain then that integers are obtained for special closed @xmath318-manifolds .
` special ' in both cases means the manifold is equipped with a rank @xmath11 vector bundle and a stable isomorphism with the tangent bundle .
[ srehyp ] finally , we assume @xmath9 is _ invertible _ , which is the hypothesis of _ no long - range entanglement _ , referred to as _ short - range entanglement_. the macroscopic definition given in some literature ( for example @xcite ) is that @xmath319 for all closed @xmath11-manifolds .
this _ is _ precisely invertibility at this level dimension @xmath11of the field theory .
it is not too much of a stretch to extrapolate that to invertibility at all levels .
the theorem mentioned in footnote [ foot ] supports this extrapolation .
suppose that a lie group @xmath121 acts as global symmetries of a condensed matter system .
this can include internal symmetries which act on the local hilbert space @xmath303 in .
it can also include time - reversal symmetry , since we consider time as external to space and so time - reversal symmetries fix the points of space .
we do not , however , consider symmetries which move points of space since we want to consider the theory on an arbitrary @xmath11-manifold @xmath22 .
for example , this rules out rotation and reflection symmetries of theories on euclidean space . just as we study the condensed matter system on arbitrary space manifolds @xmath22 to explore its low energy behavior ( coupling to gravity ) , to explore the effect of the global symmetry we attempt to construct an equivariant extension , as in [ subsec:2.4 ] .
recall that this means that we augment the set of fields to include a @xmath121-connection on a principal @xmath121-bundle . as mentioned after example
[ thm:9 ] there may be obstructions to constructing a @xmath121-invariant extension .
we assume that any obstruction , if it exists , can be expressed as an anomaly theory in one higher dimension .
passing to the low energy approximation , we stipulate that there is a topological anomaly theory @xmath129 and the original long - range effective theory @xmath9 is extended to be an anomalous theory @xmath160 with anomaly @xmath129 .
the topological theories @xmath129 and @xmath160 have a _
principal @xmath121-bundle _ as an additional background field the choice of connection does not appear in the classification of effective topological theories .
( see the discussion in [ subsec:2.7 ] . ) also , both theories are truncated only on manifolds of dimension @xmath320 , as in [ subsec:3.3](4 ) . as for any anomaly theory
, @xmath129 is invertible .
we make the hypothesis that for short - range entanglement the @xmath121-equivariant long - range topological theory @xmath160 is also invertible . when we come to classify anomalies in [ subsec:5.2 ] we use the considerations of [ subsec:3.3](2 ) to guide the choice of topology on the space of anomaly theories . in summary ,
then , we will assume one of two cases .
either the original long - range effective theory @xmath9 extends to a @xmath121-equivariant theory @xmath160 , or there is an anomaly @xmath129 and there is an anomalous @xmath121-equivariant extension @xmath160 . in both cases
the original set of fields @xmath38 in @xmath9 is augmented to @xmath321 there is an embedding @xmath322 which chooses the trivial @xmath121-bundle .
restriction along this map allows us to study the effective theory @xmath160 ignoring the symmetry @xmath121 .
this restriction is simply the theory @xmath9 .
a topological phase is said to be _ symmetry protected _ if this restriction @xmath9 is the trivial theory .
( notice that the anomaly theory @xmath129 is trivialized under this restriction since the original long - range effective theory @xmath9 is an absolute theory
it has no anomaly . )
in this section we recall the _ madsen - tillmann _ spectra which appear in the study of invertible topological field theories . in the notation of [ subsec:2.6 ]
the simplest of these is the spectrum ' denotes a shift , or suspension , and is present because of an unfortunate indexing convention .
what should appear on the left hand side of is the 0-space of the madsen - tillmann spectrum , which is standardly denoted ` @xmath323 ' , but we blur the notation . ] @xmath324 which is the geometric realization of the bordism category , obtained by inverting all morphisms .
the madsen - tillmann spectra were introduced in @xcite and versions of including nontrivial fields @xmath38are proved in @xcite , @xcite , @xcite .
those theorems treat the geometric realization of a topological 1-category , whereas the right hand side of is the geometric realization of an @xmath63-category .
the @xmath63 version is stated in @xcite and the techniques to prove it are most likely contained in the cited references .
we proceed to use the @xmath63 statement as the basis for our proposal in [ subsec:5.2 ] .
we give a brief introduction in [ subsec:4.1 ] ; the class notes @xcite contain much more detail . in [ subsec:4.2 ] we note the modifications to include symmetry and the modifications for theories of @xmath299-type discussed in remark [ thm:17 ] .
we also explain orientation - reversal and unitarity in this context .
pontrjagin studied smooth maps @xmath325 by choosing a regular value @xmath326 and focusing on the inverse image @xmath327 , an @xmath8-dimensional closed submanifold .
it carries an additional topological structure .
namely , for all @xmath328 the differential @xmath329 is an isomorphism from the normal space to @xmath91 at @xmath99 to a fixed vector space , the tangent space of @xmath330 at @xmath331 .
this is a _ framing _ of the normal bundle .
we emphasize that it is a _
, rather than tangential , framing .
the submanifolds for different regular values and homotopic maps are all _ framed bordant _ in the sense that for any two @xmath332 there exists a compact @xmath130-dimensional manifold @xmath333\times s^{n+q}$ ] with boundary @xmath334 ; the manifold @xmath163 carries a normal framing which agrees with that of @xmath335 and @xmath336 on the boundary .
this is the fundamental connection between bordism and homotopy theory @xcite , which was elaborated and given computational punch by thom in his phd thesis @xcite .
if we seek to understand not @xmath8-dimensional closed submanifolds of a fixed sphere @xmath337 but rather @xmath8-dimensional abstract closed framed manifolds , then a basic theorem of whitney tells that every abstract manifold embeds in some sphere , so it suffices to take @xmath338 large .
we increase @xmath338 by suspending ; large @xmath338 is realized by iterated suspension .
the suspension of is a map between the suspension of the domain and codomain spheres , and the suspension of a sphere is a sphere of dimension one greater .
a sequence of spaces related by suspension is a _ spectrum _ , and thom introduced special spectra to compute bordism groups .
the manifolds classified by these bordism groups may carry geometric structures ( framings , orientations , spin structures , ) on their _ stable normal _ bundle .
the bordism question for invertible topological field theory differs in a fundamental way .
a field theory has a definite spacetime dimension @xmath8 , and the geometric structures live on the @xmath8-dimensional tangent bundle .
so whereas thom s theory is _
stable normal _ , the madsen - tillmann spectra which arise encode _
unstable spectra are still part of stable homotopy theory
they _ are _ spectra but the dimension is fixed , not stabilized . ]
tangential _ bordism .
we sketch the basic example @xmath339 and indicate the modifications @xmath340 for oriented @xmath8-manifolds and @xmath341 for spin @xmath8-manifolds . formally , a _ prespectrum _
@xmath342 is a sequence @xmath343 of pointed spaces and pointed maps @xmath344 , where ` @xmath345 ' denotes suspension .
it is a _ spectrum _ if the induced maps @xmath346 are homeomorphisms , where ` @xmath347 ' denotes based loops .
any prespectrum has an associated spectrum , and furthermore it suffices to define @xmath348 for @xmath349 for some @xmath350 .
the simplest example , indicated above , is the sphere prespectrum with @xmath351 .
we now construct a prespectrum whose associated spectrum is @xmath339 for a fixed @xmath352 .
let @xmath353 be the grassmannian of @xmath8-dimensional subspaces of the real vector space @xmath354 .
a point of @xmath355\in gr_n(w)$ ] is an @xmath8-dimensional subspace @xmath356 .
the grassmannian is a smooth manifold , and there is a tautological rank @xmath8 _ universal subbundle _ @xmath357
whose fiber at @xmath355 $ ] is @xmath69 .
even better , there is a tautological exact sequence @xmath358 of vector bundles ; the fiber of the _ universal quotient bundle _
@xmath359 at @xmath360 is the quotient vector space @xmath361 , and the vector bundle @xmath362 has constant fiber @xmath354 . for any integer @xmath363
we define the space @xmath364 to be the _ thom space _ of the universal quotient bundle @xmath365 the thom space is obtained from the manifold @xmath366 by introducing an inner product on the vector bundle and collapsing the subspace of all vectors of norm @xmath367 to a point , where @xmath368 is any real number , as indicated in figure [ fig:6 ] .
the structure map @xmath369 is obtained by applying the thom space construction to the map @xmath370^ { } \ar[d ] _ { } & q(q+1)\ar[d]^ { } \\ gr_n({{\mathbb r}}^{n+q})\ar[r]^ { } & gr_n({{\mathbb r}}^{n+q+1 } ) } \end{gathered}\ ] ] the bottom arrow takes a subspace @xmath371 and regards it as a subspace of @xmath372 all of whose vectors have first coordinate zero . ]
more useful to us is the @xmath373 suspension @xmath374 , which is represented by the shift of this prespectrum with @xmath375 .
a map @xmath376 is represented by a pointed map @xmath377 for some @xmath338 large , so a map @xmath378 which sends the basepoint of @xmath379 to a vector of norm @xmath367 .
suppose the map is transverse to the zero section @xmath380 .
then the pullback @xmath381 is a submanifold of dimension @xmath99 .
its normal bundle is identified via @xmath382 with the normal bundle to @xmath383 in @xmath366 , which in turn is identified with the vector bundle .
but because of the degrees it is more natural to consider the virtual bundle @xmath384 whose pullback is then identified with the negative of the _ tangent _ bundle to @xmath91 , stabilized to have rank @xmath8 .
( recall that @xmath385 is the universal subbundle . )
the choice of embedding into a sphere @xmath379 disappears in the limit @xmath386 and , as in thom s bordism theory , we obtain abstract manifolds rather than embedded ones .
[ thm:53 ] the pullback of @xmath385 to @xmath91 is a rank @xmath8 bundle which is equipped with a stable isomorphism to the tangent bundle @xmath387 . more generally , for any smooth manifold @xmath57 , a map @xmath388 leads to a smooth proper map @xmath389 with @xmath390 and with a rank @xmath8 vector bundle @xmath391 equipped with a stable isomorphism with @xmath392 .
the galatius - madsen - tillmann - weiss theorem @xcite asserts that the spectrum @xmath374 classifies a bordism theory of proper _ fiber bundles _ , rather than arbitrary proper maps @xmath393 .
[ thm:19 ] there is a sequence of maps @xmath394 whose `` limit '' is the thom spectrum @xmath395 which classifies unoriented manifolds . in this precise sense the madsen - tillmann spectra approximate thom spectra . to construct the madsen - tillmann spectra @xmath340 ( @xmath341 )
use the grassmannian of oriented ( spin ) subspaces .
quite generally , a map @xmath396 of spectra represents a @xmath283-cohomology class on @xmath57 .
the @xmath283-cohomology of a madsen - tillmann spectrum @xmath397 is isomorphic to the @xmath283-cohomology of the space @xmath398 by the _
thom isomorphism_. for example , if @xmath399 then there is a thom isomorphism @xmath400\cong h^q(bso_n;{{\mathbb z}}).\ ] ] there is an orientation condition which is satisfied here because we consider the madsen - tillmann spectrum with group @xmath401 .
if instead we use @xmath402 , then we obtain cohomology with twisted coefficients : @xmath403\cong h^q(bo_n;{{\mathbb z}}^w),\ ] ] where @xmath404 is the nontrivial local system with holonomy @xmath405 around the nontrivial loop in @xmath406 .
the general form of the twisted thom isomorphism is @xmath407\cong t^{\tau + q}(bo_n),\ ] ] where @xmath408 is the twisting of @xmath283-cohomology defined by the virtual vector bundle @xmath409 .
there are similar statements for the groups @xmath401 and @xmath410 .
we refer to @xcite and references therein for a modern treatment of twisted cohomology , orientations , and the thom isomorphism .
let @xmath121 be a lie group . as explained in [ subsec:3.4 ]
the long - range effective topological theory approximating a condensed matter system with global symmetry group @xmath121 includes a @xmath121-bundle among its background fields ; see .
such a bundle can be obtained from a universal @xmath121-bundle @xmath411 by pullback . to incorporate the bundle into the madsen - tillmann spectrum @xmath339 replace with the pullback bundle @xmath412 over the cartesian product .
a map @xmath378 which is transverse to the zero section of gives a manifold @xmath413 and a @xmath121-bundle @xmath414 .
the structure maps defined from extend to incorporate the @xmath415 factor .
the spectrum so obtained is denoted @xmath416 the wedge , pronounced `` smash '' , is the appropriate product for pointed spaces ; the ` @xmath417 ' denotes a disjoint basepoint .
does have a basepoint which represents the trivial @xmath121-bundle , and we use it at the end of [ subsec:5.2 ] to implement the embedding described after in the discussion of spt phases .
the basepoint of @xmath415 is ignored in the construction of @xmath418 , the disjoint union of @xmath415 and a point .
] of course , there are similar spectra to for oriented and spin manifolds . in a @xmath11-space dimensional
theory we obtain numerical invariants only for @xmath312-manifolds which are essentially products : time is not mixed with space ; see remark [ thm:17 ] . in terms of an `` @xmath299-type '' theory
is defined on the subcategory of @xmath419 which only contains top dimensional bordisms which are fibered over 1-manifolds : @xmath420 an invertible theory of that form , if @xmath38 is empty , is a map out of the spectrum @xmath421 . in terms of the explicit prespectrum described in [ subsec:4.1 ] ,
the @xmath422-space of @xmath421 is the thom space of @xmath423 we explained in the paragraph after that manifolds in the usual @xmath312-spacetime dimensional theory have tangent bundles stabilized to rank @xmath5 . here in the space theory they are stabilized to rank @xmath11 .
so to get a bundle of rank @xmath5 we simply add an extra rank one trivial bundle @xmath424 ; the fiber represent time , which is visibly a product and does not mix with space .
the general picture of symmetries in quantum mechanics ( [ subsubsec:2.4.3 ] ) distinguishes time - preserving vs. time - reversing from linear vs. antilinear . often , of course , they are equal dichotomies . in that sense
an accounting of antilinearity suffices to account for time - reversal .
but more to the point , as we consider theories of @xmath299-type ( [ subsubsec:4.2.2 ] ) in which there is no explicit time , there is no need to track other than through antilinearity .
this is to implement the unitarity ( [ subsec:2.8 ] ) .
the geometric realization of the _ oriented _ bordism category @xmath425 is @xmath426 , if @xmath38 consists of an orientation .
thus orientation - reversal induces an involution on @xmath427 . introducing the notation @xmath428 for the thom spectrum of the virtual vector bundle @xmath429
, we can summarize as @xmath430 where @xmath431 is the universal rank @xmath8 bundle .
orientation - reversal on rank @xmath8 bundles is represented as the deck transformation of the double cover @xmath432 there is an induced involution on the thom spectrum . in a unitary theory orientation - reversal maps to complex conjugation .
complex conjugation on @xmath84 corresponds to the involution @xmath433 on @xmath434 via exponentiation .
let @xmath435 be the target of an invertible unitary theory @xmath436 .
this represents a twisted cohomology class of @xmath374 .
the twisted thom isomorphism theorem identifies this twisted cohomology group with @xmath437 where the coefficients are twisted by the complex conjugation action ( no sign ) .
note the short exact sequence @xmath438 which is equivariant for the @xmath439-action on @xmath219 and the action on the other two groups .
so when we map to @xmath440 or @xmath441 we use the @xmath439-action in place of the action on @xmath442 and @xmath113 .
we remark that the twisted thom isomorphism involves more complicated twistings for @xmath113 . an analogous construction works if we replace @xmath401 with @xmath410 and @xmath402 with @xmath443 . however , there are two choices for @xmath443 .
if we view @xmath443 as embedded in the real clifford algebra @xmath444 , then the choices depend on the sign in the relation @xmath445 which defines @xmath444 ; see @xcite .
so it seems there are two distinct notions of unitarity for spin theories .
we will simply write ` @xmath446 ' and not investigate the distinction further in this paper .
we now sketch precisely how we implement the @xmath439-involution on spectra .
[ thm:48 ] let @xmath447 denote the real hopf line bundle .
the thom spectrum @xmath448 of the reduced hopf bundle is a bundle , or sheaf , of spectra over @xmath449 whose typical fiber is the sphere spectrum .
the holonomy around the nontrivial loop acts on the homotopy groups @xmath450 of the sphere as multiplication by @xmath405 . in other words , for each @xmath338 the homotopy groups make a local system over @xmath449 with fiber @xmath451 and holonomy @xmath405 .
any other spectrum @xmath283 is a module over @xmath452 and there is an induced bundle of spectra over @xmath449 with fiber @xmath283 ; the holonomy acts on @xmath453 as multiplication by @xmath405 .
see @xcite for a precise definitions , statements , and proofs of these assertions about bundles of spectra and the twisted thom isomorphism .
see [ subsec:2.8 ] for a general discussion of unitarity , and [ subsubsec:4.2.4 ] for its implementation in invertible theories . here
we compute and interpret groups of low - dimensional oriented unitary theories , using .
the group of @xmath12 spacetime dimensional oriented theories with eilenberg - maclane target is @xmath454\cong h^1(bso_1;{{{\mathbb c}}/{{\mathbb z}}})=0;\ ] ] there is only the trivial theory @xmath9 which assigns the trivial line @xmath455 to the oriented point @xmath169 and the number @xmath456 to the oriented circle .
this theory is clearly unitariz__able__. a unitary structure is data , and according to the group of isomorphism classes of unitarity data is @xmath457 where the local system @xmath458 has holonomy @xmath405 around the nontrivial loop . with a single cell in each dimension ,
trivialize the local system over each cell , and derive the cochain complex @xmath459 in which @xmath286 is complex conjugation ; ( 2 ) use the short exact coefficient sequence where @xmath68 , @xmath434 are local systems with holonomy @xmath405 and @xmath219 is untwisted . ]
the two equivalence classes of unitarity data on @xmath9 are easily explained .
writing @xmath460 the unitarity data provides an isomorphism @xmath461 .
the oriented interval with both endpoints incoming is a bordism @xmath462 and applying @xmath9 and the unitarity isomorphism we obtain a nondegenerate hermitian form @xmath463 . such a form is either positive or negative , which accounts for .
[ thm:49 ] the _ positivity _ condition in a unitary theory means we should exclude the negative form , so only allow a unique isomorphism class of unitarity data .
the group only implements the correct action of orientation - reversal , not the positivity , but i do not know how to pick out the subgroup corresponding to positive unitarity data in higher dimensions . as a further illustration we consider @xmath108 spacetime dimensional oriented theories with eilenberg - maclane target .
these theories are discussed in example [ thm:4 ] and example [ thm:13 ] .
the group of oriented theories is @xmath464\cong h^2(bso_2;{{{\mathbb c}}/{{\mathbb z}}})\cong { { { \mathbb c}}/{{\mathbb z}}},\ ] ] and the theory @xmath465 corresponding to @xmath466 has @xmath467 for a closed oriented 2-manifold @xmath19 , where @xmath468 .
since @xmath19 has an orientation - reversing involution , only those theories with @xmath469 real can possibly be unitarizable , which forces @xmath470 . that is consistent with the computation of for @xmath108 : @xmath471 together with the computation of the map @xmath472 which includes @xmath473 .
for @xmath474 we have with @xmath475 , but since the euler number is even the numerical invariants are positive .
the line @xmath476 attached to the oriented circle has a real structure , since @xmath313 has an orientation - reversing involution ( reflection ) , and the cylinder with both boundaries incoming gives a nondegenerate real symmetric bilinear form @xmath477 .
glue 2-disks to each incoming boundary component to deduce that @xmath478 is positive .
the conclusion is that all theories with parameter @xmath479 are uniquely unitarizable .
there is one more preliminary before we can state our proposed topological invariant of short - range entangled ( sre ) phases : we must specify our choice of target spectrum for classifying long - range effective topological theories . recall that in general to define a topological field theory we need to specify a target higher category @xmath64 , but for an invertible theory we need the much weaker information of the sub - groupoid @xmath480 of invertibles ; see .
while concrete arguments give information about the highest few homotopy groups of @xmath221 , we can only guess at the structure lower down , which is increasingly relevant as the dimension of the theory increases . in the general fermionic case
we take a universal choice , the _ brown - comenetz dual to the sphere spectrum_. in the bosonic case we have solid arguments to determine the `` top part '' of the spectrum , but after that only sparse data points . for that reason
we take the same target spectrum as in the fermionic case
after all , theories with only bosons are special cases of general theories though in space dimension @xmath3 the `` top part '' , an _ eilenberg - maclane spectrum _ , is all that is relevant and so we use it instead .
we explain these choices in [ subsec:5.1 ] .
our main proposal is stated precisely in [ subsec:5.2 ] . in
[ subsec:5.3 ] we work out the relationship to the group cohomology classifications in the literature @xcite , @xcite .
the discussions in example [ thm:13 ] and especially [ subsec:3.3](2 ) are relevant here .
let @xmath50 be an abelian group , which might be discrete or have a topology .
( in the latter case there is an additional hypothesis : @xmath50 is locally compact . )
examples : @xmath481 , @xmath219 , @xmath482 .
there are two notions of dual group we might consider .
the first is the _ pontrjagin dual _ , which is the group of continuous homomorphisms @xmath483 .
the pontrjagin dual of @xmath481 is isomorphic to @xmath481 , the pontrjagin dual of @xmath219 is isomorphic to @xmath482 , and the pontrjagin dual of @xmath482 is isomorphic to @xmath219 . on the other hand , we can consider an integral dual @xmath484 .
however , the naive interpretation is not so well behaved .
for example , if @xmath485 there are no nonzero homomorphisms @xmath486 .
the resolution is to consider @xmath487 in the _ derived _ sense , which means that we replace @xmath50 by a free chain complex whose homology in degree 0 is @xmath50 and then compute @xmath488 .
the result is called @xmath489 and is the `` correct '' integral dual . for @xmath485
the only nonzero group is in degree @xmath405 , whereas for @xmath490 the only nonzero group is in degree 0 .
both notions of duality exist in the world of spectra ; see @xcite , ( * ? ? ?
* appendix b ) , ( * ? ? ?
* appendix b ) , @xcite for precise definitions and discussion .. recall that a spectrum @xmath342 has an associated @xmath219-graded abelian group @xmath453 . for the sphere spectrum @xmath452
the first several homotopy groups are @xmath491 the analog of the pontrjagin dual is the _ brown - comenetz dual_. for the sphere we denote it as @xmath113 .
( we replace @xmath482 with @xmath434 ; in topology it is often @xmath492 instead . )
its homotopy groups are the pontrjagin dual groups to : @xmath493 note carefully the indexing difference on the left hand side between and .
the analog of the integral dual is the _ anderson dual_. for the sphere we denote it as @xmath441 .
its homotopy groups are the integral dual to ; the torsion groups are degree shifted : @xmath494 there is a fiber sequence @xmath495 and a corresponding long exact sequence of homotopy groups .
note @xmath496 is an eilenberg - maclane spectrum with a single nonzero homotopy group .
these spectra enjoy universal properties which make them universal targets for invertible field theories . recall that the notation ` @xmath497 $ ] ' is used for the abelian group of homotopy classes of maps @xmath498 between spectra .
if @xmath19 is any spectrum , then @xmath499 = i{{{\mathbb c}}/{{\mathbb z}}}^n(x)\cong \operatorname{hom}(\pi _ nx,{{{\mathbb c}}/{{\mathbb z}}}).\ ] ] ( to compute the @xmath113 cohomology of a space @xmath22 , set @xmath500 the suspension spectrum , so @xmath501 is the @xmath373 _ stable _ homotopy group of @xmath22 . )
thus an invertible @xmath8-spacetime dimensional theory whose values on closed @xmath8-manifolds are numbers in @xmath434 pushes uniquely to a field theory with values in @xmath113 , and any theory with target @xmath113 is determined by its numerical values on @xmath8-manifolds . for maps from any spectrum @xmath19 into @xmath441 there is a short exact sequence @xmath502 which is split , but not canonically .
we continue with theories @xmath9 defined in @xmath11-space dimensions and assume first that @xmath9 is a bosonic theory .
the target is a @xmath312-category @xmath503 , but since @xmath9 is invertible we need only the groupoid @xmath504 .
standard quantum mechanics dictates that @xmath20 is a complex vector space for any closed @xmath11-manifold @xmath22 , and since @xmath9 is invertible it must be 1-dimensional .
furthermore , a diffeomorphism of @xmath22 acts as an invertible map @xmath505 , which is multiplication by a scalar in - manifold need not . ] @xmath84 .
thus the truncation @xmath506 of the target groupoid to the top two levels is the groupoid @xmath244 of complex lines . a crucial point , discussed in [ subsec:3.3](2 ) , is that we use the _ continuous _ topology on the morphism spaces , in particular on @xmath84
this gives information about some homotopy groups of @xmath507 : @xmath508 the first equality expresses the fact that any two lines are isomorphic ; the latter two that @xmath84 is connected with infinite cyclic fundamental group .
furthermore , @xmath509 for @xmath510 since there are no non - identity @xmath25-morphisms for @xmath511 .
it remains to determine the lower homotopy groups .
we have quite solid information about the next homotopy group down , as we know the nature of the higher categorical invariant usually attached to manifolds of dimension @xmath512 by the theory @xmath9 .
for example , a standard choice is to assign a linear category to a closed @xmath513-manifold , and since all invertible linear categories are isomorphic deduce @xmath514 an alternative choice is to assign an invertible complex algebra to a closed @xmath513-manifold , and the triviality of the brauer group of @xmath68 leads to the same conclusion . at this stage
we have reproduced the first part of example [ thm:13 ] , in which the target is a 2-category and there are no further homotopy groups .
note that and amount to an eilenberg maclane spectrum @xmath515 : a single nonzero homotopy group . this is the full story for @xmath516 .
what is perhaps surprising is that we postulate a nonzero homotopy group at the next level down : @xmath517 one rationale for is the 4-dimensional integral invertible oriented extended topological field theory @xmath518 for which the integer invariant @xmath519 of a closed oriented 4-manifold @xmath354 is its signature @xmath520 . in its analytic incarnation
this invariant involves only differential forms on @xmath354 , not spinor fields , so in that sense is bosonic . the theory factors through a map @xmath521 which can be viewed as the universal symbol @xcite of the signature operator in dimension 4 .
the relevant stretch of homotopy groups of the target spectrum in is @xmath522 this theory does _ not _ factor through @xmath523 , so the bottom @xmath219 in can not be replaced by @xmath524 .
another piece of evidence for at least a nonzero homotopy group in this spot ( @xmath525 ) comes from _ conformal nets_. these are a possible target for 3-dimensional theories with numerical invariants in @xmath434 , and conjecturally not all invertible conformal nets are isomorphic .
( see @xcite for a discussion of conformal nets in this context . )
we do not have any information about lower homotopy groups .
so we could take the shifted postnikov truncation @xmath526 as a reasonable choice of target spectrum @xmath507 .
but the @xmath219 at the bottom of the spectrum is overkill a smaller torsion group will do and so instead for theories in dimension @xmath527 we use the universal choice , the anderson dual of the sphere .
[ thm:20 ] the target spectrum for classifying long - range effective theories of a @xmath11-space dimensional bosonic system is @xmath528 for @xmath3 and @xmath529 for @xmath527 .
[ thm:21 ] the classification of anomalies has similar target spectra , but as indicated in [ subsec:3.3](2 ) we use instead the discrete topology on @xmath84 .
thus for @xmath527 the target which classifies anomaly theories is @xmath530 ; for @xmath3 it is @xmath531 .
the hypothesis for theories with fermions is different .
namely , the dichotomy between bosonic and fermionic states in quantum mechanical systems is encoded by stipulating that the quantum hilbert space be @xmath532-graded : states with even grading are bosonic and states with odd grading are fermionic . that persists in the long - range effective theory : vacua are either bosonic or fermionic .
so the complex lines @xmath20 in an invertible long - rang theory are either even or odd .
the existence of distinct isomorphism classes of @xmath532-graded lines modifies : @xmath533 the @xmath25-invariant is nonzero ; it is the composition @xmath534 of the integer bockstein and the steenrod square . as in the bosonic case higher homotopy groups vanish .
but now we expect many nontrivial lower homotopy groups .
for example , we expect to be replaced by @xmath535 since the super brauer group of invertible @xmath532-graded complex algebras has two elements ( represented by even and odd complex clifford algebras ) .
the idea that modules over clifford algebras may be used as the state space of a quantum system is familiar in condensed matter theory ; see also @xcite . from the super brauer category
we compute how the three eilenberg - maclane spectra fit together , though we do not do so here .
we can gather information about @xmath536 by arguments analogous to those in [ subsubsec:5.1.1 ] .
for example , there is an invertible theory @xmath537 which assigns the @xmath538-genus to a closed spin 4-manifold .
( the universal symbol is quaternionic , which explains the absence of shift in @xmath539 . )
this suggests that @xmath540 .
invertible fermionic conformal nets are another clue , but the only knowledge is conjectural .
therefore , by fiat really , we make a _ universal _ choice for the target spectrum , namely the anderson spectrum @xmath541 .
notice that its first four homotopy groups agree with and and more precisely the postnikov truncations are equivalent .
[ thm:22 ] the target spectrum for classifying long - range effective theories of a @xmath11-space dimensional fermionic system is @xmath542 .
[ thm:23 ] we use the shift @xmath543 as a target to classify anomaly theories . as explained in [ subsubsec:2.4.3 ] antilinear symmetries lead to anomalous field theories after gauging . for invertible theories we account for the anomaly using twisted cohomology .
we implement complex conjugation on the target spectra by the universal @xmath439-action defined in construction [ thm:48 ] .
consider , for example , the eilenberg - maclane spectrum @xmath440 . in degree 1 a cohomology class
is represented by a map to @xmath84 , and the @xmath439-action is complex conjugation on @xmath84 .
the classifying space for a degree 1 class is @xmath84 , and construction [ thm:48 ] gives a fiber bundle ( not principal ! ) with fiber @xmath84 over @xmath449 whose holonomy acts as complex conjugation on @xmath84 . in degree 2 ,
as just explained , a cohomology class is represented by a complex line bundle and this action is complex conjugation on complex line bundles . in degree 3
there is a similar story with bundles of complex algebras , and it is reasonable to extend this picture to all degrees .
the same story applies to @xmath442 , which one can view as flat elements in @xmath440 ( with a degree shift ) .
for example , an element in degree 0 in @xmath442 is a locally constant map to @xmath84 .
similar considerations apply to other target spectra .
the top nonzero homotopy group of @xmath541 ( respectively @xmath544 ) is the same as that of @xmath440 ( respectively @xmath442 ) , and the @xmath439-action is trivial on the @xmath532 homotopy groups in and is trivial .
our tentative grasp on lower homotopy groups puts further justification beyond reach .
we propose a home for long - range effective topological field theories of gapped systems with short - range entanglement .
we do not know if the map from the microscopic phases to the deformation classes of field theories is either injective or surjective .
regardless , the evidence presented in the remainder of the paper suggests that it is a very effective invariant .
assume the theory is @xmath11-space dimensional and has a global symmetry group @xmath121 , which is a lie group equipped with a smooth homomorphism @xmath545 which encodes linearity vs. antilinearity : an element @xmath171 with @xmath546 acts linearly and an element @xmath171 with @xmath173 acts antilinearly .
recall our assumption , stated before , that there is a @xmath121-equivariant extension of the long - range effective theory ( _ non - anomalous case _ ) or that there is an anomaly and an anomalous extension ( _ anomalous case _ ) .
the hypotheses underlying the proposal are stated in [ subsec:3.3 ] and [ subsec:5.1 ] .
we use the twisted thom isomorphism ( and its variations for @xmath547 and @xmath548 replacing @xmath549 ) . according to hypothesis [ thm:20 ]
the target spectrum is @xmath550 for @xmath3 and @xmath529 for @xmath527 .
we use the notation ` @xmath551 ' for this target spectrum . note that @xmath552 is a ring spectrum .
let @xmath553 denote the thom twisting of the ring spectrum @xmath552 associated to the virtual bundle ( see ) @xmath554 and @xmath555 the analogous thom twisting for @xmath556 in place of @xmath557 .
let @xmath558 denote the @xmath439-twist of @xmath552 ( construction [ thm:48 ] ) associated to the double cover @xmath559 .
the homomorphism determines a @xmath439-twist of @xmath552 associated to the double cover is identically @xmath560 , then @xmath561 and @xmath562 is trivial . ]
@xmath563 , where @xmath564 ; we denote it @xmath565 .
degree shifts are thom twistings of trivial bundles , whence sums of twistings and degrees are defined .
[ thm:24 ] short - range entangled phases of a @xmath11-space dimensional bosonic theory with global symmetry group @xmath121 map to the abelian group @xmath566 in the non - anomalous case .
the unitarizable theories lie in the image of the map we settle for a weaker formulation . ] @xmath567 anomalies are classified by the abelian group @xmath568 and the anomalous theories with fixed anomaly map to a torsor for the abelian group . an anomaly theory has spacetime dimension @xmath569 , but here is only defined on manifolds of dimension @xmath320 ; this explains the degrees in .
the last assertion is that any two choices of anomalous theories with the same anomaly are related by tensoring with a non - anomalous theory .
the proposal for fermionic theories is similar : we simply swap out the eilenberg - maclane spectra we used in the @xmath3 bosonic case for the anderson and brown - comenetz spectra ( [ subsubsec:5.1.2 ] ) and assume all manifolds are spin in addition to being oriented .
the corresponding thom twistings of are denoted @xmath570 and @xmath571 ; the @xmath439-twisting of @xmath541 associated to the double cover @xmath559 is @xmath572 ; and the @xmath439-twisting of @xmath541 associated to the double cover @xmath563 is @xmath573 .
[ thm:25 ] short - range entangled phases of a @xmath11-space dimensional fermionic theory with global symmetry group @xmath121 map to the abelian group @xmath574 in the non - anomalous case .
the unitarizable theories lie in the image of the map @xmath575 anomalies are classified by the abelian group @xmath576 and the anomalous theories with fixed anomaly map to a torsor for the abelian group .
now we address the question of _ symmetry protected topological _
( spt ) phases .
the `` symmetry protection '' means that the effective topological field theory @xmath9 is trivial when the symmetry is ignored . as explained at the end of [ subsec:3.4 ] this means that the @xmath121-extension @xmath160 is trivial when restricted to the trivial @xmath121-bundle .
the basepoint of @xmath415 determines an embedding @xmath577 for any spectrum @xmath19 , and so a restriction map @xmath578\longrightarrow [ x , x']\ ] ] for any spectrum @xmath579 . when @xmath19 is a madsen - tillmann spectrum we can rewrite using the twisted thom isomorphism ; then the map is pullback along the inclusion @xmath580 defined by the basepoint of @xmath415 .
[ thm:26 ] a. symmetry protected topological phases of a @xmath11-space dimensional bosonic theory with global symmetry group @xmath121 map to the kernel @xmath581 of the indicated restriction map constructed from .
b. symmetry protected topological phases of a @xmath11-space dimensional fermionic theory with global symmetry group @xmath121 map to the kernel @xmath582 of the indicated restriction map . from the definition of @xmath552 at the beginning of [ subsubsec:5.2.1 ] we construct a map of spectra @xmath583 it induces the second map in the composition @xmath584 the first is induced from the projection @xmath585 . here
@xmath586 is the local system defined by .
note that ordinary cohomology is oriented for oriented vector bundles , which explains why the twisting @xmath553 is trivialized when restricted under .
the homomorphism maps the group cohomology phases discussed , where @xmath587 has its continuous topology . ] in @xcite to @xmath588 .
furthermore , the image of lies in the subgroup @xmath589 of symmetry protected phases ; see .
[ thm:46 ] if @xmath121 is discrete , then the topological cohomology of @xmath415 is isomorphic to the group cohomology of the group @xmath121 .
more generally , if @xmath121 is a ( finite dimensional ) lie group , then a theorem of d. wigner @xcite states that the topological cohomology of @xmath415 with coefficients in a discrete @xmath121-module is isomorphic to the borel cohomology of @xmath121 ; see @xcite for more on group cohomology . for theories with fermions gu and
wen @xcite introduced a group `` super '' cohomology theory .
in fact , it can be identified with a certain generalized cohomology theory of the classifying space @xmath415 . this generalized cohomology theory , which we simply call @xmath590 ,
had already appeared in at least a few contexts in theoretical physics : ( 1 ) in spin chern - simons theories @xcite and ( 2 ) in qcd , in the wess - zumino term of the long - range effective theory of pions @xcite .
the spectrum @xmath590 has two nonzero homotopy groups : @xmath591 the @xmath25-invariant which relates them is nonzero .
we defer to @xcite for generalities on this cohomology theory .
the two nontrivial homotopy groups in @xmath590 occur in , shifted up by degree @xmath569 , and as the @xmath25-invariant match there is a map @xmath592 of spectra .
the theory @xmath590 is oriented for spin bundles , as proved in ( * ? ? ?
* proposition 4.4 ) .
therefore , there is a homomorphism so has vanishing homotopy group in degrees @xmath593 .
the spin orientation of @xmath590 explains why @xmath570 does not appear in the domain of .
] @xmath594 the projection @xmath595 induces an inclusion @xmath596 which , after composition with , induces a homomorphism of the group `` super '' cohomology into @xmath597 , as expected . note that the image of @xmath598 in @xmath597 lies in the subgroup @xmath599 of symmetry protected phases
we illustrate how the proposed invariants of gapped short - range entangled ( sre ) phases in [ subsec:5.2 ] detect phases not covered by the group cohomology classification discussed in [ subsec:5.3 ] .
we organize the discussion by space dimension @xmath11 and by whether or not the theory includes fermions .
the examples treated here are non - anomalous .
clearly there are many more computations and analyses which can be carried out .
this is the one case in which there is nothing beyond the group cohomology classification .
there are two reasons for this : ( 1 ) the group @xmath600 which governs the spatial tangential structure is trivial , and ( 2 ) for @xmath3 we have @xmath601 . more formally , from and we deduce @xmath602 since the shifted madsen - tillmann spectra are connective ( in this case the relevant spectrum is @xmath603 )
, we can replace the codomain @xmath604 of an invertible topological field theory by its connective cover .
that connective cover is a module for @xmath605-theory , which is connective real @xmath0-theory .
( see @xcite , for example , where that connective cover is the theory called ` @xmath606 ' . ) in particular , the connective cover is spin - oriented so the thom twisting is trivial .
hence @xmath607 the subgroup coming from the @xmath415 factor is @xmath608 [ thm:30 ] this already goes beyond the group `` super '' cohomology theory @xmath590 , since @xmath590 has two nonzero cohomology groups , whereas the truncation of @xmath541 we are using here has a third nonzero homotopy group @xmath609 .
consider the special case @xmath610 with @xmath103 nontrivial ; this is the case of a time - reversal symmetry which squares to the identity . then is cyclic of order 8 : @xmath611 one proof of is ( * ? ? ?
* theorem 3.13 ) .
[ thm:27 ] one interpretation of the left hand side of is the group of degree shifts of @xmath539-theory , which is the brauer group of _ real _ @xmath532-graded central simple algebras .
this is surely very closely related to the classification in @xcite . because the group @xmath612 is nontrivial
, there are additional sre phases not captured by .
one example is the truncation to 1-space dimension of the 2-spacetime dimensional `` arf theory '' .
the invariant of a 2-dimensional closed spin manifold is @xmath613 according to the arf invariant of the quadratic form defining the spin structure : even spin structures have invariant @xmath560 and odd spin structures have invariant @xmath405 .
the invariants on 1- and 0-dimensional manifolds are also explicit .
the invariant of @xmath313 is the trivial even line for the bounding spin structure and the odd line for the nonbounding spin structure .
the invariant of @xmath169 with the standard spin structure is the complex clifford algebra @xmath614 .
( for simplicity we take the target 2-category @xmath64 of the theory to be algebras - bimodules - intertwiners .
equivalently , we can assign to @xmath169 the category of @xmath532-graded @xmath614-modules . ) for more on the arf theory , and a beautiful geometric application , see @xcite .
it appears that the arf theory is realized as the long - range effective topological theory of the majorana chain @xcite in its nontrivial phase .
for example , the description of the effective theory ( * ? ? ?
* ( 15 ) ) does assign the clifford algebra @xmath614 to a point . in the remainder of this section
we illustrate some relevant computational techniques .
more elaborate techniques are needed in higher dimensions , as we illustrate in the appendix .
the arf theory is predicted by @xmath615 in fact , since the arf theory extends to a 2-spacetime dimensional theory , we can detect it in the group @xmath616 .
it is illuminating to first compute @xmath617 the first line is the thom isomorphism ; the third the defining property of @xmath544 ; the fourth the general fact @xmath618 ; and the last line the results @xmath619 ( computed in @xcite , for example ) and @xmath620 .
this is the set of 2-dimensional invertible spin theories with target @xmath544 .
it includes the family of euler theories ( example [ thm:4 ] ) parametrized by @xmath621 and the arf theory .
the group of path components of is @xmath622 one computation of this group uses the split short exact sequence @xmath623 and the computations @xmath624 and @xmath625 .
] the atiyah - hirzebruch spectral sequence provides another means to compute the generalized cohomology groups and .
the relevant portion of the @xmath626 page of the spectral sequence @xmath627 for computing is shown in figure [ fig:7 ] .
the differentials emanating from the initial column vanish , as can be seen from the splitting @xmath628 of the projection to a point .
that reasoning also applies to the spectral sequence @xmath629 a portion of which is shown in figure [ fig:8 ] , and it also applies to prove that the short exact sequence @xmath630 splits .
( one reads off from the @xmath631 page . ) ] to investigate sre phases at the other extreme from those detected by group cohomology , we set @xmath632 to be the trivial group .
then necessarily @xmath633 is the trivial homomorphism .
unwinding proposal [ thm:24 ] we have @xmath634,\ ] ] the group of @xmath299-type oriented theories with target the anderson dual of the sphere .
the following computations are carried out in appendix [ sec:8 ] .
recall that the group @xmath635 is thom s oriented bordism group of 4-manifolds , which is isomorphic to @xmath219 via homomorphism which attaches to each closed oriented 4-manifold @xmath91 its signature @xmath636 .
[ thm:50 ] 1 .
2 . @xmath638 and the composition @xmath639 maps the generators to @xmath640 .
3 . @xmath641 .
4 . @xmath642 and the composition @xmath643 maps the generators to @xmath644 .
we interpret these computations using .
first , ( i ) and ( ii ) imply @xmath645\cong { { \mathbb z}}.\ ] ] furthermore , the generating field theory extends to a @xmath219-valued invertible 4-dimensional oriented theory whose numerical invariant is _ 4 times _ the signature .
assertions ( iii ) and ( iv ) together imply @xmath646\cong { { \mathbb z}}\ ] ] and the generating field theory extends to a @xmath219-valued invertible 4-dimensional theory whose numerical invariant is _ 2 times _ the signature . a 2-dimensional theory which generates the group does not extend to 3 dimensions , and a 3-dimensional theory which generates the group does not extend to 4 dimensions .
we relate these factors to known geometric facts about @xmath647-invariants and determinant lines .
the interpretation of the homotopical computations , is not obvious at first glance ; indeed that mystery was the motivation for [ subsec:2.7 ] .
furthermore , the example discussed there in particular , the contractible choice replacing @xmath250 with @xmath240has a clear analog in our current situation .
if we include a riemannian metric as a field , then there is a 3-spacetime dimensional theory whose invariant of a closed oriented riemannian 3-manifold is the exponentiated atiyah - patodi - singer @xmath648-invariant @xcite .
the invariant is @xmath649 , where @xmath650 . ] as in remark [ thm:28 ] there is a related _ 4-dimensional _ theory ( called @xmath651 below ) whose value on a closed oriented 4-manifold
@xmath91 is a multiple of the signature @xmath636 .
if we use the @xmath647-invariant associated to the signature operator , then the multiple is 1 .
but we can use instead the @xmath647-invariant associated to the self - duality operator ; the corresponding boundary operator @xmath652 is @xmath653 that of the signature operator .
( this @xmath647-invariant appears in quantum chern - simons theory ( * ? ? ? * ( 1.27 ) ) . )
this theory represents a generator of ; the multiple of the signature is 1/2 .
the invariants of 2-manifolds are determinant lines , and the determinant line of the 2-dimensional signature operator on a closed oriented riemannian manifold has a natural @xmath2 root : the determinant line of the @xmath654-operator .
determinant lines of @xmath654 provide a generator of ; the multiple of the signature is 1/4 .
[ thm:54 ] the discussion in [ subsec:4.1 ] , especially remark [ thm:53 ] , and also remark [ thm:17 ] are relevant here .
a theory classified by gives integer invariants of 4-manifolds whose stable tangent bundle can be represented by a rank 3 vector bundle .
such manifolds have even signature .
( for example , the @xmath2 stiefel - whitney number vanishes .
it is the reduction modulo 2 of the euler number which is equal to the signature modulo 2 . )
an example of such a 4-manifold is the mapping cylinder of a 3-manifold , which fibers over the circle , but then the signature vanishes .
a nontrivial example is the connected sum @xmath655 , which has signature 2 and euler number 0 .
the vanishing euler number implies that @xmath91 admits a nonzero vector field , which splits an oriented line bundle off of @xmath387 .
similarly , 4 divides the signature of a compact oriented 4-manifold whose stable tangent bundle is 2-dimensional . in this case
there are nontrivial examples which are fiber bundles ; the base and fiber are both compact oriented 2-manifolds .
the first examples are due to atiyah @xcite ; an example with signature exactly 4 is constructed in ( * ? ? ?
* theorem 1 ) .
we now argue that the sre phase referred to in the literature as `` kitaev s @xmath1 phase '' or `` kitaev s @xmath1 state '' ( see @xcite , @xcite , @xcite ) has the field theory whose invariant is exactly the signature as its low energy approximation .
to make a first connection to @xmath1 chern - simons , we recall that the gravitational chern - simons invariant enters into the quantization of classical chern - simons theory as a counterterm ( * ? ? ?
* ( 2.20 ) ) .
its appearance means that in general quantum chern - simons theory is _ anomalous _ as an oriented theory .
the anomaly is an invertible 4-dimensional theory @xmath656 whose invariant on a closed oriented 4-manifold
@xmath91 is @xmath657 the anomaly depends only on the mod 8 reduction @xmath658 of the _ chiral central charge _
@xmath659 of the corresponding 2-dimensional chiral wess - zumino - witten model .
[ thm:29 ] walker s approach @xcite to quantum chern - simons theory uses bounding 4-manifolds to control the framing dependence . in joint work with constantin teleman ( so far unpublished )
we prove that a modular tensor category is invertible as an object in the 4-category of braided tensor categories , and we use it to define an invertible oriented 4-dimensional topological field theory which is precisely @xmath660 .
note that the modular tensor category determines @xmath661see ( * ? ? ?
* ( 172 ) ) , for example but it does not determine @xmath659 .
the usual approach to quantum chern - simons theory in the mathematics literature is to lift to a theory of manifolds with a @xmath662-structure .
a @xmath663-structure is a trivialization of @xmath663 : an orientation .
a @xmath53-structure is similar @xcite , but its geometric avatars are not as simple as an orientation .
for example , a @xmath53-structure on a 3-manifold can be given by a `` 2-framing '' @xcite . in this way one obtains chern - simons as an extended theory of 1- , 2- , and 3-dimensional manifolds , but to do so one needs to lift @xmath664 to @xmath665 .
of course , given @xmath659 there is a preferred choice , but starting from a modular tensor category there are 3 choices .
we observe that given a chiral central charge @xmath659 there is a 4-dimensional invertible theory @xmath666 whose invariant on a closed oriented 4-manifold
@xmath91 is the real number @xmath667 the anomaly theory @xmath660 in is obtained by composing with the map @xmath668 induced by the exponential map @xmath669 ; see .
if @xmath670 for some @xmath671 , then factors through an integral theory @xmath672 these integral topological theories are not part of the usual quantum chern - simons theory : only the exponential of occurs ( as the `` framing anomaly '' theory ) . for the theories in the framing anomaly
is trivial .
what does occur in chern - simons is the invertible metric 3-dimensional theory whose partition function is the exponentiated @xmath647-invariant to a suitable power @xcite . @xmath1 chern - simons at level 1 , or chern - simons theory for the maximal torus of @xmath1 ( with its cartan matrix specifying the level , or `` @xmath0-matrix '' ) , has chiral central charge @xmath673 .
the @xmath647-invariant which occurs is associated to the signature operator , and this is the class of theories we associate to kitaev s @xmath1 phase .
[ thm:55 ] proposition [ thm:50 ] implies that there are additional possibilities for an @xmath299-type 2-dimensional theory : a `` @xmath2 root '' of the effective theory of kitaev s @xmath1 phase .
such theories are associated with chiral central charge @xmath674 , and in general the 2-dimensional @xmath299-type theories are associated with chiral central charge divisible by 2 .
this matches the conformal anomaly in 2-dimensional conformal field theory
see ( * ? ? ?
* ( 5.9 ) ) , for example : if the chiral central charge is not divisible by 2 , then a @xmath53-structure is needed to define the theory .
even if we require the theory to extend to 3-manifolds , there is still the possibility of dividing by 2 , so having chiral central charge divisible by 4 .
so it appears that our proposal allows for more effective topological theories than have been seen so far by sre phases .
continuing with @xmath4 space dimensional theories , we give an example to illustrate the unitarizability restriction in .
now we allow a global symmetry group @xmath121 .
we focus on the kunneth component @xmath675 the group @xmath676 is infinite cyclic with generator the euler class @xmath677
. for simplicity let @xmath121 be finite and @xmath103 the trivial homomorphism .
then @xmath678 is isomorphic to the group of abelian characters @xmath679 .
fix one and let @xmath680 be the corresponding class .
let @xmath9 be the theory which corresponds to @xmath681 in .
we remark in passing that @xmath9 does _ not _ extend to a 3-spacetime dimensional theory : the euler class @xmath677 is not the restriction of a class in @xmath682 . the main point : this theory is _ not _ unitarizable . to see this it suffices to restrict along @xmath683 in , since we are trying to hit @xmath681 which lies in ordinary cohomology .
there is an isomorphism @xmath684 of twistings in the domain of , so the relevant group is @xmath685 and the relevant kunneth component is @xmath686 .
the euler class @xmath677 does _ _ not _ _ with twisted coefficients , but that is not relevant here . ]
drop to a class in @xmath687 .
[ thm:51 ] it is instructive to compute something nontrivial in the theory @xmath9 . let @xmath22 be a closed oriented 2-manifold and @xmath688 a principal @xmath121-bundle .
then @xmath689 is a complex line .
suppose @xmath690 is an automorphism of @xmath688 , so a map @xmath691^{\varphi } \ar[d ] _ { } & p\ar[d]^ { } \\ y\ar[r]^{\bar\varphi } & y } \end{gathered}\ ] ] of principal @xmath121-bundles covering an orientation - preserving diffeomorphism of @xmath22 .
gluing the ends of @xmath692\times p\to[0,1]\times y$ ] using @xmath690 we obtain a principal @xmath121-bundle @xmath693 over the mapping cylinder @xmath694 .
note that the mapping cylinder is a 3-manifold which is the total space of a fiber bundle @xmath695 with typical fiber @xmath22 .
the rank 2 relative tangent bundle @xmath696 is oriented so has an euler class @xmath697 .
the @xmath121-bundle @xmath693 has a characteristic class @xmath698
. then the action of @xmath690 on the line @xmath689 is multiplication by @xmath699 \rangle\in { { { \mathbb c}}^\times } .\ ] ] in we pair the cup product of the characteristic classes with the fundamental class of the oriented 3-manifold @xmath694 . a special case of note :
@xmath700 is the identity diffeomorphism , @xmath688 is the trivial bundle , and the gauge transformation @xmath690 is given by an element @xmath171 ( assuming @xmath22 is connected ) . then reduces to @xmath701 , where @xmath702 is the euler number .
set @xmath610 and @xmath703 the identity map .
the group cohomology captures a subgroup of the group of sre phases : @xmath704 as appears in @xcite .
another nontrivial sre phase of order 2 was introduced in @xcite where it was emphasized that this goes beyond the group theory computation .
this sre phase is predicted by proposal [ thm:24 ] .
[ thm:52 ] we have @xmath705 furthermore , the map @xmath706 is surjective and @xmath707 we defer most of the proof to the appendix ; here we briefly sketch two ways to compute .
the first is a direct approach .
use the chain complex . ]
@xmath708 for @xmath709 and the chain complex @xmath710 for @xmath449 with the nontrivial local system @xmath711 .
compute the cohomology of the cochain complex obtained by applying @xmath487 to the tensor product of and .
an alternative approach is to apply the kunneth formula for cohomology @xcite , which in this case gives a split short exact sequence @xmath712 ^ 5 \longrightarrow h^5(bso_3\times { { \mathbb r\mathbb p}}^{\infty};{{{\mathbb z}}_\phi } ) \\\longrightarrow \bigl[h^{\bullet } ( bso_3;{{\mathbb z}})*h^{\bullet } ( { { \mathbb r\mathbb p}}^{\infty};{{{\mathbb z}}_\phi } ) \bigr]^6 \longrightarrow 0 \end{gathered}\ ] ] here ` @xmath713 ' denotes the torsion product of abelian groups .
the tensor product in the kernel of is isomorphic to @xmath714 generated by the nonzero class in @xmath715 and the tensor product of @xmath716 and the nonzero class @xmath717 .
the quotient group in is isomorphic to @xmath532 , which is the torsion product @xmath718 .
[ thm:32 ] the image @xmath719 of @xmath720 under is the long - range effective topological theory of the sre phase identified in @xcite .
we argue for this claim in [ sec:7 ] .
[ thm:56 ] since @xmath719 is torsion we can lift it to the group @xmath721 where it is easier to identify a particular theory in this class ; see remark [ thm:44 ] .
we begin in [ subsec:7.1 ] with a general discussion of spatial boundary conditions for field theories and condensed matter systems .
we specialize to the case at hand : the bulk theory is gapped and the long - range topological theory is invertible .
we apply these general ideas in [ subsec:7.2 ] to argue for claim [ thm:32 ] , which locates the 3d @xmath1 phase with half - quantized surface thermal hall effect introduced in @xcite and further investigated in @xcite .
we recover some key aspects of that theory from our topological viewpoint .
suppose we are given a theory @xmath9 in @xmath8 spacetime dimensions .
it may be a quantum field theory or a condensed matter theory . then to a compact @xmath21-manifold @xmath22 _ with empty boundary _
we obtain a complex vector space of states .
we would like to extend to allow compact @xmath21-dimensional manifolds @xmath22 which have nonempty boundary .
these boundaries are _ spatial _ , not _
temporal_. in this case we expect to impose boundary conditions @xmath722 which essentially close off the boundary .
in other words , the pair @xmath723 behaves as a closed @xmath21-manifold for the pair of theories @xmath724 , and the @xmath725 theory attaches to it a vector space of states .
( without the boundary condition @xmath722 we expect instead a module for an algebra more complicated than @xmath68 , or an object in a category more complicated than @xmath33 . ) furthermore , we expect @xmath722 to be
_ local_. in classical physics a spatial boundary condition is typically a _
local constraint _ on fields : a boundary condition for a system of classical partial differential equations . in quantum physics a spatial boundary condition is a _ relative field theory _ ( [ subsec:2.3 ] ) @xmath726 the theory @xmath9 evaluates on @xmath22 to a map @xmath727 and the boundary condition @xmath722 evaluates on @xmath728 to a map @xmath729 .
the composition @xmath730 is tensor product with the vector space associated to @xmath22 in the theory @xmath724 . if @xmath9 is an invertible field theory , then @xmath722 is an anomalous theory with anomaly @xmath9 .
we call @xmath724 a _ bulk - boundary pair_. [ thm:36 ] a quantum boundary condition @xmath722 in quantum field theory is sometimes called a _
d - brane _ , a term most appropriate in the context of 2-spacetime dimensional conformal field theories . in condensed matter physics @xmath722 goes by a name like _ edge termination _ or _
surface termination _ , depending on the dimension of the theory .
sometimes the word ` excitation ' is used in place of ` termination ' .
[ thm:38 ] if @xmath9 is a _ topological _ field theory with values in an @xmath61-category @xmath64 , then a boundary condition is a 1-morphism @xmath731 in @xmath64 .
the dual is outgoing rather than incoming .
there is an equivalent exposition with the other choice , and we would not need the dual here . ]
map @xmath732 may be considered as an `` object in @xmath183 '' .
for example , if @xmath108 and @xmath64 is a 2-category of categories , then @xmath722 is literally an object in the 1-category @xmath183 .
boundary conditions in topological theories are a special case of a much more general construction ( * ? ? ?
* example 4.3.22 ) .
[ thm:37 ] if @xmath9 is a @xmath11-space dimensional theory of @xmath299-type , then is replaced by @xmath733 there is not a unique boundary condition for a given @xmath9 , but rather @xmath9 determines a collection of boundary conditions .
these formal considerations can lead to physical consequences .
one important example in condensed matter physics is the _ integer quantum hall effect _ ; see @xcite for an account aimed at mathematicians .
suppose the theory @xmath9 is the long - range topological approximation to a gapped @xmath11-space dimensional system of @xmath299-type .
then if a boundary condition produces a combined bulk - boundary pair which is still gapped , we expect that the long - range topological approximation is a bulk - boundary pair @xmath724 of topological theories . if @xmath9 describes a short - range entangled phase that is , @xmath9 is invertible then @xmath722 is a @xmath513-space dimensional anomalous theory with anomaly @xmath9 .
we implicitly assume that the truncation @xmath734 of @xmath9 is nontrivial .
if , furthermore , @xmath9 describes an spt phase , then we arrive at the following trichotomy .
@xmath735 possibility ( ii ) arises since by the definition of an spt phase @xmath9 restricts to a trivial theory when the symmetry is broken , and a theory relative to the trivial theory is non - anomalous . for ( iii )
we observe that an invertible theory relative to an invertible theory @xmath736 is a trivialization of @xmath736 , so if @xmath736 is not trivial then any relative theory must not be invertible . in the physics lingo it is not
short - range entangled but rather is long - range entangled it `` exhibits topological order '' .
the trichotomy is a restatement of an assertion in the introduction to @xcite .
we turn now to the spt phase identified as @xmath719 at the end of [ subsec:6.5 ] . the argument that @xmath719 corresponds to the 3d @xmath1 phase with half - quantized surface thermal hall effect
is based on ( iii ) in the trichotomy .
consider a long - range effective boundary condition
surface termination which is time - reversal symmetric and anomalous with anomaly @xmath719 .
this boundary condition is a relative 3-spacetime dimensional topological theory .
we claim that any chern - simons theory with chiral central charge @xmath737 is such an effective boundary condition : it satisfies ( iii ) in the trichotomy .
( see [ subsec:6.3 ] for a topological discussion of chiral central charge . )
one of the simplest examples is chern - simons theory for the maximal torus of @xmath738 , which was proposed for this role in @xcite and was realized as a boundary condition in the exactly soluble hamiltonian constructed in @xcite . to justify the claim begin with the short exact coefficient sequence @xmath739 in which the first map is the inclusion
identify @xmath740 , which in particular is untwisted , and so write as a class in one lower degree , and so identify it with particular field theories .
this circumvents the issues raised in [ subsec:2.7 ] about nontorsion classes ; see remark [ thm:44 ] . ] @xmath720 as the image of @xmath741 under the connecting homomorphism in the long exact sequence deduced from .
the image @xmath742 in the brown - comenetz dual of the sphere is computed via the sequence of maps @xmath743\\\longrightarrow [ \sigma
^4mtso_3,\sigma ^4i{{{\mathbb c}}/{{\mathbb z}}}]\cong { { { \mathbb c}}/{{\mathbb z}}}. \end{gathered}\ ] ] the last isomorphism follows from proposition [ thm:50](iv ) . since has order 2 , so does its image , and checking against we identify it with the anomaly theory @xmath744 with @xmath745 . as explained in [ subsec:6.3 ] , @xmath746 is the ( framing ) anomaly of any quantum chern - simons theory whose chiral central charge satisfies .
we prove proposition [ thm:50 ] and proposition [ thm:52 ] .
i thank oscar randal - williams for sharing his expertise , for correcting a mistake in a previous version of proposition [ thm:50 ] , and for providing a few arguments in the proof .
first recall some facts about madsen - tillmann spectra .
set @xmath747 . then @xmath748 . let @xmath749 denote the suspension spectrum of a pointed space @xmath22 .
the fibration @xmath750 is proved in ( * ? ? ?
* proposition 3.1 ) and ( * ? ? ?
* lemma 3.8 ) . for any space @xmath22
we have @xmath751 the stable homotopy groups of a pointed space @xmath383 are @xmath752 . the oriented version of expresses the thom spectrum @xmath753 as the colimit of a sequence of maps @xmath754 .
the homotopy groups of @xmath753 are thom s oriented bordism groups .
the long exact sequence of homotopy groups deduced from implies that @xmath755 is an isomorphism and that there is an exact sequence @xmath756 an element of @xmath757 is represented by a closed oriented @xmath130-manifold
@xmath354 , and its image under @xmath758 is the euler number of @xmath354 . for @xmath8
even the map @xmath759 is zero . for @xmath39
the map @xmath759 is surjective , since @xmath760 .
first apply with @xmath39 to derive the exact sequence @xmath761 as remarked above the euler characteristic map @xmath759 is onto , and since @xmath762 we deduce @xmath763 , which is ( iii ) in the proposition .
next , apply with @xmath161 to deduce @xmath764 and the composition @xmath765 is surjective .
then a stretch of the long exact sequence of homotopy groups deduced from with @xmath161 is @xmath766\ar@{=}[d ] & \pi _ 4x_3\ar[r ] & \pi _
4x_4\ar[r]^<<<<<\chi \ar@{=}[d ] & \pi _ 0^s ( bso_4)_+\ar[r ] \ar@{=}[d ] & \pi \mstrut _ 3x\mstrut _ 3\ar@{=}[d]\\ { { { { \mathbb z}}/2{{\mathbb z } } } } & & { { \mathbb z}}\times { { \mathbb z } } & { { \mathbb z } } & 0 } \end{gathered}\ ] ] from which @xmath767 . to see that it is the former , let @xmath9 be the fiber of the spectrum map @xmath768 which represents the generator of @xmath769
so there is a cofiber sequence @xmath770 then and the vanishing of @xmath771 imply that @xmath772 , whence the hurewicz map @xmath773 is an isomorphism .
in addition , the map @xmath774 is an isomorphism .
figure [ fig:11 ] is a schematic depiction of the long exact sequence of @xmath776-cohomology groups induced by the cofibration , where @xmath777 is the field of 2 elements .
the @xmath776-cohomology of a spectrum is a @xmath219-graded @xmath776-vector space which is a module for the steenrod algebra .
the dots indicate basis elements and the vertical arrows the action of the steenrod operations @xmath778 .
the degrees in the figure ascend from 0 to 5 .
the @xmath776-cohomology of @xmath440 is isomorphic to the group of cohomology operations @xmath779 and is computed by a theorem of serre .
the generators are the operations @xmath780 ( preceded by reduction modulo 2 ) , and the action of the steenrod operations is given by the adem relations .
for @xmath781 the generators are @xmath782 where @xmath783 is the mod 2 thom class of the virtual bundle , which stably is _ minus _ the canonical rank 3 bundle @xmath784 .
its total stiefel - whitney class is @xmath785 the inverse of the stiefel - whitney class of @xmath784 . the action of the total steenrod operation @xmath786 is @xmath787 .
the horizontal arrow in degree 0 follows from the definition of @xmath788 , and the arrows in degrees 2,3,4 from the module structure , as does the lack of a horizontal arrow in degree 5 .
exactness then implies the existence of a class in @xmath789 which maps to @xmath790 .
thus @xmath791 and we conclude @xmath792 , whence @xmath793 . for the last claim in
( iv ) we revisit . a 4-manifold which represents a class in @xmath794 has vanishing @xmath795 , since @xmath795 is a stable characteristic class and vanishes for rank 3 bundles .
thus its euler number is even , and since the euler number and signature are congruent modulo 2 its signature is also even . then observe that the 4-manifold @xmath796 represents an element of @xmath797 ( since it has vanishing euler characteristic , so a nonvanishing vector field which splits a line bundle off its tangent bundle ) and has signature 2 .
similar techniques prove ( i ) and ( ii ) .
( an alternative is to use the madsen - weiss theorem @xcite and known facts about the stable homology of mapping class groups of surfaces . )
first @xmath798 from and with @xmath108 .
there is a nontrivial @xmath25-invariant connecting these homotopy groups ; if not , then @xmath799 would be 2-dimensional .
let @xmath800 denote the spectrum with these two nonzero homotopy groups and nontrivial @xmath25-invariant .
its @xmath776-cohomology is worked out in figure [ fig:12 ] using the cofiber sequence @xmath802 .
all cohomology in degree 1 vanishes . also , the nontrivial connecting map @xmath803 is the @xmath25-invariant .
let @xmath804 be the fiber of the postnikov map @xmath805 , and consider the cofiber sequence @xmath806 .
the induced maps on @xmath776-cohomology are worked out in figure [ fig:13 ] ; the nonzero cohomology is in degrees 0,2,4,5 .
we deduce @xmath808 and from hurewicz @xmath809 and @xmath810 is an isomorphism . now
the long exact sequence of homotopy groups induced from with @xmath108 includes the stretch @xmath811\ar@{=}[d]&\pi _ 4x_2\ar[r ] & \pi _ 2^s(bso_2)_+\ar[r]\ar@{=}[d ] & \pi _ 3s^0 \ar@{=}[d ] \\ 0&&{{\mathbb z}}\times { { { { \mathbb z}}/2{{\mathbb z } } } } & { { { \mathbb z}}/24{{\mathbb z } } } } \end{gathered}\ ] ] this implies @xmath812 ; rules out the latter since @xmath813 and @xmath814 . for the last statement in ( ii ) consider the long exact sequence of homotopy groups induced from with @xmath39 : @xmath815\ar@{=}[d]&\pi _ 4x_3\ar[r ] \ar@{=}[d ] & \pi _ 1^s(bso_3)_+\ar[r]\ar@{=}[d ] & \pi _ 3x_2 \ar@{=}[d ] \\ { { \mathbb z}}&{{\mathbb z}}&{{{{\mathbb z}}/2{{\mathbb z } } } } & 0 } \end{gathered}\ ] ] and so the first homomorphism is multiplication by @xmath644 on generators . as a preliminary we prove that @xmath816 is finite .
since this group is finitely generated , an equivalent assertion is @xmath817 . to prove this
observe that @xmath818 induces an isomorphism on rational homology in degrees @xmath819 , whence also on rational homotopy groups in that range .
collating with and facts in the previous proof we have @xmath820 introduce the mapping spectrum @xmath821 from - graded homotopy group of the anderson dual to @xmath781 is the derived @xmath822 ; there is a shift of 5 in ] and we deduce @xmath823 and @xmath824 .
the cohomology group in is @xmath825 , which we compute using the atiyah - hirzebruch spectral sequence .
the rows in the @xmath626 page , shown in figure [ fig:9 ] , are twisted cohomology groups of @xmath449 .
all differentials vanish in this range , for degree reasons , whence @xmath825 is isomorphic to @xmath714 or @xmath826 , depending on whether there is a group extension .
define @xmath827 then @xmath828\cong h^{5-j}(x_3;{{\mathbb z}})\cong h^{5-j}(bso_3;{{\mathbb z}}),\ ] ] where the last step is the thom isomorphism .
hence @xmath829 the map @xmath830 induces a map @xmath831 and so a map of atiyah - hirzebruch spectral sequences .
the @xmath626 page of the spectral sequence for @xmath832 is shown in figure [ fig:10 ] .
the group @xmath833 was computed after , and it follows that @xmath834 vanishes and there is no group extension passing from the degree 0 part of the @xmath835 page to @xmath832 .
the map of spectral sequences now implies that there is no group extension in the @xmath50-spectral sequence either , that @xmath836 , and that @xmath837 is surjective .
matthew ando , andrew j. blumberg , david gepner , michael j. hopkins , and charles rezk , _ an @xmath839-categorical approach to @xmath840-line bundles , @xmath840-module thom spectra , and twisted @xmath840-homology _ , http://arxiv.org/abs/arxiv:1403.4325[arxiv:1403.4325 ] .
fj burnell , xie chen , lukasz fidkowski , and ashvin vishwanath , _ exactly soluble model of a 3d symmetry protected topological phase of bosons with surface topological order _ , http://arxiv.org/abs/arxiv:1302.7072 [ arxiv:1302.7072 ] . john c. baez and james dolan , _ higher - dimensional algebra and topological quantum field theory _ , http://dx.doi.org/10.1063/1.531236[j .
phys . * 36 * ( 1995 ) ] , no .
11 , 60736105 , http://arxiv.org/abs/arxiv:q-alg/9503002[arxiv:q-alg/9503002 ] .
xie chen , lukasz fidkowski , and ashvin vishwanath , _ symmetry enforced non - abelian topological order at the surface of a topological insulator _ , phys .
b * 89 * ( 2014 ) , 165132 , http://arxiv.org/abs/1306.3250[1306.3250 ] .
xie chen , zheng - cheng gu , zheng - xin liu , and xiao - gang wen , _ symmetry protected topological orders and the cohomology class of their symmetry group _ , http://dx.doi.org/10.1103/physrevb.87.155114[phys.rev .
* b87 * ( 2013 ) ] , 155114 , http://arxiv.org/abs/1106.4772[arxiv:1106.4772 [ cond-mat.str-el ] ] .
xie chen , zheng - cheng gu , and xiao - gang wen , _ local unitary transformation , long - range quantum entanglement , wave function renormalization , and topological order _ , phys .
b * 82 * ( 2010 ) , 155138 , http://arxiv.org/abs/arxiv:1004.3835[arxiv:1004.3835 ] .
xie chen , zheng - cheng gu , and xiao - gang wen , _ complete classification of one - dimensional gapped quantum phases in interacting spin systems _ , physical review b * 84 * ( 2011 ) , no .
23 , 235128 , http://arxiv.org/abs/arxiv:1103.3323[arxiv:1103.3323 ] .
xie chen , zheng - xin liu , and xiao - gang wen , _ two - dimensional symmetry - protected topological orders and their protected gapless edge excitations _ , physical review b * 84 * ( 2011 ) , no .
23 , 235141 , http://arxiv.org/abs/arxiv:1106.4752[arxiv:1106.4752 ] .
pierre deligne , pavel etingof , daniel s. freed , lisa c. jeffrey , david kazhdan , john w. morgan , david r. morrison , and edward witten ( eds . ) , _ quantum fields and strings : a course for mathematicians .
, 2 _ , american mathematical society , providence , ri , 1999 .
material from the special year on quantum field theory held at the institute for advanced study , princeton , nj , 19961997 . pierre deligne and daniel s. freed , _ classical field theory _ , quantum fields and strings : a course for mathematicians , vol . 1 , 2 ( princeton , nj , 1996/1997 ) , amer . math .
soc . , providence , ri , 1999 , pp .
137225 .
jacques distler , daniel s. freed , and gregory w. moore , _ spin structures and superstrings _ , surveys in differential geometry .
volume xv .
perspectives in mathematics and physics , surv .
press , somerville , ma , 2011 , pp .
http://arxiv.org/abs/arxiv:0906.0795[arxiv:0906.0795 ] .
christopher l. douglas and andr g. henriques , http://dx.doi.org/10.1090/pspum/083/2742433[_topological modular forms and conformal nets _ ] , mathematical foundations of quantum field theory and perturbative string theory , proc . sympos .
pure math .
83 , amer .
soc . , providence , ri , 2011 , pp .
http://arxiv.org/abs/arxiv:1103.4187[arxiv:1103.4187 ] .
h. endo , m. korkmaz , d. kotschick , b. ozbagci , and a. stipsicz , _ commutators , lefschetz fibrations and the signatures of surface bundles _
, http://dx.doi.org/10.1016/s0040-9383(01)00011-8[topology * 41 * ( 2002 ) ] , no . 5 , 961977 , http://arxiv.org/abs/arxiv:math/0103176[arxiv:math/0103176 ] .
daniel s. freed , _ the cobordism hypothesis _
, http://dx.doi.org/10.1090/s0273-0979-2012-01393-9[bull .
( n.s . ) * 50 * ( 2013 ) ] , no . 1 , 5792 , http://arxiv.org/abs/arxiv:1210.5100[arxiv:1210.5100 ] . to3em , _ characteristic numbers and generalized path integrals _ , geometry , topology , & physics , conf
lecture notes geom .
topology , iv , int . press , cambridge ,
ma , 1995 , pp .
http://arxiv.org/abs/arxiv:dg-ga/9406002[arxiv:dg-ga/9406002 ] .
daniel s. freed and michael j. hopkins , _ chern - weil forms and abstract homotopy theory _ , http://dx.doi.org/10.1090/s0273-0979-2013-01415-0[bull .
( n.s . ) * 50 * ( 2013 ) ] , no . 3 , 431468 , http://arxiv.org/abs/arxiv:1301.5959[arxiv:1301.5959 ] .
daniel s. freed , michael j. hopkins , and constantin teleman , http://dx.doi.org/10.1093/acprof:oso/9780199534920.003.0019[_consistent orientation of moduli spaces _ ] , the many facets of geometry , oxford univ . press , oxford , 2010 , pp .
http://arxiv.org/abs/arxiv:0711.1909[arxiv:0711.1909 ] .
lukasz fidkowski and alexei kitaev , _ topological phases of fermions in one dimension _ , http://dx.doi.org/10.1103/physrevb.83.075103[phys .
b * 83 * ( 2011 ) ] , 075103 , http://arxiv.org/abs/arxiv:1008.4138 [ arxiv:1008.4138 ] .
daniel s. freed and gregory w. moore , _ setting the quantum integrand of m - theory _ , http://dx.doi.org/10.1007/s00220-005-1482-7[commun .
* 263 * ( 2006 ) ] , 89132 , http://arxiv.org/abs/hep-th/0409135[arxiv:hep-th/0409135 ] .
to3em , _ twisted equivariant matter _ , http://dx.doi.org/10.1007/s00023-013-0236-x[ann .
henri poincar * 14 * ( 2013 ) ] , no . 8 , 19272023 , http://arxiv.org/abs/arxiv:1208.5055[arxiv:1208.5055 ] .
daniel s. freed , gregory w. moore , and graeme segal , _ the uncertainty of fluxes _ , http://dx.doi.org/10.1007/s00220-006-0181-3[commun .
* 271 * ( 2007 ) ] , 247274 , http://arxiv.org/abs/hep-th/0605198[arxiv:hep-th/0605198 ] .
daniel s. freed and frank quinn , _ chern - simons theory with finite gauge group _ , comm .
* 156 * ( 1993 ) , no . 3 , 435472 , http://arxiv.org/abs/arxiv:hep-th/911100[arxiv:hep-th/911100 ] .
daniel s. freed and constantin teleman , _ relative quantum field theory _ , http://dx.doi.org/10.1007/s00220-013-1880-1[comm .
* 326 * ( 2014 ) ] , no . 2 , 459476 , http://arxiv.org/abs/arxiv:1212.1692[arxiv:1212.1692 ] .
sren galatius , ulrike tillmann , ib madsen , and michael weiss , _ the homotopy type of the cobordism category _ , http://dx.doi.org/10.1007/s11511-009-0036-9[acta math .
* 202 * ( 2009 ) ] , no . 2 , 195239 , http://arxiv.org/abs/arxiv:math/0605249 [ arxiv : math/0605249 ] . zheng - cheng gu and xiao - gang wen , _ symmetry - protected topological orders for interacting fermions fermionic topological non - linear @xmath841-models and a group super - cohomology theory _ , http://arxiv.org/abs/1201.2648[1201.2648 ] . m. j. hopkins and i. m. singer , _
quadratic functions in geometry , topology , and m - theory _ , j. diff .
* 70 * ( 2005 ) , 329452 , http://arxiv.org/abs/math/0211216[arxiv:math/0211216 ] .
ling - yan hung and xiao - gang wen , _ universal symmetry - protected topological invariants for symmetry - protected topological states _ , phys .
b * 89 * ( 2014 ) , http://arxiv.org/abs/arxiv:1311.5539 [ arxiv:1311.5539 ] .
alexei kitaev , _ periodic table for topological insulators and superconductors _ , http://dx.doi.org/10.1063/1.3149495[aip conf.proc .
* 1134 * ( 2009 ) ] , 2230 , http://arxiv.org/abs/0901.2686 [ arxiv:0901.2686 [ cond-mat.mes-hall ] ] .
jacob lurie , _ on the classification of topological field theories _ , current developments in mathematics , 2008 , int . press , somerville , ma , 2009 ,
http://arxiv.org/abs/arxiv:0905.0465 [ arxiv:0905.0465 ] .
yuan - ming lu and ashvin vishwanath , _ theory and classification of interacting integer topological phases in two dimensions : a chern - simons approach _ , http://dx.doi.org/10.1103/physrevb.86.125119[phys .
b * 86 * ( 2012 ) ] , 125119 , http://arxiv.org/abs/arxiv:1203.3156 [ arxiv:1203.3156 ] .
ib madsen and michael weiss , _ the stable moduli space of riemann surfaces : mumford s conjecture _ , http://dx.doi.org/10.4007/annals.2007.165.843[ann . of math .
( 2 ) * 165 * ( 2007 ) ] , no . 3 , 843941 , http://arxiv.org/abs/arxiv:math/0212321[arxiv:math/0212321 ] .
scott morrison and kevin walker , _ higher categories , colimits , and the blob complex _ , http://dx.doi.org/10.1073/pnas.1018168108[proc .
usa * 108 * ( 2011 ) ] , no .
20 , 81398145 , http://arxiv.org/abs/arxiv:1108.5386[arxiv:1108.5386 ] .
eugeniu plamadeala , michael mulligan , and chetan nayak , _ short - range entangled bosonic states with chiral edge modes and @xmath283-duality of heterotic strings _
b * 88 * ( 2013 ) , 045131 , http://arxiv.org/abs/1304.0772[1304.0772 ] .
graeme segal , _ the definition of conformal field theory _ , topology ,
geometry and quantum field theory , london math .
lecture note ser .
308 , cambridge univ . press , cambridge , 2004 , pp .
421577 .
ashvin vishwanath and t senthil , _ physics of three - dimensional bosonic topological insulators : surface - deconfined criticality and quantized magnetoelectric effect _
, physical review x * 3 * ( 2013 ) , no . 1 , 011016 , http://arxiv.org/abs/arxiv:1209.3058[arxiv:1209.3058 ] . to3em , _ the verlinde algebra and the cohomology of the grassmannian _ , geometry , topology , & physics , conf
lecture notes geom .
topology , iv , int . press , cambridge ,
ma , 1995 , pp .
http://arxiv.org/abs/arxiv:hep-th/9312104[arxiv:hep-th/9312104 ] .
xiao - gang wen , _ symmetry - protected topological invariants of symmetry - protected topological phases of interacting bosons and fermions _
b * 89 * ( 2014 ) , http://arxiv.org/abs/arxiv:1301.7675[arxiv:1301.7675 ] .
juven wang , zheng - cheng gu , and xiao - gang wen , _ a field theory representation of pure gauge and mixed gauge - gravity symmetry - protected topological invariants , group cohomology and beyond _ , http://arxiv.org/abs/arxiv:1405.7689[arxiv:1405.7689 ] .
chong wang , andrew c potter , and t senthil , _ classification of interacting electronic topological insulators in three dimensions _ , science * 343 * ( 2014 ) , no . 6171 , 629631 , http://arxiv.org/abs/arxiv:1306.3238 [ arxiv:1306.3238 ] .
|
quantum field theories with an energy gap can be approximated at long - range by _ topological _ quantum field theories .
the same should be true for suitable condensed matter systems .
for those with _ short range entanglement _ ( sre ) the effective topological theory is _
invertible _ , and so amenable to study via stable homotopy theory .
this leads to concrete topological invariants of gapped sre phases which are finer than existing invariants .
computations in examples demonstrate their effectiveness .
plus4.5pt minus9pt plus4.5pt 10.5pt plus4.5pt minus6pt = 15 truept
|
in this paper , we will study a cahn - hilliard equation posed on an evolving surface with prescribed velocity .
the key methodology is to discretise the equations using the evolving surface finite element method @xcite originally proposed for a surface heat equation .
the idea is to take a triangulation of the initial surface and evolve the nodes along the velocity field .
this leads to a family of discrete surfaces on which we can pose a variational form of the cahn - hilliard equation .
there are two key results in this paper : first , we show well posedness of the continuous scheme and , second , we show convergence of a finite element scheme .
the well posedness result is proven by rigorously showing convergence , along a subsequence , of the discrete scheme .
in contrast to the planar setting , there are extra difficulties in this work since the classical bochner space set - up is unavailable to us .
the finite element method is analysed under the assumption of higher regularity of the solution and shown to converge to the true solution quadratically with respect to the mesh size in an @xmath0 norm .
the paper concludes with some numerical examples to show various properties of the methodology .
we assume we are given an evolving surface @xmath1 , for @xmath2 $ ] , which evolves according to a given underlying velocity field @xmath3 which can be decomposed into normal @xmath4 and tangential components @xmath5 so that @xmath6 .
we seek a solution @xmath7 of @xmath8 subject to the initial condition @xmath9 here @xmath10 denotes the material derivative of @xmath7 and @xmath11 the laplace - beltrami operator of @xmath7 .
the function @xmath12 is a double well potential , which we will take to be given by @xmath13 the behaviour of the cahn - hilliard equation in the planar case is well studied @xcite .
extra effects such as spatial or concentration dependent mobilities or more physically realistic potentials could also be solved with similar methods to those suggested in this paper .
such considerations are left for future work .
this cahn - hilliard equation is a simplification of the model for surface dissolution set out in @xcite arising from a conservation law .
the model @xcite takes a different approach and considers a gradient flow for an energy consisting of the sum of the ginzburg - landau functional and a helfrich energy on a stationary surface .
one could alternatively couple the evolution of the surface to the surface field @xmath7 and recover a gradient flow of the ginzburg - landau functional @xcite .
the results in this work can be seen as a generalisation of the work of @xcite to evolving surfaces .
that work considers a fully discrete approximation of a cahn - hilliard equation posed on a two - dimensional stationary surface with boundary ( with a zero dirichlet boundary condition ) under the assumption @xmath14 and @xmath15 for @xmath16 .
their method uses a triangulated surface for the spatial discretisation and a crank - nicolson scheme in time .
they show an error estimate of the form @xmath17 where @xmath18 is a partition of time with fixed time step @xmath19 and @xmath20 is the inverse lift of the continuous solution @xmath7 .
the paper is laid out as follows . in section two
, we will derive a cahn - hilliard equation on an evolving surface using a local conservation law .
we introduce the notation for partial differential equations on evolving surfaces taken from @xcite and state any assumptions on the smoothness of the surfaces and its evolution we require .
the third section introduces a finite element discretisation of the continuous equations .
we describe the process of triangulating an evolving surface and how we formulate the space discrete - time continuous problem as a system of ordinary differential equations .
this section is completed by showing some domain perturbation results relating geometric quantities on the discrete and smooth surfaces .
well posedness of the continuous equations is addressed in the fourth section .
an existence result is achieved by showing convergence , along a subsequence , of the discrete solutions as the mesh size tends to zero . in section five
, we analyse the errors introduced by our finite element scheme and go on to show an optimal order error estimate .
some numerical experiments are shown in the sixth section backing up the analytical results .
we will use a gronwall inequality as a standard tool in the analysis which leads to exponential dependence on @xmath21 in most bounds .
we are not interested in taking @xmath22 in this work so will simply write @xmath23 for a generic constant which depends on @xmath21 .
in this section , we will derive a cahn - hilliard equation on an evolving surface as a conservative advection - diffusion equation .
we will also introduce functional analytic setting and definition of solution that will be used . given a final time @xmath24 , for each time @xmath2
$ ] , we write @xmath25 for a compact , smooth , connected @xmath26-dimensional hypersurface in @xmath27 for @xmath28 or @xmath29 and @xmath30 .
we assume that @xmath25 is the boundary of an open , bounded domain @xmath31 .
it follows that @xmath25 admits a description as the zero level set of a signed distance function @xmath32 so that @xmath33 in @xmath31 and @xmath34 in @xmath35 .
we denote by @xmath36 for the space - time domain given by @xmath37 } \gamma(t ) \times \ { t \}.\ ] ] . ] for our analysis , it is sufficient to consider @xmath38 locally to @xmath25 .
we restrict our considerations to @xmath39 , an open neighbourhood of @xmath25 .
we choose @xmath39 so that @xmath40 for @xmath41 and assume that @xmath42 here @xmath43 } { \mathcal{n}}(t ) \times \ { t \}$ ] .
the orientation of @xmath25 is fixed by choosing @xmath44 as the outward pointing normal , so that @xmath45 . for @xmath46 , we denote @xmath47 the projection operator onto the tangent space @xmath48 , given by @xmath49 and by @xmath50 the ( extended ) weingarten map ( or shape operator ) , @xmath51 we will use the fact that @xmath52 .
finally , we denote by @xmath53 the mean curvature of @xmath25 @xmath54 for a function @xmath55 , we define its tangential gradient @xmath56 by @xmath57 where @xmath58 is a smooth extension of @xmath59 away from @xmath25 .
it can be shown that this definition is independent of the choice of extension .
we denote the @xmath60 components of @xmath56 by @xmath61 the laplace - beltrami operator is given by @xmath62 we will denote by @xmath63 the surface measure on @xmath25 which admits the following formula for partial integration for a portion @xmath64 ( * ? ? ? * theorem 2.10 ) : @xmath65 where @xmath66 is the co - normal to @xmath67 which is normal to @xmath67 but tangent to @xmath25 . if @xmath68 and has no boundary , the boundary term vanishes .
furthermore , we have a green s formula on @xmath25 ( * ? ? ?
* theorem 2.14 ) : @xmath69 these formulae allow the definition of weak derivatives and sobolev spaces .
we define the space @xmath70 by @xmath71 with norm @xmath72 this can be easily extended to higher order spaces .
see @xcite for details .
we will use the notation @xmath73 for @xmath74 .
we will make use of the following sobolev embeddings : [ ( * ? ? ?
* theorems 2.5 and 2.6 ) ] [ lem : sobolev - embed ] for @xmath25 as above , we have @xmath75 furthermore there exists a constant @xmath76 , independent of @xmath77 , such that for any @xmath78 , @xmath79 in particular , this allows us to embed @xmath80 in @xmath81 for all dimensions @xmath82 so that @xmath83 .
further , we assume that for each @xmath84 there exists a unique @xmath85 , such that @xmath86 see ( * ? ? ?
* chapter 14 ) for a proof .
we extend @xmath87 and @xmath88 to functions on @xmath89 by setting @xmath90 and similarly @xmath91 and @xmath92 for @xmath84 . although it is sufficient to describe the evolution of the surface through a normal velocity , we wish to consider material surfaces for which a material particle , at @xmath93 on @xmath25 , has a material velocity @xmath94 not necessarily only in the normal direction .
the normal velocity of the surface can be calculated to be @xmath95 .
we say @xmath96 is a tangential velocity field if @xmath97 in @xmath89 .
given a tangential velocity field @xmath96 , we call @xmath98 a material velocity field .
we assume that we are given a global velocity field @xmath3 so that points @xmath93 evolve with the velocity @xmath99 .
we will assume that @xmath100 .
given a family of surfaces @xmath1 evolving in time with normal velocity field @xmath101 , we define the normal time derivative @xmath102 of a function @xmath103 by @xmath104 here , @xmath58 denotes a smooth extension of @xmath59 to @xmath89 .
this derivative describes how a quantity @xmath59 evolves in time with respect to the evolution of @xmath25 .
it can be shown that this definition is an intrinsic surface derivative , independent of the choice of extension .
given a tangential vector field @xmath96 , we define the material derivative of a scalar function @xmath105 , by @xmath106 the following formula shows the significance of the material derivative .
the result is a generalisation of the classical reynolds transport formula to curved domains .
[ transport formula ( * ? ? ?
* lemma 2.1 ) ] [ lem : transport ] let @xmath107 be an evolving surface with normal velocity @xmath101 .
let @xmath96 be a tangential velocity field on @xmath107 .
let the boundary @xmath67 evolve with velocity @xmath6 .
assume that @xmath108 are functions such that all the following quantities exist .
then , we obtain the identity @xmath109 furthermore , we have @xmath110 let @xmath111 be a matrix which is positive definite on the tangent space to @xmath25 . denote by @xmath112 the rate of deformation tensor given by @xmath113 and by @xmath114 the tensor @xmath115 then we have the formula @xmath116 we conclude this subsection with a result allowing us to extend functions defined on one surface to the whole space - time domain .
[ lem : pull - back ] fix @xmath2 $ ] and let @xmath117 , respectively @xmath118 .
then there exists an extension @xmath119 such that @xmath120 and @xmath121 , resp .
@xmath122 , for all times @xmath123 $ ] and @xmath124 .
the ordinary differential equation : @xmath125 , \qquad x(t ) = x,\ ] ] determines a flow @xmath126 on @xmath36 for @xmath127 such that @xmath128 \quad \mbox { and } \phi_t(x ) = x.\ ] ] our assumptions on @xmath3 imply that @xmath129 and @xmath130 are both @xmath131 mappings ( * ? ? ? * theorem 3.1 ) .
we define the extension @xmath58 by @xmath132 it is clear that since @xmath133 , we have @xmath121 ( resp .
@xmath122 ) for all times @xmath123 $ ] .
finally , we can calculate for @xmath134 , @xmath135 which shows the result .
we will consider a conservation law on an evolving surface with a diffusive flux driven by a chemical potential .
this is the approach taken by @xcite .
in general , the ginzburg - landau functional on @xmath25 will not decrease along the trajectory of solutions .
let @xmath7 represent a density of a scalar quantity on @xmath25 .
following @xcite , we arrive at the pointwise conservation law @xmath136 here @xmath137 represents the tangential flux of @xmath7 on @xmath1 .
we will assume that the flux @xmath137 is the sum of a diffusive flux @xmath138 and an advective flux @xmath139 : @xmath140 the diffusive flux is driven by the gradient of chemical potential @xmath141 gives us the split system @xcite [ eq : ch - split ] @xmath142 this leads to the fourth order cahn - hilliard equation on @xmath36 : @xmath143 we close the system with the initial condition @xmath144 there are no boundary conditions since the boundary of @xmath25 is empty .
one can derive the cahn - hilliard equations posed in a cartesian domain as an @xmath145 gradient flow of the ginzburg - landau functional . to obtain a gradient flow on an evolving surface
, there would need to be a model for @xmath3 and which would lead to a coupled system for @xmath7 and @xmath3 . in terms of modelling
, we feel these extra terms are geometric terms determining an evolution equation for the surface , which we assume is given .
therefore , we do not consider such terms in this work
. in standard parabolic theory one looks for solutions in bochner spaces . considering our cahn - hilliard equation on a cartesian domain @xmath146 @xcite
, one would expect solutions to live in the spaces @xmath147 these spaces are constructed by considering @xmath7 as a function from @xmath148 into the hilbert space @xmath149 .
we would like to extend this definition so that @xmath150 is in the now time - dependent hilbert space @xmath80 .
we consider sobolev spaces over the space - time domain @xmath36 .
we will write @xmath151 for the space - time gradient and @xmath152 for the space - time measure on @xmath36 .
this approach is similar to the eulerian formulation of @xcite .
we contrast our approach with that of @xcite , who proposed using an equivalent formulation using a reference domain .
we start by presenting the space - time domains @xmath153 and @xmath154 defined by @xmath155 with norms @xmath156 [ ( * ? ? ?
* theorem 2.9 ) ] [ prop : compact - embed ] the space @xmath154 is compactly embedded into @xmath153 .
using the identities , @xmath157 and @xmath158 our assumptions on @xmath3 imply that the space - time norms can be replaced with the equivalent norms @xmath159 we will use the equivalent primed norms ( dropping the prime ) on @xmath153 and @xmath154 in the following .
we define the space @xmath160 by @xmath161 with the inner product @xmath162 it is clear that @xmath160 is equivalent to @xmath153 and hence is a hilbert space .
next , we define the space @xmath163 as @xmath164 with the inner product @xmath165 where @xmath56 should be interpreted in the weak sense . notice that elements of this space are weakly differentiable at almost every time .
the space @xmath163 is a hilbert space .
it is clear that @xmath163 is an inner product space and we are left to show completeness .
let @xmath166 be a cauchy sequence in @xmath163 .
this implies that @xmath166 and @xmath167 are cauchy sequences in @xmath153 and @xmath168 .
this means that there exists @xmath169 such that @xmath170 fix @xmath171 and let @xmath172 and @xmath173 . using , we can construct @xmath174 such that @xmath175 and @xmath176 for each time @xmath177 .
then , for @xmath178 , we obtain @xmath179 where we have used the fact that @xmath166 is weakly differentiable at almost every time . taking the limit @xmath180
, we infer @xmath181 since this holds for all @xmath173 , by the fundamental lemma of the calculus of variations , at @xmath182 , we have @xmath183 since the choice of @xmath184 was arbitrary , we infer that @xmath185 is the weak gradient of @xmath59 for almost every time @xmath177 and the proof is complete .
the equivalence of norms implies that @xmath186 with @xmath187 if , and only if , @xmath188 . for @xmath189
, we will define the space @xmath190 by @xmath191 with norm @xmath192 it is clear that @xmath193 and that @xmath194 finally , we define @xmath195 and @xmath196 by @xmath197 as a restriction on our analysis we will only consider @xmath10 as a function in @xmath160 since we do not wish to consider a weak material derivative .
such considerations are left to future work .
we conclude this section with a result which will take an integral in time equality into an almost everywhere in time equality .
the proof is the generalisation of a similar result given in ( * ? ? ?
* lemma 7.4 ) for planar domains .
[ lem : tiredy ] let @xmath186 with @xmath198 then for almost all times @xmath177 , @xmath199 fix @xmath200 and @xmath201)$ ] , then choosing @xmath202 and @xmath203 since the choice of @xmath204 was arbitrary , the fundamental lemma of the calculus of variations implies the result .
we start by multiplying ( [ eq : ch - w ] , [ eq : ch - u ] ) by a test function @xmath205 and apply integration by parts to the laplacian terms to give the weak form
. this will be the definition of solution used throughout this paper .
existence and uniqueness of solutions will be shown .
[ def : ch - solution ] we say that the pair @xmath206 , with @xmath207 and @xmath208 , are a weak solution of the cahn - hilliard equation if , for almost every time @xmath177 , [ eq : ch - weak ] @xmath209 for all @xmath200 , and @xmath210 pointwise almost everywhere in @xmath211 .
restricting our thoughts to @xmath212 , applying the transport formula to the first two terms in gives the variational formulation : [ eq : ch - var ] @xmath213 we remark that this formulation has no explicit mention of the velocity field @xmath3 and will be the basis of our finite element calculations .
it will be useful to write these equations using abstract bilinear forms .
we define the following three to describe the above equations for @xmath214 : @xmath215 this lets us write as @xmath216 and as @xmath217 we may also write the results of in this form : @xmath218 with the addition of @xmath219 using @xmath220 in the definition of @xmath114 .
in this section , we propose a finite element method for approximating solutions of the cahn - hilliard equation based on the evolving surface finite element method @xcite .
let @xmath221 be a polyhedral approximation of the initial surface @xmath211 with the restriction that the nodes @xmath222 of @xmath221 lie on @xmath211 .
we evolve the nodes @xmath223 by the smooth surface velocity : @xmath224 linearly interpolating between these nodes defines a family of discrete surfaces @xmath225 . at each time
, we assume that we have a triangulation @xmath226 of @xmath227 , with @xmath228 the maximum diameter of elements in @xmath226 uniformly in time : @xmath229 we assume this triangulation is quasi - uniform @xcite uniformly in time . in practical situations , assuming
a uniformly regular mesh may not be feasible .
large surface deformations can lead to poor quality triangulations with deformed elements . in such cases
, re - meshing may be required @xcite .
alternatively , one may use an arbitrary lagrangian - eulerian formulation by allowing extra tangential mesh motions @xcite .
we define @xmath230 element - wise as the unit outward pointing normal to @xmath227 and denote by @xmath231 the tangential gradient on @xmath227 defined element - wise by @xmath232 this is a vector - valued quantity and we will denote its components by @xmath233 we define the finite element space of piecewise linear functions on @xmath227 by @xmath234 we will write @xmath235 for the nodal basis of @xmath236 given by @xmath237 . the definition of a basis of @xmath236 allows us to characterise the velocity of the surface @xmath225 .
an arbitrary point @xmath93 on @xmath227 evolves according to the discrete velocity @xmath238 given by @xmath239 we will write @xmath240 as the discrete equivalent to @xmath36 : @xmath241 the discrete velocity @xmath238 induces a discrete material derivative . for a scalar quantity @xmath242 on @xmath240 , we define the discrete material derivative @xmath243 by @xmath244 where @xmath245 is an arbitrary extension of @xmath242 to @xmath89 .
this leads to the remarkable transport property of the basis functions @xmath246 .
[ transport of basis functions ( * ? ? ?
* proposition 5.4 ) ] [ lem : transport - basis ] let @xmath247 be a nodal basis function as described above , then @xmath248 from a practical view point , a key advantage of this methodology is that , since basis functions have zero discrete material velocity , there is no mention of the velocity or curvature in the resulting finite element scheme .
these discrete quantities also satisfy a variant of the transport formula from .
we label the surface measure on @xmath227 as @xmath249 .
[ transport lemma for triangulated surfaces ( * ? ? ?
* lemma 4.2 ) ] [ lem : discrete - transport ] let @xmath225 be a discrete family of triangulated surfaces evolving with velocity @xmath238 .
let @xmath250 be time - dependent finite element functions such that the following quantities exist .
then , we have @xmath251 in particular , for the @xmath0 inner product this means that @xmath252 and for the dirichlet inner product , we obtain @xmath253 where @xmath254 [ lem : vh - bound ] under our assumptions on @xmath225 , we have that @xmath255 } \left ( { \left\| \nabla_{\gamma_h } \cdot v_h \right\|}_{l^\infty(\gamma_h(t ) ) } + { \left\| { \mathcal{b}}_h(v_h ) \right\|}_{l^\infty(\gamma_h(t ) ) } \right ) \le c \sup_{t \in [ 0,t ] } { \left\| v \right\|}_{c^2({\mathcal{n}}_t)}.\ ] ] the result follows from applying the geometric estimates and along with our assumption that @xmath100 .
we will assume that there exists a mesh size @xmath256 such that @xmath257 is bounded independently of @xmath228 for @xmath258 .
this implies that there exists @xmath259 such that for all @xmath258 , we have @xmath260 one particular choice of initial condition will be to take @xmath261 as a suitable approximation of @xmath262 ( for example , @xmath263 defined in ) for @xmath264 .
our solution spaces will be @xmath265 \ } \\
s_h^t & : = \ { \phi_h \in { \widetilde}{s}_h^t : { \partial^\bullet}_h \phi_h \in c ( { \mathcal{g}}_{h , t } ) \}. \end{aligned}\ ] ] the finite element scheme is : given @xmath261 , find @xmath266 and @xmath267 such that for almost every time @xmath177 [ eq : ch - fem ] @xmath268 for all @xmath269 , subject to the initial condition @xmath270 the transport formula implies that , for @xmath271 , is equivalent to @xmath272 we can write these equations in matrix form . first , we will introduce vectors @xmath273 for the nodal values of @xmath274 and @xmath275 by @xmath276 in place of the bilinear forms , we have the mass matrix @xmath107 and stiffness matrix @xmath277 : @xmath278 and in place of the non - linear term , we will write @xmath279 using the transport of basis property ( ) , we can write as [ eq : ch - matrix ] @xmath280 alternatively , eliminating @xmath281 , this can be written as @xmath282 one could also use lumped mass integration ( * ? ? ?
* chapter 15 ) instead of the full mass matrix .
notice that this is the same structure as a finite element discretisation of a cahn - hilliard equation posed on a planar domain .
we now have time dependent matrices which need to be assembled on each time step .
various time stepping schemes have been considered for second - order parabolic problems on evolving surfaces @xcite .
next , we introduce abstract notation which permit a more compact writing of the analysis that follows : @xmath283 this lets us write as @xmath284 and as @xmath285 the transport laws from transfer to the abstract setting also : @xmath286 where @xmath287 under the above assumptions , the following estimates are possible .
[ well - posedness of the finite element scheme ] [ thm : ch - fem - wp ] under the above assumptions on @xmath261 and @xmath225 , there exists a unique solution pair @xmath288 , both with @xmath131 in time nodal values , to the finite element scheme and @xmath289 is conserved : @xmath290 furthermore , there exists @xmath291 , @xmath292 , and @xmath293 , which depend on the final time @xmath294 and the @xmath295-norm of the initial condition @xmath261 , such that for all @xmath296 the following bound is satisfied : @xmath297 the proof will be shown after we have proven some intermediate results .
the following analysis will rely on lift operators defined using a time dependent closest point operator @xmath298 .
this lifting process will also be applied to the surface triangulation .
this will induce a further discrete material velocity @xmath299 which will describe how the lifts of triangles on @xmath300 evolve .
first , for a function @xmath301 , we define its lift , @xmath302 , implicitly , by : @xmath303 and , for a function @xmath105 , we define its inverse lift , @xmath304 by @xmath305 it is clear that these operations are inverses of each other @xmath306 furthermore , allows us to define a lifted triangulation @xmath307 of @xmath25 by @xmath308 this defines an exact triangulation of @xmath25 .
[ stability of lift ( * ? ? ?
* for @xmath309 ) ] [ lem : ch - norm - equiv ] let @xmath301 , with lift @xmath302 , be such that the following quantities exist . for @xmath310 ,
there exists @xmath311 , independent of @xmath228 , but depending on @xmath137 , such that for each time @xmath2 $ ] and each element @xmath312 with associated lifted element @xmath313 , the following hold : [ eq:79 ] @xmath314 this result allows us to give sobolev embeddings for discrete surfaces : [ lem : sobolev - embedh ] for @xmath227 as above , @xmath315 furthermore there exists a constant @xmath76 , independent of @xmath228 , such that for any @xmath316 @xmath317 to see the embedding result , we apply .
the bounds then follow using the stability of the lift ( ) .
we will write @xmath318 for the space of lifted finite element functions : @xmath319 this space comes with the standard approximation property : [ approximation property ] [ prop : ch - approx ] the lagrangian interpolation operator @xmath320 is well defined and , for @xmath321 , satisfies the bound @xmath322 let @xmath189 be such that @xmath80 embeds into @xmath323 , then @xmath324 the proof is given in @xcite for the case @xmath309 and can be easily extended using standard interpolation theory ( * ? ? ?
* theorem 3.1.6 ) to the case @xmath325 . for the remainder of the paper
, we will write lower case letters for the lift finite element functions with capital letters ( i.e. @xmath326 and @xmath327 ) and @xmath328 for the lift of @xmath329 . the motion of the edges of the simplices in the triangulation @xmath330 defines a discrete material velocity for the surface @xmath1 .
let @xmath93 be the trajectory of a point on @xmath225 with velocity @xmath331 .
we set @xmath332 then define @xmath299 by @xmath333 so that for @xmath334 , using , we have @xmath335 this defines another discrete material derivative for functions @xmath336 .
we define the discrete material derivative on @xmath36 element - wise by @xmath337 a quick calculation @xcite shows that for all @xmath269 , with lift @xmath338 , @xmath339 it can be shown , similarly to , that @xmath340
. we will write @xmath341 and @xmath342 for the lifts of the spaces @xmath343 and @xmath344 defined by .
it is clear that from that @xmath345 we remark that the continuous and discrete material velocities on @xmath300 only differ in the tangential direction .
this implies that the difference between the two material derivatives on @xmath1 only depends on the tangential gradient of the original function and not on any time derivatives .
these definitions also permit transport formulae : [ transport lemma for smooth triangulated surfaces ( * ? ? ?
* lemma 4.2 ) ] let @xmath300 be an evolving surface decomposed at each time into a family curved elements @xmath330 whose edges evolve with velocity @xmath299 .
then the following relations hold for functions @xmath346 such that the following quantities exist : @xmath347 and @xmath348 before showing stability of the finite element scheme , we will show a generalised gronwall inequality : let @xmath349 and satisfy the following differential inequality for @xmath350 @xmath351 let @xmath228 be sufficiently small so that @xmath352 , then @xmath353 satisfy the bound @xmath354 let @xmath355 and @xmath356 .
we note that @xmath357 ( @xmath358 ) and @xmath359 .
then @xmath242 satisfies @xmath360 this implies @xmath361 assuming that @xmath228 is sufficiently small so that @xmath362 , integrating this inequality in time implies @xmath363 rearranging this inequality gives the desired result .
we can now show the stability result in theorem [ thm : ch - fem - wp ] .
considering , since @xmath107 is positive definite , @xmath277 positive semi - definite and @xmath364 is locally lipschitz , standard theory of ordinary differential equations gives a unique short - time solution @xmath365 ; { \mathbb{r}}^n)$ ] for some @xmath366 . from , we know @xmath277 and @xmath107 are @xmath131 in time , and @xmath367 by the inverse function theorem .
thus , we infer @xmath368 ; { \mathbb{r}}^n).\ ] ] this is easily translated into solutions @xmath369 in the appropriate spaces .
since @xmath370 is an admissible test function in , it is clear that @xmath289 is conserved . to extend to the long - term solution , we construct an energy bound .
we start by testing with @xmath275 and with @xmath371 and sum to see @xmath372 applying the transport formulae from , we obtain @xmath373 next , we introduce the @xmath374 projection @xmath375 . for @xmath376 , we define @xmath377 as the unique solution of @xmath378 for @xmath379 , we will make use of the following bounds : @xmath380 these bounds follow since our triangulation is quasi - uniform .
we first note that from our assumptions on @xmath3 , we have @xmath381 with @xmath382 .
next , we test with @xmath383 and using and the sobolev embedding ( lemma [ lem : sobolev - embedh ] ) , we see that @xmath384 similarly , testing with @xmath275 leads to @xmath385 applying the geometric bound , the two previous bounds , a poincar inequality and the fact that the mass of @xmath274 is conserved , we infer that @xmath386 where @xmath387 is a constant which only depends on the integral of @xmath261 on @xmath221 .
this leads to the estimate @xmath388 we will use the generalised gronwall inequality from with @xmath389 , @xmath390 and @xmath391 .
given @xmath294 , there exists @xmath291 , @xmath292 , such that for @xmath296 , we have @xmath392 this gives the energy bound in with @xmath393 given by @xmath394 this implies , that if @xmath296 , we have an energy bound on @xmath148 and hence can turn the short - time existence result in to existence over @xmath148 where @xmath294 is arbitrary .
in this section , we will simply state the following geometric estimates without proof .
details can be found in ( * ? ? ?
* section 5 ) except for and which can be found in ( * ? ? ?
* lemma 3.3.14 ) .
let @xmath395 denote the quotient of surface measures @xmath63 on @xmath25 and @xmath249 on @xmath227 such that @xmath396 ; then @xmath397 } \sup_{\gamma_h(t ) } { \left| 1 - \mu_h \right| } & \le c h^2 \\
\label{eq : ch - mdh - muh - bound } \sup _ { t \in [ 0,t ] } { \left\| { \partial^\bullet}_h \mu_h \right\|}_{l^\infty(\gamma_h(t ) ) } & \le c h^2 . \end{aligned}\ ] ] [ lem : ch - geometric - bounds ] let @xmath398 with lifts @xmath399 .
then the following estimates hold for the given bilinear forms : @xmath400 using the same reasoning , it is also clear that @xmath401 similar results apply if the first argument is the material derivative of a finite element function : for @xmath402 with lifts @xmath399 for each time , we have @xmath403 the next lemma bounds errors from the approximation of @xmath3 by @xmath299 : the difference between the continuous velocity @xmath3 and the discrete velocity @xmath299 on @xmath25 can be estimated by @xmath404 this allows us to bound the error between the material derivatives on @xmath25 : suppose that @xmath405 and @xmath406 and @xmath407 exist .
for @xmath117 , we have the estimate @xmath408 and for @xmath409 , we obtain @xmath410 we conclude this section by constructing a discrete projection operator , similar to an interpolation operator .
we define the ritz projection operator , @xmath411 , of @xmath412 as the unique solution of @xmath413 and @xmath414 we will write @xmath415 for the lift of the ritz projection .
this operator is the ritz projection used by @xcite , but different to that used in other surface finite element analyses such as @xcite , which use the operator @xmath416 given as the unique solution of @xmath417 the following bounds are immediate : [ thm : pih - bound ] for @xmath412 , @xmath418 for @xmath321 , @xmath419 and for @xmath189 , such that @xmath80 embeds into @xmath323 , @xmath420 the @xmath421 stability result is clear and the @xmath0 error bound for a @xmath421 function follows from an aubin - nitsche trick .
the @xmath0 results for @xmath321 follow from standard error estimates for the surface finite element @xcite .
the @xmath422 result follows from the same splitting argument along with an inverse inequality .
the ritz projection is bounded in @xmath423 and we have the bound @xmath424 let @xmath425 be such that @xmath80 embeds into @xmath323 and such that @xmath70 embeds into @xmath426 . the previous result , a poincar inequality and imply that @xmath427 for @xmath228 sufficiently small .
we use a sobolev embedding ( ) , to see @xmath428 it is clear that @xmath429 which completes the proof . since @xmath430
, we also wish to have a bound on the discrete material derivative of this error for a function .
we will assume that @xmath321 and @xmath431 for each @xmath77 . under this assumption , we may take a time derivative of , so that for all @xmath271 with lift @xmath432 , @xmath433 in fact using similar arguments to , we can construct a similar extension of a finite element function @xmath269 to a function @xmath434 by @xmath435 hence , we deduce that applies at each time @xmath177 for @xmath269 .
we start by proving two technical lemmas : given @xmath436 with @xmath321 and @xmath431 for almost every time @xmath177 , then @xmath437 exists and we have the bound @xmath438 to show the bound , we start from , using a young s inequality , and gives @xmath439 applying this bound with @xmath440 gives the estimate . define the function @xmath441 on @xmath318 by @xmath442 then we have the bound @xmath443 furthermore , for any @xmath409 , we have that @xmath444 using and , we see for @xmath269 , with lift @xmath338 , @xmath445 using our bound on the ritz projection , and two geometric estimates and , we have that @xmath446 we can improve this estimate by comparing @xmath299 to the smooth velocity @xmath3 and introducing a smooth function @xmath409 .
then , we split the first term in @xmath447 into @xmath448 using the smoothness of @xmath59 , the final term , @xmath449 , is bounded using an integration by parts argument given by @xcite : @xmath450 hence , we obtain @xmath451 combining these calculations with and , we get @xmath452 hence , we have @xmath453 which is the second estimate . these results allow us to show an estimate for the difference between the material derivative of a function and its ritz projection .
[ lem : md - pih - bound ] for @xmath436 with @xmath454 , we have @xmath455 we start by rewriting the error as @xmath456 we can bound the first term on the right - hand side using by @xmath457 the second term is bounded using the approximation property : @xmath458 finally , we use our estimate of the difference of material derivatives to bound the third term : @xmath459 combining these three bounds in , we get the desired gradient norm bound for @xmath228 sufficiently small . to show the @xmath0 bound , we use the aubin - nitsche trick .
we start by writing @xmath460 , then @xmath461 is in @xmath0 so can be set as the right - hand side for the dual problem : find @xmath462 such that @xmath463 where @xmath464 .
we know @xcite that has a unique solution and satisfies the regularity result @xmath465 we note that from @xmath466 , that @xmath467 we remark that from and , we have @xmath468 and using , we infer @xmath469 this implies @xmath470 these calculations lead to @xmath471 the first term on the right - hand side is bounded using the approximation property and the gradient norm bound on @xmath461 , together with the dual regularity result : @xmath472 the second term is estimated using the improved bound on @xmath473 with @xmath474 . applying the approximation we see @xmath475 applying these two bounds in gives the desired result .
we use this section to show some properties of the continuous scheme based on the energy estimates coming from along with further some estimates .
we will use these properties in later sections but they are also important results in their own right . in order to derive some improved bounds on @xmath371 and @xmath275 , we will assume that @xmath476 with @xmath264 .
it is clear that assumption still holds in this case .
in fact , we will make use of the bound @xmath477 where @xmath478 this implies the constant @xmath479 from theorem [ thm : ch - fem - wp ] can be bounded by @xmath480 this is not essential for well - posedness of the finite element method but will be used for the well - posedness results for the continuous problem .
first , we need a bound on @xmath481 : [ lem : ch - fem - w0-bound ] under the assumption that @xmath264 , the following bound holds for @xmath481 : @xmath482 since @xmath483 are @xmath484;{\mathbb{r}}^n)$ ] in time ( ) , we known that holds at time @xmath485 .
we see that from the choice @xmath476 , using green s formula , we have @xmath486 this implies that @xmath487 in the last line we have used and the sobolev embedding of @xmath488 ( ) . from , we see that @xmath489 , { \mathbb{r}}^n)$ ] so @xmath490 exists .
hence , we may take the time derivative of to see , for @xmath271 , @xmath491 [ lem : ch - fem - bound - improved ] under the assumption that @xmath264 , we have the bound @xmath492 with @xmath493 given by @xmath494 we start by subtracting tested with @xmath275 from tested with @xmath495 and use the transport formula to arrive at @xmath496 note that using a hlder inequality , young s inequality with @xmath21 , and the sobolev embedding ( ) we have @xmath497 applying this estimate in , we have @xmath498 integrating in time using a gronwall inequality gives us @xmath499 applying the bounds from , and completes the proof . the idea of the existence proof is to show that the lift of the solutions to finite element scheme converges , along a subsequence , to a solution of the continuous equations .
we suppose that @xmath264 is a given function . in this section
, we will take @xmath476 with @xmath500 the ritz projection defined in . since the ritz projection is stable in @xmath421 , the stability bound in holds independently of @xmath228 .
furthermore , the stability bounds from imply we may transform this bound to @xmath1 and bound the lifts @xmath501 and @xmath502 by @xmath503 our assumption that @xmath264 allows the use of the improved bounds in . using similar lifting arguments we have @xmath504 these bounds , along with the conservation of mass property , imply that @xmath505 is uniformly bounded in @xmath506 and @xmath507 in @xmath163 .
hence , we may extract subsequences ( for which we will still use the subscript @xmath228 ) , and functions @xmath508 and @xmath509 with @xmath510 , and @xmath511 such that @xmath512 we remark that these results imply @xmath513 and @xmath514 weakly in @xmath160 .
furthermore , from the compactness result ( ) we infer that we may take a further subsequence ( still denoted @xmath505 ) such that @xmath515 using a dominated convergence theorem - type argument ( * ? ? ?
* lemma 8.3 ) , since @xmath516 is bounded independently of @xmath228 , we infer that @xmath517 we will show that @xmath508 and @xmath509 satisfy . for @xmath200
, we write @xmath518 , where @xmath500 is the ritz - projection , and @xmath519 .
using , we have @xmath520 and @xmath521 we may use the geometric estimates shown in section [ sec : geometric - estimates ] and the bounds on the ritz projection from section [ sec : ritz - projection ] to see to bound the terms on the right - hand sides of these equations .
we will denote by @xmath522 a generic constant depending on @xmath228 , which may also depend on @xmath21 , such that @xmath523 as @xmath524 .
integrating in time , this implies @xmath525 and @xmath526 we may send @xmath524 in the right - hand sides of both previous equations , and use the convergence results and , so that for all @xmath200 we arrive at @xmath527 finally , we use to transform this equality into a almost everywhere in time equality so that the pair @xmath528 satisfy . to show that @xmath508 achieves the initial condition , we start by choosing @xmath529 and continue with the notation @xmath530 . using the discrete transport formula ,
the lift of the finite element solution @xmath505 satisfies @xmath531 for all @xmath532 .
using similar limiting arguments as above , with the addition of , we obtain the identity @xmath533 in fact , by density of @xmath534 functions in @xmath154 ( * ? ? ?
* theorem 2.4 ) , we see that this equality holds for all @xmath212 .
this implies that @xmath535 is weakly differentiable as a function on @xmath148 with weak derivative @xmath536 . since @xmath537 ,
this weak derivative is a function in @xmath538 , and hence we infer that @xmath535 is absolutely continuous on @xmath539 $ ] ( * ? ? ?
* section 4.9 , theorem 1 ) .
in particular , @xmath540 is absolutely continuous , which means that we can interpret @xmath541 as an @xmath542 function .
the absolute continuity of @xmath543 for @xmath544 also implies that @xmath545 next , we choose @xmath544 with @xmath546 .
it is clear that @xmath200 , hence we can use the limiting equation and to see that @xmath547 we can do the same in the finite element scheme for @xmath518 , using the transport formula : @xmath548 the above calculations show that we are able to take the limit @xmath524 ( in the appropriate sense ) to see that @xmath549 therefore , by comparing terms , we have shown that @xmath550 almost everywhere in @xmath211 by the fundamental lemma of the calculus of variations .
hence we have shown the following result : [ thm : ch - wp ] given @xmath264 there exists a weak solution pair @xmath551 of the cahn - hilliard equation in the sense of .
furthermore the solution satisfies the energy bound @xmath552 to show the uniqueness result , we require an inverse laplacian on @xmath25 . for @xmath553 with @xmath554 , we define @xmath555 the inverse laplacian of @xmath556 as the unique solution of @xmath557
we will write @xmath558 and remark that @xmath559 it is clear that if @xmath553 then @xmath560 . we also have a similar result for the material derivative of @xmath555 . if @xmath561 , with @xmath554 , then @xmath562 .
it is clear that @xmath563 for @xmath564 .
it is left to show @xmath565 .
we start by taking a time derivative of so that for @xmath566 : @xmath567 from , given @xmath568 , we can construct @xmath174 , with @xmath569 for all @xmath2 $ ] and @xmath570 .
thus , we have that @xmath571 and , in particular , at @xmath182 , @xmath572 also , we have that @xmath573 these calculations imply that @xmath574 solves the elliptic problem : @xmath575 this implies that @xmath576 with the bound @xmath577 integrating in time gives the desired result .
there is at most one solution to .
we suppose that @xmath578 and @xmath579 are solutions to . we will write @xmath580 and @xmath581 . for @xmath200 , we know that @xmath582 testing with @xmath583 tells us that @xmath584 hence , since @xmath585 is well defined and @xmath586 , we may test the first equation with @xmath585 , and apply , to obtain @xmath587 next , using the monotonicity of @xmath588 , testing the second equation with @xmath589 gives @xmath590 taking the sum of and , we obtain @xmath591 for the first term on the right - hand side , we see that @xmath592 and for the second , we have @xmath593 combining these terms , we obtain the estimate @xmath594 we next use a gronwall inequality and integration in time to see @xmath595 since @xmath596 , we apply a poincar inequality to arrive at @xmath597 this shows that @xmath598 .
now , we know that @xmath599 and thus testing with @xmath600 gives @xmath601 this shows that @xmath602 . in this section , we show that the solution enjoys @xmath603 regularity .
[ regularity ] let @xmath264 and @xmath551 be the solution pair of , then @xmath604 and @xmath605 , with the bounds @xmath606 using the improved estimates from , we have that @xmath607 now , we can translate the fact that @xmath551 are solutions of into @xmath608 for @xmath609 and @xmath610 . notice that @xmath611 the above improved bounds combined with the bounds in gives @xmath612 and @xmath613 .
standard theory of elliptic partial differential equations @xcite gives @xmath604 and @xmath605 .
the proof is completed by using the bounds in and on @xmath614 and @xmath615 .
in this section , we show an error bound for the surface finite element method described in .
the proof relies on decomposing the errors into errors between the smooth solution and ritz projection and between the ritz projection and discrete solution .
in contrast to previous studies of partial differential equations on surfaces @xcite , we show an error bound on @xmath227 instead of @xmath25 .
this allows an easier treatment of the non - linear terms .
we will assume that @xmath616 and @xmath141 are bounded in the following norms @xmath617 shows how to bound some of these terms .
again , we will assume that the initial condition of the finite element scheme is given by the ritz projection : @xmath618 the error bound we will show is stated as follows : [ thm : ch - error ] let @xmath619 solve and satisfy .
let @xmath369 solve with initial condition .
we have that @xmath620 and @xmath621 with @xmath622 given by @xmath623 in the following error analysis , a pointwise bound on the discrete solution uniformly in space and time will be extremely useful .
this will allow us to convert the local lipschitz property of @xmath12 and @xmath624 into global results .
[ thm : ptwise - bound ] the discrete solution @xmath274 is bounded uniformly in space and time , independently of @xmath228 , and we have the bound @xmath625 let @xmath626 , then @xmath627 with the estimate @xmath628 this follows immediately from and combined with a sobolev inequality ( ) and . furthermore ,
since @xmath370 is an admissible test function in , the mean value of @xmath629 is zero : @xmath630 we define @xmath631 , so that @xmath632 let @xmath633 solve @xmath634 then it is clear that @xmath635 .
standard elliptic theory @xcite and the @xmath423 bound on @xmath500 gives that @xmath636 we apply this inequality uniformly in time , with , to give the desired estimate .
we split the error into two parts using the ritz projection @xmath500 from : @xmath637 we note that from , we already have estimates for @xmath638 and @xmath639 and it is left to bound @xmath640 and @xmath641 .
notice that , the assumptions in imply that @xmath642 and @xmath643 . to derive equations for @xmath640 and @xmath641 , we start by rewriting using the definition of @xmath500 and to obtain for @xmath271 with lift @xmath432 that @xmath644 next , we rewrite using this time to see for @xmath645 with lift @xmath646 that @xmath647 the quantities @xmath648 , for @xmath649 , are consistency terms involving the approximation properties of the finite element spaces and the geometric perturbation .
[ lem : e - bound ] for @xmath271 we have @xmath650 and for @xmath645 : @xmath651 the proof is a combination of the geometric bounds from and the bounds of @xmath500 from and . in this section ,
we derive bounds on @xmath640 and @xmath641 based on the error equations derived in the previous section and natural energy methods for the partial differential equation system go on to show the final error estimate . to bound @xmath640 and @xmath641
we start by testing with @xmath652 and with @xmath641 and subtract to see that @xmath653 applying and the transport lemma , with the local lipschitz property of @xmath624 , this result gives that @xmath654 we apply a young s inequality to find that @xmath655 next , in order to bound the @xmath656 term in the previous equation , we test with @xmath640 . using and and the @xmath423 bound on @xmath7 and @xmath274 , we have for some @xmath657 , @xmath658 applying this bound in the right - hand side of , we may choose @xmath659 small enough so that @xmath660 we recall from : @xmath476 , hence we have that @xmath661 .
applying a gronwall inequality and integrating in time gives the following bounds on @xmath640 and @xmath641 : @xmath662 with @xmath663 given by @xmath664 the previous bound can then be combined with the bounds on @xmath638 and @xmath639 from to give the @xmath0 error .
one can also apply an inverse inequality to derive gradient bounds on @xmath640 and @xmath641 to give the @xmath421 error bound .
the above finite element method discretised in time using semi - implicit time stepping . given @xmath261 and a partition of time @xmath665 , for @xmath666
, we find @xmath667 as the solution the matrix system @xmath668 full analysis of the fully discrete problem is left to future work .
based on ideas from @xcite , we expect stability subject to @xmath669 and convergence rate order @xmath670 for the discrete version of the norms in .
we start by showing the derived orders of convergence can be achieved for a fourth order linear problem .
we calculate with @xmath671 and choose @xmath672 .
we couple @xmath673 to ensure we see the full order of convergence .
the surface is given by @xmath674 with @xmath675 we have chosen @xmath676 and solve for @xmath677 .
the exact solution is given by @xmath678 , where right hand side @xmath679 is calculated from @xmath680 the convergence is shown in for the errors in the @xmath0 norm .
the experimental order of convergence ( eoc ) is calculated via the formula : given an error @xmath681 and @xmath682 at two different mesh sizes @xmath683 and @xmath684 , we calculate the experimental order of convergence ( eoc ) by @xmath685 the results for the @xmath421 norm are not shown here , however we observe first order convergence in @xmath228 .
we present two plots to show the behaviour of the numerical solution .
first , in , we see that for short times we have good convergence of the solution .
the second , , demonstrates that the energy does not decrease monotonically along solutions .
running for a longer time suggests that the solution converges to a time periodic solution .
we show a plot of the solution at level 2 at different times in .
the system is solved with a fixed time step of @xmath688 .
we take a surface given by the level set function @xmath691 where @xmath692 in addition , we will prescribe a tangential velocity so that we will consider points moving according to @xmath693 we plot the solution at different times in . in particular
, we notice that under this flow the nodes remain uniformly distributed .
dziuk , g. : finite elements for the beltrami operator on arbitrary surfaces . in : s.
hildebrandt , r. leis ( eds . ) partial differential equations and calculus of variations , _ lecture notes in mathematics _ , vol . 1357 , pp .
springer - verlag , berlin ( 1988 ) elliott , c.m . : the cahn - hilliard model for the kinetics of phase separation . in : j.f .
rodrigues ( ed . ) mathematical models for phase change problems , _ international series of numerical mathematics _
birkhuser , basel ( 1989 ) lubich , c. , mansour , d. , venkataraman , c. : backward difference time discretization of parabolic differential equations on evolving surfaces .
i m a j. numer .
first published online : march 28 , 2013 .
doi : imanum.drs044 ( 2013 )
|
we use the evolving surface finite element method to solve a cahn - hilliard equation on an evolving surface with prescribed velocity .
we start by deriving the equation using a conservation law and appropriate transport formulae and provide the necessary functional analytic setting .
the finite element method relies on evolving an initial triangulation by moving the nodes according to the prescribed velocity .
we go on to show a rigorous well - posedness result for the continuous equations by showing convergence , along a subsequence , of the finite element scheme .
we conclude the paper by deriving error estimates and present various numerical examples .
|
star formation ( sf ) was processing at a much larger rate in galaxies in the first half of the universe history , and the most striking feature in the cosmic sf rate density is the decline by a factor @xmath4 10 since z=1 ( madau 1998 , hopkins & beacom 2006 ) .
several factors could be invoked to explain such a behavior : first the gas fraction in star forming galaxies is likely to have been higher in the past , as already suggested by co surveys , tracing the molecular gas content of galaxies . locally , the gas fraction for giant spirals is about 7 - 10% ( leroy 2008 , saintonge 2011a ) , while at [email protected] it increases to 34@xmath55% and at [email protected] to 44@xmath56% ( tacconi 2010 , daddi 2010 ) .
second , the star formation efficiency might have been higher in the past , due to the dynamical trigger of galaxy interactions , whose frequency increases with redshift ( e.g. conselice 2009 , kartaltepe 2010 ) , and also the more violent instabilities in more unstable disks , with lower bulge - to - disk ratios .
the star formation efficiency ( sfe ) defined as the ratio of sfr to gas content , has been observed to increase with redshift ( e.g. greve 2005 ) , even for the most extreme starbursts , represented by ultra - luminous infrared galaxies ( ulirg ) .
this tendency is however supported mainly by comparing local and high - z galaxies , at z@xmath61 , and very little is known about the molecular gas content of galaxies at intermediate redshift between z=0.2 and 1 .
this co desert is mainly due to observational difficulties , and motivated our co survey of starburst galaxies in this redshift range .
a first study at 0.2@[email protected] ( combes 2011 ) has indeed confirmed a strong increase of sfe in this redshift range . for @xmath8/@xmath9 proposed by obreschkow & rawlings ( 2009 ) .
* right : * correlation between far infrared and co luminosities for the same data .
the 3 lines are for l@xmath10/m()=10 , 100 and 1000 / from bottom to top , assuming a conversion factor @xmath0 = 0.8 m@xmath1 ( k pc@xmath2)@xmath3 .
the three lines correspond to gas depletion time - scales of 580 ( bottom ) , 58 ( middle ) and 5.8 myr ( top).,title="fig:",width=226 ] for @xmath8/@xmath9 proposed by obreschkow & rawlings ( 2009 ) . *
right : * correlation between far infrared and co luminosities for the same data .
the 3 lines are for l@xmath10/m()=10 , 100 and 1000 / from bottom to top , assuming a conversion factor @xmath0 = 0.8 m@xmath1 ( k pc@xmath2)@xmath3 .
the three lines correspond to gas depletion time - scales of 580 ( bottom ) , 58 ( middle ) and 5.8 myr ( top).,title="fig:",width=226 ]
we have now completed our co survey of 69 starburst galaxies with 0.2@[email protected] , and obtained a global detection rate of 48% .
since the conversion factor is a key parameter in this study , we have obtained co detections in several j - lines for a dozen of galaxies , and found different excitation for the gas . when the gas mass could be derived from the dust emission however , the resulting values conforted our adoption of the @xmath0=0.8 conversion factor ( combes 2012 ) .
mapping of some galaxies with the iram interferometer showed that the molecular gas is extended at kpc scales , and not only confined to a nuclear component ( combes 2006 , and in prep . ) .
figure [ fig1 ] shows that the data gap at intermediate redshift is now filled .
the co luminosities of starburst galaxies display an envelope which is constantly rising with redshift in this interval .
the fir to co correlation is non - linear , and define gas depletion time between 6 and 600 myr ( fig [ fig1 ] right ) . .
the red curve is a schematic line summarizing the evolution of cosmic star formation density , from the compilation by hopkins & beacom ( 2006 ) , complemented with recent work by kistler ( 2009 ) and bouwens ( 2008 ) .
this indicative curve is logarithmic and can be translated vertically . *
right : * the gas to stellar mass ratio vs redshift , with the same data . the points of the high - z samples ( blue asterisks and green dots ) , have been continued by a dotted line joining the two extreme values of gas fraction , obtained with conversion factors @xmath0=0.8 and 4.6 m@xmath1 ( k pc@xmath2)@xmath3 .
, title="fig:",width=226 ] .
the red curve is a schematic line summarizing the evolution of cosmic star formation density , from the compilation by hopkins & beacom ( 2006 ) , complemented with recent work by kistler ( 2009 ) and bouwens ( 2008 ) .
this indicative curve is logarithmic and can be translated vertically . *
right : * the gas to stellar mass ratio vs redshift , with the same data .
the points of the high - z samples ( blue asterisks and green dots ) , have been continued by a dotted line joining the two extreme values of gas fraction , obtained with conversion factors @xmath0=0.8 and 4.6 m@xmath1 ( k pc@xmath2)@xmath3 .
, title="fig:",width=226 ] the gas fraction requires the determination of stellar mass , which was obtained through sed - fitting of optical and near - infrared luminosities ( combes 2012 ) .
we define the star formation efficiency as the ratio of far infrared luminosity ( proxy of star formation rate ) and the molecular mass derived from the co luminosity , with a constant conversion factor @xmath0= 0.8 m@xmath1 ( k pc@xmath2)@xmath3 .
the sfe and gas fraction are plotted versus redshift and compared to available data in fig [ fig2 ] .
both display an envelope which rises with redshift in the intermediate range between 0.2 and 1.0 , the range where the cosmic star formation rate density is remarkably increasing by an order of magnitude .
these trends are also visible in the bin - averages of the data , taking or not the upper limits into account ( cf fig [ fig3 ] ) . upper limits in black ( for high - z samples only ) .
the error bars are based on poisson noise .
the red line is the same as in figure [ fig2 ] .
, title="fig:",width=226 ] upper limits in black ( for high - z samples only ) .
the error bars are based on poisson noise .
the red line is the same as in figure [ fig2 ] .
, title="fig:",width=226 ]
the increase of both star formation efficiency and gas fraction with redshift in the intermediate range for starburst galaxies suggests that both factors play a role in driving the evolution of the cosmic star formation rate density . the relative influence of each physical quantity is strongly linked to the co - to- conversion factor .
if a standard milky way ratio is used for these galaxies , their gas fraction will then dominate the evolution .
we have tried to relate the star formation efficiency to the compactness of the starburst , measured from the half - light radius in the i - band ( corresponding to the blue band in the rest frame ) .
there is indeed an anticorrelation of sfe with half - light radius , but with a large scatter .
the sfe was also plotted with respect to the specific star formation rate ( or sfr per unit stellar mass ) , and compared to the locus of local `` normal '' star forming galaxies from the cold gass survey ( saintonge 2011b ) .
the starburst galaxies are globally located below , meaning that their gas content is significantly higher .
bouwens r.j . , illingworth g.d .
, franx m. , ford h. : 2008 , _ apj _ 686 , 230 chung , a. , narayanan , g. , yun , m. s. , heyer , m.,erickson , n.r . : 2009 , _ aj _ 138 , 858 combes , f. , garca - burillo , s. , braine , j. : 2006 , _ a&a _ 460 , l49 combes , f. , garca - burillo , s. , braine , j. : 2011 , _ a&a _ 528 , a124 conselice , c. j. , yang , c. , bluck , a.f.l . : 2009 , _ mnras _ 394 , 1956 daddi e. , bournaud , f. , walter , f. : 2010 _ apj _ 713 , 686 gao y. , solomon p.m. : 2004 _ _ apj__s 152 , 63 geach , j.e .
, smail i. , coppin k. : 2009 , _ mnras _ 395 , l62 geach , j.e . , smail i. , moran s.m . : 2011 , _ apj _ 730 , l19 genzel , r. , tacconi , l. j. , gracia - carpio , j. : 2010 , _ mnras _ 407 , 2091 greve .
t.r . , bertoldi , f. , smail , i. : 2005 , _ mnras _ 359 , 1165 hopkins , a.m. , beacom j.f . : 2006 , _ apj _ , 651 , 142 iono , d. , wilson , c. d. , yun , m.s . : 2009 , _ apj _ , 695 , 1537 kartaltepe , j. s.,sanders , d. b. , le floch , e. : 2010 , _ apj _ 721 , 98 kistler m.d .
, yksel h. , beacom j.f . : 2009 , _ apj _ 705 , l104 leroy , a. k. , walter , f. , brinks , e. 2008 , _ aj _ 136 , 2782 madau p. , pozzetti l. , dickinson m.e . : 1998 , _ apj _ 498 , 106 obreschkow , d. , rawlings , s. : 2009 _ apj _ 696 , l129 saintonge , a. , kauffmann , g. , kramer , c. : 2011a , _ mnras _ 415 , 32 saintonge , a. , kauffmann , g. , wang j. : 2011b , _ mnras _ 415 , 61 solomon p. , downes d. , radford s. , barrett j. : 1997 , _ apj _ 478 , 144 solomon p. , vanden bout p.a .
: 2005 , _ araa _ 43 , 677 tacconi l.j .
, genzel r. , neri r. : 2010 , _ nature _ 463 , 781
|
star formation is evolving very fast in the second half of the universe , and it is yet unclear whether this is due to evolving gas content , or evolving star formation efficiency ( sfe ) . we have carried out a survey of ultra - luminous galaxies ( ulirg ) between z=0.2 and 1 , to check the gas fraction in this domain of redshift which is still poorly known .
our survey with the iram-30 m detected 33 galaxies out of 69 , and we derive a significant evolution of both the gas fraction and sfe of ulirgs over the whole period , and in particular a turning point around z=0.35 .
the result is sensitive to the co - to- conversion factor adopted , and both gas fraction and sfe have comparable evolution , when we adopt the low starburst conversion factor of @xmath0 = 0.8 m@xmath1 ( k pc@xmath2)@xmath3 .
adopting a higher @xmath0 will increase the role of the gas fraction .
using @xmath0 = 0.8 , the sfe and the gas fraction for [email protected] - 1.0 ulirgs are found to be significantly higher , by a factor 3 , than for local ulirgs , and are comparable to high redshift ones .
we compare this evolution to the expected cosmic abundance and the cosmic star formation history .
|
superconductors exhibit quantum coherence of electronic degrees of freedom on a macroscopic scale . for @xmath2-wave pairing they are characterized by a complex pairing amplitude @xmath3 with a well - defined phase
@xmath4 @xcite .
this is a consequence of the off - diagonal long - range order present in the two - particle density matrix @xcite .
the electrons condense into cooper pairs and form entangled two - electron singlet states .
the prospect of harvesting these useful coherence and entanglement properties for the manipulation of quantum states as required for example in quantum information processing and communication @xcite is a motivation to integrate superconducting elements in semiconductor solid - state nanostructures .
prominent specific goals are the on - demand production of entangled photon pairs due to the recombination of cooper pairs @xcite and the generation of non - classical states of light @xcite .
it further brings about the fundamental question of how to efficiently transfer the electronic coherence and entanglement naturally present in a superconductor to excitonic particle - hole pairs in a semiconductor and eventually to photons that are emitted from the hetero - structure .
-@xmath1 junction coupled to superconducting ( sc ) leads operated under forward bias voltage @xmath5 .
basic recombination processes at low temperature @xmath6 : ( i ) transfer of particle from conduction to valence band upon radiation of a photon ( red ) with energy @xmath7 , ( ii ) absorption of a photon ( blue ) with energy @xmath8 upon transfer of a particle from valence to conduction band , and ( iii ) cooper pair tunneling upon emission or absorption of two photons ( green ) with energy @xmath9 .
the superconducting gaps are denoted @xmath10 ( @xmath11 ) for valence ( conduction ) band . ]
while the experimental realization of superconductor - semiconductor hybrid nanostructures has proven to be technology challenging @xcite , there have recently been a number of successful experiments using semiconductor nanowires @xcite , quantum wells @xcite and self - organized hetero - superstructures in unconventional superconductors @xcite .
various desired electronic and optoelectronic properties have been observed such as the proximity effect @xcite and the josephson effect @xcite , the realization of a superconducting field - effect transistor @xcite and enhanced electro- and photo - luminescence rates @xcite .
enhanced luminescence rates were observed in a series of experiments using superconductor - semiconductor hetero - structures where a @xmath0-@xmath1 junction was contacted with a superconducting niobium lead on the @xmath1-side of the junction and cooled below the superconducting transition temperature @xmath12 @xcite .
in addition , it was clearly shown via the josephson effect that due to the proximity effect cooper pairs tunnel into the active region of the @xmath0-@xmath1 junction , where they recombine with normal holes @xcite .
more recently , enhanced photo - luminescence below a superconducting transition temperature was reported in self - organized superconductor - semiconductor hetero - structures that form naturallly in the iron - based superconductor @xmath13 @xcite .
the exciting case where the semiconducting region only contains a few quantum dot states has been studied in refs . . here
, we investigate a superconductor - semiconductor hetero - structure consisting of a @xmath0-@xmath1 junction that is sandwiched between two superconducting leads .
the setup is shown in fig .
[ basicprocesses ] .
we focus on the properties of the light that is emitted from such a superconducting light - emitting diode ( led ) under forward bias in the steady - state . to reach a steady - state
we consider a coupling of the photons to an external photon bath .
one of our main results is a drastic enhancement of the electro - luminescence rate in a sharp spectral window in the presence of superconductivity .
this increase , which is shown in figs .
[ lum1 ] and [ lum2 ] is a consequence of the van - hove singularities at the edges of the superconducting bands .
similar behavior was observed in a setup with one normal and one superconducting lead in refs . .
we also report the formation of an additional luminescence peak on resonance that is effectively due to transitions of cooper pairs from the conduction to the valence band .
a similar rich peak structure was reported for the case of quantum dots coupled to superconducting leads in ref . .
further , we analyze the statistical properties of the emitted photons and predict non - classical , two - mode squeezed states of light .
photons with momentum @xmath14 and @xmath15 inherit their squeezing correlations from the electrons forming cooper pairs . as shown in fig .
[ squeezing ] , the uncertainty of the corresponding two - mode quadrature operators in the squeezed state falls below the minimum uncertainty of coherent states enforced by the heisenberg limit .
even more importantly , the orientation of the squeezing ellipse can be externally manipulated by changing the relative phase between the two superconductors that are coupled to the @xmath0-@xmath1 junction @xcite .
this shows that one may employ the coherence of a superconductor to tailor the properties of a two - photon quantum state via the transfer of cooper pair entanglement to the photons .
two - photon correlations in related models are theoretically discussed in refs . . finally , we discuss conditions for lasing in this system and find that the lasing threshold is reduced in the presence of superconductivity , again due to the formation of van - hove singularities .
the structure of the remainder of the paper is as follows : in the next sec .
[ basmodsec ] , we introduce the superconductor-@xmath0-@xmath1-superconductor setup and develop the many - body non - equilibrium field theory framework that we use to extract properties of the photons that are emitted in such a device under forward bias . in sec .
[ sec : effect - phot - acti ] , we derive the effective photonic action by integrating over the electronic degrees of freedom .
we calculate the photonic self - energy due to the coupling to the electrons up to one loop .
we discuss in detail the structure and the features of the self - energy both in the presence and in the absence of superconductivity . to obtain a steady state
, we consider the leakage of photons out of the system via a coupling to an external photon bath .
this yields another contribution to the photonic self - energy . inverting the dyson equation , we sum up the complete rpa series and
obtain our main analytical result : the dressed photon keldysh propagators .
in the following sections , we extract the physical information they contain about the photonic system . in sec . [ sec : lumin - sque - sled ] , we discuss the luminescence and the light squeezing properties and in sec . [
sec : steady - state - laser ] we derive the conditions to obtain lasing in the system .
we determine the lasing threshold also in a physically very transparent way using rate equations . to focus on our main results and the physical consequences in the main text , we shift details of the calculations that yield the effective photon action , the photonic self - energy and the dressed photon propagators to the appendix .
we consider a @xmath0-@xmath1 junction that is coupled on each side to a superconducting lead .
the system is set under an external forward bias voltage @xmath5 to operate as a led . due to the proximity effect cooper pairs tunnel into the active region of the junction , and participate in recombination processes that lead to the emission of light .
we model the active region of the junction together with the photonic degrees of freedom by the hamiltonian @xmath16 the hamiltonian of the conduction and the valence electron bands takes the form @xmath17 where @xmath18 creates for @xmath19 an electron with momentum @xmath20 and spin @xmath21 in the conduction ( valence ) band .
the band dispersions are given by @xmath22 which we will later assume to be of the form @xmath23 and @xmath24 with @xmath25 , _
i.e. _ , the conduction ( valence ) band is electron ( hole ) like . we use @xmath26 here and in the following .
the proximity induced bcs @xmath2-wave gaps are denoted by @xmath27 and the chemical potentials by @xmath28 .
we have included the chemical potentials into eq . to be able to work with time - independent gap functions @xmath27 .
this requires that we measure the electronic energies with respect to the two different chemical potentials @xmath28 in the two respective bands .
their difference is equal to the applied bias voltage @xmath29 which drives the system out - of - equilibrium .
formally , this is achieved by the gauge transformation @xmath30 .
this changes the dispersions from @xmath31 to @xmath32 .
in addition , this gauge transformation leads to a time - dependence of the electron - photon coupling constant @xmath33 that appears in the coupling hamiltonian @xmath34 the operator @xmath35 creates a photon with frequency @xmath36 .
the electron - photon coupling is such that the transfer of each electron from the upper conduction to the lower valence band leads to an emission of a single ( optical ) photon with frequency @xmath37 .
we restrict ourselves to the case of spin - conserving recombinations .
the generalization to circularly polarized photons is straightforward and does not yield any qualitatively new aspects to our model .
the free photon hamiltonian is given by @xmath38 since photons are constantly produced in the led due to the presence of the bias voltage , we need to include an external photon bath to deposit the energy and reach a steady - state .
this is described by the bath hamiltonian @xcite @xmath39 the photon bath operator @xmath40 creates a photon with frequency @xmath41 in the bath .
the coupling constants between system and bath photons are denoted by @xmath42 .
we will later integrate over the bath degrees of freedom leading to the possibility to dissipate a ( system ) photon to the external bath and to create a ( system ) photon from the bath . due to the applied bias voltage the system that we are considering is in a non - equilibrium state . in the following we want to set up the required non - equilibrium keldysh field theory formalism that is used to determine observable properties of the system such as the electro - luminescence or the squeezing of the emitted light .
the small parameter that controls our ( infinite - order ) perturbative calculation is the electron - photon coupling constant @xmath43 .
more precisely , the small dimensionless parameter is given by @xmath44 for normal conducting leads and @xmath45 for superconducting leads .
here , @xmath46 denotes the electronic density of states at the leads fermi energy @xmath47 ( see fig .
[ basicprocesses ] ) .
we further assume that we are always below the lasing threshold and derive parametric conditions where lasing occurs later ( see sec . [
sec : steady - state - laser ] ) . in non - equilibrium
it is required to begin and end the time evolution at the initial state of the system .
one thus formulates the field theory on a closed time contour @xmath48 such that the action reads @xcite @xmath49 \,.\end{aligned}\ ] ] here , @xmath50 denote grassmann fields and @xmath51 denote complex fields arising in a path integral formulation of the action .
one can now clearly observe how the applied bias voltage appears by performing the gauge transformation introduced above @xmath30 .
the time derivative then produces additional terms that contain the chemical potentials such that the energy dispersions are measured with respect to the two different chemical potentials @xmath52 and @xmath33 .
we suppress the tilde notation from now on .
the contour @xmath48 starts ( and ends ) at the initial time @xmath53 , where we assume that @xmath54 and the system is in thermal equilibrium at potentially different temperatures for the electronic and photonic subsystems .
at @xmath55 we consider a non - zero electron - photon coupling @xmath43 and a non - zero bias voltage @xmath56 leading to photon production . in the following ,
we focus on steady - state properties of the system at times @xmath57 where @xmath58 is a characteristic time - scale over which transient effects associated with the switch - on decay .
we will always check self - consistently that our assumption of a steady - state holds .
it is convenient to introduce fermionic and bosonic spinors @xmath59 where @xmath60 denotes that the time variable is located on the forward @xmath61 or the backward @xmath62 branch of the contour @xmath48 .
the @xmath3 superscript denotes the fermionic combined nambu - conduction / valence space and the @xmath63 superscript describes the photon particle - hole space , in which the complex bosons can be described with real spinors , see eq . .
it is convenient to make a transformation from the contour @xmath64 basis to the rak ( retarded - advanced - keldysh ) basis . for bosons this transformation to the classical and quantum fields
is given by @xmath65 for fermions we follow larkin and ovchinnikov @xcite and perform the transformation to @xmath66 fields as @xmath67 the conjugate grassmann variable fields @xmath68 are not related to the @xmath69 fields and one may thus choose a different transformation for them @xmath70 combining the two fields as usual we introduce the keldysh vectors @xmath71 and @xmath72 .
we denote these vectors by @xmath73 superscripts .
we can then write the keldysh action as @xmath74 the factor of @xmath75 in front of the bosonic propagator arises from our choice of `` real '' bosonic spinors .
the unperturbed fermionic green s functions are given by @xmath76 the retarded and advanced blocks , which only depend on the time difference @xmath77 , explicitly read @xmath78^\dagger } \
, \rangle}_0 \\ & \quad = { \ensuremath{\left ( \begin{matrix } g_{0,{\boldsymbol{k}},v}^{(p),r / a } & 0 & p_{0,{\boldsymbol{k}},v}^{r / a } & 0 \\ 0 & g_{0,{\boldsymbol{k}},c}^{(p),r / a } & 0 & p_{0,{\boldsymbol{k}},c}^{r / a } \\
\bar p_{0,{\boldsymbol{k}},v}^{r / a } & 0 & g_{0,{\boldsymbol{k}},v}^{(h),r / a } & 0 \\ 0 & \bar p_{0,{\boldsymbol{k}},c}^{r / a } & 0 & g_{0,{\boldsymbol{k}},c}^{(h),r / a } \end{matrix } \right ) } } _ \tau \nonumber \,.\end{aligned}\ ] ] the average @xmath79 is with respect to the free action @xmath80 .
the particle and hole propagators are denoted by @xmath81 and @xmath82 . the anomalous electronic propagators which arise due to the presence of superconductivity
are denoted by @xmath83 and @xmath84 .
after fourier transformation @xmath85 they take the explicit form @xmath86 with dispersions @xmath87 for @xmath88 and pauli matrices @xmath89 where @xmath90 .
since the electrons are assumed to be in thermal equilibrium , one can express the keldysh green s function via the fluctuation - dissipation theorem as @xmath91 .
\label{mod6}\end{aligned}\ ] ] with fermionic distribution function @xmath92 at temperature @xmath93
. the free green s function for the photons in the system described by @xmath94 can be written as @xmath95 note that the expectation value contains the transpose spinor @xmath96 ( and not the hermitian conjugate ) .
the retarded and advanced blocks take the form @xmath97^t } \
, \rangle}_0 \nonumber \\ & = { \ensuremath{\left ( \begin{matrix } 0 & d_{0,{\boldsymbol{q}}}^{r / a}(t - t ' ) \\ d_{0,{\boldsymbol{q}}}^{a / r}(t'-t ) & 0 \end{matrix } \right ) } } \,.\end{aligned}\ ] ] where @xmath98^{-1}= \omega - \omega_{{\boldsymbol{q } } } \pm i 0 $ ] .
since the photonic subsystem is initially in equilibrium , we may write the keldysh component via the fluctuation - dissipation theorem as @xmath99 \ , . \label{mod7}\end{aligned}\ ] ] here , @xmath100 denotes an initial bosonic distribution function of the uncoupled ( @xmath101 ) system .
as shown below the distribution function @xmath100 will not be important in our calculation of the photon distribution in the steady state .
this is rather determined by the interplay between the photon distribution function in the external photon bath and the photon production rate in the led .
the electron - photon coupling part in the action contains the vertex expression @xmath102 with matrices @xmath103 and @xmath104 in keldysh space . using eq .
, the components of the photon field vector read explicitly @xmath105 and @xmath106 .
the coupling matrices take the form @xmath107 and @xmath108 with @xmath109 and @xmath110 denoting the complex conjugate .
finally , the part of the action that describes the coupling of the photons to the photon bath reads @xmath111 \,.\end{aligned}\ ] ] we will later in sec . [ sec : infl - phot - bath ] integrate over the external photon modes @xmath112 under the standard assumptions of an ohmic bath in the white noise limit of frequency independent couplings @xmath42 and bath density of states @xcite .
in this section , we want to derive an effective action for the photonic degrees of freedom in the system @xmath113 which takes the coupling to the superconducting leads as well as to the photon bath into account .
formally , we integrate over the fermion fields @xmath114 and the external bath photon fields @xmath115 . this integration yields an electronic contribution to the photon self - energy @xmath116 and a bath contribution to the self - energy @xmath117 .
the effective photon action takes the form @xmath118 \hat \phi_{-{\boldsymbol{q}}}(t ' ) \ , .
\label{seff1}\end{aligned}\ ] ] while we treat the bath self - energy exactly , we take the electronic self - energy up to one loop into account .
importantly , due to the presence of superconductivity there exist non - zero anomalous fermionic green s functions @xmath119 .
this leads to anomalous terms in the bosonic self - energy @xmath116 , which induce similarly anomalous photon expectation values such as @xmath120 . in the following ,
we first derive the photon self - energies and then solve the resulting dyson equation in the steady state .
this corresponds to treating the electron - photon coupling in the random - phase approximation ( rpa ) .
it provides us with explicit expressions for the photon propagators - both for normal conducting and for superconducting leads .
details of the calculation are shifted to appendices [ appepa ] , [ appelse ] and [ appendixdressedpropagators ] . due to the coupling to conduction ( @xmath121 ) and valence ( @xmath122 ) electrons ,
see eq . .
external photon propagators ( wiggly lines ) are not part of the self - energy .
each vertex is associated with a coupling constant @xmath123 .
wiggly lines denote photons , solid ( dashed ) lines denote conduction ( valence ) electron propagators .
the anomalous contributions @xmath124 appear only for superconducting leads @xmath125 . ]
let us first focus on the electronic contribution to the effective action in eq .
( [ seff1 ] ) .
it is described by the self - energy @xmath116 and contains the information about the photon absorption and emission processes that involve the transition of electrons between conduction and valence bands .
the electronic part of the photon self - energy has the following structure in the particle - hole basis of @xmath126 : @xmath127 where @xmath77 denotes the time difference .
this structure is identical for the retarded , advanced and the keldysh components of the self - energy .
the anomalous components on the diagonal , which are associated with @xmath128 and @xmath129 , depend on the absolute time @xmath130 via the phase @xmath131 here , @xmath132 denote the ( constant ) phases of the superconducting gaps @xmath133 , which depend on microscopic details at the initial time @xmath134 , and @xmath135 is the phase of the coupling constant @xmath136 .
the remaining parts @xmath137 and @xmath138 , as well as the normal components on the off - diagonals @xmath139 , only depend on the time difference @xmath77 .
they are also independent of the phases @xmath140 .
we treat these self - energies in the one - loop approximation .
as shown in detail in appendix [ sec : integr - out - electr ] the self - energies are given by @xmath141_{ij,{\boldsymbol{q}}}^{\alpha \beta}(t , t ' ) = \\ & \qquad - i \sum_{{\boldsymbol{k } } } { \text{tr}}\bigl [ \hat \gamma^\alpha { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) \hat g_{0,{\boldsymbol{k}}}(t , t ' ) \hat \gamma^\beta { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) \hat g_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}(t',t ) \bigr ] \nonumber \,.\end{aligned}\ ] ] the corresponding feynman diagrams are shown in fig . [ seel1 ] .
the remaining task is to insert the expression for the fermionic green s functions , perform the traces over the keldysh and nambu indices and to carry out the summation over electronic momenta @xmath20 . before we describe the explicit result for the self - energies in the next sections , we want to discuss the different energy scales that are present in the problem . to give a numerical estimate of the different scales
, we use realistic values of gaas .
there are five important energy scales in the problem : ( i ) the applied bias voltage @xmath142 , which is of the order of the semiconducting bandgap @xmath143 .
this is the largest energy scale in the problem .
it sets the photon energy @xmath144 and the photon momentum @xmath145 , where @xmath121 denotes the speed of light . for gaas
this energy scale is given by @xmath146 ev ; ( ii ) the semiconductor fermi energy @xmath147 .
it is determined by the effective masses @xmath148 and the carrier density @xmath149 as @xmath150 . in gaas one
finds @xmath151 and two hole bands with @xmath152 and @xmath153 , where @xmath154 is the bare electron mass .
typical values for the carrier density are @xmath155 . for this range of doping and approximately using @xmath156 , we find @xmath157 ev . the fermi velocity follows to @xmath158 ; ( iii ) the ( proximity induced ) superconducting gap @xmath159 for which we estimate @xmath160 ; ( iv ) the strength of the electron - photon coupling @xmath161 with conduction band density of states @xmath162 .
for the range of carrier densities above one finds the estimate @xmath163 ev @xcite .
this is the smallest energy scale in the system , which justifies our perturbative approach ; ( v ) the coupling to an external photon bath induces a photon decay rate @xmath164 ( for each photon mode @xmath165 ) .
our assumption of a steady - state with a finite photon number requires that @xmath164 must be larger than the led photon production rate .
a result of this estimate is that one can not neglect the photon momentum @xmath165 .
although it is much smaller than typical electronic momenta @xmath166 , its associated energy scale is of the same order as the superconducting gap @xmath167 in the following , we discuss the explicit result for the electronic part of the photon self - energy @xmath168 .
we first describe the case of normal conducting leads and then the case of superconducting leads . taking a finite photon momentum into account
serves as a physical cutoff for divergences that would otherwise occur in the imaginary part for the superconducting case . , @xmath169 denotes the fermi velocity , @xmath170 the fermi momentum , @xmath171 the filling factor of the bands and @xmath142 is the applied bias voltage . electronic transitions involving photon emission ( absorption ) are possible at photon energies @xmath172 ( @xmath173 ) .
photon momentum @xmath165 is properly taken into account and results in electronic transitions that are not vertical . in the presence of superconductivity the electrons at the fermi energies
are gapped out , resulting in allowed transitions for @xmath174 . ]
let us first discuss the electronic contribution to the photon self - energy @xmath168 in the case of normal conducting leads . in this case ,
the diagonal elements in eq . , which are proportional to the product @xmath175 , vanish : @xmath176 . at fixed momentum @xmath165 as a function of frequency @xmath177
for normal conducting leads the imaginary part is shown in dashed green .
the self - energy changes sign at @xmath178 ; positive ( negative ) sign corresponds to photon production ( absorption ) . for superconducting
leads the imaginary part ( blue solid ) exhibits jumps at @xmath179 ( here @xmath180 ) , remains constant over an interval of width @xmath181 and then falls off towards the normal state solution . for superconducting
leads the real part ( red dashed ) exhibits logarithmic divergences at the position of the jumps in the imaginary part . ] to be able to perform the summation over momenta in the off - diagonal elements analytically , we make the symmetric choice of assuming parabolic conduction and valence bands bands with effective masses @xmath182 and equal superconducting gap amplitudes @xmath180 . as a result ,
the electronic dispersions fulfill @xmath183 . the fermionic density of states ( dos ) @xmath184 is only weakly energy dependent around the fermi energy , where @xmath185 corresponds to the dos at @xmath186 .
this situation is schematically depicted in fig .
[ bandsnc ] . for general dispersions
the summation can easily be performed numerically . as shown in detail in appendix [ appelse ] , one finds for the retarded and advanced self - energies at @xmath187 @xmath188 here , the frequencies @xmath189 are measured relative to the applied voltage .
the retarded function is shown in fig .
[ normalse ] .
finite fermionic temperatures @xmath190 will smear the zero temperature results , but do not yield qualitatively different results . as the fermions are assumed to be in equilibrium at temperature @xmath93 , the corresponding keldysh self - energies are given by @xmath191 \\ \pi_{12,{\boldsymbol{q}}}^{\text{el},k}(\omega ) & = \coth\bigl(\frac{\omega_+}{2t_f } \bigr ) \bigl[\pi_{12,{\boldsymbol{q}}}^{\text{el},r}(\omega)-\pi_{12,{\boldsymbol{q}}}^{\text{el},a}(\omega ) \bigr ] \ , .
\end{split } \label{eq:18}\end{aligned}\ ] ] as can be seen easily the keldysh self - energies obey the symmetry @xmath192 . in eq .
we have neglected the real part which depends only weakly on frequency .
it gives an unimportant renormalization of the photon resonance frequency @xmath193 .
the imaginary part of the self - energy @xmath194 describes the production and decay of photons due to the coupling to the electrons . from fig .
[ bandsnc ] we see that for @xmath195 there are two possible transitions : ( i ) photons with energy @xmath196 are emitted due to transitions of an electron in the conduction band to the valence band , or ( ii ) photons with energy @xmath197 are absorbed by raising an electron from the valence band to the conduction band .
the absorption is associated with a negative imaginary part of the retarded self - energy in eq . , while the emission is associated with a positive imaginary part .
the linear dependence of @xmath198 around @xmath178 arises due to the restricted phase space of decay and absorption processes for photons with energy @xmath199 . for larger energies @xmath200 on the other hand , the self - energies reach a constant value because we linearize around the fermi energies . as a function of frequency @xmath177 .
real part ( red dashed ) and imaginary part ( blue solid ) are non - zero only for superconducting leads and exhibit features similar to the normal self - energy ( see fig .
[ normalse ] ) .
note that features occur around @xmath201 here . ]
we have seen that the production of photons is described by a retarded self - energy @xmath194 with a positive imaginary part for frequencies @xmath202 .
it changes sign at the applied bias voltage @xmath142 . if there was no further contribution to the total photon self - energy this would be nonphysical , since it does not correspond to the analytic structure required by a retarded bosonic self - energy , which must have a negative imaginary part for @xmath203 .
physically , this corresponds to the fact that the assumption of a steady - state breaks down if photons are produced at a constant rate but there is no photon decay mechanism considered .
the photon number would simply grow without bound .
therefore , in sec . [
sec : infl - phot - bath ] we take a coupling of the photons to an external photon bath into account .
the associated bath induced photon self - energy @xmath204 will be added to the electronic contribution to the self - energy .
the total self - energy will then have a negative imaginary part at positive frequencies @xmath203 as required . in the superconducting state the anomalous diagonal elements of the photon self - energy in eq .
are non - zero .
they depend on the total time @xmath205 and it is thus convenient to perform a wigner transformation @xmath206 , which yields @xmath207 with phase @xmath208 .
the self - energy elements obey the relation @xmath209 and @xmath210 .
as shown in appendix [ appsup ] , the one - loop self - energies at @xmath211 take the form @xmath212 \\
\pi_{21,{\boldsymbol{q}}}^{\text{el},r / a}(\omega)&= 2 { \left|g_0 ^ 2 \right|}\sum_{{\boldsymbol{k } } } \biggl [ \frac{v_{{\boldsymbol{k}},v}^2 u_{{\boldsymbol{k}}+{\boldsymbol{q}},c}^2 } { \omega_--e_{v}({\boldsymbol{k}})-e_{c}({\boldsymbol{k}}+{\boldsymbol{q}})\pm i0 } -\frac{u_{{\boldsymbol{k}},v}^2 v_{{\boldsymbol{k}}+{\boldsymbol{q}},c}^2}{\omega_-+e_{v}({\boldsymbol{k}})+e_{c}({\boldsymbol{k}}+{\boldsymbol{q}})\pm i0 } \biggr ] \ , , \label{nse2.0}\end{aligned}\ ] ] where @xmath213 and @xmath214 are the superconducting coherence factors for the conduction and valence band .
they contain the bogoliubov quasi - particle dispersion relation @xmath215 .
the normal component is shown in fig .
[ normalse ] and the anomalous component in fig .
[ anomalousse ] . from the retarded self - energies
@xmath216 we easily get the keldysh self - energies as @xmath217\ , , \nonumber \\
\pi_{21,{\boldsymbol{q}}}^{\text{el},k}(\omega ) & = \pi_{12,{\boldsymbol{q}}}^{\text{el},k}(-\omega)= \coth\bigl(\frac{\omega_-}{2t_f } \bigr ) \bigl [ \pi_{21,{\boldsymbol{q}}}^{\text{el},r}(\omega)- \pi_{21,{\boldsymbol{q}}}^{\text{el},a}(\omega ) \bigr ] \ , , \label{nse8 } \end{aligned}\ ] ] let us discuss the retarded self - energies in some detail .
the combination @xmath218 and the diagonal elements @xmath219 are therefore proportional to the product @xmath220 , _ i.e. _ , they are non - zero only if both leads exhibit superconductivity . in the following ,
we assume for convenience a momentum independent gap function which is identical for the two bands @xmath221 .
note that the phases of the superconducting order parameters @xmath222 and @xmath223 are factored out explicitly in eq . .
both diagonal and off - diagonal elements of the self - energy show similar behavior .
their imaginary part vanishes in a region of width @xmath224 .
while the normal components are zero for @xmath225 the anomalous components vanish for @xmath226 . at the border of these regions
the functions exhibit a jump in the imaginary part , then they remain constant over a frequency window of the order of @xmath227 and finally decay towards the result for normal conducting leads ( see eq . ( [ nse1 ] ) ) further away from the gapped region . explicitly , the real ( imaginary )
part of the normal self - energy @xmath198 is ( anti)symmetric around the region @xmath228 .
the imaginary part is zero for @xmath229 . at the border it exhibits a jump of size @xmath230 here and in the following we consider @xmath231 , which holds for our realistic choice of parameters ( see sec . [
sec : energy - scales - system ] ) .
a jump in the imaginary part yields via the kramers - kronig relations a logarithmic divergence in the real part at @xmath232 .
if we define the functions @xmath233 \ , , \nonumber \\
i(w,\chi ) & = - \frac{\pi}{2 } { \left|g_0 \right|}^2 \frac{\rho_f}{\chi } \theta({\left|w \right| } - 2 ) \arcsin \biggl [ \frac{\min[\chi,\sqrt{{\left|w \right|}-2 } } { \sqrt{{\left|w \right| } - 2 } } \biggr ] \ , , \label{nse3 } \end{aligned}\ ] ] the self - energy for frequencies @xmath234 can be expressed as @xmath235 since @xmath236 ( see sec .
[ sec : energy - scales - system ] ) , the real part is only important around its logarithmic divergence .
it quickly decays towards the normal conducting result away from the resonance .
the imaginary part of @xmath237 is given by eq .
close to the jump and also approaches the constant result that we found for normal conducting leads ( see eq . ) .
the vanishing of the imaginary part for @xmath225 can easily be understood if we look at fig .
[ basicprocesses ] . for zero temperature
there is no possibility for one photon to be either absorbed or emitted within this energy range @xmath238 due to the superconducting gaps in the conduction and the valence bands . on the other hand , in a superconductor the dos diverges at energies @xmath239 relative to the fermi energy and there are thus many electronic states leading to an enhanced emission and absorption of photons coupling to those states . the imaginary part of the self - energy is therefore enhanced in this region compared to the normal conductor .
we note that it is absolutely essential that we take the finite photon momentum @xmath165 into account .
otherwise , the imaginary part would exhibit a square - root divergence at @xmath240 .
this divergence is cut - off by finite @xmath165 at @xmath241 leading to a finite jump instead .
since the imaginary part corresponds to the photon production rate in the superconducting led , this rate would diverge if one neglects the photon momentum .
the anomalous diagonal components of the self - energy @xmath242 are shown in fig .
[ anomalousse ] .
close to @xmath243 they can also be expressed by the functions defined in eqs.([nse3 ] ) as @xmath244 away from @xmath245 the imaginary part decays like @xmath246 to zero .
this is a faster decay than predicted by eq .
( [ nse4b ] ) . at finite but small temperatures @xmath247
, these results acquire small corrections such as as an exponentially suppressed imaginary part in the gapped regions . at higher temperatures
the described features are suppressed , but this also leads to the breakdown of superconductivity and one approaches the results for normal conducting leads . to achieve a steady - state in the system it is required that the photons , which are produced by electrons making a transition from conduction to valence band , may also be absorbed .
we thus consider the coupling to an external photon bath as a decay mechanism @xcite .
as shown in appendix [ couplingtophotonpath ] , the coupling to the bath ( see eq . ) gives rise to an additional contribution to the photon self - energy , which corresponds to a finite photon lifetime @xmath248 .
if we neglect the unimportant real part , which has no divergent features , the bath induced photon self - energy reads @xmath249 with @xmath250 we assumed that the coupling @xmath251 depends only on the photon frequency and introduced the dos of the external photon bath @xmath252 .
the imaginary part is determined by the spectral function of the external bath , which we assume to be of ohmic form @xmath253 with @xmath254 .
the spectral function is constant @xmath255 for @xmath256 and a vanishing function for @xmath257 .
the fact that it decays to zero at small frequency is important since otherwise the keldysh self - energy @xmath258 with component @xmath259 \end{aligned}\ ] ] would diverge as @xmath260 for small @xmath177 . here
, we have assumed that the external photon bath is in thermal equilibrium with photon temperature @xmath261 .
one can easily incorporate a different external photon distribution by replacing @xmath262 with an arbitrary bath distribution function @xmath263 , where @xmath264 denotes the number of bath photons in a state of energy @xmath177 . to calculate observable quantities such as the luminescence or the statistical properties of the light emitted from the superconducting led , we need to find the dressed photon propagator @xmath265 .
the photon self - energies @xmath266 and @xmath267 determine the dressed photon propagator @xmath265 via the dyson equation , which reads in keldysh space @xcite @xmath268 where @xmath269 . here
, @xmath63 denotes a convolution in time .
explicitly , this corresponds to three coupled integral equations for the retarded , advanced and keldysh components of the dressed propagator @xmath270^{-1}-{\mathring}\pi_{{\boldsymbol{q}}}^{r / a } \bigr ) \circ { \mathring}d_{{\boldsymbol{q}}}^{r / a } & = { \mathring}{\openone } \\ \label{dpp3 } \bigl ( [ { \mathring}d_{0,{\boldsymbol{q}}}^{r}]^{-1}-{\mathring}\pi_{{\boldsymbol{q}}}^{r } \bigr ) \circ { \mathring}d_{{\boldsymbol{q}}}^k & = { \mathring}\pi_{{\boldsymbol{q}}}^{k } \circ \hat d_{{\boldsymbol{q}}}^a \ , .
\end{aligned}\ ] ] one can omit the infinitesimal component @xmath271^k\sim i0 $ ] , which acts as a regularization in the non - interacting system , since the coupling to the electrons and the bath induces a finite keldysh self - energy .
since the anomalous components of the self - energy @xmath272 depend on the absolute time @xmath273 , we perform a wigner transformation with the ansatz for the full propagator ( see appendix [ appendixdressedpropagators ] for details ) @xmath274 with the phase factor given in eq . .
the dependence on the absolute time is similar but of opposite sign to the one found in the self - energy matrices .
the explicit forms of the full retarded and advanced photon propagators read @xmath275\bigl[\omega - e v_0+\omega_{{\boldsymbol{q } } } + \pi_{12,{\boldsymbol{q}}}^r(\omega - ev_0 ) \bigr]+ \bigl [ \tilde \pi_{11,{\boldsymbol{q}}}^r(\omega)\bigr]^2 } \\
\label{eq:21 } d_{12,{\boldsymbol{q}}}^r(\omega ) & = d_{21,{\boldsymbol{q}}}^a(-\omega ) = \frac{\omega-2 e v_0+\omega_{{\boldsymbol{q } } } + \pi_{12,{\boldsymbol{q}}}^r(\omega-2ev_0)}{\bigl[\omega-\omega_{{\boldsymbol{q } } } - \pi_{21,{\boldsymbol{q}}}^r(\omega ) \bigr]\bigl[\omega-2e v_0+\omega_{{\boldsymbol{q } } } + \pi_{12,{\boldsymbol{q}}}^r(\omega-2ev_0 ) \bigr]+ \bigl [ \tilde \pi_{11,{\boldsymbol{q}}}^r(\omega - v)\bigr]^2}\end{aligned}\ ] ] the keldysh components are given by @xmath276^ * & = \tilde
d_{11,{\boldsymbol{q}}}^r(\omega)\bigl [ \tilde \pi_{11,{\boldsymbol{q}}}^k(\omega ) \tilde d_{11,{\boldsymbol{q}}}^a(\omega ) + \pi_{12,{\boldsymbol{q}}}^k(\omega - v ) d_{21,{\boldsymbol{q}}}^a(\omega - v ) \bigr ] \nonumber \\ & \qquad + d_{12,{\boldsymbol{q}}}^r(\omega+v ) \bigl [ \pi_{21,{\boldsymbol{q}}}^k(\omega+v ) \tilde d_{11,{\boldsymbol{q}}}^a(\omega ) + \tilde\pi_{22,{\boldsymbol{q}}}^k(\omega ) d_{21,{\boldsymbol{q}}}^a(\omega - v ) \bigr ] \\
d_{12,{\boldsymbol{q}}}^k(\omega ) = -[d_{21,{\boldsymbol{q}}}^k(-\omega ) ] ^ * & = \tilde d_{11,{\boldsymbol{q}}}^r(\omega - v ) \bigl[\tilde \pi_{11,{\boldsymbol{q}}}^k(\omega - v ) d_{12,{\boldsymbol{q}}}^a(\omega ) + \pi_{12,{\boldsymbol{q}}}^k(\omega-2v ) \tilde d_{22,{\boldsymbol{q}}}^a(\omega - v ) \bigr ] \nonumber\\ & \qquad + d_{12,{\boldsymbol{q}}}^r(\omega)\bigl [ \pi_{21,{\boldsymbol{q}}}^k(\omega ) d_{12,{\boldsymbol{q}}}^a(\omega ) + \tilde \pi_{22,{\boldsymbol{q}}}^k(\omega - v ) \tilde d_{22,{\boldsymbol{q}}}^a(\omega - v ) \bigr ] \label{prop}\end{aligned}\ ] ] by inverting the dyson equation we have summed up the complete rpa series of bubble diagrams with the electronic contribution to the photon self - energy given in fig . [ seel1 ] and
the bath contribution given by eq . .
in fact , the expressions for the dressed propagators are formally exact if the self - energy was known exactly .
this follows from the fact that the structure of the self - energy in eq
. holds to all orders in perturbation theory .. by inverting the dyson equation we have only employed this general structure . from eq .
we find that the photon particle propagator @xmath277 shows features at @xmath278 and @xmath234 . in the case that the photon is off - resonant with the applied voltage , which is determined by the semiconductor bandgap , @xmath279 the photon can just propagate through the system and is only weakly interacting with the electrons .
in contrast , if the photon is resonant @xmath280 , the enhanced spectral weights around @xmath234 correspond to photon - exciton bound states or polaritons . here , a photon can excite a quasi - particle from the superconducting edges of the valence band to the conduction band , which again recombines under emission of a photon . this process may repeat itself an arbitrary number of times thus forming an electron - photon bound state , a polariton .
since the dos of the two superconducting bands diverge at @xmath239 ( measured from the the fermi energy ) , only scattering processes with a photon matching the energy difference @xmath280 give rise to a large effective coupling between photons and electrons and to the formation of polaritons . the unperturbed fermionic and bosonic
part of the action in eq . can be written in the basis of the r , a , k fields as @xmath443 where we defined the fermionic and bosonic spinors and matrix green s functions in section [ basmodsec ] .
the electron - photon interaction and photon - bath interaction can also be written in terms of these spinors .
let us consider first the electron - photon coupling @xmath444 { \stackrel{\delta}\psi { } \hspace{-1mm}}_{{\boldsymbol{k}}'}(t ) \label{epa2}\end{aligned}\ ] ] where @xmath445 and we define the matrices in the extended nambu - space as @xmath446 the keldysh contour can now be expressed by two integrations on the real axis for the @xmath447 and @xmath448 fields @xmath449^\dagger \bigl[\sum_{i=1,2 } \sqrt{2 } { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) \phi_{{\boldsymbol{k}}-{\boldsymbol{k}}',i}^\alpha ( t)\bigr ] { \stackrel{\delta}\psi { } \hspace{-1mm}}_{{\boldsymbol{k}}'}^\alpha ( t ) \nonumber \\ & = \sum_{{\boldsymbol{k}},{\boldsymbol{k } } ' } { \int_{-\infty}^\infty}dt \ , \hat \psi_{{\boldsymbol{k}}}^\dagger(t ) \underbrace{\bigl[\sum_{\alpha={\text{q } } , { \text{cl}}\atop i=1,2 } \hat \gamma^\alpha { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) \phi_{{\boldsymbol{k}}-{\boldsymbol{k}}',i}^\alpha ( t)\bigr]}_{\hat v_{{\boldsymbol{k}},{\boldsymbol{k}}'}(t ) } \hat \psi_{{\boldsymbol{k } } ' } ( t ) \label{epa4}\end{aligned}\ ] ] and finally we have rewritten the action in the r , a , k basis using @xmath450 and @xmath451 in keldysh space . here , we use the convention that we always first evaluate the keldysh matrix structure and after that the structure of the fermionic nambu - space @xmath3 or bosonic particle hole space @xmath63 .
now , we can integrate out the fermions easily as we have a quadratic action @xmath452
e^{i s_{ph } } e^{i [ s_c+s_v+s_{\text{int } } ] } & = e^{i s_{ph } } \int d[\psi,\bar \psi ] e^{i \sum_{{\boldsymbol{k}},{\boldsymbol{k } } ' } { \int_{-\infty}^\infty}dt dt'\ , \hat \psi_{{\boldsymbol{k}}}^\dagger(t ) \biggl ( \hat g_{0,{\boldsymbol{k}}}^{-1}(t , t ' ) \delta_{{\boldsymbol{k}},{\boldsymbol{k}}'}+ \hat v_{{\boldsymbol{k}},{\boldsymbol{k}}'}(t ) \delta(t - t ' ) \biggr ) \hat \psi_{{\boldsymbol{k}}}(t ' ) } \nonumber \\ & = e^{i s_{ph}+ { \text{tr}}\ln \bigl [ 1 + \hat g_0 \hat v ] } = e^{i s_{ph,\text{eff}}^{\text{el } } } \label{epa5}\end{aligned}\ ] ] expanding the trace log we find in leading order in @xmath453 that @xmath454 & = \sum_{n=1}^\infty \frac{(-1)^{n+1}}{n } { \text{tr}}\bigl [ ( \hat g_0 \hat v)^n \bigr ] = - \frac{1}{2}{\text{tr}}\bigl [ ( \hat g_0 \hat v)^2 \bigr ] + \ldots \nonumber \\ & \approx - \frac{1}{2 } \int dt dt ' \sum_{{\boldsymbol{k } } , { \boldsymbol{k } } ' } { \text{tr}}\bigl [ \hat g_{0,{\boldsymbol{k}}}(t',t ) \hat v_{{\boldsymbol{k}},{\boldsymbol{k}}'}(t ) \hat g_{{\boldsymbol{k}}'}(t , t ' ) \hat v_{{\boldsymbol{k}}',{\boldsymbol{k}}}(t ' ) \bigr ] \nonumber \\ & = - \frac{1}{2 } \int dt dt ' \sum_{{\boldsymbol{k } } , { \boldsymbol{q } } } { \text{tr}}\biggl [ \hat g_{0,{\boldsymbol{k}}}(t',t ) \bigl[\sum_{\alpha={\text{q } } , { \text{cl}}\atop i=1,2 } \hat \gamma^\alpha { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) \phi_{{\boldsymbol{q}},i}^\alpha ( t)\bigr ] \hat g_{0,{\boldsymbol{k}}-{\boldsymbol{q}}}(t , t ' ) \bigl[\sum_{\beta={\text{q } } , { \text{cl}}\atop j=1,2 } \hat \gamma^\beta { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) \phi_{-{\boldsymbol{q}},j}^\beta ( t')\bigr ] \biggr ] \nonumber \\ & = - \frac{1}{2 } \int dt dt ' \sum_{{\boldsymbol{q } } } \sum_{i , j=1,2 \atop \alpha,\beta = { \text{q } } , { \text{cl } } } \phi_{{\boldsymbol{q}},i}^\alpha ( t ) \underbrace{\biggl ( \sum_{{\boldsymbol{k } } } { \text{tr}}\bigl [ \hat g_{0,{\boldsymbol{k}}}(t',t ) \hat \gamma^\alpha { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) \hat g_{0,{\boldsymbol{k}}-{\boldsymbol{q}}}(t , t ' ) \hat \gamma^\beta { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) \bigr ] \biggr ) } _ { : = i [ \pi^{\text{el}}]_{ij,{\boldsymbol{q}}}^{\alpha \beta}(t , t ' ) } \phi_{-{\boldsymbol{q}},j}^\beta(t ' ) \nonumber \\ & = - \frac{i}{2 } \int dt dt ' \sum_{{\boldsymbol{q } } } \hat \phi_{{\boldsymbol{q}}}^t(t ) \hat \pi_{{\boldsymbol{q}}}^{\text{el}}(t , t ' ) \hat \phi_{-{\boldsymbol{q}},j}(t ' ) \label{epa6}\end{aligned}\ ] ] where we still have to evaluate the trace over the keldysh and nambu structure .
we get @xmath455 \hat \phi_{-{\boldsymbol{q}}}(t ' ) \label{epa7}\end{aligned}\ ] ] with @xmath456^{{\text{q } } , { \text{cl}}}_{11,{\boldsymbol{q } } } ( t , t ' ) & \bigl [ \pi^{\text{el } } \bigr]^{{\text{q } } , { \text{cl}}}_{12,{\boldsymbol{q } } } ( t , t ' ) \\ \bigl [ \pi^{\text{el } } \bigr]^{{\text{q } } , { \text{cl}}}_{21,{\boldsymbol{q } } } ( t , t ' ) & \bigl [ \pi^{\text{el } } \bigr]^{{\text{q } } , { \text{cl}}}_{22,{\boldsymbol{q } } } ( t , t ' ) \end{matrix } \right ) } } \nonumber \\ { \mathring}\pi_{{\boldsymbol{q}}}^{\text{el},a}(t , t ' ) & = { \ensuremath{\left ( \begin{matrix } \bigl [ \pi^{\text{el } } \bigr]^{{\text{cl}},{\text{q}}}_{11,{\boldsymbol{q } } } ( t , t ' ) & \bigl [ \pi^{\text{el } } \bigr]^{{\text{cl}},{\text{q}}}_{12,{\boldsymbol{q } } } ( t , t ' ) \\ \bigl [ \pi^{\text{el } } \bigr]^{{\text{cl}},{\text{q}}}_{21,{\boldsymbol{q } } } ( t , t ' ) & \bigl [ \pi^{\text{el } } \bigr]^{{\text{cl}},{\text{q}}}_{22,{\boldsymbol{q } } } ( t , t ' ) \end{matrix } \right ) } } \nonumber \\ { \mathring}\pi_{{\boldsymbol{q}}}^{\text{el},a}(t , t ' ) & = { \ensuremath{\left ( \begin{matrix } \bigl [ \pi^{\text{el } } \bigr]^{{\text{cl}},{\text{cl}}}_{11,{\boldsymbol{q } } } ( t , t ' ) & \bigl [ \pi^{\text{el } } \bigr]^{{\text{cl}},{\text{cl}}}_{12,{\boldsymbol{q } } } ( t , t ' ) \\ \bigl [ \pi^{\text{el } } \bigr]^{{\text{cl}},{\text{cl}}}_{21,{\boldsymbol{q } } } ( t , t ' ) & \bigl [ \pi^{\text{el } } \bigr]^{{\text{cl}},{\text{cl}}}_{22,{\boldsymbol{q } } } ( t , t ' ) \end{matrix } \right ) } } \nonumber \\ [ \pi^{\text{el}}]_{ij,{\boldsymbol{q}}}^{\alpha \beta}(t , t ' ) & = - i \sum_{{\boldsymbol{k } } } { \text{tr}}\bigl [ \hat \gamma^\alpha { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) \hat g_{0,{\boldsymbol{k}}}(t , t ' ) \hat \gamma^\beta { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) \hat g_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}(t',t ) \bigr ] \label{epa8}\end{aligned}\ ] ] following the same lines as the previous calculation , we want to integrate out the bath photons in ( [ mod3.1 ] ) to find the corresponding self - energy for the sled photons , see also ref . [ ] .
it is straightforward to show that the corresponding self - energy is then given by @xmath457\end{aligned}\ ] ] the assumed ( sled photon ) momentum - independent coupling @xmath458 then leads to the simple relation ( we omit the unimportant since featureless real part ) @xmath459\end{aligned}\ ] ] with @xmath460 = \pi \sum_{{\boldsymbol{p } } } { \left|\lambda(\nu_{{\boldsymbol{p } } } ) \right|}^2 \delta(\omega-\omega_{{\boldsymbol{p } } } ) = \pi \int d\epsilon \
, n_{\text{bath}}(\epsilon ) { \left|\lambda(\epsilon ) \right|}^2 \delta(\omega-\epsilon ) \nonumber \\ & = \pi { \left|\lambda(\omega ) \right|}^2 n_{\text{bath}}(\omega)\end{aligned}\ ] ] where @xmath461 is the dos of the photon bath .
we choose an markovian ( or ohmic ) bath , where the dos and the coupling are constant in frequency for all @xmath462 ( where @xmath463 ) and thus give rise to a constant absorption rate of the bath for the photons produced by the sled ( with characteristic energy @xmath464 ) .
we use the parametrization @xmath465 with @xmath466 ( which is important as explained in the main text ) .
the derivation from the constant behavior for small @xmath467 does not change the results and features of the sled presented in this paper as we are interested at photons with @xmath468 .
the total effective action then reads @xmath469 \hat \phi_{-{\boldsymbol{q}}}(t ' ) \label{epa10}\end{aligned}\ ] ]
in this section , we investigate the photon luminescence @xmath281 of the superconducting led , _ i.e _ , the number of photons present in the system in the steady state .
we consider both the case of normal conducting leads and the one of superconducting leads .
superconductivity leads to a strong enhancement of the luminescence in a frequency window close to @xmath282 .
we demonstrate that the superconducting led emits entangled photon pairs and produces squeezed light of frequency @xmath283 .
the squeezing occurs in certain two - mode quadrature operators of the light field defined below and implies that the fluctuations in one of the quadrature components falls below the minimal uncertainty of coherent states .
the photon luminescence is defined as the expectation value @xmath284 , which can be expressed via the lesser photon green s function @xmath285 as @xmath286 the lesser propagator @xmath287 is a linear combination of the retarded , advanced and keldysh propagators given in eqs . - .
[ lum1 ] we present the luminescence in the normal and in the superconducting state for an external photon bath at @xmath288 . in fig .
[ lum2 ] we show the luminescence in the presence of bath photons for a bath kept at temperature @xmath289 . those photons can be absorbed in the semiconductor junction and transfer electrons from valence to conduction bands . as a function of photon frequency @xmath36 .
we show the luminescence of the led with normal conducting leads ( blue ) and superconducting leads ( red ) .
temperature of the external photon bath is kept at @xmath288 .
enhanced luminescence around @xmath282 in presence of superconductivity is due to quasi - particles tunneling from conduction to valence band and larger density of states at the edges of the superconducting gap ( see fig . [ basicprocesses ] ) .
inset enlarges region around resonance @xmath283 showing luminescence peak due to cooper pair tunneling . ] as a function of photon frequency @xmath36 .
external photon bath is kept at temperature @xmath289 .
thermal bath photons ( green dashed ) with frequency @xmath290 can be absorbed by the @xmath0-@xmath1 junction both for normal conducting ( blue ) and for superconducting leads ( red ) .
inset enlarges frequency region where the cooper pair peak and the absorption can be seen . ]
let us first focus the case of @xmath288 . for normal conducting leads
we then observe that only photons with frequency @xmath172 are present in the system .
this follows from the fact that electronic transitions are only available with energy difference @xmath291 ( see fig . [ bandsnc ] ) .
the frequency scale on which the luminescence increases from zero to its constant value is given by the photon momentum @xmath292 with @xmath293 . in the case of superconducting leads , we clearly observe a strong enhancement of the number of photons with frequency close to @xmath282 .
this derives from the large number of electronic states that are pushed out of the gapped region in the superconductor to the border .
photons are produced via a recombination of ( bogoliubov ) quasi - particles from the conduction band to the valence band as depicted in fig .
[ basicprocesses ] .
the behavior of the luminescence can be traced back to the photon self - energy shown in fig .
[ normalse ] , if one notices that the imaginary parts of the self - energies correspond to photon production rate ( for positive imaginary part ) and decay rate ( for negative imaginary part ) .
in addition to the enhanced photon production at the band edges , the luminescence also exhibits a cooper pair peak at @xmath294 in the presence of superconductivity .
this stems from cooper pairs that are transferred from the conduction to the valence band .
this process is of the order @xmath295 and arises from the diagram shown in fig .
[ fig:10 ] that contains two anomalous self - energy bubbles .
this contributions is taken into account in the rpa summation .
the luminescence peaks are characterized by a width @xmath296 $ ] which is also the width of the plateau of the imaginary part of the retarded photon self - energies in eq . .
to obtain the numerical result in fig .
[ lum1 ] , we have used parameters that are consistent with our general discussion of energy scales in sec . [
sec : energy - scales - system ] .
specifically , we have expressed all energies in units of the superconducting gap @xmath159 ( a realistic value is @xmath297 ) .
we have set @xmath298 , @xmath299 , and @xmath300 .
we further assume a simple quadratic conduction band with a band edge that lies at a distance @xmath301 below the chemical potential @xmath302 .
the electronic density of states thus reads @xmath303 . for a non - lasing steady - state to exist ,
the photon decay rate due to the bath @xmath164 must be larger than ( see eq . ) @xmath304 we choose @xmath305 as even larger decay rates ( in particular @xmath306 ) result in a photon linewidth that is larger than the superconducting gap which smears the features observed in fig .
as soon as @xmath307 the system exhibits lasing and the photon number diverges at a particular frequency .
we discuss this possibility in detail in sec .
[ sec : steady - state - laser ] .
if the bath contains real photons they can be absorbed by the hetero - structure .
this case is shown in fig .
[ lum2 ] where we assume a thermal photon bath at temperature @xmath289 .
photons of frequency @xmath308 may be absorbed promoting electrons from the valence to the conduction band .
this occurs both for normal conducting as well as for superconducting leads . here
, we neglect the effect on the fermionic distribution and assume that the fermionic distribution functions remains at @xmath187 .
this can be realized in practice by a fermionic bath of this temperature .
in addition to the photon emission peaks , which are unaffected by the presence of bath photons , we now clearly observe an additional absorption dip at @xmath309 in the presence of superconductivity .
that contribute to the luminescence @xmath310 .
part ( b ) contains two anomalous self - energy loops @xmath311 and gives rise to the cooper pair peak seen at @xmath283 . ]
[ sec : squeezing ] in this section , we investigate the statistical properties of the light that is emitted from the superconducting led .
we find that the photons inherit the anomalous correlations that are present between electrons in a superconductor .
in particular , the photon expectation value @xmath312 is non - zero , which reduces the quantum fluctuations of certain two - mode quadrature operators of the light field to a value below that found for a coherent state @xcite .
the fluctuations of the conjugate operator are of course increased such that the heisenberg uncertainty limit is obeyed .
the superconducting led therefore emits two - mode squeezed light , which can act as a resource for quantum information processing and metrology @xcite .
the light squeezing in this setup was studied in ref . within lowest order perturbation theory in the electron - photon coupling @xmath43 .
to this order , the luminescence @xmath313 vanishes on resonance @xmath283 ( at @xmath314 ) where the lowest order terms are @xmath315 .
the amount of squeezing is then solely determined by the anomalous expectation value @xmath312 . here , we consider the effect of the diagonal ( photon number ) expectation values @xmath313 on the squeezing by summing the complete rpa series .
we make a quantitative prediction for the maximal reduction of quantum fluctuations that is achievable in this setup . as a function of photon frequency @xmath36 .
we show the real part ( red ) , imaginary part ( blue ) and absolute value ( green ) . external bath temperature is set to @xmath288 . ]
squeezing occurs in the two - mode quadrature operators @xmath316 \ , , \label{sq1}\end{aligned}\ ] ] where @xmath317 and @xmath318 .
the photon operators are defined in the rotating frame @xmath319 .
the fluctuations of these two - mode quadrature operators are given by @xcite @xmath320 \ , , \label{sq2}\end{aligned}\ ] ] where @xmath321 .
the state is truly squeezed if the fluctuations in one of the quadrature operators fall below the heisenberg uncertainty limit : @xmath322 for either @xmath323 or @xmath324 .
if the number of photons @xmath325 in a mode @xmath165 is zero , a finite anomalous expectation value of @xmath326 always results in squeezing . on the other hand ,
in the presence of photons one must compare the expectation values @xmath327 and @xmath328 to each other .
since the number of photons is smaller at low temperature , the resulting squeezing amplitudes are larger .
we thus focus on the case of @xmath314 in the following .
a crucial observation is that close to resonance @xmath283 it holds that the luminescence @xmath329 while the anomalous expectation values are already of order @xmath330 .
we obtain the anomalous photon expectation value @xmath331 } \int \frac{d\omega}{2\pi}i \tilde d_{11,{\boldsymbol{q}}}^<(\omega)\end{aligned}\ ] ] by an integral over the lesser anomalous propagator @xmath332 .
we define the anomalous luminescence by @xmath333 to arrive at @xmath334 |\mathcal{l}_a(\omega_{{\boldsymbol{q}}})| \bigr ) \nonumber \ , .
\end{aligned}\ ] ] here , the initial phase of the last terms reads @xmath335
= \phi_v-\phi_c + 2 \phi_g + l_{{\boldsymbol{q}}}$ ] and the time dependence vanishes for photons on resonance @xmath283 .
the anomalous luminescence @xmath336 is shown in fig . [ la ] .
at @xmath314 it exhibits two main peaks : one at @xmath282 corresponding to the transition of a bogoliubov quasi - particle from the conduction to the valence band and one at @xmath283 corresponding to the transition of a cooper pair .
it is important to note that the peak at @xmath282 implies that breaking up ( two ) different cooper pairs still leads to the emission of correlated and phase coherent photons due to the presence of macroscopic electronic bcs condensates .
both processes are depicted in fig .
[ basicprocesses ] . due to the finite number of photons at @xmath309 which are produced in the system ( see luminescence in fig . [ lum1 ] )
we observe a small peak of @xmath336 at @xmath309 .
this peak corresponds to an absorption process of a bogoliubov quasi - particle from the valence to the conduction band . in order to obtain squeezing for the mode with photon momentum @xmath14
, it is required that @xmath337 . comparing the normal luminescence @xmath338 in fig .
[ lum1 ] with the anomalous luminescence @xmath336 in fig .
[ la ] , we find that squeezing is maximal for photons on resonance @xmath283 , which corresponds to the transitions involving cooper pairs . in fig .
[ squeezing ] we show the amount of squeezing on resonance at @xmath288 .
it shows that squeezing can be controlled by the relative phase between the two superconductors @xmath339 .
the fluctuations of either @xmath340 or @xmath341 fall below the heisenberg uncertainty limit for a broad range of relative phases .
the maximal amount of squeezing for our realistic choice of parameters is about @xmath342 percent . in the experimental setup of the @xmath0-@xmath1 junction
the relative phase @xmath343 depends on microscopic details such as the initial switch - on time . in an experiment
it will be random from experiment to experiment but fixed within one run .
if one places two superconducting @xmath0-@xmath1 junctions in parallel in a squid geometry one can control the relative phase @xmath339 between the two junctions .
since this will essentially change the individual relative phases @xmath344 , @xmath345 as well , this provides a way to change @xmath346 with a magnetic field . of the two - mode quadrature operators @xmath347 for @xmath293 ( on resonance ) as a function of superconducting phase difference @xmath348 .
the phase @xmath135 denotes the constant phase of the electron - photon coupling constant @xmath43 . for a range of phase differences one of the fluctuation
amplitudes falls below the heisenberg uncertainty minimum for the symmetric case @xmath349 .
this shows that a superconducting led emits squeezed light .
inset shows mean and uncertainty of two - mode quadrature operators @xmath347 .
the orientation of the squeezing ellipse is controlled by @xmath339 . ]
in our calculation so far we have assumed that the system reaches a steady - state with a finite number of photons in the system .
this requires that the bath absorbs the photons that are produced in the led sufficiently fast . in this section
, we derive the exact requirements that have to be fulfilled for this to be the case .
otherwise , the luminescence diverges at certain frequencies and the system exhibits lasing .
we discuss the lasing conditions and give the frequency window where lasing occurs .
finally , we calculate the steady - state photon distribution using rate equations . to derive the conditions that are required for a steady - state with a finite photon number to exist , we follow a discussion given in ref .
the key idea is to determine when and for which frequencies the luminescence exhibits a divergence .
this corresponds to a transition into a lasing regime .
we first consider the case of normal conducting leads and then the one of superconducting leads .
the luminescence @xmath350 can be obtained via the lesser green s function @xmath287 ( see eq . ) .
we can parametrize the inverse retarded , advanced and keldysh green s functions as @xmath351^{-1 } & = a_{{\boldsymbol{q}}}(\omega ) \pm i b_{{\boldsymbol{q}}}(\omega ) \\ \label{lm2 } [ d_{12,{\boldsymbol{q}}}^k(\omega)]^{-1 } & = i c_{{\boldsymbol{q}}}(\omega ) \,.\end{aligned}\ ] ] using the matrix structure in keldysh space ( see eq . )
one obtains immediately @xmath352^{-1 } \\
\label{lm2b } d_{12,{\boldsymbol{q}}}^k(\omega ) & = - \frac { [ d_{12,{\boldsymbol{q}}}^k(\omega)]^{-1}}{[d_{12,{\boldsymbol{q}}}^{r}(\omega)]^{-1}[d_{12,{\boldsymbol{q}}}^{a}(\omega)]^{-1 } } \ , .
\end{aligned}\ ] ] the zeros of @xmath353 describe the excitations of the system and @xmath354 their linewidth .
let us assume that there exists a resonance at the renormalized photon frequency @xmath355 such that @xmath356 .
it is then required that the imaginary part fulfills @xmath357 to obtain a proper retarded green s function with the poles lying in the lower complex frequency plane .
( green dashed ) of inverse retarded propagator @xmath358^{-1}$ ] .
lasing occurs for momenta larger than @xmath359 .
corresponding real part @xmath353 is shown for different momenta @xmath14 around @xmath360 .
the photon bath decay rate is chosen @xmath361 to fulfill the laser threshold relation for frequencies @xmath362 .
lasing occurs for mode with momentum @xmath363 where both real part @xmath364 ( red ) and imaginary part @xmath365 have simultaneous zeros .
lower panel shows the luminescence @xmath366 as a function of photon energy @xmath367 . at @xmath368
the luminescence diverges , which denotes a violation of our assumption of a ( non - lasing ) steady - state with finite photon number . in the lasing regime @xmath369 ( light blue region )
the system produces photons at a faster rate than the absorption @xmath164 due to the bath occurs . ] ( red dashed ) and real part @xmath353 of inverse retarded propagator @xmath358^{-1}$ ] for different momenta @xmath14 with energy close to @xmath370 . the photon decay rate @xmath164 is chosen to fulfill the laser threshold relation in eq . in the superconducting state .
the lasing regime is bounded by momenta @xmath371 and @xmath372 which correspond to the photon modes where @xmath353 and @xmath354 have simultaneous zeros ( for some frequency @xmath177 ) .
note the strong renormalization of the photon frequency close to @xmath373 which increases the lasing frequency window shown in the lower panel ( in light blue ) .
lower panel shows divergence of luminescence for photon modes @xmath374 and @xmath375 and lasing regime ( light blue ) in between . ] in eqs .
- we have given the propagators within the rpa approximation .
for normal conducting leads , the retarded function reduces to @xmath358^{-1 } = \omega-\omega_{{\boldsymbol{q } } } - \pi_{21,{\boldsymbol{q}}}^r(\omega)$ ] and thus @xmath376 here , @xmath377 denotes the renormalized photon frequency , where the real part of the self - energy is only weakly frequency dependent
. the imaginary part @xmath378 is given in eq . and
@xmath379 in eq . .
the bath contribution to the self - energy fulfills @xmath380 for all frequencies @xmath177 .
in contrast , the electronic contribution @xmath378 changes sign at @xmath178 and is positive for frequencies below resonance @xmath381 ( see fig . [ normalse ] ) .
this follows from the electronic population inversion and describes photon production via transitions of conduction electrons to the valence band ( see fig .
[ bandsnc ] ) . depending on the size of the bath decay rate @xmath164
the total imaginary part @xmath354 may now be negative for frequencies @xmath381 .
photon excitations with a resonance energy @xmath382 for which this is the case violate the analytical requirements of a retarded propagator .
this indicates the breakdown of our assumption of a steady - state with a finite photon number . at this point , the system exhibits lasing @xcite .
this breakdown can also be observed as a divergence of the luminescence @xmath383 , where @xmath384 if the bath decay rate @xmath385 , the imaginary part vanishes @xmath386 for some frequency @xmath387 .
around @xmath388 one can linearize @xmath389 with @xmath390 . if the excitation energy @xmath391 is close to this critical energy , we can approximate the luminescence as @xmath392 we have used that the dominant part of the integral comes from the region around @xmath393 .
we note that the renormalization of the photon frequency is a small effect for normal conducting leads @xmath394 . as shown in fig .
[ ans0 ] , the photon number diverges like @xmath395 from above , where the luminescence is finite .
this behavior is reminiscent of bose - einstein condensation in an ideal gas .
the divergence occurs for photons with momentum @xmath396 where the zeros of the real part @xmath353 and of the imaginary part @xmath354 occur at the same frequency @xmath397 . neglecting any frequency renormalization @xmath398
, we find @xmath399 due to its population inversion the electronic system induces an effective chemical potential @xmath400 for the photons . within our approach
, we can not calculate the luminescence for frequencies @xmath369 since our assumption of a steady - state with a finite photon number breaks down .
the led produces more photons in these modes as the external bath can absorb . for these frequencies one needs to perform perturbation theory around a lasing state instead . to summarize
, the system exhibits lasing if the imaginary part of the retarded self - energy becomes positive for positive frequencies .
the luminescence diverges for that photon mode @xmath360 for which both real and imaginary part of the retarded photon propagator exhibits zeros @xmath401 for at least one @xmath177 .
turning to the case of superconducting leads , we observe that the electronic contribution to the self - energy is strongly increased around @xmath402 ( see fig .
[ normalse ] ) .
the laser threshold thus corresponds to a larger value of the bath absorption rate @xmath403 where @xmath404 for our realistic choice of parameters ( see sec . [
sec : energy - scales - system ] ) . in fig .
[ ans1 ] we show @xmath405 and the luminescence @xmath406 for superconducting leads for @xmath407 such that there is no lasing in the absence of superconductivity .
as the simultaneous zeros in the upper panel of the figure indicate , superconductivity leads to lasing in the frequency range @xmath408 .
the luminescence clearly diverges as @xmath409 at the border of this region . within this region
one needs to perform perturbation theory around a lasing state , which is beyond the scope of this work . to conclude , superconductivity leads to a sharp increase of the photon production rate of the led around the frequency @xmath410 .
this induces a sharp increase of the electroluminescence @xmath338 and a smaller laser threshold @xmath411 .
the lasing condition in eq . can also be derived using rate equations .
this approach is appealing due to its simplicity and physical transparency .
it begins with calculating the change of the number of bosons @xmath412 , which is given by @xmath413 \rangle } \
, , \label{rp1}\ ] ] where the time - dependent operators @xmath414 are defined in the heisenberg picture .
the expectation value @xmath415 of an operator @xmath416 is defined with respect to the full hamiltonian @xmath417 in eq . .
the commutator can be evaluated to @xmath418 & = \sum_{{\boldsymbol{k}},{\boldsymbol{k } } ' , \sigma } \bigl[\bar{g}(t ) v^\dag_{{\boldsymbol{k}},\sigma}(t )
b^\dag_{{\boldsymbol{k}}'-{\boldsymbol{k}}}(t ) c_{{\boldsymbol{k}}',\sigma}(t ) \\ & \quad - g(t ) v_{{\boldsymbol{k}},\sigma}(t ) b_{{\boldsymbol{k}}'-{\boldsymbol{k}}}(t ) c^\dag_{{\boldsymbol{k}}',\sigma}(t ) \bigr ] \nonumber \\ & \quad + \sum_{{\boldsymbol{p}},{\boldsymbol{q } } } \bigl [ \lambda_{{\boldsymbol{p}},{\boldsymbol{q } } } b^\dag_{{\boldsymbol{p}}}(t ) a_{{\boldsymbol{q}}}(t)- \bar \lambda_{{\boldsymbol{q}},{\boldsymbol{p } } } a^\dag_{{\boldsymbol{p}}}(t ) b_{{\boldsymbol{q}}}(t ) \bigr ] \nonumber \ , .\end{aligned}\ ] ] the expectation value of this commutator can be conveniently evaluated within perturbation theory in @xmath43 using the keldysh formalism @xcite .
the total photon production rate @xmath419 is composed of two contributions : first , @xmath420 which describes the emission and absorption of photons due to electronic transitions in the led and second , @xmath421 , which characterizes the photon exchange with the external bath . an explicit calculation , which takes the diagrams shown in fig .
[ rate ] into account , yields @xmath422 \\
\label{rp3 } \gamma _ { \text{sled } } & = 2 \sum_{{\boldsymbol{q } } } \int \frac{d\omega}{\pi } { \text{im}\,}\pi_{21,{\boldsymbol{q}}}^{\text{el},r}(\omega ) { \text{im}\,}d_{12,{\boldsymbol{q}}}^a(\omega ) \biggl ( \bigl[1+n_{{\boldsymbol{q}}}(\omega)\bigr ] \theta(ev_0-\omega ) + n_{{\boldsymbol{q}}}(\omega ) \theta(\omega - ev_0 ) \biggr ) \nonumber \\ & \qquad - \frac{1}{\pi } \sum_{{\boldsymbol{q } } } \int \frac{d\omega}{\pi } { \text{re}\ , } \bigl [ \tilde \pi_{11,{\boldsymbol{q}}}^{\text{el},r}(\omega ) d_{11,{\boldsymbol{q}}}^a(\omega ) \bigr ] \ , .\end{aligned}\ ] ] here , @xmath423 denotes the photon distribution function in the ( superconducting ) led system and @xmath264 denotes the photon distribution in the external bath . for a thermal bath ,
@xmath264 is given by the bose - einstein distribution function .
the coupling to the bath drives the photon distribution in the system @xmath424 towards the external bath distribution @xmath425 .
in contrast , the led produces photons via spontaneous as well as stimulated emission for @xmath426 and absorbs them for @xmath427 .
we evaluate the integrals in eq . and to leading order in the electron - photon coupling @xmath43 .
to this order , we only get an on - shell contribution @xmath428 , because higher order terms in @xmath429 are neglected as the self - energies are already of order @xmath430 . as a result
, the frequency integration enforces @xmath431 and the total photon production rate simplifies to @xmath432 + { \text{im}\,}\pi_{21,{\boldsymbol{q}}}^{\text{el},r}(\omega_{{\boldsymbol{q } } } ) \biggl ( \bigl[1+n(\omega_{{\boldsymbol{q}}})\bigr ] \theta(ev_0-\omega_{{\boldsymbol{q } } } ) + n(\omega_{{\boldsymbol{q } } } ) \theta(\omega_{{\boldsymbol{q}}}-ev_0 ) \biggr ) \biggr ] \end{aligned}\ ] ] in a steady state the photon emission and absorption are balanced and the total rate @xmath433 is zero for all photon modes @xmath14 . employing that @xmath434 and @xmath435 \sim { \text{sign}}(ev_0 - \omega_{{\boldsymbol{q}}})$ ]
we can infer the photon distribution in the steady state as @xmath436 for @xmath437 the distribution function is positive and finite . for @xmath438 , however , the imaginary part of the total self - energy @xmath439 in the denominator of eq . may become zero . at this point , the production rate of the led is exactly equal in magnitude to the decay rate into the bath .
this results in the divergence of the photon distribution for energies @xmath193 for which @xmath440 .
it follows that the laser threshold in the ( superconducting ) led is given by our previous result in eq .
( [ lm4 ] ) .
we have investigated the electroluminescence and photonic properties of a forward biased @xmath0-@xmath1 junction in proximity to superconducting leads .
we have shown that superconductivity leads to a significant enhancement of the luminescence of the light - emitting diode in a frequency window of the order of @xmath441 .
this effect stems from the increased density of states at the edges of the superconducting gap in the electronic bandstructure of the valence and conduction bands .
the increased photon production rate also reduces the lasing threshold in the system . by summing the complete infinite order rpa perturbation series in the photon - electron coupling
, we were able to show that an additional luminescence peak occurs on resonance due to the tunneling of cooper pairs from the conduction to the valence band .
in addition , such a superconducting light - emitting diode emits two - mode squeezed light .
the squeezing angle is controlled by the superconducting phase difference @xmath442 between conduction and valence band .
this proves that one may transfer the macroscopic coherence of an electronic superconducting condensate to photon pairs and manipulate the photonic coherence electronically .
we acknowledge useful discussions with p. baireuther , u. briskot , b. narozhny , m. schtt and j. schmalian .
the young investigator group of p.p.o . received financial support from the `` concept for the future '' of the karlsruhe institute of technology ( kit ) , germany , within the framework of the german excellence initiative .
we already derived the expressions for the bosonic self - energy caused by the superconducting electrons in the system in ( [ epa8 ] ) .
let us now first derive the causal structure of this self - energy by performing the trace over the keldysh indices @xmath471_{ij,{\boldsymbol{q}}}^{r}(t , t ' ) & = \bigl [ \pi^{\text{el } } \bigr]^{{\text{q } } , { \text{cl}}}_{ij,{\boldsymbol{q } } } ( t , t ' ) = - i \sum_{{\boldsymbol{k } } } { \text{tr}}\bigl [ \hat \gamma^{\text{q}}{\stackrel{\delta}g { } \hspace{-1mm}}_i(t ) \hat g_{0,{\boldsymbol{k}}}(t , t ' ) \hat \gamma^{\text{cl}}{\stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) \hat g_{{\boldsymbol{k}}+{\boldsymbol{q}}}(t',t ) \bigr ] \nonumber \\ & { \ensuremath{\stackrel{(\ref{mod6})}{= } } } - i \sum_{{\boldsymbol{k } } } { \text{tr}}_\delta { \text{tr}}_{\wedge } \biggl [ { \ensuremath{\left ( \begin{matrix } 0 & 1\\ 1 & 0 \end{matrix } \right ) } } { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) { \ensuremath{\left ( \begin{matrix } { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^r(t , t ' ) & { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^k(t , t ' ) \\ 0 & { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^a(t , t ' ) \end{matrix } \right ) } } { \ensuremath{\left ( \begin{matrix } 1 & 0 \\ 0 & 1 \end{matrix } \right ) } } { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) { \ensuremath{\left ( \begin{matrix } { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^r(t',t ) & { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^k(t',t ) \\ 0 & { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^a(t',t ) \end{matrix } \right ) } } \biggr ] \nonumber \\ & = - i \sum_{{\boldsymbol{k } } } { \text{tr}}_\delta \bigl [ { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^r(t , t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^k(t',t ) + { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^k(t , t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^a(t',t ) \bigr ] \label{cse1 } \\ [ \pi^{\text{el}}]_{ij,{\boldsymbol{q}}}^{a}(t , t ' ) & = - i \sum_{{\boldsymbol{k } } } { \text{tr}}_\delta \bigl [ { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^a(t , t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^k(t',t ) + { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^k(t , t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^r(t',t ) \bigr ] \label{cse2 } \\ [ \pi^{\text{el}}]_{ij,{\boldsymbol{q}}}^{a}(t , t ' ) & = - i \sum_{{\boldsymbol{k } } } { \text{tr}}_\delta \biggl [ { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^k(t , t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^k(t',t ) \nonumber \\ & \hspace{2cm}- { \stackrel{\delta}g { } \hspace{-1mm}}_i(t ) \bigl[{\stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^r(t , t ' ) - { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^a(t , t')\bigr ] { \stackrel{\delta}g { } \hspace{-1mm}}_j(t ' ) \bigl [ { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^r(t',t)- { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^a(t',t)\bigr ] \biggr ] \label{cse3 } \end{aligned}\ ] ] which are just the usual forms of the retarded , advanced and keldysh self - energies as known from standard text books
. the next step will be the evaluation of the nambu trace .
let us first write define the _ absolute _ anomalous propagator ( @xmath472 ) @xmath473 where we separated the phase of the superconducting gaps @xmath474 . defining also the phase of the electron - photon coupling @xmath475 , the absolute time @xmath476 and relative time @xmath477 we can evaluate @xmath478 & = -\frac{g(t ) g(t')}{2 } \bigl [ p_{0,{\boldsymbol{k}},v}^\alpha(t , t ' ) \bar p_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^\beta(t',t ) + \bar p_{0,{\boldsymbol{k}},c}^\alpha(t , t ' ) p_{0,{\boldsymbol{k}}+{\boldsymbol{q}},v}^\beta(t',t ) \bigr ] \\ & = - e^{i \phi(2 t ) } { \left|g_0 \right|}^2 \bigl [ { \left|p \right|}_{0,{\boldsymbol{k}},v}^\alpha(\tau ) { \left| p \right|}_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^\beta(-\tau ) + { \left|p \right|}_{0,{\boldsymbol{k}},c}^\alpha(\tau ) { \left|p \right|}_{0,{\boldsymbol{k}}+{\boldsymbol{q}},v}^\beta(-\tau ) \bigr ] \\ { \text{tr}}_\delta \bigl[{\stackrel{\delta}g { } \hspace{-1mm}}_1(t ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^\alpha(t , t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_2(t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^\beta(t',t ) \bigr ] & = \frac{g(t ) \bar g(t')}{2 } \bigl [ g_{0,{\boldsymbol{k}},v}^{(p),\alpha}(t , t ' ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{(p),\beta}(t',t ) + g_{0,{\boldsymbol{k}},c}^{(h),\alpha}(t , t ' ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},v}^{(h),\beta}(t',t ) \bigr ] \\ & = e^{i e v_0 \tau } { \left|g_0 \right|}^2 \bigl [ g_{0,{\boldsymbol{k}},v}^{(p),\alpha}(\tau ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{(p),\beta}(-\tau ) + g_{0,{\boldsymbol{k}},c}^{(h),\alpha}(\tau ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},v}^{(h),\beta}(-\tau ) \bigr ] \\ { \text{tr}}_\delta \bigl[{\stackrel{\delta}g { } \hspace{-1mm}}_2(t ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^\alpha(t , t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_1(t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^\beta(t',t ) \bigr ] & = \frac{\bar g(t ) g(t')}{2 } \bigl [ g_{0,{\boldsymbol{k}},v}^{(h),\alpha}(t , t ' ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{(h),\beta}(t',t ) + g_{0,{\boldsymbol{k}},c}^{(p),\alpha}(t , t ' ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},v}^{(p),\beta}(t',t ) \bigr ] \\ & = e^{-i e v_0 \tau } { \left|g_0 \right|}^2 \bigl [ g_{0,{\boldsymbol{k}},v}^{(h),\alpha}(\tau ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{(h),\beta}(-\tau ) + g_{0,{\boldsymbol{k}},c}^{(p),\alpha}(\tau ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},v}^{(p),\beta}(-\tau ) \bigr ] \\ { \text{tr}}_\delta \bigl[{\stackrel{\delta}g { } \hspace{-1mm}}_2(t ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}}^\alpha(t , t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_2(t ' ) { \stackrel{\delta}g { } \hspace{-1mm}}_{0,{\boldsymbol{k}}+{\boldsymbol{q}}}^\beta(t',t ) \bigr ] & = -\frac{\bar g(t ) \bar g(t')}{2 } \bigl [ \bar p_{0,{\boldsymbol{k}},v}^\alpha(t , t ' ) p_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^\beta(t',t ) + p_{0,{\boldsymbol{k}},c}^\alpha(t , t ' ) \bar p_{0,{\boldsymbol{k}}+{\boldsymbol{q}},v}^\beta(t',t ) \bigr ] \\ & = - e^{-i \phi(2 t ) } { \left|g_0 \right|}^2 \bigl [ { \left|p \right|}_{0,{\boldsymbol{k}},v}^\alpha(\tau ) { \left| p \right|}_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^\beta(-\tau ) + { \left|p \right|}_{0,{\boldsymbol{k}},c}^\alpha(\tau ) { \left|p \right|}_{0,{\boldsymbol{k}}+{\boldsymbol{q}},v}^\beta(-\tau ) \bigr ] \end{split } \label{cse5 } \end{aligned}\ ] ] with the rotating phase @xmath479 = 2 e v_0 t -\phi_c+\phi_v+2 \phi_g \label{cse6}\end{aligned}\ ] ] from ( [ epa8 ] ) and ( [ cse5 ] )
we see that we indeed find the structure presented in eq .
( [ ef1 ] ) in the photon particle - hole space @xmath63 . performing the wigner transformation @xmath480 , which is just a fourier transformation in relative time @xmath481
, we have the identity @xmath482 in total , we can write down the photon self - energy part from the coupling to the electrons as @xmath483 with @xmath484 \\ \label{eq:35 } \pi_{12,{\boldsymbol{q}}}^{\text{el},r / a}(\omega ) & = - i { \left|g_0 ^ 2 \right| } \sum_{{\boldsymbol{k } } } { \int_{-\infty}^\infty}\frac{d\omega_1}{2\pi } \biggl [ g_{0,{\boldsymbol{k}},v}^{(p),r / a}(\omega_1 ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{(p),k}(\omega_1-\omega_+ ) + g_{0,{\boldsymbol{k}},v}^{(p),k}(\omega_1 ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{(p),a
/ r}(\omega_1-\omega_+ ) \biggr ] \\
\label{eq:36 } \pi_{21,{\boldsymbol{q}}}^{\text{el},r / a}(\omega ) & = - i { \left|g_0 ^ 2 \right| } \sum_{{\boldsymbol{k } } } { \int_{-\infty}^\infty}\frac{d\omega_1}{2\pi } \biggl [ g_{0,{\boldsymbol{k}},v}^{(h),r / a}(\omega_1 ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{(h),k}(\omega_1-\omega_- ) + g_{0,{\boldsymbol{k}},v}^{(h),k}(\omega_1 ) g_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{(h),a / r}(\omega_1-\omega_- ) \biggr ] \\ \label{cse9 } \tilde \pi_{22,{\boldsymbol{q}}}^{\text{el},r / a}(\omega ) & = i { \left|g_0 ^ 2 \right| } \sum_{{\boldsymbol{k } } } { \int_{-\infty}^\infty}\frac{d\omega_1}{2\pi } \biggl [ { \left|p \right|}_{0,{\boldsymbol{k}},v}^{r / a}(\omega_1 ) { \left| p \right|}_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^k(\omega_1-\omega ) + { \left|p \right|}_{0,{\boldsymbol{k}},v}^k(\omega_1 ) { \left| p \right|}_{0,{\boldsymbol{k}}+{\boldsymbol{q}},c}^{a / r}(\omega_1-\omega ) \biggr]\end{aligned}\ ] ] where we defined @xmath485 and used several symmetry properties @xmath486 of the fermionic propagators .
the corresponding feynman graphs are presented in fig .
[ seel1 ] . using the symmetries ( [ cse9 ] )
it is easy to show that @xmath487 the keldysh self - energies can be easily shown to be given by the thermal equilibrium expressions ( [ nse8 ] ) . in the normal conductor the propagators
simplify to @xmath488 and @xmath489 \delta\bigl[\omega-\xi_{v / c}({\boldsymbol{k } } ) \bigr ] \end{split } \label{senc1}\end{aligned}\ ] ] where @xmath490 is the fermi function of the system , such that we find the standard results for the occurring particle - hole bubbles ( the anomalous self - energies @xmath491 vanish in the normal conductor ) @xmath492-n_f\bigl[\xi_c({\boldsymbol{k}}+{\boldsymbol{q}})\bigr ] } { \omega_- + \xi_v({\boldsymbol{k } } ) -\xi_c({\boldsymbol{k}}+{\boldsymbol{q}})\pm i0 } \label{senc2}\end{aligned}\ ] ] as depicted in fig .
[ bandsnc ] , let us now consider an isotropic electron band @xmath493 and a symmetric hole band @xmath494 and linearize around the fermi surface @xmath495 = v_f ( k - k_f ) \label{senc3}\end{aligned}\ ] ] where @xmath496 parametrize the orbital direction of @xmath497 in spherical coordinates and we can define the dos as usual @xmath498 $ ] .
let us now consider a finite momentum transfer with @xmath499 . due to the rotational invariance of the fermi surface we can choose @xmath14 to lie along the @xmath500 axis , see fig .
[ fermisurface ] , such that @xmath501 = v_f ( k - k_f ) + { \boldsymbol{v}}_f(\phi,\theta ) \cdot { \boldsymbol{q } } = v_f ( k - k_f ) + v_f q \cos(\theta ) \nonumber \\ & = \xi_c({\boldsymbol{k } } ) + v_f q \cos(\theta ) \label{senc4}\end{aligned}\ ] ] which allows us to write the above sum of momenta as @xmath502 & = \frac{1}{(2\pi)^3 } \int_0^{2\pi } d\phi \int_{-1}^1 d\cos \theta \int_0^\infty dk \cdot k^2 \ , f\bigl[-\xi_c(k),\xi_c(k)+v_f q \cos ( \theta ) \bigr ] \nonumber \\ & = \int_{-1}^1 \frac{d\cos \theta}2 \int_{-b}^\infty d\xi \ , \rho(\xi ) f\bigl [ - \xi , \xi+ v_f
q \cos(\theta ) \bigr ] \nonumber \\ & = \int_{-v_f q}^{v_f q } \frac{d\alpha}{2 v_f q } \int_{-b}^\infty d\xi \ , \rho(\xi ) f\bigl [ - \xi , \xi+ \alpha \bigr ] \label{senc5}\end{aligned}\ ] ] where @xmath171 is the energy distance from the lower band edge to the fermi energy in the conduction band .
we will now explicitly calculate the imaginary part of ( [ senc2 ] ) for the symmetric particle and hole bands in the normal conducting state and omit the real part since it is featureless and gives just a small unimportant correction to the dispersion of the photons .
we find for zero fermionic temperature @xmath211 the imaginary part @xmath503-n_f\bigl[\xi+\alpha\bigr]\biggr ) \delta \bigl[\omega_- - 2\xi-\alpha \bigr ] \nonumber \\ & = \mp \frac { \pi { \left|g_0 ^ 2 \right| } } { 2 v_f q } \int_{-v_f q}^{v_f q } d\alpha \underbrace{\rho\bigl(\frac{\omega_- -\alpha}{2 } \bigr)}_{\approx \rho(\omega_-/2 ) } \biggl ( n_f\bigl[\frac{-\omega_- + \alpha}{2 }
\bigr]-n_f\bigl[\frac{\omega_- + \alpha}{2 } \bigr]\biggr ) \nonumber \\ & \approx \mp \frac { \pi { \left|g_0 ^ 2 \right| } } { 2 v_f q}\rho(\omega_-/2 ) \int_{-v_f q}^{v_f q } d\alpha \biggl [ \theta(\omega_- -\alpha ) - \theta(-\omega_- -\alpha ) \biggr ] \nonumber \\ & = \mp \pi { \left|g_0 ^ 2 \right|}\rho(\omega_-/2 ) \ , { \text{sign}}(\omega_- ) \ , \min\bigl[1 , \frac{{\left|\omega_- \right|}}{v_f q } \bigr ] \label{senc6}\end{aligned}\ ] ] the real part will be neglected in the normal state as it is only weakly energy and momentum dependent and gives no important renormalization of the photon excitation energies @xmath504 . in the superconducting state
we can use the standard parametrization of the normal and anomalous greens functions ( @xmath211 ) @xmath505 \biggl ( a_{0,{\boldsymbol{k}},v / c}^{r}(\omega)-a_{0,{\boldsymbol{k}},v / c}^{a}(\omega ) \biggr ) = - 2 \pi i \biggl ( \alpha_{a,{\boldsymbol{k}},v / c } \delta\bigl[\omega - e_{v / c}({\boldsymbol{k } } ) \bigr ] + \beta_{a,{\boldsymbol{k}},v / c } \delta\bigl[\omega+e_{v / c}({\boldsymbol{k } } ) \bigr ] \biggr ) \nonumber\end{aligned}\ ] ] where we defined the superconducting dispersion @xmath506 and the coherence factors for the different green s functions are given by @xmath507 an explicit calculation of the normal and anomalous self - energy show that @xmath508 \\ \pi_{21,{\boldsymbol{q}}}^{\text{el},r / a}(\omega)&= 2 { \left|g_0 ^ 2 \right|}\sum_{{\boldsymbol{k } } } \biggl [ \frac{v_{{\boldsymbol{k}},v}^2 u_{{\boldsymbol{k}}+{\boldsymbol{q}},c}^2 } { \omega_--e_{v}({\boldsymbol{k}})-e_{c}({\boldsymbol{k}}+{\boldsymbol{q}})\pm i0 } -\frac{u_{{\boldsymbol{k}},v}^2 v_{{\boldsymbol{k}}+{\boldsymbol{q}},c}^2}{\omega_-+e_{v}({\boldsymbol{k}})+e_{c}({\boldsymbol{k}}+{\boldsymbol{q}})\pm i0 } \biggr ] \end{split}\label{sesc3}\end{aligned}\ ] ] and the other two self - energies are related via the symmetries ( [ cse11 ] ) .
let us now consider first the anomalous self - energy @xmath509 that can only occur in the presence of superconducting quasi - particles .
we can use the linearization ( [ senc5 ] ) for the symmetric bands to write the retarded self - energy as @xmath510 \nonumber \\ & = \frac{{\left|g_0 ^ 2 \right| } \rho_f \delta}{4 v_f q } \int \limits_{-v_f q/\delta}^{v_f q/\delta } d\alpha \int \limits_{-b/\delta}^\infty dx \frac{1}{\sqrt{x^2 + 1}\sqrt{(x+\alpha)^2 + 1 } } \biggl [ \frac{1}{\tilde \omega-\sqrt{x^2 + 1}-\sqrt{(x+\alpha)^2 + 1 } + i0 } - \ldots \biggr ] \label{sesc5a}\end{aligned}\ ] ] where we assumed @xmath511 , defined @xmath512 and approximated the dos to lie near the fermi surface since the dominant part of the integral will come from there .
the real part of the integral is obviously symmetric and the imaginary part anti - symmetric in frequency @xmath513 such that we will continue to focus on positive frequencies @xmath514 .
let us first calculate the real part @xmath515 \label{sesc5c}\end{aligned}\ ] ] for @xmath516 .
we see that the integrand will have a singularity if @xmath517 and we therefore expect that at @xmath518 some important feature can occur .
let us now consider this case @xmath519 , where it is convenient to separate the singular region @xmath520 of the integrand @xmath521 \label{sesc5e}\end{aligned}\ ] ] the contribution @xmath522 will just give a small unimportant term as @xmath523 is small and therefore our dispersion is just changed slightly . the important contributions will come from the divergences at @xmath524 , which are captured well by the expression ( [ sesc5e ] ) , such that we approximate @xmath525 \label{sesc5f}\end{aligned}\ ] ] the imaginary part can be calculated using the same methods as just explained for the real part and is given by @xmath526 \label{sesc5g}\end{aligned}\ ] ] obviously , there is a gap of @xmath527 for the imaginary part of the particle - hole bubble @xmath528 .
let us again focus on the region @xmath529 , where the dominant contributions come again from @xmath530 such that @xmath531 \nonumber\\ & = - \frac{{\left|g_0 ^ 2 \right| } \pi \rho_f \delta}{2\sqrt{2 } v_f q } \int \limits_{-v_f q/\delta}^{v_f q/\delta } d\alpha \int dy \frac{1}{\sqrt{\tilde \omega_\alpha } } \delta \bigl [ \tilde \omega-\tilde \omega_\alpha -y^2 \bigr ] \nonumber \\ & = - \frac{{\left|g_0 ^ 2 \right| } \pi \rho_f \delta}{2\sqrt{2 } v_f q } \int \limits_{-v_f q/\delta}^{v_f q/\delta } d\alpha \frac{\theta(\tilde \omega-\tilde \omega_\alpha)}{\sqrt{\tilde \omega_\alpha}\sqrt{\tilde \omega-\tilde \omega_\alpha } } \approx - \frac{{\left|g_0 ^ 2 \right| } \pi \rho_f}{2 \bigl ( \frac{v_f q}{2\delta}\bigr ) } \arcsin\biggl [ \frac{\min \bigl[\frac{v_f q}{2\delta},\sqrt{\tilde \omega-2 } \bigr]}{\sqrt{\tilde \omega - 2 } } \biggr ] \label{sesc5h } \end{aligned}\ ] ] where in the end we approximated the integral for small @xmath532 due to the integrand restrictions
. far away from the gap region @xmath533 it is easy to show that the imaginary part @xmath534 falls down rapidly and is therefore not important , such that we can approximate @xmath535 for all @xmath177 with ( [ sesc5h ] ) as stated in the main text .
the normal self - energy @xmath536 around @xmath537 can easily be shown to behave exactly like the anomalous self - energy @xmath509 around @xmath524 . far away from this region
@xmath538 it just behaves like in the normal conductor ( [ senc6 ] ) . in between ,
we interpolate between these two by replacing the self - energy as the maximum of the two limits @xmath539 \nonumber \\ & \quad \mp
i \max\biggl [ \frac{{\left|g_0 ^ 2 \right| } \pi \rho_f}{2 \bigl ( \frac{v_f q}{2\delta}\bigr ) } \arcsin\biggl ( \frac{\min \bigl[\frac{v_f q}{2\delta},\sqrt{{\left|\tilde \omega \right|}-2}\bigr ] } { \sqrt{{\left|\tilde \omega \right|}- 2 } } \biggr ) , \pi { \left|g_0 \right|}^2 \rho(\omega_-/2 ) \biggr ] { \text{sign}}(\omega_- ) \theta({\left|\tilde \omega \right|}-2)\end{aligned}\ ] ]
let us first consider the wigner transformation of a convolution of two bosonic matrices @xmath540 with time dependence @xmath541 where as usual we set @xmath542 , defined @xmath543 and @xmath477 .
the wigner - transform of a convolution @xmath540 can be expressed conveniently by the wigner transforms of @xmath544 and @xmath545 as @xmath546 } { \mathring}b(\omega , t ) \label{dp2}\end{aligned}\ ] ] using the identities @xmath547 we get @xmath548 } { \ensuremath{\left ( \begin{matrix } e^{-i\phi(2 t ) } b_{11}(\omega ) & b_{12}(\omega ) \\
b_{21}(\omega ) & e^{i\phi(2 t ) } b_{22}(\omega ) \end{matrix } \right ) } } \nonumber \\ & = { \ensuremath{\left ( \begin{matrix } e^{i\phi(2 t ) } a_{11}(\omega ) e^ { -\frac{i}{2 } \overleftarrow \partial_t \overrightarrow \partial_\omega } & a_{12}(\omega ) \\
a_{21}(\omega ) & e^{-i\phi(2 t ) } a_{22}(\omega ) e^{- \frac{i}{2 } \overleftarrow \partial_t \overrightarrow \partial_\omega } \end{matrix } \right ) } } { \ensuremath{\left ( \begin{matrix } e^ { \frac{i}{2 } \overleftarrow \partial_\omega \overrightarrow \partial_t } e^{-i\phi(2 t ) } b_{11}(\omega ) & b_{12}(\omega ) \\
b_{21}(\omega ) & e^ { \frac{i}{2 } \overleftarrow \partial_\omega \overrightarrow \partial_t } e^{i\phi(2 t ) } b_{22}(\omega ) \end{matrix } \right ) } } \nonumber \\ & = { \ensuremath{\left ( \begin{matrix } e^{i\phi(2 t ) } a_{11}(\omega ) e^ { e v_0 \overrightarrow \partial_\omega } & a_{12}(\omega ) \\
a_{21}(\omega ) & e^{-i\phi(2 t ) } a_{22}(\omega ) e^ { -e v_0 \overrightarrow \partial_\omega } \end{matrix } \right ) } } { \ensuremath{\left ( \begin{matrix } e^ { e v_0 \overleftarrow \partial_\omega } e^{-i\phi(2 t ) } b_{11}(\omega ) & b_{12}(\omega ) \\
b_{21}(\omega ) & e^ { -e v_0 \overleftarrow \partial_\omega } e^{i\phi(2 t ) } b_{22}(\omega ) \end{matrix } \right ) } } \nonumber \\ & = { \ensuremath{\left ( \begin{matrix } a_{11}(\omega+ e v_0 ) b_{11}(\omega+ev_0 ) + a_{12}(\omega ) b_{21}(\omega ) & e^{i \phi(2 t ) } \bigl [ a_{11}(\omega ) b_{12}(\omega+ e v_0)+ a_{12}(\omega- e v_0 ) b_{22}(\omega ) \bigr ] \\
e^{-i \phi(2 t ) } \bigl [ a_{21}(\omega+ e v_0 ) b_{11}(\omega)+ a_{22}(\omega ) b_{21}(\omega- e v_0 ) \bigr ] & a_{21}(\omega ) b_{12}(\omega ) + a_{22}(\omega- e v_0 ) b_{22}(\omega- e v_0 ) \end{matrix } \right ) } } \label{dp4}\end{aligned}\ ] ] the dyson equation in keldysh space is formally given by the convolution @xmath549 where @xmath550 is the full self - energy of the system .
the three coupled equations for the full retarded , advanced and keldysh propagators of the photon @xmath551^{-1}-{\mathring}\pi_{{\boldsymbol{q}}}^{r / a } \bigr)\circ { \mathring}d_{{\boldsymbol{q}}}^{r / a } & = { \mathring}1 \ , , \\ \bigl ( [ { \mathring}d_{0,{\boldsymbol{q}}}^{r}]^{-1}-{\mathring}\pi_{{\boldsymbol{q}}}^{r } \bigr ) \circ { \mathring}d_{{\boldsymbol{q}}}^k & = { \mathring}\pi_{{\boldsymbol{q}}}^{k } \circ \hat d_{{\boldsymbol{q}}}^a \ , .
\end{split } \label{dp6}\end{aligned}\ ] ] can now be solved with the ansatz that @xmath552 has the structure of @xmath545 . since both @xmath553^{-1}(\omega , t)-{\mathring}\pi_{{\boldsymbol{q}}}^{r } ( \omega , t)\bigr ) & = { \ensuremath{\left ( \begin{matrix } - e^{i \phi(2 t ) } \tilde \pi_{11,{\boldsymbol{q}}}^{r}(\omega ) & \overbrace{-\omega-\omega_{{\boldsymbol{q } } } - i0}^{[d_{0,{\boldsymbol{q}}}^a(-\omega)]^{-1}}- \pi_{12,{\boldsymbol{q}}}^{r}(\omega ) \\
\underbrace{\omega-\omega_{{\boldsymbol{q } } } + i0}_{[d_{0,{\boldsymbol{q}}}^r(\omega)]^{-1}}- \pi_{21,{\boldsymbol{q}}}^{r}(\omega ) & -e^{-i \phi(2 t ) } \tilde \pi_{22,{\boldsymbol{q}}}^{r}(\omega ) \end{matrix } \right ) } } \\ { \mathring}\pi_{{\boldsymbol{q}}}^{k } ( \omega , t ) & = { \ensuremath{\left ( \begin{matrix } e^{i \phi(2 t ) } \tilde \pi_{11,{\boldsymbol{q}}}^{k}(\omega ) & \pi_{12,{\boldsymbol{q}}}^{k}(\omega ) \\
\pi_{21,{\boldsymbol{q}}}^{k}(\omega ) & e^{-i \phi(2 t ) } \tilde \pi_{22,{\boldsymbol{q}}}^{k}(\omega ) \end{matrix } \right ) } } \end{split } \label{dp8 } \end{aligned}\ ] ] have exactly the structure of @xmath544 we can use ( [ dp4 ] ) to calculate the wigner transform of the convolutions in ( [ dp6 ] ) exactly , which leads to the propagators defined in ( [ prop ] )
. 42ifxundefined [ 1 ] ifx#1 ifnum [ 1 ]
# 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop _ _ ( , ) link:\doibase 10.1103/revmodphys.34.694 [ * * , ( ) ] @noop _ _ ( , ) @noop _ _ , ed .
( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.104.156802 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.094508 [ * * , ( ) ] @noop ( ) @noop ( ) link:\doibase 10.1103/physreva.13.2226 [ * * , ( ) ] @noop _ _ ( , , ) link:\doibase 10.1103/physrevlett.112.077003 [ * * , ( ) ] in link:\doibase 10.1007/978 - 94 - 015 - 8839 - 3_13 [ _ _ ] , , vol . ,
( , ) pp .
link:\doibase 10.1126/science.1113523 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop ( ) http://stacks.iop.org/0953-2048/23/i=3/a=034025 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.103.187001 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.157403 [ * * , ( ) ] http://stacks.iop.org/0957-4484/21/i=27/a=274004 [ * * , ( ) ] link:\doibase 10.1103/physrevb.87.094511 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.073901 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.96.230602 [ * * , ( ) ] link:\doibase 10.1103/physrevb.75.195331 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) link:\doibase 10.1103/physrevlett.59.278 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.59.2153 [ * * , ( ) ] @noop * * , ( ) in link:\doibase 10.1007/978 - 3 - 642 - 12491 - 4_12 [ _ _ ] , , vol . , ( , ) pp . in @noop
_ _ , ( , )
|
we investigate a semiconductor @xmath0-@xmath1 junction in contact with superconducting leads that is operated under forward bias as a light - emitting diode .
the presence of superconductivity results in a significant increase of the electroluminescence in a sharp frequency window .
we demonstrate that the tunneling of cooper pairs induces an additional luminescence peak on resonance .
there is a transfer of superconducting to photonic coherence that results in the emission of entangled photon pairs and squeezing of the fluctuations in the quadrature amplitudes of the emitted light .
we show that the squeezing angle can be electrically manipulated by changing the relative phase of the order parameters in the superconductors .
we finally derive the conditions for lasing in the system and show that the laser threshold is reduced due to superconductivity .
this reveals how the macroscopic coherence of a superconductor can be used to control the properties of light .
|
exclusive hadron production in high energy electron - positron annihilation is a very interesting task for theoretical and experimental investigations .
the presence of high energy scale @xmath7 that is much greater than typical hadronic scale allows one to separate the amplitude of such processes into hard part ( creation of quarks at very small distances ) and soft part ( subsequent hadronization of these quarks into experimentally observed mesons at larger distances ) .
the first part of the amplitude can be calculated within perturbative qcd .
the second part of the amplitude is described by distribution amplitudes ( da ) , which contain nonperturbative properties of final hadrons .
the description of hard exclusive hadron production within this pattern gives some very interesting predictions of the properties of hard exclusive processes @xcite .
one of such prediction is the asymptotic behavior of the amplitudes and cross sections of hard exclusive processes in the limit @xmath8 @xcite .
it turns out that this behavior is determined by the perturbative part of amplitude and quantum numbers of final hadrons and does not depend on das of final hadrons . to give quantitative prediction for the cross section of hard exclusive process one needs to know das .
it is interesting to note that if das are known the theory can equivalently well predict the cross sections for the production of hadrons composed of light ( @xmath9 quarks ) or heavy quarks ( @xmath10 quarks ) .
there is well known example of exclusive process with heavy quarkonia production @xmath11 measured at belle @xcite and babar @xcite collaborations .
this process was extensively studied in many papers @xcite within different approaches , what led to a better understanding of charmonia properties and production processes .
it is important to note that the approach to hard exclusive processes described above leads to a reasonable agreement with the experiments . in this paper
we study the processes @xmath12 , where @xmath1 .
experimental measurement of the cross section of the process @xmath2 at babar collaboration at large center mass energy @xmath3 gev and some low energy experimental data @xmath4 gev allow us to study the cross section of this process in the broad energy region .
our first purpose is to use these data in order to determine the asymptotic behavior of the cross section of the process @xmath13 . assuming that the same asymptotic behavior is valid for the other processes under consideration and using low energy experimental data one can predict the cross sections of the @xmath12 at energies @xmath6 gev .
in addition , we apply perturbative qcd approach to estimate the values of the cross sections for the processes under consideration .
this paper is organized as follows . in the next section
we analyze the experimental data for the process @xmath2 and determine the asymptotic behavior of this cross section .
then we apply the result of this study to predict the cross sections of the other processes under consideration at the center mass energies @xmath6 gev . in section [ sec : light - cone ] we give theoretical predictions for the cross sections @xmath14
, @xmath15 , @xmath16 and @xmath17 at @xmath18 gev and 10.6 gev and compare them with available experimental data .
the final section is devoted to the discussion of the results of this paper .
the amplitude of the process involved can be written in the following form : @xmath19 where @xmath20 is the electromagnetic coupling constant , @xmath21 and @xmath22 are electron and positron bispinors , @xmath23 is the invariant mass of @xmath24 system squared and @xmath25 is electromagnetic current .
the matrix element @xmath26 can be parameterized by the only formfactor @xmath27 : @xmath28 where @xmath29 is the polarization vector of meson @xmath30 .
the cross section of the process under consideration equals @xmath31 in the last formula @xmath32 is the momentum of the vector meson @xmath30 in the center mass frame of final mesons .
process.[fig : diags ] ] in this section we will be interested in the asymptotic behavior of the formfactor @xmath27 in the high energy region .
typical diagrams of the process under consideration are shown in fig .
[ fig : diags ] . the asymptotic behavior of the diagram shown in fig .
[ fig : diags]a is @xmath33 . at extremely large energies
this diagram gives the dominant contribution .
however , the amplitude of this diagram is suppressed by the smallness of the electromagnetic coupling constant and our study shows that in the energy region analyzed in this paper the contribution of this diagram is negligible .
further let us consider the diagrams shown in fig .
[ fig : diags]b , c . according to perturbative qcd @xcite amplitudes ( [ eq : matr ] ) from such diagrams
have the following asymptotic behavior @xmath34 where @xmath35 and @xmath36 are mesons with momenta @xmath37 , @xmath38 and helicities @xmath39 and @xmath40 . for the process under consideration
@xmath35 and @xmath36 are the vector and pseudoscalar mesons respectively .
the helicity of the pseudoscalar meson is , obviously , @xmath41 .
because of antisymmetric tensor in ( [ eq : matr ] ) , longitudinal polarization of the vector meson is forbidden and it is transversely polarized ( @xmath42 .
so , we have @xmath43 , @xmath44 .
on the other hand , in papers @xcite , it was stated that experimental data can be described only by the dependence @xmath45 .
to clarify this situation we will parameterize the formfactor @xmath27 by the expression @xmath46 where @xmath47 slightly depends on @xmath48 due to power and logarithmic corrections to the leading asymptotic behavior . later in this section
we will neglect such dependence .
this approximation allows us to fix the constants @xmath49 from the low energy data and predict the cross sections at @xmath18 gev and 10.6 gev .
we check three different hypothesis ( @xmath50 and 5 ) and compare the results with existing experimental data .
first we are going to consider the process @xmath51 . to fix the constants @xmath49
one can use low energy data @xcite .
when the constants @xmath49 for different hypothesis are fixed , one can use them to predict the cross section in the high energy region and then compare this prediction with available high energy data @xcite measured by babar collaboration @xmath52 our results are shown in fig .
[ fig : phieta ] and table [ tab : params ] . from these figure and table
it can be clearly seen , that only @xmath53 describes satisfactory low and high energy experimental babar data @xcite .
this result is in agreement with the predictions of perturbative qcd .
[ fig : phieta ] one can see that low energy cleo - c data @xcite are in disagreement with hypothesis @xmath50 .
only hypotheses @xmath54 does not contradict to the experimental results .
however , if we assume that this hypotheses is correct we will faced with dramatic contradiction with the high energy babar data ( see fig . [
fig : phieta ] and table [ tab : params ] ) .
it should be also noted that in papers @xcite it was stated that the energy dependence @xmath55 describes experimental data more accurately .
we believe that the disagreement of this statement with our conclusion arises from the fact that the authors of papers @xcite did not take into account low energy experimental result @xcite . in view of this
the question arises : is it possible to use asymptotic behavior of the cross section ( [ asympt ] ) in the region @xmath56 gev .
first , one can estimate the cross uncertainty due to the power corrections as @xmath57 .
even , for the heaviest meson @xmath58 and the smallest @xmath59 from the region @xmath56 the error is @xmath60 .
we can also determine the size of power corrections from the fitting data @xcite by the asymptotic form ( [ asympt ] ) plus next - to - leading - order power correction .
our analysis shows that the uncertainty is not greater than @xmath61 .
we believe that these arguments confirm the applicability of the asymptotic expression for the cross section .
now let us consider the processes @xmath62 , @xmath63 .
unfortunately at the moment only the low energy cleo - c data @xcite are available for these processes . as it is seen from tab .
[ tab : params ] the experimental error of these data are rather large .
more precise experimental data can be obtained from the decays @xmath64 .
corresponding branching fraction are equal to @xcite @xmath65 generally speaking , these decays can proceed both via strong and electromagnetic interaction ( see fig .
[ diag : psi ] for the typical diagrams ) . because of the isospin violation , however , the gluon induced diagrams are strongly suppressed , and purely electromagnetic diagram gives the dominant contribution .
the branching fraction of these decays are equal to @xmath66 from this relation we can determine the value of the function @xmath67 ) and , neglecting energy dependence in @xmath47 ( this assumptions leads to additional errors , which can be estimated as @xmath68 ) , predict the cross sections of the processes @xmath62 and @xmath63 over the large energy region .
the results are shown in figs . [
fig : rhoeta ] , [ fig : rhoetap ] and table [ tab : params ] . from these results
it is seen that the low energy data do not allow to understand what hypotheses gives the best agreement with the experiments .
so , to determine the asymptotic behavior unambiguously the high energy data are needed .
however , if we assume that the asymptotic behavior for the cross section of the process @xmath51 is the same as that for the processes @xmath62 , @xmath63 , one can predict the cross sections of these processes at energies @xmath69 gev @xmath70 the cross sections at the energy @xmath71 gev are in agreement with the cleo - c data . .
the constants @xmath72 are fixed from the experimental values of this cross section in the low energy region ( @xmath73).[fig : phieta ] ] decay[diag : psi],title="fig : " ] decay[diag : psi],title="fig : " ] .
the constants @xmath72 are fixed from the @xmath74 branching fraction ( the rightmost point represents the cross section at @xmath75 calculated using relations ( [ eq : sigma ] ) , ( [ eq : f ] ) and ( [ eq : br ] ) . ) [ fig : rhoeta ] ] .
the constants @xmath72 are fixed from the @xmath76 branching fraction ( the rightmost point represents the cross section at @xmath75 calculated using relations ( [ eq : sigma ] ) , ( [ eq : f ] ) and ( [ eq : br ] ) .
[ fig : rhoetap ] ] .
the constants @xmath72 are fixed from the experimental values of this cross section at @xmath77 gev .
[ fig : phietap ] ] let us now proceed with the reaction @xmath78 . in this case
experimental information is even more poor . in the case
one can not use the branching fraction of the decay @xmath79 , since in this case isospin is conserved and the diagram shown in fig .
[ diag : psi]a gives the main contribution to @xmath80 decay .
in addition , cleo - c gives us only upper bound on the cross section @xmath17 , and no experimental information in high energy region is available . using @xmath81 mixing it is possible , however , to estimate the cross section @xmath82 from the value of the @xmath83 cross section .
the mixing of pseudoscalar mesons can be described by different parameterizations @xcite . in our paper
we will use the parametrization of @xmath81 mixing in quark flavor basis with one mixing angle @xcite : @xmath84 where @xmath85 and @xmath86 represents the basis of the quark mixing scheme . using this mixing scheme
it is easy to obtain the following relation between @xmath83 and @xmath87 cross sections : @xmath88 there are plenty of theoretical and experimental works dedicated to the determination of the mixing angle .
the well known estimation , based on gell - mann - okubo mass formulas give the value of @xmath89 about @xmath90 for linear mass formula and about @xmath91 for a quadratic case .
the analysis of the axial anomaly generated decays @xmath92 was performed in papers @xcite and the estimate @xmath93 was obtained . in papers
@xcite another anomaly based investigation of a large set of decay processes was performed and the value @xmath94 was presented .
the authors of the recent work @xcite used dispersive approach to @xmath81 mixing and obtained the value @xmath95 it is interesting to note that this value is close to phenomenological value @xmath96 , presented in pioneering work @xcite .
we will use this value of the mixing angle in our article .
the error of this angle is small in comparison with the error in @xmath83 cross section . knowing the value of the cross section @xmath97 one can determine the value of the cross section @xmath82 at the energy @xmath3 gev .
@xmath98 from this we can calculate the value of the constants @xmath72 for different hypothesis and predict the cross section of the process @xmath87 in the low energy region .
we show the energy dependence of this cross section in fig .
[ fig : phietap ] . from this figure
one sees that the value @xmath54 contradicts experimental results at @xmath18 gev , while @xmath99 and @xmath53 do not . from theoretical arguments and presented above analysis of @xmath83 reaction we think , that the value @xmath53 is more preferable .
for this hypothesis we can predict the value of the cross section at energy @xmath71 gev @xmath100 which does not contradict to the cleo - c data . in table
[ tab : params ] we present the results of this section .
second column contains constants @xmath101 for different final states , obtained from low energy fits . in the third and fifth columns experimental results for the cross sections @xmath102 at the center
mass energies @xmath18 gev and 10.6 gev are presented .
the results of the calculation are shown in the forth and sixth columns . from this table
it is clear , that only relation @xmath103 agrees with experiment . [
cols="^,^,^,^,^,^ " , ] the following point deserves consideration .
the models for the das that we use in our calculation are truncated series in gegebauer polynomials .
this means that the end point behavior ( @xmath104 ) of these das coincides with the end point behavior of the corresponding asymptotic das . from this
it is not difficult to see that the integral @xmath105 is logarithmically divergent .
one way to regularize this divergence is to introduce cut off parameter @xmath106 . in our calculation
we take the following value of cut off parameter @xmath107 .
@xmath108 is of order of typical hadronic scale , which is of order of several hundreds mev .
physical meaning of this cut off can be understood as follows : if @xmath109 quark momentum is of order of @xmath110 and in this region one must take into account transverse motion in hadron what regularizes the whole integral @xmath105 .
the parameter @xmath108 can be determined from available experimental results .
we determined this parameter from the cross section of the process @xmath111 at @xmath112 gev .
thus we get @xmath113 mev .
the variation of parameter @xmath108 corresponds @xmath114 deviation from the central value measured at belle collaboration .
the results of the calculation are shown in table [ res1 ] .
the second and fifth columns contain the experimental results for the cross sections at the energies @xmath18 gev and @xmath77 gev correspondingly .
the result obtained in paper @xcite are shown in the third and sixth columns .
the results obtained in this paper are shown in the fourth and seventh columns .
the variation of the results are due to the variation in parameter @xmath108 .
it is seen from this table that the results of the calculation are in satisfactory agreement with the experiment .
it should be noted here that in addition to the diagrams shown in fig .
[ fig : diags ] , there is additional contribution to the formfactor @xmath27 which was not considered in this paper .
this contribution appears if one takes into account higher fock state of the vector and pseudoscalar mesons @xmath115 and it s asymptotic behavior is also @xmath103 .
this contribution was considered in papers @xcite . in paper
@xcite the authors asserted that the fock state @xmath115 gives very important contribution to the cross section and must be taken into account .
contrary to this conclusion , the author of paper @xcite asserted that the main contribution arises from the diagrams shown in fig .
[ fig : diags]c .
so , the question about the role of higher fock state @xmath115 in the total cross section deserves separate consideration and it will not be considered here .
in this work we have studied the production of light mesons @xmath116 , @xmath117 , @xmath118 and @xmath119 in the high energy electron - positron annihilation .
the first question studied in this paper is the asymptotic behavior of the cross sections of the processes under consideration .
perturbative qcd predicts , that the cross section of the reaction @xmath12 has the following asymptotic behavior @xmath103 in the limit @xmath5 . experimental measurement of the cross section of the process @xmath2 at the large center mass energy @xmath3 gev @xcite and the low energy experimental data @xmath4 gev
@xcite give us the possibility to study the cross section in the broad energy region .
as the result , we have determined the asymptotic behavior of the cross section of @xmath2 in the limit @xmath5 , which is in agreement with perturbative qcd prediction . as to the other processes under study , there are no high energy experimental data , which allow us to confirm perturbative qcd prediction for these processes .
we would like to stress here that the high energy experimental data turned out to be crucial in the determination of the asymptotic behaviour of the cross sections . assuming that the asymptotic behavior predicted by perturbative qcd is valid for the other processes under consideration , we have calculated the cross sections of the processes @xmath120 at the energies @xmath121 gev . in addition , we have applied perturbative qcd approach to calculate the cross sections of the processes under study at the energies @xmath122 gev and @xmath112 gev .
the results of this calculation are in satisfactory agreement with available experimental data .
the authors would like to thank a.a .
sokolov and m.m .
shapkin for useful and stimulating discussions .
this work was partially supported by russian foundation of basic research under grant 07 - 02 - 00417 .
the work of v. braguta was partially supported by crdf grant y3-p-11 - 05 and president grant mk-2996.2007.2 .
the work of a. luchinsky was partially supported by president grant mk-110.2008.2 and russian science support foundation .
b. aubert _ et al . _ [ babar collaboration ] , phys .
d * 74 * , 111103 ( 2006 ) [ arxiv : hep - ex/0611028 ] .
n. e. adam _ et al . _
[ cleo collaboration ] , phys .
* 94 * , 012005 ( 2005 ) [ arxiv : hep - ex/0407028 ] .
w. m. yao _ et al .
_ [ particle data group ] , j. phys .
g * 33 * , 1 ( 2006 ) .
j. f. donoghue , b. r. holstein and y. c. r. lin , phys .
* 55 * ( 1985 ) 2766 [ erratum - ibid .
* 61 * ( 1988 ) 1527 ] .
f. j. gilman and r. kauffman , phys .
d * 36 * , 2761 ( 1987 ) [ erratum - ibid .
d * 37 * , 3348 ( 1988 ) ] .
r. akhoury and j. m. frere , phys .
b * 220 * , 258 ( 1989 ) .
p. ball , j. m. frere and m. tytgat , phys .
b * 365 * , 367 ( 1996 ) [ arxiv : hep - ph/9508359 ] .
y. n. klopot , a. g. oganesian and o. v. teryaev , arxiv:0810.1217 [ hep - ph ] .
|
this paper is devoted to consideration of the hard exclusive processes @xmath0 , where @xmath1 .
experimental measurement of the cross section of the process @xmath2 at babar collaboration at large center mass energy @xmath3 gev and some low energy experimental data @xmath4 gev give us the possibility to study the cross section in the broad energy region . as the result
, we have determined the asymptotic behavior of the cross section of @xmath2 in the limit @xmath5 , which is in agreement with perturbative qcd prediction .
assuming that the same asymptotic behavior is valid for the other processes under consideration and using low energy experimental data we have predicted the cross sections of these processes at energies @xmath6 gev . in addition , we have calculated the cross sections of these processes at the same energies within perturbative qcd .
our results are in agreement with available experimental data .
|
in the past decade , distributed consensus algorithms have attracted a significant amount of research attention @xcite , due to their wide applications in distributed control and estimation @xcite , distributed signal processing @xcite , and distributed optimization methods @xcite .
the basic idea is that a network of interconnected nodes with different initial values can reach a common value , namely an agreement or a consensus , via local information exchange as long as the communication graph is well connected .
the consensus value is informative in the sense that it can depend on all nodes initial states even if the value is not the exact average @xcite ; the consensus processes are robust against random or deterministic switching of network interactions as well as against noises @xcite . as a generalized notion ,
constrained consensus seeks to reach state agreement in the intersection of a number of convex sets , where each set serves as the supporting state space for a particular node @xcite .
it was shown that with doubly stochastic arc weights , a projected consensus algorithm , where each node iteratively projects a weighted average of its neighbor s values onto its supporting set , can guarantee convergence to constrained consensus when the intersection set is bounded and contains an interior point @xcite .
this idea was then extended to the continuous - time case under which the graph connectivity and arc weight conditions can be relaxed with convergence still being guaranteed @xcite .
the projected consensus algorithm was shown to be convergent even under randomized projections @xcite .
in fact , those developments are closely related to a class of so - called alternating projection algorithms , which was first studied by von neumann in the 1940s @xcite and then extensively investigated across the past many decades @xcite .
a related but different intriguing problem lies in how to develop distributed algorithms that solve a linear algebraic equation @xmath7 with respect to variable @xmath8 , to which tremendous efforts have been devoted with many algorithms developed under different assumptions of information exchanges among the nodes @xcite .
naturally , we can assume that a node @xmath1 has access to a row vector , @xmath2 of @xmath3 as well as the corresponding entry , @xmath4 in @xmath5 .
each node will then be able to determine with straightforward calculations an affine subspace whose elements satisfy the equation @xmath9 .
this means that , as long as the original equation @xmath0 has a nonempty solution set , solving the equation would be equivalent to finding a point in the intersection of all the affine subspaces and therefore can be put into the constrained consensus framework . if however the linear equation has no exact solution and its least squares solutions are of interest , we ended up with a convex optimization problem with quadratic cost and linear constraints .
many distributed optimization algorithms , such as @xcite , developed for much more complex models , can therefore be directly applied .
it is worth emphasizing that the above ideas of distributed consensus and optimization were traced back to the seminal work of bertsekas and j. tsitsiklis @xcite . in this paper
, we study two distributed network flows as distributed solvers for such linear algebraic equations in continuous time .
the first so - called consensus + projection " flow consists of two additive terms , one from standard consensus dynamics and the other contributing to projection onto each affine subspace .
the second projection consensus " flow on the other hand simply replaces the relative state feedback in consensus dynamics with projected relative state feedback .
essentially only relative state information is exchanged among the nodes for both of the two flows , which justifies their full _
distributedness_. to study the asymptotic behaviours of the two distributed flows with respect to the solutions of the linear equation , new challenges arise in the loss of intersection boundedness and interior points for the exact solution case as well as in the complete loss of intersection sets for the least squares solution case . as a result
, the analysis can not be simply mapped back to the studies in @xcite .
the contributions of the current paper are summarized as follows .
* under mild conditions on the communication graphs ( without requiring a dwell - time for switching graphs and only requiring a positively lower bound on the integral of arc weights over certain time intervals ) , we prove that all node states asymptotically converge to a common solution of the linear algebraic equation for the two flows , if there is any .
the convergence is global for the consensus + projection " flow , and local for the projection consensus " flow in the sense that the initial values must be put into the affine subspaces ( which is a very minor restriction indeed ) .
we manage to characterize the node limits for balanced or fixed graphs . *
if the linear equation has no exact solutions , we show that the node states can be forced to converge to a ball of fixed but arbitrarily small radius surround the least squares solution by taking the gain of the consensus dynamics to be sufficiently large for consensus + projection " flow under fixed and undirected graphs .
semi - global convergence to approximate least squares solutions is established for switching balanced directed graphs under suitable conditions . a minor , but more explicit result arises where it is also shown that the projection consensus " flow drives the average of the node states to the least squares solution with complete communication graphs .
these results rely on our development of new technique of independent interest for treating the interplay between consensus dynamics and the projections onto affine subspaces in the absence of intersection boundedness and interior point assumptions .
all the convergence results can be sharpened to provide exponential convergence with suitable conditions on the switching graph signals .
the remainder of this paper is organized as follows .
section [ sec : problem ] introduces the network model , presents the distributed flows under consideration , and defines the problem of interest .
section [ sec : exact ] discusses the scenario where the linear equation has at least one solution .
section [ sec : least - squares ] then turns to the least squares case where the linear equation has no solution at all .
finally , section [ sec : numerical ] presents a few numerical examples illustrating the established results and section [ sec : conclusions ] concludes the paper with a few remarks .
a directed graph ( digraph ) is an ordered pair of two sets denoted by @xmath10 @xcite . here
@xmath11 is a finite set of vertices ( nodes ) .
each element in @xmath12 is an ordered pair of two distinct nodes in @xmath13 , called an arc .
a directed path in @xmath14 with length @xmath15 from @xmath16 to @xmath17 is a sequence of distinct nodes , @xmath18 , such that @xmath19 , for all @xmath20 .
a digraph @xmath21 is termed _ strongly connected _ if for any two distinct nodes @xmath22 , there is a path from @xmath1 to @xmath23 .
a digraph is called _ bidirectional _ when @xmath24 if and only if @xmath25 for all @xmath1 and @xmath23 .
a strongly connected bidirectional digraph is simply called _
connected_. all vectors are column vectors and denoted by bold , lower case letters , i.e. , @xmath26 , etc . ; matrices are denoted with bold , upper case letters , i.e. , @xmath27 , etc . ; sets are denoted with @xmath28 , etc .
depending on the argument , @xmath29 stands for the absolute value of a real number or the cardinality of a set .
the euclidean inner product between two vectors @xmath30 and @xmath31 in @xmath32 is denoted as @xmath33 , and sometimes simply @xmath34 .
the euclidean norm of a vector is denoted as @xmath35 .
consider the following linear algebraic equation : @xmath36 with respect to variable @xmath37 , where @xmath38 and @xmath39 .
we know from the basics of linear algebra that overall there are three cases .
* there exists a unique solution satisfying eq .
( [ linearequation ] ) : @xmath40 and @xmath41 .
* there is an infinite set of solutions satisfying eq .
( [ linearequation ] ) : @xmath42 and @xmath41 .
* there exists no solution satisfying eq .
( [ linearequation ] ) : @xmath43 .
we denote @xmath44 with @xmath2 being the @xmath1-th row vector of @xmath3 . for the ease of presentation and with inessential loss of generality we assume throughout the rest of the paper that @xmath45 introduce @xmath46 for each @xmath47 , which is an affine subspace .
it is then clear that case ( i ) is equivalent to the condition that @xmath48 is a singleton , and that case ( ii ) is equivalent to the condition that @xmath48 is an affine space with a nontrivial dimension . for case ( iii ) , a least squaress solution of ( [ linearequation ] ) can be defined via the following optimization problem : @xmath49 which yields a unique solution @xmath50 if @xmath40 .
consider a network with nodes indexed in the set @xmath51 .
each node @xmath1 has access to the value of @xmath6 and @xmath52 without the knowledge of @xmath53 or @xmath54 from other nodes .
each node @xmath1 holds a state @xmath55 and exchanges this state information with other neighbor nodes , these being determined by the edges of the graph of the network .
we are interested in distributed flows for the @xmath56 that asymptotically solve the equation ( [ linearequation ] ) , i.e. , @xmath56 approaches some solution of ( [ linearequation ] ) as @xmath57 grows .
let @xmath58 denote the set of all directed graphs associated with node set @xmath59 .
node interactions are described by a signal @xmath60 .
the digraph that @xmath61 defines at time @xmath57 is denoted as @xmath62 , where @xmath63 is the set of arcs .
the neighbor set of node @xmath1 at time @xmath57 , denoted @xmath64 , is given by @xmath65 this is to say , at any given time @xmath57 , node @xmath1 can only receive information from the nodes in the set @xmath64 .
let @xmath66 and @xmath67 be the sets of nonnegative and positive real numbers , respectively .
associated with each ordered pair @xmath68 there is a function @xmath69 representing the weight of the possible connection @xmath68 .
we impose the following assumption , which will be adopted throughout the paper without specific further mention . *
weights assumption .
* the function @xmath70 is continues except for at most a set with measure zero over @xmath66 for all @xmath22 ; there exists @xmath71 such that @xmath72 for all @xmath73 and all @xmath22 .
denote @xmath74 as an indicator function for all @xmath75 , where @xmath76 if @xmath77 and @xmath78 otherwise .
we impose the following definition on the connectivity of the network communication structures .
* an arc @xmath68 is said to be a @xmath79-arc of @xmath80 for the time interval @xmath81 if @xmath82 * @xmath80 is @xmath79-uniformly jointly strongly connected ( @xmath79-@xmath83 ) if there exists @xmath84 such that the @xmath79-arcs of @xmath80 on time interval @xmath85 form a strongly connected digraph for all @xmath86 ; * @xmath80 is @xmath79-bidirectionally infinitely jointly connected ( @xmath79-@xmath87 ) if @xmath80 is bidirectional for all @xmath88 and the @xmath79-arcs of @xmath80 on time interval @xmath89 form a connected graph for all @xmath86 .
define the mapping @xmath90 as the projection onto the affine subspace @xmath91 .
let @xmath92 be a given constant .
we consider the following continuous - time network flows : [ * consensus + projection " flow * ] : @xmath93 [ * projection consensus " flow * ] : @xmath94 the first term of ( [ 1 ] ) is a standard consensus flow @xcite , while along @xmath95 @xmath96 will be asymptotically projected onto @xmath97 since ( [ r1 ] ) is a gradient flow with @xmath98 , where @xmath99 is a @xmath100 convex function .
the flow ( [ 2 ] ) simply replaces the relative state in standard consensus flow with relative projective state .
notice that a particular equilibrium point of these equations is given by @xmath101 , where @xmath102 is a solution of eq .
( [ linearequation ] ) .
the aim of the two flows is to ensure that the @xmath56 asymptotically tend to a solution of eq .
( [ linearequation ] ) .
note that , the projection consensus flow ( [ 2 ] ) can be rewritten as . ]
@xmath103 therefore , in both of the consensus + projection " and the projection consensus " flows , a node @xmath1 essentially only receives the information @xmath104 from its neighbors , and the flows can then be utilized in addition with the @xmath6 and @xmath4 it holds . in this way , the flows ( [ 1 ] ) and ( [ 2 ] ) are _
distributed_. without loss of generality we assume the initial time is @xmath105 .
we denote by @xmath106 a trajectory along any of the two flows .
the two network flows are intrinsically different in their geometries .
in fact , the consensus + projection " flow is a special case of the optimal consensus flow proposed in @xcite consisting of two parts , a consensus part and another projection part .
the projection consensus " flow , first proposed and studied in @xcite for fixed bidirectional graphs , is the continuous - time analogue of the projected consensus algorithm proposed in @xcite . because it is a gradient descent on a riemannian manifold
if the communication graph is undirected and fixed , there is guaranteed convergence to an equilibrium point .
the two flows are closely related to the alternating projection algorithms , first studied by von neumann in the 1940s @xcite .
we refer to @xcite for a thorough survey on the developments of alternating projection algorithms .
we illustrate the intuitive difference between the two flows in figure [ fig : twoflows ] .
the dynamics described in systems ( [ 1 ] ) and ( [ 2 ] ) are linear , in contrast to the nonlinear dynamics associated with the general convex sets studied in @xcite .
however , we would like to emphasize that new challenges arise with the systems ( [ 1 ] ) and ( [ 2 ] ) compared to the work of @xcite .
first of all , for both of the cases ( i ) and ( ii ) , @xmath48 contains no interior point .
this interior point condition however is essential to the results in @xcite .
next , for case ( ii ) where @xmath48 is an affine space with a nontrivial dimension , the boundedness condition for the intersection of the convex sets no longer holds , which plays a key role in the analysis of @xcite .
finally , for case ( iii ) , @xmath48 becomes an empty set .
the least squares solution case then completely falls out from the discussions of constrained consensus in @xcite .
we have assumed hitherto that each node @xmath1 only has access to the value of @xmath6 and @xmath52 from the equation ( [ linearequation ] ) and therefore there are a total of @xmath107 nodes .
alternatively and as a generalization , there can be @xmath108 nodes with node @xmath1 holding @xmath109 rows , denoted @xmath110 of @xmath111 with @xmath112 . in this case
, we can revise the definition of @xmath91 to @xmath113 which is nonempty if ( [ linearequation ] ) has at least one solution .
then the two flows ( [ 1 ] ) and ( [ 2 ] ) can be defined in the same manner for the @xmath114 nodes .
of course , if @xmath115 is large , finding an initial condition consistent with @xmath91 becomes computationally more burdensome .
let to be disjoint . ]
@xmath116 note that , the @xmath91 are still affine subspaces , while solving ( [ linearequation ] ) exactly continues to be equivalent to finding a point in @xmath117 .
consequently , all our results for cases ( i ) and ( ii ) apply also to this new setting with row grouping .
in this section , we show how the two distributed flows asymptotically solve the equation ( [ linearequation ] ) under quite general conditions for cases ( i ) and ( ii ) .
we first focus on the case when eq .
( [ linearequation ] ) admits a unique solution @xmath118 , or equivalently , @xmath119 is a singleton . for the consensus + projection " flow ( [ 1 ] ) , the following theorem holds
. [ thm1 ] let ( i ) hold .
then along the consensus + projection " flow ( [ 1 ] ) , there holds @xmath120 for all initial values if @xmath80 is either @xmath79-@xmath83 or @xmath79-@xmath87 .
the projection consensus " flow ( [ 2 ] ) , however , can only guarantee local convergence for a particular set of initial values .
the following theorem holds .
[ thm2 ] let ( i ) hold .
suppose @xmath121 for all @xmath1 .
then along the projection consensus " flow ( [ 2 ] ) , there holds @xmath120 if @xmath80 is either @xmath79-@xmath83 or @xmath79-@xmath87 . convergence along the projection consensus " flow relies on specific initial values due to the existence of equilibriums other than the desired consensus states within the set @xmath122 : if @xmath123 are all equal , then obviously they will stay there for ever along the flow ( [ 2 ] ) .
it was suggested in @xcite that one can add another term in the projection consensus " flow and arrive at @xmath124 for @xmath125 , then convergence will be global under ( [ 100 ] ) .
we would like to point out that ( [ 100 ] ) has a similar structure as the consensus + projection " flow with the consensus dynamics being replaced by projection consensus .
in fact , our analysis of the consensus + projection " flow developed in this paper can be refined and generalized to the flow ( [ 100 ] ) and then leads to results under the same graph conditions for both exact and least squares solutions .
we now turn to the scenario when eq .
( [ linearequation ] ) has an infinite set of solutions , i.e. , @xmath48 is an affine space with a nontrivial dimension .
we note that in this case @xmath122 is no longer a bounded set ; nor does it contain interior points .
this is in contrast to the situation studied in @xcite . for the consensus + projection " flow ( [ 1 ] )
, we present the following result .
[ thm3 ] let ( ii ) hold . then along the consensus + projection " flow ( [ 1 ] ) and for any initial value @xmath126 , there exists @xmath127 , which is a solution of ( [ linearequation ] ) , such that @xmath128 if @xmath80 is either @xmath79-@xmath83 or @xmath79-@xmath87 . for the projection consensus " flow ,
convergence relies on restricted initial nodes states .
[ thm4 ] let ( ii ) hold . then along the projection consensus
" flow ( [ 2 ] ) and for any initial value @xmath126 with @xmath121 for all @xmath1 , there exists @xmath127 , which is a solution of ( [ linearequation ] ) , such that @xmath128 if @xmath80 is either @xmath79-@xmath83 or @xmath79-@xmath87 . for any given graph signal @xmath80 , the value of @xmath127 in theorems [ thm3 ] and [ thm4 ] depends only on the initial value @xmath126 . we manage to provide a characterization to @xmath127 for balanced switching graphs or fixed graphs .
denote @xmath129 as the projection operator over @xmath122 .
[ prop1 ] the following statements hold for both the consensus + projection " and the projection consensus " flows .
* suppose @xmath80 is balanced , i.e. , @xmath130 for all @xmath88 and for all @xmath22 .
suppose in addition that @xmath80 is either @xmath79-@xmath83 or @xmath79-@xmath87 .
then @xmath131 * suppose @xmath132 for some fixed , strongly connected , digraph @xmath133 and for any @xmath22 , @xmath134 for some constant @xmath135 .
let @xmath136 with @xmath137 be the left eigenvector corresponding to the simple eigenvalue zero of the laplacian associated with the graph @xmath133 under the given arc weights is defined as @xmath138 where @xmath139 $ ] and @xmath140 in fact , we have @xmath141 for all @xmath142 if @xmath133 is strongly connected @xcite . ] @xmath143 of the digraph @xmath133 .
then we have @xmath144 due to the linear nature of the systems ( [ 1 ] ) and ( [ 2 ] ) , it is straightforward to see that the convergence stated in theorems [ thm1 ] , [ thm2 ] , [ thm3 ] and [ thm4 ] is exponential if @xmath80 is periodic and @xmath79-@xmath83 .
it is however difficult to provide tight approximations of the exact convergence rates because of the general structure with switching interactions adopted by our network model . for time - invariant networks with constant edge weights ,
the two flows become linear time - invariant , and then the convergence rate is determined by the spectrum of the state transition matrix : the rate of convergence for the consensus + projection " flow is determined together by both the graph laplacian and the structure of the linear manifolds @xmath91 and one in fact can not dominate another ; the rate of convergence for the projection consensus " flow is upper bounded by the smallest nonzero eigenvalue of the graph laplacian @xcite . before presenting the detailed proofs for the stated results , we recall some preliminary theories from affine subspaces , convex analysis , invariant sets , and dini derivatives , as well as a few auxiliary lemmas which will be useful for the analysis .
an affine space is a set @xmath145 that admits a free transitive action of a vector space @xmath146 .
a set @xmath147 is said to be _ convex _ if @xmath148 whenever @xmath149 and @xmath150 @xcite . for any set @xmath151 , the intersection of all convex sets containing @xmath152
is called the _ convex hull _ of @xmath152 , denoted by @xmath153 .
let @xmath154 be a closed convex subset in @xmath155 and denote @xmath156 as the distance between @xmath157 and @xmath154 .
there is a unique element @xmath158 satisfying @xmath159 associated to any @xmath157 @xcite .
the map @xmath160 is called the _ projector _ onto @xmath154 and @xmath129 introduced earlier are consistent with this definition . ] .
the following lemma holds @xcite .
[ lemconvex ] ( i ) @xmath161 .
\(ii ) @xmath162 .
\(iii ) the function @xmath163 is continuously differentiable at every @xmath164 with @xmath165 .
consider the following autonomous system @xmath166 where @xmath167 is a continuous function .
let @xmath168 be a solution of ( [ i1 ] ) with initial condition @xmath169 .
then @xmath170 is called a _ positively invariant set _ of ( [ i1 ] ) if , for any @xmath171 and any @xmath172 , we have @xmath173 , @xmath174 , along every solution @xmath168 of ( [ i1 ] )
. the _ upper dini derivative _ of a continuous function @xmath175 ( @xmath176 ) at @xmath57 is defined as @xmath177 when @xmath178 is continuous on @xmath179 , @xmath178 is non - increasing on @xmath179 if and only if @xmath180 for any @xmath181 .
hence if @xmath180 for all @xmath88 and @xmath182 is lower bounded then the limit of @xmath182 exists as a finite number when @xmath57 tends to infinity .
the next result is convenient for the calculation of the dini derivative @xcite .
[ lemdini ] let @xmath183 be a continuously differentiable function with respect to both @xmath57 and @xmath184 and @xmath185 .
let @xmath186 be an absolutely continuous function . if @xmath187 is the set of indices where the maximum is reached at @xmath57 , then @xmath188 the following lemmas , lemmas [ lem : affineprojection ] , [ lem : subspace - affine - projection ] , [ lem : doubleprojection ] can be easily proved using the properties of affine spaces , which turn out to be useful throughout our analysis . we therefore collect them below and the details of their proofs
are omitted .
[ lem : affineprojection ] let @xmath189 be an affine subspace , where @xmath190 with @xmath191 , and @xmath192 . denote @xmath193 as the projection onto @xmath154 .
then @xmath194 for all @xmath37 .
consequently there holds @xmath195 for all @xmath196 .
[ lem : subspace - affine - projection ] let @xmath154 be an affine subspace in @xmath32 . take an arbitrary point @xmath197 and define @xmath198 , which is a subspace in @xmath32 .
denote @xmath193 and @xmath199 as the projectors onto @xmath154 and @xmath200 , respectively .
then @xmath201 for all @xmath196 .
[ lem : doubleprojection ] let @xmath202 and @xmath203 be two affine subspaces in @xmath32 with @xmath204 . denote @xmath205 and @xmath206 as the projections onto @xmath202 and @xmath203 , respectively
. then @xmath207 [ lem : invariant ] suppose either ( i ) or ( ii ) holds .
take @xmath208 as an arbitrary solution of ( [ linearequation ] ) and let @xmath209 be arbitrary .
define @xmath210 then * @xmath211 is a positively invariant set for the consensus + projection " flow ( [ 1 ] ) ; * @xmath212 is a positively invariant set for the projection consensus " flow ( [ 2 ] ) . _
_ let @xmath213 be a trajectory along the flows ( [ 1 ] ) or ( [ 2 ] ) .
define @xmath214 denote @xmath215 .
\(i ) from lemma [ lemdini ] we obtain that along the flow ( [ 1 ] ) @xmath216\nonumber\\ & \stackrel{b)}{\leq } \max_{i\in\mathcal{i}(t)}\bigg [ k \sum_{j\in \mathrm{n}_i(t)}\frac{a_{ij}(t)}{2 } \big(\big\|\mathbf{x}_j(t)-\mathbf{y}^\sharp\big\|^2 - \big\|\mathbf{x}_i(t)-\mathbf{y}^\sharp\big\|^2 \big)-\big|\mathbf{h}_i^t ( \mathbf{x}_i(t)-\mathbf{y}^\sharp)\big|^2 \bigg]\nonumber\\ & \stackrel{c)}{\leq } \max_{i\in\mathcal{i}(t ) } \bigg[-\big|\mathbf{h}_i^t ( \mathbf{x}_i(t)-\mathbf{y}^\sharp)\big|^2 \bigg]\nonumber\\ & \leq 0,\end{aligned}\ ] ] where @xmath217 follows from the fact that @xmath218 in view of lemma [ lem : affineprojection ] and the fact that @xmath208 is a solution of ( [ linearequation ] ) ; @xmath219 makes use of the elementary inequality @xmath220 @xmath221 follows from the definition of @xmath222 and @xmath223 .
we therefore know from ( [ 3 ] ) that @xmath222 is always a non - increasing functions .
this implies that , if @xmath224 , i.e. , @xmath225 for all @xmath1 , then @xmath226 for all @xmath1 and all @xmath174 .
equivalently , @xmath227 is a positively invariant set for the flow ( [ 1 ] ) .
\(ii ) let @xmath228 .
the structure of the flow ( [ 2 ] ) immediately tells that @xmath229 for all @xmath174 .
furthermore , again by lemma [ lemdini ] we obtain @xmath230\nonumber\\ & \stackrel{d)}{\leq } \max_{i\in\mathcal{i}(t)}\bigg [ \sum_{j\in \mathrm{n}_i(t)}\frac{a_{ij}(t)}{2 } \big(\big\|\mathbf{x}_j(t)-\mathbf{y}^\sharp\big\|^2 - \big\|\mathbf{x}_i(t)-\mathbf{y}^\sharp\big\|^2 \big)\bigg]\nonumber\\ & \leq 0,\end{aligned}\ ] ] where @xmath217 follows from lemma [ lem : affineprojection ] , @xmath219 holds due to the fact that @xmath231 is a projection matrix , @xmath221 is again based on the inequality ( [ 4 ] ) , and @xmath232 holds because @xmath233 ( so that @xmath234 ) and that @xmath231 is a projection matrix ( so that @xmath235 ) .
we can therefore readily conclude that @xmath212 is a positively invariant set for the projection consensus " flow ( [ 2 ] ) .
@xmath236 we now have the tools in place to present the proofs of the stated results .
when ( i ) holds , @xmath237 is a singleton , which is obviously a bounded set .
the consensus + projection " flow ( [ 1 ] ) is a special case of the optimal consensus flow proposed in @xcite .
theorem [ thm1 ] readily follows from theorems 3.1 and theorems 3.2 in @xcite by adapting the treatments to the weights assumption adopted in the current paper using the techniques established in @xcite .
we therefore omit the details .
being a special case of theorem [ thm4 ] , theorem [ thm2 ] holds true as a direct consequence of theorem [ thm4 ] , whose proof will be presented below .
recall that @xmath208 is an arbitrary solution of ( [ linearequation ] ) .
with lemma [ lem : invariant ] , for any initial value @xmath126 , the set @xmath238 which is obviously bounded , is a positively invariant set for the consensus + projection " flow ( [ 1 ] ) .
define @xmath239 and @xmath240 . as a result , recalling theorem 3.1 and theorem 3.2 from @xcite .
] , if @xmath80 is either @xmath79-@xmath83 or @xmath79-@xmath87 , there hold \(i ) @xmath241 for all @xmath125 ; \(ii ) @xmath242 for all @xmath1 and @xmath23 .
we still need to show that the limits of the @xmath56 indeed exist .
introduce @xmath243 and @xmath244 .
with lemma [ lem : doubleprojection ] , we see that @xmath245 as a result , taking @xmath246 from both the left and right sides of ( [ 1 ] ) is a projector onto a subspace . ] , we obtain @xmath247 this is to say , if @xmath80 is either @xmath79-@xmath83 or @xmath79-@xmath87 , we can invoke theorem 4.1 ( for @xmath79-@xmath83 graphs ) and theorem 5.2 ( for @xmath79-@xmath87 graphs ) in @xcite to conclude : for any initial value @xmath126 , there exists @xmath248 such that @xmath249 while on the other hand @xmath250 leads to @xmath251 where the last equality follows from lemma [ lem : subspace - affine - projection ] .
we conclude from ( [ 5 ] ) and ( [ 6 ] ) that @xmath252 we have now completed the proof of theorem [ thm3 ] .
we continue to use the notation @xmath239 and @xmath240 introduced in the proof of theorem [ thm3 ] . in view of lemma [ lem : invariant ] , for any initial value @xmath126 , @xmath253 is a positively invariant set for the projection consensus " flow ( [ 2 ] ) .
define @xmath254 let @xmath255 .
we obtain from lemma [ lemconvex].(iii ) that @xmath256 in ( [ 7 ] ) , the argument used to establish ( [ 8 ] ) can be carried through when @xmath208 is replaced by @xmath257 .
accordingly , we obtain ) hold without relying on the fact that @xmath208 is a constant . ] , @xmath258 .
this immediately implies that there is a constant @xmath259 such that @xmath260 . as a result ,
for any @xmath261 , there exists @xmath262 such that @xmath263 now that @xmath264 is a nonnegative constant satisfying @xmath260 , we can in fact show that @xmath265 with suitable graph conditions , as summarized in the following two technical lemmas .
details of their proofs can be found in the appendix .
[ lem : directed ] let @xmath80 be @xmath79-@xmath83
. then @xmath265 .
in fact , @xmath265 due to the following two contradictive conclusions : \(i ) if @xmath266 , then there holds @xmath267 for all @xmath142 along the projection consensus " flow ( [ 2 ] ) .
\(ii ) if @xmath267 for all @xmath142 along the projection consensus " flow ( [ 2 ] ) , then @xmath265 .
[ lemma - bidirectional ] suppose @xmath80 is @xmath79-@xmath87
. then @xmath265 along the projection consensus " flow ( [ 2 ] ) .
recall @xmath268 .
applying @xmath246 to both the left and right sides of ( [ 2 ] ) , we have @xmath269 where we have used lemma [ lem : subspace - affine - projection ] to derive @xmath270 we can again make use of the argument in the proof of theorem [ thm3 ] and conclude that the limits of the @xmath271 exist and they are obviously the same .
we have now completed the proof of theorem [ thm4 ] .
we provide detailed proof for the projection+consensus " flow .
the analysis for the projection consensus flow can be similarly established in view of ( [ 25 ] ) .
\(i ) with ( [ 20 ] ) , we have @xmath272 along the projection+consensus " flow .
note that ( [ 22 ] ) is a standard consensus flow with arguments being @xmath273 , @xmath47 . as a result
, there holds @xmath274 if @xmath80 is balanced for all @xmath57 @xcite . as a result , similar to ( [ 21 ] ) , we have @xmath275 \(ii ) if @xmath132 for some fixed digraph @xmath133 and for any @xmath22 , @xmath134 for some constant @xmath135 , then along ( [ 22 ] ) we have @xcite @xmath276 where @xmath136 is the left eigenvector corresponding to eigenvalue zero for the laplacian matrix @xmath143
. this immediately gives us @xmath277 which completes the proof .
in this section , we turn to case ( iii ) and consider that ( [ linearequation ] ) admits a unique least squares solution @xmath278 .
evidently , neither of the two continuous - time distributed flows ( [ 1 ] ) and ( [ 2 ] ) in general can yield exact convergence to the least squares solution of ( [ linearequation ] ) since , even for a fixed interaction graph , @xmath278 is not an equilibrium of the two network flows .
it is indeed possible to find the least squares solution using double - integrator node dynamics @xcite .
however , the use of double integrator dynamics was restricted to networks with fixed and undirected ( or balanced ) communication graphs @xcite . in another direction
, one can also borrow the idea of the use of square - summable step - size sequences with infinite sums in discrete - time algorithms @xcite and build the following flow @xmath279 for @xmath125 .
the least squares case can then be solved under graph conditions of connectedness and balance @xcite , but the convergence rate is at best @xmath280 .
this means ( [ 9 ] ) will be fragile against noises . for the projection+consensus " flow , we can show that under fixed and connected bidirectional interaction graphs , with a sufficiently large @xmath281 , the node states will converge to a ball around the least squares solution whose radius can be made arbitrarily small .
this approximation is global in the sense that the required @xmath281 only depends on the accuracy between the node state limits and the @xmath278 .
[ thm5 ] let ( iii ) hold with @xmath282 and denote the unique least squares solution of ( [ linearequation ] ) as @xmath278 .
suppose @xmath283 for some bidirectional , connected graph @xmath133 and for any @xmath22 , @xmath284 for some constant @xmath135 .
then along the flow ( [ 1 ] ) , for any @xmath261 , there exists @xmath285 such that @xmath286 exists and @xmath287 for any initial value @xmath126 if @xmath288 .
the intuition behind theorem [ thm5 ] can be described briefly as follows ( see its proof presented later for a full exposure of this intuition ) . with fixed and undirected network topology , the consensus + projection " flow ( [ 1 ] )
is a gradient flow in the form of @xmath289 where @xmath290 there holds @xmath291 if each @xmath6 is normalized with a unit length .
therefore , for large @xmath281 , the trajectory of ( [ 1 ] ) will asymptotically tend to be close to the solution of the following optimization problem : @xmath292 any solution of the above problem consists of @xmath107 copies of the solution to @xmath293 which is the unique least squares solution @xmath50 if @xmath40 .
therefore , a large @xmath281 can drive the node states to somewhere near @xmath278 along the flow ( [ 1 ] ) as stated in theorem [ thm5 ] . from the above argument
, for the case when @xmath6 is not normalized , we can replace the consensus + projection " flow ( [ 1 ] ) with @xmath294 for all @xmath125 .
the statement of theorem [ thm5 ] will continue to hold for the flow ( [ r3 ] ) . under the assumptions of theorem [ thm5 ]
, the flow defines a linear time - invariant system @xmath295 where @xmath143 is the laplacian of the graph @xmath133 , @xmath296 is a block - diagonal matrix , and @xmath297 .
therefore , the rate of convergence is given by @xmath298 which is influenced by both @xmath143 ( the graph ) and @xmath299 ( the linear equation ) .
we also know that @xmath300 in general can not grow to infinity by increasing @xmath281 due to the presence of the zero eigenvalue in @xmath143 .
one can however speed up the convergence by adding a weight @xmath301 to the projection term in the flow ( [ 1 ] ) and obtain @xmath302 certainly the convergence rate ( to a ball centered at @xmath278 with radius @xmath303 ) of ( [ r9 ] ) can be arbitrarily large by selecting sufficiently large @xmath281 and @xmath301 .
however , it is clear from the proof of theorem [ thm5 ] that the required @xmath281 for a given accuracy @xmath303 will in turn depend on @xmath301 and require a large @xmath281 for a large @xmath301 .
we also manage to establish the following semi - global result for switching but balanced graphs under two further assumptions .
* [ a1 ] * the set @xmath304 is a bounded set . *
[ a2 ] * @xmath305 .
[ thm6 ] let ( iii ) hold with @xmath282 and denote the unique least squares solution of ( [ linearequation ] ) as @xmath278
. assume [ a1 ] and [ a2 ] .
suppose @xmath80 is balanced for all @xmath306 and @xmath79-@xmath83 with respect to @xmath84 .
then for any @xmath261 and any @xmath307 , there exist two constants @xmath308 and @xmath309 such that when @xmath310 and @xmath311 , there holds @xmath312 along the flow ( [ 1 ] ) with the initial value @xmath126 . the two assumptions , [ a1 ] and [ a2 ] are indeed rather strong assumptions . in general , @xmath313 $ ] holds if @xmath314 is a nonempty bounded set @xcite , which is exactly opposite to the least squares solution case under consideration .
we conjecture that at least for @xmath315 case , @xmath313 $ ] should hold when @xmath316 are distinct vectors .
the assumption @xmath317 $ ] requires a strong symmetry in the affine spaces @xmath91 , which turns out to be essential for the result to stand . for the projection consensus " flow , we present the following result . [ thm7 ] let ( iii ) hold with @xmath282 and denote the unique least squares solution of ( [ linearequation ] ) as @xmath278 .
suppose @xmath318 is fixed , complete , and @xmath319 for all @xmath22 .
then for any initial value @xmath307 , there holds @xmath320 along the flow ( [ 2 ] ) . based on simulation experiments , the requirement that @xmath318 does not have to be complete for the theorem [ thm7 ] to be valid for certain choices of the @xmath91 .
moreover , because the flow ( [ 2 ] ) is a gradient flow over the manifold @xmath321 , the @xmath56 will converge but perhaps to different limits at different nodes .
suppose @xmath318 for some bidirectional , connected graph @xmath133 and for any @xmath22 , @xmath284 for some constant @xmath135 .
fix @xmath261 .
note that , we have @xmath322 therefore , the flow ( [ 1 ] ) is a gradient flow written as @xmath323 where @xmath324 is a @xmath325 convex function .
denote @xmath326 .
there must hold @xmath327 the following lemma holds .
suppose @xmath282 .
then * @xmath328 is a singleton , which implies that @xmath168 converges to a fixed point asymptotically ; * @xmath329 is a bounded set for all @xmath330 .
_ ( i ) recall that @xmath143 is the laplacian of the graph @xmath133 , @xmath296 is a block - diagonal matrix , and @xmath297 .
with lemma [ lem : affineprojection ] , the equation @xmath331 can be written as @xmath332 or , in a compact form , @xmath333 consider @xmath334 we immediately know from the second term of @xmath335 that @xmath336 only if @xmath337 .
on the other hand , obviously @xmath338 for any @xmath339 if @xmath282 . therefore , @xmath340 is positive - definite , which yields @xmath341 \(ii ) by the courant - fischer theorem ( see theorem 4.2.11 in @xcite ) , we have @xmath342.\end{aligned}\ ] ] this immediately implies @xmath343 for all @xmath344 and consequently , @xmath329 is obviously a bounded set .
this proves the desired lemma .
@xmath236 now we introduce @xmath345 .
let @xmath346 with @xmath347 and define @xmath348 and @xmath349 with @xmath278 being the unique least squares solution of ( [ linearequation ] ) .
we see that @xmath350 and @xmath351 are finite numbers due to the boundedness of @xmath352 .
the remainder of the proof contains two steps . *
step 1*. let @xmath353
. then @xmath354 satisfies ( [ 200 ] ) , or in equivalent form , @xmath355 denote in @xmath356 , @xmath357 , and @xmath358 in order to simplify the presentation .
one should however keep in mind that they always depend on @xmath281 . ]
let @xmath360 be the consensus manifold .
denote @xmath361 as the second smallest eigenvalue of @xmath143 .
we can now conclude that @xmath362 where @xmath217 holds from the fact that the zero eigenvalue of @xmath143 corresponds to a unique unit eigenvector whose entries are all the same , @xmath219 is from ( [ 32 ] ) , and @xmath221 holds from the definition of @xmath350 and the fact that @xmath363 .
this allows us to further derive @xmath364 and then @xmath365 therefore , for any @xmath366 we can find @xmath367 that @xmath368 for all @xmath369 . *
step 2*. applying lemma [ lem : affineprojection ] , we have @xmath370 then with ( [ 31 ] ) and noticing the continuity of @xmath371 , for any @xmath366 , there exists @xmath372 such that @xmath373 if @xmath374 consequently , we can conclude without loss of generality then we can replace @xmath375 with @xmath376 in ( [ 201 ] ) with ( [ 51 ] ) continuing to hold ; if @xmath377 then both ( [ 31 ] ) and ( [ 51 ] ) hold directly . ]
that for any @xmath366 we can find @xmath367 so that both ( [ 31 ] ) and ( [ 51 ] ) hold when @xmath369 . now noticing @xmath378 , from ( [ 32 ] ) we have @xmath379 therefore , with ( [ 31 ] ) , we can find another @xmath380 such that @xmath381 for all @xmath382 .
take @xmath383 .
we can finally conclude that @xmath384 for all @xmath125 , where @xmath217 is from ( [ 51 ] ) , @xmath219 is from the convexity of @xmath385 , @xmath221 is based on direct computation of @xmath386 in ( [ 52 ] ) , @xmath232 is due to ( [ 53 ] ) , and @xmath387 holds because @xmath388 from the definition of @xmath351 and @xmath389 .
this completes the proof of theorem [ thm5 ] .
consider the following dynamics for the considered network model : @xmath390 where @xmath391 , @xmath92 is a given constant , @xmath392 are weight functions satisfying our standing assumption , and @xmath393 is a piecewise continuous function . the proof is based on the following lemma on the robustness of consensus subject to noises , which is a special case of theorem 4.1 and proposition 4.10 in @xcite .
[ lemrobust ] let @xmath80 be @xmath79-@xmath83 with respect to @xmath84 . then along ( [ 60 ] ) , there holds that for any @xmath261 , there exist a sufficiently small @xmath394 and sufficiently large @xmath395 such that @xmath396 for all initial value @xmath397 when @xmath398 and @xmath399 , where @xmath400 . with assumption [ a1 ] , the set @xmath401^n\ ] ] is a compact set , and is obviously positively invariant along the flow ( [ 1 ] ) .
therefore , we can define @xmath402 as a finite number . now along the trajectory @xmath168 of ( [ 1 ] )
, we have @xmath403 for all @xmath88 and all @xmath125 . invoking lemma [ lemrobust ] we have for any @xmath261 and any initial value @xmath126
, there exist @xmath404 and @xmath405 such that @xmath406 if @xmath407 and @xmath408 .
furthermore , with lemma [ lem : subspace - affine - projection ] , we have @xmath409 which implies @xmath410 if assumption [ a2 ] holds and @xmath80 is balanced . while if @xmath307 , then @xmath411 this certainly guarantees @xmath412 for all @xmath88 in view of (
the proof for the fact that @xmath413 can then be built using exactly the same analysis as the final part of the proof of theorem [ thm5 ] .
we have now completed the proof of theorem [ thm6 ] .
the proof is based on the following lemma .
there holds @xmath414 for all @xmath415 .
_ from lemma [ lem : subspace - affine - projection ] we can easily know @xmath416 for all @xmath196 . as a result
, we have @xmath417 on the other hand , @xmath418 the desired lemma thus holds .
@xmath236 suppose @xmath318 is fixed , complete , and @xmath319 for all @xmath22 .
now along the flow ( [ 2 ] ) , we have @xmath419 \nonumber\\ & = a^\s \sum_{i=1}^n \bigg[\mathpzc{p}_{\mathcal{a}_i}\big(\frac { \sum_{j=1}^n\mathbf{x}_j}{n}\big ) -\frac{\sum_{j=1}^n \mathbf{x}_j}{n}\bigg]\end{aligned}\ ] ] for any initial value @xmath307 .
the desired result follows straightforwardly and this concludes the proof of theorem [ thm7 ] .
in this section , we provide a few examples illustrating the established results .
* example 1*. consider three nodes in the set @xmath420 whose interactions form a fixed , three - node directed cycle .
let @xmath315 , @xmath421 , and @xmath422 for all @xmath22 .
take @xmath423 , @xmath424 , @xmath425 and @xmath426 , @xmath427 , @xmath428 so the linear equation ( [ linearequation ] ) has a unique solution at @xmath429 corresponding to case ( i ) . with the same initial value @xmath430 , @xmath431 , and @xmath432 , we plot the trajectories of @xmath168 , respectively , for the consensus + projection " flow ( [ 1 ] ) and the projection consensus " flow ( [ 2 ] ) in figure [ fig : fig1 ] .
the plot is consistent with the result of theorems [ thm1 ] and [ thm2 ] .
* example 2*. consider three nodes in the set @xmath420 whose interactions form a fixed , three - node directed cycle .
let @xmath433 , @xmath421 , and @xmath422 for all @xmath22 .
take @xmath434 , @xmath435 , @xmath436 and @xmath426 , @xmath427 , @xmath428 so corresponding to case ( ii ) , the linear equation ( [ linearequation ] ) admits an infinite solution set @xmath437 for the initial value @xmath438 , @xmath439 , and @xmath440 , we plot the trajectories of @xmath441 respectively , for the consensus + projection " flow ( [ 1 ] ) and the projection consensus " flow ( [ 2 ] ) in figure [ fig : fig2 ] .
the plot is consistent with the result of theorems [ thm3 ] , [ thm4 ] , and [ prop1 ] .
* example 3*. consider four nodes in the set @xmath442 whose interactions form a fixed , four - node undirected cycle .
let @xmath315 , and @xmath422 for all @xmath22 .
take @xmath423 , @xmath443 , @xmath444 , @xmath445 and @xmath426 , @xmath428 , @xmath446 , @xmath446 so corresponding to case ( iii ) the linear equation ( [ linearequation ] ) has a unique least squares solution @xmath447 .
for the initial value @xmath448 , @xmath449 , @xmath450 , and @xmath451 we plot the trajectories of @xmath452 along the consensus + projection " flow ( [ 1 ] ) for @xmath453 , respectively in figure [ fig : fig3 ] .
the plot is consistent with the result of theorem [ thm5 ] . for @xmath454 , respectively along the consensus + projection " flow.,width=307 ]
two distributed network flows were studied as distributed solvers for a linear algebraic equation @xmath0 , where a node @xmath1 holds a row @xmath2 of the matrix @xmath3 and the entry @xmath4 in the vector @xmath5 .
a consensus + projection " flow consists of two terms , one from standard consensus dynamics and the other as projection onto each affine subspace specified by the @xmath6 and @xmath4 .
projection consensus " flow simply replaces the relative state feedback in consensus dynamics with projected relative state feedback . under mild graph conditions , it was shown that that all node states converge to a common solution of the linear algebraic equation , if there is any .
the convergence is global for the consensus + projection " flow while local for the projection consensus " flow .
when the linear equation has no exact solutions but has a well defined least squares approximate solution , it was proved that the node states can converge to somewhere arbitrarily near the least squares solution as long as the gain of the consensus dynamics is sufficient large for consensus + projection " flow under fixed and undirected graphs .
it was also shown that the projection consensus " flow drives the average of the node states to the least squares solution if the communication graph is complete .
numerical examples were provided verifying the established results .
interesting future direction includes more precise comparisons of the existing continuous - time or discrete - time linear equation solvers in terms of convergence speed , computational complexity , and communication complexity .
suppose @xmath266 .
we show @xmath267 for all @xmath455 by a contradiction argument .
suppose ( to obtain a contradiction ) that there exists a node @xmath456 with @xmath457 .
therefore , we can find a time instant @xmath458 with @xmath459 in other words , there is an absolute positive distance between @xmath460 and the limit @xmath264 of @xmath461 .
let @xmath80 be @xmath79-@xmath83 .
consider the @xmath462 time intervals @xmath463,\cdots,[t_\ast+(n-2)t , t_\ast + ( n-1)t]$ ] . in view of the arguments in ( [ 8 ] )
, we similarly obtain @xmath464 which leads to @xmath465 for all @xmath466 noticing ( [ 10 ] ) . denoting @xmath467 and applying grnwall s inequality
we can further conclude that @xmath468 for all @xmath469 $ ] , where @xmath470 .
now that @xmath80 is @xmath79-@xmath83 , there must be a node @xmath471 for which @xmath472 is a @xmath79-arc over the time interval @xmath473 .
thus , there holds @xmath474 for @xmath475 $ ] . noticing the definition of @xmath79-arcs and that @xmath476 , we invoke the grnwall s inequality again and conclude from ( [ 11 ] ) that @xmath477\nonumber\\ & + \frac{\mu ( h_\s+\epsilon)^2}{2 } \bigg[1-e^{-\int_{t_\ast}^{t_\ast+t } \sum_{j\in \mathrm{n}_{i_1}(s)}{a_{i_1j}(s)}ds } \cdot\int_{t_\ast}^{t_\ast+t}e^{\int_{t_\ast}^{t } f_1(s)ds } f_1(t ) dt\bigg]\nonumber\\ & \leq \frac{\mu \gamma}{2 } \cdot l_\s^2+\big(1-\frac{\mu\gamma}{2}\big ) \cdot(h_\s + \epsilon)^2\end{aligned}\ ] ]
where @xmath478 and @xmath479 .
this further allows us to apply the estimation of @xmath480 over the interval @xmath481 $ ] to node @xmath482 for the interval @xmath483 $ ] and obtain @xmath484 for all @xmath485 $ ] .
since @xmath80 is @xmath79-@xmath83 , the above analysis can be recursively applied to the intervals @xmath486 , for which nodes @xmath487 can be found , respectively , with @xmath488 such that @xmath489 for @xmath490 this implies @xmath491 when @xmath303 is sufficiently small .
however , we have known that non - increasingly there holds @xmath260 , and therefore such @xmath492 does not exist , i.e. , @xmath267 for all @xmath455 .
the statement of lemma [ lem : directed ] .
( i ) is proved .
suppose @xmath267 for all @xmath455 .
then for any @xmath261 , there exists @xmath493 such that @xmath494 moreover , for the ease of presentation we assume that @xmath316 are distinct vectors since otherwise we can always combine the nodes with the same @xmath6 as a cluster to be treated together in the following arguments .
denote the angle between the two unit vectors @xmath495 as @xmath496 .
then for all @xmath88 . ]
@xmath497 this leads to ( cf . , ( [ 8 ] ) ) @xmath498 with @xmath499 for all @xmath88 .
this will in turn give us @xmath500 for all @xmath501 .
now we have @xmath502 if @xmath80 is @xmath79-@xmath83 . combining ( [ 16 ] ) and ( [ 17 ] )
we arrive at @xmath503 which leaves @xmath265 the only possibility since @xmath303 can be arbitrary number .
this proves lemma [ lem : directed ] .
( ii ) .
again without loss of generality we can assume that @xmath316 are distinct vectors . with ( [ 19 ] )
, we have @xmath504 where @xmath505 .
it is easy to see from ( [ 19 ] ) using a contradiction argument that if @xmath80 is @xmath79-@xmath87 , there must hold @xmath506 therefore , we conclude that @xmath265 immediately and this proves the desired lemma .
the authors thank zhiyong sun for his generous help in the preparation of the numerical examples .
99 g. shi and b. d. o. anderson , distributed network flows solving linear algebraic equations , " _ american control conference _ , boston , ma , pp . 2864 - 2869 , 2016 . c. godsil and g. royle , new york : springer - verlag , 2001 .
s. kar , j. m. f. moura and k. ramanan , distributed parameter estimation in sensor networks : nonlinear observation models and imperfect communication , " _ ieee trans .
information theory _ , 58(6 ) , pp .
3575 - 52 , 2012 .
f. deutsch , rate of convergence of the method of alternating projections , " in _ parametric optimization and approximation _ , b. brosowski and f. deutsch , eds.,ol .
96 - 107 , basel , switzerland : birkhuser , 1983 .
j. lu and c. y. tang , distributed asynchronous algorithms for solving positive definite linear equations over networks - part ii : wireless networks . in _
proc first ifac workshop on estimation and control of networked systems _ , pages 258 - 264 , 2009 .
j. lu and c. y. tang , distributed asynchronous algorithms for solving positive definite linear equations over networks - part i : agent networks , " in _ proc first ifac workshop on estimation and control of networked systems _ , pages 22 - 26 , 2009 .
j. liu , s. mou , and a. s. morse , an asynchronous distributed algorithm for solving a linear algebraic equation , " in _ proceedings of the 2013 ieee conference on decision and control _ , pp .
5409 - 5414 , 2013 .
b. d. o. anderson , s. mou , a. s. morse , and u. helmke , decentralized gradient algorithm for solution of a linear equation , " _ numerical algebra , control and optimization _ , to appear ( preprint arxiv:1509.04538 ) .
d. jakoveti , j. xavier and j. m. f. moura , cooperative convex optimization in networked systems : augmented lagrangian algorithms with directed gossip communication , " _ ieee trans .
signal processing _ , vol .
59 , no . 8 , pp . 3889 - 3902 , 2011 .
|
we study distributed network flows as solvers in continuous time for the linear algebraic equation @xmath0 .
each node @xmath1 has access to a row @xmath2 of the matrix @xmath3 and the corresponding entry @xmath4 in the vector @xmath5 .
the first consensus + projection " flow under investigation consists of two terms , one from standard consensus dynamics and the other contributing to projection onto each affine subspace specified by the @xmath6 and @xmath4 .
the second projection consensus " flow on the other hand simply replaces the relative state feedback in consensus dynamics with projected relative state feedback . without dwell - time assumption on switching graphs , we prove that all node states converge to a common solution of the linear algebraic equation , if there is any .
the convergence is global for the consensus + projection " flow while local for the projection consensus " flow in the sense that the initial values must lie on the affine subspaces .
if the linear equation has no exact solutions , we show that the node states can converge to a ball around the least squares solution whose radius can be made arbitrarily small through selecting a sufficiently large gain for the consensus + projection " flow for a fixed bidirectional graph .
semi - global convergence to approximate least squares solutions is also demonstrated for switching balanced directed graphs under suitable conditions .
it is also shown that the projection consensus " flow drives the average of the node states to the least squares solution with a complete graph .
numerical examples are provided as illustrations of the established results .
|
we study markov processes on a complex of half - lines @xmath12 , where @xmath13 is finite , and all half - lines are connected at a common origin . on a given half - line
, a particle performs a random walk with a heavy - tailed increment distribution , until it would exit the half - line , when it switches ( in general , at random ) to another half - line to complete its jump . to motivate the development of the general model ,
we first discuss informally some examples ; we give formal statements later .
the _ one - sided oscillating random walk
_ takes place on two half - lines , which we may map onto @xmath8 . from the positive half - line ,
the increments are negative with density proportional to @xmath14 , and from the negative half - line the increments are positive with density proportional to @xmath15 , where @xmath16 .
the walk is transient if and only if @xmath17 ; this is essentially a result of kemperman @xcite .
the oscillating random walk has several variations and has been well studied over the years ( see e.g. @xcite ) .
this previous work , as we describe in more detail below , is restricted to the case of _ two _ half - lines .
we generalize this model to an arbitrary number of half - lines , labelled by a finite set @xmath13 , by assigning a rule for travelling from half - line to half - line .
first we describe a deterministic rule .
let @xmath18 be a positive integer .
define a _ routing schedule of length _
@xmath18 to be a sequence @xmath19 of @xmath18 elements of @xmath13 , dictating the sequence in which half - lines are visited , as follows .
the walk starts from line @xmath20 , and , on departure from line @xmath20 jumps over the origin to @xmath21 , and so on , until departing @xmath22 it returns to @xmath20 ; on line @xmath23 , the walk jumps towards the origin with density proportional to @xmath24 where @xmath25 .
one simple example takes a cyclic schedule in which @xmath26 is a permutation of the elements of @xmath13 . in any case , a consequence of our results is that now the walk is transient if and only if @xmath27 where @xmath0 is the number of times @xmath3 appears in the sequence @xmath26 . in particular ,
in the cyclic case the transience criterion is @xmath28 . it is easy to see that , if @xmath13 contains two elements , the _ cotangent criterion _ is equivalent to the previous one for the one - sided oscillating walk ( @xmath29 ) . for more than two half - lines
the criterion is non - linear , and it was necessary to extend the model to more than two lines in order to see the essence of the behaviour .
more generally , we may choose a random routing rule between lines : on departure from half - line @xmath30 , the walk jumps to half - line @xmath31 with probability @xmath32 .
the deterministic cyclic routing schedule is a special case in which @xmath33 for @xmath34 the successor to @xmath35 in the cycle . in fact
, this set - up generalizes the arbitrary deterministic routing schedule described above , as follows .
given the schedule sequence @xmath26 of length @xmath18 , we may convert this to a cyclic schedule on an extended state - space consisting of @xmath0 copies of line @xmath3 , and then reading @xmath26 as a permutation .
so the deterministic routing model is a special case of the model with markov routing , which will be the focus of the rest of the paper .
our result again will say that is the criterion for transience , where @xmath0 is now the stationary distribution associated to the stochastic matrix @xmath32 .
our general model also permits _ two - sided _
increments for the walk from some of the lines , which contribute terms involving @xmath36 to the cotangent criterion .
these two - sided models also generalize previously studied classical models ( see e.g. @xcite ) .
again , it is only in our general setting that the essential nature of the cotangent criterion becomes apparent . rather than @xmath12
, one could work on @xmath37 instead , with mass functions replacing probability densities ; the results would be unchanged .
the paper is organized as follows . in section [ sec : results ] we formally define our model and describe our main results , which as well as a recurrence classification include results on existence of moments of return times in the recurrent cases . in section [ sec : oscillating ] we explain how our general model relates to the special case of the oscillating random walk when @xmath13 has two elements , and state our results for that model ; in this context the recurrence classification results are already known , but the existence of moments results are new even here , and are in several important cases sharp .
the present work was also motivated by some problems concerning many - dimensional , partially homogeneous random walks similar to models studied in @xcite : we describe this connection in section [ sec : many - dim ] .
the main proofs are presented in sections [ sec : proofs ] , [ sec : moments ] , and [ sec : critical ] , the latter dealing with the critical boundary case which is more delicate and requires additional work .
we collect various technical results in appendix [ sec : appendix ] .
consider @xmath38 , a discrete - time , time - homogeneous markov process with state space @xmath12 , where @xmath13 is a finite non - empty set .
the state space is equipped with the appropriate borel sets , namely , sets of the form @xmath39 where @xmath40 is a borel set in @xmath41 , and @xmath42 . the process will be described by : * an irreducible stochastic matrix labelled by @xmath13 , @xmath43 ; and * a collection @xmath44 of probability density functions , so @xmath45 is a borel function with @xmath46
. we view @xmath12 as a complex of half - lines @xmath47 , or _ branches _ , connected at a central origin @xmath48 ; at time @xmath49 ,
the coordinate @xmath50 describes which branch the process is on , and @xmath51 describes the distance along that branch at which the process sits .
we will call @xmath51 a _ random walk _ on this complex of branches . to simplify notation , throughout we write
@xmath52 $ ] for @xmath53 $ ] , the conditional probability starting from @xmath54 ; similarly we use @xmath55 for the corresponding expectation .
the transition kernel of the process is given for @xmath54 , for all borel sets @xmath56 and all @xmath31 , by @xmath57 & = { { \mathbb p}}_{x , i } \left [ ( x_{1 } , \xi_{1 } ) \in b \times \ { j \ } \right ] \nonumber\\ & { } \hskip -4.7 cm { } = p ( i , j ) \int_b w_i ( -z - x ) { { \mathrm d}}z + { { \mathbf 1}{\ { i = j \ } } } \int_b w_i ( z - x ) { { \mathrm d}}z .
\end{aligned}\ ] ] the dynamics of the process represented by can be described algorithmically as follows .
given @xmath58 , generate ( independently ) a spatial increment @xmath59 from the distribution given by @xmath60 and a random index @xmath61 according to the distribution @xmath62 .
then , * if @xmath63 , set @xmath64 ; or * if @xmath65 , set @xmath66 . in words , the walk takes a @xmath67-distributed step . if this step would bring the walk beyond the origin , it passes through the origin and switches onto branch @xmath68 ( or , if @xmath68 happens to be equal to @xmath50 , it reflects back along the same branch ) .
the finite irreducible stochastic matrix @xmath69 is associated with a ( unique ) positive invariant probability distribution @xmath70 satisfying @xmath71 for future reference , we state the following . : : let @xmath43 be an irreducible stochastic matrix , and let @xmath70 denote the corresponding invariant distribution .
our interest here is when the @xmath60 are _ heavy tailed_. we allow two classes of distribution for the @xmath60 : _ one - sided _ or _ symmetric_. it is convenient , then , to partition @xmath13 as @xmath72 where @xmath73 and @xmath74 are disjoint sets , representing those branches on which the walk takes , respectively , one - sided and symmetric jumps .
the @xmath75 are then described by a collection of positive parameters @xmath76 . for a probability density function @xmath77 , an exponent @xmath78 , and a constant @xmath79 , we write @xmath80 to mean that there exists @xmath81 with @xmath82 and @xmath83 for which @xmath84
if @xmath80 is such that holds and @xmath85 satisfies the stronger condition @xmath86 for some @xmath87 , then we write @xmath88 .
our assumption on the increment distributions @xmath60 is as follows .
: : suppose that , for each @xmath23 , we have an exponent @xmath89 , a constant @xmath90 , and a density function @xmath91 .
then suppose that , for all @xmath92 , @xmath75 is given by @xmath93 we say that @xmath51 is _ recurrent _ if @xmath94 , a.s .
, and _ transient _ if @xmath95 , a.s .
an irreducibility argument shows that our markov chain @xmath96 displays the usual recurrence / transience dichotomy and exactly one of these two situations holds ; however , our proofs establish this behaviour directly using semimartingale arguments , and so we may avoid discussion of irreducibility here . throughout we
define , for @xmath23 , @xmath97 our first main result gives a recurrence classification for the process .
[ thm : recurrence ] suppose that and hold .
* suppose that @xmath98 .
then @xmath51 is recurrent . *
suppose that @xmath99 .
* * if @xmath100 , then @xmath51 is recurrent . * * if @xmath101 , then @xmath51 is transient . * * suppose in addition that the densities @xmath102 of belong to @xmath103 for each @xmath3 .
then @xmath104 implies that @xmath51 is recurrent . in the recurrent cases , it is of interest to quantify recurrence via existence or non - existence of passage - time moments . for @xmath105 ,
let @xmath106 , where throughout the paper we adopt the usual convention that @xmath107 .
the next result shows that in all the recurrent cases , excluding the boundary case in theorem [ thm : recurrence](b)(iii ) , the tails of @xmath108 are _
polynomial_. [ thm : moments ] suppose that and hold . in cases ( a ) and ( b)(i ) of theorem [ thm :
recurrence ] , there exists @xmath109 such that for all @xmath110 and all @xmath111 , @xmath112 < \infty , \
\textrm { for } q < q_\star \ \text { and } \operatorname{\mathbb{e}}_{x , k } [ \tau_a^{q } ] = \infty , \ \text { for } q > q_\star.\ ] ] one would like to precisely locate @xmath113 ; here we only locate @xmath113 within an interval , i.e. , we show that there exist @xmath114 and @xmath115 with @xmath116 such that @xmath117 < \infty$ ] for @xmath118 and @xmath117 = \infty$ ] for @xmath119 .
the existence of a critical @xmath113 follows from the monotonicity of the map @xmath120 $ ] , but obtaining a sharp estimate of its value remains an open problem in the general case .
we do have sharp results in several particular cases for _ two _ half - lines , in which case our model reduces to the _ oscillating random walk _ considered by kemperman @xcite and others .
we present these sharp moments results ( theorems [ thm : symmetric_random_walk_moments ] and [ thm : one - sided_moments ] ) in the next section , which discusses in detail the case of the oscillating random walk , and also describes how our recurrence results relate to the known results for this classical model .
in the case of our general model in which @xmath13 consists of two elements , @xmath121 , say , it is natural and convenient to represent our random walk on the whole real line @xmath8 .
namely , if @xmath122 for @xmath123 and @xmath124 , we let @xmath125 .
the simplest case has no reflection at the origin , only transmission , i.e. @xmath126 , so that @xmath127
. then , for @xmath128 a borel set , @xmath129 = \\ & \qquad \qquad { } { { \mathbb p}}_{x , i } [ ( x_{1 } , \xi_{1 } ) \in b^+\ ! \times \!\ { + 1 \ } ] + { { \mathbb p}}_{x , i } [ ( x_{1 } , \xi_{1 } ) \in b^-\ ! \times \!\ { -1 \ } ] , \end{aligned}\ ] ] where @xmath130 and @xmath131 .
in particular , by , writing @xmath132 for @xmath133 , for @xmath123 , @xmath134 & = \int_{b^+ } w_{+ } ( z - x ) { { \mathrm d}}z + \int_{b^- } w_{+ } ( -z - x ) { { \mathrm d}}z \\ & = \int_b w_{+ } ( z - x ) { { \mathrm d}}z , \end{aligned}\ ] ] and , similarly , writing @xmath135 for @xmath136 , for @xmath123 , @xmath137 = \int_b w_{- } ( z+x ) { { \mathrm d}}z .
\end{aligned}\ ] ] for @xmath138 , @xmath139 is invertible with @xmath140 and hence we have for @xmath141 and borel @xmath128 , @xmath142 = \begin{cases } \int_b w_{+ } ( z - x ) { { \mathrm d}}z & \text{if } x > 0 \\ \int_b w_{- } ( z - x ) { { \mathrm d}}z & \text{if } x < 0 .\end{cases } \end{aligned}\ ] ] we may make an arbitrary non - trivial choice for the transition law at @xmath143 without affecting the behaviour of the process , and then shows that @xmath144 is a time - homogeneous markov process on @xmath8 . now @xmath144 is _ recurrent _ if @xmath145 , a.s . , or _
transient _ if @xmath146 , a.s .
the one - dimensional case described at has received significant attention over the years .
we describe several of the classical models that have been considered .
the most classical case is the following .
: : let @xmath78 .
for @xmath80 , suppose that @xmath147 for @xmath148 . in this case , @xmath144 describes a random walk with i.i.d .
symmetric increments .
[ thm : symmetric_random_walk ] suppose that holds
. then the symmetric random walk is transient if @xmath149 and recurrent if @xmath150 . if , in addition , @xmath151 , then the case @xmath152 is recurrent .
theorem [ thm : symmetric_random_walk ] follows from our theorem [ thm : recurrence ] , since in this case @xmath153 since it deals with a sum of i.i.d .
random variables , theorem [ thm : symmetric_random_walk ] may be deduced from the classical theorem of chung and fuchs @xcite , via e.g. the formulation of shepp @xcite .
the method of the present paper provides an alternative to the classical ( fourier analytic ) approach that generalizes beyond the i.i.d . setting .
( note that theorem [ thm : symmetric_random_walk ] is not , formally , a consequence of shepp s most accessible result , theorem 5 of @xcite , since @xmath154 does not necessarily correspond to a unimodal distribution in shepp s sense . ) with @xmath108 as defined previously , in the setting of the present section we have @xmath155 .
use @xmath156 $ ] as shorthand for @xmath157 $ ] .
we have the following result on existence of passage - time moments , whose proof is in section [ sec : moments ] ; while part ( i ) is well known , we could find no reference for part ( ii ) .
[ thm : symmetric_random_walk_moments ] suppose that holds , and that @xmath150 .
let @xmath158 $ ] .
* if @xmath159 , then @xmath160 < \infty$ ] if @xmath161 and @xmath160 = \infty$ ] if @xmath162 . *
if @xmath11 , then @xmath163 < \infty$ ] if @xmath164 and @xmath160 = \infty$ ] if @xmath165 . our main interest concerns spatially inhomogeneous models , i.e. , in which @xmath166 depends on @xmath167 , typically only through @xmath168 , the sign of @xmath167 .
such models are known as _ oscillating random walks _ , and were studied by kemperman @xcite , to whom the model was suggested in 1960 by anatole joffe and peter ney ( see @xcite ) .
the next example , following @xcite , is a _ one - sided oscillating random walk _ : : : let @xmath169 . for
@xmath170 and @xmath171 , suppose that @xmath172 in other words , the walk always jumps in the direction of ( and possibly over ) the origin , with tail exponent @xmath173 from the positive half - line and exponent @xmath174 from the negative half - line .
the following recurrence classification applies .
[ thm : one - sided ] suppose that holds
. then the one - sided oscillating random walk is transient if @xmath17 and recurrent if @xmath175 . if , in addition , @xmath176 and @xmath177 , then the case @xmath178 is recurrent .
theorem [ thm : one - sided ] was obtained in the discrete - space case by kemperman @xcite ; it follows from our theorem [ thm : recurrence ] , since in this case @xmath179 the special case of in which @xmath180 was called _
antisymmetric _ by kemperman ; here theorem [ thm : one - sided ] shows that the walk is transient for @xmath181 and recurrent for @xmath182 .
we have the following moments result , proved in section [ sec : moments ] .
[ thm : one - sided_moments ] suppose that holds , and that @xmath183 .
let @xmath158 $ ] .
* if @xmath184 , then @xmath163 < \infty$ ] if @xmath185 and @xmath163 = \infty$ ] if @xmath186 . *
if @xmath187 , then @xmath163 < \infty$ ] if @xmath188 and @xmath163 = \infty$ ] if @xmath189 .
obtain sharp moments results for general @xmath16 .
another model in the vein of @xcite is a _ two - sided oscillating random walk _ : : : let @xmath169 . for
@xmath170 and @xmath171 , suppose that @xmath190 now the jumps of the walk are symmetric , as under , but with a tail exponent depending upon which side of the origin the walk is currently on , as under .
the most general recurrence classification result for the model is due to sandri @xcite .
a somewhat less general , discrete - space version was obtained by rogozin and foss ( theorem 2 of @xcite ) , building on @xcite .
analogous results in continuous time were given in @xcite .
here is the result .
[ thm : two - sided ] suppose that holds .
then the two - sided oscillating random walk is transient if @xmath191 and recurrent if @xmath192 . if , in addition , @xmath176 and @xmath177 , then the case @xmath193 is recurrent .
theorem [ thm : two - sided ] also follows from our theorem [ thm : recurrence ] , since in this case @xmath194 obtain sharp moments results for @xmath195 .
a final model is another oscillating walk that mixes the one- and two - sided models : : : let @xmath169 . for @xmath170 and @xmath171 , suppose that @xmath196 in the discrete - space case , theorem 2 of rogozin and foss @xcite gives the recurrence classification .
[ thm : mixture ] suppose that holds .
then the mixed oscillating random walk is transient if @xmath197 and recurrent if @xmath198 . if , in addition , @xmath176 and @xmath177 , then the case @xmath199 is recurrent .
theorem [ thm : mixture ] also follows from our theorem [ thm : recurrence ] , since in this case @xmath200 it is possible to generalize the model further by permitting the local transition density to vary _ within _ each half - line .
then we have the transition kernel @xmath201 = \int_b w_x ( z -x ) { { \mathrm d}}z,\ ] ] for all borel sets @xmath128 . here
the local transition densities @xmath202 are borel functions
. variations of the oscillating random walk , within the general setting of , have also been studied in the literature .
sandri @xcite supposes that the @xmath166 satisfy , for each @xmath203 , @xmath204 as @xmath205 for some measurable functions @xmath206 and @xmath173 ; he refers to this as a _ stable - like markov chain_. under a uniformity condition on the @xmath166 , and other mild technical conditions , sandri @xcite obtained , via foster
lyapunov methods similar in spirit to those of the present paper , sufficient conditions for recurrence and transience : essentially @xmath207 is sufficient for recurrence and @xmath208 is sufficient for transience . these results can be seen as a generalization of theorem [ thm : symmetric_random_walk ] . some related results for models in continuous - time ( lvy processes ) are given in @xcite .
further results and an overview of the literature are provided in sandri s phd thesis @xcite .
the next two examples show how versions of the oscillating random walk of section [ sec : oscillating ] arise as embedded markov chains in certain two - dimensional random walks .
[ ex : cp1 ] consider @xmath209 , @xmath210 , a nearest - neighbour random walk on @xmath211 with transition probabilities @xmath212 = p ( x_1 , x_2 ; y_1 , y_2 ) .\ ] ] suppose that the probabilities are given for @xmath213 by , @xmath214 ( the rest being zero ) and for @xmath215 by @xmath216 for all @xmath217 , @xmath218 for all @xmath219 , and @xmath220 .
see figure [ fig : camp_pet2 ] for an illustration .
iin 0, ... ,12 ( 0.4*i,0 ) ( 0.4*i,0.4 * 12 ) ; iin 0, ... ,12 ( 0,0.4*i ) ( 0.4 * 12,0.4*i ) ; ( 0 * 6,0.4 * 6 ) ( 0.4 * 12,0.4 * 6 ) ; ( 0.4 * 6,0 ) ( 0.4 * 6,0.4 * 12 ) ; ( 0.4 * 9,0.4 * 9 )
( 0.4 * 9,0.4 * 8 ) ; ( 0.4 * 9,0.4 * 9 )
( 0.4 * 9,0.4 * 10 ) ; ( 0.4 * 9,0.4 * 9 ) ( 0.4 * 8,0.4 * 9 ) ; ( 0.4 * 3,0.4 * 9 ) ( 0.4 * 3,0.4 * 8 ) ; ( 0.4 * 3,0.4 * 9 )
( 0.4 * 3,0.4 * 10 ) ; ( 0.4 * 3,0.4 * 9 ) ( 0.4 * 2,0.4 * 9 ) ; iin 1, ... ,5 ( 0.4*i,0.4 * 6 ) ( 0.4*i,0.4 * 5 ) ; iin 7, ... ,11 ( 0.4*i,0.4 * 6 ) ( 0.4*i,0.4 * 7 ) ; ( 0.4 * 6,0.4 * 5 ) ( 0.4 * 6,0.4 * 7 ) ; ( 0.4 * 9,0.4 * 3 )
( 0.4 * 9,0.4 * 2 ) ; ( 0.4 * 9,0.4 * 3 ) ( 0.4 * 9,0.4 * 4 ) ; ( 0.4 * 9,0.4 * 3 ) ( 0.4 * 10,0.4 * 3 ) ; ( 0.4 * 3,0.4 * 3 ) ( 0.4 * 3,0.4 * 2 ) ; ( 0.4 * 3,0.4 * 3 )
( 0.4 * 3,0.4 * 4 ) ; ( 0.4 * 3,0.4 * 3 ) ( 0.4 * 4,0.4 * 3 ) ; pictorial representation of the non - homogeneous nearest - neighbour random walk on @xmath211 of example [ ex : cp2 ] , plus a simulated trajectory of 5000 steps of the walk .
we conjecture that the walk is recurrent .
, scaledwidth=40.0% ] set @xmath221 and define recursively @xmath222 for @xmath223 ; consider the embedded markov chain @xmath224 .
we show that @xmath51 is a discrete version of the oscillating random walk described in section [ sec : oscillating ] .
indeed , @xmath225 is a reflecting random walk on @xmath226 with increments taking values @xmath227 each with probability @xmath228 .
we then ( see e.g. @xcite ) have that for some constant @xmath79 , @xmath229 = ( c + o(1 ) ) r^{-1/2 } , ~\text{as}~ r \to \infty.\ ] ] suppose that @xmath230 . since between times @xmath231 and @xmath232 , @xmath233 is monotone , we have @xmath234 \geq { { \mathbb p } } [ \tau_1 > 3r + r^{3/4 } ] - { { \mathbb p } } [ \xi^{(1)}_{3r + r^{3/4 } } - \xi^{(1)}_0 \geq -r ] .\ ] ] here , by the azuma hoeffding inequality , for some @xmath235 and all @xmath236 , @xmath237 \leq \exp \ { - { \varepsilon}r^{1/2 } \ } .\ ] ] similarly , @xmath234
\leq { { \mathbb p } } [ \tau_1 > 3r - r^{3/4 } ] + { { \mathbb p } } [ \xi^{(1)}_{3r- r^{3/4 } } - \xi^{(1)}_0 \leq -r ] , \ ] ] where @xmath238 \leq \exp \ { - { \varepsilon}r^{1/2 } \}$ ] . combining these bounds , and using the symmetric argument for @xmath239 when @xmath240 , we see that for @xmath241 , @xmath242 & = u(r ) , \text { if } x > 0 , \text { and } \nonumber\\ { { \mathbb p } } [ x_{n+1 } - x_n >
r \mid x_n = x ] & = u(r ) , \text { if } x < 0 , \end{aligned}\ ] ] where @xmath243 . thus @xmath51 satisfies a discrete - space analogue of with @xmath244 .
this is the critical case identified in theorem [ thm : one - sided ] , but that result does not cover this case due to the rate of convergence estimate for @xmath245 ; a finer analysis is required .
we conjecture that the walk is recurrent .
[ ex : cp2 ] we present two variations on the previous example , which are superficially similar but turn out to be less delicate .
first , modify the random walk of the previous example by supposing that holds but replacing the behaviour at @xmath215 by @xmath246 for all @xmath247 .
see the left - hand part of figure [ fig : camp_pet1 ] for an illustration .
the embedded process @xmath51 now has , for all @xmath248 and for @xmath249 , @xmath250 = { { \mathbb p } } [ x_{n+1 } - x_n > r \mid x_n = x ] = u(r ) , \ ] ] where @xmath251 .
thus @xmath51 is a random walk with symmetric increments , and the discrete version of our theorem [ thm : symmetric_random_walk ] ( and also a result of @xcite ) implies that the walk is transient .
this walk was studied by campanino and petritis @xcite , who proved transience via different methods .
next , modify the random walk of example [ ex : cp1 ] by supposing that holds but replacing the behaviour at @xmath215 by @xmath246 if @xmath252 , and @xmath253 for @xmath219 .
see the right - hand part of figure [ fig : camp_pet1 ] for an illustration .
this time the walk takes a symmetric increment as at when @xmath254 but a one - sided increment as at when @xmath255 . in this case
the discrete version of our theorem [ thm : mixture ] ( and also a result of @xcite ) shows that the walk is transient .
one may obtain the general model on @xmath37 as an embedded process for a random walk on complexes of half - spaces , generalizing the examples considered here ; we leave this to the interested reader .
our proofs are based on demonstrating appropriate lyapunov functions ; that is , for suitable @xmath256 we study @xmath257 such that @xmath258 has appropriate local supermartingale or submartingale properties for the one - step mean increments @xmath259 \\ & = \operatorname{\mathbb{e}}_{x , i } \left [ \varphi(x_{1 } , \xi_{1 } ) - \varphi ( x_0 , \xi_0 ) \right ] .
\end{aligned}\ ] ] first we note some consequences of the transition law .
let @xmath256 be measurable .
then , we have from that , for @xmath260 , @xmath261 our primary lyapunov function is roughly of the form @xmath262 , @xmath263 , but weighted according to an @xmath13-dependent component ( realised by a collection of multiplicative weights @xmath264 ) ; these weights provide a crucial technical tool . for @xmath263 and @xmath203
we write @xmath265 then , for parameters @xmath266 for each @xmath23 , define for @xmath123 and @xmath23 , @xmath267 now for this lyapunov function , gives @xmath268 depending on whether @xmath269 or @xmath270 , the above integrals can be expressed in terms of @xmath271 as follows .
for @xmath272 , @xmath273 for @xmath274 , @xmath275 in the course of our proofs , we need various integral estimates that can be expressed in terms of classical transcendental functions . for the convenience of the reader , we gather all necessary integrals in lemmas [ lem : i - integrals ] and [ lem : j - integrals ] ; the proofs of these results are deferred until section [ sec : integrals ] . recall that the euler gamma function @xmath276 satisfies the functional equation @xmath277 , and the hypergeometric function @xmath278 is defined via a power series ( see @xcite ) .
[ lem : i - integrals ] suppose that @xmath279 and @xmath280 .
then @xmath281 suppose that @xmath282 and @xmath283 .
then @xmath284 suppose that @xmath285 and @xmath283 .
then @xmath286 recall that the digamma function is @xmath287 , which has @xmath288 where @xmath289 is euler s constant .
[ lem : j - integrals ] suppose that @xmath290 .
then @xmath291 suppose that @xmath282 .
then @xmath292 suppose that @xmath285 .
then @xmath293 the @xmath294 integrals can be obtained as derivatives with respect to @xmath295 of the @xmath35 integrals , evaluated at @xmath296 .
the next result collects estimates for our integrals in the expected functional increments and in terms of the integrals in lemma [ lem : i - integrals ] .
[ lem : all - integrals ] suppose that @xmath80 .
for @xmath290 and @xmath297 we have @xmath298 for @xmath282 and @xmath283 we have @xmath299 for @xmath285 and @xmath283 we have @xmath300 moreover , if @xmath88 then stronger versions of all of the above estimates hold with @xmath301 replaced by @xmath302 for some @xmath303 .
these estimates are mostly quite straightforward , so we do not give all the details .
we spell out the estimate in ; the others are similar .
we have @xmath304 with the substitution @xmath305 , this last expression becomes @xmath306 let @xmath307 .
then there exists @xmath308 such that @xmath309 for all @xmath310 , so that @xmath311 for all @xmath245 in the range of integration , provided @xmath312 .
writing @xmath313 for the duration of the proof , we have that @xmath314 for @xmath315 and @xmath312 , we have @xmath316 so that @xmath317 for a constant @xmath318 not depending on @xmath312 or @xmath319
. moreover , it is easy to see that @xmath320 for a constant @xmath321 depending only on @xmath295 and @xmath173 , provided @xmath322 .
hence lemma [ lem : integral_approx ] shows that @xmath323 for all @xmath312 . since also @xmath324 as @xmath325
, it follows that for any @xmath235 we may choose @xmath167 sufficiently large so that @xmath326 and since @xmath235 was arbitrary , we obtain from .
we also need the following simple estimates for ranges of @xmath173 when the asymptotics for the final two integrals in lemma [ lem : all - integrals ] are not valid .
[ lem : big - alpha ] suppose that @xmath80 . * for @xmath159 and any @xmath327 , there exist @xmath235 and @xmath328 such that , for all @xmath329 , @xmath330 * for @xmath184 and any @xmath331 , there exist @xmath235 and @xmath328 such that , for all @xmath329 , @xmath332 for part ( i ) , set @xmath333 , so that @xmath334 suppose that @xmath159 and @xmath327 . for @xmath327 , calculus shows that @xmath335 has a single local maximum at @xmath336 , so that @xmath337 for all @xmath338 .
moreover , taylor s theorem shows that for any @xmath327 there exists @xmath339 such that @xmath340 for all @xmath341 $ ] .
also , @xmath342 for all @xmath310 sufficiently large . hence , for all @xmath343 , @xmath344 which yields part ( i ) of the lemma . for part ( ii ) , suppose that @xmath184 and @xmath331 . for any @xmath331 ,
there exists @xmath339 such that @xmath345 for all @xmath346 $ ] .
moreover , @xmath347 for all @xmath310 sufficiently large . hence ,
since the integrand is non - positive , for @xmath348 , @xmath349 and part ( ii ) follows . [
lem : df ] suppose that holds and @xmath350 .
then for @xmath351 , @xmath352 , and @xmath353 , as @xmath354 , @xmath355 where @xmath356 , @xmath357 moreover , if @xmath358 then , for some @xmath303 , @xmath359 finally , as @xmath360 , @xmath361 the above expressions for @xmath362 follow from and with lemma [ lem : all - integrals ] .
additionally , we compute that @xmath363 . for @xmath295 in a neighbourhood of @xmath364 uniformity of convergence of the integrals over @xmath365 enables us to differentiate with respect to @xmath295 under the integral sign to get @xmath366 using lemma [ lem : j - integrals ] and the digamma reflection formula ( equation 6.3.7 from @xcite ) , and then the first formula in follows by taylor s theorem .
similarly , for the second formula in , @xmath367 we conclude this subsection with two algebraic results .
[ lem : finding - theta ] suppose that holds .
given @xmath368 with @xmath369 for all @xmath3 , there exists a solution @xmath370 with @xmath371 for all @xmath3 to the system of equations @xmath372 if and only if @xmath373 .
moreover , if a solution to exists , we may take @xmath374 for all @xmath23 .
as column vectors , we write @xmath375 for the stationary probabilities as given in , @xmath376 , and @xmath377
. then in matrix - vector form , reads @xmath378 , while @xmath379 satisfies , which reads @xmath380 , the homogeneous system adjoint to .
( here @xmath381 is the identity matrix and @xmath382 is the vector of all 0s . )
a standard result from linear algebra ( a version of the fredholm alternative ) says that @xmath378 admits a solution @xmath383 if and only if the vector @xmath384 is orthogonal to any solution @xmath385 to @xmath386 ; but , by , any such @xmath385 is a scalar multiple of @xmath379 .
in other words , a solution @xmath383 to exists if and only if @xmath387 , as claimed .
finally , since @xmath69 is a stochastic matrix , @xmath388 , where @xmath389 is the column vector of all 1s ; hence if @xmath383 solves @xmath378 , then so does @xmath390 for any @xmath391 .
this implies the final statement in the lemma .
[ lem : algebraic2 ] let @xmath392 be a given upper triangular matrix having all its upper triangular elements non - negative ( @xmath393 for @xmath394 and vanishing all other elements ) and @xmath395 a vector with positive components .
then there exists a unique lower triangular matrix @xmath396 ( so diagonal and upper triangular elements vanish ) satisfying * @xmath397 for @xmath398 ; * @xmath399 for @xmath400 . also , all lower triangular elements of @xmath401 are positive , i.e. @xmath402 for @xmath403 .
we construct @xmath401 inductively .
item ( i ) demands @xmath404 in the case @xmath405 , with the usual convention that an empty sum is @xmath364 , the demand is simply @xmath406 .
so we can start our construction taking @xmath407 , which is positive by assumption .
( item ( ii ) makes no demands in the case @xmath408 , @xmath409 . )
suppose now that all matrix elements @xmath410 have been computed in the lower - right corner @xmath411
( @xmath412 ) : @xmath413 the elements of @xmath401 involved in statement ( i ) ( for given @xmath414 ) and in statement ( ii ) for @xmath415 are all in @xmath411 ; thus as part of our inductive hypothesis we may suppose that the elements of @xmath411 are such that ( i ) holds for the given @xmath414 , and ( ii ) holds with @xmath415 and all @xmath416
. we have shown that we can achieve this for @xmath417 .
the inductive step is to construct from @xmath411 ( @xmath418 ) elements @xmath419 for @xmath420 and hence complete the array @xmath421 in such a way that ( i ) holds for @xmath422 replacing @xmath414 , that ( ii ) holds for @xmath423 , and that all elements are positive .
now reveals the demand of item ( i ) as @xmath424 which we can achieve since the elements of @xmath401 on the right - hand side are all in @xmath411 , and since @xmath425 we get @xmath426 .
the @xmath423 case of ( ii ) demands that for @xmath416 we have @xmath427 which involves only elements of @xmath411 in addition to @xmath428 , which we have already defined , and positivity of all the @xmath419 follows by hypothesis .
this gives us the construction of @xmath421 and establishes the inductive step .
this algorithm can be continued down to @xmath429 .
but then the lower triangular matrix @xmath401 is totally determined .
the diagonal and upper triangular elements of @xmath401 do not influence the construction , and may be set to zero .
[ cor : algebraic2-open ] let the matrix @xmath430 and the vector @xmath431 be as in lemma [ lem : algebraic2 ] .
let @xmath432 be the set of lower triangular matrices @xmath433 satisfying * @xmath434 for @xmath398 ; * @xmath435 for @xmath400 , viewed as subset of the positive cone @xmath436 . then @xmath432 is a non - empty , open subset of @xmath437 .
we use the notation @xmath438 we start with the case @xmath99 .
we will obtain a local supermartingale by choosing the @xmath439 carefully .
lemma [ lem : finding - theta ] , which shows how the stationary probabilities @xmath0 enter , is crucial ; a similar idea was used for random walks on strips in section 3.1 of @xcite .
next is our key local supermartingale result in this case .
[ pro : main - supermartingale ] suppose that and hold , and that @xmath99 . * if @xmath100 , then there exist @xmath440 , @xmath441 ( @xmath23 ) , @xmath235 , and @xmath328 such that @xmath442 \leq - { \varepsilon}x^{\nu - \bar \alpha } , ~\text{for all } x \geq x_0 \text { and all } i \in { { \mathcal s}}.\ ] ] * if @xmath101 , then there exist @xmath443 , @xmath441 ( @xmath23 ) , @xmath235 , and @xmath328 such that @xmath442 \leq - { \varepsilon}x^{\nu - \bar \alpha } , ~\text{for all } x \geq x_0 \text { and all } i \in { { \mathcal s}}.\ ] ] write @xmath444 , @xmath23 .
first we prove part ( i ) . by hypothesis ,
@xmath445 for some @xmath446 .
set @xmath447 for all @xmath23 , so that @xmath373 .
lemma [ lem : finding - theta ] shows that for these @xmath448 , we can find a collection of @xmath449 so that @xmath450 fix these @xmath451 for the rest of the proof .
we then choose @xmath452 of the form @xmath453 , for some @xmath454 .
since @xmath455 for all @xmath456 and all @xmath457 , lemma [ lem : df ] shows that @xmath362 will be negative for all @xmath35 and all @xmath167 sufficiently large provided that we can find @xmath295 such that @xmath458 for all @xmath35 . by , writing @xmath377 , @xmath459 by .
therefore , by lemma [ lem : df ] , we can always find sufficiently small @xmath460 and a vector @xmath461 with strictly positive elements for which @xmath462 for some @xmath235 , all @xmath35 , and all @xmath167 sufficiently large .
maximizing over @xmath35 gives part ( i ) .
the argument for part ( ii ) is similar .
suppose @xmath443 .
this time , @xmath463 for some @xmath446 , and we set @xmath464 , so that @xmath373 once more .
lemma [ lem : finding - theta ] now shows that we can find @xmath451 so that @xmath465 with this choice of @xmath451 we again set @xmath466 ; note we may assume @xmath467 for all @xmath3 for @xmath295 sufficiently small .
again , @xmath362 will be non - negative for all @xmath35 and all @xmath167 sufficiently large provided that we can find @xmath468 and @xmath295 such that @xmath458 for all @xmath35 . following a similar argument to before ,
we obtain with that @xmath469 thus we can find for @xmath470 close enough to @xmath364 a vector @xmath461 with strictly positive elements for which @xmath471 for all @xmath35 and all @xmath167 sufficiently large . now we examine the case @xmath472 . [
pro : recurrence - i ] suppose that and hold , and @xmath98 .
then there exist @xmath473 , @xmath441 ( @xmath23 ) , @xmath235 , and @xmath328 such that for all @xmath474 , @xmath475 \leq - { \varepsilon}x^{\nu - \alpha^\star } , ~\text{for all } x \geq x_0.\ ] ] before starting the proof of this proposition , we introduce the following notation . for @xmath111 denote by @xmath476 and define the vector @xmath477 . for @xmath474 and @xmath42 , write @xmath478 .
define @xmath479 and recursively , for @xmath480 , @xmath481 denote by @xmath482 .
since @xmath69 is irreducible , the collection @xmath483 is a partition of @xmath13 .
it suffices to find @xmath484 , @xmath235 , @xmath473 , and an open , non - empty subset @xmath485 of the positive cone @xmath486 such that @xmath487 now for @xmath488 , inequality is satisfied thanks to , and lemmas [ lem : all - integrals ] and [ lem : big - alpha ] for every choice of @xmath468 ( with positive components ) and @xmath440 .
hence the previous condition reduces to the requirement @xmath489 the rest of the proof is devoted into establishing this fact .
suppose that @xmath490 with @xmath491 .
then @xmath492 by construction , and lemma [ lem : df ] shows that condition will be satisfied if the system of inequalities @xmath493 have non - trivial solutions @xmath468 for sufficiently small @xmath295 .
thanks to the lemma [ lem : df ] , we have @xmath494
. we will obtain if , for @xmath295 sufficiently small , @xmath495 we seek a solution @xmath496 ( for sufficiently small @xmath295 ) , under the ansatz that the @xmath497 are constant on every @xmath498 , i.e. , the vector @xmath468 has the form @xmath499 with @xmath500 for all @xmath501 .
suppose that @xmath502 .
then @xmath503 for @xmath504 with @xmath505 , so that @xmath506 we introduce the auxiliary matrix @xmath507 defined by @xmath508 . by construction , @xmath509 and @xmath510 .
let @xmath511 it suffices to determine the upper triangular part of @xmath512 , or , equivalently , the lower triangular array @xmath513 .
we do so recursively , starting with @xmath514 . in the case @xmath515 , the condition in reads , by , @xmath516 on introducing the constant @xmath517 , it is enough to choose @xmath518 .
in other words , with @xmath519 , we see that the choice @xmath520 satisfies the condition in for @xmath521 .
suppose now that we have determined the condition in for @xmath522 .
then , for @xmath523 the condition amounts , via , to @xmath524 using the fact that for @xmath525 we have @xmath526 and for @xmath527 we have @xmath528 , the above expression becomes , up to @xmath529 terms @xmath530 introducing the upper triangular matrix @xmath531 defined by @xmath532 for @xmath533 , and the vector @xmath534 for @xmath398 , the condition in is satisfied if we solve the recursion @xmath535 with initial condition @xmath536 and condition @xmath537 for @xmath538 .
additionally , we have from that @xmath539 hence by corollary [ cor : algebraic2-open ] , there exist non - trivial solutions for the lower triangular matrix @xmath401 within an algorithmically determined region @xmath540 .
the positivity of the lower triangular part of @xmath401 implies that the components of @xmath468 are ordered : @xmath541 for @xmath542 . by choosing @xmath543 sufficiently large
, we can guarantee that the first term ( and consequently the entire sequence ) satisfies @xmath544 .
we are almost ready to complete the proof of theorem [ thm : recurrence ] , excluding part ( b)(iii ) ; first we need one more technical result concerning non - confinement .
[ l : orw_non - confinement ] suppose that and hold . then @xmath545 , a.s .
we claim that for each @xmath123 , there exists @xmath546 such that @xmath547 \geq { \varepsilon}_x , ~\text{for all}~ y \in [ 0,x ] \text { and all } i\in { { \mathcal s}}.\ ] ] indeed , given @xmath54 , we may choose @xmath31 so that @xmath548 and we may choose @xmath549 sufficiently large so that , for some @xmath235 , @xmath550 for all @xmath551 . then if @xmath552 $ ] , @xmath553 & \geq p(i , j ) \int_{2z_0 + 1}^\infty { \varepsilon}z^{-1-\alpha_i } { { \mathrm d}}z = { \varepsilon}_{x , i } > 0 , \end{aligned}\ ] ] which gives . the local escape
property implies the @xmath554 result by a standard argument .
we are not yet ready to prove part ( b)(iii ) : we defer that part of the proof until section [ sec : critical ] .
the other parts of the theorem follow from the supermartingale estimates in this section together with the technical results from section [ sec : criteria ] .
indeed , under the conditions of part ( a ) or ( b)(i ) of the theorem , we have from proposition [ pro : recurrence - i ] or proposition [ pro : main - supermartingale](i ) respectively that for suitable @xmath331 and @xmath439 , @xmath555 \leq 0 , ~~\text{on}~ \ { x_n \geq x_0\ } .\ ] ] thus we may apply lemma [ l : adapted_rec_r ] , which together with lemma [ l : orw_non - confinement ] shows that @xmath556 , a.s .
thus there exists an interval @xmath557 $ ] such that @xmath558 i.o .
, where @xmath35 is some fixed element of @xmath13 .
let @xmath559 and for @xmath560 define @xmath561 .
given @xmath30 , we may choose @xmath562 such that @xmath563 and @xmath564 for some @xmath565 ; let @xmath566 .
then we may choose @xmath567 and @xmath568 such that @xmath569 and @xmath570 for all @xmath551 .
then for any @xmath571 , @xmath572 & \geq { { \mathbb p } } [ x_{\tau_k + 2 } < { \varepsilon } , x_{\tau_k + 1 } \in [ z_0 + 1 , z_0 + 2 ] , \eta_{\tau_k+1 } = j \mid x_{\tau_k } ] \\ & \geq \delta_1 ^ 2 \delta_2 ^ 2 { \varepsilon}(z_0 + 3)^{-1-\gamma } ( x_0 + z_0 + 3)^{-1-\gamma } , \end{aligned}\ ] ] uniformly in @xmath3 .
thus lvy s extension of the borel
cantelli lemma shows @xmath573 infinitely often .
thus , since @xmath571 was arbitrary , @xmath574 , a.s .
on the other hand , under the conditions of part ( b)(ii ) of the theorem , we have from proposition [ pro : main - supermartingale](ii ) that for suitable @xmath575 and @xmath439 , @xmath555 \leq 0 , ~~\text{on}~ \ { x_n \geq x_1\ } , \ ] ] for any @xmath576 sufficiently large .
thus we may apply lemma [ l : adapted_good_escape_r ] , which shows that for any @xmath235 there exists @xmath577 for which , for all @xmath578 , @xmath579 \geq 1 - { \varepsilon } , \text { on } \ { x_n \geq x \ } .\ ] ] set @xmath580 . then , on @xmath581 , @xmath582 \geq 1 - { \varepsilon } , { \ \text{a.s.}}\ ] ] but on @xmath583 we have @xmath584 , so @xmath585 & \geq \operatorname{\mathbb{e}}\left [ { { \mathbb p}}\bigl [ \inf_{m \geq \sigma_x } x_m > x_1 { \ ; \bigl| \;}{{\mathcal f}}_{\sigma_x } \bigr ] { { \mathbf 1}{\ { \sigma_x < \infty \ } } } \right ] \\ & \geq ( 1 -{\varepsilon } ) { { \mathbb p } } [ \sigma_x < \infty ] = ( 1-{\varepsilon}),\end{aligned}\ ] ] by lemma [ l : orw_non - confinement ] . since @xmath235 was arbitrary , we get @xmath584 , a.s . , and since @xmath576 was arbitrary we get @xmath586 , a.s .
the following result is a straightforward reformulation of theorem 1 of @xcite .
[ lem : aim ] let @xmath258 be an integrable @xmath587-adapted stochastic process , taking values in an unbounded subset of @xmath41 , with @xmath588 fixed .
for @xmath589 , let @xmath590 .
suppose that there exist @xmath87 , @xmath589 , and @xmath591 such that for any @xmath578 , @xmath592 \leq -\delta y_n^{\gamma } , \text { on } \{n < \sigma_x \}.\ ] ] then , for any @xmath593 , @xmath594 < \infty$ ] .
the following companion result on non - existence of moments is a reformulation of corollary 1 of @xcite .
[ lem : aim2 ] let @xmath258 be an integrable @xmath587-adapted stochastic process , taking values in an unbounded subset of @xmath41 , with @xmath588 fixed .
for @xmath589 , let @xmath590 .
suppose that there exist @xmath595 , @xmath589 , @xmath596 and @xmath597 such that for any @xmath578 , on @xmath598 the following hold : @xmath599 & \geq - c_1 ; \\ \label{eq : aim2b } \operatorname{\mathbb{e } } [ y^r_{n+1}-y^r_n \mid { { \mathcal f}}_n ] & \leq c_2 y_n^{r-1 } ; \\ \label{eq : aimc } \operatorname{\mathbb{e } } [ y_{n+1}^p - y_n^p \mid { { \mathcal f}}_n ] & \geq 0 .\end{aligned}\ ] ] then for any @xmath600 , @xmath601 = \infty$ ] for @xmath602 . under conditions ( a ) or ( b)(i ) of theorem [ thm : recurrence ]
, we have from propositions [ pro : recurrence - i ] or [ pro : main - supermartingale ] respectively that there exist positive @xmath264 and constants @xmath235 , @xmath603 and @xmath604 such that , @xmath605 let @xmath606 . then @xmath258 is bounded above and below by positive constants times @xmath607 , so we have that holds for @xmath167 sufficiently large with @xmath608 .
it follows from lemma [ lem : aim ] that @xmath594 < \infty$ ] for @xmath609 , which gives the claimed existence of moments result .
it is not hard to see that _ some _ moments of the return time fail to exist , due to the heavy - tailed nature of the model , and an argument is easily constructed using the ` one big jump ' idea : a similar idea is used in @xcite .
we sketch the argument . for any @xmath610 , for all @xmath611
sufficiently large we have @xmath612 \geq { \varepsilon}y^{-\bar \alpha}$ ] .
given such a first jump , with uniformly positive probability the process takes time at least of order @xmath613 to return to a neighbourhood of zero ( where @xmath174 can be bounded in terms of @xmath614 ) ; this can be proved using a suitable maximal inequality as in the proof of theorem 2.10 of @xcite . combining these two facts
shows that with probability of order @xmath615 the return time to a neighbourhood of the origin exceeds order @xmath613 .
this polynomial tail bound yields non - existence of sufficiently high moments .
we now restrict attention to the case @xmath616 with @xmath617 and @xmath618 , so both half - lines are of the same type .
take @xmath619 and @xmath620 , so that @xmath621 . then lemma [ lem : df ] shows that , for @xmath622 , @xmath623 , @xmath624 where @xmath625 the two cases we are interested in are the recurrent two - sided symmetric case , where @xmath626 ( i.e. , @xmath627 ) with @xmath150 , and the recurrent one - sided antisymmetric case , where @xmath628 ( i.e. , @xmath629 ) with @xmath630 .
[ lem : c - properties ] let @xmath623 and @xmath631 .
the function @xmath632 is continuous for @xmath633 with @xmath634 and @xmath635 .
there exists @xmath636 such that @xmath637 for @xmath638 , @xmath639 , and @xmath640 for @xmath641 we give the proof only in the case @xmath642 ; the other case is very similar .
thus @xmath643 , and , for ease of notation , we write just @xmath644 instead of @xmath645
. clearly @xmath646 . for @xmath647 , convexity of the function @xmath648 on @xmath41 shows that @xmath649 for all @xmath650 $ ] , so that @xmath651 ; clearly , @xmath652 and @xmath653 are also non - negative . hence , by the expression for @xmath654 in lemma [ lem : i - integrals ] , @xmath655 which is @xmath656 . moreover , by lemma [ lem : j - integrals ] and the subsequent remark , @xmath657 which is negative for @xmath11 .
hence @xmath658 for @xmath460 small enough .
since @xmath659 is a non - constant analytic function on @xmath660 , its zeros can accumulate only at @xmath173 ; but this is ruled out by the fact that @xmath661 as @xmath662 .
hence @xmath663 has only finitely many zeros in @xmath660 ; one is at @xmath364 , and there must be at least one zero in @xmath664 , by rolle s theorem .
define @xmath665 and @xmath666 to be the smallest and largest such zeros , respectively .
suppose @xmath667 . by jensen s inequality , @xmath668 \geq \left ( \operatorname{\mathbb{e}}_{x , i } [ ( 1 + x_1 ) ^{\nu_1 } ] \right)^{\nu_2/\nu_1}$ ] .
hence @xmath669 using taylor s theorem and the fact that @xmath670 , by . by another application of
, it follows that @xmath671 multiplying by @xmath672 and taking @xmath325 , we obtain @xmath673 in particular , ( i ) if @xmath674 then @xmath675 for all @xmath676 ; and ( ii ) if @xmath677 and @xmath678 , we have @xmath679 .
it follows from these two observations that @xmath680 for @xmath681 $ ] , which is not possible unless @xmath682 .
hence there is exactly one zero of @xmath663 in @xmath683 ; call it @xmath684 .
[ lem : nu0 ] the positive zero of @xmath685 described in lemma [ lem : c - properties ] is given by @xmath686 or @xmath687 .
first suppose @xmath688 .
then from lemma [ lem : i - integrals ] we verify that for @xmath689 , @xmath690 since @xmath691 .
now suppose that @xmath642 and @xmath11 .
to verify @xmath692 it is simpler to work with the integral representations directly , rather than the hypergeometric functions .
after the substitution @xmath693 , we have @xmath694 similarly , after the same substitution , @xmath695 which we may evaluate as @xmath696 finally we have that @xmath697 , so altogether we verify that @xmath698 .
we can now complete the proofs of theorems [ thm : symmetric_random_walk_moments ] and [ thm : one - sided_moments ] .
let @xmath699 .
first suppose that @xmath11 . then we have from together with lemmas [ lem : c - properties ] and [ lem : nu0 ] that , for any @xmath700 , @xmath701 \leq - { \varepsilon}y_n^{1-(\alpha/\nu ) } , \text { on } \ { y_n \geq y_0 \ } , \ ] ] for some @xmath235 and @xmath308 .
it follows from lemma [ lem : aim ] that @xmath702 < \infty$ ] for @xmath703 and since @xmath704 was arbitrary we get @xmath702 <
\infty$ ] for @xmath705 . for
the non - existence of moments when @xmath11 , we will apply lemma [ lem : aim2 ] with @xmath706 for some @xmath707 . then condition follows from , which also shows that for @xmath708 , @xmath709 = \frac{c_i}{2 } x^{r \nu - \alpha } c^{\rm sym } ( r \nu , \alpha ) + o ( x^{r \nu -\alpha } ) , \end{aligned}\ ] ] so that @xmath710 \leq c_i c^{\rm sym } ( r \nu , \alpha ) y_n^{r - ( \alpha / \nu)}$ ] , for all @xmath258 sufficiently large . since @xmath711 condition follows .
finally , we may choose @xmath712 close enough to @xmath173 and then take @xmath713 so that from , with lemmas [ lem : c - properties ] and [ lem : nu0 ] , @xmath714 \geq 0 $ ] , for all @xmath258 sufficiently large .
thus we may apply lemma [ lem : aim2 ] to obtain @xmath702 = \infty$ ] for @xmath715 , and taking @xmath716 close to @xmath717 and @xmath295 close to @xmath173 we can achieve any @xmath718 , as claimed .
next suppose that @xmath159 . a similar argument to before but this time using lemmas [ lem : all - integrals ] and [ lem : big - alpha ] shows that for any @xmath327 , @xmath701 \leq - { \varepsilon}y_n^{1-(2/\nu ) } , \text { on } \ { y_n \geq y_0 \ } , \ ] ] for some @xmath235 and @xmath308 .
lemma [ lem : aim ] then shows that
@xmath702 < \infty$ ] for @xmath719 and since @xmath327 was arbitrary we get @xmath702 < \infty$ ] for @xmath720 .
we sketch the argument for @xmath721 = \infty$ ] when @xmath722 . for @xmath723 , it is not hard to show that @xmath724 for all @xmath167 sufficiently large , and we may verify the other conditions of lemma [ lem : aim2 ] to show that @xmath702 = \infty$ ] for @xmath722 .
most of the proof is similar to that of theorem [ thm : symmetric_random_walk_moments ] , so we omit the details .
the case where a different argument is required is the non - existence part of the case @xmath184 .
we have that for some @xmath235 and all @xmath611 sufficiently large , @xmath725 \geq { \varepsilon}(x + y)^{-\alpha}$ ] . a similar argument to lemma [ lem : big - alpha ]
shows that for any @xmath327 , for some @xmath726 , @xmath727 \geq - c$ ] .
then a suitable maximal inequality implies that with probability at least @xmath4 , started from @xmath728 it takes at least @xmath729 steps for @xmath51 to return to a neighbourhood of @xmath364 , for some @xmath730 .
combining the two estimates gives @xmath731 \geq \frac{1}{2}(x + y)^{-\alpha } , \ ] ] which implies @xmath702 = \infty$ ] for @xmath732 , and since @xmath327 was arbitrary we can achieve any @xmath733 .
in this section we prove theorem [ thm : recurrence](b)(iii ) . throughout this section
we write @xmath734 and suppose that @xmath99 , that @xmath735 , and that @xmath736 for all @xmath30 , that is , for @xmath737 , @xmath738 , with @xmath739 and @xmath740 , where @xmath303 may be chosen so as not to depend upon @xmath35 . to prove recurrence in the critical cases , we need a function that grows more slowly than any power ; now the weights @xmath264 are additive rather than multiplicative . for @xmath203 write @xmath741
then , for @xmath123 and @xmath23 , define @xmath742 where @xmath467 for all @xmath23 . also write @xmath743 lemma [ lem : finding - theta ] shows that there exist @xmath266 ( @xmath23 ) such that @xmath744 we fix such a choice of the @xmath264 from now on .
we prove recurrence by establishing the following result .
[ lem : h - lem ] suppose that the conditions of theorem [ thm : recurrence](b)(iii ) hold , and that @xmath745 are such that holds
. then there exists @xmath328 such that @xmath746 \leq 0 , ~\text{for } ~ x \geq x_0 \text { and all } i \in { { \mathcal s}}.\ ] ] it is not easy to perform the integrals required to estimate @xmath747 ( and hence establish lemma [ lem : h - lem ] ) directly .
however , the integrals for @xmath748 are computable ( they appear in lemma [ lem : j - integrals ] ) , and we can use some analysis to relate @xmath747 to @xmath748 .
thus the first step in our proof of lemma [ lem : h - lem ] is to estimate @xmath749 . for the lyapunov function @xmath750 ,
gives for @xmath123 and @xmath30 , @xmath751 \nonumber\\ & = \sum_{j \in { { \mathcal s } } } p ( i , j ) ( \lambda_j - \lambda_i ) \int_{-\infty}^{-x } w_i ( y ) { { \mathrm d}}y + \int_{-\infty}^\infty \left ( g(x+y ) - g(x ) \right ) w_i ( y ) { { \mathrm d}}y \nonumber \\ & = \chi_i \sum_{j \in { { \mathcal s } } } p ( i , j ) ( \lambda_j - \lambda_i ) t_{i } ( x ) + g^\flat_{i } ( x),\end{aligned}\ ] ] where we have used the fact that @xmath752 is defined for all @xmath203 and symmetric about @xmath364 , and we have introduced the notation @xmath753 and @xmath754 the next lemma , proved in the next subsection , estimates the integrals in . [ lem : g - estimate ] suppose that @xmath736 . then , for @xmath755 , for @xmath756 , for some @xmath757 , as @xmath325 , @xmath758 note that lemma [ lem : g - estimate ] together with shows that @xmath759 by .
this is not enough by itself to establish recurrence , since the sign of the @xmath760 term is unknown .
this is why we need the function @xmath761 .
to ease notation , we drop the subscripts @xmath35 everywhere for the duration of this proof . from we
obtain , for @xmath589 , @xmath762 v ( y ) { { \mathrm d}}y .\end{aligned}\ ] ] let @xmath282 . the claim in the lemma for @xmath763 will follow from the estimates @xmath764 v ( y ) { { \mathrm d}}y & = c x^{-\alpha } j_1^\alpha + o ( x^{-\alpha-\eta } ) ; \end{aligned}\ ] ] since then we obtain from with and lemma [ lem : j - integrals ] that @xmath765 which yields the stated result via the digamma reflection formula ( equation 6.3.7 from @xcite ) .
we present here in detail the proof of only the final estimate in ; the others are similar .
some algebra followed by the substitution @xmath766 shows that the third integral in is @xmath767 there is a constant @xmath726 such that for all @xmath167 sufficiently large , @xmath768 using taylor s theorem for @xmath769 and the fact that @xmath770 is uniformly bounded . on the other hand , for @xmath771 we have @xmath772 , so that @xmath773 here we have that @xmath774 combining these estimates and using the fact that @xmath282 , we obtain the final estimate in .
the claim in the lemma for @xmath775 follows after some analogous computations , which we omit .
now we relate @xmath776 to @xmath777 , by comparing the individual integral terms .
[ lem : h - vs - g ] suppose that @xmath736 .
then , for all @xmath167 and all @xmath35 , @xmath778 finally , there exists @xmath235 such that , for all @xmath779 and all @xmath167 sufficiently large , @xmath780 the proof is based on the observation that , since @xmath781 , @xmath782 thus , for @xmath783 , @xmath784 since @xmath785 and @xmath786 .
this gives the first inequality in the lemma .
similarly , for @xmath552 $ ] , @xmath787 since @xmath788 and @xmath789 .
this gives the second inequality , and also yields the third inequality once combined with the @xmath552 $ ] case of . finally , for @xmath790 note that @xmath791 also note that , for @xmath792 , @xmath793 and @xmath794 , so @xmath795 so the sign of the expression in is non - positive for @xmath796 and non - negative for @xmath797 , and @xmath798 by the monotonicity in @xmath611 of the denominator , the expression in satisfies @xmath799 both for @xmath800 $ ] and for @xmath801 . here
@xmath802 , by .
hence we obtain the bound @xmath803 to improve on this estimate , suppose that @xmath804 where @xmath805 is such that @xmath806 . then , using the fact that the numerator in is positive , we may choose @xmath805 such that for all @xmath294 and all @xmath807 , @xmath808 for @xmath809 sufficiently large ( depending on @xmath810 ) and and all @xmath167 sufficiently large .
here @xmath811 it follows that , for some @xmath235 , for all @xmath167 sufficiently large , for @xmath807 , @xmath812 the final inequality in the lemma now follows since , for all @xmath167 sufficiently large , @xmath813 where we used the substitution @xmath814 . for @xmath809 sufficiently large
the term inside the logarithm is uniformly positive , and the claimed bound follows .
now we may complete the proofs of lemma [ lem : h - lem ] and then theorem [ thm : recurrence](b)(iii ) .
lemma [ lem : h - vs - g ] together with shows that , @xmath815 for all @xmath167 sufficiently large .
lemma [ lem : h - lem ] with lemma [ l : adapted_rec_r ] shows that @xmath556 , a.s . , and then a similar argument to that in the proof of parts ( a ) and ( b)(i ) of theorem [ thm : recurrence ] shows that @xmath816 , a.s .
[ l : adapted_rec_r ] let @xmath817 be an @xmath587-adapted process taking values in @xmath12 .
let @xmath818 be such that @xmath819 for all @xmath30 , and @xmath820 .
suppose that there exist @xmath328 and @xmath821 for which , for all @xmath578 , @xmath822 & \leq 0 , \text { on } \ { x_n >
x_0 \ } , { \ \text{a.s . } } ; \\
\operatorname{\mathbb{e } } [ f(x_{n+1 } , \xi_{n+1 } ) - f(x_n , \xi_n ) \mid { { \mathcal f}}_n ] & \leq c , \text { on } \ { x_n \leq x_0 \ } , { \ \text{a.s.}}\end{aligned}\ ] ] then @xmath823 = 1 .\ ] ] fix @xmath210 .
for @xmath328 in the hypothesis of the lemma , write @xmath826
. let @xmath827 .
then @xmath828 is an @xmath829-adapted non - negative supermartingale .
hence , by the supermartingale convergence theorem , there exists @xmath830 such that @xmath831 , a.s .
in particular , @xmath832 set @xmath833 , which has @xmath834 a.s . since @xmath835 .
then @xmath836 on @xmath837 .
hence @xmath838 = 1 .\ ] ] since @xmath839 was arbitrary , the result follows : @xmath840 = 1 .
\qedhere\ ] ] [ l : adapted_good_escape_r ] let @xmath817 be an @xmath587-adapted process taking values in @xmath12 .
let @xmath818 be such that @xmath841 and @xmath842 for all @xmath30 .
suppose that there exists @xmath843 for which @xmath844 for all @xmath35 and @xmath845 \leq 0 , \text { on } \{x_n
\geq x_1 \ } , \text { for all } n \geq 0.\ ] ] then for any @xmath235 there exists @xmath577 for which , for all @xmath578 , @xmath579 \geq 1 - { \varepsilon } , \text { on } \
{ x_n \geq x \ } .\ ] ] fix @xmath210 . for @xmath843 in the hypothesis of the lemma , write @xmath846 and set @xmath847
. then @xmath848 is an @xmath829-adapted non - negative supermartingale , and so converges a.s .
as @xmath849 to some @xmath830 . moreover , by the optional stopping theorem for supermartingales , @xmath850 \geq \operatorname{\mathbb{e } } [ y_\infty { { \mathbf 1}{\ { \lambda <
\infty \ } } } \mid { { \mathcal f}}_{n } ] , { \ \text{a.s.}}\ ] ] here we have that , a.s . , @xmath851 combining these inequalities we obtain @xmath852 \leq y_{n } , { \
\text{a.s.}}\ ] ] in particular , on @xmath853 , we have @xmath854 and so @xmath855 \leq f ( x_n , \xi_n ) \leq \max_i \sup_{y \geq x } f(y , i ) .\ ] ] since @xmath856 and @xmath844 , given @xmath235 we can choose @xmath857 large enough so that @xmath858 the choice of @xmath167 depends only on @xmath859 , @xmath576 , and @xmath319 , and , in particular , does not depend on @xmath49 . then , on @xmath860 , @xmath861
< { \varepsilon}$ ] , as claimed .
consider @xmath862 . with the change of variable @xmath863
, we get @xmath864 by the integral representation for the gauss hypergeometric function ( see equation 15.3.1 of @xcite ) . the given expression for @xmath862 follows .
the integral @xmath865 is trivial .
consider @xmath654 .
the substitution @xmath863 gives @xmath866 by the integral formula for the beta function ( see equation 6.2.1 of @xcite ) , provided @xmath867 and @xmath868 . hence we obtain the given expression for @xmath654 .
next consider @xmath869 . by considering separately the asymptotics of the integrand as @xmath870 and @xmath871
, we see that @xmath869 is finite provided @xmath282 and @xmath283 . for @xmath872 $ ]
, we have the taylor series expansion @xmath873 .
hence @xmath874 here , the power series for @xmath875 converges normally ( hence uniformly ) over @xmath876 .
this remark allows interchanging summation and integration to obtain @xmath877 , where for @xmath878 we define @xmath879 \nonumber\\ & = \frac{1}{2-\alpha } \nu(\nu-1)z\left[\sum_{n\geq 0 } c_n z^n\right ] , \nonumber\end{aligned}\ ] ] where @xmath880 , for @xmath881 and @xmath882 .
an elementary computation yields @xmath883 therefore ( see @xcite or ( * ? ? ?
* equation 5.81 , page 207 ) for a more easily accessible reference ) , @xmath884 .
now , the series defining the generalized hypergeometric function @xmath885 for @xmath886 converges for all @xmath338 with @xmath887 ; for @xmath888 the series converges for @xmath889 , a condition that reduces to @xmath283 in the present case . hence @xmath890 finally consider @xmath891 . due to the singularity at @xmath892
, we compute @xmath893 for the last integral , we get , for @xmath279 and @xmath894 , @xmath895 with the substitution @xmath896 , the last integral becomes @xmath897 provided @xmath283 , by the integral representation for the gauss hypergeometric function ( equation 15.3.1 of @xcite ) .
now , by equation 15.3.6 from @xcite , @xmath898 as @xmath899 , provided @xmath285 . combining these results we get @xmath900 letting @xmath899 and using the fact that @xmath901 we obtain the given expression for @xmath902 .
we appeal to tables of standard integrals ( mellin transforms ) from section 6.4 of @xcite .
in particular , the given formulae for @xmath903 , @xmath904 , and @xmath905 follow from , respectively , equations 6.4.17 , 6.4.20 , and 6.4.19 of @xcite ; also used are formulas 6.3.7 and 6.3.8 from @xcite and the fact that @xmath906 .
lastly , for @xmath907 we use the substitution @xmath908 to obtain @xmath909 [ lem : integral_approx ] let @xmath910 be a borel set , and let @xmath859 and @xmath750 be measurable functions from @xmath431 to @xmath8 .
suppose that there exist constants @xmath911 and @xmath912 with @xmath913 such that @xmath914 for all @xmath915 .
then , @xmath916 s. aspandiiarov , r. iasnogorodski , and m. menshikov , passage - time moments for nonnegative stochastic processes and an application to reflected random walks in a quadrant , _ ann .
_ * 24 * ( 1996 ) 932960
. m. campanino and d. petritis , on the physical relevance of random walks : an example of random walks on a randomly oriented lattice , pp .
393411 in _ random walks and geometry _
, walter de gruyter , berlin , 2004 .
o. hryniv , i.m .
macphee , m.v .
menshikov , and a.r .
wade , non - homogeneous random walks with non - integrable increments and heavy - tailed random walks on strips , _ electron .
j. probab .
_ * 17 * ( 2012 ) article 59 .
|
we study a random walk on a complex of finitely many half - lines joined at a common origin ; jumps are heavy - tailed and of two types , either one - sided ( towards the origin ) or two - sided ( symmetric ) .
transmission between half - lines via the origin is governed by an irreducible markov transition matrix , with associated stationary distribution @xmath0 .
if @xmath1 is @xmath2 for one - sided half - lines @xmath3 and @xmath4 for two - sided half - lines , and @xmath5 is the tail exponent of the jumps on half - line @xmath3 , we show that the recurrence classification for the case where all @xmath6 is determined by the sign of @xmath7 . in the case of two half - lines ,
the model fits naturally on @xmath8 and is a version of the _ oscillating random walk _ of kemperman . in that case , the cotangent criterion for recurrence becomes linear in @xmath9 and @xmath10 ; our general setting exhibits the essential non - linearity in the cotangent criterion .
for the general model , we also show existence and non - existence of polynomial moments of return times .
our moments results are sharp ( and new ) for several cases of the oscillating random walk ; they are apparently even new for the case of a homogeneous random walk on @xmath8 with symmetric increments of tail exponent @xmath11 . _
key words : _ random walk , heavy tails , recurrence , transience , passage time moments , lyapunov functions , oscillating random walk , cotangent criterion .
_ ams subject classification : _
60j05 ( primary ) ; 60j10 , 60g50 ( secondary )
|
originally proposed by paczyski ( 1986 ) as a method to detect the presence of massive compact objects in the halo of our galaxy , microlensing has increasingly been recognized as a tool for studying a broad range of astrophysical phenomena .
various applications include detection and characterization of binary and planetary systems ( mao & paczyski 1991 ; gould & loeb 1992 ) , reconstruction of the stellar mass function down to masses below the hydrogen burning limit ( paczyski 1991 ; griest et al .
1991 ; gould 1996 ) , measurement of the rotation speed of giants ( gould 1997 ) , measuring the transverse velocity of galaxies ( gould 1995a ) , and probing the central engines of quasars ( gould & gaudi 1997 ) .
recently , valls - gabaud ( 1994 , 1997 ) and sasselov ( 1996 ) have proposed using microlensing to study stellar atmospheres .
here we analyze this application in detail
. currently , three collaborations ( macho , alcock et al . 1997a ; ogle , udalski et al . 1997 ; eros , ansari et al .
1996 ) have ongoing projects that survey the galactic bulge with roughly nightly sampling in order to detect microlensing events . over
60 events per year are being detected .
these data are being analyzed real time which has allowed macho to issue `` alerts '' , notifications of ongoing events detected before the peak .
two follow - up collaborations ( planet , albrow et al.1997 , 1998a ; gman , alcock et al .
1997b ) have formed in order to monitor these alerts around the clock with high precision and high temporal sampling with the aim of detecting ( and further alerting on ) deviations from the standard microlensing light curve , such as would be expected from binary lenses , binary sources , finite sources , and parallax .
these deviations are useful in that they can provide additional information about the lens and/or source .
in addition , a collaboration has formed to conduct _ spectroscopic _ monitoring of alerted events in order to study the source stars in detail and has observed several events to date ( lennon et al .
thus , the prospect for the real - time detection and monitoring of light curve anomalies both photometrically and spectroscopically is promising .
all gravitational microlenses have caustics , defined as the set of points in the source plane where the magnification of a point source is formally infinite .
when a finite source crosses a caustic , the gradient of the magnification over the source is large , and therefore different parts of the source are magnified by different amounts .
hence , the source is partially resolved .
different parts of the source are resolved at different times during the caustic crossing , and thus , by taking a series of measurements during the course of the crossing , one can recover the intensity profile of the source .
several workers realized that finite source effects could be useful for breaking or partially breaking the degeneracy among microlensing event parameters ( gould 1994 ; nemiroff & wickramasinghe 1994 ; witt & mao 1994 ; maoz & gould 1994 ) and that variations in the surface profile could be exploited to this end ( witt 1995 ; loeb & sasselov 1995 ; gould & welch 1996 ) .
however , valls - gabaud ( 1994 , 1997 ) , sasselov ( 1996 ) , and heyrovsky , sasselov , & loeb ( 1998 ) proposed to exploit the same effects for radically different aims . instead of using finite - source effects to learn more about the lens , they sought to learn more about the source .
the basic idea is as follows
: imagine that one could image separately different annular rings on the surface of a star .
in effect one would be sampling different depths of the photosphere .
since the temperature varies as a function of depth , the broad spectral energy distribution would change with annulus , with more blue light near the center ( greater depths ) and more red light near the outer limb ( lesser depths ) .
that is , the star would be limb - darkened , and more in the blue than the red . since different spectral lines form at different depths , one would expect that the detailed spectral profile would vary as a function of annular radius .
hence , the entire atmosphere could be studied as a function of depth by resolving the two - dimensional ( radius and wavelength ) spectral profile of the star .
currently , it is only possible to study stellar atmospheres in this way for the sun and eclipsing binaries .
however , since the surface of the source star is partially resolved during a microlensing caustic crossing , one can also probe the atmospheres of the source stars for these types of events .
both valls - gabaud ( 1994 , 1997 ) and sasselov ( 1996 ) used specific stellar atmosphere models to construct broad - band and spectral line brightness profiles , and then used these profiles to predict in detail the variations in the broad - band color or equivalent width of specific lines that one would expect during the course of a point - mass microlensing event .
sasselov ( 1996 ) , and heyrovsky et al .
( 1997 ) also consider the effects of star spots on the microlensing light curve .
in addition , they compared their predictions to macho alert event 95 - 30 , a point - mass caustic crossing event for which spectra were taken during the course of the crossing , and for which variations in the optical tio bands were detected ( alcock et al .
all of these authors predict that the color and spectral - line variations during the caustic crossing should be significant and note that this provides an entirely new method of studying stellar atmospheres . although caustic crossing events are in principle useful for studying the atmospheres of stars , these events are rare , and they typically last for only about 7 hours ( for a giant source ) . for this method to be successful , it is essential that observers have a clear sense of what can be accomplished with these events , since substantial telescope resources are likely to be expended . to this end
, we approach this topic from another perspective .
we quantify the intrinsic ability of both point - mass and binary lens microlensing caustic crossings to recover the radial variation of the intensity of the star for any arbitrary wavelength , and hence , for any spectral line .
specifically , we calculate the fractional error in the recovered intensity profile as a function of the unlensed flux of the source , the duration of the measurements , the size of the telescope , the magnification of the event , and the spatial resolution of the recovered intensity profile .
this information will be useful to observers in making rational decisions about which events to follow , and what resources are required to address specific questions .
microlenses come mostly in two flavors : point lenses and binary lenses .
the magnification structure of these two types of lenses is essentially different , so we consider them separately . the flux of a point source being microlensed by a point lens , @xmath2 , can be expressed in terms of five parameters , @xmath3 , the unlensed flux of the source , @xmath4 , the flux of any unresolved background light , @xmath5 , the time of maximum magnification , @xmath6 , the impact parameter of the event , and @xmath7 , the time scale of the event .
these are related by @xmath8 , where @xmath9 is the magnification , which is itself a function of the parameters @xmath5 , @xmath7 , and @xmath6 , and is given by , @xmath10= { x^2 + 2 \over { x(x^2 + 4)^{1/2 } } } , \,\,\,\,\ , x(t)=\biggl[\beta^2 + { ( t - t_0)^2\over t_e^2}\biggr]^{1/2}. \eqno(2.1)\ ] ] here @xmath11 and @xmath6 are in units of the einstein ring radius , @xmath12 , given by , @xmath13^{1/2 } \over c^2 } , \eqno(2.2)\ ] ] where @xmath14 is the mass of the lens , and @xmath15 and @xmath16 are the distances to the lens and source respectively .
the time scale is related to the einstein ring radius by @xmath17 , where @xmath18 is the transverse speed of the lens relative to the observer - source line of sight .
note that for @xmath19 , @xmath20 , and that the magnification therefore diverges for a point source .
thus the point @xmath21 corresponds to the caustic for the point - lens case .
consider a source of radius @xmath22 .
the radius of this star , projected onto the lens plane , in units of @xmath12 is given by , @xmath23 for @xmath24 , @xmath25 , @xmath26 , and @xmath27 , @xmath28 .
for a source of this size , equation ( 2.1 ) remains valid for the majority of the event .
however , when the lens comes within @xmath29 of the source , i.e. , for events with @xmath30 , the magnification deviates from equation ( 2.1 ) , and the finite size of the source must be considered .
the time scale for this deviation is roughly the crossing time for the source , @xmath31 for typical bulge events . for these events ,
the source parameter @xmath32 enters into the calculation of the magnification .
the probability of such an event is @xmath33 , where @xmath34 is the average radius of the source stars being monitored .
if @xmath35 giant - source events were detected during a year , then the lens would transit the source for about 5 of them and these would exhibit finite - source deviations ( gould 1995b ) . the magnification of a point source being lensed by a binary depends on the same parameters as the point lens , @xmath5 , @xmath7 , and @xmath6 , along with the additional parameters @xmath36 , the separation of the lenses in units of @xmath12 , @xmath37 , the mass ratio of the lenses , and @xmath38 , the angle between the axis of the binary and the trajectory .
unfortunately , the dependence of the magnification on these parameters in the general case has no analytic form .
since the method of determining the magnification in the general case has been described elsewhere ( see , e.g. witt & mao 1995 ) , and since we are only interested in caustic crossings , we will employ a simplified formalism for these crossings . for nearly equal mass binary lenses ,
the caustics are composed of curved lines ( called folds ) whose extent is of order the einstein ring radius .
thus the probability of encountering a fold caustic during a binary - lens event is almost unity . for a point source , the excess magnification near
the fold caustic is approximately @xmath39 ( schneider , ehlers , & falco 1992 ) , where @xmath11 is the distance from the caustic , and the magnification diverges at @xmath21 ( at the caustic ) . for a finite source of radius @xmath32 , the magnification deviates from this form for distances @xmath40 , and the finite size of the source must be considered .
binary lenses also have cusps , points where two fold caustics merge , and the magnification structure near a fold is quite different from that near a cusp .
in particular , it can not be described by the same equations . while the probability of encountering a fold caustic is nearly unity ,
the probability of encountering a cusp is @xmath41 , where @xmath42 is the number of cusps . for typical binary lenses , @xmath43 , and
thus the probability is @xmath44 .
if @xmath35 events were discovered towards the bulge per year , and @xmath45 of these were binaries , we would expect only one cusp crossing per year , while we would expect @xmath46 events with fold crossings .
hence , we will consider only fold crossings here . before continuing to the next section , where we consider the magnification one expects when a finite source crosses the two types of caustics considered above
, we include a brief discussion concerning notation .
note that there are two characteristic scales in this problem , @xmath12 and @xmath32 , which are related by equation ( 2.3 ) .
either of these could be used as our fiducial scale .
for definiteness , we choose to scale all quantities by @xmath12 . for the remainder of the discussion
, we will also assume that @xmath47 is the time when the center of the source crosses ( or comes closest to ) the caustic .
for the point - mass case , this simply means setting @xmath48 .
finally , for the binary lens , the caustic crossings come in pairs . throughout
we will be referring to the second caustic crossing ( since it is this crossing that can be predicted , see 5 ) , when the source is moving from the inside of the caustic structure to the outside .
the magnification of an extended source is given by : @xmath49 where @xmath50 is the intensity profile of the source , @xmath51 is the magnification of a point source at @xmath52 , and the integral is over the area of the source . the numerator and denominator are respectively the lensed and unlensed flux of the source .
assuming that the intensity profile has azimuthal symmetry , and using polar coordinates , this becomes , @xmath53 where @xmath54 is the angle averaged magnification function , @xmath55 this definition is useful because the magnification geometry of the lens is entirely isolated in this function .
once the function has been calculated , it can be convolved with any source intensity profile to give the total magnification @xmath56 .
furthermore , it is the shape of this function that determines how well one can resolve the source , i.e. if @xmath57 is highly peaked at a particular value of @xmath58 , then the majority of the lensed flux will be coming from a small range of radii near @xmath58 , and therefore the lens is resolving the source . for both fold and point caustic crossings , the approximate form of the function @xmath57 can be calculated analytically . when a source crosses a fold caustic ,
two images appear or disappear . for typical total binary - lens mass ,
the size of the source , @xmath32 , is considerably smaller than the einstein ring radius of the lens , @xmath59 , and thus the magnification of the other three images changes very little as the source crosses the caustic . also , the curvature of the caustic is typically very small across the source . in this regime
, the magnification in the vicinity of the fold caustic can be approximated by ( schneider et al .
1992 ) , @xmath60 where @xmath61 is the total magnification of the three unrelated images , @xmath11 is the distance to the caustic , and @xmath62 describes the scale of the caustic . for
approximately equal mass binaries , @xmath63 . for @xmath64 ( outside the caustic ) , @xmath65 .
defining @xmath66 , where @xmath11 is now the distance from the center of the source to the caustic , and setting @xmath67 , the angle - averaged magnification function is , @xmath68 where , @xmath69&$|z|<1 $ \cr { 2\over \pi}(1+z)^{-1/2 } k\left [ \left({2\over 1+z}\right)^{1/2}\right ] & $ z>1 $ \cr } .
\eqno(2.9)\ ] ] here @xmath70 is the complete elliptic integral of the first kind .
figure 1 shows @xmath71 versus @xmath72 for @xmath73 , @xmath74 , and @xmath75 , @xmath32 , @xmath76 , @xmath77 , @xmath78 , @xmath79 , and @xmath80 . for @xmath81 ( when the source is entirely contained within the caustic ) , the gradient of the magnification across the face of the source is small , and furthermore @xmath71 has no maximum
this implies that these times are not useful for resolving the source . for @xmath82
, @xmath71 shows a sharp peak at @xmath83 , and thus the source is being resolved .
however , other annuli are being significantly magnified and are thus contributing significantly to the total light , and therefore the resolution will be degraded somewhat .
by contrast , for @xmath84 , and in the limit @xmath85 , only those annuli that have just crossed the caustic contribute to the total light .
it is therefore these times that are most useful for resolving the source . for a source of uniform brightness ,
the total magnification is ( schneider & weiss 1987 ) , @xmath86 where @xmath87 + 2z e\left[\left({1+z}\over 2\right)^{1/2}\right ] \right ) & $ |z|<1$\cr { 8 \over 3\pi}(1+z)^{1/2 } \left(z e\left[\left(2 \over 1+z \right)^{1/2}\right ] -(z-1)k\left[\left(2\over { 1+z}\right)^{1/2}\right]\right ) & $ z > 1$\cr } , \eqno(2.11)\ ] ] and @xmath88 is the complete elliptic integral of the second kind .
note that in equations ( 2.8 ) and ( 2.10 ) , the dependences on @xmath89 and @xmath58 are separable for @xmath73 .
this implies that , for a fixed value of @xmath89 , both @xmath57 and @xmath56 scale simply as @xmath90 .
we now turn to the point lens case .
when the separation between the lens and the source is much smaller than the einstein ring radius , @xmath91 , the magnification can be approximated by , @xmath92 the angle - averaged magnification function is then , @xmath93 where , @xmath94 , \eqno(2.14)\ ] ] and @xmath70 is the complete elliptic integral of the first kind .
figure 2 shows @xmath71 versus @xmath72 for @xmath74 , and trajectories with @xmath95 , for impact parameters @xmath96 ( solid curves ) and @xmath97 ( dashed curves ) , and @xmath98 , @xmath99 , @xmath100 , @xmath101 , @xmath102 , @xmath103 , and @xmath104 .
here @xmath105 achieves a local maximum whenever @xmath106 , implying that these radii are being partially resolved at these times .
unfortunately , for any impact parameter @xmath107 , there is a range of source radii at which @xmath105 is never at a maximum , those for which @xmath108 .
these radii are not resolved during the caustic crossing , and thus very little information about the radial intensity profile will be gained for this range of source radii . the total magnification for a source of uniform brightness is ( schneider et al .
1992 ) , @xmath109 where @xmath110 & $ z\ge 1$\cr } .
\eqno(2.16)\ ] ] as in the fold case , the dependences on @xmath89 and @xmath58 in equations ( 2.13 ) and ( 2.15 ) are separable .
thus , for a fixed value of @xmath89 , @xmath105 and @xmath56 scale as @xmath111 .
the flux measured from a star of finite size being lensed is @xmath112 , where the general form of @xmath56 is given in equation ( 2.4 ) . for most caustic crossing events ,
the magnification is high , and @xmath113 , thus we will henceforth use the approximation that @xmath114 . consider a star divided up into @xmath115 radial bins ,
each located at @xmath116 . defining @xmath117 as the width of this bin , the flux of the star is given by , @xmath118 where @xmath119 is the angle averaged magnification in bin @xmath120 .
for the case of a large number of bins , or @xmath121 , the magnification in each bin is well approximated by equations ( 2.8 ) and ( 2.13 ) .
for a small number of bins , however , these forms are not good approximations , and one should use the total magnification in each bin , which can be obtained by either integrating equations ( 2.8 ) and ( 2.13 ) over the radii of the bin , or , equivalently , by using equations ( 2.10 ) and ( 2.15 ) , and subtracting the flux within the inner radius of the bin from the flux within the outer radius of the bin , and dividing by the area of the annulus . in practice ,
the latter is simpler .
the parameters one would like to recover are @xmath122 , the mean intensity in bin @xmath120 for a certain wavelength . in order to do this , however , one must first know @xmath123 , which implies knowing the parameters @xmath32 , @xmath6 , @xmath7 , and @xmath5 for the point - lens case , and the additional parameters @xmath38 , @xmath36 , and @xmath37 for the binary case . in principle , one could determine these parameters entirely within the context of the spectrophotometric measurements .
however , variations in the seeing and extinction make this difficult .
fortunately , this is not a significant hindrance because independent broad - band measurements can be used to constrain these parameters .
since the follow - up collaborations already monitor these events , and , in one case , have already used broad - band measurements to constrain these parameters _ and _ measure the limb - darkening of the source star ( albrow et al .
1998b ) , this should not be a problem in practice , although it does imply that the spectroscopic and photometric follow - up collaborations should closely coordinate their efforts .
now suppose that a series of spectrophotometric measurements @xmath124 are made at times @xmath125 , with uncertainties @xmath126 , and these measurements are fitted to equation ( 3.1 ) .
the parameters of this fit are @xmath127 , and the covariance matrix of the errors in these parameters is given by @xmath128 , where @xmath129 and @xmath130 .
the variances in the parameters are then the diagonal elements of @xmath128 , and thus the fractional error in each parameter is @xmath131 . assuming photon - limited precision , the fractional errors scale as the square root of the total number of photons received , which in turn scales as @xmath132 . from equation ( 2.10 ) ,
@xmath133 , for the point caustic , and from equation ( 2.15 ) , @xmath134 for the fold caustic .
let @xmath135 be the rate at which a telescope receives photons from the unmagnified star in a certain spectral range , and assume that measurements are made continuously during a source - radius crossing time @xmath136 .
then the square root of the total number of photons scales as @xmath137 , where @xmath138 for the point caustic and @xmath139 for the fold caustic .
the fractional errors will also depend on the spatial resolution of the recovered intensity profile , @xmath115 .
assuming poisson statistics and no correlations between the parameters @xmath127 , one would expect the fractional errors in @xmath127 to scale simply as @xmath140 .
however , we find that the parameters @xmath127 are correlated , and the errors in fact scale approximately as @xmath141 .
we justify this assertion in the next section . taking these scalings out of @xmath142
, we have @xmath143 where @xmath144 is the normalized fractional error for @xmath145 , @xmath74 , @xmath146 .
the specific form of @xmath147 will depend on the details of the caustic encounter and will therefore vary on an event - by - event basis . in the next section , we consider specific cases , and
evaluate @xmath144 directly .
in order to justify the assertion that the errors scale as @xmath141 , we consider a specific example for each caustic case .
for the point case , we consider an event with impact parameter @xmath96 , and we assume that this event is observed from @xmath148 to @xmath149 , i.e. for one source crossing time .
we then evaluate the normalized error , @xmath147 for this event as a function @xmath116 , the radius of the bin , for different values of @xmath115 .
this is shown in figure 3a , for @xmath150 . if the errors scaled exactly as @xmath141 , these curves would be indistinguishable .
in fact , the normalized errors decrease slightly for larger @xmath115 , indicating that the errors do not increase quite as fast as @xmath141 , although this depends on @xmath116 . for the fold case ,
we assume that the event is observed from @xmath151 when the center crosses the caustic until @xmath152 when the source moves completely outside the caustic . in figure 3b
, we show @xmath144 for @xmath150 . here
the scaling is nearly perfect .
we will henceforth assume that the scaling is perfect for both cases , and therefore that @xmath153 has no dependence on @xmath115 . in order to evaluate the fractional errors explicitly , we adopt some fiducial parameters . for a giant in the bulge , and a typical microlensing event , @xmath154 , and
we assume that observations are taken continuously for a time , @xmath155 .
we use @xmath156 for our fiducial number of radial bins , and , for the moment , we assume a uniformly bright source , as this is the most general way to quantify the error in a model - independent way . in the next section ,
we consider a source with limb darkening .
the fractional error in the intensity profile for these parameters is , @xmath157 where @xmath138 for the point caustic and @xmath139 for the fold caustic , and @xmath158 is approximately the flux of photons per spectral resolution element from a star with @xmath159 that can be acquired with a 2 m telescope and a spectrograph with @xmath160 resolution . in figure
4a we show @xmath161 as a function of @xmath116 for the point - lens case , for @xmath148 to @xmath32 , and for four different impact parameters , @xmath162 , and @xmath163 ( solid , dotted , short dashed and long dashed lines ) .
it is clear from this diagram , that for @xmath164 , the fractional error is small , @xmath165 .
one could , in principle , improve the fractional error by a factor @xmath166 by observing from @xmath167 , i.e. for two source crossings . however , this will prove difficult in practice ( see 5 ) . while the errors are small for @xmath164 , there is almost no information for radii @xmath168 , as expected , since the lens is not passing over this region of the star . in figure 4b
we show @xmath161 as a function of @xmath116 for the fold caustic with @xmath73 , and four different cases .
the short dashed line corresponds to measurements taken from @xmath169 to @xmath21 , one crossing time from when the edge of the source first crosses the caustic to when the source is exactly centered on the caustic , i.e. , the first half of the caustic crossing .
the solid line is for @xmath21 to @xmath152 ( the second half of the caustic crossing ) ; the dotted line is for @xmath169 to @xmath152 ( the entire caustic crossing ) , and the long - dashed line is for @xmath170 to @xmath152 . from this figure
one learns three things .
first , the errors are reasonable , @xmath171 for typical parameters .
second , times immediately after the center of the source crosses the caustic are the most crucial for recovering the intensity profile accurately . finally , it is only the times just before until just after the caustic crossing that are useful for resolving the source .
we also show the errors for the case that @xmath172 ( dashed - dot line ) . clearly , the magnification of the images not associated with the caustic does not greatly affect the resultant errors . in this section
we examine the effect limb darkening has on the fractional error in the recovered intensity profile .
we can anticipate that limb darkening will serve to increase the errors in the outer annuli relative the constant surface - brightness case because there will be fewer photons coming from outer annuli , and thus the errors will be larger . for the inner annuli , the fractional errors will be smaller compared to the constant surface - brightness case , because the source is , in essence , more compact , and thus the net magnification will be larger , and therefore the errors smaller . to quantify this effect
, we apply the same formalism as in 4.3 , except that we adopt the following parameterization of the surface brightness as a function of radius , @xmath173 where @xmath174 and @xmath175 are the limb - darkening coefficients .
we adopt the coefficients for a cool ( 4500 k ) giant ( @xmath176 ) of solar metallicity from manduca , bell , & gustafsson ( 1977 ) and manduca ( 1979 ) .
these are @xmath177 and @xmath178 for the v , i and k bands , respectively .
since stars are less limb darkened in the infrared , the results for the k band will be the most similar to the uniform surface brightness case , while those for v will be least similar . the results for the same fiducial parameters used in 4.2 ( @xmath179 , @xmath180 , @xmath181 , @xmath156 ) are shown in figure 5 along with the uniform surface - brightness case .
for the point - caustic case [ panel ( a ) ] , we have used a trajectory with @xmath96 and assumed that observations are taken for @xmath182 . for larger values of @xmath6
, the fractional errors will deviate more dramatically from the uniform surface brightness case , since for large values of @xmath6 , only the outer annuli are being effectively resolved , and it is these annuli that are affected most by limb darkening .
for the fold caustic case [ panel ( b ) ] , we have assumed @xmath183 . as expected , for both the point - caustic and fold - caustic cases ,
the fractional errors are somewhat larger than for the uniform source at larger radii , and somewhat smaller at smaller radii .
also , the differences between the uniform source and limb - darkened source decrease for longer wavelengths . in general
, however , for both the point and fold caustics , the differences between uniform and limb - darkened sources is modest . for fold caustics
the effect is @xmath184 over the entire range of radii . for point caustics
the effect is small over most of the star but the error can more than double at the very limb of the star when observed in @xmath185 band , the bluest color considered here .
while both point - mass lenses and binary lenses can in principle be used to resolve the two - dimensional ( radial + spectral ) profile of a star , binary lenses are substantially easier to use .
first , for a binary one always has warning of the second caustic crossing . when the source crosses the caustic the first time it is suddenly magnified by a factor @xmath186 and hence is easily recognized .
the second crossing can then be expected in several days to several weeks .
intensive photometric monitoring ( now routinely undertaken by planet and gman ) can then be analyzed to make a more precise prediction . from figure 4b
it is clear that the most useful portion of the second caustic crossing is the final @xmath187 of the time that the source actually straddles the caustic .
the onset of this optimal period can be judged extremely accurately if photometric monitoring is proceeding simultaneously , and reasonably accurately even one day in advance . by contrast
, there is no way to guarantee a priori that a point - mass caustic crossing will occur because one does not know the size of the einstein ring projected onto the source plane beforehand , and hence one does not know @xmath32 . using optical photometry alone , one can `` predict '' a source crossing only at about the time it begins . using optical / infrared photometry ,
it could be predicted at @xmath188 ( gould & welch 1995 ) , but this would leave only a few hours warning . second , fold caustics
generically provide information about the entire radial profile of the star while point caustics provide information only for annuli of the star that are greater than the impact parameter , @xmath6 .
see figure 4 .
again , it is virtually impossible to predict in advance for which events @xmath189 ( and hence for which events one can resolve essentially the whole star ) , although once the peak occurred , these events could be recognized provided that the star was being monitored photometrically . in any event , of all point - caustic transits , only a fraction @xmath190
have impact parameters smaller than @xmath6 .
third , binary - lens events are probably more common than point - mass caustic crossings . the fraction of events with binary - caustic crossings has not yet been established empirically , but 5% appears to be a plausible estimate .
the fraction of point - mass caustic crossings is @xmath34 , but by the argument of the previous paragraph , only about half of these are really useful .
the mean radius of a giant is @xmath191 ( gould 1995b ) , about 2.2 times larger than the fiducial value used in equation ( 4.1 ) .
thus , the fraction of events with useful point - mass caustic crossings is @xmath192 . on the other hand ,
as shown by figure 4 , over the range probed by the point lens ( @xmath108 ) , the point - lens errors are less than half as large as those of the binary .
this advantage diminishes as @xmath193 for larger stars , but is still substantial for most giants .
the problem of recognizing events with @xmath194 sufficiently early to permit spectroscopic monitoring is formidable .
however , if they can be monitored beginning at their peak , they are the best events to follow .
|
microlensing is a powerful tool for studying stellar atmospheres because as the source crosses regions of formally infinite magnification ( caustics ) the surface of the star is resolved , thereby allowing one to measure the radial intensity profile , both photometrically and spectroscopically .
however , caustic crossing events are relatively rare , and monitoring them requires intensive application of telescope resources .
it is therefore essential that the observational parameters needed to accurately measure the intensity profile are quantified .
we calculate the expected errors in the recovered radial intensity profile as a function of the unlensed flux , source radius , spatial resolution the recovered intensity profile , and caustic crossing time for the two principle types of caustics : point - mass and binary lenses .
we demonstrate that for both cases there exist simple scaling relations between these parameters and the resultant errors .
we find that the error as a function of the spatial resolution of the recovered profile , parameterized by the number of radial bins , increases as @xmath0 , considerably faster than the naive @xmath1 expectation .
finally , we discuss the relative advantages of binary caustic - crossing events and point - lens events .
binary events are more common , easier to plan for , and provide more homogeneous information about the stellar atmosphere .
however , a sub - class of point - mass events with low impact parameters can provide dramatically more information provided that they can be recognized in time to initiate observations .
# 1#2#1 10^#2 = + 0.2 in = 6.0 in = 8.0 in = 0.10 in = 0.5 in = 0.5 in # 1 = 0.4 in = 1 # 1 # 1 ccc 1 & 1 & 1 + 1 & 1 & 1 + 4 & 3 & 8 submitted to _ the astrophysical journal _ : feb 16 , 1998 preprint : osu - ta-2/98
|
molecular photonic wires are of considerable interest as biomimetic models for photosynthetic light harvesting systems and as molecular devices @xcite .
lindsey et .
al . have been the first to demonstrate directed energy transfer through a porphyrin - based wire after selective excitation of a boron - dipyrrin donor @xcite .
subsequently , they described a number of different systems on the basis of multiporphyrin arrays @xcite including , for instance , shape persistent cyclic architectures @xcite or optoelectronic gates @xcite ( for a review , see also ref .
the energy transfer in these structures is dominated by a through - bond mechanism with some dependence on the nature of the bridge linking the chromophores @xcite .
a different design strategy has been followed by sauer and coworkers who have used the backbone of double - stranded dna as a scaffold for positioning highly fluorescent chromophores @xcite ( see also ref .
@xcite ) . here
, frster type dipole - dipole interaction driven transfer is the responsible mechanism for energy transfer with efficiencies close to 100% over a range of @xmath014 nm . the utilization of molecular wires in optoelectronic devices requires to have at hand a means for switching the energy transfer dynamics .
various methods such as optical excitation using uv radiation @xcite , electrochemical oxidation @xcite or changing the ph value of the medium @xcite have been suggested ( for an overview , see also refs .
this includes even the realization of photochemical logic gates as reported in ref .
recently , we have proposed another route to switch excitation energy transfer which could operate in the ultrafast regime of a few picoseconds @xcite .
the idea builds on the observation that the direction of the electronic transition dipole moment , which is responsible for the coupling between chromophores in the frster mechanism @xcite , depends on the positions of the two h atoms in the porphyrin like building blocks of molecular wires .
thus switching between the ground state tautomers , i.e. triggering the double proton transfer ( dpt ) by means of an appropriately designed laser field in the infrared ( ir ) domain enables one to modify the strength of the frster transfer coupling .
the aim of the present contribution is to extend the model study of ref .
@xcite towards a specific molecular system and to show exemplarily to what extent laser - triggered dpt can influence the frster transfer coupling .
specifically , we will focus on porphycene like structures , which have rather strong hydrogen bonds ( hbs ) whose properties are widely adjustable by the structural design @xcite . for the laser - driven dpt
the shape of the potential energy landscape is rather important . besides energy barriers which need to be surmounted by excitation with a few ir photons , reactants and products
should be energetically different to prevent an efficient backreaction via tunneling .
asymmetrically substituted porphycenes seem to meet these requirements .
below we will consider 9-actetoxy-2,7,12,17-tetra - n - propylporphycene ( * 3 * ) whose spectroscopy had been studied by waluk and coworkers @xcite .
thus , besides the specific goal of demonstrating a laser - driven switch for molecular photonic wires , our quantum chemical studies of the ground state properties and the excitation spectrum will shed some light on the effect of symmetric ( * 1 * ) and asymmetric ( * 3 * ) substitution of the porphycene parent compound ( * 1 * ) and hence will provide the theoretical basis for the assignment of the absorption spectrum .
the paper is organized as follows : section [ sec : methods ] gives a brief summary of the quantum chemical and quantum dynamical methods .
the results section [ sec : res ] starts with the discussion of the geometries and energetics of the stationary points of porphycene and its derivatives ( see fig .
[ fig:1 ] ) . based on this information a two - dimensional ( 2d ) model hamiltonian
is constructed in section [ sec : ham ] which comprises coordinates of concerted and sequential dpt .
subsequently , the excitation spectrum is discussed in terms of the involved molecular orbitals and transition dipole moments .
the laser driven dpt dynamics will be presented in section [ sec : control ] and the final section [ sec : concl ] concludes with a summary .
geometry optimization for the electronic ground state has been performed using density functional theory ( dft ) employing the b3lyp hybrid functional together with a 6 - 31+g(d , p ) basis set .
stationary points have been validated by means of frequency calculations .
excitation energies are obtained from time - dependent dft ( tddft ) calculations at the optimized ground state geometries ( vertical excitation ) and using the same functional and basis set .
all quantum chemical calculations are performed with the gaussian 03 suite of programs @xcite .
the two - dimensional time - dependent schrdinger equation has been solved using the multiconfiguration time - dependent hartree ( mctdh ) approach @xcite as implemented in the heidelberg program package @xcite .
the grid was chosen according to the harmonic oscillator discrete variable representation ( 64 points within [ -2.5:2.5]a@xmath1 ) .
the actual propagation is performed using the variable mean field scheme in combination with a 6-th order adams - bashforth - moulton integrator . using 20 single particle functions per degree of freedom
the largest natural orbital populations have been typically on the order of @xmath2 . selected eigenstates of the time - independent hamiltonian , @xmath3 ,
are obtained by improved relaxation @xcite .
ground state optimized structures of * 2 * show that this molecule like * 1 * is planar in accord with ref .
asymmetric substitution , however , causes * 3 * to become nonplanar with respect to the acetoxy group .
overall * 3 * has two global trans minima , tr1 and tr2 , two local cis minima , cis1 and cis2 , four local maxima which correspond to first order saddle points , ts11 to ts22 , and one global maximum being a second order saddle point , sp2 .
all these structures are shown in fig .
[ fig:3struct ] and the energetics is compiled in tab .
[ tab : energies ] . in terms of the double proton transfer
, sp2 is passed in the concerted mechanism , while pathways like tr1-ts11-cis1-ts21-tr2 are followed during sequential transfer . finally , we note from the dipole moments given in tab . [
tab : energies ] that * 3 * is considerably more polar than * 2*. the energetic difference between trans and cis , which is 804 for * 1 * , is largely unaffected by the symmetric substitution in * 2*. the same holds true for the second order transition state sp2 , i.e. the barrier height for concerted dpt is about 2260 .
the barrier for stepwise proton transfer which is 1637 for * 1 * decreases slightly to 1514 in * 2*. upon asymmetric substitution with an acetoxy group in the 9 position in * 3 * most notably in comparison with * 2 * is the asymmetry of the two trans tautomers which amounts to 38 .
this rather small value is in good accord with the experimental finding of indistinguishable tautomers @xcite .
further , the sp2 transition state is lowered by about 4% and the first order saddle points and cis minima are changed in a way such as to make a stepwise dpt via cis1 energetically more preferable .
[ tp ] .energies and absolute value of dipole moments of stationary points for the tautomerism of the molecules shown in fig .
[ fig:1 ] at the b3lyp/6 - 31+g(d , p ) level of theory ( values for * 1 * from ref .
@xcite ) . [ cols="^,^,>,^",options="header " , ] in the following we will discuss the , which contributes to the @xmath4 absorption band .
results of tddft calculations are compared with experimental data in tab . 3 for the different compounds .
first , we notice that tddft gives excitation energies which exceed the experimental ones by about 1500 - 2100 .
most of this difference must be attributed to the method itself ( see , e.g. , systematic study in ref .
further deviations might be due to the fact that the experimental values correspond to the 0 - 0 transitions whereas the calculations are performed for a vertical excitation . comparing the experimental spectra of porphycences in gas and condensed phases the environmental effects ( e.g. , a nitrogen matrix in ref .
@xcite ) is negligible @xcite .
let us focus on the tautomerization of * 3*. as discussed in section [ sec : geom ] the energetic asymmetry in the electronic ground state is rather small . the measured splitting in the transition , however , gave evidence for a considerable asymmetry in the excited states of 285 @xcite .
the assignment of the most stable tautomer in the s@xmath5 excited state has been based on a semiempirical zindo / s calculation .
considering tab .
3 we notice that the agreement of the splitting in the transitions ( 278 ) is rather good thus giving strong support for the experimental conclusions in ref . @xcite .
the transitions are of @xmath6 type with participation of the @xmath7 ( homo , homo-1 ) and @xmath8 ( lumo , lumo+1 ) molecular orbitals ( mos ) .
these orbitals are shown for the different molecules in fig .
[ fig : mo ] and orbital coefficients are compiled in tab .
3 . the main contributions to the transition in * 1 * and *
2 * is of homo-1 @xmath9 lumo type ( see also ref .
the same holds true for the less stable tautomer tr2 of * 3*. interestingly , in tautomer tr1 of * 3 * this transition is of homo @xmath9 lumo type .
comparison of the mos of tr1 and tr2 shows that the order of homo and homo-1 is _ reversed _ upon tautomerization
. a closer look at the orbitals reveals the effect of symmetric and asymmetric substitutions .
going from * 1 * to * 2 * some electron density appears at those alkyl chains which are opposite to the position of the hydrogen bonded h atom .
overall , however , the mos are rather similar , i.e. in the homo there is a @xmath10-orbital at those nitrogens , which are bonded to the central hydrogens whereas there is no electron density at the `` free '' nitrogen atoms . in the homo-1
this situation is reversed .
asymmetric substitution causes a disturbance of the porphycene @xmath7-system by the acetoxy group with the effect being different for the two tautomers .
tr2 of * 3 * resembles the homo in * 1 * and * 2 * with some electron density at the acetoxy group . in tr1 of * 3 * the presence of the latter electron density renders the tr2-homo-1 to become more stable than the tr2-homo .
the directions of the transition dipole moments for the transitions of the different molecules are compiled in fig .
[ fig : dipole ] . comparing * 1 * and * 2
* we notice that symmetric substitution has almost no effect on the direction of the transition dipole moments .
furthermore , both cases show the well - known 90 degree change of the transition dipole moment upon tautomerization .
the presence of the acetoxy group in * 3 * , however , changes this picture due to the polarity introduced by the oxygen atoms .
most notably is the fact that the tautomerization causes a tilt of the transition dipole moment by about 50 degrees only .
let us assume that * 3 * is built into a molecular photonic wire .
the frster energy transfer coupling between different molecular subunits of the wire called donor ( d ) and acceptor ( a ) is given by @xcite @xmath11 here , @xmath12 is the magnitude of the transition dipole moment for the transition and @xmath13 is the distance between donor and acceptor .
most important for the present discussion is the orientational factor given by @xmath14 where @xmath15 and @xmath16 are the unit vectors pointing in the directions of the transition dipole moment of d / a and the center to center da distance , respectively . in order to estimate the effect of dpt on the frster transfer coupling and in particular on the orientational factor
we suppose that * 3 * is built into the wire such that the neighboring unit has a transition dipole moment like structure * 1*. this could be the case , for instance , for the widely used porphyrins .
let us further assume that @xmath16 points along the @xmath17-direction . in this case we obtain @xmath18(tr1)=1.43 and @xmath18(tr2)=0.60 .
in other words the transfer coupling changes by about 60% as a consequence of the tautomerization .
the green line corresponds to the energy of the sp2 barrier .
the vibrational eigenstates have been obtained using the improved relaxation scheme within mctdh @xcite .
the pathway of laser - driven pump - dump control is marked by vertical arrows .
the density plots cover the range @xmath19a_0 $ ] along the vertical @xmath20 and the horizontal @xmath21 axes . ] in the following we will investigate the possibility to control the dpt in * 3 * using a simple two pulse pump - dump scheme involving an intermediate state which is energetically above the reaction barrier for concerted dpt ( cf . fig .
[ fig : levels ] ) .
the initial state is localized mostly in the tr1 well , while the final state lies mostly in the tr2 well .
due to the low barrier and the small asymmetry both states have finite probabilities in the respective other wells .
the goal is to find an electric field of the form given by eq .
( [ eq : laser ] ) , for which the first ( pump ) pulse populates the intermediate state and the second ( dump ) pulse triggers a transition from this intermediate to the reactant state .
this scheme operates in complete analogy to other isomerization reactions studied previously , such as the single h - atom transfer in malonaldehyde @xcite or the cope rearrangement in semibullvalenes @xcite .
the parameters for the present case of dpt were chosen as follows : first , the field strength is fixed at a certain value and the pulse length is increased starting from 800 fs until the intermediate state population reaches a maximum .
for this pulse length the field amplitude is changed around its starting value to check whether the intermediate state population can be increased .
the carrier frequency is chosen to correspond to the respective transition frequency .
the same procedure is followed independently for the dump pulse by letting it start immediately after the pump pulse .
this gave a total pulse duration of 5 ps during which the population of the intermediate state showed a plateau , which allowed us to shift the dump pulse such that pump and dump pulse partly overlap ( cf . fig .
[ fig : pdv4 ] ) .
the progress of the laser - driven reaction can be followed by calculating the populations of vibrational eigenstates of the model hamiltonian . a representation of the energy level scheme is given in fig .
[ fig : levels ] which also shows the probability density of those states which dominate the dynamics .
the chosen above - barrier transition state involves an excitation of the asymmetric stretching coordinate @xmath20 which triggers the reaction to occur mostly along the stepwise pathway ( see also ref .
furthermore , it is seen that the probability density has a higher weight in that part of the potential where @xmath22 , implying that there is a preference for the cis1 minimum to be visited during the dynamics . according to our previous study @xcite
this is due to the higher stability of cis1 as compared with cis2 .
finally , we note that there are several states in the region of the chosen intermediate state @xmath23 , thus state selective population will be more easily achieved using long pulses as compared , e.g. , with the nh - vibrational stretching frequency . the optimized field and selected state populations
are shown in fig .
[ fig : pdv4 ] .
the pump pulse excites the system from the ground state ground state , @xmath24 , to the intermediate state @xmath23 , while the dump pulse de - excites this population to the target state @xmath25 .
the maximum intermediate state population is 91 % , reached around 3.4 ps . at this time
, however , the dump pulse is already transferring population from @xmath23 to the target state @xmath25 .
the overall duration of the laser - induced dpt process is 4.5 ps and the final population of the target state is 95 % .
two minor features seen in fig . [ fig : pdv4 ] deserve a comment . around 2.4 ps
there is some transient population of the target state already due to the pump pulse .
later on around 3.6 ps the dump pulse also is seen to further depopulate the initial ground state .
both effects can be traced to the fact that the energetic separation between the two transitions is only 56 , i.e. they can be simultaneously addressed by the laser pulse spectra .
me@xmath26/ea@xmath1 ) , @xmath27 @xmath28 , @xmath29 fs , and @xmath30 fs ; @xmath31 me@xmath26/ea@xmath1 ) , @xmath32 @xmath28 , @xmath33 fs , and @xmath34 fs ) .
lower panel : population dynamics of the states @xmath24 , @xmath25 , and @xmath23 ]
in summary , we have investigated the possibility to use ir laser - controlled double proton transfer in the electronic ground state of asymmetrically substituted porphycene as a way of influencing the direction of the transition dipole moment .
this gives a means for changing the dipole - dipole interaction between neighboring chromophores in a molecular photonic wire .
the estimated effect amounts to a 60% change of the transfer coupling which could be used to design a molecular switch operating on a time scale of a few picoseconds and being , in principle , reversible . needless to say that this proof - of - principle study calls for several extensions .
first , the two - dimensional model of double proton transfer needs to be extended to account for heavy atom motions of the scaffold , e.g. , by using the cartesian reaction surface approach @xcite .
indeed waluk and coworkers showed that the coupling to this type of motions may have a strong influence on the dynamics @xcite .
it is to be expected that intramolecular vibrational energy redistribution is effective on the time scale of a few picoseconds covered in the present study , i.e. either substantially shorter pulses have to be used or more sophisticated pulse forms need to be devised , which , e.g. , make use of the promoting character of certain scaffold modes .
second , a suitable design for a molecular photonic wire containing a dpt switching unit has to be found .
this implies not only a favorable orientation of the transition dipoles of neighboring units but also the dominance of through space frster transfer over through bond energy transfer .
g. sanchez - mosteiro , e.m.h.p . van dijk , j. hernando , m. heilemann , p. tinnefeld , m. sauer , f. koberlin , m. patting , m. wahl , r. erdmann , n.f .
van hulst , m.f .
garcia - parajo , j. phys .
b * 110 * , 26349 ( 2006 )
|
electronic excitation energy transfer along a molecular wire depends on the relative orientation of the electronic transition dipole moments of neighboring chromophores . in porphycenes
this orientation is changed upon double proton transfer in the electronic ground state .
we explore the possibility to trigger such a double proton transfer reaction by means of an infrared pump - dump laser control scheme . to this end a quantum chemical characterization of an asymmetrically substituted porphycene is performed using density functional theory .
ground state geometries , the topology of the potential energy surface for double proton transfer , and transition energies are compared with the parent compound porphycene and a symmetric derivative . employing a simple two - dimensional model for the double proton transfer , which incorporates sequential and concerted motions , quantum dynamics simulations of the laser driven dynamics
are performed which demonstrate tautomerization control .
based on the orientation of the transition dipole moments this tautomerization may lead to an estimated change in the frster transfer coupling of about 60% .
|
strong bending of dna plays a critical biological role in genome maintenance and regulation , most famously evident in dna - protein complexes such as those formed in conjunction with the lac repressor @xcite , the tata binding protein @xcite and the nucleosome @xcite .
experiments have confirmed that these large bending fluctuations are not an artefact of dna - protein co - crystallization , and are observed in a variety of sensitive solution - phase assays @xcite . in the regime
where dna is not strongly bent , worm - like chain ( wlc ) models @xcite that treat dna as a semi - flexible rod are sufficient to describe its configurational behaviour @xcite .
it has been claimed , however , that strongly bent configurations can be reached much more easily than would be predicted by the wlc as parameterized to small fluctuations . in particular , deviations from wlc behaviour have been reported in experiments on cyclization @xcite , multimerization @xcite , in dna minicircles @xcite , in a stressed ring system @xcite and a molecular vice @xcite .
much of the evidence for super - wlc flexibility , has , however , been contested .
for example , in dna cyclization assays , cyclized molecules with @xmath0 base pairs ( bp ) @xcite were reported to be orders of magnitude more likely than predicted by the wlc model as normally parameterized .
however , other authors have criticized the interpretation of the experimental results and have suggested that non - wlc behavior only begins at shorter lengths @xcite .
similarly , apparent evidence of high flexibility of dna on short length scales obtained via afm @xcite has since been attributed to artefacts of the curve - fitting algorithms @xcite . the wlc model , which assumes a harmonic cost on the local degree of bending , favours the distribution of bending strain throughout a dna molecule .
however , above a certain degree of local bending the wlc model may break down because it becomes favourable for dna systems to localize strain within small regions , allowing the remainder to relax and reducing the overall cost of adopting highly bent configurations .
a disruption of the b - dna structure ( including broken base pairs and/or stacking interactions ) , or `` kink '' , is a candidate motif that would allow such enhanced flexibility and localization of bending @xcite .
nevertheless , although there is experimental evidence for kinking in strongly bent configurations ( reviewed for example in reference @xcite ) , it is not clear whether all of the proposed violations of wlc behaviour in the experiments listed above can be explained in this way .
one reason for the lack of clarity is that a number of fundamental questions about the thermodynamic cost and basic physics associated with the formation of kinks remain open .
it is therefore important to develop a sound theoretical underpinning of the underlying physics .
doing so would enable a systematic comparison between distinct systems and help establish whether experimental results are self - consistent and also consistent with basic dna thermodynamics .
here we use oxdna @xcite , a coarse - grained model of dna , to address these questions .
oxdna incorporates a physical description of single- and double - stranded dna , including the thermodynamics of hybridization and basic mechanical properties such as persistence length and torsional modulus @xcite . as such
, it is an ideal model for exploring strong bending , highlighting general properties and providing evidence as to whether the reported observations of enhanced dna flexibility are consistent with kinking as implied by the known thermodynamic and mechanical properties of dna .
further , as the model is generally applicable , oxdna can be used to compare strong bending in a variety of distinct systems . here , in the first of a series of papers investigating the biophysics of highly bent dna , we consider a `` molecular vice '' @xcite .
in an accompanying paper we study cyclization @xcite , and in an upcoming future publication we consider minicircles @xcite . in particular , in this paper , we first investigate the general thermodynamics of the strongly bent regime of dna . using this insight
, we corroborate the results of recent `` molecular vice '' experiments @xcite , which provide one of the most direct probes of strong dna bending . specifically , we reproduce the experimental high - salt buckling transition and approximate critical buckling length .
we then directly probe the microscopic configurations responsible for this transition , validating the experimental interpretation of a kink - induced buckling transition .
in an elegant set of experiments , fields _ et al . _ deployed a `` molecular vice '' to probe the strong bending regime of dna , providing evidence of a buckling transition at high salt that can not be explained by a wlc model @xcite .
they infer kinking as the physical mechanism underlying this transition .
the system they devised consists of a dna hairpin and complement strand ( ) ; the hairpin has two regions , a loop composed of @xmath1 nucleotides and a predominantly poly(at ) stem composed of 49 base pairs ( 39 a
t base pairs adjacent to the loop and 10 terminal g c base pairs ) .
complement strands of different lengths are hybridized to the hairpin loop to form a duplex of @xmath2 base pairs .
the hairpin acts as a force clamp , imposing a bending stress on the duplex .
formation of a duplex within the loop may require part of the hairpin stem to be opened , which requires work against a salt - dependent unzipping force . stem unzipping is monitored via fret ( fluorescence resonance energy transfer ) , reporting the separation between the fluorophores attached to the first and last bases of the hairpin loop ( ) .
this setup is powerful because it allows the observation of stem unzipping as a function of both the length of the complement strand and loop size .
for example , at short @xmath3 , there is minimal stress on the system it is necessary neither for the duplex to bend nor the stem of the hairpin to unzip . at long @xmath3 ( comparable to @xmath1 )
, the system can not maintain full base - pairing without bending the duplex . at high salt fields
_ et al . _
observe for short loop lengths unzipping of the stem as @xmath3 approaches @xmath1 base pairs .
for some longer loop lengths , they then see a sharp re - zipping when @xmath3 gets even closer to @xmath1 base pairs .
fields _ et al .
_ argue that the high - salt behaviour can not be explained by the wlc , and infer the formation of a sharp kink in the duplex that allows re - zipping @xcite .
oxdna @xcite was developed to bridge the gap between all - atom molecular dynamics simulations and continuum or analytic approaches .
it has been employed successfully for many systems , reviewed in reference @xcite .
most importantly for this work , oxdna has previously been shown to reproduce a range of structural transitions in response to applied stress @xcite .
the philosophy of the model is to encode the well - understood and extensively characterized aspects of dna physics , such as its basic thermodynamic , mechanical and structural properties , into a model with physically - motivated interactions .
briefly , each nucleotide ( sugar , phosphate and base group ) is modelled as a rigid body with three interaction sites
. a complex set of anisotropic interactions between the nucleotides have been parametrized to reproduce structural , thermodynamic and mechanical properties of single- and double - stranded dna .
the model is parameterized to a monovalent salt concentration of , where the electrostatic interactions are strongly screened .
the thermodynamic , structural and dynamic properties of the model are fully characterized in references @xcite and @xcite .
these include basic mechanical properties such as persistence length and torsional rigidity .
the model has two parameterizations , average - base @xcite and sequence - dependent @xcite . in the average - base parameterization , the base pairing and stacking interactions of each base are identical ; whereas in the sequence - dependent parameterization @xcite , stacking and hydrogen - bonding interactions vary by base identity ( e.g. a t base pairs weaker than g c base pairs )
. we use the average - base parameterization to elucidate the fundamental thermodynamics of dna bending since this illustrates the fundamental physics more clearly .
sequence - dependent thermodynamics are used for a more direct comparison with the molecular vice experiments and to elucidate sequence effects on the free - energy profile for bending .
simulations to probe the free - energy cost of bending were performed with the virtual - move monte carlo ( vmmc ) algorithm ( the variant in the appendix of ref .
@xcite ) in the nvt ensemble at .
as the entire dna bending regime is of interest , including strongly bent states that are high in free energy , umbrella sampling @xcite was employed to efficiently sample the free - energy landscape of dna bending .
the end - to - end distance @xmath4 between the base interaction sites of the first and last bases along a strand was used as an order parameter for the umbrella sampling procedure , whereby the model hamiltonian is augmented with an empirical potential that is iteratively adjusted to yield a roughly uniform sampling of the relevant values of @xmath4 .
the statistical bias of this artificial potential is removed at the analysis stage .
further details of simulation procedure are given in supplementary .
to simplify interpretation , base pairing at the ends of the duplex were enforced through the umbrella potential to explicitly disallow `` fraying '' of duplexes .
fraying is a possible mechanism for relaxation of stress in bent systems @xcite , but we do not consider it in this section as we are interested in the bending of the main body of the duplex , not its ends .
we allow fraying in the molecular vice system , but , as discussed by fields _
et al . _
@xcite , it is explicitly set up to inhibit fraying of the @xmath3 base pairs .
results are essentially identical for any constraint in which nucleotides at either end of the duplex are constrained to be base paired ( supplementary ) ; in this article we quote results in which we constrain only the final base pairs .
imposing such a constraint allows us to sensibly define and measure the end - to - end distance between the centre of the helical axis , @xmath5 , as the first and last base pairs are guaranteed to be present .
error bars in reported data represent the standard error of the mean from 5 independent simulations .
simulations of the molecular vice were conducted with the sequence - dependent parameterization of oxdna using the vmmc algorithm in the nvt ensemble at .
error bars represent the standard error of the mean from 6 independent simulations .
these simulations were performed without any biasing through the use of umbrella potentials .
further details of simulation procedure are given in supplementary .
our study investigates the role of several structural defects , the detection and reporting of which rely on practical definitions .
namely , we require definitions of fraying , bubble formation and kinking .
the identification of these defects requires , in turn , a definition of a formed base pair . following previous work @xcite
, we define a base pair as formed if the hydrogen - bonding interaction between the two complementary bases is stronger than ( @xmath6 at @xmath7 ) .
we define fraying as the loss of base pairs at the duplex ends , while bubbles are defined as the opening of base pairs in otherwise continuous stretches away from the duplex ends .
a practical definition of a kink is less straightforward .
qualitatively , a kink is an area of pronounced curvature localized to a small region .
there is some diversity of opinion on the quantitative description and classification of kinks @xcite ; however , they have in common the breaking of base pairs and/or stacking . other requirements may include a change in the relative orientation of nucleotides . due to the ambiguity in the definition of a kink , we explore both structural and energetic definitions .
structurally , we define a kink using the relative orientation of consecutive nucleotides . in oxdna ,
the orientation of each nucleotide is unequivocally determined by two orthogonal unit vectors , the base - backbone vector and the base - normal vector .
the base - backbone vector connects the backbone and base interaction sites of each nucleotide .
the model is designed such that , in a relaxed duplex , the base - normal vector at index @xmath8 , @xmath9 , and at the consecutive index @xmath10 , @xmath11 , are approximately parallel ; that is , @xmath12 .
a kink is defined to be present if @xmath13 , a condition implying that the orientation of consecutive nucleotides differs by more than .
energetically , we define a kink as a disruption in stacking between adjacent base pairs , where the stacking potential energy is less than one - tenth its typical value in a duplex ( ) . for the oxdna model , the structural and energetic criteria for a kink behave very similarly when kinks are reasonably probable due to strong bending of the duplex , as shown in ( c ) .
in this regime , the detection of a kink is reasonably unambiguous and straightforward as favourable kinks are necessarily strongly bent , and both criteria do a reasonable job of identifying them . at larger end - to - end distances ,
the two criteria give distinct results , and caution should be used when interpreting the data ; a fuller discussion is provided in supplementary .
we start our discussion by characterising the basic thermodynamics of dna bending in oxdna .
we simulate a 30-bp dna duplex and report the free - energy profile @xmath14 as a function of the end - to - end distance @xmath5 ( ( b ) ) .
the free energy of duplex bending has two distinct regimes , indicated by a change in slope at @xmath15 .
the change in slope of the free - energy profile in ( b ) correlates with a crossover from continuously bent to kinked macrostates ( ( c ) ) .
representative oxdna structures for the continuously bent and kinked 30-bp duplex ( ( a ) ) reveal that duplex kinking typically induces a bubble at the kink ( further quantified in supplementary ) and that this kink is located near the centre of the duplex ( supplementary ) . and @xmath16 , respectively .
* ( b ) * free energy of bending @xmath17 for oxdna ( ) , with the decomposed free energies of the continuously bent ( ) and kinked ( ) states overlaid .
oxdna is compared to the wlc model approximation of becker _
et al . _
@xcite ( ) . the comparison is fit - free , using known oxdna parameters for persistence length ( ) and rise per base pair along the helical axis ( ) @xcite note that , rather than setting the minimum @xmath17 to be zero as elsewhere in the text , we normalize @xmath17 so that @xmath18 in both cases . * ( c ) * probability of kinking @xmath19 using energetic ( ) and structural ( ) criteria for a kink , which are based upon disruption of stacking and base orientation , respectively . ,
width=317 ] in ( b ) we also compare oxdna results to the wlc model ; specifically to the analytic approximation to the wlc end - to - end distance developed by becker _
et al . _
this approach provides a good description of wlc behaviour across the near - rigid , semiflexible and flexible regimes .
as expected , agreement between the wlc ( when using the base pair rise and persistence length appropriate to oxdna ) and oxdna is excellent for intermediate @xmath5 .
the deviations most pertinent to this paper arise at @xmath20 where the oxdna model develops a kink that allows the system to lower its free energy compared to the wlc model .
deviations also arise at longer @xmath5 near the dna contour length @xmath21 , because the wlc neglects the finite intrinsic extensibility of dna . additionally ,
the wlc does not account for the finite thickness of dna , which causes excluded volume effects at very short @xmath5 , reported by oxdna as a sharp rise in @xmath22 at @xmath23 . ,
the duplex dna bending free energy , as a function of reduced end - to - end distance @xmath24 .
error bars are comparable to the line thickness . *
( b ) * length dependence of duplex dna kinking .
the structural criterion for kinking is shown ; results using the energetic criterion are very similar in the kink transition region .
, width=317 ] according to the wlc model , the enthalpic cost of bending a duplex to some fraction of its contour length decreases with increasing duplex length .
by contrast , there is no obvious reason why the free - energetic cost of forming a kink at the centre of the duplex should depend strongly on duplex length .
we therefore expect that , as duplex length is increased , the dna will need to be relatively more bent ( smaller @xmath24 ) to favour the formation of kinks . for sufficiently long dna , we expect to see unkinked duplexes even for small @xmath24 .
these expectations are confirmed in ; long duplexes ( @xmath25 ) do not show any substantial kinking at any @xmath5 , and for shorter duplexes the length - dependence of the kink transition is as expected . the re - entrant behavior where @xmath19 drops again at extremely small @xmath24 arises from a change in geometry that is enforced by excluded volume effects ; for a detailed discussion see supplementary .
we next consider the question of what is the free - energy cost of kink formation , a question which turns out not to be so easy to answer unambiguously .
one natural approach is to define a free - energy cost @xmath26 that corresponds to that at the middle of the kink transition , e.g. @xmath27 for the 30-bp duplex in .
this quantity is essentially a measure of work that must be done on the duplex by pulling its ends together before kink formation occurs .
this numerical value is comparable to the estimate of fields _ et al .
_ for `` @xmath28 '' @xcite , the bending stress released upon kinking for a similar system ( in that kink formation is driven by pulling two ends of a duplex together )
. however , free energies must be carefully defined if quantitative comparisons are to be made . in particular ,
neither @xmath26 nor @xmath28 is a generic free - energy cost of kink formation that can be directly compared with other systems in which kink transitions may occur .
furthermore , even when the bending stress is applied in the same way , @xmath29 is not a constant , but , as we show in supplementary for the current system , is a function of duplex length ( because the degree of bending at the kink transition depends on length ) .
one might instead define a quantity @xmath30 through the probability of spontaneous kink formation ( in the absence of stress ) at a specific location in the duplex .
@xmath30 would in principle be directly measurable from kinking probabilities of relaxed duplexes .
although in principle this would be expected to be a generic ( sequence - dependent ) quantity , its value will be strongly dependent on the criteria used to identify kinks ( supplementary ) .
this is in sharp contrast to @xmath29 , which is almost identical for both criteria used to identify kinking in this work . in the presence of sharp bending ,
the overwhelming majority of disruptions to the basic duplex structure are sharply bent regions with broken stacking interactions , which are unambiguously `` kinks '' . in an unstressed system , however , the most common deviations from the basic duplex are not unambiguously identifiable as kinks , making the definition of a meaningful @xmath30 problematic . even given well - defined criteria for identifying kinks , @xmath31 is not directly relatable to @xmath26 .
two reasons for this difference are particularly evident in the context of bending duplexes .
firstly , to be a generic quantity , @xmath31 must be defined through the probability that a kink occurs that is centred on a specific base pair or ( base pair step ) in a duplex . in the context of duplex bending ,
however , kink formation can occur over a finite range of locations ( supplementary ) , although kinks near to the centre are preferred , and this fact is reflected in @xmath29 .
secondly , in pulling the ends of the duplex together to @xmath32 , we do work on both the unkinked _ and _ the kinked state .
although kinks are more flexible than unkinked duplexes they still have a finite resistance to bending , due both to entropic effects and the need to strain the kink itself , as evidenced from the post - kinking slope in @xmath17 in ( b ) . to compensate , we must apply even more stress ( than if the kink was completely flexible ) to the system to reach the point at which kinks become favourable relative to unkinked duplexes , tending to increase @xmath26 . that the bending stress at which kinking occurs in a specific system ( @xmath29 )
is not equivalent to the intrinsic cost of kink formation ( @xmath30 ) emphasizes the difficulty in comparing between distinct experimental setups , in which the factors contributing to @xmath29 may well be different .
for example , in some setups , it may be possible to relieve stress by kinking at one of a number of positions , as observed in simulations of short minicircles @xcite . in others , such as the molecular vice @xcite
, the kinking location may be constrained to be near the mid - point of a duplex .
the free - energetic cost of bending the kinked state may also be distinct in different systems .
unless care is taken , such differences will hamper quantitative comparison of free energies .
analytic models @xcite used to extract free energies from data must therefore be carefully chosen if quantitative comparisons are to be drawn .
indeed , one of the advantages of explicit simulation with coarse - grained models is that direct comparisons can be made between distinct systems at the level of physical observables , rather than indirectly via free energies derived from analytic models . in this work ,
we will only make direct comparisons of @xmath29 for systems that are identical but for the dna sequences , in which case differences will largely reflect differences in the intrinsic cost of kinking , @xmath30 . given that the relative free energies of bending and kinking can also be perturbed with sequence variation @xcite
, we expect that sequence variation should affect the kink transition .
we report sequence dependence in the free energy @xmath17 of bending for a 28-bp duplex ( ( a , d ) ) .
it is clear from these data that the poly(at ) sequence undergoes kinking more easily than the poly(gc ) sequence ; @xmath29 is lower , presumably due to the weaker base - pairing in the poly(at ) sequence .
a secondary observation is that the poly(at ) sequence is slightly more flexible prior to the kink transition in oxdna .
these two effects should , in principle , have opposing consequences for @xmath32 , tending to increase and decrease @xmath32 respectively .
it is clear from ( d ) , however , that the ease of kinking dominates and poly(at ) has a longer @xmath32 .
we also observe that a gc sequence has a slightly lower @xmath29 than the average - base model ( supplementary ) .
this behavior is expected since the duplex will kink and bend preferentially in at - rich regions . due to the lack of easily kinkable
regions , the average - base model is harder to kink than would be navely expected from an ensemble of gc - content sequences .
having investigated sequence variation , we now turn our attention to a system in which we introduce a mismatch , a structural motif that substitutes a base pair with a non - watson - crick pair .
it is reasonable to expect that kinking will localize to the mismatch , provided it is conveniently positioned , and @xmath29 will be reduced . following fields _ et al . _
@xcite , a mismatch was introduced at the mid - point of an otherwise intact duplex .
we investigated gg , gt , aa and cc mismatches .
as expected , oxdna reports that mismatched sequences have a lower @xmath29 and undergo the kink transition at longer @xmath32 ( ( b , e ) ) .
the change in @xmath29 is consistent with typical free energy costs associated with mismatches @xcite : oxdna reports for the perfectly - matched duplex versus for the mismatches , a deviation of @xmath33 ( see ) .
fields _ et al .
_ also found evidence that ease of kinking was correlated with the sequence - dependent mismatch cost predicted by the santalucia model of dna thermodynamics @xcite .
oxdna s treatment of mismatches is extremely simple , neglecting any residual interactions or possible steric clashes between geometrically incompatible bases .
consequently , it does not strongly distinguish between different mismatches , and does not reproduce sequence - dependent mismatch variation .
like a mismatch , a nick , which is a break in an otherwise contiguous stretch of dna whereby the backbone phosphodiester bond is severed , is a structural motif expected to impact dna bending . in the relaxed state , base pairing is preserved and coaxial stacking may occur between the adjacent , but now non - covalently bonded , nucleotides .
it is relatively easy for a nicked system to give sharp kinks whilst only disrupting stacking and not hydrogen - bonding ( ( f ) ) .
we observe , unsurprisingly , that the presence of a nick dramatically reduces the bending free energy at the kink transition ( ( c ) ) . in accordance with this
lower bending free energy ( @xmath34 ) , the crossover from continuously bent to kinked regimes is shifted to larger @xmath32 by about @xmath35 .
it is also worth noting that dna with a nick is far easier to bend then normal duplex dna after the kink transition : the post - kinking slope for the nicked system ( ( c ) ) is much shallower than for the intact duplex ( ( b ) ) .
the absence of the constraint of the additional backbone bond in the nicked system makes extremely sharp bending very easy once coaxial stacking has been broken ; this contrast highlights the work that must be done to strongly bend a kinked duplex without a nick .
we note that for the nicked system , a more substantial difference between energetic and structural criteria for kinks can be seen ( ( f ) ) .
this distinction arises because nicked systems undergo significant kinking at relatively large @xmath5 , when it is less advantageous for kinked sections to bend extremely sharply ; hence , because the structural criterion implies an effective angle @xmath36 between consecutive bases , it can underestimate the prevalence of kinks .
having rigorously explored bending in a generic context , we turn to the `` molecular vice '' of fields _ et al . _
@xcite , which provides an exquisite experimental probe of the behaviour of duplex dna under bending stress .
we use oxdna to explore the microscopic configurations that underlie their experimental observations .
we will show that our findings are compatible with the observations of fields _ et al .
_ at high salt concentrations . in our simulations ,
the thymines located at the first and last bases of the loop were chosen as the reporter for hairpin separation since they are the closest analogues to the experimental setup .
specifically , we define @xmath37 as the separation between these bases centres of mass .
we begin our discussion by considering the system with a loop length of @xmath38 nucleotides and a variable length duplex of @xmath3 base pairs ( ) .
oxdna reveals distinct behaviour for small , medium and large values of @xmath3 .
these behaviours are related to the most prevalent deformation modes from the canonical relaxed duplex dna structure at each given @xmath3 length .
possible deformations to consider involve unzipping of the hairpin stem , and fraying , bending or kinking of the duplex . and variable @xmath3 . *
( a)(i - iv ) * representative structures for @xmath39bp , respectively . stem - end thymines ( * yellow * ) are used for fluorophore attachment and reporting hairpin separation .
a kink is highlighted with an * arrow*. * ( b ) * @xmath37 is the separation of the two thymines bases at the interface between the hairpin stem and loop . *
( c ) * @xmath19 is the probability of kinking in the duplex via the structural criterion . *
( d ) * @xmath40 and @xmath41 are the number of disrupted base pairs via fraying in the duplex ( ) and unzipping of the hairpin stem ( ) , respectively . , width=317 ] . * ( a ) * kinking probability @xmath19 via the structural criterion . *
( b ) * hairpin separation @xmath37 .
maxima in @xmath37 are highlighted ( black circles ) .
* ( c ) * the experimental fret efficiency @xmath42 as reported in ref .
the available data is for @xmath43 .
, width=317 ] for short duplexes ( @xmath44 ) relative to the loop length @xmath38 , the hairpin separation @xmath37 is comparable to that of a hairpin without an annealed complement strand ( ( b ) ) .
( c ) and ( d ) show that kinking and fraying of the duplex , and hairpin unzipping , are small , and simulation snapshots such as ( a)(ii ) do not show significantly bent duplexes . in this regime ,
the duplexes are too short to exert significant stress on the system . for intermediate - length duplexes ( @xmath45 )
, @xmath37 increases due to stem unzipping ( ( b ) ) .
the duplex is sufficiently long that the system can not maintain full base - pairing without experiencing considerable mechanical stress .
the system must either bend the duplex , or suffer a reduction in base pairing ( either via unzipping the stem or fraying at the ends of the duplex ) .
in fact , the system adopts configurations in which the deformation is shared between continuous bending of the duplex and unzipping , as can be seen from ( a)(iii ) .
the bending energy released for each base pair unzipped decreases as the duplex gets straighter , whereas the cost of unzipping each base pair in the stem is approximately constant .
the system can therefore minimize its free energy by adopting a state which combines unzipping with moderate bending . unzipping the hairpin stem is favoured over fraying the duplex for two reasons ; firstly , stem unzipping contributes two nucleotides worth of single - stranded dna to the hairpin loop ( thereby reducing the tension ) , versus a single base for fraying the duplex . secondly , the weak poly(at ) stem is easier to melt than the strongly binding gc - rich regions at both ends of the duplex ( this was a deliberate design choice of fields _ et al . _
@xcite ) . for long duplexes ( @xmath46 ) , @xmath37 decreases by re - zipping the stem and kinking , as reported in ( b - d ) ; a typical kinked structure is shown in ( a)(iv ) . increasing @xmath3 increases the stress in the system , leading to stronger unzipping and bending costs if the duplex does not kink
the cost of kinking , however , is much less strongly dependent on @xmath3 ( as was discussed in the general analysis of bending ) .
thus at sufficiently large @xmath3 , it is favourable to form a kink in the centre of the duplex , relieving the bending stress elsewhere and reducing the need to unzip .
as well as considering the behaviour at fixed @xmath47 , it is instructive to analyse the system as @xmath1 is varied , focusing on the highly stressed regime of @xmath48 ( ( a ) and ( b ) ) .
we observe three regimes : for small @xmath49 , unzipping and continuous bending are primarily observed , with almost no signal of kinking in @xmath37 . for @xmath50 , we see a sharp kink transition in which kinked states become dominant as @xmath51 ( @xmath52 sits near the crossover between these regimes ) . for larger
@xmath1 , this transition is less pronounced , eventually fading away .
additional simulations at @xmath53 , with @xmath54 , show the kinking probability decaying from @xmath55 to @xmath56 ( supplementary ) . why does a kink transition occur only for a finite range of @xmath1 ?
unzipping of base pairs is extremely effective at relaxing stress for short duplexes : each unzipped base pair opened allows the duplex to straighten substantially .
thus at short lengths , stress is primarily relaxed by unzipping , duplex bending is small and kinking does not come into play . as the duplex gets longer , each unzipped base pair has less of a relaxing effect on the duplex , as it provides a smaller contribution relative to the overall length discrepancy between the intrinsic length of the duplex and the end - to - end distance enforced by the vice
. consequently , the bending of the duplex is enhanced , and eventually a kink transition is observed when this bending is sufficiently severe . for longer duplexes , the enthalpic cost of continuously bending to satisfy the separation imposed by the vice drops with duplex length ( this can be shown directly using the wlc expression of becker _
et al . _
@xcite ) , whereas kink formation does not get easier .
thus the driving force for kinking is reduced , and kinking is absent at sufficiently long loop lengths .
another notable point is that the maximum in @xmath57 prior to the kink transition ( indicated by black dots in ( b ) ) drops with increasing @xmath1 at values of @xmath1 for which kinking is observable .
why should this be ?
as mentioned above , unzipping of stem base pairs is increasingly ineffective for larger systems , and bending of the duplex becomes increasingly easy .
therefore the degree to which the system unzips as @xmath51 is reduced for larger systems , and @xmath57 does not grow as significantly prior to kinking .
we now compare the oxdna results to those of fields _ et al . _
@xcite . unlike oxdna , fields _ et al .
_ could not directly observe kinked macrostates .
instead , they inferred kinking from the non - monotonic behavior of the fret signal between dyes attached to the first two bases of the loop ( a proxy for unzipping - ( c ) ) .
they defined a critical buckling length at high - salt , which they interpreted as the onset of localized bending via kinking , as a local minimum in the fret signal .
we can define an analogous quantity in oxdna , the maximum in @xmath58 for a given @xmath1 . in their higher salt experiments , fields _ et al .
_ saw evidence of unzipping and no kinking for @xmath59 , identified non - monotonic behavior of fret for @xmath60 and saw no substantial change in the fret signal with @xmath3 for larger values of @xmath1 . broadly speaking ,
these three regimes are consistent with our observations : kink transitions occur for intermediate values of @xmath1 , with pure unzipping at smaller values and no unzipping or kinking at larger values .
quantitatively , we observe the beginning of a kink transition for @xmath61 , slightly earlier than fields _
et al . _ , and see evidence of kinking at larger values of @xmath1 .
the @xmath3 values at which kink transitions occur for @xmath62 and @xmath63 in oxdna are within two base pairs of those inferred by fields _
, as can be seen in ( c ) ( clearly , however , inferring the ` exact ' onset of kinking is not trivial from the experimental data )
. the quantitative consistency of results may be even stronger than immediately apparent .
oxdna is parametrized at , meaning that the stem of the vice should be somewhat more resistant to unzipping than in the experiments , which were performed at , perhaps explaining the onset of kinking for @xmath52 that is not apparent in the work of fields _ et al . _
( for additional discussion , please see supplementary ) consistent with this argument , fields _ et al .
_ do not find evidence for kinking at all at much lower salt concentrations , when the stem of the vice is less stable .
slightly smaller values of @xmath3 at which buckling occurs for a given @xmath1 in oxdna may also be explicable in this way .
our results for larger values of @xmath1 also suggest that although kinking occurs , it does so following an increasingly small amount of unzipping .
this is actually consistent with the observations of fields _ et al .
_ compare the drop in fret prior to kinking for @xmath38 and @xmath64 .
we posit that unzipping prior to kinking may be undetectable through the methodology of fields _ et al . _ for @xmath65 .
there is a highly non - linear relationship between fret signal and fluorophore separation , which is roughly equivalent to @xmath37 .
fret is most sensitive near the frster radius , @xmath66 for the cy3b / alexa647 fluorophore pair used to probe stem unzipping .
oxdna reports @xmath67 for @xmath68 , suggesting that it is possible that kinking occurs in experiment at larger values of @xmath1 without sufficient change in fret signal for it to be observable .
oxdna does suggest , however , that kinking is never observed when @xmath1 is large enough that @xmath48 duplexes undergo continuous bending .
in this paper we have provided a full characterization of the thermodynamics of dna bending in a coarse - grained model , oxdna . when weakly bent , oxdna is well - described by the worm - like chain ( wlc ) model .
however , if a sufficiently strong degree of bending is imposed , oxdna tends to adopt an alternative conformation in which the strain is localized in a small region of disrupted base - pairing and stacking interactions .
the cost of such local kinks are compensated by the relaxation of the rest of the molecule .
such alternative modes of relaxation are not captured by the wlc model and lead to an enhanced probability of sharp bends .
although a kink is far more flexible than canonical b - dna , bending is not completely free .
conformations are more restricted than for a duplex with a nick , for example , which is not constrained by two continuous backbones .
our results enable us to identify the end - to - end constraint that is sufficient to cause kinking ( or localized bending ) .
short duplexes exhibit localized bending at relatively larger end - to - end distance than long duplexes , while sufficiently long duplexes ( @xmath69 ) do not undergo kinking even if the ends are held as close together as possible ; they are long enough to achieve such a conformation through reasonably gentle bending distributed throughout the molecule . additionally , at - rich sequences exhibit localized bending at longer end - to - end distance than gc - rich sequences , due to the lower cost of disrupting weaker interactions .
our analysis of kinking reveals that one must be extremely careful in inferring free energies of kinking from specific systems .
a natural definition of @xmath30 would be through the probability of spontaneous kink formation in relaxed dna at a specific site
this should be a generic property depending on the local sequence , and would be the quantity measurable if relaxed duplexes were examined for kinking . if kinking is enforced through external bending of duplex dna , however , the amount of work that must be done on the duplex state to enforce a transition , @xmath29 , is potentially quite distinct from @xmath30 .
reasons include the fact that work may also be done in bending the kinked state to reach the point of transition , and the fact that specific experimental systems ( such as the molecular vice @xcite ) may favour kink formation at certain locations whereas other systems ( such as dna minicircles @xcite ) may allow kinks anywhere .
these considerations make direct comparisons of free - energies of kinking as inferred through simple analytic models difficult , hindering the comparison of kinking in distinct experimental systems .
computational models such as oxdna have the virtue that they allow comparison between different experimental systems in which kinking has been hypothesized , as they can be directly simulated with the same model and experimentally comparable observables can be computed . in particular , physical factors such as the work done in bending a kink , or the preference for some kink locations over others in an experimental setup , should be naturally incorporated ( at least at a semi - quantitative level ) .
coarse - grained models parametrized to reproduce basic dna mechanics and thermodynamics therefore offer a valuable tool for establishing whether proposed observations of kinking are consistent with each other and our current knowledge of dna . to this end , we have simulated the `` molecular vice '' of fields _ et al . _
@xcite , a direct probe of dna bending . in this experiment ,
the hybridization of a strand to a hairpin loop can lead to a range of behaviours interpreted by fields _
_ as duplex bending , kinking and hairpin stem unzipping dependent on the length of the strand and hairpin loop .
our results are consistent with the observations of fields _ et al .
_ , support their interpretation that kinking is responsible for rezipping at high salt , and provide a more detailed mechanistic perspective on the data .
we also propose that duplex kinking might be present for longer duplexes than can be directly inferred from the experimental data , due to the limits of fret precision and minimal unzipping prior to kinking .
oxdna is a simplified model , and the level of agreement with fields _ et al .
_ should not be over - interpreted .
although reproduction of the data is evidence that the model is well - founded , a number of modelling simplifications , and the slight difference in salt conditions in experiment and oxdna , might give rise to small quantitative deviations .
the important point is that the general agreement between the experimental data and oxdna helps to further validate fields _
et al . _
s inferral of kinking , and to show that this behaviour is consistent with the known thermodynamics and mechanics of dna as captured by the oxdna model .
furthermore , we have recently also performed careful comparisons between oxdna simulations and experiments on dna cyclization @xcite and minicircles @xcite .
reasonable consistency with experiment in these studies allows us to conclude that the experimental evidence of kinking in these systems is broadly consistent , and gives us further confidence that oxdna describes the basic physics of kinking well .
this consistency is particularly important given the controversy surrounding the existence of kinks and enhanced dna flexibility @xcite .
we note that our observation that kinking may go undetected for intermediate loop sizes in the setup of fields _ et al .
_ is potentially important in this regard ; otherwise the absence of kinking for fairly short @xmath1 might be surprising when compared to digestion assays on minicircles @xcite , particularly as simple mechanical arguments suggest that the geometry of torsionally relaxed minicircles is less conducive to kinking than the geometry of the molecular vice .
this work was supported by the engineering and physical sciences research council [ ep / i001352/1 ] , the national science foundation graduate research fellowship program , the national institutes of health national heart , lung and blood institute , wolfson college , oxford , and university college , oxford .
_ conflict of interest statement_. none declared .
the authors acknowledge the use of the computing facilities of oxford advanced research computing and the e - infrastructure south iridis high performance computing facility . 10 lewis , m. , chang , g. , horton , n. c. , kercher , m. a. , pace , h. c. , schumacher , m. a. , brennan , r. g. , and lu , p. ( 1996 ) crystal structure of the lactose operon repressor and its complexes with dna and inducer .
_ science , _ * 271 * , 12471254 .
nikolov , d. b. , chen , h. , halay , e. d. , hoffman , a. , roeder , r. g. , and burley , s. k. ( 1996 ) crystal structure of a human tata box - binding protein / tata element complex .
usa , _ * 93 * , 48624867 .
richmond , t. j. and davey , c. a. ( 2003 ) the structure of dna in the nucleosome core .
_ nature , _ * 423 * , 145150 .
widom , j. ( 2001 ) role of dna sequence in nucleosome stability and dynamics .
biophys . ,
_ * 34 * , 269324 .
haeusler , a. r. , goodson , k. a. , lillian , t. d. , wang , x. , goyal , s. , perkins , n. c. , and kahn , j. d. ( 2012 ) fret studies of a landscape of lac repressor - mediated dna loops .
acids res . , _ * 40 * , 44324445 .
davis , n. a. , majee , s. s. , and kahn , j. d. ( 1999 ) tata box dna deformation with and without the tata box - binding protein .
biol . , _ * 291 * , 249265 .
shimada , j. and yamakawa , h. ( 1984 ) ring - closure probabilities for twisted wormlike chains .
application to dna .
_ macromolecules , _ * 17 * , 689698 .
thirumalai , d. and ha , b .- y .
( 1997 ) statistical mechanics of semiflexible chains : a meanfield variational approach .
_ arxiv : cond - mat/9705200,_. mazur , a. k. ( 2007 ) wormlike chain theory and bending of short dna .
lett . , _ * 98 * , 218102 .
becker , n. b. , rosa , a. , and everaers , r. ( 2010 ) the radial distribution function of worm - like chains .
j. e , _ * 32 * , 5369 .
shore , d. and baldwin , r. l. ( 1983 ) energetics of dna twisting : i. relation between twist and cyclization probability .
, _ * 170 * , 957981 .
bustamante , c. , marko , j. f. , siggia , e. d. , and smith , s. b. ( 1994 ) entropic elasticity of @xmath70-phage dna .
_ science , _ * 265 * , 15991600 .
du , q. , smith , c. , shiffeldrim , n. , vologodskaia , m. , and vologodskii , a. ( 2005 ) cyclization of short dna fragments and bending fluctuations of the double helix .
natl . acad .
usa , _ * 102 * , 53975402
. cloutier , t. e. and widom , j. ( 2004 ) spontaneous sharp bending of double - stranded dna . _
cell , _ * 14 * , 355362 .
vafabakhsh , r. and ha , t. ( 2012 ) extreme bendability of dna less than 100 base pairs long revealed by single - molecule cyclization .
_ science , _ * 337 * , 10971101 .
le , t. t. and kim , h. d. ( 2014 ) probing the elastic limit of dna bending .
acids res .
, _ * 42 * , 1078610794 .
podtelezhnikov , a. a. , mao , c. , seeman , n. c. , and vologodskii , a. ( 2000 ) multimerization - cyclization of dna fragments as a method of conformational analysis .
_ biophys .
j. , _ * 79 * , 26922704 .
du , q. , kotlyar , a. , and vologodskii , a. ( 2008 ) kinking the double helix by bending deformation .
acids res . , _ * 36 * , 11201128 .
demurtas , d. , amzallag , a. , rawdon , e. j. , maddocks , j. h. , dubochet , j. , and stasiak , a. ( 2009 ) bending modes of dna directly addressed by cryo - electron microscopy of dna minicircles .
acids res . , _
* 37 * , 112 .
qu , h. , tseng , c .- y . , wang , y. , levine , a. j. , and zocchi , g. ( 2010 ) the elastic energy of sharply bent nicked dna .
_ epl , _ * 90 * , 18003 .
qu , h. , wang , y. , tseng , c .- y . , and zocchi , g. ( 2011 ) critical torque for kink formation in double - stranded dna
x , _ * 1 * , 021008 .
qu , h. and zocchi , g. ( 2011 ) the complete bending energy function for nicked dna .
_ epl , _ * 94 * , 18003 .
kim , c. , lee , o .- c . , kim , j .- y . , sung , w. , and lee , n. k. ( 2015 ) dynamic release of bending stress in short dsdna by formation of a kink and forks .
, _ * 54 * , 16
. fields , a. p. , meyer , e. a. , and cohen , a. e. ( 2013 ) euler buckling and nonlinear kinking of double - stranded dna .
acids res . , _ * 41 * , 98819890 .
vologodskii , a. , du , q. , and frank - kamenetskii , m. d. ( 2013 ) bending of short dna helices .
_ artificial dna : pna & xna , _ * 4 * , 13 .
vologodskii , a. and frank - kamenetskii , m. d. ( 2013 ) strong bending of the dna double helix .
acids res . , _
* 41 * , 67856792 .
wiggins , p. a. , van der heijden , t. , moreno - herrero , f. , spakowitz , a. , phillips , r. , widom , j. , dekker , c. , and nelson , p. c. ( 2006 ) high flexibility of dna on short length scales probed by atomic force microscopy .
nanotechnol . , _ * 1 * , 137141
. mazur , a. k. and maaloum , m. ( 2014 ) dna flexibility on short length scales probed by atomic force microscopy .
_ * 112 * , 068104 .
crick , f. h. c. and klug , a. ( 1975 ) kinky helix .
_ nature , _ * 255 * , 530533 .
yan , j. and marko , j. f. ( 2004 ) localized single - stranded bubble mechanism for cyclization of short double helix dna .
lett . , _ * 93 * , 108108 .
mitchell , j. s. , laughton , c. a. , and harris , s. a. ( 2011 ) atomistic simulations reveal bubbles , kinks and wrinkles in supercoiled dna .
acids res .
, _ * 39 * , 39283938 .
sivak , d. a. and geissler , p. l. ( 2012 ) consequences of local inter - strand dehybridization for large - amplitude bending fluctuations of double - stranded dna .
_ j. chem . phys . , _ * 136 * , 045102 .
ouldridge , t. e. , louis , a. a. , and doye , j. p. k. ( 2011 ) structural , mechanical , and thermodynamic properties of a coarse - grained dna model . _ j. chem .
_ * 134 * , 085101 .
ouldridge , t. e. ( 2012 ) coarse - grained modelling of dna and dna self - assembly , springer , berlin .
ulc , p. , romano , f. , ouldridge , t. e. , rovigatti , l. , doye , j. p. k. , and louis , a. a. ( 2012 ) sequence - dependent thermodynamics of a coarse - grained dna model . _ j. chem .
_ * 137 * , 135101 .
harrison , r. m. , romano , f. , ouldridge , t. e. , louis , a. a. , and doye , j. p. k. ( 2015 ) coarse - grained modelling of strong dna bending ii : cyclization .
acids res . , _ _
submitted_. harrison , r. m. , romano , f. , ouldridge , t. e. , louis , a. a. , and doye , j. p. k. ( 2015 ) coarse - grained modelling of dna minicircles .
_ to be submitted,_. doye , j. p. k. , ouldridge , t. e. , louis , a. a. , romano , f. , ulc , p. , matek , c. , snodin , b. e. k. , rovigatti , l. , schreck , j. s. , harrison , r. m. , and smith , w. p. j. ( 2013 ) coarse - graining dna for simulations of dna nanotechnology .
, _ * 15 * , 2039520414 .
matek , c. , ouldridge , t. e. , levy , a. , doye , j. p. k. , and louis , a. a. ( 2012 ) dna cruciform arms nucleate through a correlated but asynchronous cooperative mechanism .
_ j. phys .
b , _ * 116 * , 1161611625 .
romano , f. , chakraborty , d. , doye , j. p. k. , ouldridge , t. e. , and louis , a. a. ( 2013 ) coarse - grained simulations of dna overstretching .
_ j. chem .
_ * 138 * , 085101 .
matek , c. , ouldridge , t. e. , doye , j. p. k. , and louis , a. a. ( 2015 ) plectoneme tip bubbles : coupled denaturation and writhing in supercoiled dna . _ scientific reports , _ * 5 * , 7655 .
whitelam , s. , feng , e. h. , hagan , m. f. , and geissler , p. l. ( 2009 ) the role of collective motion in examples of coarsening and self - assembly .
_ soft matter , _ * 5 * , 12511262 .
torrie , g. and valleau , j. ( 1977 ) nonphysical sampling distributions in monte carlo free - energy estimation : umbrella sampling .
_ j. comput .
* 23 * , 187199 .
geggier , s. and vologodskii , a. ( 2010 ) sequence dependence of dna bending rigidity .
usa , _ * 107 * , 1542115426 .
santalucia , j. and hicks , d. ( 2004 ) the thermodynamics of dna structural motifs .
biom . , _ * 33 * , 415440 . * supplementary material for `` coarse - grained modelling of strong dna bending i : thermodynamics and comparison to an experimental molecular vice '' *
bending simulations were performed in three phases : exploratory , equilibration and production . in the exploratory phase , we iteratively adjusted the umbrella sampling biasing potential to yield a flat probability distribution @xmath71 , where @xmath4 is the end - to - end distance between the base interaction sites of the first and last bases on a single strand using a discrete potential with increments of . as stated in the main text ,
the duplexes were constrained not to fray , allowing us to sample in terms of @xmath5 , the end - to - end distance between the centres of the helical axis at the first and last base pairs , choosing an @xmath5 bin - width of .
non - native base - pairing was also forbidden . as the iteration was performed manually , simulation times varied in the range vmmc steps per particle . in the equilibration phase , we equilibrated the system for vmmc steps per particle using the aforementioned umbrella biasing potential .
adequate sampling ( number of transitions between end - to - end distance bins ) and decorrelation ( via a block average decorrelation method ) of potential energy , bubble size , fraying ( where applicable ) and structural kinking were checked . for the production phase , each production simulation ( five independent simulations per measurement ) was initialized with a statistically independent starting configuration and a unique random seed .
this is accomplished by randomly drawing starting configurations from the equilibration phase , one for each production run , ignoring the first vmmc steps per particle of equilibration .
the production run time for each simulation was vmmc steps per particle . for reference ,
the characteristic decorrelation time for the potential energy is @xmath72 vmmc steps per particle , while the decorrelation time for kinking is @xmath73 vmmc steps per particle .
the `` seed moves '' used to build clusters in the vmmc algorithm were : * rotation of a nucleotide about its backbone site , with an axis chosen uniformly on the unit sphere , and with an angle drawn from a normal distribution with a mean of zero and a standard deviation of 0.10 radians .
* translation of a nucleotide , where the displacement along each cartesian axis is drawn from a normal distribution with a mean of zero and a standard deviation of .
additional simulations were performed to measure the free energy of kinked states for a 30-bp duplex at larger values of @xmath5 . in this case ,
as well as biasing @xmath5 , the umbrella potential was used to forbid states without kinks ( separate simulations were performed for each kink criterion ) .
umbrella potentials for @xmath5 were obtained via manual iteration at a preliminary stage .
simulations were started from kinked configurations and thermalized for vmmc steps per particle .
subsequent data collection lasted for vmmc per particle .
five independent simulations were performed .
the vmmc `` seed moves '' differ slightly from those mentioned previously : * rotation of a nucleotide about its backbone site , with the axis chosen from a uniform random distribution and the angle from a normal distribution with mean of zero and a standard deviation of 0.15 radians ; * translation of a nucleotide with the direction chosen from a uniform random distribution and the distance from a normal distribution with mean of zero and a standard deviation of .
to isolate the effects of dna bending , as opposed to alternative deformation modes such as fraying of the terminal base pairs , we impose a condition preventing fraying in our dna bending simulations .
note that this constraint was not applied in simulations of the molecular vice system . using an additional order parameter ,
the 30-bp system was constrained such that the terminal base pairs remained hydrogen - bonded , thereby eliminating the possibility of fraying at duplex ends .
we compared constraints where 1 , 3 and 5 base pairs at the duplex ends were prevented from opening , finding that the various conditions yield comparable results ( ) . and kinking in an @xmath74 duplex . limiting cases allowing fraying and enforcing hydrogen bonding along the entire duplex
are also shown .
* ( a ) * free energy of bending .
probability of kinking based on the * ( b ) * energetic or * ( c ) * structural criteria for kinking . in all panels , for the frayed system @xmath5
( the end - to - end distance between the centres of the helical axis at the first and last base pairs ) is ill - defined , so the very similar @xmath4 ( the end - to - end distance between the base interaction sites of the first and last bases along the umbrella sampled strand ) is used instead .
, width=321 ] to further probe the behaviour of the 30-bp system , in separate simulations we allowed the system to fray freely or preserved hydrogen bonding along the entire duplex . unsurprisingly , as frayed base pairs can adopt a large range of conformations including many that reduce @xmath5 , allowing fraying as an additional deformation mode reduces the bending free - energy penalty and delays the onset of kinking . preserving hydrogen bonding throughout
the duplex also delays kinking , but by limiting kinked states to those that do not involve disrupted base pairing ; only those that break stacking but not base pairing remain , and thus the free energy cost of kinking is higher . since a system in which fraying is permitted may not have terminal base pairs ,
the usual definition of @xmath5 reporting the separation between the helical axis centres of the terminal base pairs , is not well defined .
therefore , for this system we instead use @xmath4 , defined as the separation between the base interaction sites of the first and last bases along a single strand .
akin to the bending simulations of duplex dna ( ) , the molecular vice simulations were performed in three phases : assembly , equilibration and production .
an exploratory phase is not required since the simulations were unbiased , i.e. umbrella sampling is not used . as in the bending simulations ,
non - native base pairing was disallowed for simplicity . in the assembly phase ,
the system was initialized as two single strands .
harmonic traps were used to bring the stem ends , as well as the loop and complement strand , into proximity .
assembly with traps requires @xmath75 vmmc steps per particle , significantly faster than free assembly .
the assembled systems were equilibrated for vmmc steps per particle . for the production phase , each production simulation ( six independent simulations per point ) was initialized with a statistically independent starting configuration and a unique random seed .
this was accomplished by randomly drawing starting configurations from the equilibration phase , one for each production run , ignoring the first vmmc steps per particle of the equilibration phase .
the length of each production run was @xmath76 vmmc steps per particle .
all simulations were performed with the sequence - dependent oxdna parameterization .
where available ( @xmath43 ) , the sequences are identical to those used in the experiments of fields _ et al . _
@xcite ; where not available , similar sequences are used .
all sequences are written @xmath77 to @xmath78 .
the fields _ et al . _
stem sequence is used for all systems : ` gcc cgg cgg ctt ata aaa ttt att aat tat ata ttt tat tta ata taa t - loop ` .
a complete list of loop sequences is available in .
[ cols= " < , < " , ] in the main text , we used two ` kink detectors ' to identify the emergence of kinked states : one with an energetic criterion and one with a structural criterion .
we saw that both measures located the transition to a kink - dominated ensemble at approximately the same @xmath5 . to understand the properties of these detectors further , we performed additional simulations for a 30-bp duplex in which a second umbrella bias was used to forbid unkinked states ( separate simulations were performed for each kink criterion ) . in , we report @xmath17 ( the free energy of the kinked state ) versus @xmath5 obtained from these simulations .
curves were normalized by manually matching to the data obtained for small @xmath5 in the original simulations in which unkinked states were permitted the zero of free energy is then the same as in ( a ) .
both detectors identify deviations from the b - dna structure of the duplex that are extremely rare in the ensemble of typical states , as can be seen from . in snapshots of simulations around @xmath32
, they also appear to do a good job of distinguishing between clearly kinked and clearly unkinked structures .
they are not , however , perfect reporters for configurations that would be intuitively interpreted as ` kinks ' .
firstly , there is always a degree of arbitrariness involved in defining a kink : is a disruption to the duplex structure of the type that allows enhanced bending a kink even if the net bending angle around the disruption is less than , or less than ?
of our two detectors , the structural criterion is more exacting in this regard , which explains why it reports slightly fewer kinks at @xmath79 than the energetic criterion .
although the two detectors do not agree exactly , however , they are reasonably consistent at @xmath80 .
this is because duplex disruptions are only common at @xmath80 _ because _ they allow large bending .
we do not observe a significant number of configurations that meet the energetic criterion without large , localized bending because such states do not help to relax the stress applied , and hence remain rare .
thus most configurations that meet the energetic criterion meet the structural criterion ( although a few are not quite bent enough ) . at @xmath81 ,
the number of structures that satisfy the energetic criterion but not the structural criterion is larger , partly because less bending must be localized at kinks to relax the system elsewhwere .
note that for the nicked system , kinking occurs at much larger @xmath5 due to the lower cost ; hence systems can relax the bending stress without meeting the structural criterion , leading to a bigger discrepancy between the two detectors at @xmath79 in this case .
there is a second reason why the curves for the energetic and structural detectors diverge at @xmath81 , which is related to a second reason why the detectors are not perfect reporters for kinking .
as can be seen in , the free energy curves for systems constrained by the energetic and structural detectors exhibit a noticeable change in slope as @xmath5 gets larger .
this change in slope is analogous to the change in slope of the free energy of the unconstrained system , also shown in .
such a slope change is characteristic of a crossover from one typical conformation ( with a certain flexibility ) to a qualitatively different one with a distinct flexibility . for our unconstrained system ,
the slope change is simply the transition between kinked and unkinked states .
thus at @xmath82 nm , the typical states that satisfy the energetic and structural criteria for kinking are qualitatively distinct from those that are common at @xmath80 . for example , the energetic criterion can be met by stretching a base pair step , and the structural criterion can be met by flipping out a single base from the helix and rotating it ; configurations of this kind are more probable than those that we would intuitively describe as kinked _ if the duplex is not sharply bent_. these configurations highlight an additional difficulty in unambiguously defining a metric and cost for kinking .
however , such states are only common relative to other configurations that satisfy the kink criteria at large @xmath5 ; they are still rare compared to normal states at large @xmath5 and remain rare at all values of @xmath5 as they do not help to alleviate bending stress . in summary , despite the ambiguities in defining and identifying kinks , only duplex disruptions associated with localized bending through large angles are common at @xmath80 for an intact duplex .
thus the two kink detectors are sufficiently consistent and provide meaningful results . for a 30-bp duplex constrained to kink by either the energetic or structural criteria .
the free energy is normalized by manually matching to the data reported in the main text ( ( a ) ) for kinked states at small @xmath5 .
error bars are comparable to line thickness . also shown for comparison
is the data for a system not constrained to show kinking .
, width=321 ]
the distribution of bubble sizes reveals the frequent presence of a bubble in kinked duplexes ( ) .
briefly , bubbles are defined as loss of base pairs away from the duplex ends , while loss of base pairs at duplex ends is classified as fraying .
unsurprisingly , bubbles are almost always co - localized with kinks , so we take total bubble size along the duplex as a proxy for bubbles at a kink .
results for total bubble size given the energetic and structural criteria for a kink are very similar . ) of the total bubble size in base pairs ( @xmath83 ) given the presence of a kink via either the energetic ( ) or structural ( ) criteria for @xmath74 at @xmath84 from the umbrella potential bin @xmath85 .
error bars are of the order of the point size . , width=321 ] nomenclature has been developed in all - atom md simulations @xcite to distinguish between type - i ( disruption of stacking but not base pairing ) and type - ii kinks ( disruption of both base pairing and stacking ) . due to its coarse - grained nature , we have less confidence that oxdna has sufficient molecular detail to contribute reliably to this discussion ; furthermore , the distinction between type - i and type - ii kinks are of less direct relevance to this study .
we therefore do not focus on the classification of kink subtypes , although we note that we see configurations which could be assigned to both types ( ) and that most kinks involve broken base pairs ( i.e. are of type - ii ) for example , only 10% of the kinks for the states contributing to only involve disruption of stacking .
when the ends of a duplex are pulled together , it would be expected that kinking occurs near the midpoint of the duplex .
kinks far from the midpoint would result in incommensurate relaxed duplex sections , which are then less likely to have their ends in close proximity .
analysis of the kink probability distribution confirms that kinks are localized near the midpoint ( ) .
since structural and energetic kinks are a property defined between pairs of consecutive bases ( see definitions section of main text ) , we report the kink location distribution in base steps , where base step @xmath8 refers to a kink between base index @xmath8 and @xmath10 . for every sampled state ,
all base steps along one strand that satisfy the relevant kink criterion are included in the histogram .
to normalise , we divide by the total number of detected kinks along both strands in all sampled states .
kinks are considered distinct if they are separated by more than @xmath86 base steps .
thus the normalized kink distribution gives the probability that a given pair of neighbours contributes to an arbitrarily chosen kink . to compare multiple lengths
, we report the probability of kinking versus the _ adjusted _ base pair index , centred around the mid - point of the duplex , @xmath87 . the @xmath88 offset accounts for zero - based numbering and indexing the base step to base index @xmath8 instead of @xmath10 . ) and nicked ( @xmath89 ) duplexes .
we sample all duplexes at constant end - to - end distance from the umbrella potential bin @xmath85 .
, width=321 ] for the intact duplexes , the distribution is quite wide , @xmath90 base pairs , suggesting that entropic effects are quite prevalent and some degree of incommensurately can be tolerated .
the distribution also appears to become slightly broader for longer duplexes .
importantly , the preference for localisation about the midpoint of the duplex is because this is the position that provides the maximum decrease in end - to - end distance : in relaxed duplex dna , or in dna minicircles , one would not necessarily expect a preferred site for kinking with the average - base parameterisation of oxdna . for the nicked duplex at @xmath74 ,
the distribution is quite narrow , @xmath91 base pairs , reflecting the strong preference for localisation of kinking at the nick ( ) .
contrary to the intact duplexes , the preference for kinking at a nick is not just due to the bending stress imposed on the system : nicked sites have a lower free energy cost for kinking because base pairing need not be disrupted for a kink to form ; only stacking must be compromised . at @xmath92
, kinking appears to be less effective at relieving bending stress ; for all but the shortest duplexes , there is re - entrant behaviour from kinking back to continuous bending .
a typical kinked structure is shown in ( a ) ; it consists of two long ( relatively relaxed ) duplex sections and a highly bent kink region .
given that @xmath5 is measured in terms of the distance between terminal base pair centres , however , such a configuration can not bring the ends into extremely close proximity ( @xmath93 , the approximate duplex diameter ) due to steric effects .
instead , the blunt ends of the duplex must face each other to achieve @xmath93 and this fundamentally changes the constraints placed on the rest of the system , and makes kinking opposite the blunt ends less effective at relieving the stress .
the dna duplex adopts minicircle - like configurations for extremely small values of @xmath5 , as shown in ( b ) .
minicircles can , of course , undergo kinking ; however , the geometrical constraints mean that it is impossible to localise the majority of bending stress into a single kink @xcite , unlike in systems with a slightly larger separation of the blunt ends . thus we obtain the seemingly paradoxical result that kinking is more likely for systems that have some flexibility in their ends than for systems , such as minicircles , where the ends are constrained to be in extremely close proximity . .
* ( a ) * a kinked state , with @xmath5 small but greater than 2 nm . *
( b ) * an unkinked , minicircle - like state with @xmath94 nm . with this constraint , a single kink can not relax the majority of the bending stress , unlike in case ( a ) . , width=321 ] for clarity , we present the bending free energy and end - to - end separation at the mid - point of the kink transition for all simulations in .
.4@rrr & & + + ( l)1 - 3 20 & 15.9 & 4.7 + 30 & 18.1 & 5.5 + @xmath9530 & 6.0 & 8.0 + 41 & 19.7 & 5.3 + 52 & 21.4 & 4.3 + + + ( l)1 - 3 poly(gc ) & 21.1 & 5.1 + avg - base & 18.0 & 5.4 + poly(at ) & 14.8 & 5.8 + + + ( l)1 - 3 intact & 16.1 & 5.7 + @xmath96gg & 13.0 & 6.0 + @xmath96gt & 13.3 & 6.0 + @xmath96aa & 12.8 & 6.1 + @xmath96cc & 13.5 & 5.9 + + + ( l)1 - 3 @xmath95poly(gc ) & 7.9 & 7.2 + @xmath95avg - base & 6.7 & 7.3 + @xmath95poly(at ) & 4.8 & 7.6 + +
in the main text , we show the structural properties for variable loop length ( @xmath1 ) and duplex length ( @xmath3 ) ( ) .
the data suggest that kinked configurations are gradually superseded by continuously bent states as @xmath1 increases .
this behaviour is also observed in our investigation of length - dependence in the free - energy of bending of intact duplexes ( ) , and would therefore be expected for the molecular vice .
we make the length - dependence of kinking , hairpin separation , fraying and stem unzipping in the molecular vice more explicit in , by profiling the longest @xmath3 at a range of @xmath1 lengths . at the lowest values of @xmath1 , unzipping ( and a small amount of fraying )
is the dominant method of relieving stress . at intermediate @xmath1 ,
kinked states dominate .
a transition from kinked to continuously bent states is observed at long @xmath3 ( and therefore long @xmath1 ) , with the midpoint at @xmath97 . by @xmath98bp ,
duplexes are almost exclusively continuously bent ( @xmath99 ) .
unsurprisingly , the length associated with this crossover is consistent with our bending simulations ( ) .
we can not currently study the exact salt concentration used by fields _
et al . _
@xcite with oxdna .
however , we can probe the impact of increasing temperature , which may also be somewhat analogous to decreasing salt concentration , as both serve to weaken base pairing ; in particular , reducing the unzipping force associated with the hairpin @xcite .
we also note that temperature is not perfectly controlled in gel - based experiments due to electrophoresis - induced heating of the gel running buffer .
temperature - dependence was explored with additional simulations at ( data not shown ) and ( ) . we find negligible kinking at @xmath102 for @xmath52 , consistent with the reasoning that weakened base pairing in the stem leads to easier unzipping .
huguet , j. m. , bizarro , c. v. , forns , n. , smith , s. b. , bustamante , c. , and ritort , f. ( 2010 ) single - molecule derivation of salt dependent base - pair free energies in dna .
usa , _ * 107 * , 1543115436 .
|
dna bending is biologically important for genome regulation and is relevant to a range of nanotechnological systems .
recent results suggest that sharp bending is much easier than implied by the widely - used worm - like chain model ; many of these studies , however , remain controversial .
we use a coarse - grained model , previously fitted to dna s basic thermodynamic and mechanical properties , to explore strongly bent systems . we find that as the end - to - end distance is decreased sufficiently short duplexes undergo a transition to a state in which the bending strain is localized at a flexible kink that involves disruption of base - pairing and stacking . this kinked state ,
which is not well - described by the worm - like chain model , allows the duplex to more easily be sharply bent .
it is not completely flexible , however , due to constraints arising from the connectivity of both dna backbones .
we also perform a detailed comparison to recent experiments on a `` molecular vice '' that probes highly bent dna .
close agreement between simulations and experiments strengthens the hypothesis that localised bending via kinking occurs in the molecular vice and causes enhanced flexibility of duplex dna .
our calculations therefore suggests that the cost of kinking implied by this experiment is consistent with the known thermodynamic and mechanical properties of dna .
|
computational `` thermostats '' arose as a means for controlling numerical simulations of both equilibrium and nonequilibrium stationary states . _ without _ thermostats systems driven away from equilibrium heat up . _ with _ thermostats the heat generated by irreversible processes can be steadily removed , making it possible to simulate nonequilibrium steady states@xcite . because most hamiltonian - based mechanics problems conserve energy , novel _ nonhamiltonian _
ideas are typically required when thermostats are to be included .
nevertheless , several approaches to hamiltonian thermostats have been developed .
here we consider various approaches pioneered by bill ashurst , carl dettmann , denis evans , wm .
g. hoover , tom leete , gary morriss , shuichi nos , and les woodcock over a span of about 25 years@xcite . the simplest thermostat type maintains a ( nearly ) constant kinetic energy by `` rescaling '' the velocities at the end of each computational timestep . for a more elegant , but equivalent , _ continuous _ approach to rescaling , see the `` gaussian isokinetic '' method described in section iib and illustrated in section vii . for @xmath1
cartesian degrees of freedom the kinetic - theory relation , @xmath2 defines the kinetic temperature @xmath3 .
thermostats based on @xmath3 can be applied to an entire system , or separately to two or more subsystems . with this _ ad hoc _ rescaling approach , @xmath4 there is no difficulty in maintaining several temperatures in specified parts of a single nonequilibrium system , as did ashurst in his ph d thesis research@xcite . kinetic theory and
the gibbs - boltzmann development of thermodynamics based on ideal - gas thermometry certainly suggest that the kinetic temperature , @xmath3 , is both the simplest and the most fundamental of the many possible `` temperature '' types .
there are alternatives . among the many possible temperatures based on particle coordinates , rather than velocities , the simplest on which we focus here
is based on the mean - squared forces , @xmath5 .
this definition of a configurational temperature is derived directly from the potential energy @xmath6 and its space derivatives : @xmath7 this somewhat cumbersome relation appears as an aside in landau and lifshitz classic 1958 text@xcite .
it was also recently rediscovered by rugh , and has subsequently been much discussed@xcite .
this configurational representation of temperature follows most simply from an integration by parts , in the canonical ensemble : @xmath8 @xmath9 about ten years after woodcock and ashurst s introduction of the rescaling isokinetic thermostat in the early 1970s , hoover and evans discovered that the isokinetic velocity - rescaling equations of motion can be derived from gauss principle of least constraint@xcite .
fifteen years after this discovery , dettmann and morriss@xcite found a straightforward hamiltonian approach to these same motion equations .
previously , leete and hoover had derived a hamiltonian formulation which likewise maintained the velocity - based kinetic energy @xmath10 constant@xcite . a hamiltonian generating the gauss-principle isokinetic equations of motion was not found until 1996@xcite . in the meantime
nos had discovered his canonical - ensemble thermostated dynamics , which is related to two rather different hamiltonians@xcite .
the purpose of the present work is twofold .
first , we explore the relation of the hoover - leete kinetic thermostat work to its successors .
a 2007 literature search shows only a single reference to it@xcite .
we also explore the consequences of a `` landau - lifshitz '' configurational thermostat based on the canonical - ensemble definition rediscovered by rugh .
the nos - hoover equations of motion@xcite are a standard approach to carrying out canonical - ensemble dynamics at a particular temperature @xmath11 .
this approach uses _ feedback _ , controlling the fluctuating kinetic energy with a `` friction coefficient '' or `` control variable '' @xmath12 .
the frictional feedback force is @xmath13 .
the time rate of change of the friction coefficient is based on the current value of the kinetic energy @xmath14 relative to the desired mean value @xmath15 .
the equations of motion also include a characteristic time @xmath16 which determines the time - range over which the feedback acts : @xmath17/\tau ^2 \ .\ ] ] in most cases , including our @xmath18 model studies described in detail in sections v - ix , a reasonable choice of @xmath16 corresponds to a typical collision time or vibration time .
such a choice typically provides gibbs equilibrium canonical phase - space distribution with a gaussian distribution for the friction coefficient @xmath12 : @xmath19 evidently the amplitude of the fluctuations of @xmath12 varies as @xmath20 , where @xmath1 is the number of degrees of freedom .
thus its effect on the dynamics vanishes in the large - system limit .
_ nonequilibrium _ situations are quite different . by using two or more different friction coefficients ( or `` thermostat variables '' ) temperature differences
can be established , leading to heat flow .
then the corresponding phase - space distributions are no longer smooth and gibbsian , but instead become fractal , with the underlying phase - space trajectories satisfying the second law of thermodynamics infinitely more probable than those violating the law@xcite . in the equilibrium case there is a close connection of nos - hoover mechanics to hamiltonian mechanics .
we describe this connection next . in 1984
shuichi nos discovered a hamiltonian consistent with gibbs constant - temperature canonical ensemble , @xmath21 he was able to derive the nos - hoover equations of motion described above from his hamiltonian by an artificial `` time scaling '' .
in july of 1996 carl dettmann@xcite discovered a simpler approach to the nos - hoover equations of motion .
dettmann s _ vanishing _ hamiltonian is @xmath22 \equiv 0 \ .\ ] ] provided that dettmann s hamiltonian is set equal to the special value , zero , nos s time - scaling variable @xmath23 can be eliminated .
then the nos - hoover equations of motion result _ without the need for any time scaling_. it needs to be emphasized that this @xmath24 trick does _ not _ work for the prototypical _ nonequilibrium _ case of a system with two different temperatures . in a two - temperature system the two different variables @xmath25 are necessarily coupled and can not both be eliminated , so that dettmann s @xmath24 trick does not work . there is a special case of nos - hoover dynamics that deserves special mention , the `` isokinetic '' case in which the temperature is constant .
this corresponds to the choice @xmath26 . for isokinetic mechanics dettmann and
morris discovered the special hamiltonian detailed next .
soon after dettmann discovered a hamiltonian route to the nos - hoover motion equations , dettmann and morriss@xcite discovered a related approach to isokinetic ( constant temperature , provided that temperature is defined by the kinetic energy ) hamiltonian dynamics : @xmath27 as usual , the equations of motion follow by differentiation : @xmath28 the accelerations which follow , using the identity , @xmath29 are then exactly the same as those from the isokinetic equations of motion given above : @xmath30
the hamiltonians corresponding to the isokinetic gauss-principle and nos - hoover velocity - based approaches are not the only such means of thermostating equilibrium systems . a straightforward application of nonholonomic lagrangian mechanics , as well as the familiar hamiltonian mechanics outlined in leete s master s thesis@xcite , lead to another type of isokinetic mechanics .
these two equivalent forms of mechanics , both of which can be used to keep the _
velocity_-based kinetic energy @xmath31 constant , while allowing the momentum - based kinetic energy @xmath32 to vary , proceed by modifying the @xmath33 , rather than the @xmath34 , equations . to distinguish the two different kinetic energies we use the notation :
@xmath35 let us begin with the lagrangian case : @xmath36 \ , \ ] ] where @xmath31 is the _ velocity_-based kinetic energy , @xmath37 , and the lagrange multiplier @xmath38 has the task of maintaining @xmath31 at its initial value , @xmath39 , as the motion proceeds .
lagrange s equations of motion follow from the usual textbook differentiations of the lagrangian with respect to the velocities and coordinates : @xmath40 now multiply the @xmath41 equation by @xmath42 and sum : @xmath43 the value of the lagrange multiplier @xmath44 follows : @xmath45 where the last expression , which completes the identity , follows from the hamiltonian given just below . with leete s help , hoover discovered , in 1979 , that the _ velocity_-based kinetic energy @xmath31 can alternatively be kept constant by using the ( constant in time ) hoover - leete hamiltonian : @xmath46 where @xmath6 is again the usual potential energy and @xmath47 is its initial value .
the constancy of the kinetic energy is easy to see .
the equations of motion , @xmath48 imply that the velocity - based kinetic energy does not vary : @xmath49 \equiv k_0 \ .\ ] ] at the same time , except in the equilibrium case with a large number of degrees of freedom , there is no guarantee that the momentum - based kinetic energy @xmath32 is similar in magnitude to @xmath39 .
but , provided that @xmath50 initially , leete s equations of motion do match the gaussian isokinetic ones to second order , @xmath51 through the second derivatives , @xmath52 , but differ in the _ third _ derivatives , where the meansquared constraint force in the isokinetic case is less than the hoover - leete analog , in accordance with gauss principle of least constraint@xcite .
the constraint of constant temperature in hamiltonian mechanics contradicts the thermodynamic notion that the energy and temperature can not be varied independently in a system of fixed composition and volume .
both @xmath53 and @xmath54 are _ constants _ of the motion , using leete s approach .
provided that the initial conditions are wisely chosen , with @xmath53 and @xmath54 corresponding to the _ same _ thermodynamic state , this approach _ can _ certainly be used to determine equilibrium properties .
we will see , in section viii , what the ( rather strange ) consequences of this thermostat are away from equilibrium .
a ( much ) more complicated hamiltonian thermostat , conserving the force - based _ configurational _ temperature @xmath55 , can be based on straightforward ( though quite tedious ) holonomic hamiltonian mechanics .
a proper configurationally thermostated hamiltonian conserves not only the hamiltonian but also the configurational temperature , @xmath56 here @xmath57 is boltzmann s constant , which we set equal to unity in the numerical work .
a molecular dynamics simulation based on this definition of temperature follows standard hamiltonian mechanics , as the temperature constraint is just a ( complicated ) holonomic ( coordinates only ) constraint .
the simplest procedure begins with the system lagrangian , augmented with a lagrange multiplier @xmath44 which constrains @xmath58 .
the corresponding hamiltonian has the form , @xmath59 for success , the initial conditions have to be chosen to correspond to both the desired temperature , @xmath60 , and to the condition @xmath61 .
then , to begin the analytic work , differentiate the temperature equation , @xmath62 , twice with respect to time .
the two time differentiations , using the chain rule , give first @xmath63 and then @xmath64 . by substituting the constrained equations of motion : @xmath65 for the @xmath66 , we obtain the lagrange multiplier @xmath44 . for further details of this calculation ,
see section ix .
the numerical work can be checked by noting that both @xmath67 and @xmath68 are constants of the motion when the calculation is error free .
the @xmath0 model gets its name from the functional form of a quartic `` tethering potential '' , @xmath69 , which links each particle to a fixed lattice site with a cubic restoring force : @xmath70 in the pedagogical simulations which follow we will choose the tethering force constant equal to unity , @xmath71 .
nearest - neighbor particle pair interactions in the @xmath0 model are governed by hooke s law : @xmath72 aoki and kusnezov have carried out a variety of pedagogical heatflow simulations for this model in both one and three space dimensions@xcite .
the @xmath0 model has a finite fourier conductivity in one dimension .
the numerical work carried out by aoki and kusnezov established the temperature dependence of the thermal conductivity , @xmath73 .
for simplicity in the one - dimensional work which we carry out here , we always choose the nearest - neighbor separation @xmath74 equal to unity .
we also choose the strength of the hookes - law interaction @xmath75 equal to unity . as was abundantly demonstrated by aoki and kusnezov , the combination of a site - based tethering potential with a hookes - law pair potential provides the usual fourier conductivity , with the heat flux proportional to the ( sufficiently small ) temperature gradient .
in what follows we apply four different thermostat constraints to this potential model , using four different dynamical approaches : nos - hoover , gaussian isokinetic , hoover - leete isokinetic , and landau - lifshitz isoconfigurational .
a numerical solution of the heat flow equation for the one - dimensional chain , @xmath76 can be obtained by runge - kutta integration , with a rescaling of the temperatures within the reservoir regions at the end of each timestep accounting for the source and sink terms @xmath77 in the flow equation .
figure 1 shows a fully - converged temperature profile obtained in this way , using hot and cold temperatures of 0.26 and 0.24 , with 400 mesh points .
the spatial gradient operations were replaced by finite - difference approximations : @xmath78 with appropriate subscript changes accounting for the periodic boundary conditions .
with a periodic system composed of four parts , `` hot '' , `` newton@xmath79 '' , `` cold '' , and `` newton@xmath80 '' , each part containing @xmath81 particles , the equations of motion are @xmath82/\tau ^2 \ ; \ ] ] @xmath83/\tau ^2 \ ; \ ] ] in the hot and cold regions , and @xmath84 in the two newtonian regions .
the @xmath18 force @xmath85 depends upon the three coordinates @xmath86 and also includes the tethering force , @xmath87 , where @xmath88 is the lattice site for the @xmath89th particle in the perfect lattice .
the initial velocities and displacements were selected randomly within ranges : @xmath90 with @xmath91 or @xmath92 chosen to reproduce the desired stationary reservoir temperatures , 0.26 and 0.24 .
the _ average _ temperatures , both kinetic and configurational , from the last half of a 400-particle , 40-million timestep simulation , @xmath93 are shown in figure 2 , along with a comparison solution of the macroscopic heat flow equation . the agreement is quite good , as would be expected from aoki and kusnezov s work .
simulations with two _ gaussian isokinetic thermostats _ , with the same timestep , @xmath94 , and the same equilibration and averaging intervals , produced similar results , maintaining the initial kinetic temperatures throughout .
the averaged profiles , along with a corresponding continuum solution , are shown in figure 3 . again
the agreement , for both the kinetic and the configurational temperatures , is perfectly satisfactory .
the instantaneous ( at time @xmath95 ) values of the total heat added ( from time 0 to time @xmath95 ) to the `` hot '' reservoir as well as that removed from the `` cold '' reservoir are shown in figure 4 .
the sum of the two necessarily oscillates about zero as @xmath95 increases .
straight lines drawn through the data provide an estimate for the fourier conductivity in agreement with the results of aoki and kusnezov s work .
the two conventional thermostat types , nos - hoover and gaussian isokinetic , show that either method can be used to simulate the simple two - temperature heat flow problem .
results with the hamiltonian isokinetic thermostats : @xmath96 where the fixed hot and cold kinetic temperatures @xmath97 are fixed by the initial conditions , but come to differ from the unconstrained momentum - based temperatures @xmath98 . unlike the nos - hoover and gaussian isokinetic profiles ,
the hamiltonian - based profiles show _ no temperature gradients at all_. see figure 5 .
instead there are discontinuities between the fixed kinetic temperatures and the adjacent constant values of the unconstrained newtonian regions . in the latter regions the kinetic and configurational temperatures equilibrate , and match the unconstrained configurational temperature of the thermostated regions . in the thermostated regions
there is no such equilibration .
evidently there are no heat fluxes in the system either . by constraining not only the energy ( through the constant hamiltonian driving the motion ) but also the kinetic temperatures , the system is evidently prevented from supporting heat flow .
the hamiltonian nature of the motion equations also prevents the formation of the multifractal phase - space distributions associated with nonequilibrium stationary states@xcite . a more detailed analysis of the reasons for the failure of this hamiltonian approach
could perhaps be based on a detailed analysis of the microscopic equations describing heat flux@xcite .
for the special @xmath0 model considered here , with @xmath99 , the force @xmath85 is @xmath100 \ .\ ] ] the _ pair - potential _ contribution to @xmath101 is @xmath102 , for the two nearest - neighbor interactions , @xmath103 the _ tethering _ potential contribution is @xmath104 : @xmath105 the time derivative of the configurational temperature is a quotient of products of single - particle sums : @xmath106 ^ 2 } \ , \ ] ] where @xmath107 - 3\kappa \delta _ i^2v_i \ .\ ] ] another time differentiation gives the second derivative : @xmath108 ^ 2 } { [ \sum(2\mu + 3\kappa \delta ^2)]^3 } \ -\ ] ] @xmath109)}{[\sum(2\mu + 3\kappa \delta ^2)]^2 } \ .\ ] ] where @xmath110 - \kappa [ 3 \delta _ i^2a_i + 6\delta _ iv_i^2 ] \ .\ ] ] the _ second _ derivative , @xmath111 , introduces the lagrange multiplier through the equations of motion : @xmath112 though complicated , the resulting equation for @xmath44 is linear , and gives an explicit expression for the lagrange multiplier that stabilizes @xmath54 : @xmath113 - [ \sum f^2\sum 6\kappa f_i\delta _ i]\right ) } \ .\ ] ] because @xmath54 can not change , both @xmath114 and @xmath115 need to be properly specified in the initial conditions . a convenient way to do
this is to choose the velocities equal to zero so that @xmath116 .
a typical profile using two configurational thermostats is shown in figure 6 .
just as in the hoover - leete isokinetic case there is no equilibration of the constrained temperature ( here configurational ) .
the unconstrained kinetic temperature equilibrates throughout the system to a value dependent upon the initial conditions .
evidently the landau - lifshitz constrained configurational thermostat is quite useless for nonequilibrium work .
the results obtained with the gaussian isokinetic and nose - hoover thermostats are quite consistent with the past history of their use over the last quarter century@xcite . for these familiar thermostats
there was no problem in reaching a nonequilibrium steady state with temperature profiles and heat fluxes very similar to those expected from macroscopic hydrodynamics using fourier s law .
on the other hand , the results obtained with _ purely hamiltonian _ mechanics , using either the hoover - leete constrained isokinetic or the landau - lifshitz constrained configurational thermostats were both unexpected and thought - provoking . both hamiltonian approaches stably constrained their target temperatures in the two reservoir regions .
nevertheless temperature differences led to no net heat flow and the computed temperature profiles bore no resemblance to the predictions of hydrodynamics . the kinetic and configurational temperatures , though _ unequal _ in the constrained reservoirs , @xmath117 \ , \ ] ] _ were _ equal ( to yet a third nearly - constant temperature ) in the unconstrained bulk newtonian regions : @xmath118 \ .\ ] ] these results emphasize the fact that _ nonhamiltonian _ dynamics is absolutely necessary to the realistic modeling of nonequilibrium systems .
a direct proof / understanding of the failure of the hamiltonian systems to show a nonvanishing heat flux would be most welcome .
this work was first presented at a mexico city workshop organized by francisco uribe at universidad autnoma metropolitana - iztapalapa in january , 2007 .
subsequently , a shorter version was presented at the schrdinger instutite workshop , in wien , honoring the 65th birthday of herr professor doktor harald posch .
we are specially grateful to professors uribe , posch , and dellago for making this work possible .
temperature profiles ( kinetic and configurational ) for a simulation using constrained hoover - leete isokinetic reservoirs .
the timestep is @xmath94 , with a total elapsed time of @xmath121 divided equally between equilibration and averaging portions .
temperature profiles ( kinetic and configurational ) for a simulation using constrained landau - lifshitz isoconfigurational reservoirs .
the timestep is @xmath122 , with a total elapsed time of @xmath123 divided equally between equilibration and averaging portions .
|
we consider and compare four hamiltonian formulations of thermostated mechanics , three of them kinetic , and the other one configurational .
though all four approaches `` work '' at equilibrium , their application to many - body nonequilibrium simulations can fail to provide a proper flow of heat .
all the hamiltonian formulations considered here are applied to the same prototypical two - temperature `` @xmath0 '' model of a heat - conducting chain .
this model incorporates nearest - neighbor hookes - law interactions plus a quartic tethering potential .
physically correct results , obtained with the isokinetic gaussian and nos - hoover thermostats , are compared with two other hamiltonian results .
the latter results , based on constrained hamiltonian thermostats , fail correctly to model the flow of heat .
|
the falicov - kimball model ( fkm ) was first proposed by l. m. falicov and kimball @xcite to analyze the thermodynamics of semiconductor - metal transitions in smb@xmath1 and transition - metal oxides @xcite .
the model incorporates two types of electrons : one type can undergo hopping between sites and the other type can not hop , thereby being localized at the sites .
thus , in its introduction , fkm described the coulomb interaction between mobile @xmath2 band electrons and localized @xmath3 band electrons .
there have been a multitude of subsequent physical interpretations based on this interaction , including that of localized ions attractively interacting with mobile electrons , which yields crystalline formation @xcite .
another physical interpretation of the model is as a binary alloy , in which the localized degree of freedom reflects a or b atom occupation @xcite . in this paper
we employ the original language , with @xmath2 and @xmath3 electrons as conduction and localized electrons with a repulsive interaction between them . since there is no interacting spin degree of freedom in the hamiltonian , the model
is traditionally studied in the spinless case , commonly referred as the spinless fkm ( sfkm ) and which is in fact a special case of the hubbard model in which one type of spin ( _ e.g. _ , spin - up ) can not hop @xcite . in spite of its simplicity , this model is able to describe many physical phenomena in rare - earth and transition metal compounds , such as metal transitions , charge ordering , _
etc_. beyond the introduction of the spin degree of freedom for both electrons @xcite , there also exist many extensions of the original model .
the most widely studied extensions include multiband hybridization @xcite , @xmath4 hopping @xcite , correlated hopping @xcite , non - bipartite lattices @xcite , hard - core bosonic particles @xcite , magnetic fields @xcite , and next - nearest - neighbor hopping @xcite .
exhaustive reviews are available in refs .
the wider physical application of both the basic fkm and its extended versions have aimed at explaining valence transitions @xcite , metal - insulator transitions @xcite , mixed valence phenomena @xcite , raman scattering @xcite , colossal magnetoresistance @xcite , electronic ferroelectricity @xcite , and phase separation @xcite . after the initial works on the fkm @xcite , the literature had to wait 14 years for the celebrated first rigorous results .
two independent studies , by kennedy and lieb @xcite and by brandt and schmidt @xcite , proved for dimensions @xmath5 that , at low temperatures , fkm has long - range charge order with the formation of two sublattices .
various methods have been used in the study of the fkm . in most of these studies , either the @xmath6 infinite - dimensional
limit or @xmath7 low - dimensional cases have been investigated .
studies include limiting cases such as ground - state analysis or the large interaction limit .
renormalization - group theory @xcite offers fully physical and fairly easy techniques to yield global phase diagrams and other physical phenomena .
this non - trivial nature of sfkm motivated us to determine the global phase diagram of the model , which resulted in a richly complex phase diagram involving charge ordering and phase coexistence , as exemplified in fig . 1 .
we use the general method for arbitrary dimensional quantum systems developed by a. falicov and berker @xcite to obtain the global phase diagram of the sfkm in @xmath8 , in terms of both the chemical potentials and the densities of the two types of electrons , for all temperatures .
the outline of this paper is as follows : in sec.ii , we introduce the sfkm and , in sec.iii , we present the method @xcite
. calculated phase diagrams are presented in sec.iv , for the non - hopping ( @xmath9 ) classical submodel and for the hopping ( @xmath10 ) quantum regimes of small , intermediate , and large @xmath11 .
we conclude the paper in sec.v .
the sfkm is defined by the hamiltonian @xmath12 where @xmath13 and @xmath14 denotes that the sum runs over all nearest - neighbor pairs of sites .
note that , as in all renormalization - group studies , the hamiltonian has absorbed the inverse temperature .
the dimensionless hopping strength @xmath15 can therefore be used as the inverse temperature . here
@xmath16 and @xmath17 are respectively creation and annihilation operators for the conduction electrons at lattice site @xmath18 , obeying the anticommutation rules @xmath19 and @xmath20 , while @xmath21 and @xmath22 are electron number operators for conduction and localized electrons respectively .
the operator @xmath22 takes the values @xmath23 or @xmath24 , for site @xmath18 being respectively occupied or unoccupied by a localized electron .
the particles are fermions , so that the pauli exclusion principle forbids the occupation of a given site by more than one localized electron or by more than one conduction electron .
the first term of the hamiltonian is the kinetic energy term , responsible for the quantum nature of the model .
the system being invariant under sign change of @xmath15 ( via a phase change of the local basis states in one sublattice ) , only positive @xmath15 values are considered .
the second term is the screened on - site coulomb interaction between localized and conduction electrons , with positive and negative @xmath25 values corresponding to attractive and repulsive interactions .
we consider only the repulsive case , since the attractive case can be connected to the repulsive one by the particle - hole symmetry possessed by either type of electrons .
particle - hole symmetries are achieved by the transformations of @xmath26 for the localized electrons and @xmath27 , @xmath28 for the conduction electrons , where , for a bipartite lattice , @xmath29 for one sublattice and @xmath30 for the other @xcite .
the last two terms of the hamiltonian are the chemical potential terms with @xmath31 and @xmath32 being the chemical potential for a localized and conduction electron . in order to carry out a renormalization - group transformation easily ,
we trivially rearrange the hamiltonian given in eq.([eq:1 ] ) into the equivalent form of @xmath33\\ \equiv&\sum_{\langle ij\rangle}\left[-\beta\mathcal{h}_{i , j}\right ] , \end{split}\ ] ] where , for a @xmath2-dimensional hypercubic lattice , @xmath34 , @xmath35 , @xmath36 , and @xmath37 is the two - site hamiltonian involving only nearest - neighbor sites @xmath18 and @xmath38 .
in @xmath39 the hamiltonian in eq.([eq:2 ] ) is @xmath40.\ ] ] the renormalization - group procedure traces out half of the degrees of freedom in the partition function @xcite , @xmath41}\\ = & \mbox{tr}_{\mbox{\tiny odd } } e^{\sum_i^{\mbox{\tiny odd}}\left[-\beta\mathcal{h}_{i-1,i}-\beta\mathcal{h}_{i , i+1}\right]}\\ \simeq&\prod_i^{\mbox{\tiny odd}}\mbox{tr}_ie^{\left[-\beta\mathcal{h}_{i-1,i}-\beta\mathcal{h}_{i , i+1}\right]}\\ = & \prod_{i}^{\mbox{\tiny odd}}e^{-\beta^{\prime } \mathcal{h}^{\prime}_{i-1,i+1}}\\ \simeq&e^{\sum_i^{\mbox{\tiny odd}}\left[-\beta^{\prime}\mathcal{h}^{\prime}_{i-1,i+1}\right]}=e^{-\beta^{\prime}\mathcal{h}^{\prime}}. \end{split}\ ] ] here and throughout this paper primes are used for the renormalized system .
thus , as an approximation , the non - commutativity of the operators beyond three consecutive sites is ignored at each successive length scale , in the two steps indicated by @xmath42 in the above equation .
earlier studies @xcite have established the quantitative validity of this procedure .
the above transformation is algebraically summarized in @xmath43 where @xmath44 are three successive sites .
the operator @xmath45 acts on two - site states , while the operator @xmath46 acts on three - site states .
thus we can rewrite eq.([eq:5 ] ) in matrix form as @xmath47 where state variables @xmath48 , @xmath49 , @xmath50 , @xmath51 , and @xmath52 can be one of the four possible single - site @xmath53 states at each site @xmath54 , namely one of @xmath55 , @xmath56 , @xmath57 , and @xmath58 .
eq.([eq:6 ] ) indicates that the unrenormalized @xmath59 matrix on the right - hand side is contracted into the renormalized @xmath60 matrix on the left - hand side .
we use two - site basis states , @xmath61 , and three - site basis states , @xmath62 , in order to block - diagonalize the matrices in eq.([eq:6 ] ) .
these basis states are the eigenstates of total localized and conduction electron numbers .
the set of @xmath61 and @xmath62 are given in tables [ tab:1 ] and [ tab:2 ] respectively .
the corresponding block - diagonal hamiltonian matrices are given in appendices a and b. .the two - site basis states that appear in eq.([eq:7 ] ) , in the form @xmath63 .
the total localized and conduction electron numbers @xmath64 and @xmath65 , the eigenvalue @xmath66 of the operator @xmath67 defined after eq.([eq:8 ] ) are indicated .
@xmath68 are respectively obtained from @xmath69 by the action of @xmath67 , while the corresponding hamiltonian matrix elements are multiplied by the @xmath66 values of the states . [ cols="^,^,^,^",options="header " , ] c c c c c c c c c c c c c phase ' '' '' & + sink ' '' '' ' '' '' & @xmath15 & @xmath70 & @xmath71 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 & @xmath77 & @xmath78 & @xmath79 & @xmath80 + @xmath81 & @xmath24 & @xmath24 & @xmath82 & @xmath82 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 + @xmath83 & @xmath24 & @xmath84 & @xmath84 & @xmath82 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 + @xmath85 & @xmath24 & @xmath84 & @xmath82 & @xmath84 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 + @xmath86 & @xmath24 & @xmath24 & @xmath84 & @xmath84 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 + co@xmath87 & @xmath84 & @xmath84 & @xmath82 & @xmath82 & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath82 & @xmath84 & @xmath84 & @xmath82 + co@xmath88 & @xmath84 & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath84 + co@xmath89 & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath90 & @xmath82 & @xmath84 & @xmath82 & @xmath91 & @xmath84 & @xmath84 & @xmath82 + co@xmath92 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & @xmath93 & @xmath82 & @xmath84 & @xmath84 & @xmath94 & @xmath84 & @xmath84 & @xmath84 + c c c c c c c c c c c c c phase ' '' '' & + sink ' '' '' ' '' '' & @xmath95 & @xmath96 & @xmath97 & @xmath98 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 + @xmath81 , @xmath86 & @xmath107 & @xmath107 & @xmath108 & @xmath108 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 + @xmath83 , @xmath85 & @xmath107 & @xmath108 & @xmath108 & @xmath108 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 & @xmath107 + co@xmath87 , co@xmath88 & @xmath109 & @xmath109 & @xmath109 & @xmath108 & @xmath110 & @xmath109 & @xmath110 & @xmath109 & @xmath110 & @xmath109 & @xmath109 & @xmath109 + co@xmath89 , co@xmath92 & @xmath109 & @xmath109 & @xmath109 & @xmath108 & @xmath23 & @xmath109 & @xmath110 & @xmath109 & @xmath23 & @xmath109 & @xmath109 & @xmath109 + c c c c c c c c c c c c c c phase ' '' '' & & character + sink ' '' '' ' '' '' & @xmath111 & @xmath112 & @xmath113 & @xmath114 & @xmath115 & @xmath116 & @xmath117 & @xmath118 & @xmath119 & @xmath120 & @xmath121 & @xmath122 & + @xmath81 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & dilute - dilute + @xmath83 & @xmath24 & @xmath24 & @xmath109 & @xmath24 & @xmath23 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & dilute - dense + @xmath85 & @xmath24 & @xmath24 & @xmath24 & @xmath109 & @xmath24 & @xmath23 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath23 & @xmath109 & dense - dilute + @xmath86 & @xmath24 & @xmath109 & @xmath109 & @xmath109 & @xmath23 & @xmath23 & @xmath23 & @xmath109 & @xmath109 & @xmath109 & @xmath123 & @xmath124 & dense - dense + co@xmath87 & @xmath125 & @xmath24 & @xmath126 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & dilute - charge ord .
dilute + co@xmath88 & @xmath125 & @xmath24 & @xmath127 & @xmath24 & @xmath128 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & @xmath24 & dilute - charge ord .
dense + co@xmath89 & @xmath125 & @xmath126 & @xmath126 & @xmath109 & @xmath24 & @xmath23 & @xmath24 & @xmath126 & @xmath24 & @xmath126 & @xmath23 & @xmath109 & dense - charge ord .
dilute + co@xmath92 & @xmath125 & @xmath127 & @xmath127 & @xmath109 & @xmath128 & @xmath23 & @xmath128 & @xmath127 & @xmath129 & @xmath127 & @xmath130 & @xmath131 & dense - charge ord .
dense + c c c c c c c c c c c c c c c c phase ' '' '' & boundary & & relevant & eigenvalue + boundary ' '' '' ' '' '' & type & @xmath15 & @xmath70 & @xmath71 & @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 & @xmath77 & @xmath78 & @xmath79 & @xmath80 & & exponent @xmath132 + co@xmath88/co@xmath89 ' '' '' & 1st & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath84 & @xmath82 & @xmath82 & @xmath84 & @xmath84 & @xmath82 & @xmath133 & 3 + ' '' '' & order & & & & & & & & & & & & & @xmath134 & + co@xmath87/@xmath81 ' '' '' & 2nd & @xmath84 & @xmath84 & @xmath82 & @xmath82 & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath82 & @xmath84 & @xmath84 & @xmath82 & @xmath135 & 0.273873 + ' '' '' & order & & & & & & & & & & & & & & + co@xmath88/@xmath83 ' '' '' & 2nd & @xmath84 & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath84 & @xmath82 & @xmath136 & 0.273873 + ' '' '' & order & & & & & & & & & & & & & & + co@xmath89/@xmath85 ' '' '' & 2nd & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath82 & @xmath84 & @xmath84 & @xmath82 & @xmath137 & 0.273873 + ' '' '' & order & & & & & & & & & & & & & @xmath138 & + co@xmath92/@xmath86 ' '' '' & 2nd & @xmath84 & @xmath84 & @xmath84 & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath84 & @xmath139 & 0.273873 + ' '' '' & order & & & & & & & & & & & & & @xmath140 & + co@xmath87/co@xmath88 ' '' '' & 2nd & @xmath84 & @xmath84 & @xmath82 & @xmath82 & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath82 & @xmath84 & @xmath84 & @xmath82 & @xmath141 & 1.420396 + ' '' '' & order & & & & & & & & & & & & & & + co@xmath89/co@xmath92 ' '' '' & 2nd & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath82 & @xmath84 & @xmath82 & @xmath84 & @xmath84 & @xmath84 & @xmath82 & @xmath142 & 1.420396 + ' '' '' & order & & & & & & & & & & & & & @xmath143 & + with these basis states , eq.([eq:6 ] ) can be rewritten as @xmath144 once written in the basis states @xmath61 , the block - diagonal renormalized matrix has 13 independent elements , which means that renormalization - group transformation of the hamiltonian generates 9 more interaction constants apart from @xmath15 , @xmath70 , @xmath71 , and @xmath72 . in this 13-dimensional interaction space
, the form of the hamiltonian stays closed under renormalization - group transformations .
this hamiltonian is @xmath145 where @xmath67 is a local operator that switches the conduction electron states of sites @xmath18 and @xmath38 : @xmath146 with @xmath147 for @xmath148 and @xmath149 otherwise .
when @xmath150 is applied , further below , to three consecutive sites @xmath18 , @xmath38 , @xmath151 , @xmath152 with @xmath147 for @xmath153 and @xmath149 otherwise .
classical submodel , in the chemical potentials ( upper panel ) and densities ( lower panel ) of the localized and conduction electrons . in phase subscripts throughout this paper ,
the first and second subscripts respectively describe localized and conduction electron densities , as dilute ( d ) or dense ( d ) .
the dashed lines are not phase transitions , but smooth changes between the four different density regions of the disordered ( @xmath154 ) phase . ] to extract the renormalization - group recursion relations , we consider the matrix elements @xmath155 . with @xmath156 , @xmath157 , @xmath158 , and @xmath159 ,
12 out of 16 diagonal elements are independent and , with @xmath160 , only one of the 4 off - diagonal elements is independent , summing up to 13 independent matrix elements .
thus we obtain the renormalized interaction constants in terms of @xmath161 , defining @xmath162 for the diagonal elements and @xmath163 for the only independent off - diagonal element : @xmath164 the matrix elements @xmath161 of the exponentiated renormalized hamiltonian are connected , by eq.([eq:7 ] ) , to the matrix elements , @xmath165 of the exponentiated unrenormalized hamiltonian , @xmath166 the matrix elements @xmath167 can be obtained in terms of the unrenormalized interactions via exponentiating the unrenormalized hamiltonian matrix whose elements are given in appendix b. , @xmath168 phase diagram . ]
equations ( [ eq:9 ] ) and ( [ eq:10 ] ) , together with appendix b , constitute the renormalization - group recursion relations for @xmath39 , in the form @xmath170 , where @xmath171 . to generalize to higher dimension @xmath169
, we use the migdal - kadanoff procedure @xcite , @xmath172 where @xmath173 is the rescaling factor and @xmath80 is the renormalization - group transformation in @xmath39 for the interaction constants vector @xmath174 .
this procedure is exact for @xmath2-dimensional hierarchical lattices @xcite and a very good approximation for hypercubic lattices for obtaining complex phase diagrams .
each phase in the phase diagram has its own ( stable ) fixed point(s ) , which is called a phase sink ( table [ tab:3 ] ) .
all points within a phase flow to the sink(s ) of that phase under successive renormalization - group transformations .
phase boundaries also have their own ( unstable ) fixed points ( table [ tab:4 ] ) , where the relevant exponent analysis gives the order of the phase transition .
thus , the repartition of the renormalization - group flows determine the phase diagram in thermodynamic - field space .
matrix multiplications , along the renormalization - group trajectory , with the derivative matrix of the recursion relations relate the expectation values at the starting point of the trajectory to the expectations values at the phase sink .
the latter are determined ( table iii ) by the left eigenvector , with eigenvalue @xmath175 , of the recursion matrix at the sink , where @xmath173 is the length - rescaling factor of the renormalization - group transformation .
when the expectation values are thus calculated for the points of the phase boundary , the phase diagram in density space is determined.@xcite
the global phase diagram of sfkm is calculated , as described above , for the whole range of the interactions ( @xmath15 , @xmath70 , @xmath72 , @xmath71 ) .
the global phase diagram is thus 4-dimensional .
@xmath176 can be taken as the temperature variable .
we present the calculated global phase diagram in four subsections : the first subsection gives the @xmath9 classical submodel .
the other subsections are devoted to small , intermediate , and large values of the interaction @xmath11 .
we present constant @xmath177 cross sections in terms of the localized and conduction electron chemical potentials @xmath178 and @xmath179and in terms of the localized and conduction electron densities @xmath180 and @xmath181 . , @xmath168 phase diagram .
the coexistence tie lines of the critical endpoints e and of the double critical endpoints e@xmath182 are shown . inside each region
delimited by dotted lines and the endpoint tie lines , phase separation , _
i.e. _ , phase coexistence occurs between phases as identified on this figure . ]
setting the quantum effect to zero , @xmath9 , yields the classical submodel , closed under the renormalization - group flows .
the global flow basins in @xmath178 and @xmath179 are the same for all @xmath70 , given in fig .
[ fig:1 ] .
there exist four regions of a disordered phase within this submodel , which are localized - dilute - conduction - dilute , localized - dilute - conduction - dense , localized - dense - conduction - dilute , and localized - dense - conduction - dense regions , denoted by @xmath81 , @xmath83 , @xmath85 , and @xmath86 .
[ in phase subscripts throughout this paper , the first and second subscripts respectively describe localized and conduction electron densities , as dilute ( d ) or dense ( d ) . ] in the renormalization - group flows , each @xmath154 region is the basin of attraction of its own sink . the dashed lines between the different regions are not phase boundaries , but smooth transitions ( such as the supercritical liquid - gas or up - magnetized - down - magnetized transitions ) , which are controlled by zero - coupling null fixed points.@xcite it should be noted that the suzuki - takano and migdal - kadanoff methods are actually exact for this classical submodel , and yield exactly the same picture as obtained in @xcite . phase diagram . ] phase diagram . ] in this subsection , we present our results for @xmath183 , representative of the weak - interaction regime . the @xmath9 phase diagram of fig .
2 evolves under the introduction of quantum effects via a non - zero hopping strength @xmath15 .
it should be noted that increasing the dimensionless hamiltonian parameter @xmath15 is equivalent to reducing temperature , as in all renormalization - group studies .
the first effect is the decrease and elimination ( left panels of fig .
3 ) of the ( smooth ) passage between the @xmath85 and @xmath83 regions . with this elimination ,
all four regions meet at @xmath184 and @xmath185 , the half filling of both localized and conduction electrons . with increasing @xmath15 ( equivalent to decreasing temperature ) ,
four new , charge - ordered ( co ) phases emerge at @xmath186 .
the co phases occur at and near half filling of conduction electrons for the entire range of localized electron densities .
the co phases grow with increasing @xmath15 ( decreasing temperature ) until saturation at high @xmath15 ( right panels of fig .
all of the new co phases have non - zero hopping density @xmath187 at their phase sinks .
the expectation values at the sinks are evaluated as the left eigenvector of the recursion matrix with eigenvalue @xmath175 @xcite . in the co phases ,
the hopping strength @xmath15 diverges to infinity under repeated renormalization - group transformations ( whereas in the @xmath154 phases , @xmath15 vanishes under repeated renormalization - group transformations ) .
the localized electron density is @xmath188 at the sinks of co@xmath87 and co@xmath88 , while @xmath189 at the sinks of co@xmath89 and co@xmath92 , which throughout the corresponding phases calculationally translates @xcite as low ( @xmath2 ) and high ( @xmath190 ) localized electron densities , respectively .
recall that on phase labels ( co and @xmath154 ) throughout this paper , the first and second subscripts respectively describe localized and conduction electron densities .
the conduction electron density is @xmath191 at the sinks of co@xmath87 and co@xmath89 , while @xmath192 at the sinks of co@xmath88 and co@xmath92 .
the nearest - neighbor conduction electron number correlation is @xmath193 at the sinks of co@xmath87 and co@xmath89 , while @xmath194 at the sinks of co@xmath88 and co@xmath92 .
consequently , for conduction electrons , if a given site is occupied , its nearest - neighbor site is empty at the sinks of co@xmath87 and co@xmath89 .
the co@xmath88 and co@xmath92 phases are connected to the co@xmath87 and co@xmath89 phases by particle - hole interchange on the conduction electrons .
thus , in the co phases , the lattice can be divided into two sublattices with different electron densities .
the behavior at the co sinks therefore indicates charge ordered phases at finite temperatures , as also previously seen in ground - state studies @xcite .
note that this charge ordering is a purely quantum mechanical effect caused by hopping , since the sfkm hamiltonian [ eq.([eq:1 ] ) ] studied here does not contain an interaction between electrons at different sites . in the small @xmath11 regime ,
all phase boundaries around the co phases are second order . as seen in the expanded fig .
4 , all four co phases and all four regions of the @xmath154 phase ( as narrow slivers ) meet at @xmath184 and @xmath185 , half - filling point of both localized and conduction electrons .
all characteristics of the sinks and boundary fixed points are given in tables [ tab:3 ] and [ tab:4 ] .
in this subsection , the phase diagram for @xmath195 , representative of the intermediate - interaction regime , is presented .
[ fig:4 ] gives constant @xmath177 cross - sections .
first - order phase boundaries appear in the central region of the phase diagram , at and near the half filling of both localized and conduction electrons . for low values of @xmath15 ( left panels of fig .
5 ) , equivalent to high temperatures , two first - order phase boundaries , bounded by four critical points c , pinch at a quadruple point q. in the ( left - lower ) density - density phase diagram , four phase separation ( coexistence ) regions mark the first - order phase transitions . inside these regions , coexistence ( phase separation ) occurs between the phases on each side of these regions , as indicated on the figure . the tie line of the quadruple point is shown as a thin straight line .
all four @xmath154 phases coexist ( phase separate ) on this line .
as @xmath15 increases ( temperature decreases ) , the four charge - ordered co phases appear again at @xmath186 , as seen in the leftmost panels of fig .
1 . the co phases again occur at and near half filling of conduction electrons for the entire range of localized electron densities . in the right panels of fig . 5
, the second - order transition lines bounding the co phases terminate at two critical endpoints e @xcite and two double critical endpoints e@xmath182 on the first - order line in the central region ( zoomed in fig . 6 ) .
thus , first - order transitions and phase separation occur between the pairs of @xmath85 and @xmath81,@xmath85 and co@xmath87,co@xmath89 and co@xmath88,co@xmath92 and @xmath83 , @xmath86 and @xmath83 phases , as indicated on fig . 6 , at and near the half filling of both localized and conduction electrons .
the evolution of the phase diagrams between right and left panels of fig . 5 are shown in fig . 1 .
the evolution of the global phase diagram , as the interaction strength is increased , is seen in the phase diagrams in fig .
the co phases emerge again at @xmath186 . with increasing @xmath15 ( decreasing temperature ) , the co phases grow , until saturation seen in fig .
the topology of the phase diagram with five phases stays the same for all @xmath196 .
the constant @xmath177 cross - sections of the phase diagram are given in fig .
7 . for @xmath197 , the double critical endpoints e@xmath182 have split into pairs of simple critical endpoints e , resulting in six separate critical endpoints . for @xmath198 ,
the inner two critical endpoints have merged into a single double critical endpoint . for @xmath199 ,
the double critical endpoint has split into two critical endpoints and the critical lines in the low - density and high - density localized electrons regions have disconnected from each other . in this strong interaction limit , the homogenous ( non - phase - separated ) charge - ordered phases occur again at and near half filling of conduction electrons , but at the low- or high - density limit of the localized electrons . away from these limits ,
the charge - ordered phases occur in coexistence ( phase - separated from ) the disordered phases . at and near the half filling of both localized and conduction electrons , the coexistence of the disordered phases @xmath85 and
@xmath83 occurs .
two sets of three critical lines terminate in separate endpoints , as seen in the zoomed fig .
9 . in this case , a characteristic shape of the density phase diagrams , which we dub _ chimaera _ coexistence , emerges . in the _ chimaera _ phase diagram , coexistence can be found for essentially the entire range of conduction electrons densities or for most of the range of localized electron densities . in the upper - right chemical - potential panel of fig .
7 , the first - order phase boundary exhibits , from the left to right , a sequence of the maximum , minimum , maximum , minimum points ; the four corresponding tie lines are also shown in the lower - right density panel .
these tie lines abut , on one end , very near maxima and minima of the lower and upper branches of the coexistence boundaries , thereby underpinning the distinctive _ chimaera _ topology .
with this research , we have obtained the global phase diagram of the @xmath8 sfkm , which exhibits a fairly rich collection of phase diagram topologies : for the @xmath9 classical submodel , we have obtained disordered ( @xmath154 ) regions , dilute and dense separately for localized and conduction electrons , but no phase transition between them .
the repartition of these regions , delimited by renormalization - group flows , quantitatively stays the same for the whole @xmath11 range and is exactly as obtained in ref .
for the whole @xmath11 range and @xmath200 , the classical submodel phase diagram is perturbed in such a way that regions @xmath81 and @xmath86 intercede between regions @xmath83 and @xmath85 , resulting in the shrinking and disappearing of the @xmath83 to @xmath85 passage .
all @xmath154 regions have vanishing hopping density at their corresponding sinks .
for the whole @xmath11 range , upon increasing @xmath15 ( lowering temperature ) , at @xmath186 four new phases ( co ) emerge with non - zero hopping density of @xmath201 at their sinks .
these co phases are also either dilute or dense , separately , in the localized and conduction electrons ( co@xmath87 , co@xmath88 , co@xmath89 , and co@xmath92 ) and are all charge ordered in the conduction electrons , a wholly quantum mechanical effect . in these co phases
the bipartite lattice is divided into two sublattices of alternating electron density .
the co phases occur at or near the half filling of conduction electrons .
the phase diagrams with all five phases for @xmath196 exhibit different topologies , for the small , intermediate , and large @xmath11 regimes : for the small @xmath11 ( weak - interaction ) regime , all phase boundaries are second order .
all five phases meet at @xmath184 and @xmath202 , the half - filling point of both localized and conduction electrons . for the intermediate @xmath11 ( intermediate - interaction ) regime ,
a first - order phase boundary emerges in the central region of the phase diagram .
this first - order boundary is centered at @xmath184 and is bounded by two critical points c. the second - order lines bounding the co phases terminate at critical endpoints e and double critical endpoints e@xmath182 on the first - order boundary . due to this first - order phase transition at and near the half filling of both localized and conduction electrons ,
a rich variety of phase separation ( phase coexistence ) occurs , as indicated on figs .
1,5,6,7 . for the large @xmath11 ( strong - interaction ) regime , as @xmath11 is increased , the critical endpoints pass through each other by merging and unmerging as double critical endpoints . for large @xmath11 , the co@xmath203 and co@xmath204 phases are detached from the co@xmath205 and co@xmath206 phases , forming two separate bundles , at high- and low - densities of localized electrons respectively .
first - order transitions occur between the variously dense and dilute @xmath154 .
the global phase diagram underpinning all of these cross - sections is decidedly quite complex .
we thank j. l. lebowitz for suggesting this problem to us .
support by the alexander von humboldt foundation , the scientific and technological research council of turkey ( tbitak ) , and the academy of sciences of turkey is gratefully acknowledged .
o.s.s . gratefully acknowledges support from the scientist supporting office tbitak - bideb .
the matrix elements of the block - diagonal renormalized 2-site hamiltonian in the @xmath207 basis are given in eq.([eq : a1 ] ) , where @xmath208 for the 12 independent diagonal elements and @xmath209 for the only independent off - diagonal element : @xmath210
the matrix elements of the block - diagonal unrenormalized 3-site hamiltonian in the @xmath211 basis are given in eq.([eq : a2 ] ) , where @xmath212 for the diagonal elements and @xmath213 for the off - diagonal elements : @xmath214 the matrix elements for the states connected by the exchange of the outer conduction electrons are obtained by multiplication with the eigenvalues @xmath66 of @xmath150 . the matrix elements @xmath167 that enter the recursion relations via eq.([eq:10 ] ) are obtained by exponentiating the block - diagonal hamiltonian given here .
r. ramirez , l. m. falicov , and j. c. kimball , phys .
b * 2 * , 3383 ( 1970 ) .
r. ramirez and l. m. falicov , phys .
b * 3 * , 2425 ( 1971 ) . c. e.
t. gonalves da silva and l. m. falicov , j. phys .
c * 5 * , 906 ( 1972 ) .
m. plischke , phys .
* 28 * , 361 ( 1972 ) .
u. brandt , a. fledderjohann , and g. hlsenbeck , z. phys .
b * 81 * , 409 ( 1990 ) .
p. farkaovsk , phys .
b * 60 * , 10776 ( 1999 ) .
v. zlati , j. k. freericks , r. lemaski , and g. czycholl , phil . mag .
b * 81 * , 1443 ( 2001 ) .
t. minh - tien , phys .
b * 67 * , 144404 ( 2003 ) .
r. lemaski , phys .
b * 71 * , 035107 ( 2005 ) .
p. farkaovsk and h. enarikov , eur .
j. b * 47 * , 517 ( 2005 ) .
p. m. r. brydon and m. gulcsi , phys .
b * 73 * , 235120 ( 2006 ) . v. zlati and j. k. freericks , acta phys . pol .
b * 32 * , 3253 ( 2001 ) . v. zlati and j. k. freericks , in _ open problems in strongly correlated electron systems _ , edited by j. bona , p. prelovek , a. ramak , and s. sarkar , nato arw , sci . ser .
ii , vol . 15 ( kluwer , dordrecht 2001 ) , p. 371 .
v. zlati and j. k. freericks , in _ concepts in electron correlation _ , edited by a. c. hewson and v. zlati , nato arw , sci
110 ( kluwer , dordrecht 2003 ) , p. 287 .
p. miller and j. k. freericks , j. phys . : condens .
matter * 13 * , 3187 ( 2001 ) .
j. k. freericks , b. n. nikoli , and p. miller , phys .
b * 64 * , 054511 ( 2001 ) ; * 68 * , 099901(e ) ( 2003 ) . j. k. freericks , b. n. nikoli , and p. miller , int .
b * 16 * , 531 ( 2002 ) .
r. allub and b. alascio , solid state commun . * 99 * , 613 ( 1996 ) .
r. allub and b. alascio , phys .
b * 55 * , 14113 ( 1997 ) . b. m. letfulov and j. k. freericks , phys . rev .
b * 64 * , 174409 ( 2001 ) . t. v. ramakrishnan , h. r. krishnamurthy , s. r. hassan , and g. venketeswara pai , in _ colossal magnetoresistive manganites _ , edited by t. chatterji ( kluwer academic publishers , dordrecht 2003 ) , p. 417
. h. j. leder , solid state commun .
* 27 * , 579 ( 1978 ) .
j. m. lawrence , p. s. riseborough , and r. d. parks , rep .
* 44 * , 1 ( 1981 ) .
w. hanke and j. e. hirsch , phys .
b * 25 * , 6748 ( 1982 ) .
e. baeck and g. czycholl , solid state commun .
* 43 * , 89 ( 1982 ) .
s. h. liu and k .-
m . ho , phys .
b * 28 * , 4220 ( 1983 ) .
s. h. liu and k .- m .
ho , phys . rev .
b * 30 * , 3039 ( 1984 ) . t. portengen , th .
streich , and l. j. sham , phys .
b * 54 * , 17452 ( 1996 ) . p. m. r. brydon and m. gulcsi , phys .
* 96 * , 036407 ( 2006 ) .
r. koteck and d. ueltschi , commun .
206 * , 289 ( 1999 ) . c. d. batista , phys .
lett . * 89 * , 166403 ( 2002 ) ; * 90 * , 199901(e ) ( 2003 ) .
yin , w. n. mei , c .- g .
duan , h .- q .
lin , and j. r. hardy , phys .
b * 68 * , 075111 ( 2003 ) . c. d. batista , j. e. gubernatis , j. bona , and h. q. lin , phys .
lett . * 92 * , 187601 ( 2004 ) .
d. ueltschi , j. stat .
phys . * 116 * , 681 ( 2004 ) .
k. michielsen and h. de raedt , phys .
b * 59 * , 4565 ( 1999 ) .
d. thanh - hai and t. minh - tien , j. phys . : condens .
matter * 13 * , 5625 ( 2001 ) .
z. gajek and r. lemaski , acta .
phys . pol .
b * 32 * , 3473 ( 2001 ) .
j. wojtkiewicz and r. lemaski , phys .
b * 64 * , 233103 ( 2001 ) .
h. enarikov and p. farkaovsk , int . j. mod .
b * 18 * , 357 ( 2004 ) .
ch . gruber and n. macris , helv .
acta * 69 * , 850 ( 1996 ) .
j. jdrzejewski and r. lemaski , acta phys .
b * 32 * , 3243 ( 2001 ) .
j. k. freericks and v. zlati , rev .
phys . * 75 * 1333 ( 2003 ) .
ch . gruber and d. ueltschi , in _ encyclopedia of mathematical physics _ , edited by j .-
franoise , g. l. naber , and t. s. tsun , vol .
2 ( academic press , oxford 2006 ) , p. 283 . j. k. freericks , ch .
gruber , and n. macris , phys .
b * 53 * , 16189 ( 1996 ) . j. k. freericks , e. h. lieb , and d. ueltschi , phys . rev
. lett . * 88 * , 106401 ( 2002 ) .
j. k. freericks , e. h. lieb , and d. ueltschi , commun .
phys . * 227 * , 243 ( 2002 ) .
p. tomczak , phys .
b * 53 * , r500 ( 1996 ) . p. tomczak and j. richter , phys .
b * 54 * , 9004 ( 1996 ) .
p. tomczak and j. richter , j. phys .
a * 36 * , 5399 ( 2003 ) .
m. hinczewski and a. n. berker , eur .
j. b * 48 * , 1 ( 2005 ) .
m. hinczewski and a. n. berker , eur .
j. b * 51 * , 461 ( 2006 ) .
m. hinczewski and a. n. berker , phys .
b * 78 * , 064507 ( 2008 ) . c. n. kaplan , a. n. berker , and m. hinczewski , phys . rev .
b * 80 * , 214529 ( 2009 ) . c. n. kaplan and a. n. berker , phys . rev .
lett . * 100 * , 027204 ( 2008 ) .
o. s. sariyer , a. n. berker , and m. hinczewski , phys .
b * 77 * , 134413 ( 2008 ) . ch . gruber , j. iwanski , j. jedrzejewski , and p. lemberger , phys .
b * 41 * , 2198 ( 1990 ) .
i. stasyuk , in _ order , disorder and criticality : advanced problems of phase transition theory _ , edited by y. holovatch , vol .
2 ( world scientific , singapore 2007 ) , p. 231 .
|
the global phase diagram of the spinless falicov - kimball model in @xmath0 spatial dimensions is obtained by renormalization - group theory .
this global phase diagram exhibits five distinct phases .
four of these phases are charge - ordered ( co ) phases , in which the system forms two sublattices with different electron densities .
the co phases occur at and near half filling of the conduction electrons for the entire range of localized electron densities .
the phase boundaries are second order , except for the intermediate and large interaction regimes , where a first - order phase boundary occurs in the central region of the phase diagram , resulting in phase coexistence at and near half filling of both localized and conduction electrons .
these two - phase or three - phase coexistence regions are between different charge - ordered phases , between charge - ordered and disordered phases , and between dense and dilute disordered phases .
the second - order phase boundaries terminate on the first - order phase transitions via critical endpoints and double critical endpoints .
the first - order phase boundary is delimited by critical points .
the cross - sections of the global phase diagram with respect to the chemical potentials and densities of the localized and conduction electrons , at all representative interactions strengths , hopping strengths , and temperatures , are calculated and exhibit ten distinct topologies .
pacs numbers : 71.10.hf , 05.30.fk , 64.60.de , 71.10.fd
|
measurements of particle electric - dipole moments ( edms ) @xcite are of significant interest because they provide some of the tightest constraints on extensions to the standard model , such as supersymmetry , that attempt to explain the mechanisms underlying cp violation @xcite . this neutron - edm experiment , and the performance of its cohabiting mercury magnetometer ,
have been discussed in earlier publications @xcite .
the final result presented in this letter incorporates a comprehensive analysis of systematic errors , some of which were undiscovered at the time of the earlier measurements .
the measurement was made with ultracold neutrons ( ucns ) stored in a trap permeated by uniform @xmath0- and @xmath1-fields .
this adds terms @xmath2 and @xmath3 to the hamiltonian determining the states of the neutron .
given parallel @xmath4 and @xmath1 fields , the larmor frequency @xmath5 with which the neutron spin polarization precesses about the field direction is @xmath6 for antiparallel fields , @xmath7 .
thus the experiment aimed to measure any shift in the transition frequency @xmath8 as an applied @xmath9 field alternated between being parallel and then antiparallel to @xmath1 .
a schematic of the apparatus is shown in [ fig : edm_app ] .
the ucns were prepared in a spin - polarized state by transmission through a thin , magnetized iron foil , and entered a cylindrical 21-liter trap within a 1 @xmath10 t uniform vertical magnetic field .
approximately 20 s were needed to fill the trap with neutrons , after which the entrance door was closed pneumatically . the electric field , of approximately 10 kv / cm , was generated by applying high voltage to the electrode that constituted the roof of the trap , while keeping the floor electrode grounded .
the electrodes were made of diamond - like - carbon coated al , and the side wall was sio@xmath11 .
the transition frequency @xmath8 of the neutrons was measured using the ramsey separated oscillatory field magnetic resonance method . during the storage period ,
the neutrons interacted coherently with two 2 s intervals of oscillating magnetic field having a chosen frequency close to the larmor frequency .
the two intervals were separated by a period @xmath12 s of free precession . the last step was to count the number of neutrons @xmath13 and @xmath14 that finished in each of the two polarization states .
this was achieved by opening the entrance door to the trap and allowing the neutrons to fall down onto the polarizing foil , which then acted as a spin analyzer . only those in the initial spin state could pass through to the detector , which was a proportional counter in which neutrons were detected via the reaction @xmath15he @xmath16h@xmath17 during one half of the counting period , an r.f .
magnetic field was applied in the region above the polarizing foil ; this flipped the spins of the neutrons , thereby also allowing those in the opposite spin state to be counted .
each batch cycle yielded about 14,000 ucn counts . within a run
the data - taking operations were cycled continuously for 1 - 2 days . periodically , after a preset number ( normally 16 ) of batches , the direction of @xmath9 was reversed .
all other settings were held constant during a run .
every 10 - 20 runs , was reversed so that half of the full data set was taken with upwards and half with downwards .
we adopt a system as in @xcite where the @xmath18 vector of our @xmath19 axis follows the direction of .
hence , @xmath20 is always positive , while the gravitational displacement of the ucns changes sign .
the magnetometer used the precession frequency of @xmath21 atoms of @xmath22hg ( @xmath23 atoms/@xmath24 ; @xmath25 ) stored simultaneously in the same trap as the neutrons . using ( [ eqn : hnu ] ) for both ucns and hg , and assuming that both experience the same @xmath26 , we find that to first order in the edms @xmath27 , @xmath28 for each run , @xmath29 was obtained from a linear fit to the ratio @xmath30 versus @xmath31 .
( [ eqn : freq_ratio ] ) shows that @xmath29 contains a contribution from @xmath32 .
the true @xmath32 has been shown to be @xmath33 @xcite , so it introduces a systematic error of @xmath34 into @xmath29 .
to the true @xmath35 and @xmath32 within @xmath29 there will also be added coefficients of fractional shifts in @xmath36 and @xmath37 , from other causes , which are linear in @xmath31 and thus constitute additional systematic errors .
the most important of these involves a geometric phase ( gp ) arising when the trapped particles experience a gradient @xmath38 in the presence of @xmath31 @xcite .
fortunately , the centre of gravity of our ucns is @xmath39 cm lower than that of the ( warmer ) hg atoms , so an observed shift of @xmath30 away from @xmath40 gives a measure of the volume - averaged @xmath41 , via the result ( 86 in @xcite ) @xmath42 where the + sign corresponds to downwards . in this experiment
the contribution of the gp effect in the hg to @xmath29 is 50 times larger than the gp effect of the ucns .
writing the gp false contribution to @xmath29 from the hg as @xmath43 , it is shown in @xcite ( 87 ) that @xmath44 where @xmath45 is the trap radius and the + sign again corresponds to downwards .
it follows that we can write @xmath46 where @xmath47 is the true @xmath35 plus all other systematic effects discussed below , and @xmath48 is the value of @xmath49 where @xmath50 .
[ eqn : d_meas ] defines two straight lines , one with positive slope for down and one with a negative slope for up .
the crossing point @xmath51 provides an estimator of @xmath47 free of @xmath52 .
each run was made at a chosen value of @xmath49 by pre - adjusting currents in field - trimming coils .
[ fig : edm_vs_r ] shows the data ( binned for clarity ) for @xmath29 as a function of @xmath49 for each direction of .
the lines represent a least - squares fit to all 554 of the ( unbinned ) run results , using as free parameters the two intercepts and a common absolute slope @xmath53 .
this yields @xmath54 = 652/551 and @xmath55 , which is within 1.3@xmath56 of the expected value of from ( [ eqn : daf_hgn ] ) .
the slope @xmath53 can be altered by a few percent ( although still remaining highly symmetric under reversal ) by various mechanisms including the ucns own gp signal ( a 2% effect ) ; uncertainty in @xmath57 ( 4% ) ; a slight reduction in mean free path due to cavities and grooves in the electrodes as well as to the presence of 10@xmath58 torr of he gas to prevent sparks @xcite ( 1% ) ; and asymmetric surface relaxation of the hg ( up to 5% ) .
there are some processes that can interfere with the above gp error removal essentially any process that changes @xmath49 and/or @xmath52 without conforming to the ratio between the two given by ( [ eqn : daf_hgn ] ) , and where , in addition , the changes differ with the direction of .
first , there are several processes that shift @xmath49 but not @xmath52 . changing @xmath59 , for example , shifts the two lines of 2 in the same direction by the same amount , leaving @xmath47 unaffected but changing @xmath48 .
we note at this point that our final @xmath48 is consistent with the @xmath59 value from the literature ( 1@xmath60 ppm ) @xcite , after allowing for our observed @xmath61 and @xmath62 fields with finite @xmath63 and/or @xmath64 but with ( @xmath65 ( e.g. , a quadrupole aligned with z , with @xmath66 ) , which cause @xmath49 to increase @xcite without contributing to @xmath52 .
any change in such fields when is reversed can result in a differential shift of all @xmath49 values , and thus of the two lines , thereby changing @xmath48 and @xmath47 .
below , we describe our measurements of the differential shift in @xmath48 and state the correction to @xmath47 .
@xmath26-field averaging in the trap is affected by localised loss of ucn and hg particles , and by polarization loss in the presence of the @xmath67 fractional @xmath20 inhomogeneities , which may change with direction .
however , we estimate that the resulting @xmath49 shifts are @xmath68 ppm and @xmath69 ppm for the ucn and hg respectively , and that they will be indistinguishable from the quadrupole shifts . light shifts @xcite in @xmath37 will shift @xmath49 .
they are produced by any small component , parallel to , of the @xmath70hg probe light beam passing through the precessing @xmath22hg atoms .
this component , and the @xmath49 shift , reverse sign on reversal of .
a slight dependence of @xmath49 on the incident light intensity was indeed found , the magnitude @xmath71 ppm being in agreement with theory .
a correction to @xmath29 was made on a run - by - run basis , leading to an overall correction of @xmath72 .
second , there are processes that generate an enhanced @xmath73 .
the field of a permanent magnetic dipole ( pmd ) close to the trap makes a non - uniform @xmath74 and adds an ( enhanced ) gp @xmath75 @xcite to @xmath73 , but shifts @xmath49 in accord with ( [ eqn : freq_ratio ] ) and in opposite senses for the two directions .
the two changes are in a ratio greater than that given by ( [ eqn : daf_hgn ] ) .
this shifts the lines of 2 upwards , adding @xmath76 to @xmath47 , where @xmath77 is the prediction of ( [ eqn : daf_hgn ] ) for the @xmath78 of this pmd .
our fluxgate magnetometer surveys of the trap can not rule out pmd fields of less than 1 nt at 2 cm from the inner surface .
large areas of the trap are sio@xmath11 or al , backed by large voids , and do not come under suspicion ; but the hg and ucn doors involve a heterogeneous collection of small parts close to the trap .
we allow a @xmath47 uncertainty of @xmath79 to allow for an undetected 1 nt pmd at the hg door . in the case of the ucn door
we have better diagnostics .
the trap used when taking edm data has a small cavity in the lower electrode , 4.0 cm deep and 6.8 cm in diameter , sealed from below by the door .
a pmd in the door mechanism can contribute a @xmath74 field to the cavity and to the rest of the trap .
the field in the cavity contributes a strong shift in @xmath49 , while contributing negligibly via the gp to @xmath73 due to the small cavity radius . as a result ,
the lines of 2 are shifted in opposite directions , again adding a systematic error to @xmath47 .
our additional diagnostics came from separate @xmath36 , @xmath37 and @xmath49 measurements ( without @xmath31 fields ) in two auxiliary traps having roofs that could be raised or lowered to change the height @xmath80 .
the traps were built on the same lower electrode and door mechanism that were used for edm data taking .
assuming @xmath81 , one can show that @xmath78 = 0 for a trap height @xmath80 when @xmath37 is the same for roof settings at @xmath82 and @xmath80 .
with such a field established , @xmath48 was measured for several heights @xmath80 .
the resulting forms of @xmath83 for up and down led to the conclusion that there was a dipole field of strength @xmath84 nt penetrating into the door cavity .
they also confirmed the presence of quadrupole fields , which the fluxgate scans showed vary little with @xmath80 .
polarization data from the edm runs further substantiated this : ucns of different energies have different @xmath57s , so the surviving ucn polarization decreases in proportion to @xmath85 , and the values of @xmath48 where the @xmath86 up and down polarizations peak are in excellent agreement with the auxiliary trap results .
the first auxiliary trap used had a smaller radius , 18.5 cm rather than 23.5 cm , and a cavity depth of 6.0 rather than 4.0 cm .
these differences amplify the @xmath49 shifts from a dipole field in the cavity by 1.50 and reduce the quadrupole shifts by a factor 1/1.8 .
our systematic error correction to @xmath47 to allow for the combined door dipole and quadrupole fields is @xmath87 .
the auxiliary trap used to measure @xmath57 by obtaining @xmath49 as a ( linear ) function of a series of known @xmath78 was made as similar to the data - taking trap as possible in dimensions and materials so as to reproduce the same ucn velocity spectrum .
.summary of systematic errors and their uncertainties , in units of 10@xmath88 cm .
[ cols="<,<,<",options="header " , ]
we now consider the systematic errors not involving gp .
if the ucn have a net translational motion , any perpendicular component of the @xmath90 field will be seen in their rest frame as a combination of @xmath90 and @xmath91 fields . during the 130 s ramsey measurement period ,
the ucn ensemble may warm slightly due to vibrations .
this causes the center of mass to rise by up to 0.1 cm . if the volume - averaged angles between @xmath90 , @xmath91 and @xmath92 are each as high as 0.05 radians , the induced false edm will be @xmath93 . in a similar manner to the translational effect ,
any net rotational flow of the ucn in conjunction with a radial component of the @xmath90 field may lead to an induced edm signal .
however , any such flow of ucn is expected to be attenuated by wall collisions before the first ramsey pulse is applied .
we calculate that the maximum error to be expected from this source and from higher - order @xmath94 effects is below @xmath95 .
analysis of the data suggested a possible small correlation between the intensity of the hg reading light and the value of the applied @xmath90 field . through the light shift , this could directly create an edm signal .
however , to within its uncertainty , the dependence of @xmath49 on the light intensity has been removed .
the residual systematic uncertainty from this source is @xmath96 .
there may also be residual effects from @xmath91 field fluctuations .
for example , a dipole - like field @xmath97 originating from the @xmath10-metal in the region of the hv feedthrough would be sensed by both neutrons and hg , but with a difference given by @xmath98 where @xmath99 cm is the distance from the source of the field to the center of the bottle .
thus , fluctuations in @xmath91 that are correlated with the hv can be expected to be compensated up to a factor of about 70 . in order to study this
, the hg and neutron channels were analysed independently .
the neutrons yielded an edm signal of @xmath100 ; the hg , once the gp contribution ( as calculated from the average @xmath101 at which the data were taken ) was subtracted , yielded @xmath102 .
these results are consistent with a common source of magnetic fluctuations correlated with the hv .
we therefore expect the hg compensation to shield us from this systematic effect to a level of @xmath103 .
another possible source of systematic error arises from electrostatic forces , which may move the electrodes slightly . in conjunction with a magnetic field gradient , an hv - dependent shift in the ratio would then appear .
this was sought by looking for an edm - like signal but with a frequency shift proportional to @xmath104 instead of @xmath90 .
the @xmath104 signal was consistent with zero , with an uncertainty of @xmath105 .
if the hv magnitudes were slightly different for the two signs of @xmath90 , this effect would generate a false edm signal .
study of the measured hv and of the charging currents show that the hv magnitude was the same for both polarities to within about 1% .
this systematic uncertainty is therefore @xmath106 .
analysis of the edm as a function of leakage current shows no measurable effect .
leakage currents are typically of order 1 na .
if this current were to travel 10 cm azimuthally around the bottle , the resultant @xmath91 field would result in an apparent edm of @xmath107 .
ac fields are another possible cause of concern .
there is no differential ripple visible on the hv at the level of a few volts .
sampling is done at 5 hz with a bandwidth of 20 khz , so any 50 hz ripple would show up as beats .
this is certainly absent at the level of 50 v , which would give a false edm of @xmath108 .
low - frequency ac fields were sought by means of a pickup coil in conjunction with a phase - sensitive detector .
shifts in @xmath49 from this source at the level of 0.02 ppm could not be ruled out .
cancellations in the corresponding edm signal from reversals of the electric and magnetic fields would reduce any net contribution to below the level of @xmath108 .
two approaches were adopted in studying the data . in the first ,
more straightforward , analysis , only the 293 runs with an uncertainty on @xmath49 of less than 0.05 and with -10 ppm @xmath109 21 ppm were accepted . as the two resulting average @xmath101 values ( 8.948 , 8.943 ppm for up , down respectively ) were almost identical , with approximately equal amounts of data in each field direction , a simple weighted average of the @xmath29 data was used as an estimator of @xmath35 .
the value obtained was @xmath110 with an additional uniformly distributed systematic uncertainty of @xmath111 allocated to it in order to accommodate any potential systematic biases arising from the effects listed in table 1 .
the resulting distribution of possible values implies an upper limit of @xmath112 ( 90% cl ) .
the second analysis began with the [ fig : edm_vs_r ] fitted - lines crossing - point value @xmath113 ( which includes the run - by - run light shift correction ) and then applied the systematic - error corrections given in table 1 . the final result from this approach is @xmath114 in this analysis , the systematic uncertainty is normally distributed .
the result implies an upper limit of @xmath115 ( 90% cl ) .
the data set analysed here , which excludes data that have already been published @xcite , incorporates all neutron edm measurements undertaken between the autumn of 1998 and the end of 2002 .
the results overall may be interpreted as an upper limit on the absolute value of the neutron edm of ( 90% cl ) .
the authors are grateful to n.f .
ramsey for many useful discussions ; to y. chibane , m. chouder , and i.a .
kilvington for their contributions during the development period ; to e.n .
fortson s group for information and components relating to the atomic mercury magnetometry ; and to r. baskin for computer simulation work .
the ill has been generous in provision of neutron facilities .
the work was funded by grants from the uk particle physics and astronomy research council .. support from the rffi , via grant no.03 - 02 - 17305 , is gratefully acknowledged by s.n.i .
|
an experimental search for an electric - dipole moment ( edm ) of the neutron has been carried out at the institut laue - langevin ( ill ) , grenoble .
spurious signals from magnetic - field fluctuations were reduced to insignificance by the use of a cohabiting atomic - mercury magnetometer .
systematic uncertainties , including geometric - phase - induced false edms , have been carefully studied .
two independent approaches to the analysis have been adopted .
the overall results may be interpreted as an upper limit on the absolute value of the neutron edm of ( 90% cl ) .
|
meridional circulation plays a critical role in models of solar dynamo , magnetic flux transport , and the solar cycle @xcite .
it is well established observationally that meridional flow is poleward in each hemisphere with an amplitude of about @xmath0 in the near - surface layers , peaking in strength at mid latitudes @xcite . since mass does not pile up at the poles , it is believed that a return equatorward flow in both hemispheres is operating somewhere in the convection zone , likely near its base .
one of the most promising and complete attempts to measure this meridional circulation was during the graduate work of p. giles @xcite . using the soho spacecraft s michelson doppler imager ( mdi ) helioseismic data
, giles found that the poleward meridional flows continued throughout almost all of the convection zone and that there was indirect evidence of a return equatorward flow near the tachocline of a few @xmath1 .
his methods and analysis imposed a constraint of mass conservation .
thus , the picture that emerged was of two closed circulating flows , one cell in each hemisphere , that diverge from the equator at the surface and converge toward the equator in the deep interior . since then
, other helioseismology studies using a variety of techniques have offered many differing views .
for example , @xcite observe an additional `` cell '' of meridional circulation at mid latitudes near the location of the active sunspot latitudes , which is divergent and varies in time .
also , @xcite found that such a cell has a convergent flow field @xcite .
indeed , large - scale flow profiles ( in both meridional and zonal directions ) have been found to vary rather strongly with the solar cycle , and several studies have found that the amplitude of the flow is anti - correlated with the strength of the cycle ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the latitudinal extent of the surface poleward flow has widely varied in the two previous cycles , and some helioseismic measurements indicate a high - latitude , reverse , _ equatorward _ surface component @xcite . to add to the complexity
, recent observations have shown an increasing polar flow magnitude as one probes deep into the convection zone @xcite , and @xcite place the equatorward return flow at a depth of 70 mm .
recently @xcite observed a new systematic center - to - limb signal in time - distance measurements @xcite , which may play a key role in obtaining reliable deep meridional flow measurements and be one of the sources of the discrepant results mentioned above .
the approach of @xcite was to remove the systematic travel - time shifts found in the east - west measurements , after rotation is removed , from the meridional ( north - south ) measurements .
this correction led to consistent helioseismic measurements using several different observables .
while the source of this signal is not completely understood , it could be related to existing observational limitations like changes of the line formation heights across the solar disk , which produce additional acoustic travel - time shifts in cross - correlation measurements between different locations .
@xcite showed that the effect of the vertical flows from convection in the outer solar convection zone can similarly affect travel - time measurements .
subsequently , @xcite applied their local techniques to measure two meridional circulation cells in the solar convection zone , while @xcite implemented a new global helioseismic analysis that resulted in evidence of a complex multicellular velocity structure .
these new and exciting findings from space - based data present a potentially revised view of these important large - scale flows .
this paper is the first in a series where we explore meridional circulation using time - distance helioseismology applied to global oscillation network group ( gong ) data .
here we describe in detail the travel - time measurement procedure we implement , which is non - standard and differs from the methods of @xcite , for example .
we use more than 600 daily sets of gong velocity images to probe deep into the convection zone . in order to decrease possible geometric and observational artifacts
we have selected dates with a duty cycle of more than 85% and time periods when the solar tilt angle @xmath2 .
these strict requirements substantially decrease the amount of data that can be used .
we assume that the center - to - limb systematic mentioned above is the same in any direction on the solar disk and we compute it only using the equatorial region of the observations .
travel - time differences are computed for north - south flows and corrected by subtracting the east - west signal .
we find strong evidence of a change of sign in the travel - time differences at mid latitudes and depths of about 200 mm beneath the surface , and compelling evidence that travel times may change sign ( thus signaling a flow reversal ) at shallower depths of about 50 - 60 mm . in section [ sec : data ]
we describe the data and analysis procedure , with results and discussion provided in section [ sec : res ] .
the five dotted lines are plotted in @xmath3 increments . the depth scale in mm for the lower turning point of the rays
is given on the @xmath4-axis for reference . ]
@ r c c c c c & min .
@xmath5&max .
@xmath5&@xmath6&@xmath7&depth + index&(deg)&(deg)&(@xmath8 ) & & ( @xmath9 ) + 1&2.75&5.75&52.5&250&0.98 + 2&5.00&8.75&74.3&200&0.96 + 3&8.00&11.75&96.2&150&0.94 + 4&12.50&17.75&122.4&100&0.91 + 5&17.00&23.00&148.7&85&0.88 + 6&23.00&27.50&183.7&70&0.84 + 7&28.25&33.50&223.0&60&0.80 + 8&33.50&38.75&258.0&50&0.77 + 9&38.00&43.25&284.3&45&0.74 + 10&42.50&47.00&319.2&40&0.71 + in this study we utilize gong doppler velocity images , and have selected 652 daily sets of observations with duty cycle higher than 85% during the 2004 - 2012 time period .
the typical travel - time measurements are based on cross - correlation ( ccf ) analysis between two locations on the solar surface separated by certain distances [ @xmath5 ] @xcite .
it is well known that waves with the same phase speed form a wave packet and propagate along approximately the same ray path . in order to increase the signal - to - noise ratio of cross - correlation functions , phase velocity filters
are used to isolate particular wave packets . to infer the meridional flow signal we measure wave travel times for waves propagating in opposite directions along the same ray path that lies between a pair of points ( more precisely , a point and an arc ) at constant longitude . in this case , the travel - time difference for waves propagating in the two opposite directions is sensitive only to the meridional ( north - south ) component of the flow @xcite .
below we itemize the main steps of the analysis and computation procedure : 1 .
daily velocity time series ( 1440 images / day ) were tracked according to the surface synodic differential rotation profile @xcite relative to 12:00 ( noon ) and remapped into @xmath11 coordinates , where @xmath12 and @xmath13 are latitude and longitude , respectively .
spherical harmonic ( sh ) decomposition is performed for each image and sh coefficient time series @xmath14 are obtained for modes @xmath15 and @xmath16 , where @xmath7 is the sh degree and @xmath17 is the azimuthal order .
ten different phase speed filters are employed and designed to cover the approximate depths of 0.98 - 0.70r@xmath18 .
details of their parameters are given in table [ tab : filt ] .
only modes within temporal frequencies between 1.8 and 4.5 mhz are retained .
the approximate ray paths of the filtered wave packets considered here are illustrated in fig .
phase speed filters are applied to the @xmath14 time series for each of the 10 cases to yield @xmath19 .
velocity images are reconstructed from the @xmath19 using the inverse sh decomposition relation : @xmath20 here , @xmath21 are the associated legendre polynomials , @xmath12 denotes latitude , @xmath13 longitude , and @xmath22 .
the fraction of the reconstructed solar disk was @xmath23 in latitude and @xmath24 in longitude relative to the center of the disk .
the ccf between a point and the signal averaged over a 30@xmath25 arc for a given longitude was computed as @xmath26 where @xmath27 is the filtered velocity time series , and @xmath28 is the angular distance between two spatial locations on the solar surface ( @xmath29 ) and ( @xmath30 ) .
arcs in the four cardinal directions are considered .
for each filtered set of data the cross - correlations were computed for some range of travel distances with increments of @xmath31 around the maximum of the first bounce in the ccf . in total 72 correlation functions
are computed for travel distances covering @xmath32 ( see table [ tab : filt ] )
the cross correlations were averaged over about 250 longitude bins in the range @xmath33 .
using smaller bands in longitude provides more proper center - to - limb corrections , but leads to a decrease in the signal - to - noise ratio of cross correlation measurements .
simple comparison using narrower bands showed significant increase of the variance of individual measurements . since we are interested in travel time differences of about 1 s we decided to use a wider longitude range
northward and southward travel times were obtained by fitting a gabor wavelet function to both the positive and negative lags ( @xmath34 ) of the cross correlations as @xmath35 where parameters @xmath36 , @xmath37 , @xmath38 , @xmath39 and @xmath40 are the amplitude , group and phase travel times , mean frequency , and width of wave packet , respectively . 9 .
the difference between two oppositely directed travel times is computed for travel distances @xmath5 corresponding to each phase velocity filter .
we use the convention of `` south minus north '' ( sn ) travel - time differences .
in addition , travel - time differences in the `` east minus west '' ( ew ) direction have also been computed using all of the exact steps of the data processing procedure outlined above .
the travel times for these measurements were averaged over @xmath41 in latitude . a constant shift due to internal solar rotation is observed and removed for each travel distance measurement .
these measurements are used to correct the systematics for the meridional observations .
* * * * the top left panel of fig . [ maps ] shows an average over 652 days of sn travel - time differences presented as a function of latitude and travel distance .
each point at a given travel distance corresponds to the middle position between a point and an arc in our cross - correlation scheme . to avoid very high latitude information where the endpoints of the cross correlations lie ,
the measurements are cut off as a function of distance .
the uncertainties are given in the second column , computed from the dispersion in individual measurements for each longitude and each day .
these are typically a very small percentage of the averaged signal .
signatures of poleward meridional flow in each hemisphere are clearly seen in fig .
[ maps](a ) .
the color convention in this figure is such that blue is consistent with a flow toward the north pole , and red a flow toward the south pole . indeed , in addition to a peak at mid latitudes as expected ,
an increase in the travel - time difference with depth ( i.e. , travel distance ) is also observed .
we expect this to be due to one or several systematics . to explore this further , ew travel times computed from the same dataset are shown in the middle panel of fig .
[ maps ] as a function of longitude on the @xmath42 axis .
the ew map has been symmetrized about the central meridian , as we expect there to be no significant differences between the two ( east / west ) hemispheres since the data have been tracked to account for differential rotation .
these measurements show a similar pattern of center - to - limb variation as the sn map .
@xcite reported a very detailed analysis of travel - time measurements from different observables . since they found that the shape and magnitude of center - to - limb variations is quite different for doppler , continuum , line core and line depth of hmi measurements
, one might conclude that these variations are not caused by any large - scale sub - surface flow of solar origin .
here we follow the same procedure and `` correct '' the sn measurements by subtraction of the ew measurements , the result shown in the bottom panel of fig .
[ maps](a ) .
this correction removes the tendency of the travel times to increase with depth .
furthermore , some evidence of sign changes can be seen .
figure [ cuts ] shows various cuts through the travel - time difference maps .
panels ( a ) and ( c ) are cuts at lower latitudes and shorter travel distances , while panels ( b ) and ( d ) are for mid latitudes and larger travel distances .
these figures confirm that travel - time differences are strongest at mid latitudes around @xmath43 for a range of depths , as has been observed in past studies .
this representation shows a clear yet peculiar asymmetry between the northern and southern hemispheres .
most importantly , we also observe evidence that a change in sign occurs in the measurements for two cases : ( 1 ) at high latitudes in each hemisphere for travel distances greater than about @xmath44 ; and ( 2 ) for large distances for most latitudes greater than about @xmath45 in each hemisphere . indeed , if large - scale flows are responsible for these signals , figs .
[ cuts](a)-(b ) show a tendency for the flow to approach a change of sign at skip distances of @xmath46 for a broad latitude range . at larger distances
this signal then resurrects its poleward sense , eventually reversing again at the deepest probe depths .
this very broadly suggests a multicellular structure as discussed in @xcite and @xcite , who found poleward flows down to @xmath47 , equatorward flows in the @xmath48 range , and then poleward again beneath that .
very recent work by @xcite reports yet another measurement of multicellular structure of the meridional flow using a different , global approach .
we caution that the change in sign at all distances at the maximal latitudes considered here ( most evident in fig . [ maps ] ) could be due to a systematic caused by the solar @xmath49 variation , as demonstrated recently in @xcite
. however , in the measurements here such an artifact is somewhat puzzling since we have restricted the data coverage to epochs when this angle is small .
another possible cause could simply be the use of the ad hoc correction method and any of its inherent systematics .
also evident in the measurements is a north - south hemispheric asymmetry from gong and space - based data that has been noted in previous works ( e.g. , * ? ? ?
* ; * ? ? ?
the real origin of the center - to - limb variation across the solar disk is not well understood at present .
the work of @xcite in explaining it is promising .
anomalous artifacts were even identified as early as @xcite , who consider effects due to the finite speed of light in meridional - flow measurements as causing an overall inflow towards the disk center . while only several seconds , the proper correction for this effect individually actually tends to add to the already unphysical increasing travel - time difference signal with depth as observed in the top panel of fig .
[ maps](a ) .
presumably this systematic is already accounted for in the east - west subtraction correction implemented here , although more confidence in such an approach is certainly needed and is the focus of current work .
nevertheless , we find strong evidence of variations in depth of the large - scale flow in the solar convection zone . the deepest measurements where the ray path is horizontal and less sensitive to surface flows show strong evidence of a change of sign .
the tendency for the travel - time shifts to approach zero at mid - convection zone depths is also intriguing , as one must recall that these measurements are integrated over depth and smoothed to some degree .
inversions may separate the two directional components of the flow and provide amplitudes and a more accurate depth structure .
we have shown robust travel - time measurements of the meridional flow signature in the solar convection zone using gong data and independent measurement techniques .
preliminary evidence of a change of sign , indicating an equatorward return flow at one or several depths in the convection zone is observed , and is approximately consistent with results found in other recently published work .
overall the findings might suggest multicellular structure in the large - scale flows in the sun . only a consistent inversion procedure and a very careful treatment of the systematics can unravel the significance of these trends in the measurements .
a forthcoming paper will show such inversions and discuss the implications for convection - zone dynamics .
we thank an anonymous referee for helpful comments .
is supported by a grant of the uzbek academy of sciences fa - f02-f028 . j.j .
acknowledges support from nasa contract # nnx09ap76a and nsf ast-0849986 .
s.k . was supported by nasa grant nnx11aq57 g to the national solar observatory .
this work utilizes gong data obtained by the nso integrated synoptic program ( nisp ) , managed by the national solar observatory , which is operated by aura , inc . under a cooperative agreement with the national science foundation .
the data were acquired by instruments operated by the big bear solar observatory , high altitude observatory , learmonth solar observatory , udaipur solar observatory , instituto de astrofsica de canarias , and cerro tololo inter - american observatory . , jr . , t. l. , & hanasoge , s. m. 2009 , in astronomical society of the pacific conference series , vol .
416 , solar - stellar dynamos as revealed by helio- and asteroseismology : gong 2008/soho 21 , ed .
m. dikpati , t. arentoft , i. gonzlez hernndez , c. lindsey , & f. hill , 103
|
large - scale plasma flows in the sun s convection zone likely play a major role in solar dynamics on decadal timescales .
in particular , quantifying meridional motions is a critical ingredient for understanding the solar cycle and the transport of magnetic flux .
because the signal of such features can be quite small in deep solar layers and be buried in systematics or noise , the true meridional velocity profile has remained elusive .
we perform time - distance helioseismology measurements on several years worth of gong doppler data .
a spherical harmonic decomposition technique is applied to a subset of acoustic modes to measure travel - time differences to try to obtain signatures of meridional flows throughout the solar convection zone .
center - to - limb systematics are taken into account in an intuitive , yet ad hoc manner .
travel - time differences near the surface that are consistent with a poleward flow in each hemisphere and are similar to previous work are measured .
additionally , measurements in deep layers near the base of the convection zone suggest a possible equatorward flow , as well as partial evidence of a sign change in the travel - time differences at mid - convection zone depths .
this analysis on an independent data set using different measurement techniques strengthens recent conclusions that the convection zone may have multiple `` cells '' of meridional flow .
the results may challenge the common understanding of one large conveyor belt operating in the solar convection zone .
further work with helioseismic inversions and a careful study of systematic effects are needed before firm conclusions of these large - scale flow structures can be made .
|
the discovery with iras of a large infrared excess around vega ( aumann et al .
1984 ) initiated the detailed study of other planetary systems .
we can now use the dust emission from main sequence stars to constrain models for the origin and evolution of these environments ( lagrange , backman & artymowicz 2000 , zuckerman 2001 ) with the long - term goal of developing a more comprehensive understanding of the formation and evolution of planets and related minor bodies .
most previous studies of the dust around main - sequence stars have been restricted to using broad - band infrared fluxes ( e. g. habing et al .
2001 , spangler et al .
2001 ) . here , we report measurements with the infrared spectrograph ( irs , houck et al .
2004 ) on the _ spitzer space telescope _
( werner et al . 2004 ) of 19 main - sequence stars .
we obtained spectra of main sequence stars with previously reported iras excesses with both the short - low ( 5.2 14.5 @xmath1 m ) and long - low ( 14.0 38.0 @xmath1 m ; @xmath6 @xmath3 90 ) modules on the irs . table 1 lists the six targets in our sample with strong infrared excesses .
we did not detect a substantial excess around 13 other stars .
an example of a star without a strong excess is hr 4732 ; its spectrum ( from _ spitzer _ aor 0003579392 ) is shown in figure 1 .
the other stars without strong excesses are hd 16157 ( aor 0003554560 ) , hd 221354 ( aor 0003565842 ) , hr 818 ( aor 0003554816 ) , hr 1338 ( aor 0003555584 ) , hr 2015 ( aor 0003556608 ) , hr 3220 ( aor 0003557888 ) , hr 3862 ( aor 0003558400 ) , hr 5447 ( aor 0003560192 ) , hr 8085 ( aor 0003564544 ) , hr 8549 ( aor 0003588608 ) , hr 8799 ( aor 0003565568 ) and ross 128 ( aor 0003559168 ) . in order to avoid time - consuming peak - up on our relatively bright targets with accurately known positions
, we operated the observatory in irs spectral mapping mode where a 2x3 raster ( spatial x dispersion ) centered on the star is performed .
the slit positions were separated by half of the slit width in the dispersion direction , and by a third of the slit length in the spatial direction .
we carried out the bulk of the reduction and analysis of our spectra with the irs teams smart program ( higdon et al 2004 ) .
we started with intermediate products from the _ spitzer _ irs data - reduction pipeline , that lacked only stray light and flat - fielding corrections .
we extracted point - source spectra for each nod position using a uniformly weighted column profile for which the variable width scales with the width of the instrumental pointspread function .
we also made similar spectral extractions for observations of our photometric standard star , @xmath7 lac ( a1v ) .
then we divided each target spectrum , nod position by nod position , by the spectrum of the standard star , multiplied the quotient by the appropriate template spectrum ( cohen et al 2003 ) , and averaged the resulting spectra from corresponding nod positions .
we plot the resulting spectra of our targets with ir excesses in figure 1 . based upon comparisons of irs spectra of non - variable calibration sources to ground - based , iras , and _ spitzer _ irac flux densities , we estimate that the absolute spectrophotometric accuracy of these spectra is 10% ( 1@xmath8 ) .
it is important to distinguish this absolute photometric accuracy , which describes the vertical positions of the spectra on our plots relative to the flux density scale , from the point - to - point fluctuation in the spectra , which is a measure of the precision of our spectral feature strengths and therefore of our ability to identify spectral features and compare them to our models . in general ,
the accuracy of our spectral feature identifications and feature strengths is limited by the point - to - point scatter ( currently much larger than the system noise , in bright objects ) that is visible in the spectra .
therefore almost all features that significantly exceed the fluctuations , and are wider than the two - pixel irs spectral resolution element , are significant .
the lack of spectral features that we refer to in this paper is based on this criterion .
to interpret the data for the infrared excess , we need luminosities , @xmath9 , and ages , @xmath10 , of the stars .
we take distances , @xmath11 , from _ hipparcos _ , apparent magnitudes from the yale bright star catalog , assume no interstellar reddening , and use the bolometric corrections given by flower ( 1996 ) to infer the total luminosities . for four stars ,
we adopt published estimated ages . for hr 506 , the age is highly uncertain since zuckerman & song ( 2004 ) and decin et al .
( 2000 , 2003 ) estimate 0.3 gyr and 3.0 gyr , respectively . in view of the recently measured lithium abundance of 2.6 on the usual 12-point logarithmic scale ( israelian et al .
2004 ) , we adopt the younger age . for hr 333 and hr 3927 , we assume ages of 0.1 gyr , representative of dusty main sequence a - type stars ( song et al . 2001 ) . the dust continua for @xmath0 @xmath12 20 @xmath1 m show no spectral features .
if the particles are spheres of diameter @xmath13 , then any spectral features that might be present are weakened if @xmath13 @xmath12 @xmath14 ( e. g. spitzer 1978 , wolf & hillenbrand 2003 ) .
our data thus suggest that the diameter of the grains is larger than 10 @xmath1 m , as predicted by models such as those of krivov , mann & krivova ( 2000 ) and found for the particles that produce the zodiacal light in the solar system ( fixsen & dwek 2002 ) .
a standard model of dust debris systems assumes that collisions among a population of large , unseen parent bodies produces smaller , detectable dust particles ( lagrange et al . 2000 , zuckerman 2001 ) . some of the best studied systems such as hr 4796a exhibit a well defined ring of material ( schneider et al .
1999 ) . very approximately ,
for circumstellar envelopes with @xmath15 @xmath12 10@xmath16 ( where @xmath17 is the characteristic luminosity of the infrared excess in the 30 @xmath1 m spectral window , or @xmath18 at 30 @xmath1 m ) , particle - particle collisions typically are the most important grain destruction mechanism ( e. g. krivov , mann & krivova 2000 , wyatt & dent 2001 ) . in these relatively dusty environments , the grains do not migrate far before they are fragmented into sufficiently small pieces that they are driven out of the system by radiation pressure .
thus , the spatial distribution of the dust reproduces the spatial distribution of the parent bodies which might be determined by the history of planet formation in the system . in contrast , in less dusty envelopes with @xmath15 @xmath19 10@xmath16 , mutual collisions are not so important , and the grains spiral inwards under the poynting - robertson drag . in such models ,
the spatial distribution of the grains need not reproduce that of the parent bodies .
motivated by these models of dust debris , we consider two possible fits to the reduced data .
first , we fit the spectra by a function of the form : @xmath20 where @xmath21 and @xmath22 are constants .
the first term in equation ( 1 ) describes the photospheric emission while the second term represents the infrared excess .
we show in figure 1 the comparison of our best fits with the data , and we list in table 1 our values for @xmath22 and @xmath15 where @xmath17 is @xmath23 . equation ( 1 ) fits the spectra for the two stars with the lowest values of @xmath15 , hr 1570 and @xmath0 boo . however , the infrared excesses of four stars require another model since fits using equation ( 1 ) either underestimate the observed photospheric fluxes measured at @xmath0 @xmath24 12 @xmath1 m ( hr 333 and hr 3927 ) or the measured iras and/or iso fluxes ( decin et al .
2000 ) at @xmath0 = 60 @xmath1 m ( hr 506 and hr 1082 ) . our second method to model the infrared excess is to fit a single temperature to the dust particles so the total flux is fit with a function of the form : @xmath25 where @xmath26 is the planck function . in principle
, @xmath27 must equal @xmath21 , but the absolute level of our spectra is somewhat uncertain because of imprecise calibration .
as can be seen in figure 1 , we can fit our infrared spectra adequately with equation ( 2 ) with the values of @xmath28 given in table 1 ranging from 72 k to 114 k. the spectral energy distribution of the infrared excess can test the models . if poynting - robertson drag dominates , the spectrum varies as @xmath29 ( e. g. buitrago & mediavilla 1985 ) , the dust is smeared over a wide range of radii and equation ( 1 ) should fit the data .
in contrast , in a model where collisions limit the lifetime of the particles , then inward drift is not as important and the dust is more concentrated near its birthplace . in environments with @xmath15 @xmath12 10@xmath16 , there might be a dust ring or at least a well defined inner hole to the dust distribution which achieves a characteristic maximum temperature . within the limited irs spectral range , equation ( 2 ) can fit the data .
inner dust truncations might be the result of the formation of low mass companions which might be planets ( kenyon & bromley 2002 ) , but other models can produce rings as well ( takeuchi & artymowicz 2002 ) . also , since @xmath28 is near 100 k , if water ice is an important constituent of the material , there may be substantial grain destruction by sublimation which contributes to the observed deficiency of warm grains ( jura et al .
1998 ) .
we now estimate some parameters for the circumstellar dust .
we assume that the particles radiate like black bodies so that particles at temperature @xmath30 lie at distance @xmath31 from the star : @xmath32 for the two stars where the spectra can be fit by equation ( 1 ) , there is no constraint on the maximum temperature and thus no evidence for any inner truncation in the dust distribution .
there are four stars where we use equation ( 2 ) to fit the data , and we thus find a characteristic grain temperature , which , from equation ( 3 ) , we can translate into a distance from the star . using the values of @xmath28 and @xmath9 given in table 1 ,
we find values for the inner boundary of the dust of 48 au , 11 au , 52 au and 54 au around hr 333 , hr 506 , hr 1082 and hr 3927 , respectively .
consistent with models for those dust debris systems where frequent collisions lead to smaller particles which are then ejected by radiation pressure , the stars with the strongest evidence for inner holes in the dust distribution have the highest values of @xmath15 . if the inner holes are caused by inner planets , then in the future , with sufficiently high angular resolution , it may be possible to identify azimuthal asymmetries in the dust which might indicate gravitational perturbations by this unseen system of planets ( e. g. wyatt 2003 ) .
we can also use our data to make a rough estimate of the minimum mass of the parent bodies , @xmath33 . since poynting - robertson drag provides a maximum grain lifetime , then in an approximate steady state ,
we can write ( chen & jura 2001 ) : @xmath34 this estimate of @xmath33 is insensitive to the grain size and composition ; the minimum values of @xmath33 derived from equation ( 4 ) are shown in table 1 . in systems with @xmath15 @xmath35 10@xmath16 , collisions among dust grains occur often , the poynting - robertson lifetime is thus much longer than the collision time , and we may severely underestimate @xmath33 .
there may also be large objects such as planets that are not experiencing rapid collisions and which do not contribute to the detected dust .
we show in table 1 that the inferred lower bounds to @xmath33 range from 0.004 m@xmath2 to 0.06 m@xmath2 .
the summed mass of the kbos in the solar system is uncertain and is currently estimated to be between 0.02 m@xmath2 and 0.1 m@xmath2 ( bernstein et al . 2003 ; luu & jewitt 2002 ) .
typical analogs of the kuiper belt may have total masses @xmath24 0.1 m@xmath2 ( jura 2004 ) . since we observed stars with notable infrared excesses , we probably selected objects with particularly massive systems of parent bodies . in the solar system ,
the value of the infrared luminosity of the dust is @xmath3 10@xmath36 l@xmath37 ( e. g. fixsen & dwek 2002 ) .
either the mass of parent bodies around our sample of stars is much larger than the inferred minimum , or the collision rate of these objects is substantially greater than in the kuiper belt of the solar system .
two of the stars in table 1 , @xmath0 boo itself and hr 1570 , are @xmath0 boo stars , a subclass of @xmath32% of all main - sequence a - type stars that is defined as having nearly solar abundances of c , n and o , but marked deficiencies of fe elements by factors of 10100 ( paunzen et al . 2002 ) .
the explanation for this abundance pattern is not known .
one theory to account for their surface abundances is that they have preferentially accreted some circumstellar matter ( venn & lambert 1990 ) , and within the star s atmosphere there is an element - dependent chemical separation controlled by gravitational settling and radiative acceleration at the base of the thin outer convection zone of the star ( turcotte & charbonneau 1993 ) . as shown in jura ( 2004 ) , the rate at which circumstellar dust is being produced scales as the infrared excess luminosity such that @xmath38 @xmath3 @xmath39 where @xmath40 is the speed of light .
we therefore find from the results in table 1 that @xmath38 @xmath310@xmath4 g s@xmath5 for both @xmath0 boo and hr 1570 . according to turcotte & charbonneau ( 1993 ) , hydrogen - rich gas infall rates of @xmath4110@xmath42 g s@xmath5 with selective accretion of @xmath4110@xmath4 g
s@xmath5 of metals in dust is required to explain the surface abundances of @xmath0 boo stars .
while more sophisticated treatments are required to the accretion / diffusion models , our results at least show that metal infall at the required rate appears to be occuring around @xmath0 boo stars . in models for the first 100 myr of the solar system , petit , morbidelli & chambers ( 2001 )
have computed that approximately 50% of the @xmath35 m@xmath2 of planetesimals initially within the asteroid belt had their orbits perturbed by jupiter so they hit the sun .
murray et al .
( 2001 ) have argued that such models can be generally extended to other main sequence stars .
assuming that this accretion of @xmath35 @xmath43 10@xmath42 g s@xmath5 of rocky material often occurs , the atmospheres of the a stars should be substantially polluted .
we speculate that the main sequence stars with inner holes in their dust distributions may also be systems with asteroidal belts where gravitational perturbations by giant planets leads to a significant supply of metals to an atmosphere where they would be otherwise depleted .
an infall of large rocks or asteroids may lead to observable variations in the abundance of the elements in the star s atmosphere ( cowley 1977 ) which could be different from the effects of dust inflows occuring around the @xmath0 boo stars .
we have obtained _ spitzer _ mid - infrared spectra of 19 main - sequence stars .
our data are most useful for constraining the evolution of the solid material in regions between @xmath310 au and @xmath360 au from the stars .
we conclude the following : * there are no strong spectral features ; we suggest that the typical diameter of a dust particle is greater than 10 @xmath1 m . * the infrared excesses can be explained by dust arising from collisions between parent bodies whose minimum total mass lies between 0.004 m@xmath2 and 0.06 m@xmath2 .
these derived minimum masses are comparable to the total mass of kbos in the solar system .
* we estimate dust production rates of @xmath310@xmath4 g s@xmath5 around @xmath0 boo and hr 1570 ; selective accretion of this material may help explain their peculiar surface abundances .
* there appear to be inner truncations in the dust clouds at 48 au , 11 au , 52 au and 54 au around hr 333 , hr 506 , hr 1082 and hr 3927 , respectively .
this work is based on observations made with the _
spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under nasa contract 1407 .
support for this work was provided by nasa through contract number 1257184 issued by jpl / caltech .
is supported by the _ spitze_r fellowship program under award 011 808 - 001 .
work at cornell and the university of rochester has been supported by jpl contract 960803 to cornell and cornell subcontract 31419 - 5714 to the university of rochester .
we are grateful to the entire team of scientists and engineers that have made _ spitzer _ and the irs function so well .
aumann , h. h. et al .
1984 , , 278 , 32 bernstein , g. m. , trilling , d. e. , allen , r. l. , brown , m. e. , holman , m. & malhorta , r. 2004 , , in press buitrago , j. & mediavilla , e. 1985 , , 148 , l8 chen , c. h. & jura , m. 2001 , , 560 , l171 cohen , m. , megeath , t. g. , hammersley , p. l. , martin - lui , f. & stauffer , j. 2003 , , 125 , 2645 cowley , c. 1977 , ap .
space sci .
51 , 349 decin , g. , dominik , c. , malfait , k. , mayor , m. & waelkens , c. 2000 , , 357 , 533 decin , g. , dominik , c. , waters , l. b. f. m. & waelkens , c. 2003 , , 598 , 636 fixsen , d. j. & dwek , e. 2002 , , 578 , 1009 flower , p. j. 1996 , , 469 , 355 habing , h. j. et al .
2001 , , 365 , 545 heiter , u. , weiss , w. w. & paunzen , e. 2002 , , 381 , 971 higdon , s. j. u. 2004 , , submitted houck , j. 2004 et al .
, , submitted israelian , g. , santos , n. c. , mayor , m. & rebolo , r. 2004 , , 414 , 601 jura , m. , malkan , m. , white , r. , telesco , c. , pina , r. & fisher , r. s. 1998 , , 505 , 897 jura , m. 2004 , , 603 , 729 kalas , p. , graham , j. r. , beckwith , s. v. w. , jewitt , d. c. & lloyd , j. p. 2002
, , 567 , 999 kenyon , s. j. & bromley , b. c. 2002 , , 577 , l35 bromley , krivov , a. v. , mann , i. & krivova , n. a. 2000 , , 362 , 1127 lagrange , a .- m . , backman , d. e. & artymowicz , p. 2000 , in protostars and planets iv , mannings , v. , boss a. & russell , s. s. eds , ( tucson : university of arizona ) , 639 luu , j. x. & jewitt , d. c. 2002 , , 40 , 63 murray , n. , chaboyer , b. , arras , p. , hansen , b. & noyes , r. w. 2001 , , 555 , 801 paunzen , e. , iliev , i. kh . , kamp , i. & barzova , s. 2002 , , 336 , 1030 petit , j .- m .
, morbidelli , a. & chambers , j. 2001 , icarus , 153 , 338 schneider , g. et al . 1999 , , 513 , l127 song , i. , caillault , j .-
, barrado y navascues , d. & stauffer , j. r. 2001 , , 546 , 352 spangler , c. , sargent , a. i. , silverstone , m. d. , becklin , e. e. , & zuckerman , b. 2001 , , 555 , 932 spitzer , l. 1978 , physical processes in the interstellar medium ( j. wiley : new york ) takeuchi , t. & artymowicz , p. 2001
, , 557 , 990 turcotte , s. & charbonneau , p. 1993
, , 413 , 376 uchida , k. et al .
2004 , , submitted venn , k. a. & lambert , d. l. 1990 , , 363 , 234 werner , m. w. et al .
2004 , , submitted wolf , s. & hillenbrand , l. a. 2003 , , 596 , 603 wyatt , m. c. 2003 , , 598 , 1321 wyatt , m. c. & dent , w. r. f. 2002 , , 334 , 589 zuckerman , b. 2001 , , 39 , 549 zuckerman , b. & song , i. 2004 , , 603 , 738 + + @xmath44 zuckerman & song ( 2004 ) ; @xmath45heiter , weiss & paunzen ( 2002 ) . in this table , @xmath22 ( 10@xmath46 jy hz ) is defined in equation ( 1 ) .
@xmath47 the _ spitzer _ aor data file identification number .
|
we report spectra obtained with the _ spitzer space telescope _ in the @xmath0 = 14 35 @xmath1 m range of 19 nearby main - sequence stars with infrared excesses . the six stars with strong dust emission show no recognizable spectral features , suggesting that the bulk of the emitting particles have diameters larger than 10 @xmath1 m .
if the observed dust results from collisional grinding of larger solids , we infer minimum masses of the parent body population between 0.004 m@xmath2 and 0.06 m@xmath2 .
we estimate grain production rates of @xmath310@xmath4 g s@xmath5 around @xmath0 boo and hr 1570 ; selective accretion of this matter may help explain their peculiar surface abundances .
there appear to be inner truncations in the dust clouds at 48 au , 11 au , 52 au and 54 au around hr 333 , hr 506 , hr 1082 and hr 3927 , respectively .
|
planetary nebulae ( pns ) in the outskirts of elliptical galaxies offer valuable kinematic information that can be used to trace the angular momentum distribution and mass distribution in these galaxies .
the classic example is the study of pns in ngc 5128 by hui et al .
( 1995 ) , later extended by peng et al .
( 2004 ) to 780 pns .
a dark matter halo and significant outer rotation were detected .
a recent dynamical study of 340 globular clusters in ngc 5128 has shown additional evidence of a dark matter halo ( woodley et al . 2007 ) .
another case where a dark matter halo was immediately apparent from pn kinematics is ngc 1344 ( teodorescu et al .
2005 ) . in a previous work ( mndez et al .
2001 ) , hereafter paper i , we presented photometry and slitless radial velocity information for 535 pns in the flattened elliptical galaxy ngc 4697 .
these pns provided kinematic information up to a distance of 3 effective radii from the center of ngc 4697 . the most important result in paper
i was a substantial decline of line - of - sight velocity dispersion ( losvd ) as a function of angular distance from the center , which can be interpreted as indicating the absence of a dark matter halo around this galaxy , if we assume an isotropic velocity distribution
. however , some dark matter can be present if the velocity distribution is radially anisotropic ; we explicitly warned about this alternative interpretation in paper i , and the point has been made in much more detail by dekel et al .
( 2005 ) and de lorenzi et al .
( 2008a ) .
it is important to clarify which of the two interpretations , lack of dark matter or pronounced radial anisotropy , is correct . in the current context , with dark matter
firmly established as one of the basic components of the universe , the existence of one or several substantial , medium - size elliptical galaxies without a dark matter halo ( which is expected to have made possible its very formation ) is surprising . on the other hand ,
if we are able to confirm radial anisotropy , then we have learned about an important constraint concerning the formation of elliptical galaxies .
for example , athanassoula ( 2005 ) has made numerical experiments in which a quick succession of disk mergers is more likely to produce significant radial anisotropy . de lorenzi et al .
( 2008a ) made a detailed dynamical analysis of ngc 4697 , using the pns of paper i and long - slit absorption line kinematic data .
their best - fitting models were radially anisotropic , and included potentials with a variety of dark matter halos .
the models with no dark matter were not consistent with the pn losvds .
however , more pn velocities at larger angular distances from the galaxy s center were required to narrow down the range of acceptable dark matter models . in this paper
we provide more such pn velocities .
romanowsky et al . (
2003 ) showed similarly declining losvds in ngc 3379 , ngc 4494 and ngc 821 , claiming that their `` orbit library analysis '' ruled out extreme orbital anisotropies at least in the case of ngc 3379 .
more recent modeling ( de lorenzi et al
. 2008b ) indicates again that there are mass - shape - anisotropy degeneracies that need to be resolved to reach a firm conclusion .
in addition , a study of two dozens of globular clusters in ngc 3379 ( pierce et al .
2006 ) suggests that some dark matter may be present there
. a recent long - slit spectroscopic study by forestell & gebhardt ( 2008 ) seems to indicate the presence of a dark matter halo around ngc 821 , again in contradiction to what romanowsky et al .
( 2003 ) concluded from the pns in that galaxy . in summary
, the case for no dark matter appears to be weakening , but it is still quite possible that less massive ellipticals have , on average , lower dark matter fractions , i.e. less easily detected dark matter halos ( e.g. napolitano et al . 2005 ) .
we seem to be in a situation that will require more careful studies of several galaxies before we can arrive to more solid conclusions .
the need to collect large pn samples means that we need to detect faint pns .
this can be better done using the largest telescopes available ; so we had a strong motivation to extend the slitless radial velocity method , used in paper i with vlt+fors , to other large telescopes in the northern hemisphere .
the focas spectrograph of subaru proved to be a good choice .
as a first application of the slitless technique with focas and subaru , we have extended our search for pns around ngc 4697 to larger angular distances .
in this paper we present the resulting data and a preliminary analysis . in section 2
we review the basic idea of slitless radial velocities and describe the subaru + focas observations in the outskirts of ngc 4697 .
section 3 describes the reduction and calibration procedures .
section 4 deals with quality control of the slitless radial velocities using observations of the local pn ngc 7293 . in section 5
we give a catalog of 218 pns detected with focas , and compare the focas with fors slitless radial velocities for 158 pns .
having verified the quality of the data , in section 6 we discuss the line - of - sight velocity dispersion , and we analyze a plot of the escape velocity as a function of projected distance from the center of the galaxy .
we adopt a distance of 10.5 mpc for ngc 4697 , taken from paper i. section 7 deals with rotation in the outskirts of ngc 4697 .
section 8 gives a summary of results .
since the slitless radial velocity method was described in paper i , here we give only a brief summary of the basic ideas .
planetary nebulae can be detected in the light of [ o iii ] @xmath05007 using the traditional on - band , off - band filter technique . having taken the on - band image , insertion of a grism as dispersing element produces not only dispersion but also a displacement of all images .
the dispersed images of pns remain point sources , which permits an accurate measurement of the displacement .
calibration of the displacement as a function of wavelength and position in the ccd detector offers an efficient way of measuring radial velocities for all pns in the field , irrespective of their number and distribution .
since no slits are used , there are no light losses , and it is not necessary to go through the complex selection and preparation procedures typical of multi - object slit ( or fiber ) spectroscopy .
of course the quality of the slitless velocities will depend on the seeing and will decrease for very faint sources because of the larger position uncertainty .
the project reported here involves the 8.2 m subaru telescope on mauna kea , hawaii , and its cassegrain imaging spectrograph focas .
we wanted to extend our search for pns to larger angular distances from the center of ngc 4697 , but allowing for partial overlap with the fors e and w fields described in paper i , to be able to compare the radial velocities measured with both instruments .
figure 1 shows the fors fields studied in paper i , and the two focas fields observed for the present work .
the subaru + focas observations were made on the nights of 2004 may 12/13 , 13/14 , and 2005 may 11/12 , 12/13 and 13/14 .
the 2004 nights were partly clouded .
the first two 2005 nights were photometric , and the last one was affected by thin cirrus . during both observing runs the seeing was stable , oscillating around 0.6 arcsec .
focas has been described by kashikawa et al .
the circular field of view of focas has a diameter of 6 arc minutes .
this field is covered by two ccds of 2k@xmath14k ( pixel size 15@xmath2 m ) with an unexposed gap of 5 arcsec between them .
the image scale is 0.1 arcsec pix@xmath3 .
two fields , e and w of ngc 4697 , were observed , as shown in figure 1 .
on - band direct imaging was done through the interference filter n502 , which has the following characteristics : central wavelength 5025 , peak transmission 0.68 , and fwhm 60 .
off - band imaging was done through the broad - band standard visual filter .
dispersed images were obtained by exposing through filter n502 and an echelle grism with 175 grooves mm@xmath3 , operating in the 4th order . in this way
the echelle grism gives a dispersion of 0.5 pix@xmath3 , with good efficiency and a rather small displacement .
dispersed pn images of 0.5 arcsec have a size of 5 pixels on the ccd .
this implies a radial velocity resolution of 140 km s@xmath3 , i.e. the pn s internal velocity field is not resolved . in other words
, pns should always appear as single point sources .
assuming position errors of 0.4 pixel , the expected uncertainty in radial velocity is 10 km s@xmath3 .
table 1 lists the most important ccd images obtained for this project .
it includes the exposures of the focas e and w ngc 4697 fields , the spectrophotometric standard g138 - 31 ( oke 1990 ) , and the calibration and quality control exposures to be described in next section . since there are two ccd images for each field ( chips 1 and 2 ) , for brevity we have listed only the exposure number corresponding to chip 1 . for chip 2 , add 1 to the listed exposure number .
all images were binned 2@xmath11 in the horizontal direction ( perpendicular to dispersion ) .
this was done to increase the signal from the very faint sources we wanted to detect , but at the same time without compromising the spectral resolution . as a consequence of this binning
, each point source looks like an ellipse .
unbinning is required to bring all stars back to their normal appearance .
the basic ccd reductions ( bias subtraction , flat - field correction using twilight flats ) were made using iraf standard tasks .
then we unbinned the images , recovering the original ccd pixel numbers in the horizontal direction . for image registration and combination we first selected , for each of the e and w fields , a pair ( undispersed , dispersed ) of on - band individual images , of the best possible quality , taken consecutively at the telescope , and adopted them as reference images . for field
e , chip 1 , the reference images were 60989 and 60991 . for field
w , chip 1 , the reference images were 52471 and 52473 .
obviously , for chip 2 the reference images were 60990 , 60992 , 52472 and 52474 .
all the other images , including the off - band ones , were registered upon the corresponding reference image .
registration and combination were made using iraf standard tasks .
in particular the dispersed images were registered using the brightest pn images as reference point sources , avoiding in this way any dependence on ambient temperature variations ( see the corresponding discussion in section 3 of paper i ) .
we still have to describe the calibration images for radial velocities .
they were obtained using an engineering mask with punched holes .
the mask produces an array of point sources covering the whole field when illuminated by the internal focas lamps or by any extended astronomical source .
there are almost 1000 calibration points , separated by about 100 pixels in @xmath4 and @xmath5 .
two kinds of calibration images were obtained illuminating the engineering mask with a th - ar comparison lamp : ( u1 ) undispersed , taken through filter n502 ; ( d1 ) dispersed , taken through n502 and the echelle grism .
in addition , there were `` quality control '' images , obtained illuminating the engineering mask with the local pn ngc 7293 : ( u2 ) undispersed , taken through n502 ; ( d2 ) dispersed , taken through n502 and the echelle grism .
figure 2 shows examples of images u2 , d2 and d1 .
a comparison between the left and central images in figure 2 shows the effect of inserting the grism ; comparing to the position of a fixed ccd defect , easily recognizable , it is clear that all the point sources have been displaced upwards .
the right image in figure 2 shows what happens if we keep the grism in the light path and illuminate the mask with the thorium - argon comparison lamp ; now each point source gives a th - ar spectrum .
the positions of the th emission lines permit to obtain the wavelength calibration for each grid point .
the th - ar lamp is ideally suited for our purposes , because its spectrum shows several emission lines within the on - band filter transmission curve . in paper
i we needed to take the calibration images through a broader filter , because the fors spectrographs do not have th - ar lamps .
measurement of the undispersed and dispersed calibration images permits to obtain the displacement produced by the insertion of the grism as a function of wavelength and position on the ccd .
the slitless radial velocity calculation proceeds in the following way : ( 1 ) identify the four calibration points closest to the undispersed position of the pn ; ( 2 ) from the displacement between undispersed and dispersed pn images , calculate the redshifted wavelength of the [ o iii ] emission at each of the four calibration points ; ( 3 ) obtain the final wavelength by bilinear interpolation , weighting the four wavelength values according to the distance from each of the four calibration points to the pn ; ( 4 ) calculate the radial velocity from the final wavelength and the heliocentric correction .
the ngc 7293 images we obtained permit to measure radial velocities in two ways : ( a ) slitless , using images u2 and d2 to obtain the displacement , and using the displacement as a measure of wavelength and therefore velocity , in the way we described at the end of the previous section ; ( b ) `` classical '' , using images d1 and d2 , and treating each mask hole as a slit .
figure 3 shows a comparison of slitless vs. classical radial velocities of ngc 7293 .
there is good agreement , which indicates that the calibration of the displacement works very well .
in addition , the average heliocentric velocity of ngc 7293 from all the grid points turns out to be approximately @xmath630 km s@xmath3 , in good agreement with the known systemic velocity ( @xmath627 km s@xmath3 , measured by meaburn et al . 2005 ) .
furthermore , individual grid points show a range of velocities , from @xmath615 to @xmath650 km s@xmath3 . clearly in some cases the mask holes have isolated gas parcels that are moving predominantly towards or away from us , while in most cases we see gas moving in both directions , so that the unresolved emission line profile gives us a velocity closer to the systemic velocity . from the observed range
we infer a lower limit of about 20 km s@xmath3 for the expansion velocity in [ o iii ] , again perfectly consistent with the 25 km s@xmath3 measured by meaburn et al .
no grid point has given a radial velocity departing more from the systemic velocity than we should expect from the expansion velocity .
from all these results we estimate that the uncertainty in focas slitless velocities is of the order of 10 kms@xmath3 . in the next section
we will provide more evidence about the quality of the focas slitless velocities .
we searched for pns using the traditional blinking method . in the same way as in paper i ,
a pn candidate was found by blinking on - band vs. off - band , and confirmed by blinking undispersed vs. dispersed on - band .
the source was accepted as a pn candidate if it was absent in the off - band image , and if it was a point source .
a total of 218 pns were found , 162 of which had been discovered previously with fors .
consequently , we have added 56 new detections .
most of the new detections ( 36 in fact ) are at angular distances larger than 230 arcsec from the center of ngc 4697 . after measuring the positions of the 218 pns on the undispersed and dispersed images ,
the radial velocities were calculated as described in sections 2 and 3 .
we also measured the jacoby magnitudes @xmath7(5007 ) using the same spectrophotometric standard ( g 138 - 31 , oke 1990 ) and the same procedures described in paper i. for a definition of jacoby magnitudes , see jacoby ( 1989 ) or paper i. the standard g 138 - 31 has @xmath7(5007)=19.36 ( this number depends on the characteristics of the on - band filter ) .
the faintest objects detected both in paper i and in the present search have @xmath8 .
we expect both pn searches to have similar limiting magnitudes , because the telescope sizes , instrumental and ccd efficiencies , and total exposure times , are similar .
j2000 equatorial coordinates for all pns were calculated using a set of astrometric programs written by david tholen and kindly provided by fabrizio bernardi .
a first program , when given an input file with field and camera parameters , identifies all the usno - b1 catalog stars available within the desired field and produces a list of reference stars with rough estimates of their position in the chip .
the next programs make an improved centroid fitting for reference stars , producing an output file with a list of reference stars with pixel coordinates @xmath9 accurate to a few hundredths of a pixel .
the last program performs the final astrometric fit , rejecting outliers and iterating , and provides the pn equatorial coordinates with uncertainties of about 0.5 arcsec . in table 2 we list the following for all detected pns : identification number ; ( @xmath4 , @xmath5 ) coordinates , in arcsec , relative to the optical center of ngc 4697 ; j2000 equatorial coordinates ; jacoby magnitude , @xmath7(5007 ) ; heliocentric radial velocity in km s@xmath3 ; and e and w fors field identification numbers , if the pn is present in the fors pn catalog ( mndez et al .
focas identification numbers that begin with 20 , 21 , 22 and 23 correspond respectively to focas fields w chip 1 , w chip 2 , e chip 1 , and e chip 2 .
a few numbers are missing for a variety of reasons : object 2008 s radial velocity is 281 km s@xmath3 , if the emission line is identified as [ o iii ] @xmath05007 .
since this velocity is incompatible with ngc 4697 , we assume that object 2008 is a background emission - line galaxy , at such a redshift that some other emission line falls into the on - band filter transmission curve ( e.g. ly@xmath10 at @xmath11=3.1 ) .
object 2023 was a rediscovery of 2014 .
object 2049 was not convincingly identified in the grism image , and was rejected as a pn .
since we purposely reobserved many pns discovered previously with fors , we can test the accuracy of equatorial coordinates , jacoby magnitudes , and , most important , radial velocities .
figure 4 shows the differences in declination as a function of the differences in right ascension .
we conclude that our coordinates are reliable to within one arcsec , which is adequate for our purposes .
figure 5 shows the differences in magnitudes @xmath7(5007 ) plotted as a function of the fors @xmath7(5007 ) , which is taken from mndez et al .
there is no systematic difference , and the dispersion increases , as expected , towards fainter magnitudes .
we find no need to rediscuss the photometric results and pnlf distance of 10.5 mpc obtained in paper i. figure 6 ( left panel ) shows fors radial velocities as a function of focas radial velocities for 158 pns .
there is satisfactory agreement .
we should explain that four of the 162 pns could not be measured in paper i , because these pns were too far to the right of the w fors field ( see section 6 in paper i ) .
we have generated simulations of radial velocity sets affected by random errors that follow a gaussian distribution . in paper
i we estimated errors of about 35 or 40 km s@xmath3 for the fors velocities .
the right panel in figure 6 shows what happens if we assume focas errors of the same magnitude .
the dispersion is too large .
the central panel in figure 6 was obtained assuming focas errors of 20 km s@xmath3 .
this is a much better representation of what we observe in the left panel , and we conclude that the focas errors are not larger than 20 km s@xmath3 ; in other words , the focas errors are significantly smaller than the fors errors .
this result is expected .
the focas radial velocities are better mainly because the spectral dispersion is larger : 0.5 per pixel versus the fors dispersion of 1.2 per pixel . for similar position measurement errors
, the larger dispersion gives more accurate velocities .
figure 7 shows the radial velocity difference focas @xmath6 fors as a function of the apparent jacoby magnitude @xmath7(5007 ) for 158 pns .
there is a small systematic difference of about 15 km s@xmath3 , in the sense that fors velocities are larger , but it is well within the uncertainty attributed to fors velocities .
note also that large differences occur only for faint pns .
this is expected , because the quality of the slitless radial velocities depends on the quality of the position measurements .
for very faint sources , the positions in undispersed and dispersed images are less reliable because the sources are barely detectable and the s / n ratio is low .
as mentioned in the introduction , the detailed dynamical analysis by de lorenzi et al .
( 2008a ) showed that the best - fitting models are radially anisotropic , and require some dark matter ; but it is not clear how much .
what we need is more pn radial velocities at large angular distances from the center of ngc 4697 . here
we present a preliminary discussion of the new focas pn velocities . to illustrate how much we have improved the information about pns distant from the center
, we can mention that we have discovered a total ( fors + focas ) of 42 pns at angular distances larger than 280 arcsec from the center of ngc 4697 .
we have fors velocities for 10 of these outlying pns , and we have focas velocities for 41 of them .
figure 8 shows the positions of all fors and focas pns relative to the center of light of ngc 4697 ( left panel ) , and their radial velocities as a function of the @xmath4 coordinates ( right panel ) .
our coordinate system is defined with the @xmath4-axis coincident with the major axis of ngc 4697 , and with the origin at the center of the galaxy , measuring @xmath4 and @xmath5 in arcsec . since the fors data were discussed by de lorenzi et al .
( 2008a ) , here we concentrate mostly on the new focas velocities . in a first attempt we defined six data groups : pns with @xmath12 , @xmath13 , @xmath14 , @xmath15 , @xmath16 , and @xmath17 .
the numbers of pns within each group were , respectively , 14 , 28 , 49 , 47 , 30 and 8 .
we calculated the line - of - sight velocity dispersions ( losvds ) , and corrected them for the small effect of measurement errors by subtracting 10 km s@xmath3 in quadrature .
figure 9 shows the resulting losvd for each group , as a function of average angular distance from the center of ngc 4697 .
we also show dispersions derived from absorption - line spectra along the major axis by binney et al .
( 1990 ) , and dispersions derived from fors pns in paper i. all these data are in good agreement with each other where they overlap .
de lorenzi et al . ( 2008a ) defined zones with elliptical boundaries for losvd calculations ; see their figure 6 .
their outermost datapoint , at a radius of 300 arcsec , has a losvd of 110 km s@xmath3 .
we have calculated the losvd for a group involving all the 48 focas pns that lie outside an ellipse with semi - major and minor axes of 280 and 180 arcsec , respectively .
the resulting losvd is 93@xmath1813 km s@xmath3 , somewhat smaller than the de lorenzi losvd .
we made a similar calculation for the 21 focas pns lying outside a larger ellipse , with semi - major and minor axes of 350 and 225 arcsec .
the corresponding losvd is even smaller : 81@xmath1818 km s@xmath3 .
these values are similar to those shown in figure 9 .
we conclude , first , that our focas losvds in the outskirts of ngc 4697 are not very sensitive to the shape of the regions used to define the samples ; and second , that the new focas data indicate a slightly smaller losvd beyond 250 arcsec than obtained from the fors data in de lorenzi et al .
( 2008a ) .
following their discussion , we conclude that the best fitting models are still those including a not very massive dark matter halo , but it is still difficult to narrow down the range of acceptable dark matter models , because the new focas data lie somewhat closer to the models without dark matter than the earlier fors data .
we also built a new sample by adding , to the 21 focas pns mentioned above , the six other fors pns that lie outside of the larger ellipse .
we did not make any correction for the larger errors of those few fors velocities .
still , the losvd for this new 27-object sample was 95@xmath1818 km s@xmath3 , again lower than the de lorenzi et al .
we can also consider an argument used in the pioneering work of hui et al .
( 1995 ) to confirm the existence of a dark matter halo around ngc 5128 .
when we plot pn radial velocities as a function of angular distance from the center of that galaxy , we find several pns with velocities exceeding the local escape velocity for a hernquist ( 1990 ) model with constant @xmath19 ratio ; see their figure 20 .
the argument was repeated and improved in peng et al .
( 2004 ) ; see their figure 12 . using a constant @xmath19 hernquist model , several pns in ngc 5128 are unbound ; we would expect such stars to escape in a very short time . instead , using a two - component hernquist model , which includes a dark matter halo ,
even the most distant pns become bound , because the escape velocity becomes much larger .
this is a test we can easily apply to ngc 4697 .
figure 10 shows the result .
we calculated the escape velocity for a constant @xmath19 hernquist ( 1990 ) model at a distance @xmath20 from the center of the mass distribution : @xmath21 where @xmath22 is the total mass ( for our test we use two masses , 1.5@xmath23 and 9@xmath24 ) , and @xmath25 is a scale length equal to @xmath26 .
the choice of galaxy mass needs to be explained .
de lorenzi et al . ( 2008a ) found a @xmath19 ratio of about 5 in the r band , which is equivalent to approximately 7.5 in the b band ( see e.g. maraston 1998 ) . knowing the extinction - corrected @xmath27=10.0 , the distance modulus 30.1 from paper i , and the solar @xmath28 absolute magnitude 5.48 , we obtain for ngc 4697 a blue luminosity of @xmath29 , which then gives a mass of @xmath30 .
note that the @xmath31 ratio of 11 determined in paper i needs to be changed , because we have reduced the effective radius @xmath32 of ngc 4697 after the photometric study by de lorenzi et al .
( 2008a ) .
the new value of @xmath32 is 66 arcsec , which is equivalent to 3.36 kpc at the galaxy s distance of 10.5 mpc .
this forces a decrease in the total mass required for a fit of the losvd with the hernquist model .
please refer to formulas ( 32 ) and ( 41 ) in hernquist ( 1990 ) .
the dispersion depends on the mass and the scale length @xmath25 .
if we change @xmath32 , which is proportional to @xmath25 , then we need to change @xmath33 .
this is perhaps more easily appreciated in appendix b of hernquist ( 1990 ) , where all the formulas for the dispersion express it as a function of @xmath34 .
figure 9 shows , for illustrative purposes only , the result of fitting the observed losvd with a hernquist model .
clearly the nmagic modeling by de lorenzi et al . is to be preferred , but it is useful to show that the hernquist model is not far off ; the total mass of this hernquist nodel is 1.5@xmath23 , which implies @xmath35 .
this differs from the nmagic solution by less than 20% in the mass .
let us go back to figure 10 .
we have plotted two escape velocity curves , corresponding to @xmath35 and 5 . for a @xmath31 ratio of 9 we do not find any unbound pns ,
but if we decrease @xmath31 to 5 , then a few pns become unbound .
it should be clear that this kind of diagram can not _
prove _ the absence of dark matter .
since we are dealing with projected distances to the center , some pns can in fact be located at larger distances , where the escape velocity is lower .
it is also possible that some pns have large tangential velocities , so that the total spatial velocity can exceed the escape velocity . in summary , for the larger @xmath19 ratio ,
no dark matter is necessary ; but , for the smaller @xmath19 ratio , the presence of several unbound pns requires the presence of a dark matter halo .
a curve based on the nmagic mass of ngc 4697 falls between the two curves we plotted , and therefore the evidence is inconclusive .
we can only say that the dark matter halo around ngc 4697 is less conspicuous than the one around ngc 5128 .
what is needed now is to repeat the nmagic modeling globally , taking into account both the old fors and new focas pn velocities .
this will require considerable work by other researchers , and is outside the scope of the present paper , which is limited to presenting the new information we have obtained , and exploring its possible significance .
in paper i we reported that the rotation curve of ngc 4697 appeared to drop beyond about 100 arcsec from the center .
we can reinvestigate the rotation at large angular distances from the center of ngc 4697 using our focas pn velocities , which extend further out and have smaller errors than fors velocities .
figure 11 shows histograms for pns progressively more distant from the center .
we show separate histograms for pns with negative and positive @xmath4 coordinates . in the left panel
there is a peak of positive relative radial velocities at negative @xmath4 ,
but anyway the signal of rotation ( most of the galaxy moving toward us at negative @xmath4 , and away from us at positive @xmath4 ) , is clear .
the same sense of rotation is also present in the central panel .
the right panel shows that , at @xmath4 coordinates larger than 300 arcsec in absolute value , the rotation is no longer evident . from paper
i and de lorenzi et al .
( 2008a ) we know already that there is no rotation along the minor axis ( that is to say , around the major axis ) . from long - slit absorption - line studies ( binney et al .
1990 , de lorenzi et al .
2008a ) we know how the stars are rotating .
figure 12 shows that the pns rotate in the same sense .
we have selected pns within 50 arcsec of the major axis , and divided them in eight groups along the major axis , with limits at @xmath6300 , @xmath6200 , @xmath6100 , 0 , 100 , 200 and 300 arcsec .
the two extreme groups turned out to have too few pns , and therefore we removed the @xmath5 limit in those two cases . in order of increasing average @xmath4 ,
the numbers of pns in each group are 14 , 16 , 19 , 20 , 11 , 20 , 13 and 8 .
we averaged the pn radial velocities and plotted the averages as a function of the average @xmath4 of each group .
the drop in rotation velocity at large radii is clear .
it seems that ngc 4697 is hardly rotating at all at @xmath36 .
the rotational signal must be somewhat diluted , because we were forced to include pns with rather large @xmath5 coordinates ; but still the behavior of rotation is obvious .
figures 11 and 12 confirm what we suspected in paper i. in relation to this lack of outer rotation , it may be interesting to mention a recent study of the s0 galaxy ngc 1023 ( noordermeer et al .
2008 ) , which has revealed a similar drop in the outer rotation velocity .
sufficiently massive dark matter halos are usually expected to generate significant rotation in the outskirts , as discussed by de lorenzi et al .
( 2008a ) .
indeed , for example ngc 5128 shows clear evidence of outer rotation ( woodley et al .
2007 ) . that this is not the case in ngc 4697 should not be interpreted as necessarily indicating the lack of a dark matter halo .
instead , it may be a problem of how the specific angular momentum is distributed in the system , independently of its potential well and , by this , its total gravitational mass . in the same way as with the interpretation of the losvd
, the interpretation of rotation will require more careful global modeling with nmagic or similar codes .
for the moment , the conclusion we want to extract is that there appears to exist a variety of cases concerning how much angular momentum resides in the outskirts of elliptical galaxies .
pns are a good tool that will allow to collect angular momentum information for several other galaxies .
we hope that the information derived from pn radial velocities will make it possible to explore if this variety in angular momentum distribution can be traced to different galaxy formation mechanisms involving mergers .
we have demonstrated how subaru and focas can be used to obtain accurate slitless radial velocities of extragalactic pns . working in two fields in the outskirts of ngc 4697
, we discovered and provided radial velocities for 218 pns , of which 162 had been previously observed with vlt+fors .
where possible , we compared the focas radial velocities with those previously obtained with fors , finding good agreement .
we now have pn kinematic information extending out to 5 effective radii from the center of ngc 4697 ( its effective radius is 66 arcsec ) . in particular , we have significantly improved the kinematic information at angular distances larger than 230 arcsec from the center of ngc 4697 .
we have found the following : ( 1 ) the losvd at radii between 300 and 400 arcsec from the center of ngc 4697 is below 100 km s@xmath3 , i.e. a bit lower than obtained from the much smaller fors outer pn sample by de lorenzi et al .
( 2008a ) .
this makes it difficult to narrow down the range of acceptable dark matter models .
at least we can say that the dark matter halo around ngc 4697 is not as conspicuous as the one around ngc 5128 .
( 2 ) the rotational signal , which is obvious close to the center , is no longer visible at about 5 effective radii from the center . by comparison
to ngc 5128 , it seems that much less angular momentum resides in the outskirts of ngc 4697 .
we need similar kinematic information about more elliptical galaxies . a more quantitative interpretation of these results concerning ngc 4697 will require careful global modeling of the full , enlarged , focas+fors pn data set , plus existing absorption - line kinematic data , using nmagic or similar codes , following procedures described by de lorenzi et al .
( 2008a ) .
this work was supported by the national science foundation under grant no . 0307489 .
it is a pleasure to acknowledge the help provided by the subaru staff , in particular the support astronomers youichi ohyama , takashi hattori and kentaro aoki .
we would like to thank fabrizio bernardi for his help using david tholen s astrometric programs for the calculation of j2000 equatorial coordinates .
we also acknowledge comments by r. saglia , o. gerhard and j. barnes , as well as some very useful remarks by an anonymous referee .
athanassoula , e. 2005 , in aip conference proceedings 804 , planetary nebulae as astronomical tools , eds .
r. szczerba , g. stasinska & s.k .
gorny , american institute of physics , p. 333
binney , j. j. , davies , r. l. , & illingworth , g. d. 1990 , apj , 361 , 78 de lorenzi , f. , gerhard , o. , coccato , l. , et al .
2008b , astro - ph arxiv:0804.3350 de lorenzi , f. , gerhard , o. , saglia , r.p . ,
2008a , mnras , 385 , 1729 dekel , a. , stoehr , f. , mamon , g.a . ,
2005 , nature , 437 , 707 forestell , a.d . , &
gebhardt , k. 2008 , astro - ph arxiv:0803.3626 hernquist , l. 1990 , apj , 356 , 359 hui , x. , ford , h.c . ,
freeman , k.c . , & dopita , m.a .
1995 , apj , 449 , 592 jacoby , g.h .
1989 , apj , 339 , 39 kashikawa , n. , aoki , k. , asai , r. , et al .
2002 , pasj , 54 , 819 maraston , c. 1998 , mnras , 300 , 872 meaburn , j. , boumis , p. , lpez , j.a . ,
2005 , mnras , 360 , 963 mndez , r.h . ,
riffeser , a. , kudritzki , r.p . , et al .
2001 , apj , 563 , 135 ( paper i ) mndez , r.h . , teodorescu , a.m. , & kudritzki , r.p .
2008 , apjs , 175 , 522 napolitano , n.r . ,
capaccioli , m. , romanowsky , a.j . , et al .
2005 , mnras , 357 , 691 noordermeer , e. , merrifield , m.r . , coccato , l. , et al .
2008 , mnras , 384 , 943 oke , j.b .
1990 , aj , 99 , 1621 peng , e.w .
, ford , h.c . , & freeman , k.c .
2004 , apj , 602 , 685 pierce , m. , beasley , m.a . ,
forbes , d.a .
, et al . , 2006 , mnras , 366 , 1253 romanowsky , a.j . , douglas , n.g . , arnaboldi , m. , et al .
2003 , science , 301 , 1696 teodorescu , a.m. , mndez , r.h . ,
saglia , r.p . , et al .
2005 , apj , 635 , 290 woodley , k.a . , harris , w.e . ,
beasley , m.a .
2007 , aj , 134 , 494 lrlrc ngc 7293 + mask & on - band & 52385 & 200 & 1.88 + ngc 7293 + mask & on + grism & 52387 & 300 & 1.83 + th - ar + mask & on + grism & 52391 & 10 & 1.77 + ngc 4697 w & off - band & 52445 & 120 & 1.17 + ngc 4697 w & on - band & 52447 & 1200 & 1.15 + ngc
4697 w & on + grism & 52449 & 1800 & 1.12 + th - ar + mask & on + grism & 52453 & 10 & 1.11 + th - ar + mask & on - band & 52455 & 4 & 1.11 + ngc 4697 w & off - band & 52457 & 120 & 1.11 + ngc 4697 w & on - band & 52459 & 1200 & 1.11 + ngc 4697 w & on + grism & 52461 & 1800 & 1.12 + th - ar + mask & on + grism & 52465 & 10 & 1.14 + th - ar + mask & on - band & 52467 & 4 & 1.14 + ngc 4697 w & off - band & 52469 & 120 & 1.15 + ngc
4697 w & on - band & 52471 & 1200 & 1.17 + ngc 4697 w & on + grism & 52473 & 1800 & 1.23 + ngc 4697 w & off - band & 52481 & 120 & 1.31 + ngc 4697 w & on - band & 52483 & 1200 & 1.36 + ngc 4697 w & on + grism & 52485 & 1800 & 1.70 + ngc 4697 e & off - band & 60793 & 120 & 1.24 + ngc 4697 e & on - band & 60795 & 1200 & 1.21 + ngc 4697 e & on + grism & 60797 & 1800 & 1.16 + ngc 4697 e & off - band & 60803 & 120 & 1.13 + ngc 4697 e & on - band & 60805 & 1200 & 1.12 + ngc 4697 e & on + grism & 60807 & 1800 & 1.11 + ngc 4697 e & off - band & 60813 & 120 & 1.12 + ngc 4697 e & on - band & 60815 & 1200 & 1.13 + ngc 4697 e & on + grism & 60817 & 1800 & 1.16 + ngc 4697 e & off - band & 60823 & 120 & 1.22 + ngc 4697 e & on - band & 60825 & 1200 & 1.26 + ngc 4697 e & on + grism & 60827 & 1800 & 1.35 + ngc 4697 e & on + grism & 60833 & 2000 & 1.59 + g 138 - 31 & on - band & 60843 & 60 & 1.02 + g 138 - 31 & on - band & 60845 & 120 & 1.02 + g 138 - 31 & on - band & 60847 & 60 & 1.02 + g 138 - 31 & on - band & 60849 & 60 & 1.02 + g 138 - 31 & on - band & 60851 & 120 & 1.02 + g 138 - 31 & on - band & 60879 & 60 & 1.34 + g 138 - 31 & on - band & 60881 & 120 & 1.36 + g 138 - 31 & on - band & 60883 & 60 & 1.38 + g 138 - 31 & on - band & 60885 & 120 & 1.40 + ngc 4697 w & off - band & 60967 & 120 & 1.24 + ngc 4697 w & on - band & 60969 & 1200 & 1.21 + ngc 4697 w & on + grism & 60971 & 1800 & 1.16 + ngc 4697 w & off - band & 60977 & 120 & 1.12 + ngc 4697 w & on - band & 60979 & 1200 & 1.12 + ngc 4697 w & on + grism & 60981 & 1800 & 1.11 + ngc 4697 e & off - band & 60987 & 120 & 1.12 + ngc
4697 e & on - band & 60989 & 1200 & 1.13 + ngc 4697 e & on + grism & 60991 & 1800 & 1.16 + th - ar + mask & on + grism & 60993 & 10 & 1.20 + th - ar + mask & on - band & 60995 & 4 & 1.20 + ngc 4697 e & off - band & 60997 & 120 & 1.21 + ngc 4697 e & on - band & 60999 & 1200 & 1.24 + ngc 4697 e & on + grism & 61001 & 1800 & 1.35 + ngc 4697 w & off - band & 61011 & 120 & 1.55 + ngc 4697 w & on - band & 61013 & 1200 & 1.65 + ngc 4697 w & on + grism & 61015 & 2000 & 1.96
+ g 138 - 31 & on - band & 61017 & 120 & 1.02 + g 138 - 31 & on - band & 61019 & 120 & 1.02 + g 138 - 31 & on - band & 61021 & 120 & 1.02 + g 138 - 31 & on - band & 61023 & 120 & 1.03 + ngc 4697 e & off - band & 61081 & 140 & 1.24 + ngc
4697 e & on - band & 61083 & 1400 & 1.20 + ngc 4697 e & on + grism & 61085 & 2200 & 1.15 + ngc 4697 e & off - band & 61091 & 140 & 1.12 + ngc 4697 e & on - band & 61093 & 1400 & 1.11 + ngc 4697 e & on + grism & 61095 & 2200 & 1.12 + ngc 4697 w & off - band & 61101 & 140 & 1.13 + ngc 4697 w & on - band & 61103 & 1400 & 1.15 + ngc 4697 w & on + grism & 61105 & 2200 & 1.22 + ngc 4697 w & off - band & 61111 & 140 & 1.30 + ngc 4697 w & on - band & 61113 & 1400 & 1.36 + ngc 4697 w & on + grism & 61115 & 2200 & 1.54 + ngc 4697 w & on + grism & 61119 & 2200 & 1.93 + rrrrrrrrrrrrr 2001 & 376 & -2 & 12 & 48 & 13.13 & -5 & 50 & 43.3 & 27.02 & 1256 & -1 & -1 + 2002 & 378 & -128 & 12 & 48 & 16.59 & -5 & 52 & 38.1 & 27.01 & 1240 & -1 & -1 + 2003 & 386 & -129 & 12 & 48 & 16.12 & -5 & 52 & 42.5 & 26.98 & 1282 & -1 & -1 + 2004 & 287 & -191 & 12 & 48 & 23.89 & -5 & 52 & 56.8 & 26.67 & 1314 & -1 & -1 + 2005 & 312 & -16 & 12 & 48 & 17.45 & -5 & 50 & 28.9 & 27.84 & 1134 & -1 & -1 + 2006 & 338 & -142 & 12 & 48 & 19.37 & -5 & 52 & 34.1 & 28.33 & 1362 & -1 & -1 + 2007 & 286 & -13 & 12 & 48 & 18.91 & -5 & 50 & 15.3 & 25.90 & 1367 & -1 & 1746 + 2009 & 281 & -69 & 12 & 48 & 20.79 & -5 & 51 & 3.5 & 26.49 & 1178 & -1 & 1742 + 2010 & 278 & -94 & 12 & 48 & 21.71 & -5 & 51 & 25.4 & 27.51 & 1400 & -1 & -1 + 2011 & 281 & -136 & 12 & 48 & 22.70 & -5 & 52 & 4.1 & 27.60 & 1251 & -1 & -1 + 2012 & 260 & -155 & 12 & 48 & 24.48 & -5 & 52 & 12.9 & 27.90 & 1346 & -1 & -1 + 2013 & 250 & -130 & 12 & 48 & 24.40 & -5 & 51 & 45.7 & 28.06 & 1326 & -1 & -1 + 2014 & 236 & -80 & 12 & 48 & 23.82 & -5 & 50 & 54.6 & 26.47 & 1218 & -1 & 1734 + 2015 & 260 & -61 & 12 & 48 & 21.84 & -5 & 50 & 48.0 & 26.95 & 1370 & -1 & 1743 + 2016 & 269 & -34 & 12 & 48 & 20.52 & -5 & 50 & 27.2 & 26.93 & 1312 & -1 & 1745 + 2017 & 271 & -44 & 12 & 48 & 20.71 & -5 & 50 & 36.5 & -1 .
& 1450 & -1 & 1744 + 2018 & 239 & -8 & 12 & 48 & 21.59 & -5 & 49 & 50.9 & 27.43 & 1208 & -1 & -1 + 2019 & 236 & -5 & 12 & 48 & 21.71 & -5 & 49 & 46.5 & 27.97 & 1309 & -1 & -1 + 2020 & 224 & -38 & 12 & 48 & 23.38 & -5 & 50 & 11.6 & 27.09 & 1472 & -1 & 1736 + 2021 & 216 & -28 & 12 & 48 & 23.62 & -5 & 49 & 59.0 & 25.97 & 1332 & -1 & 1737 + 2022 & 211 & -45 & 12 & 48 & 24.37 & -5 & 50 & 12.8 & 27.19 & 1280 & -1 & 1735 + 2024 & 235 & -92 & 12 & 48 & 24.26 & -5 & 51 & 4.5 & 27.16 & 1368 & -1 & 1733 + 2025 & 220 & -137 & 12 & 48 & 26.43 & -5 & 51 & 40.1 & 26.03 & 1265 & -1 & 1732 + 2026 & 196 & -150 & 12 & 48 & 28.23 & -5 & 51 & 41.1 & 27.51 & 1312 & -1 & -1 + 2027 & 191 & -80 & 12 & 48 & 26.55 & -5 & 50 & 35.9 & 27.23 & 1135 & -1 & 1722 + 2028 & 202 & -45 & 12 & 48 & 24.90 & -5 & 50 & 9.0 & 26.05 & 1315 & -1 & 1723 + 2029 & 187 & -16 & 12 & 48 & 25.01 & -5 & 49 & 36.3 & 27.57 & 1399 & -1 & -1 + 2030 & 184 & -51 & 12 & 48 & 26.20 & -5 & 50 & 6.1 & 27.10 & 1322 & -1 & -1 + 2031 & 165 & -120 & 12 & 48 & 29.26 & -5 & 51 & 0.8 & 26.56 & 1286 & -1 & 1700 + 2032 & 130 & -27 & 12 & 48 & 28.79 & -5 & 49 & 21.4 & 26.15 & 1430 & -1 & 1668 + 2033 & 153 & -21 & 12 & 48 & 27.18 & -5 & 49 & 25.8 & 25.96 & 1234 & -1 & 1707 + 2034 & 141 & -65 & 12 & 48 & 29.19 & -5 & 50 & 1.0 & 26.16 & 1097 & -1 & 1703 + 2035 & 126 & -78 & 12 & 48 & 30.46 & -5 & 50 & 6.5 & 25.64 & 1243 & -1 & 1661 + 2036 & 108 & -38 & 12 & 48 & 30.41 & -5 & 49 & 22.8 & 26.51 & 1193 & -1 & 1663 + 2037 & 198 & -13 & 12 & 48 & 24.25 & -5 & 49 & 38.3 & 27.44 & 1517 & -1 & -1 + 2038 & 74 & -25 & 12 & 48 & 32.14 & -5 & 48 & 55.9 & 25.72 & 1348 & -1 & 1627 + 2039 & 108 & -10 & 12 & 48 & 29.62 & -5 & 48 & 57.6 & 26.29 & 1365 & -1 & 1666 + 2040 & 92 & -27 & 12 & 48 & 31.10 & -5 & 49 & 5.3 & 26.85 & 1400 & -1 & 1664 + 2041 & 126 & -21 & 12 & 48 & 28.84 & -5 & 49 & 15.1 & 26.63 & 1464 & -1 & 1667 + 2042 & 85 & -51 & 12 & 48 & 32.21 & -5 & 49 & 24.5 & 26.20 & 953 & -1 & 1623 + 2043 & 120 & -54 & 12 & 48 & 30.15 & -5 & 49 & 41.7 & 27.67 & 1449 & -1 & -1 + 2044 & 104 & -54 & 12 & 48 & 31.09 & -5 & 49 & 35.0 & 27.08 & 1425 & -1 & 1662 + 2045 & 120 & -103 & 12 & 48 & 31.52 & -5 & 50 & 26.9 & 27.30 & 1101 & -1 & 1660 + 2046 & 147 & -84 & 12 & 48 & 29.34 & -5 & 50 & 21.2 & 26.78 & 1312 & -1 & 1701 + 2047 & 133 & -78 & 12 & 48 & 30.01 & -5 & 50 & 9.1 & 26.85 & 1222 & -1 & 1702 + 2048 & 156 & -72 & 12 & 48 & 28.48 & -5 & 50 & 13.5 & 27.67 & 1245 & -1 & -1 + 2050 & 69 & -13 & 12 & 48 & 32.10 & -5 & 48 & 43.1 & -1 .
& 1407 & -1 & 1628 + 2051 & 81 & -14 & 12 & 48 & 31.38 & -5 & 48 & 49.7 & 26.77 & 1388 & -1 & 1630 + 2101 & 273 & 112 & 12 & 48 & 16.16 & -5 & 48 & 15.8 & 26.93 & 1261 & -1 & -1 + 2102 & 270 & 102 & 12 & 48 & 16.62 & -5 & 48 & 23.9 & 27.53 & 1207 & -1 & -1 + 2103 & 279 & 100 & 12 & 48 & 16.16 & -5 & 48 & 29.7 & 27.12 & 1540 & -1 & -1 + 2104 & 329 & 89 & 12 & 48 & 13.42 & -5 & 49 & 0.2 & 27.61 & 1375 & -1 & -1 + 2105 & 313 & 60 & 12 & 48 & 15.18 & -5 & 49 & 20.5 & 26.30 & 1270 & -1 & -1 + 2106 & 356 & 162 & 12 & 48 & 9.72 & -5 & 48 & 6.1 & 27.15 & 1276 & -1 & -1 + 2107 & 281 & 164 & 12 & 48 & 14.18 & -5 & 47 & 32.2 & 28.16 & 1335 & -1 & -1 + 2108 & 278 & 30 & 12 & 48 & 18.18 & -5 & 49 & 33.1 & 28.45 & 1329 & -1 & -1 + 2109 & 241 & 66 & 12 & 48 & 19.41 & -5 & 48 & 44.2 & 26.92 & 1380 & -1 & 1738 + 2110 & 225 & 69 & 12 & 48 & 20.30 & -5 & 48 & 34.7 & 27.67 & 1339 & -1 & 1739 + 2111 & 211 & 79 & 12 & 48 & 20.87 & -5 & 48 & 19.9 & -1 .
& 1185 & -1 & 1740 + 2112 & 199 & 67 & 12 & 48 & 21.94 & -5 & 48 & 25.4 & 26.23 & 1359 & -1 & 1728 + 2113 & 207 & 36 & 12 & 48 & 22.31 & -5 & 48 & 57.3 & 26.08 & 1290 & -1 & 1727 + 2114 & 207 & 26 & 12 & 48 & 22.63 & -5 & 49 & 6.2 & 27.40 & 1205 & -1 & 1725 + 2115 & 194 & 31 & 12 & 48 & 23.24 & -5 & 48 & 55.9 & 27.65 & 1403 & -1 & 1726 + 2116 & 203 & 15 & 12 & 48 & 23.14 & -5 & 49 & 14.5 & 27.44 & 1125 & -1 & 1724 + 2117 & 186 & 45 & 12 & 48 & 23.35 & -5 & 48 & 40.5 & 27.68 & 1316 & -1 & -1 + 2118 & 157 & 37 & 12 & 48 & 25.32 & -5 & 48 & 35.7 & 27.04 & 1349 & -1 & 1714 + 2119 & 165 & 41 & 12 & 48 & 24.69 & -5 & 48 & 34.9 & 26.48 & 1204 & -1 & 1715 + 2120 & 175 & 108 & 12 & 48 & 22.20 & -5 & 47 & 38.6 & 27.30 & 1511 & -1 & 1729 + 2121 & 168 & 85 & 12 & 48 & 23.29 & -5 & 47 & 56.0 & 27.27 & 1339 & -1 & 1717 + 2122 & 159 & 66 & 12 & 48 & 24.39 & -5 & 48 & 9.3 & 27.88 & 1437 & -1 & 1716 + 2123 & 125 & 72 & 12 & 48 & 26.25 & -5 & 47 & 49.7 & 26.81 & 1401 & -1 & 1686 + 2124 & 173 & 70 & 12 & 48 & 23.38 & -5 & 48 & 11.8 & 27.99 & 1323 & -1 & -1 + 2125 & 114 & 116 & 12 & 48 & 25.67 & -5 & 47 & 5.6 & 26.39 & 1357 & -1 & 1692 + 2126 & 110 & 86 & 12 & 48 & 26.80 & -5 & 47 & 30.9 & 27.68 & 1076 & -1 & 1688 + 2127 & 121 & 85 & 12 & 48 & 26.16 & -5 & 47 & 36.5 & 28.00 & 1554 & -1 & 1689 + 2128 & 97 & 96 & 12 & 48 & 27.27 & -5 & 47 & 16.3 & 26.80 & 1268 & -1 & 1691 + 2129 & 99 & 82 & 12 & 48 & 27.58 & -5 & 47 & 29.7 & 26.48 & 1286 & -1 & 1687 + 2130 & 100 & 75 & 12 & 48 & 27.73 & -5 & 47 & 36.4 & 27.05 & 1290 & -1 & 1685 + 2131 & 94 & 70 & 12 & 48 & 28.23 & -5 & 47 & 38.2 & 26.63 & 1415 & -1 & 1684 + 2132 & 149 & 142 & 12 & 48 & 22.85 & -5 & 46 & 56.8 & 27.51 & 1301 & -1 & 1718 + 2133 & 152 & 138 & 12 & 48 & 22.75 & -5 & 47 & 1.2 & 27.57 & 1471 & -1 & -1 + 2134 & 194 & 133 & 12 & 48 & 20.33 & -5 & 47 & 24.0 & 27.62 & 1265 & -1 & 1730 + 2135 & 138 & 42 & 12 & 48 & 26.32 & -5 & 48 & 22.3 & 26.31 & 1064 & -1 & 1713 + 2136 & 136 & 33 & 12 & 48 & 26.70 & -5 & 48 & 29.5 & 26.28 & 1175 & -1 & 1711 + 2137 & 133 & 34 & 12 & 48 & 26.86 & -5 & 48 & 27.6 & 26.44 & 1063 & -1 & 1712 + 2138 & 122 & 31 & 12 & 48 & 27.63 & -5 & 48 & 25.5 & 27.08 & 1323 & -1 & 1677 + 2139 & 117 & 59 & 12 & 48 & 27.13 & -5 & 47 & 58.8 & 25.63 & 1095 & -1 & 1683 + 2140 & 109 & 59 & 12 & 48 & 27.60 & -5 & 47 & 55.6 & 26.59 & 1240 & -1 & 1682 + 2141 & 135 & 14 & 12 & 48 & 27.31 & -5 & 48 & 46.8 & 27.33 & 1321 & -1 & -1 + 2142 & 135 & 2 & 12 & 48 & 27.67 & -5 & 48 & 58.0 & 27.00 & 1462 & -1 & 1710 + 2143 & 85 & 61 & 12 & 48 & 28.98 & -5 & 47 & 42.9 & 27.03 & 1410 & -1 & -1 + 2144 & 95 & 53 & 12 & 48 & 28.62 & -5 & 47 & 54.5 & 26.51 & 1087 & -1 & 1681 + 2145 & 99 & 42 & 12 & 48 & 28.67 & -5 & 48 & 6.1 & 26.69 & 1315 & -1 & 1679 + 2146 & 95 & 42 & 12 & 48 & 28.95 & -5 & 48 & 5.0 & 26.71 & 1351 & -1 & 1678 + 2147 & 98 & 50 & 12 & 48 & 28.55 & -5 & 47 & 58.4 & 27.55 & 1287 & -1 & 1680 + 2148 & 87 & 35 & 12 & 48 & 29.64 & -5 & 48 & 7.7 & 27.22 & 1222 & -1 & 1644 + 2149 & 112 & 2 & 12 & 48 & 29.02 & -5 & 48 & 48.2 & 26.08 & 1303 & -1 & 1671 + 2150 & 109 & 8 & 12 & 48 & 29.04 & -5 & 48 & 41.4 & 26.70 & 1180 & -1 & 1673 + 2151 & 104 & 16 & 12 & 48 & 29.15 & -5 & 48 & 31.3 & 26.53 & 1377 & -1 & 1676 + 2152 & 78 & 30 & 12 & 48 & 30.33 & -5 & 48 & 7.8 & 26.61 & 1160 & -1 & 1643 + 2153 & 85 & 27 & 12 & 48 & 29.97 & -5 & 48 & 13.7 & 27.20 & 1170 & -1 & 1642 + 2154 & 86 & 19 & 12 & 48 & 30.11 & -5 & 48 & 21.2 & 27.28 & 1060 & -1 & 1641 + 2155 & 74 & 3 & 12 & 48 & 31.32 & -5 & 48 & 30.8 & 25.87 & 1414 & -1 & 1636 + 2201 & -265 & -109 & 12 & 48 & 55.10 & -5 & 47 & 48.8 & 27.77 & 1107 & 27 & -1 + 2202 & -393 & -128 & 12 & 49 & 3.37 & -5 & 47 & 12.2 & 27.07 & 1297 & -1 & -1 + 2203 & -340 & -157 & 12 & 49 & 0.98 & -5 & 48 & 0.7 & 26.23 & 1257 & -1 & -1 + 2204 & -349 & -51 & 12 & 48 & 58.55 & -5 & 46 & 21.6 & 25.70 & 1289 & -1 & -1 + 2205 & -342 & -46 & 12 & 48 & 57.98 & -5 & 46 & 19.6 & 27.24 & 1349 & -1 & -1 + 2206 & -331 & -57 & 12 & 48 & 57.60 & -5 & 46 & 34.6 & 26.64 & 1244 & -1 & -1 + 2207 & -318 & -54 & 12 & 48 & 56.76 & -5 & 46 & 37.0 & 26.74 & 1173 & -1 & -1 + 2208 & -258 & -38 & 12 & 48 & 52.63 & -5 & 46 & 47.8 & 26.55 & 1220 & 65 & -1 + 2209 & -251 & -53 & 12 & 48 & 52.66 & -5 & 47 & 4.0 & 26.25 & 1153 & 66 & -1 + 2210 & -244 & -52 & 12 & 48 & 52.16 & -5 & 47 & 6.2 & 26.88 & 1189 & 67 & -1 + 2211 & -242 & -47 & 12 & 48 & 51.96 & -5 & 47 & 2.8 & 27.78 & 1094 & 68 & -1 + 2212 & -228 & -26 & 12 & 48 & 50.48 & -5 & 46 & 49.4 & 27.77 & 1314 & -1 & -1 + 2213 & -338 & -16 & 12 & 48 & 56.88 & -5 & 45 & 54.5 & 26.12 & 1261 & -1 & -1 + 2214 & -303 & -28 & 12 & 48 & 55.06 & -5 & 46 & 19.5 & 26.72 & 1310 & 114 & -1 + 2215 & -221 & -113 & 12 & 48 & 52.56 & -5 & 48 & 11.3 & 28.09 & 1245 & -1 & -1 + 2216 & -261 & -131 & 12 & 48 & 55.45 & -5 & 48 & 10.9 & 27.55 & 1167 & 11 & -1 + 2217 & -255 & -167 & 12 & 48 & 56.13 & -5 & 48 & 45.6 & 26.73 & 1085 & 2 & -1 + 2218 & -294 & -179 & 12 & 48 & 58.82 & -5 & 48 & 40.5 & 26.65 & 1363 & 1 & -1 + 2219 & -215 & -157 & 12 & 48 & 53.42 & -5 & 48 & 53.4 & 27.60 & 1289 & 3 & -1 + 2220 & -217 & -33 & 12 & 48 & 50.00 & -5 & 47 & 0.7 & 26.78 & 1189 & 69 & -1 + 2221 & -209 & -34 & 12 & 48 & 49.53 & -5 & 47 & 5.0 & -1 .
& 1224 & 70 & -1 + 2222 & -201 & -13 & 12 & 48 & 48.49 & -5 & 46 & 49.1 & 27.06 & 1356 & 118 & -1 + 2223 & -178 & -30 & 12 & 48 & 47.59 & -5 & 47 & 14.1 & 27.03 & 1388 & 121 & -1 + 2224 & -172 & -18 & 12 & 48 & 46.88 & -5 & 47 & 6.3 & 27.12 & 1297 & 120 & -1 + 2225 & -143 & -9 & 12 & 48 & 44.83 & -5 & 47 & 10.4 & 27.43 & 1051 & 128 & -1 + 2226 & -150 & -38 & 12 & 48 & 46.12 & -5 & 47 & 33.8 & 27.45 & 1124 & 73 & -1 + 2227 & -124 & -11 & 12 & 48 & 43.76 & -5 & 47 & 20.3 & 27.17 & 1245 & 129 & -1 + 2228 & -107 & -14 & 12 & 48 & 42.80 & -5 & 47 & 29.4 & 25.93 & 1495 & 130 & -1 + 2229 & -95 & -29 & 12 & 48 & 42.46 & -5 & 47 & 48.5 & 26.35 & 1409 & 142 & -1 + 2230 & -83 & -23 & 12 & 48 & 41.59 & -5 & 47 & 47.8 & 26.51 & 1074 & 140 & -1 + 2231 & -86 & -23 & 12 & 48 & 41.77 & -5 & 47 & 46.9 & -1 .
& 1190 & 139 & -1 + 2232 & -85 & -25 & 12 & 48 & 41.77 & -5 & 47 & 49.3 & 27.36 & 1486 & 141 & -1 + 2233 & -89 & -11 & 12 & 48 & 41.65 & -5 & 47 & 35.0 & 27.50 & 980 & 137 & -1 + 2234 & -79 & -14 & 12 & 48 & 41.09 & -5 & 47 & 41.8 & 26.73 & 1474 & 138 & -1 + 2235 & -76 & -47 & 12 & 48 & 41.87 & -5 & 48 & 12.8 & 26.56 & 1015 & 82 & -1 + 2236 & -79 & -43 & 12 & 48 & 41.94 & -5 & 48 & 7.7 & 27.31 & 1115 & 81 & -1 + 2237 & -88 & -58 & 12 & 48 & 42.90 & -5 & 48 & 17.3 & 26.89 & 1164 & 83 & -1 + 2238 & -99 & -67 & 12 & 48 & 43.83 & -5 & 48 & 21.1 & 26.54 & 1249 & 84 & -1 + 2239 & -118 & -69 & 12 & 48 & 45.03 & -5 & 48 & 15.4 & 26.83 & 1304 & 78 & -1 + 2240 & -144 & -80 & 12 & 48 & 46.93 & -5 & 48 & 13.8 & 25.82 & 1165 & 36 & -1 + 2241 & -132 & -107 & 12 & 48 & 46.92 & -5 & 48 & 43.6 & 26.21 & 1016 & 41 & -1 + 2242 & -126 & -106 & 12 & 48 & 46.59 & -5 & 48 & 45.5 & 26.96 & 1333 & 42 & -1 + 2243 & -110 & -113 & 12 & 48 & 45.77 & -5 & 48 & 57.9 & 26.50 & 1250 & 15 & -1 + 2244 & -137 & -87 & 12 & 48 & 46.71 & -5 & 48 & 23.3 & 27.48 & 1170 & 37 & -1 + 2245 & -113 & -85 & 12 & 48 & 45.19 & -5 & 48 & 32.1 & 27.26 & 1019 & 40 & -1 + 2246 & -120 & -82 & 12 & 48 & 45.49 & -5 & 48 & 26.4 & 27.63 & 1068 & 39 & -1 + 2247 & -139 & -71 & 12 & 48 & 46.37 & -5 & 48 & 8.4 & 27.98 & 1150 & -1 & -1 + 2248 & -155 & -95 & 12 & 48 & 47.99 & -5 & 48 & 23.1 & 26.99 & 1246 & 35 & -1 + 2249 & -157 & -89 & 12 & 48 & 47.96 & -5 & 48 & 17.0 & 27.51 & 1283 & 34 & -1 + 2250 & -166 & -106 & 12 & 48 & 49.02 & -5 & 48 & 28.6 & 26.51 & 1159 & 31 & -1 + 2251 & -169 & -109 & 12 & 48 & 49.25 & -5 & 48 & 29.5 & 27.52 & 1076 & 30 & -1 + 2252 & -162 & -108 & 12 & 48 & 48.80 & -5 & 48 & 32.2 & 27.52 & 1307 & 32 & -1 + 2253 & -172 & -133 & 12 & 48 & 50.11 & -5 & 48 & 50.2 & 27.06 & 1343 & 12 & -1 + 2254 & -133 & -123 & 12 & 48 & 47.48 & -5 & 48 & 57.3 & -1 .
& 1241 & 13 & -1 + 2255 & -198 & -81 & 12 & 48 & 50.21 & -5 & 47 & 51.8 & 26.60 & 1351 & 28 & -1 + 2256 & -175 & -101 & 12 & 48 & 49.38 & -5 & 48 & 20.0 & 27.40 & 1214 & 29 & -1 + 2257 & -179 & -64 & 12 & 48 & 48.58 & -5 & 47 & 44.9 & 27.98 & 1139 & 71 & -1 + 2258 & -111 & -43 & 12 & 48 & 43.89 & -5 & 47 & 54.8 & 27.47 & 1176 & 76 & -1 + 2259 & -103 & -38 & 12 & 48 & 43.24 & -5 & 47 & 53.0 & 27.34 & 1096 & 79 & -1 + 2260 & -66 & -6 & 12 & 48 & 40.07 & -5 & 47 & 39.9 & -1 .
& 1140 & 136 & 1024 + 2261 & -59 & -15 & 12 & 48 & 39.92 & -5 & 47 & 50.5 & 26.61 & 1207 & 162 & 1043 + 2262 & -61 & -14 & 12 & 48 & 40.00 & -5 & 47 & 49.3 & -1 .
& 1123 & 161 & 1042 + 2301 & -88 & 71 & 12 & 48 & 39.25 & -5 & 46 & 20.6 & 26.61 & 1319 & 300 & -1 + 2302 & -89 & 50 & 12 & 48 & 39.92 & -5 & 46 & 39.0 & 26.73 & 1218 & 301 & -1 + 2303 & -92 & 45 & 12 & 48 & 40.22 & -5 & 46 & 43.3 & 27.45 & 1148 & 220 & -1 + 2304 & -105 & 31 & 12 & 48 & 41.39 & -5 & 46 & 49.7 & 26.78 & 1402 & 219 & -1 + 2305 & -86 & 29 & 12 & 48 & 40.34 & -5 & 47 & 0.1 & 26.95 & 1332 & 222 & -1 + 2306 & -94 & 20 & 12 & 48 & 41.09 & -5 & 47 & 4.7 & 27.05 & 1341 & 221 & -1 + 2307 & -74 & 24 & 12 & 48 & 39.72 & -5 & 47 & 9.5 & 26.55 & 973 & 226 & -1 + 2308 & -76 & 25 & 12 & 48 & 39.79 & -5 & 47 & 7.6 & 27.24 & 1220 & 225 & -1 + 2309 & -70 & 32 & 12 & 48 & 39.22 & -5 & 47 & 4.0 & 27.15 & 1254 & 228 & -1 + 2310 & -63 & 34 & 12 & 48 & 38.78 & -5 & 47 & 5.1 & 26.55 & 1069 & 229 & 1050 + 2311 & -117 & 129 & 12 & 48 & 39.34 & -5 & 45 & 16.5 & 26.68 & 1157 & 362 & -1 + 2312 & -129 & 127 & 12 & 48 & 40.17 & -5 & 45 & 13.1 & 26.27 & 1324 & 343 & -1 + 2313 & -137 & 119 & 12 & 48 & 40.85 & -5 & 45 & 16.3 & 27.30 & 1389 & 342 & -1 + 2314 & -113 & 96 & 12 & 48 & 40.07 & -5 & 45 & 47.4 & 27.15 & 1011 & 346 & -1 + 2315 & -112 & 64 & 12 & 48 & 40.87 & -5 & 46 & 16.8 & 26.78 & 1124 & 298 & -1 + 2316 & -125 & 28 & 12 & 48 & 42.74 & -5 & 46 & 44.5 & 26.38 & 1161 & 217 & -1 + 2317 & -130 & 26 & 12 & 48 & 43.04 & -5 & 46 & 44.3 & -1 .
& 1320 & 216 & -1 + 2318 & -124 & 6 & 12 & 48 & 43.26 & -5 & 47 & 4.3 & 26.88 & 1035 & 126 & -1 + 2319 & -156 & 19 & 12 & 48 & 44.83 & -5 & 46 & 39.8 & 27.38 & 1375 & -1 & -1 + 2320 & -172 & 7 & 12 & 48 & 46.13 & -5 & 46 & 43.4 & 26.03 & 1313 & 119 & -1 + 2321 & -168 & 63 & 12 & 48 & 44.32 & -5 & 45 & 54.5 & 26.54 & 1464 & 295 & -1 + 2322 & -149 & 80 & 12 & 48 & 42.68 & -5 & 45 & 46.5 & 27.47 & 1211 & 296 & -1 + 2323 & -146 & 102 & 12 & 48 & 41.90 & -5 & 45 & 28.3 & 28.22 & 1336 & 340 & -1 + 2324 & -156 & 99 & 12 & 48 & 42.55 & -5 & 45 & 26.4 & 27.44 & 1279 & 339 & -1 + 2325 & -151 & 93 & 12 & 48 & 42.41 & -5 & 45 & 34.0 & 27.87 & 1327 & 341 & -1 + 2326 & -166 & 125 & 12 & 48 & 42.47 & -5 & 44 & 58.7 & 27.79 & 1349 & 337 & -1 + 2327 & -181 & 19 & 12 & 48 & 46.37 & -5 & 46 & 28.7 & 26.68 & 1030 & -1 & -1 + 2328 & -183 & 13 & 12 & 48 & 46.65 & -5 & 46 & 33.3 & 27.71 & 1193 & -1 & -1 + 2329 & -188 & 27 & 12 & 48 & 46.55 & -5 & 46 & 18.9 & 26.42 & 1331 & -1 & -1 + 2330 & -188 & 26 & 12 & 48 & 46.57 & -5 & 46 & 19.5 & 26.30 & 1223 & -1 & -1 + 2331 & -189 & 31 & 12 & 48 & 46.53 & -5 & 46 & 14.7 & 27.69 & 1251 & -1 & -1 + 2332 & -201 & 17 & 12 & 48 & 47.64 & -5 & 46 & 21.6 & 26.90 & 1037 & 214 & -1 + 2333 & -206 & 19 & 12 & 48 & 47.86 & -5 & 46 & 18.4 & 26.85 & 1275 & 213 & -1 + 2334 & -216 & 12 & 12 & 48 & 48.70 & -5 & 46 & 20.2 & 26.84 & 1019 & 212 & -1 + 2335 & -223 & 6 & 12 & 48 & 49.27 & -5 & 46 & 22.5 & 26.64 & 1316 & 116 & -1 + 2336 & -204 & 77 & 12 & 48 & 46.11 & -5 & 45 & 25.9 & 27.66 & 1148 & 293 & -1 + 2337 & -252 & 16 & 12 & 48 & 50.74 & -5 & 46 & 1.8 & 25.87 & 1225 & 211 & -1 + 2338 & -247 & 5 & 12 & 48 & 50.78 & -5 & 46 & 12.9 & 27.39 & 1273 & 115 & -1 + 2339 & -232 & 20 & 12 & 48 & 49.44 & -5 & 46 & 6.1 & 27.08 & 1285 & -1 & -1 + 2340 & -272 & 161 & 12 & 48 & 47.88 & -5 & 43 & 41.5 & 27.30 & 1239 & 357 & -1 + 2341 & -272 & 39 & 12 & 48 & 51.31 & -5 & 45 & 32.3 & 27.70 & 1267 & -1 & -1 + 2342 & -277 & 35 & 12 & 48 & 51.73 & -5 & 45 & 33.5 & 27.74 & 1158 & -1 & -1 + 2343 & -283 & 46 & 12 & 48 & 51.76 & -5 & 45 & 21.5 & -1 .
& 1267 & 210 & -1 + 2344 & -290 & 111 & 12 & 48 & 50.37 & -5 & 44 & 19.5 & 26.97 & 1110 & 336 & -1 + 2345 & -294 & 107 & 12 & 48 & 50.75 & -5 & 44 & 21.5 & 26.55 & 1101 & 335 & -1 + 2346 & -338 & 143 & 12 & 48 & 52.37 & -5 & 43 & 29.6 & 27.21 & 1120 & -1 & -1 + 2347 & -327 & 19 & 12 & 48 & 55.21 & -5 & 45 & 26.9 & 26.05 & 1425 & -1 & -1 + 2348 & -361 & 81 & 12 & 48 & 55.54 & -5 & 44 & 16.6 & 26.62 & 1360 & -1 & -1 + 2349 & -373 & 96 & 12 & 48 & 55.84 & -5 & 43 & 57.7 & 26.93 & 1238 & -1 & -1 + 2350 & -403 & 101 & 12 & 48 & 57.51 & -5 & 43 & 40.8 & 26.49 & 1105 & -1 & -1 + 2351 & -409 & 95 & 12 & 48 & 58.08 & -5 & 43 & 43.5 & 26.77 & 1229 & -1 & -1 + 2352 & -83 & 17 & 12 & 48 & 40.48 & -5 & 47 & 11.6 & 27.39 & 1286 & 224 & -1 + 2353 & -72 & 19 & 12 & 48 & 39.74 & -5 & 47 & 14.3 & 27.61 & 1165 & 227 & -1 +
|
we describe the implementation of slitless radial velocity measurements of extragalactic planetary nebulae ( pns ) with the 8.2 m subaru telescope and its cassegrain imaging spectrograph , focas . as a first application
, we have extended a previous search for pns in ngc 4697 to larger angular distances from its center .
a total of 218 pns were detected , and their radial velocities were measured .
we have added 56 new pn detections to the existing sample of 535 , observed previously with the eso vlt + fors imaging spectrograph ; 36 of these new 56 pns are located at angular distances larger than 230 arcsec from the center of ngc 4697 .
we compare the new focas velocities with the earlier fors velocities , for 158 of the 162 reobserved sources , finding good agreement .
we now have kinematic information extending out to 5 effective radii from the center .
the outer line - of - sight velocity dispersion is a bit lower than estimated earlier .
this result is compatible with the existence of a dark matter halo plus some degree of radial anisotropy , but the dark matter halo is rather inconspicuous , and it is still unclear how massive it can be . a more detailed global dynamical study of the whole set of pn velocities will be required to decide if they permit to narrow down the range of possible dark matter distributions in ngc 4697 .
the new radial velocities reveal no evidence of rotation at 5 effective radii .
|
many data analysis tasks can be facilitated by high - dimensional data visualization .
for example , it is often helpful to know what sets of dimensions in the data are correlated or have a significant impact in processes such as clustering , sampling , or labeling .
another important task is the discovery of hidden relationships between labels and numeric values in the analysis of labeled high - dimensional datasets .
several machine learning techniques such as deep neural networks or association rule mining are useful for this purpose but may require significant computational time when discovering relationships between labels and large arbitrary combinations of numeric dimensions
. moreover , in many application domains , high - dimensional data analysis requires interactive analysis and decision making by a domain expert to specify the rules between the labels and the numeric values .
such circumstances arise in important application domains such as biomedicine , finance , and social media analysis , where the a priori characterization and detection of interesting patterns is difficult and limited due to a lack of domain - specific knowledge as well as the complexity and dynamic nature of the data .
the application of a data visualization framework is often helpful for these types of problems .
high - dimensional data visualization continues to be an important and active research field with several survey papers dedicated to this area @xcite @xcite .
one of these surveys @xcite divided the available multi - dimensional data visualization techniques into three categories : animations , two - variate displays , and multivariate displays .
animation techniques facilitate the dynamic display of multiple configurations of the high - dimensional data and are commonly applied to both two - variate and multivariate techniques .
two - variate techniques only visualize the relationships between two variables ; an example of such a technique is the well - known scatterplot ( sp ) .
multivariate visualization techniques attempt to represent the distribution of all the dimensions in a given dataset on a single display space .
the increasing dimensionality of modern datasets is spurring the use of multivariate visualization techniques , such as sp matrices and parallel coordinate plots ( pcps ) @xcite .
a sp matrix consists of multiple adjacent sps in a grid - like arrangement , with each sp being identified by its row and column index .
the greatest advantage of sp matrices is the high level of familiarity that non - expert users have with it .
however , this visualization technique suffers from few major drawbacks .
firstly , individual sps within the display space may be very small if the number of dimensions of the dataset is very large .
in addition , it can be difficult for humans to visually compare arbitrary pairs of sps that are distantly placed in the display space .
several studies aimed at selecting meaningful sets of sps and effectively arrange them onto the display spaces @xcite @xcite @xcite @xcite @xcite ; however , these studies often had difficulty to visually compare large number of sps .
pcps are an alternative multivariate visualization technique that display high - dimensional datasets as a set of polylines that intersect with parallel axes ; this visualization better enables the observation of correlations between pairs of dimensions .
specifically , the existence of parallel polylines between two axes indicates positive correlation .
conversely , polylines crossing between axes are indicative of negative correlation . while pcps have been well studied for high - dimensional data visualization and are a valuable tool for the analysis of multidimensional data , they have some shortcomings in practice . pcps may require a large horizontal display space even for a modest number of dimensions . within this horizontal display space
the ordering of the dimensions is relevant for the detection of pairwise correlations .
moreover , it is difficult to visually represent the correlation of a particular dimension with three or more different dimensions .
an example of this situation can be illustrated given a dataset which has four dimensions @xmath0 , @xmath1 , @xmath2 , and @xmath3 , which are visualized by a single pcp .
the correlations between @xmath0 and @xmath1 , or @xmath1 and @xmath2 , can be easily observed if the dimensions are arranged in the order @xmath0 , @xmath1 , @xmath2 , and @xmath3 .
however , the correlation between @xmath1 and @xmath3 can not be easily observed within this visualization .
several techniques @xcite @xcite have attempted to address this last issue by only displaying pcps for subsets of the dimensions that are highly correlated .
given the example dataset and situation specified above , these techniques would construct two pcps : one displaying @xmath0 , @xmath1 , and @xmath2 , and the other displaying @xmath1 and @xmath3 .
this enables users to view correlations between @xmath1 and all of the other dimensions .
however , adjusting the optimal visualization parameters ( subset of dimensions to display ) is a challenging issue that often requires interactive input by an expert user .
we hypothesize that new high - dimensional data visualization capabilities can be enabled by combining multivariate visualization techniques ( e.g. pcps ) with an innovative coupling to interactive dimension selection techniques
@xcite @xcite @xcite @xcite .
this paper presents a new technique for high - dimensional data visualization in which a set of low - dimensional pcps are interactively constructed .
figure [ fig : snapshot0 ] shows a snapshot of our proposed technique , which features two visualization components : _ low - dimensional pcps _ and a _ dimension graph_. the low - dimensional pcps display the values of the selected dimensions .
we provide two approaches to interactive dimension selection , derived from the correlation between numeric dimensions or from data mined association rules between numeric dimensions and categorical labels .
the dimension graph displays the relationships between the numeric dimensions and allows interactive dimension selection .
this representation is visually similar to correlation map proposed by zhang et al .
@xcite ; however , our technique differs in the application of the dimension graph where we enabled simultaneous extraction of a set of low - dimensional subspaces with a simple threshold adjustment operation , thus offering an interactive mechanism that is more intuitive , when compared with existing techniques .
we allow duplication in the selection of dimensions and the consequent display across multiple low - dimensional pcps .
this enables visualization of relationships between a particular dimension and three or more other dimensions , which can be often difficult to understand when all dimensions are shown in a single pcp . at the same time
, our method flexibly saves display space because many unnecessary dimensions can be removed from the visualization results .
the rest of this paper is organized as follows .
we introduce related work in section 2 . in section 3
, we present the framework that describes our new high - dimensional visualization technique .
we present an experimental validation of our technique in section 4 and draw conclusions from the outcomes in section 5 .
in this section , we survey pcps and interactive dimension selection techniques . our visualization technique builds upon a combination of these approaches .
as mentioned previously , pcps @xcite display high - dimensional datasets as polylines intersecting with parallel axes .
the improvement of pcps is a very active research topic and one of the well - known challenges in this domain is that of polyline cluttering , i.e. , a reduction of line crossings and overlaps for visual comprehensibility .
several techniques have attempted to improve the comprehensibility of the results obtained by pcps by applying clustering or sampling of the polylines @xcite
@xcite @xcite @xcite .
in addition , the effectiveness of pcps are highly dependent on the order of the dimensions and various dimension ordering techniques have recently been proposed to address this issue @xcite @xcite @xcite @xcite .
the last major challenge is the difficulty in representing all correlations in one display space , especially when a particular dimension is strongly correlated with many other dimensions . in these circumstances , pcps can represent only a subset of all possible relationships between the dimensions .
when a multi - dimensional dataset contains a very large number of dimensions , existing visualization techniques ( e.g. , pcps or sps ) may need very large display spaces to represent them completely .
this problem can be solved by dividing the high - dimensional data space into smaller subsets .
ten caat et al . applied multiple pcps to represent time - varying multidimensional data @xcite .
suematsu et al .
@xcite also converted high - dimensional datasets into low - dimensional subsets and visualized these subsets using multiple pcps arranged on display spaces based upon their similarity and correlation . using similar ideas ,
zheng et al .
@xcite selected sps based upon the meaningfulness of the dimensions being displayed and adjusted their layout based upon their similarity .
claessen et al .
@xcite presented a technique to visualize high - dimensional datasets by selecting sets of low - dimensional subspaces and representing them as a combination of pcps and sps .
these techniques provide static results by the decomposition of high - dimensional spaces into multiple pre - selected low - dimensional spaces .
however , the lack of any interactive mechanisms to select the sets of dimensions denies domain experts the ability to use their prior experience and knowledge to specify rules about the data .
several studies have demonstrated that subsets or subspaces of high - dimensional spaces can be effectively visualized by a user s interactive selection of the subspace ( subset of dimensions ) .
elmqvist et al .
@xcite presented an interactive mechanism to select pairs of dimensions and to smoothly switch between the sps , which was effective in preserving users mental maps of the high - dimensional spaces .
lee et al .
@xcite and liu et al .
@xcite applied dimension reduction schemes to interactively select subsets of the high - dimensional data .
nohno et al .
@xcite presented a technique to interactively contract highly - correlated dimensions to adjust the number of axes displayed in pcps .
several recent studies have applied sps for the representation of _ dimension spaces _ in which each dot in the sp represents a single dimension in the space .
turkay et al .
@xcite @xcite presented a dual sp model to visualize both the items and dimensions spaces .
similarly , yuan et al .
@xcite presented a interactive mechanism to select low - dimensional subspaces on the sp display in which each dot corresponds to a different dimension .
we also represent the relationships among the dimensions in a 2d space ; however , our technique applies a graph rather than a sp .
the technique recently proposed by zhang et al .
@xcite uses a similar representation to that applied by our technique .
they construct a `` correlation map '' in which the dataset dimensions are represented by dots where the connection between the dots is derived from pairwise correlations .
our technique includes two characteristics that differ fundamentally from the method of zhang et al .
firstly , the dimension graph in our technique is used as an interactive mechanism to simultaneously control the dimensionality of the set of pcps , thereby allowing the pcps to act as a visual representation of a set of low - dimensional subspaces ; in contrast , users need to find interesting dimensions and select them individually while using the method of zhang et al .
in addition , our technique uses association rule mining to extract low - dimensional subspaces in contrast with the correlation - based technique used by zhang et al .
our technique can therefore extract complex multi - variate relationships in comparison to pairwise correlations .
this section describes our new high - dimensional data visualization technique .
figure [ fig : overview1 ] illustrates the processing flow of our technique .
our technique treats the values of numeric dimensions as vectors , and calculates the distances between arbitrary pairs of dimensions .
it selects the sets of dimensions with an interactive threshold setting .
our current implementation provides two types of dimension selection schemes .
the first approach is based on distances between arbitrary pairs of dimensions , as described in section [ sec : dimselect1 ] .
this approach forms a graph by connecting the pairs of dimensions if their distances are smaller than the threshold , and finds cliques of the graph as sets of dimensions .
the second approach is based on concentration or separateness of the labels , as described in section [ sec : dimselect2 ] .
the approach treats the value of a user - specified categorical dimension as _ labels _ , and extracts sets of dimensions if particular labels purely concentrate on the particular portions of the axes of pcps .
the technique allows the user to modify the threshold via a slider widget on the window system , and selects the sets of dimensions and updates the sets of pcps accordingly .
we formalize the high - dimensional datasets visualized by our technique as follows .
the dataset has @xmath4 items , which contain @xmath5-dimensional values , including @xmath6 numeric dimensions and @xmath7 categorical dimensions .
the dataset @xmath8 and the @xmath9-th item @xmath10 are described as : @xmath11 where @xmath12 denotes the @xmath13-th numeric dimension value of the @xmath9-th item , and @xmath14 denotes the @xmath15-th categorical dimension value of the @xmath9-th item .
the value of @xmath14 should be one of the categorical values @xmath16 , where @xmath17 denotes the number of possible categorical values for the @xmath15-th categorical dimension .
dimension set selection is a key component for our visualization technique .
the requirements for the dimension set selection are strongly motivated by what individual users want to see and how they wish to explore the data . in this paper , we describe two approaches for interactive dimension set selection .
it is important to understand correlation or relationships among variables in many high - dimensional data analysis applications .
thus we provide a scheme to select sets of dimensions in which similar or highly correlated dimensions are grouped into the same set .
this process generates a dimension graph by connecting highly correlated dimensions , and extracts cliques of the dimension graph as sets of highly correlated dimensions .
the dimension graph is visualized by applying a dimensionality reduction scheme .
we treat each numeric dimension as a vector , described as @xmath18 for the @xmath13-th dimension , where @xmath5 is the number of items .
we first calculate the distances between all possible pairs of the numeric dimensions .
the distance between the @xmath13-th and @xmath15-th numeric dimensions is defined as : @xmath19 where @xmath20 denotes spearman s rank correlation coefficients .
this definition means that positively or negatively correlated numeric dimensions have similar distances and are thus close .
we then form a graph by connecting pairs of numeric dimensions if their distances @xmath21 are smaller than user - defined threshold @xmath22 . consequently
, more connections are generated if the user set larger thresholds .
next , we extract the set of _ cliques _ , or complete subgraphs , by performing a maximum clique detection .
our implementation applies the bron - kerbosch algorithm @xcite .
the sets of dimensions corresponding to the cliques are displayed as pcps .
figure [ fig : distanceclique ] shows a basic example of dimension set selection result . the right side of the window displays the dots corresponding to the numeric dimensions and their connections .
we calculate the positions of the dots based on the distances @xmath21 by applying mds ( multi - dimensional scaling ) .
the left side of the window displays the pcps generated from the dimensions contained in the cliques of the graph .
our implementation allows two pcps to be merged if their corresponding cliques share only a single dimension , e.g. , pcps ( b) and ( c) in this figure .
we reorder the numeric dimensions so that well correlated pairs of dimensions are adjacently placed in the pcps .
we attempt to minimize the sum of @xmath21 between dimensions adjacently drawn in the pcps , applying an approximate solution of the traveling salesman problem @xcite .
categorical dimensions in datasets can be used as labels that indicate particular characteristics of data elements .
the correlations or relationships between these labels and the numeric values in other dimensions quantify important information in many high - dimensional data analysis and exploration applications .
we therefore also also provide a scheme to select sets of dimensions where particular labels are either concentrated or distinctly separated .
this scheme firstly applies an association rule mining method to extract well related combinations of labels and ranges of numeric dimensions , and visualizes them as a set of pcps .
note that the distance - based dimension selection ( section [ sec : dimselect1 ] ) was a general approach to select dimensions and therefore required the calculation of the correlation between all pairs of features .
in contrast , the presence of labels through categorical dimensions allows the opportunity to reduce the dimensional space to only those dimensions that have strong relationships with the labels . to extract association rules between labels and ranges of numeric dimensions , we first divide each range of numeric dimensions into a same number of subranges .
let the range of one subrange of the @xmath13-th numeric dimension @xmath23 , and the minimum and maximum values be given by @xmath24 and @xmath25 . here
, the @xmath15-th subrange of the @xmath13-th numeric dimension is described as follows : @xmath26\ ] ] we count the number of items where the value of the @xmath13-th numeric dimension is in the @xmath15-th subrange .
we then apply a quantitative association rule mining algorithm @xcite @xcite to discover the rules given by : @xmath27 or @xmath28 where @xmath29 is a particular label corresponding to a particular value of a categorical dimension of the input dataset .
the first rule ( equation [ eq : rule1 ] ) denotes the case where the ranges of particular numeric dimensions are predicted from labels of items .
in other words , this rule describes the subspaces where specific labels are concentrated .
the second rule ( equation [ eq : rule2 ] ) denotes the case where labels of items are predicted based upon values of particular numeric dimensions .
in other words , this rule describes the subspaces that separate labels .
we extract the sets of dimensions that correspond to these discovered rules and visualize them as sets of low - dimensional pcps .
users can interactively adjust the thresholds of support ( @xmath30 ) and confidence ( @xmath31 ) , which are commonly used as criteria of association rule mining .
figure [ fig : labelseparateness ] illustrates this process applying equation [ eq : rule2 ] .
it shows an example of pcps of numeric dimensions where different poly lines ( indicated by color ) are well - separated .
we define the order of dimensions by using an approximate solution for the traveling salesman as described in section [ sec : dimselect1 ] .
it is often important to render appropriate number of nodes in a limited display space to comprehensively represent graphs .
meanwhile , many multidimensional datasets consist of groups of very similar dimensions where it is not necessary to observe every dimension .
our technique supports a dimension sampling scheme to render dimensions invisible when they are sufficiently close to one or more visible dimensions . for this process
, we again use equation [ eq : dist ] to calculate the distance between dimensions .
we extract pairs of dimensions where these distances are smaller than a user - defined threshold @xmath32 , where @xmath33 , and randomly select one of these dimensions to make it invisible .
the visible dimensions are then visualized by pcps .
figure [ fig : sampling ] illustrates how dimensions are made invisible .
figure [ fig : snapshot1 ] shows a snapshot of our visualization technique .
the upper - left side of the window features the radio buttons to select one of the categorical dimensions .
after selection , the lower - left side of the window displays a list of categorical values and their associated colors , to assist users to recognize the numerical distributions with the categorical values . in this snapshot ,
the selected categorical dimension contains two values , `` false '' associated to red , and `` true '' associated to cyan .
the right side of the drawing area displays the dimension graph .
each vertex ( dot ) corresponds to a numeric dimension .
edges ( connections ) between pairs of numeric dimensions are formed if their distances are smaller than the user - specified threshold @xmath22 .
the threshold can be smoothly controlled by users through slider widgets featured at the side of the window .
the slider widgets are also used to control the support and confidence thresholds while applying the association rule mining .
users may also draw rectangles ( using mice or other pointing devices ) to specify the dimensions that should be forcibly included or excluded from the selected sets of dimensions .
the left side of the drawing area displays the sets of pcps .
the pcps are updated when users select different categorical values using the radio buttons , adjust the thresholds @xmath22 or @xmath32 with the slider widgets , or draw rectangles within the drawing area .
the visualization of the low - dimensional subspace occurs after the interactive selection of numeric dimensions described in section [ sec : interact ] . while we use pcps for our visualization , other techniques can also be applied , e.g. , sps can be used instead of pcps by limiting the number of dimensions in each of the selected subsets . in our visualization , the axes of the pcps are ordered as described in section [ sec : dimselect1 ] .
polylines are then drawn ; the transparency of the polylines in our current implementation can be controlled with a slider widget .
polyline coloring is powerful visual tool for assisting users to distinguish polylines in pcps . in our implementation
, the colors are defined according to the values of the user - selected categorical dimension . however , there are many datasets without any categorical dimensions thus precluding color assignment based on categories . in these cases ,
we therefore divide the items into several groups using @xmath15-means clustering @xcite and assign the colors accordingly ; the number of clusters can be specified by the user .
we performed two case studies with our visualization technique : optimization of airplane wing shape design , and knowledge mining between features and annotations of medical images .
each case study used a different form of dimension set selection : distances among dimensions for the airplane wing shape case study and the concentration and separateness of labels ( association rules ) for the medical imaging case study . both of these problem domains have high - dimensional attributes and
hence it is difficult to apply existing low - dimensional approaches such as those mentioned in section 2 .
we implemented our technique using java development kit ( jdk ) 1.7.0 , and executed it on a lenovo thinkpad t450 ( 2.60mhz dual core , ram 8 gb ) with windows 7 ( 64 bit ) .
our first case study examined our visualization technique in its application to analyzing the variables used for the optimization of airplane wing shape design @xcite @xcite .
the dataset consists of 72 design variables of whole wing shapes and 4 objective functions obtained from fluid dynamics simulations .
the dataset contains 776 pareto optimal solutions obtained by a multi - objective genetic algorithm .
these solutions satisfy pareto efficiency , a state of allocation of resources in which it is impossible to make any one item better off without making at least one item worse off .
we visualized this dataset as 776 items of a 76 dimensional dataset .
note that this dataset does not contain any categorical dimensions ; we used the clustering technique described in section [ sec : pcprender ] to assign four colors to the polylines of the pcps .
one of the motivations for visualizing this dataset is to observe and understand the trade - offs among the objective functions ; such trade - offs can be found in many multi - objective optimization problems .
we believe that visualization can contribute to the careful analysis of the distribution of the objective functions thereby enabling a better subjective selection of pareto solutions and the design of better optimization processes . another motivation is to discover unknown relationships among all the variables , not limited to a subset ( @xmath34 to @xmath35 as described below ) .
such discoveries will be useful in narrowing down the variables to be optimized during the design of better wing shapes .
the design variables in the dataset are given by the range from @xmath34 to @xmath36 .
it is well - known in aerospace design community that the following six variables were most important for the optimum solution discovery : * @xmath34 , @xmath37 : span lengths of the inboard / outboard wing panels .
* @xmath38 , @xmath39 : leading - edge sweep angles . * @xmath40 , @xmath35 : root - side chord lengths .
other design variables included the following : * @xmath41 to @xmath42 : variables to define the inner surface connecting corresponding points on upper and lower surfaces of the wing , to design the warping of the wing .
* @xmath43 to @xmath44 : variables to design the twist of the wing .
* @xmath45 to @xmath36 : variables to design the thickness of the wing .
the objective functions are as follows : * @xmath46 : drag coefficient during transonic cruise .
* @xmath47 : drag coefficient during supersonic cruise .
* @xmath48 : bending moment at the wing root during supersonic cruise . *
@xmath49 : pitching moment during supersonic cruise .
we first observed the positions of dots in the right side of the window as shown in figure [ fig : multiopt1 ] .
we found that six dots , corresponding to the four design variables @xmath34 , @xmath37 , @xmath40 , and @xmath35 , and the objective functions @xmath46 and @xmath48 , were closely placed in the right end as shown in figure [ fig : multiopt1](a ) .
we also found that four dots , corresponding to the design variables @xmath38 and @xmath39 , and the objective functions @xmath47 and @xmath49 , were similarly concentrated as shown in figure [ fig : multiopt1](b ) .
we then interactively adjusted the threshold @xmath22 and observed the pcps to determine the correlations among the dimensions . as a result
, we discovered correlations between pairs of the variables , including negative correlation between @xmath46 and @xmath48 as in figure [ fig : multiopt1](a). this negative correlation denotes a typical trade - off between the two objective functions .
meanwhile , @xmath34 and @xmath37 had negative correlations with @xmath40 , while @xmath40 had a positive correlation with @xmath35 .
figure [ fig : multiopt1](b) shows that another pair of objective functions @xmath50 and @xmath49 has a trade - off .
it also shows that the positive correlation between @xmath38 and @xmath39 brings pareto solutions .
to @xmath35 were known to be dominant for the design optimization , and these variables are strongly correlated with the four objective functions .
a clique ( a ) is represented by pcps ( a) , while another clique ( b ) is represented as another pcp ( b). ] figure [ fig : multiopt1 ] also demonstrates that our technique effectively visualizes the subset of dimensions that are strongly correlated .
however , the correlation visualized in this figure was already a well - known property of airplane wing design .
it is perhaps a more interesting and important problem to use our visualization technique to uncover unknown relationships among the variables .
we therefore visualized more relationships by interactively adjusting the threshold parameter @xmath22 ; the resulting visualization is shown in figure [ fig : multiopt2 ] .
we found negative correlations between @xmath48 and @xmath51 , @xmath52 , and @xmath53 , @xmath40 and @xmath54 , and @xmath54 and @xmath55 . here
, @xmath54 is one of the design variables to define the shape of curved surface connecting corresponding positions on upper and lower surfaces of the wing , which controls the crook of the wing ; @xmath51 is one of the variables to design the twist of the wing ; and , @xmath52 and @xmath53 control the thickness of the wing .
the dataset owner , an expert in aerospace fluid simulations , communicated that he did not know this result and it might bring new knowledge in the field of airplane design . several related visualization techniques @xcite @xcite apply dimension - reduced scatterplots for representation of low - dimensional subspaces .
this representation is suitable to find similar simulation results ; on the other hand , our technique is more convenient for carefully observing how dimensions correlate to each other .
the correlation map @xcite is also useful for observing the relationships among the dimension ; however , users need to find interesting dimensions and select them individually on the correlation map .
our technique demonstrated its ability to simultaneously extract all sets of well - correlated low - dimensional subspaces with a simple threshold adjustment operation . .
] figure [ fig : multioptonepcp ] shows an example of visualization of pareto solutions using common pcp .
it is hard to recognize relationships between adjacent dimensions unless we have very horizontally wide display spaces .
dimension selection is useful to compactly visualize sets of well - correlated dimensions . also , it is hard to analyze relationship between a particular dimension and each of more than two other dimensions using regular pcps , because they are useful mainly while observing relations between adjacent dimensions .
it is more useful to observe complex relationships between such dimensions using our technique , which allows to duplicate a single dimension to appear in multiple pcps .
medical data informatics is an increasingly important research area in healthcare @xcite .
a key challenge in this area is the analysis and interpretation of multidimensional data for retrieval and classification applications .
most existing approaches rely on automatic `` black - box '' approaches that use machine - learning to select the most discriminative features , with limited opportunity for input or verification by domain experts , e.g. , artificial neural networks to predict the characteristics of lung nodules @xcite , decision tree committees for medical case retrieval @xcite , etc .
however , it has been shown that black - box learning can train models that do not have meaningful reasons for their predictions .
for example , computer - selected chemicals did not perform as predicted in real - world physical experiments @xcite .
transparency is therefore an important aspect , especially for medical informatics , as it engenders trust in the system by allowing human experts the opportunity to verify and validate the choices that have been automatically made @xcite .
visualization introduces an element of transparency to the feature selection process and provides deeper insights into the correlations and associations between different dimensions .
visualizations can improve verifiability by showing that the combinations of inter - related dimensions correspond to a particular clinical outcome .
furthermore , visualizations also offer the opportunity for an exploration of the dataset that be used to identify outliers and new unknown patterns @xcite .
clinical experts usually interpret volumetric ( 3d ) medical images through multi - planar views that show the images from the three standard orientations or planes ; each plane shows similar or related information from different points of view @xcite . in the literature , many medical image informatics systems use features extracted from the axial ( _ x_-y ) viewing plane only @xcite . in this case study , we will visualize the image features extracted from a large dataset of volumetric medical images to validate whether axial features are sufficient to predict diagnostic outcomes .
we obtained 933 computed tomography ( ct ) images of lung cancer patients from the lung image database consortium and image database resource initiative ( lidc / idri ) @xcite .
the dataset also included four annotations assigned by clinical experts ; two annotations described the diagnostic outcomes of the images and the other two indicated the method used to determine the diagnosis .
our visualizations plotted the image features ( dimensions ) that were related to the following annotations : * _ nodule - level diagnosis _ : diagnosis assigned to the first lung growth or lesion .
( 0 = unknown , 1 = benign or non - malignant disease , 2 = malignant primary lung cancer , 3 = malignant metastatic diseases . ) * _ method for nodule - level diagnosis _ : the clinical method used to determine the diagnosis for the first / primary nodule . ( 0 = unknown , 1 = review of radiological images to show 2 years of stable nodule 2 = biopsy , 3 = surgical resection , 4 = progression or response . ) we used automatic image processing techniques to calculate well - established image features from the nodules in each image @xcite . for each image
, we extracted the same set of 316 dimensional features from the three standard viewing planes ( axial : _
x_-y plane , coronal : _
x_-z plane , sagittal : _ y_-z plane ) , resulting in 948 dimensions in total .
our method is capable of operating in the cases where the number of dimensions are larger than the number of images . while image features from the three standard planes enables consideration of the complete information about the image it may also introduce redundant information through highly correlated features extracted from different planes .
the features we extracted were categorized as follows : * nodule size and shape properties ( e.g. , area , convex hull , circularity , etc . ) .
* pixel intensities inside the nodule ( e.g. , range , mean , standard deviation , etc . ) .
* texture of the nodule , described using grey - level co - occurrence ( glc ) features ( e.g. , entropy , contrast , etc . ) and gabor filter responses .
figure [ fig : lidc1 - 2 ] shows an initial visualization of the dataset . the high degree of inter - relationships between the image features means that it is quite difficult to understand numeric distribution of individual dimensions by a simple observation of the pcps .
this remains true even after we visualized a subset of the 364 features selected by the distance among dimensions sampling ( section [ sec : sampling ] ) .
we therefore visualized a separate set of pcps for the subspace of image features that were associated with each of the labels within the annotations .
we interactively determined the low - dimensional subspace as follows .
we first selected an annotation and data mined the rules that indicated which numeric dimensions were responsible for separation or concentration of the labels within the annotation ( see section [ sec : dimselect2 ] ) .
we then adjusted the thresholds @xmath30 and @xmath31 to vary the strength of the concentration or separation until visually clear pcps were discovered .
the resultant visualizations are depicted in figures [ fig : lidc3a ] and [ fig : lidc3b ] .
the axes of these pcps represented the most meaningful image features for each different label .
the ids of individual image features are indicated in the figures , and their names are shown in tables [ tab : annotation3 ] and [ tab : annotation4 ] . figure [ fig : lidc3a ] shows the most meaningful features for the annotations of _ nodule - level diagnosis _ , in which the polylines for label `` 0 '' are drawn in red , `` 1 '' in green , `` 2 '' in cyan , and `` 3 '' in purple ; the names of these features are summarized in table [ tab : annotation3 ] . figure [ fig : lidc3b ] shows the most meaningful features for the annotations of _ method for nodule - level diagnosis _ , in which polylines for label `` 0 '' are drawn in red , `` 1 '' in green , `` 2 '' in cyan , and `` 4 '' in purple ; these features are summarized in table [ tab : annotation4 ] .
note that no discriminatory features were found for label `` 3 '' of _ method for nodule - level diagnosis _ , indicating that perhaps new image features need to be developed to account for diagnosis after ` surgical resection ' .
visually , adjacent axes connected by densely placed parallel polylines indicate the features that are most important to a particular label .
both the visualization and table [ tab : annotation3 ] show that while axial features are useful , the introduction of features from the other viewing planes could augment the classification of lung abnormalities .
in particular , the thick , parallel polylines for the `` coronal glc texture cluster shade '' feature in the plots for both primary ( label 2 ) and metastatic ( label 3 ) disease suggest that this feature is useful for distinguishing between benign ( labels 0 and 1 ) and malignant disease ( labels 2 and 3 ) .
other features have polylines that are only parallel in one plot and so are useful for separating one label from the others , e.g. , `` coronal shape solidity '' and `` sagittal shape roughness '' are useful for separating primary disease ( initial site of cancer ) indicated by label 2 from metastatic disease ( cancer that has spread ) indicated by label 3 . similarly ,
the visualization and table [ tab : annotation4 ] show that image features from all planes are relevant to different diagnostic methods .
axial intensity and shape axis length features are related to diagnosis by biopsy ( label 2 ) and exhibit many parallel polylines .
similarly , coronal glc texture contrast , coronal intensity , and sagittal entropy would be useful for emulating radiological review ( label 1 ) .
our findings indicate that a combination of features from different viewing planes offer better discriminatory power for distinguishing different labels and annotations .
this is in contrast to existing work on the same dataset that only uses axial features @xcite .
our visualization technique has thus identified new image features that will be more meaningful to a human reader and this could enable the creation of computer - aided diagnostic tools that emulate different diagnostic processes .
for example , it could facilitate the use of optimal multi - planar image features ( table [ tab : annotation3 ] ) in computer - aided diagnosis systems , thereby allowing these applications to mimic the image interpretation of a clinical expert .
.list of meaningful features that distinguish labels from `` nodule - level diagnosis '' discovered by interactively adjusting @xmath30 and @xmath31 . [ cols="^,^,<",options="header " , ] [ tab : time ]
our technique allowed users to intuitively define and select multiple low - dimensional spaces using an interactive dimension selection mechanisms .
both our case studies discovered new patterns within the data that were not apparent when visualizing the entire high - dimensional space .
thus our case studies showed how an effective visualization of a subspace of high - dimensional data could be constructed by leveraging a user s prior knowledge of the application domain .
we suggest that our visualization technique can be adapted and expanded through the modification of its constituent components . in our current approach ,
we use classical mds to place the vertices ( dots corresponding to numeric dimensions ) of the dimension graph ( section [ sec : dimselect1 ] ) . however , under mds it is possible that there may be a distortion of the distances caused by the projection to a two - dimensional space .
thus in the view of the dimension graph it is possible that vertices in close proximity may not represent dimensions that are close . while mds was sufficient for our chosen applications ,
we suggest several other schemes that may be more suited to other applications .
non - linear dimensionality reduction algorithms ( e.g. isomap ) may create better layouts but require more computation time .
graph layout techniques such as force - directed @xcite or stress minimization @xcite layouts may also improve the visualization of the dimension graph .
similarly , alternate methods could also be used to define the order of dimensions in the pcps . in our approach
we used an approximate solution for the traveling salesman problem ( section [ sec : dimselect1 ] ) .
other approaches could also potentially be used to derive a meaningful ordering of the dimensions , especially when the dataset contains categorical values .
for example , for some applications it would be meaningful to sort the axes according to the quality of the separation of categorical values , i.e. , the dimensions that give the best separation appear first and those that give the least clear separation appear last . sorting axes according to the confidence and/or support values
could also be an effective approach for some datasets .
these approaches may assist users in understanding the importance of the displayed dimensions .
another alternative adaptation would be the integration of our two interactive dimension selection techniques to enable more focused lower - dimensional pcp visualization . in this adaptation
, the category - based rule mining would be used to plot the numeric dimensions as a dimension graph , in which the edges would be derived from the support and confidence of the association rules .
our clique - based dimension selection could then be applied to select sets of mined dimensions . as future work
, we will investigate the use of different graph layout algorithms and alternate dimension reordering approaches . also , we aim to introduce several additional functions to the current visualization including : * brushing and filtering operations on the pcps to specify the region of interest , and recalculation of dimension graph , as some related studies @xcite @xcite implement brushing interaction for both items and dimensions . * applying various definitions of distances for dimension graph . * applying various layout algorithms to dimension graph , as discussed earlier in section [ sec : discuss ] . * applying various dimension reordering for pcps , as discussed earlier in section [ sec : discuss ] .
* improvement of rendering for pcps , including cluttering avoidance , and coloring of overlapped semitransparent polylines . after implementing these functions
, we will investigate application - specific optimizations of our technique through user experiments that evaluate the effectiveness of the technique in different domains with a variety of datasets .
we would also like to integrate our visualization into a data visualization tool for medical imaging informatics @xcite as a non `` blackbox '' method of feature selection ; this would enable clinical users to conduct an expanded analysis of the relationships between ct image features and patient- or nodule - level diagnosis .
all of these experiments will be supported by detailed analysis of the visualization results by domain experts .
we presented an interactive technique for the visualization of high - dimensional data spaces .
our technique displays numeric values of the selected sets of dimensions by a set of pcps , while the dimension graph component displays the relations between the numeric dimensions and provides interactive dimension selection mechanisms .
our technique features two types of dimension selection criteria : distances among dimensions , and separateness of labels .
the former is used to visualize highly correlated sets of dimensions while the latter is used to visualize the sets of dimensions which sufficiently separate items with specific labels .
we demonstrated the effectiveness of our technique using two case studies .
we discovered important knowledge about unknown relations among design variables of airplane wing shape and objective functions , and about relations between image feature values derived from ct images and nodule - level diagnosis .
this work was supported in part by arc grants .
99 s. g. armato , et al . ,
the lung image database consortium ( lidc ) and image database resource initiative ( idri ) : a completed reference database of lung nodules on ct scans , _ medical physics _ , 38(2 ) , 915931 , 2011 .
w. k. chooi , s. matthews , m. j. bull , s. k. morcos , multislice computed tomography in staging lung cancer : the role of multiplanar image reconstruction _ journal of computed assisted tomography _ , 29(3 ) , 357360 , 2005 .
n. elmqvist , p. dragicevic , j. fekete , rolling the dice : multidimensional visual exploration using scatterplot matrix navigation , _ ieee transactions on visualization and computer graphics _ , 14(6 ) , 11411148 , 2008 .
r. kim , g. dasovich , r. bhaumik , r. brock , j. d. furst , d. s. raicu , an investigation into the relationship between semantic and content based similarity using lidc , _ acm international conference on multimedia information retrieval _
, 185192 , 2010 .
a. kumar , f. nette , k. klein , m. fulham , j. kim a visual analytics approach using the exploration of multi - dimensional feature spaces for content - based medical image retrieval , _ ieee journal of biomedical and health informatics_. in press .
dii : 10.1109/jbhi.2014.2361318 j. h. lee , k. t. mcdonell , a. zelenyuk , d. imre , k. muller , a structure - based distance metric for high - dimensional space exploration with multidimensional scaling .
_ ieee transaction on computer graphics _
, 20(3 ) , 351364 , 2013 .
g. quellec , m. lamard , l. bekri , g. cazuguel , c. roux , b. cochener medical case retrieval from a committee of decision trees , _ ieee transactions on information technology in biomedicine _ , 12271235 , 2010 .
i. solt , d. tikk , v. gl , z. t. kardkovcs , semantic classification of diseases in discharge summaries using a context - aware rule - based classifier , _ journal of the american medical informatics association _ , 580584 , 2009 .
h. suematsu , y. zheng , t. itoh , r. fujimaki , s. morinaga , y. kawahara , arrangement of low - dimenional parallel coordinate plots for high - dimensional data visualization , _
17th international conference on information visualisation _
, 5965 , 2013 . c. turkay , a. lundervoid , h. hauser , representative factor generation for the interactive visual analysis of high - dimensional data _ ieee transactions on visualization and computer graphics _ , 18(12 ) , 26212630 , 2012 .
j. yang , d. hubball , m. o. ward , e. a. rundensteiner , w. ribarsky , value and relation display : interactive visual exploration of large data sets with hundreds of dimensions , _ ieee transations on visualization and computer graphics _ , 13(3 ) , 494507 , 2007 .
x. yuan , d. ren , z. wang , c. guo , dimension projection matrix / tree : interactive subspace visual exploration and analysis of high dimensional data , _ ieee transactions on visualization and computer graphics _ , 19(12 ) , 26252633 , 2013 .
z. zhang , k. t. mcdonnel , e. zadak , k. muller , visual correlation analysis of numerical and categorical data on the correlation map , _ ieee transactions on visualization and computer graphics _ , 21(2 ) , 289303 , 2015 .
|
parallel coordinate plots ( pcps ) are among the most useful techniques for the visualization and exploration of high - dimensional data spaces .
they are especially useful for the representation of correlations among the dimensions , which identify relationships and interdependencies between variables . however , within these high - dimensional spaces , pcps face difficulties in displaying the correlation between combinations of dimensions and generally require additional display space as the number of dimensions increases . in this paper , we present a new technique for high - dimensional data visualization in which a set of low - dimensional pcps are interactively constructed by sampling user - selected subsets of the high - dimensional data space . in our technique , we first construct a graph visualization of sets of well - correlated dimensions .
users observe this graph and are able to interactively select the dimensions by sampling from its cliques , thereby dynamically specifying the most relevant lower dimensional data to be used for the construction of focused pcps .
our interactive sampling overcomes the shortcomings of the pcps by enabling the visualization of the most meaningful dimensions ( i.e. , the most relevant information ) from high - dimensional spaces .
we demonstrate the effectiveness of our technique through two case studies , where we show that the proposed interactive low - dimensional space constructions were pivotal for visualizing the high - dimensional data and discovering new patterns .
visualization , high - dimensional data , parallel coordinate plots .
|
noncommutative @xmath0-theory is widely believed to be not renormalisable in four dimensions . to underline the belief one usually draws the non - planar one - loop two - point function resulting from the noncommutative @xmath0-action .
the corresponding integral is finite , but behaves @xmath2 for small momenta @xmath3 of the two - point function .
the finiteness is important , because the @xmath3-dependence of the non - planar graph has no counterpart in the original @xmath0-action , and thus ( if divergent ) can not be absorbed by multiplicative renormalisation .
however , if we insert the non - planar graph declared as finite as a subgraph into a bigger graph , one easily builds examples ( with an arbitrary number of external legs ) where the @xmath4 behaviour leads to non - integrable integrals at small inner momenta .
this is the so - called uv / ir - mixing problem @xcite .
the heuristic argumentation can be made exact : using a sophisticated mathematical machinery , chepelev and roiban have proven a power - counting theorem @xcite which relates the power - counting degree of divergence to the topology of _ ribbon graphs_. the rough summary of the power - counting theorem is that noncommutative field theories with quadratic divergences become meaningless beyond a certain loop order .
for example , in the real noncommutative @xmath0-model there exist three - loop graphs which can not be integrated .
in this paper we prove that the real noncommutative @xmath0-model is _ renormalisable to all orders_. at first sight , this seems to be a grave contradiction .
however , we do not say that the famous papers @xcite are wrong .
in fact , we reconfirm their results , it is only that we take their message serious .
the message of the uv / ir - entanglement is that _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ noncommutativity relevant at very short distances modifies the physics of the model at very large distances . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ at large distances we have approximately a free theory .
thus , we have to alter the free theory , whereas the quasi - local @xmath0-interaction could hopefully be left unchanged . but how to modify the free action ?
we found the distinguished modification in the course of a long refinement process of our method . but
knowing the result , it can be made plausible .
it was pointed out by langmann and szabo @xcite that the @xmath5-product interaction is invariant under a duality transformation between positions and momenta , @xmath6 where @xmath7 .
using the definition of the @xmath5-product given in ( [ starprod ] ) and the reality @xmath8 one obtains @xmath9 with @xmath10 passing to quantum field theory , @xmath11 and @xmath12 become the feynman rules for the vertices in position space and momentum space , respectively .
multiplicative renormalisability of the four - point function requires that its divergent part has to be self - dual , too .
this requires an appropriate feynman rule for the propagator .
building now two - point functions with these feynman rules , it is very plausible that if the two - point function is divergent in momentum space , also the duality - transformed two - point function will be divergent .
that divergence has to be absorbed in a multiplicative renormalisation of the initial action .
however , the usual free scalar field action is not invariant under that duality transformation and therefore can not absorb the expected divergence in the two - point function .
in order to cure this problem we have to extend the free scalar field action by a harmonic oscillator potential : @xmath13 the action @xmath14 according to ( [ sfreedual ] ) and ( [ sintdual ] ) is now preserved by duality transformation , up to rescaling . from the previous considerations we can expect that also the renormalisation flow preserves that action @xmath14 .
we prove in this paper that this is indeed the case .
thus , the duality - covariance of the action @xmath14 implements precisely the uv / ir - entanglement .
of course , we can not treat the quantum field theory associated with the action ( [ sfreedual ] ) in momentum space .
fortunately , there is a matrix representation of the noncommutative @xmath15 , where the @xmath5-product becomes a simple product of infinite matrices and where the duality between positions and momenta is manifest .
the matrix representation plays an important rle in the proof that the noncommutative @xmath15 is a spectral triple @xcite .
it is also crucial for the exact solution of quantum field theories on noncommutative phase space @xcite .
coincidently , the matrix base is also required for another reason . in the traditional feynman graph approach ,
the value of the integral associated to non - planar graphs is not unique , because one exchanges the order of integrations in integrals which are not absolutely convergent . to avoid this problem
one should use a renormalisation scheme where the various limiting processes are better controlled .
the preferred method is the use of flow equations .
the idea goes back to wilson @xcite .
it was then used by polchinski @xcite to give a very efficient renormalisability proof for commutative @xmath0-theory .
several improvements have been suggested in @xcite . applying polchinski s method to the noncommutative @xmath0-model
there is , however , a serious problem in momentum space .
we have to guarantee that planar graphs only appear in the distinguished interaction coefficients for which we fix the boundary condition at the renormalisation scale @xmath16 .
non - planar graphs have phase factors which involve inner momenta .
polchinski s method consists in taking norms of the interaction coefficients , and these norms ignore possible phase factors .
thus , we would find that boundary conditions for non - planar graphs at @xmath16 are required .
since there is an infinite number of different non - planar structures , the model is not renormalisable in this way . a more careful examination of the phase factors is also not possible , because the cut - off integrals prevent the gauian integration required for the parametric integral representation @xcite .
as we show in this paper , the bare action @xmath17 according to ( [ sfreedual ] ) and ( [ sintdual ] ) corresponds to a quantum field theory which is renormalisable to all orders .
together with the duality argument , this is now the conclusive indication that the usual noncommutative @xmath0-action ( with @xmath18 ) has to be dismissed in favour of the duality - covariant action of ( [ sfreedual ] ) and ( [ sintdual ] ) .
our proof is very technical .
we do not claim that it is the most efficient one .
however , it was for us , for the time being , the only possible way .
we encountered several `` miracles '' without which the proof had failed .
the first is that the propagator is complicated but numerically accessible .
we had thus convinced ourselves that the propagator has such an asymptotic behaviour that all non - planar graphs and all graphs with @xmath19 external legs are irrelevant according to our general power - counting theorem for dynamical matrix models @xcite . however
, this still leaves an infinite number of planar two- or four - point functions which would be relevant or marginal according to @xcite . in the first versions of @xcite we had , therefore , to propose some consistency relations in order to get a meaningful theory .
miraculously , all this is not necessary .
we have further found numerically that the propagator has some universal locality properties suggesting that the infinite number of relevant / marginal planar two- or four - point functions can be decomposed into four relevant / marginal base interactions and an irrelevant remainder .
of course , there must exist a reason for such a coincidence , and the reason are orthogonal polynomials . in our case ,
it means that the kinetic matrix corresponding to the free action ( [ sfreedual ] ) written in the matrix base of the noncommutative @xmath15 is diagonalised by orthogonal meixner polynomials @xcite - model we had originally termed these polynomials `` deformed laguerre polynomials '' , which we had only constructed via its recursion relation .
the closed formula was not known to us .
thus , we are especially grateful to stefan schraml who provided us first with @xcite , from which we got the information that we were using meixner polynomials , and then with the encyclopaedia @xcite of orthogonal polynomials , which was the key to complete the renormalisation proof .
] now , having a closed solution for the free theory in the preferred base of the interaction , the desired local and asymptotic behaviour of the propagator can be derived .
we stress , however , that some of the corresponding estimations of section [ bounds ] are , so far , verified numerically only .
there is no doubt that the estimations are correct , but for the purists we have to formulate our result as follows : the quantum field theory corresponding to the action ( [ sfreedual ] ) and ( [ sintdual ] ) is renormalisable to all orders provided that the estimations given in section [ bounds ] hold .
already this weaker result is a considerable progress , because the elimination of the last possible doubt amounts to estimate a single integral .
this estimation will be performed in @xcite .
the method , further motivation and an outlook to constructive applications are already presented in @xcite . finally , let us recall that the noncommutative @xmath0-theory in two dimensions is different .
we also need the harmonic oscillator potential of ( [ sfreedual ] ) in all intermediate steps of the renormalisation proof , but at the end it can be switched off with the removal of the cut - off @xcite .
this is in agreement with the common belief that the uv / ir - mixing problem can be cured in models with only logarithmic divergences .
as the renormalisation proof is long and technical , we list here the most important steps and the main ideas . a flow chart of these steps is presented in fig .
[ fig : flowchart ] .
a more detailed introduction is given in @xcite .
( 130,130 ) ( 0,120 ) ( 80,122 ) ( 85,105 ) ( 21,95 ) ( 85,89 ) ( 0,78 ) ( 0,51 ) ( 77,70 ) ( 15,32 ) ( 81,20 ) ( 0,9 ) ( 75,-5 ) ( 80,123)(-2,-1)9.5 ( 68,117)(-2,-1)10.5 ( 56,111)(-2,-1)18 ( 40,125)(4,-1)50 ( 40,120)(2,-1)27 ( 70,105)(2,-1)10.5 ( 82,99)(2,-1)7 ( 85,110)(-4,-1)33 ( 40,90.5)(-2,-1)11.5 ( 10,115)(0,-1)30 ( 85,105)(-2,-3)17.7 ( 65.7,76)(-2,-3)11 ( 130,100)(0,-1)21 ( 105,84)(0,-1)5 ( 77,74)(-1,-1)5 ( 70,67)(-1,-1)8 ( 55,90.5)(-1,-4)7.7 ( 35,73)(0,-1)13.5 ( 60,90.5)(1,-2)30.5 ( 67,35)(3,-1)16 ( 81,25)(-4,-1)13.5 ( 66,21.25)(-4,-1)9 ( 138,100)(0,-1)70.5 ( 7,45)(0,-1)26.5 ( 61,27)(1,-1)26 ( 63,7.7)(3,-1)20 the first step is to rewrite the @xmath0-action ( [ sfreedual])+([sintdual ] ) in the harmonic oscillator base of the moyal plane , see ( [ action ] ) and ( [ g4d ] ) . the free theory is solved by the propagator ( [ noalpha ] ) , which we compute in appendix [ appmeixner ] using meixner polynomials in an essential way .
the propagator is represented by a finite sum which enables a fast numerical evaluation .
unfortunately , we can offer analytic estimations only in a few special cases .
the propagator is so complicated that a direct calculation of feynman graphs is not practicable .
therefore , we employ the renormalisation method based on flow equations @xcite which we have previously adapted to non - local ( dynamical ) matrix models @xcite .
the modification @xmath20 $ ] of the weights of the matrix indices in the kinetic term is undone in the partition function by a careful adaptation of effective action @xmath21 $ ] , which is described by the matrix polchinski equation ( [ poll ] ) . for a modification given by a cut - off function @xmath20 $ ] , renormalisation of the model amounts to prove that the matrix polchinski equation ( [ poll4 ] ) admits a regular solution which depends on a finite number of initial data . in a perturbative expansion ,
the matrix polchinski equation is solved by ribbon graphs drawn on riemann surfaces .
the existence of a regular solution follows from the general power - counting theorem proven in @xcite together with the numerical determination of the propagator asymptotics in appendix [ appb ] . however , the general proof involved an infinite number of initial conditions , which is physically not acceptable .
therefore , the challenge is to prove the reduction to a finite number of initial data for the renormalisation flow .
the answer is the integration procedure given in definition [ defint ] , sec .
[ intpoleq ] , which entails mixed boundary conditions for certain planar two- and four - point functions .
the idea is to introduce four types of reference graphs with vanishing external indices and to split the integration of the polchinski equation for the distinguished two- and four - point graphs into an integration of the difference to the reference graphs and a different integration of the reference graphs themselves .
the difference between original graph and reference graph is further reduced to differences of propagators , which we call `` composite propagators '' .
see sec .
[ seccomposite ] .
the proof of the power - countung estimations for the interaction coefficients ( proposition [ power - counting - prop ] in sec .
[ pct ] ) requires the following extensions of the general case treated in @xcite : * we have to prove that graphs where the index jumps along the trajectory between incoming and outgoing indices are suppressed .
this leaves 1pi planar four - point functions with constant index along the trajectory and 1pi planar two point functions with in total at most two index jumps along the trajectories as the only graphs which are marginal or relevant . * for these types of graphs we have to prove that the leading relevant / marginal contribution is captured by reference graphs with vanishing external indices , whereas the difference to the reference graphs is irrelevant .
this is the discrete analogue of the bphz taylor subtraction of the expansion coefficients to lowest - order in the external momenta .
thus , proposition [ power - counting - prop ] provides bounds for the interaction coefficients for the effective action at a scale @xmath22 $ ] . here , @xmath16 is the renormalisation scale where the four reference graphs are normalised , and @xmath23 is the initial scale for the integration which has to be sent to @xmath24 in order to scale away possible initial conditions for the irrelevant functions .
the estimations of proposition [ power - counting - prop ] are actually independent of @xmath23 so that the limit @xmath25 can be taken .
this already ensures the renormalisation of the model .
however , one would also like to know whether the interaction coefficients converge in the limit @xmath25 and if so , with which rate .
that analysis is performed in section [ sec : convergence ] which culminates in theorem [ final - theorem ] , confirming convergence with a rate @xmath26 .
the noncommutative @xmath1 is defined as the algebra @xmath27 which as a vector space is given by the space @xmath28 of ( complex - valued ) schwartz class functions of rapid decay , equipped with the multiplication rule @xcite @xmath29 the entries @xmath30 in ( [ starprod ] ) have the dimension of an area .
we place ourselves into a coordinate system in which @xmath31 has the form @xmath32 we use an adapted base @xmath33 where the base @xmath34 is given in @xcite .
this base satisfies @xmath35 more information about the noncommutative @xmath15 can be found in @xcite .
we are going to study a duality - covariant @xmath0-theory on @xmath27 .
this means that we add a harmonic oscillator potential to the standard @xmath0-action , which breaks translation invariance but is required for renormalisation . expanding the scalar field in the matrix base , @xmath36 , we have @xmath37 & = \int d^4x \big ( \frac{1}{2 } g^{\mu\nu } \big(\partial_\mu \phi \star \partial_\nu \phi + 4 \omega^2 ( ( \theta^{-1})_{\mu\rho } x^\rho \phi ) \star ( ( \theta^{-1})_{\nu\sigma } x^\sigma \phi ) \big ) + \frac{1}{2 } \mu_0 ^ 2 \,\phi \star \phi \nonumber \\ & \hspace*{10em } + \frac{\lambda}{4 ! }
\phi \star \phi \star
\phi \star \phi\big)(x ) \nonumber \\ & = 4\pi^2 \theta_1 \theta_2 \sum_{m , n , k , l \in \mathbb{n}^2 } \big ( \frac{1}{2 } g_{mn;kl } \phi_{mn } \phi_{kl } + \frac{\lambda}{4 ! } \phi_{mn } \phi_{nk } \phi_{kl } \phi_{lm}\big)\ ; , \label{action}\end{aligned}\ ] ] where according to @xcite @xmath38 we need , in particular , the inverse of the kinetic matrix @xmath39 , the propagator @xmath40 , which solves the partition function of the free theory ( @xmath41 ) with respect to the preferred base of the interaction .
we present the computation of the propagator in appendix [ appmeixner ] .
the result is @xmath42 here , @xmath43 is the beta - function and @xmath44 the hypergeometric function .
we have developed in @xcite the wilson - polchinski renormalisation programme @xcite for non - local matrix models where the kinetic term ( taylor coefficient matrix of the part of the action which is bilinear in the fields ) is neither constant nor diagonal . introducing a cut - off in the measure @xmath45 of the partition function @xmath46 , the resulting effect is undone by adjusting the effective action @xmath47 $ ] ( and other terms which are easy to evaluate ) .
if the cut - off function is a smooth function of the cut - off scale @xmath48 , the adjustment of @xmath21 $ ] is described by a differential equation , @xmath49}{\partial \lambda } & = \sum_{m , n , k , l } \frac{1}{2 } \lambda \frac{\partial \delta^k_{nm;lk}(\lambda)}{\partial \lambda } \bigg ( \frac{\partial l[\phi,\lambda]}{\partial \phi_{mn}}\frac{\partial l[\phi,\lambda]}{\partial \phi_{kl } } - \frac{1}{4\pi^2\theta_1\theta_2 } \big[\frac{\partial^2 l[\phi,\lambda]}{\partial \phi_{mn}\,\partial \phi_{kl}}\big]_\phi \bigg ) \ ; , \label{poll}\end{aligned}\ ] ] where @xmath50\big]_\phi:= f[\phi]-f[0]$ ] and @xmath51 here , @xmath52 is a smooth monotonous cut - off function with @xmath53 for @xmath54 and @xmath55 for @xmath56 .
the differential equation ( [ poll ] ) is referred to as the polchinski equation . in @xcite
we have derived a power - counting theorem for @xmath21 $ ] by integrating ( [ poll ] ) perturbatively between the initial scale @xmath23 and the renormalisation scale @xmath57 .
the power - counting degree is given by topological data of ribbon graphs and two scaling exponents of the ( summed and differentiated ) cut - off propagator .
the power - counting theorem in @xcite is model independent , but it relied on boundary conditions for the integrations which do not correspond to a physically meaningful model . in this paper we will show that the four - dimensional duality - covariant noncommutative @xmath0-theory given by the action ( [ action ] ) admits an improved power - counting behaviour which only relies on a finite number of physical boundary conditions for the integration .
the first step is to extract from the power - counting theorem @xcite the set of relevant and marginal interactions , which on the other hand is used as an input to derive the power - counting theorem . to say it differently
: one has to be lucky to make the right ansatz for the initial interaction which is then reconfirmed by the power - counting theorem as the set of relevant and marginal interactions .
we are going to prove that the following ansatz for the initial interaction is such a lucky choice : @xmath58 \nonumber \\ * & = \sum_{m^1,m^2,n^1,n^2 \in \mathbb{n}}\frac{1}{2\pi\theta } \big(\frac{1}{2 } \big(\rho_1 ^ 0 { + } ( n^1{+}m^1{+ } n^2{+}m^2)\rho_2 ^ 0 \big ) \phi_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } \phi_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } \nonumber \\ * & \hspace*{5em } - \rho_3 ^ 0 \big ( \sqrt{n^1m^1 } \phi_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } \phi_{{\genfrac{}{}{0pt}{}}{n^1 - 1}{n^2}{\genfrac{}{}{0pt}{}}{m^1 - 1}{m^2 } } + \sqrt{n^2m^2 } \phi_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } \phi_{{\genfrac{}{}{0pt}{}}{n^1}{n^2 - 1}{\genfrac{}{}{0pt}{}}{m^1}{m^2 - 1}}\big)\big ) \nonumber \\ * & + \sum_{m^1,m^2,n^1,n^2,k^1,k^2 , l^1,l^2 \in \mathbb{n } } \frac{1}{4 ! } \rho_4 ^ 0 \,\phi_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } \phi_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } } \phi_{{\genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2 } } \phi_{{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}\;. \label{ct4}\end{aligned}\ ] ] for simplicity we impose a symmetry between the two components @xmath59 of matrix indices @xmath60 , which could be relaxed by taking different @xmath61-coefficients in front of @xmath62 and @xmath63 .
accordingly , we choose the same weights in the noncommutativity matrix , @xmath64 .
the differential equation ( [ poll ] ) is non - perturbatively defined .
however , we shall solve it perturbatively as a formal power series in a coupling constant @xmath65 which later on will be related to a normalisation condition at @xmath66 , see ( [ initrho ] ) .
we thus consider the following expansion : @xmath67 \nonumber \\ * & = \sum_{v = 1}^\infty \lambda^v \sum_{n=2}^{2v+2 } \frac{(2\pi\theta)^{\frac{n}{2}-2}}{n ! } \sum_{m_i , n_i } a^{(v)}_{m_1n_1;\dots;m_nn_n}[\lambda,\lambda_0,\rho^0 ] \phi_{m_1n_1}\cdots \phi_{m_nn_n}\;. \label{lg4}\end{aligned}\ ] ] inserting ( [ lg4 ] ) into ( [ poll ] ) we obtain @xmath68 \nonumber \\ * & = \ !
\sum_{n_1=2}^n \sum_{v_1=1}^{v-1 } \sum_{m , n , k , l \in \mathbb{n}^2 } \frac{1}{2 } q_{nm;lk}(\lambda ) a^{(v_1)}_{m_1n_1;\dots;m_{n_1 - 1}n_{n_1 - 1};mn}[\lambda ] a^{(v - v_1)}_{m_{n_1}n_{n_1};\dots;m_{n}n_{n};kl}[\lambda ] \nonumber \\*[-1ex ] & \hspace*{15em } + \big(\binom{n}{n_1{-}1 } -1\big ) \text { permutations } \nonumber \\ * & - \sum_{m , n , k , l\in \mathbb{n}^2 } \frac{1}{2 } q_{nm;lk}(\lambda ) a^{(v)}_{m_1n_1;\dots;m_{n}n_{n};mn;kl}[\lambda]\ ; , \label{poll4}\end{aligned}\ ] ] with @xmath69 we are going to compute the functions @xmath70 by iteratively integrating the polchinski equation ( [ poll4 ] ) starting from boundary conditions either at @xmath16 or at @xmath23
. the right choice of the integration direction is an art : the boundary condition influences crucially the estimation , which in turn justifies or discards the original choice of the boundary condition . at the end of numerous trial - and - error experiments with the boundary condition ,
one convinces oneself that the procedure described in definition [ defint ] below is up to finite re - normalisations discussed later the unique possibility to renormalise the model .
first we have to recall @xcite the graphology resulting from the polchinski equation ( [ poll4 ] ) .
the polchinski equation is solved by _ ribbon graphs _ drawn on a riemann surface of uniquely determined _ genus _
@xmath71 and uniquely determined number @xmath72 of _ boundary components _ ( holes ) .
the ribbons are made of double - line propagators @xmath73 attached to vertices @xmath74 under certain conditions verified by our model , the rough power - counting behaviour of the ribbon graph is determined by the topology @xmath75 of the riemann surface and the number of vertices @xmath76 and external legs @xmath77 . however , in order to prove this behaviour we need some auxiliary notation : the number @xmath78 of external vertices ( vertices to which at least one external leg is attached ) , a certain segmentation index @xmath79 and a certain summation of graphs with appropriately varying indices .
we recall in detail the index summation , because we need it for a refinement of the general proof given in @xcite . viewing the ribbon graph as a set of single - lines , we can distinguish closed and open lines .
the open lines are called _ trajectories _ starting at an incoming index @xmath80 , running through a chain of inner indices @xmath81 and ending at an outgoing index @xmath82 .
each index belongs to @xmath83 , its components are labelled by superscripts , e.g. @xmath84 .
we define @xmath85=\mathfrak{i}[k_j]$ ] and @xmath86=\mathfrak{o}[k_j]$ ] .
there is a conservation of the total amount of indices , @xmath87 ( as vectors in @xmath83 ) .
an index summation @xmath88 is a summation over the graphs with outgoing indices @xmath89 where @xmath90,\dots,\mathfrak{i}[m_s]$ ] are kept fixed .
the number of these summations is restricted by @xmath91 . due to the symmetry properties of the propagator
one could equivalently sum over @xmath92 where @xmath93 $ ] are fixed .
a graph @xmath94 is produced via a certain _ history _ of contractions of ( in each step ) either two smaller subgraphs ( with fewer vertices ) or a self - contraction of a subgraph with two additional external legs . at a given order @xmath76 of vertices
there are finitely many graphs ( distinguished by their topology and the permutation of external indices ) contributing the part @xmath95 to a function @xmath70 .
it is therefore sufficient to prove estimations for each @xmath95 separately .
a ribbon graph is called one - particle irreducible ( 1pi ) if it remains connected when removing a single propagator .
the first term on the rhs of the polchinski equation ( [ poll4 ] ) leads always to one - particle reducible graphs , because it is left disconnected when removing the propagator @xmath96 in ( [ poll4 ] ) .
according to the detailed properties a graph @xmath94 , which is possibly a generalisation of the original ribbon graphs as explaned in section [ seccomposite ] below , we define the following recursive procedure ( starting with the vertex ( [ vertex ] ) which does not have any subgraphs ) to integrate the polchinski equation ( [ poll4 ] ) : [ defint ] we consider generalised . ]
ribbon graphs @xmath94 which result via a history of contractions of subgraphs which at each contraction step have already been integrated according to the rules given below . 1
. [ defint4 ] let @xmath94 be a planar ( @xmath97 , @xmath98 ) one - particle irreducible graph with @xmath99 external legs , where the index along each of its trajectories is constant ( this includes the two external indices of a trajectory and the chain of indices at contracting inner vertices in between them ) .
then , the contribution @xmath100 $ ] ( using the natural cyclic order of legs of a planar graph ) of @xmath94 to the effective action is integrated as follows : @xmath101 \nonumber \\ * & : = -\int_{\lambda}^{\lambda_0 } \!\frac{d \lambda'}{\lambda ' } \ ; \bigg\ { \widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2};{\genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2 } ; { \genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda ' ] } \nonumber \\ * & \hspace*{10em } - ~ \parbox{17\unitlength}{\begin{picture}(17,17 ) \put(0,0){\epsfig{file = h04,scale=0.9,bb=71 638 117 684 } } \put(-1,7.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(6,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(12,7.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{k^1}{k^2}$ } } \put(6,14){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{l^1}{l^2}$ } } \end{picture } } \cdot\left[\widehat{\lambda ' \frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda']}\right ] \bigg\ } \nonumber \\ * & + ~~ \parbox{19\unitlength}{\begin{picture}(17,17 ) \put(0,0){\epsfig{file = vert1,scale=0.9,bb=71 638 117 684 } } \put(-1,7.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(6,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(12,7.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{k^1}{k^2}$ } } \put(6,14){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{l^1}{l^2}$ } } \end{picture } } \left [ \int_{\lambda_r}^\lambda \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda ' \frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda']}\big ) + a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_r]\right]\;. \label{inta4}\end{aligned}\ ] ] here ( and in the sequel ) , the wide hat over the @xmath102-derivative of an @xmath103-function indicates that the rhs of the polchinski equation ( [ poll4 ] ) has to be inserted .
the two vertices in the third and fourth lines of ( [ inta4 ] ) are identical ( both are equal to @xmath104 ) .
the four - leg graph in the third line of ( [ inta4 ] ) indicates that the graph corresponding to the function in brackets right of it has to be inserted into the holes .
the result is a graph with the same topology as the function in the second line , but different indices on inner trajectories .
the graph in the fourth line of ( [ inta4 ] ) is identical to the original vertex ( [ vertex ] ) .
the different symbol shall remind us that in the analytic expression for subgraphs containing the vertex of the last line in ( [ inta4 ] ) we have to insert the value in brackets right of it .
+ 1ex + remark : we use here ( and in all other cases discussed below ) the convention ( its consistency will be shown later ) that at @xmath105 the contribution to the initial four - point function is independent of the external matrix indices , @xmath106= a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2};{\genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2 } ; { \genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda_0]$ ] .
this is not really necessary , we could admit @xmath107-a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0]=c_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2};{\genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2 } ; { \genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda_0]$ ] with @xmath108\big| \leq \frac{\mathrm{const}}{\theta \lambda_0
^ 2}$ ] . + 2ex 2 .
[ defint0 ]
let @xmath94 be a planar ( @xmath97 , @xmath98 ) 1pi graph with @xmath109 external legs , where either the index is constant along each trajectory , or one component of the index jumps be the sequence of indices at inner vertices on the considered trajectory @xmath110 , in correct order between @xmath80 and @xmath82 .
then , for either @xmath111 or @xmath112 we require @xmath113 for all @xmath114 \cup [ q , a{-}1]$ ] and @xmath115 ( fixed sign ) for all @xmath116 $ ] . the cases @xmath117 , @xmath118 and @xmath119 are admitted .
the other index component is constant along the trajectory . ]
once by @xmath120 and back on one of the trajectories , whereas the index along the possible other trajectory remains constant .
then , the contribution @xmath121 $ ] of @xmath94 to the effective action is integrated as follows : @xmath122 \nonumber \\ * & : = -\int_{\lambda}^{\lambda_0 } \!\frac{d \lambda'}{\lambda ' } \ ; \left\ { \widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda']}\right .
\nonumber \\ * & \quad - ~~\parbox{23\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = h01,scale=0.9,bb=71 666 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \end{picture } } \cdot \left[\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } \right .
\nonumber \\*[-1ex ] & \hspace*{10em } + m^1 \big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\ * & \hspace*{10em } + m^2 \big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\ * & \hspace*{10em } + n^1 \big(\lambda'\widehat{\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\ * & \hspace*{10em } + n^2 \left.\left .
\big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \right]\right\ } \nonumber \\ & + ~~\parbox{24\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = at3,scale=0.9,bb=71 666 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \end{picture } } \left [ \int_{\lambda_r}^{\lambda } \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } \big ) + a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r ] \right .
\nonumber \\ * & \hspace*{8em } + m^1 \bigg(\int_{\lambda_r}^{\lambda } \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\*[-0.5ex ] & \hspace*{15em } + a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0}}[\lambda_r ] { -}a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r]\bigg ) \nonumber \\ * & \hspace*{8em } + m^2 \bigg(\int_{\lambda_r}^{\lambda } \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\*[-0.5ex ] & \hspace*{15em } + a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1}}[\lambda_r ] { - } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r]\bigg ) \nonumber \\ * & \hspace*{8em } + n^1 \bigg(\int_{\lambda_r}^{\lambda } \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\*[-0.5ex ] & \hspace*{15em } + a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r ] { -}a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r]\bigg ) \nonumber \\ * & \hspace*{8em } + n^2 \bigg(\int_{\lambda_r}^{\lambda } \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\*[-0.5ex ] & \hspace*{15em } + \left .
a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r ] { -}a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r]\bigg ) \right ] .
\label{inta1}\end{aligned}\ ] ] 3 .
[ defint1 ] let @xmath94 be a planar ( @xmath97 , @xmath98 ) 1pi graph having @xmath109 external legs with external indices @xmath123 ( equal sign ) or @xmath124 , with a single and sequences @xmath125 ( @xmath126 ) of indices at inner vertices on the trajectory @xmath127 ( @xmath128 ) this means that there exist labels @xmath129 with @xmath130 for all @xmath114 $ ] , @xmath131 for all @xmath132 $ ] and @xmath133 for all @xmath134 $ ] on one trajectory and @xmath135 for all @xmath136 $ ] , @xmath137 for all @xmath138 $ ] and @xmath139 for all @xmath140 $ ] on the other trajectory .
the cases @xmath141 and @xmath142 are admitted . ]
jump in the index component of each trajectory . under these conditions
the contribution of @xmath94 to the effective action is integrated as follows : @xmath143 \nonumber \\ * & : = -\int_{\lambda}^{\lambda_0 } \!\frac{d \lambda'}{\lambda ' } \ ; \left\ { \widehat{\lambda ' \frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda ' ] } \right . \nonumber
\\*[-2ex ] & \hspace*{7em } \left .
- \sqrt{(m^1{+}1)(n^1{+}1 ) } ~~\parbox{24\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = h03,scale=0.9,bb=71 666 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \end{picture } } \cdot \left [ \widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } \right ] \right\ } \nonumber \\ * & + \sqrt{(m^1{+}1)(n^1{+}1 ) } ~~\parbox{24\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = at3b , scale=0.9,bb=71 665 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \end{picture } } \bigg[\int_{\lambda_r}^{\lambda } \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda ' \frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } \big ) \nonumber \\ * & \hspace*{22em } + a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r ] \bigg]\ , , \label{inta2 } \\ & a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2 + 1}{\genfrac{}{}{0pt}{}}{n^1}{n^2 + 1 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda ] \nonumber \\ * & : = -\int_{\lambda}^{\lambda_0 } \!\frac{d \lambda'}{\lambda ' } \ ; \left\ { \widehat{\lambda ' \frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2 + 1}{\genfrac{}{}{0pt}{}}{n^1}{n^2 + 1 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda ' ] } \right .
\nonumber \\*[-2ex ] & \hspace*{7em } \left .
- \sqrt{(m^2{+}1)(n^2{+}1 ) } ~~\parbox{24\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = h03,scale=0.9,bb=71 666 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2 + 1}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2 + 1}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \end{picture } } \cdot \left [ \widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } \right ] \right\ } \nonumber \\ * & + \sqrt{(m^2{+}1)(n^2{+}1 ) } ~~\parbox{24\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = at3b , scale=0.9,bb=71 665 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2 + 1}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2 + 1}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \end{picture } } \bigg[\int_{\lambda_r}^{\lambda } \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda ' \frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } \big ) \nonumber \\ * & \hspace*{22em } + a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_r ] \bigg]\ , .
\label{inta3}\end{aligned}\ ] ] 4 .
[ defintnp ] let @xmath94 be any other type of graph .
this includes non - planar graphs ( @xmath144 and/or @xmath145 ) , graphs with @xmath146 external legs , one - particle reducible graphs , four - point graphs with non - constant index along at least one trajectory and two - point graphs where the integrated absolute value of the jump along the trajectories is bigger than @xmath147 .
then the contribution of @xmath94 to the effective action is integrated as follows : @xmath148 & : = -\int_{\lambda}^{\lambda_0 } \!\frac{d \lambda'}{\lambda ' } \ ; \big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{m_1n_1;\dots;m_nn_n}[\lambda']}\big ) \;.\end{aligned}\ ] ] the previous integration procedure identifies the following distinguished functions @xmath149 $ ] : [ rhoa ] @xmath150 & : = \sum_{\text{$\gamma$ as in def.~\ref{defint}.\ref{defint0 } } } a^\gamma_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0]\ ; , \label{rho1 } \\
\rho_2[\lambda,\lambda,\rho^0 ] & : = \sum _ { \text{$\gamma$ as in def.~\ref{defint}.\ref{defint0 } } } \big(a^\gamma_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0}}[\lambda,\lambda_0,\rho^0 ] - a^\gamma_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0]\big)\ ; , \label{rho2 } \\
\rho_3[\lambda,\lambda_0,\rho^0 ] & : = \sum_{\text{$\gamma$ as in def.~\ref{defint}.\ref{defint1 } } } \big(- a^\gamma_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0 ] \big ) \ ; , \label{rho3 } \\
\rho_4[\lambda,\lambda_0,\rho^0 ] & : = \sum_{\text{$\gamma$ as in def.~\ref{defint}.\ref{defint4 } } } a^\gamma_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda,\lambda_0,\rho^0]\;. \label{rho4}\end{aligned}\ ] ] this identification uses the symmetry properties of the @xmath151-functions when summed over all contributing graphs .
it follows from definition [ defint ] and ( [ ct4 ] ) that @xmath152
\equiv \rho_a^0\;,\qquad a=1,\dots,4\;. \label{rhoa0}\end{aligned}\ ] ] as part of the renormalisation strategy encoded in definition [ defint ] , the coefficients ( [ rhoa ] ) are kept constant at @xmath66 .
we define @xmath153 & = 0 \qquad \text{for } a=1,2,3\ ; , & \rho_4[\lambda_r,\lambda_0,\rho^0 ] & = \lambda\;. \label{initrho}\end{aligned}\ ] ] the normalisation ( [ initrho ] ) for @xmath154 identifies @xmath155 as the cut - off propagator related to the normalised two - point function at @xmath16 .
this entails a normalisation of the mass @xmath156 , the oscillator frequency @xmath157 and the amplitude of the fields @xmath158 .
the normalisation condition for @xmath159 $ ] defines the coupling constant used in the expansion ( [ lg4 ] ) .
it is convenient to write the linear combination of the functions in braces @xmath160 in ( [ inta4])([inta3 ] ) as a ( non - unique ) linear combination of graphs in which we find at least one of the following composite propagators : [ comp - prop ] @xmath161 & \mathcal{q}^{(1)}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } \nonumber \\ * & : = \mathcal{q}^{(0)}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } - m^1 \mathcal{q}^{(0)}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{1}{0 } } - m^2 \mathcal{q}^{(0)}_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{0}{1 } } \hspace*{-1em } & = \parbox{23\unitlength}{\begin{picture}(20,8 ) \put(0,2){\epsfig{file = q1,scale=0.9,bb=71 667 131 677 } } \put(1,-1.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(16,-1.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(2,7.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(17,7.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \end{picture } } \label{comp - prop-1 } \\[2ex ] & \mathcal{q}^{(+\frac{1}{2})}_{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } \nonumber \\ * & : = q_{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } - \sqrt{m^1{+}1 } q_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{0}{0 } } & = \parbox{23\unitlength}{\begin{picture}(20,8 ) \put(0,2){\epsfig{file = qp12,scale=0.9,bb=71 667 132 677 } } \put(1,-1.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}$ } } \put(16,-1.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(2,7){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}$ } } \put(17,7){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \end{picture } } \label{comp - prop - p12 } \\[2ex ] & \mathcal{q}^{(-\frac{1}{2})}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2 + 1}{\genfrac{}{}{0pt}{}}{n^1}{n^2 + 1 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } \nonumber \\*[-1ex ] & : = q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2 + 1}{\genfrac{}{}{0pt}{}}{n^1}{n^2 + 1};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } - \sqrt{m^2{+}1 } q_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{n^1}{n^2 + 1 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{0}{0 } } & = \raisebox{1ex}{\parbox{23\unitlength}{\begin{picture}(20,8 ) \put(0,2){\epsfig{file = qm12,scale=0.9,bb=71 667 132 677 } } \put(1,-1.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2 + 1}$ } } \put(16,-1.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(2,7){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2 + 1}$ } } \put(17,7){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \end{picture } } } \label{comp - prop - m12}\end{aligned}\ ] ] to obtain the linear combination we recall how the graph @xmath94 under consideration is produced via a history of contractions and integrations of subgraphs .
for a history @xmath162-@xmath163-@xmath164-@xmath80 ( @xmath162 first ) we have @xmath165 \nonumber \\ * & = \sum_{m_a , n_a , k_a , l_a,\dots , m_n , n_n , k_n , l_n } \int_{(\lambda)_a}^{(\lambda)_b } \ !
\frac{d \lambda_n}{\lambda_n } \int_{(\lambda_{n})_a}^{(\lambda_{n})_b } \ !
\frac{d \lambda_{n-1}}{\lambda_{n-1 } } \dots \int_{(\lambda_b)_a}^{(\lambda_b)_b } \
! \frac{d \lambda_a}{\lambda_a } \nonumber \\ * & \qquad \times q_{m_nn_n;k_nl_n}(\lambda_n ) \dots q_{m_bn_b;k_bl_b}(\lambda_b ) q_{m_an_a;k_al_a}(\lambda_a)\ , v^{m_an_ak_al_a\dots m_nn_nk_nl_n } _ { m_1n_1\dots m_n n_n}\ ; , \label{inthist}\end{aligned}\ ] ] where @xmath166 is the vertex operator and either @xmath167 or @xmath168 .
the graph @xmath169 $ ] is obtained via the same procedure ( including the choice of the integration direction ) , except that we use the vertex operator @xmath170 .
this means that all propagator indices which are not determined by the external indices are the same .
therefore , we can factor out in the difference of graphs all completely inner propagators and the integration operations .
we first consider the difference in ( [ inta4 ] ) .
since @xmath94 is one - particle irreducible with constant index on each trajectory , we get for a certain permutation @xmath171 ensuring the history of integrations @xmath172 } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma}_{00;00;00;00}[\lambda ' ] } \nonumber \\ * & = \dots \left\ { \prod_{i=1}^a q_{m_{\pi_i}k_{\pi_i};k_{\pi_i}m_{\pi_i}}(\lambda_{\pi_i } ) - \prod_{i=1}^a q_{0k_{\pi_i};k_{\pi_i}0}(\lambda_{\pi_i } ) \right\ } \nonumber \\ * & = \dots \bigg\ { \sum_{b=1}^a \big(\prod_{i=1}^{b-1 } q_{0k_{\pi_i};k_{\pi_i}0}(\lambda_{\pi_i } ) \big ) \mathcal{q}^{(0)}_{m_{\pi_b}k_{\pi_b};k_{\pi_b}m_{\pi_b}}(\lambda_{\pi_b } ) \nonumber \\*[-1ex ] & \hspace*{10em } \times \big(\prod_{j = b+1}^a q_{m_{\pi(j)}k_{\pi(j ) } ; k_{\pi(j)}m_{\pi(j)}}(\lambda_{\pi(j ) } ) \big ) \bigg\}\ ; , \label{deca4}\end{aligned}\ ] ] where @xmath94 contains @xmath162 propagators with external indices and @xmath173 .
the parts of the analytic expression common to both @xmath174}$ ] and @xmath175}$ ] are symbolised by the dots .
the @xmath176 are inner indices .
we thus learn that the difference of graphs appearing in the braces in ( [ inta4 ] ) can be written as a sum of graphs each one having a composite propagator ( [ comp - prop-0 ] ) . of course , the identity ( [ deca4 ] ) is nothing but a generalisation of @xmath177 .
there are similar identities for the differences appearing in ( [ inta1])([inta3 ] ) .
we delegate their derivation to appendix [ appcompositeidentity ] . in appendix
[ app : graph ] we show how the difference operation works for a concrete example of a two - leg graph . differentiating the cut - off propagator ( [ gkdk1 ] ) with respect to @xmath48 and recalling that the cut - off function @xmath52 is constant unless @xmath178 $ ] , we notice that for our choice @xmath64 the indices are restricted as follows : @xmath179 in particular , the volume of the support of the differentiated cut - off propagator ( [ qc4 ] ) with respect to a single index @xmath180 equals @xmath181 , which is the correct normalisation of a four - dimensional model @xcite .
we compute in appendix [ appb ] the @xmath48-dependence of the maximised propagator @xmath182 , which is the application of the sharp cut - off realising the condition ( [ qc4 ] ) to the propagator , for selected values of @xmath183 and @xmath157 , which is extremely well reproduced by ( [ form4 - 1 ] ) .
we thus obtain for the maximum of ( [ q4 ] ) @xmath184 \displaystyle \frac{c_0}{\sqrt{\lambda^2\theta}}\ , \delta_{m+k , n+l } \qquad & \text{for } \omega=0\ ; , \end{array}\right .
\label{est4 - 0}\end{aligned}\ ] ] where @xmath185 .
the constant @xmath186 corrects the fact that ( [ form4 - 1 ] ) holds asymptotically only .
next , from ( [ form4 - 2 ] ) we obtain @xmath187 where @xmath188 .
the product of ( [ est4 - 0 ] ) by the volume @xmath189 of the support of the cut - off propagator with respect to a single index leads to the following bound : @xmath190 \displaystyle 4c_0 \,\big(\sqrt{\theta}\lambda\big)^3\ , \qquad & \text{for } \omega=0\;. \end{array}\right .
\label{est4 - 2}\end{aligned}\ ] ] according to ( [ delta - ss ] ) there is the following refinement of the estimation ( [ est4 - 0 ] ) : @xmath191 this property will imply that graphs with big total jump along the trajectories are suppressed , provided that the indices on the trajectory are `` small '' .
however , there is a potential danger from the presence of completely inner vertices , where the index summation runs over `` large '' indices as well .
fortunately , according to ( [ form4 - 3 ] ) this case can be controlled by the following property of the propagator : @xmath192 where we have defined the following norms : @xmath193 moreover , we define @xmath194 finally , we need estimations for the composite propagators ( [ comp - prop ] ) and ( [ comp - prop - rest ] ) : @xmath195 these estimations follow from ( [ delta-00 ] ) and ( [ delta-01 ] ) .
now we are going to prove a power - counting theorem for the @xmath0-model in the matrix base , generalising the theorem proven in @xcite .
the generalisation concerns 1pi planar graphs and their subgraphs .
a subgraph of a planar graph has necessarily genus @xmath196 and an even number of legs on each boundary component .
we distinguish one boundary component of the subgraph which after a sequence of contractions will be part of the unique boundary component of an 1pi planar graph . for a trajectory @xmath110 on the distinguished boundary component , which passes through the indices
@xmath197 when going from @xmath80 to @xmath86 $ ] , we define the total jump as @xmath198 clearly , the jump is additive : if we connect two trajectories @xmath110 and @xmath199 to a new trajectory @xmath200 , then @xmath201 .
we let @xmath202 be a set of trajectories @xmath203}$ ] on the distinguished boundary component for which we measure the total jump . by definition
, the end points of a trajectory in @xmath202 can not belong to @xmath204 .
moreover , we consider a second set @xmath205 of @xmath206 trajectories @xmath203}$ ] of the distinguished boundary component where one of the end points @xmath207 or @xmath92 is kept fixed and the other end point is summed over .
however , we require the summation to run over @xmath208 } \rangle \geq 5 $ ] only , see ( [ est4 - 4 ] ) .
we let @xmath209 be the corresponding summation operator .
additionally , we have to introduce a new notation in order to control * the behaviour for large indices and given @xmath48 , * the behaviour for given indices and large @xmath48 . for this purpose
we let @xmath210 $ ] denote a function of the indices @xmath211 @xmath212 and the scale @xmath48 which is bounded as follows : @xmath213 \leq \left\ { \begin{array}{ll } c_a m^{a } & \text{for } m \geq 1\ ; , \\ c_b m^{b } & \text{for } m \leq 1 \ ; , \end{array } \right .
\label{polab } \\ * m & : = \max_{m_i , n_i \notin \mathcal{e}^s,\mathcal{e}^{t ' } } \big ( \frac{m_1^r+1}{2\theta \lambda^2 } , \frac{n_1^r+1}{2\theta \lambda^2 } , \dots , \frac{n_n^r+1}{2
\theta \lambda^2}\big ) \ ; , \nonumber\end{aligned}\ ] ] for some constants @xmath214 .
the maximisation over the indices @xmath215 excludes the summation indices @xmath216 . fixing the indices and varying @xmath48 we have @xmath217 \leq p^{a}_{b}\big [ \frac{m_1n_1;\dots;m_nn_n}{\theta \lambda^2}\big ] \ ; , \label{polab - range}\end{aligned}\ ] ] for @xmath218 and @xmath219 , assuming appropriate @xmath214 .
moreover , @xmath220 & p^{a_2}_{b_2}\big [ \frac{m_{n_1 + 1}n_{n_1 + 1};\dots;m_nn_n } { \theta \lambda^2}\big ] \nonumber \\ * & \leq p^{a_1+a_2}_{b_1+b_2}\big [ \frac{m_1n_1;\dots;m_nn_n}{\theta \lambda^2}\big ] \;. \label{pabring}\end{aligned}\ ] ] we are going to prove : [ power - counting - prop ] let @xmath94 be a ribbon graph having @xmath77 external legs , @xmath76 vertices , @xmath78 external vertices and segmentation index @xmath79 , which is drawn on a genus-@xmath71 riemann surface with @xmath72 boundary components .
we require the graph @xmath94 to be constructed via a history of subgraphs and an integration procedure according to definition [ defint ] .
then the contribution @xmath221 of @xmath94 to the expansion coefficient of the effective action describing a duality - covariant @xmath0-theory on @xmath27 in the matrix base is bounded as follows : 1 .
[ prp - planar - a4 ] if @xmath94 is as in definition [ defint].[defint4 ] , we have + [ prp - a4 ] @xmath222 & - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda,\lambda_0,\rho_0]\big| \nonumber \\ * & \hspace*{-13em } \leq p^{4v - n}_1 \bigg[\frac{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } ; { \genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2};{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2 } \bigg ] \big(\frac{1}{\omega}\big)^{3v-2-v^e } \,p^{2v-2}\big[\ln\frac{\lambda}{\lambda_r}\big ] \ ; , \label{hata4 } \\ \big|a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] \big| & \leq \big(\frac{1}{\omega}\big)^{3v-2-v^e } p^{2v-2}\big[\ln\frac{\lambda}{\lambda_r}\big ] \;.
\label{hata4 - 0 } \end{aligned}\ ] ] 2 .
[ prp - planar - a0 ] if @xmath94 is as in definition [ defint].[defint0 ] , we have + [ prp - a0 ] @xmath223 & - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] \nonumber \\ * & - m^1 \big ( a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0 } } [ \lambda,\lambda_0,\rho_0 ] -a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] \big ) \nonumber \\ * & - n^1\big ( a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] -a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] \big ) \nonumber \\ * & - m^2\big ( a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1 } } [ \lambda,\lambda_0,\rho_0 ] -a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] \big ) \nonumber \\ * & - n^2 \big ( a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] -a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] \big ) \big| \nonumber \\ * & \hspace*{-8em } \leq ( \theta \lambda^2)\ , p^{4v - n}_2 \bigg[\frac{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2 } \bigg ] \big(\frac{1}{\omega}\big)^{3v-1-v^e } \ , p^{2v-1}\big[\ln\frac{\lambda}{\lambda_r}\big ] \ ; , \label{hata0 } \\
\big|a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0]\big| & \leq ( \theta \lambda^2)\big(\frac{1}{\omega}\big)^{3v-1-v^e } \ , p^{2v-1}\big[\ln\frac{\lambda}{\lambda_r}\big ] \ ; , \label{hata0 - 0 } \\ \big|a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0 } } [ \lambda,\lambda_0,\rho_0 ] & -a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0]\big| \nonumber \\ * & \leq \big(\frac{1}{\omega}\big)^{3v-1-v^e } \ , p^{2v-1}\big[\ln\frac{\lambda}{\lambda_r}\big ] \;. \label{hata0 - 1}\end{aligned}\ ] ] 3 .
[ prp - planar - a1 ] if @xmath94 is as in definition [ defint].[defint1 ] , we have + [ prp - a1 ] @xmath224 - \sqrt{(m^1{+}1)(n^1{+}1 ) } a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0 ] \big| \nonumber \\ * & \quad \leq ( \theta \lambda^2)\ , p^{4v - n}_2 \bigg[\frac{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2 } \bigg ] \big(\frac{1}{\omega}\big)^{3v-1-v^e } \ , p^{2v-1}\big[\ln\frac{\lambda}{\lambda_r}\big ] \ ; , \label{hata1 } \\ & \big| a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho_0]\big| \leq \big(\frac{1}{\omega}\big)^{3v-1-v^e } \ , p^{2v-1}\big[\ln\frac{\lambda}{\lambda_r}\big ] \;. \label{hata1 - 0}\end{aligned}\ ] ] 4 .
[ prp - sub - planar ] if @xmath94 is a subgraph of an 1pi planar graph with a selected set @xmath202 of trajectories on one distinguished boundary component and a second set @xmath205 of summed trajectories on that boundary component , we have @xmath225 \big| \nonumber \\ * & ~~
\leq \big(\theta \lambda^2\big)^{(2-\frac{n}{2})+2(1-b ) } p^{4v - n}_{\big(2 t ' + \sum_{\overrightarrow{n_j\mathfrak{o}[n_j ] } \in t } \min(2 , \frac{1}{2 } \langle \overrightarrow{n_j\mathfrak{o}[n_j ] } \rangle ) \big ) } \big [ \frac{m_1n_1;\dots;m_nn_n}{\theta \lambda^2 } \big ] \nonumber \\ * & \qquad \times \big(\frac{1}{\omega}\big)^{3v-\frac{n}{2}-1 + b - v^e-\iota+s+t ' } p^{2v-\frac{n}{2}}\big[\ln \frac{\lambda}{\lambda_r}\big]\;. \label{an4 - 1plan}\end{aligned}\ ] ] 5 .
[ prp - non - planar ] if @xmath94 is a non - planar graph , we have @xmath226 \big| \nonumber \\ * & \quad \leq \big(\theta \lambda^2\big)^{(2-\frac{n}{2})+2(1-b-2g)}\ , p^{4v - n}_0 \big [ \frac{m_1n_1;\dots;m_nn_n}{\theta \lambda^2 } \big ] \nonumber \\ * & \qquad \times \big(\frac{1}{\omega}\big)^{3v-\frac{n}{2}-1+b+2g - v^e-\iota+s}\ , \,p^{2v-\frac{n}{2}}\big[\ln \frac{\lambda}{\lambda_r}\big]\;. \label{an4 - 1}\end{aligned}\ ] ] _ proof .
_ we prove the proposition by induction upward in the vertex order @xmath76 and for given @xmath76 downward in the number @xmath77 of external legs . 1 .
we start with with the proof for non - planar graphs , noticing that due to ( [ polab - range ] ) the estimations ( [ prp - a4 ] ) , ( [ prp - a0 ] ) , ( [ prp - a1 ] ) and ( [ an4 - 1plan ] ) can be further bounded by ( [ an4 - 1 ] ) .
the proof of ( [ an4 - 1 ] ) reduces to the proof of the general power - counting theorem given in @xcite , where we have to take for @xmath227 @xmath228 and @xmath229 the estimations ( [ est4 - 0 ] ) , ( [ est4 - 1 ] ) and ( [ est4 - 2 ] ) with both their @xmath48- and @xmath157-dependence
. independent of the factor ( [ polab ] ) , the non - planarity of the graph guarantees the irrelevance of the corresponding function so that the integration according to definition [ defint ] agrees with the procedure used in @xcite .
the dependence on @xmath230 through ( [ polab ] ) is preserved in its structure , because for @xmath231 we have + [ ccb ] @xmath232 \leq \int^{\lambda_0}_\lambda \frac{d\lambda'}{\lambda ' } \frac{c}{\lambda^{\prime \omega } } c_b \big(\frac{m{+}1}{2\theta \lambda^{\prime 2}}\big)^b \leq \frac{1}{\omega{+}2b } \frac{c}{\lambda^\omega } c_b \big(\frac{m{+}1}{2\theta \lambda^2}\big)^b\end{aligned}\ ] ] for @xmath233 and @xmath234 \nonumber \\ * & \leq \int^{\lambda_0}_{\sqrt{\frac{m+1}{2\theta } } } \frac{d\lambda'}{\lambda ' } \frac{c}{\lambda^{\prime \omega } } c_b \big(\frac{m{+}1}{2\theta \lambda^{\prime 2}}\big)^b + \int^{\sqrt{\frac{m+1}{2\theta}}}_\lambda \frac{d\lambda'}{\lambda ' } \frac{c}{\lambda^{\prime \omega } } c_a \big(\frac{m{+}1}{2\theta \lambda^{\prime 2}}\big)^a \nonumber \\ * & \leq \frac{1}{\omega{+}2a } \frac{c}{\lambda^\omega } c_a \big(\frac{m{+}1}{2\theta \lambda^2}\big)^a + \frac{(\omega{+}2a ) c_b{-}(\omega{+}2b ) c_a}{(\omega{+}2a)(\omega{+}2b ) } \frac{c}{\big(\frac{m+1}{2\theta}\big)^{\frac{\omega}{2 } } } \label{ccbb } \end{aligned}\ ] ] + for @xmath235 . for @xmath236
we can omit the last term in the second line of ( [ ccbb ] ) , and for @xmath237 we estimate it by @xmath238 times the first term . taking a polynomial in @xmath239 into account , the spirit of ( [ ccb ] ) is unchanged according to @xcite .
+ the general power - counting theorem in @xcite uses analogues of the bounds ( [ est4 - 0 ] ) and ( [ est4 - 1 ] ) of the propagator , which do not add factors @xmath240 . since two legs of the subgraph(s ) are contracted , the total @xmath162-degree of ( [ polab ] ) becomes @xmath241 , which due to ( [ polab - range ] ) can be regarded as degree @xmath242 , too .
+ 1ex 2 .
the proof of ( [ an4 - 1plan ] ) is essentially a repetition of the proof of ( [ an4 - 1 ] ) , with particular care when contracting trajectories on the distinguished boundary component .
the verification of the exponents of @xmath243 , @xmath244 and @xmath239 in ( [ an4 - 1plan ] ) is identical to the proof of ( [ an4 - 1 ] ) .
it remains to verify the @xmath245-degrees of the factor ( [ polab ] ) .
+ we first consider the contraction of two smaller graphs @xmath246 ( left subgraph ) and @xmath247 ( right subgraph ) to the total graph @xmath94 .
we first assume additionally that all indices of the contracting propagator are determined ( this is the case for @xmath248 and @xmath249 ) , e.g.@xmath250 as a subgraph of an 1pi planar graph , at most one side @xmath251 or @xmath252 ( @xmath253 or @xmath254 ) of the contracting propagator @xmath255 ( @xmath256 ) can belong to a trajectory in @xmath202 . in the left graph of ( [ graph - aa1 ] ) let us assume that the side @xmath257 connects two trajectories @xmath258m}\in t_1 $ ] and @xmath259 } \in t_2 $ ] to a new trajectory @xmath258\mathfrak{o}[l ] } \in t$ ] .
the proof for the small-@xmath48 degree @xmath260 in ( [ an4 - 1plan ] ) is immediate , because the contraction reduces the number of external legs by @xmath147 and we are free to estimate the contracting propagator by its global maximisation ( [ est4 - 0 ] ) .
concerning the large-@xmath48 degree @xmath163 , there is nothing to prove if already @xmath261m}\rangle + \langle\overrightarrow{l\mathfrak{o}[l]}\rangle\geq 4 $ ] . for @xmath261m}\rangle + \langle\overrightarrow{l\mathfrak{o}[l]}\rangle
< 4 $ ] we use the refined estimation ( [ est4 - 3 ] ) for the contracting propagator , which gives a relative factor @xmath262 compared with ( [ est4 - 0 ] ) , where @xmath263 .
now , the result follows from ( [ jump ] ) .
because of @xmath261m}\rangle + \langle\overrightarrow{l\mathfrak{o}[l]}\rangle < 4 $ ] , the indices @xmath264 from the propagator can be estimated by @xmath265^r$ ] and @xmath266^r$ ] .
if the resulting jump leads to @xmath267 \mathfrak{o}[l]}\rangle > 2 $ ] , we use ( [ polab - range ] ) to reduce it to @xmath147 . in this way
we can guarantee that the @xmath163-degree does not exceed the @xmath162-degree . alternatively ,
if @xmath268 we can avoid a huge @xmath261\mathfrak{o}[l]}\rangle$ ] by estimating the contracting propagator via ( [ est4 - 4 ] ) instead of in ( [ est4 - 1 ] ) compared with ( [ est4 - 0 ] ) .
it is plausible that this is due to the summation , which we do not need here .
however , we do not prove a corresponding formula without summation . in order to be on a safe side
, one could replace @xmath157 in the final estimation ( [ lima4 ] ) by @xmath269 .
since @xmath157 is finite anyway , there is no change of the final result .
we therefore ignore the discrepancy in @xmath244 . ]
( [ est4 - 3 ] ) , because a single propagator @xmath270 is clearly smaller than the entire sum over @xmath271 .
+ if @xmath258\mathfrak{o}[l ] } \in t'$ ] , then the sum over @xmath266 $ ] with @xmath272\mathfrak{o}[l]}\rangle \geq 5 $ ] can be estimated by the combined sum * over finitely many combinations of @xmath273 with @xmath274m}\rangle,\langle \overrightarrow{ml } \rangle,\langle \overrightarrow{l\mathfrak{o}[l]}\rangle)\leq 4 $ ] , which via ( [ est4 - 3 ] ) and the induction hypotheses relative to @xmath275 contributes a factor @xmath276\mathfrak{o}[l]}\rangle}$ ] to the rhs of ( [ an4 - 1plan ] ) , where @xmath277^r+1}{\theta \lambda^2 } , \frac{\mathfrak{i}[m]^r+1}{\theta \lambda^2},\frac{l^r+1}{\theta \lambda^2 } , \frac{m^r+1}{\theta \lambda^2}\big)$ ] .
we use ( [ polab - range ] ) to reduce the @xmath163-degree from @xmath278\mathfrak{o}[l]}\rangle $ ] to @xmath147 . * over @xmath82 via the induction hypothesis relative to @xmath258 m } \in t_1'$ ] , combined with the usual maximisation ( [ est4 - 0 ] ) of the contracting propagator and an estimation of @xmath247 where @xmath259 } \notin t_2,t_2'$ ] , * over @xmath279 for fixed @xmath280 $ ] , via ( [ est4 - 4 ] ) , taking @xmath258 m } \notin t_1,t_1'$ ] and @xmath259 } \notin t_2,t_2'$ ] . * over @xmath266 $ ] for fixed @xmath82 and @xmath279 , with @xmath265 \approx m \approx l$ ] , via the induction hypothesis relative to @xmath259 } \in t_2'$ ] , the bound ( [ est4 - 0 ] ) of the propagator and @xmath258 m } \notin t_1,t_1'$ ] .
+ a summation over @xmath265 $ ] with given @xmath266 $ ] is analogous .
+ in conclusion , we have proven that the integrand for the graph @xmath94 is bounded by ( [ an4 - 1plan ] ) .
since we are dealing with a @xmath281-point function , the total @xmath48-exponent is negative . using ( [ ccb ] )
we thus obtain the same bound ( [ an4 - 1plan ] ) after integration from @xmath23 down to @xmath48 . if @xmath282 or @xmath283 we get ( [ an4 - 1plan ] ) directly from ( [ est4 - 3 ] ) or ( [ est4 - 4 ] ) .
+ the discussion of the right graph in ( [ graph - aa1 ] ) is similar , showing that the integrand is bounded by ( [ an4 - 1plan ] ) .
as long as the integrand is irrelevant ( i.e. the total @xmath48-exponent is negative ) , we get ( [ an4 - 1plan ] ) after @xmath48-integration , too .
however , @xmath94 might have two legs only with @xmath284 . in this case
the integrand is marginal or relevant , but according to definition [ defint].[defintnp ] we nonetheless integrate from @xmath23 down to @xmath48 .
we have to take into account that the cut - off propagator at the scale @xmath48 vanishes for @xmath285 .
assuming two relevant two - leg subgraphs @xmath286 bounded by @xmath287 times a polynomial in @xmath288 each , we have @xmath289 & \leq \frac{c_0}{\omega } \int_\lambda^{\sqrt{\|m_1m;k_1k\|_\infty/\theta } } \frac{d \lambda'}{\lambda ' } ( \theta \lambda^{\prime 2 } ) p^{2v-2}\big[\ln\frac{\lambda'}{\lambda_r}\big ] \nonumber \\ * & \leq \frac{c_0}{\omega } \|m_1m;k_1k\|_\infty p^{2v-2}\bigg[\ln \big(\frac{\|m_1m;k_1k\|_\infty}{\theta\lambda_r^2 } \big)^{\frac{1}{2}}\bigg ] \nonumber \\ * & \leq \frac{c_0}{\omega } ( \theta \lambda^2 ) p^2_1\big[\frac{m_1m;k_1k}{2\theta\lambda^2}\big ] p^{2v-2}\big[\frac{\lambda}{\lambda_r}\big]\;. \label{1pr}\end{aligned}\ ] ] here , i have inserted the estimation ( [ est4 - 0 ] ) for the propagator , restricted to its support . in the logarithm we expanded @xmath290 and
estimated @xmath291 .
thus , the small-@xmath48 degree @xmath162 of the total graph is increased by 2 over the sum of the small-@xmath48 degrees of the subgraphs ( taken @xmath292 here ) , in agreement with ( [ an4 - 1plan ] ) .
the estimation for the logarithm is not necessary for the large-@xmath48 degree @xmath163 in ( [ polab ] ) . using ( [ polab - range ] )
we could reduce that degree to @xmath293 .
we would like to underline that the integration of 1pr graphs is one of the sources for the factor ( [ polab ] ) in the power - counting theorem .
taking the factors ( [ polab ] ) in the bounds for the subgraphs @xmath294 into account , the formula modifies accordingly .
we confirm ( [ an4 - 1plan ] ) in any case .
+ it is clear that all other possibilities with determined propagator indices as discussed in @xcite are treated similarly .
next , let one index of the contracting propagator be an undetermined summation index , e.g.@xmath295 + let @xmath296\mathfrak{o}[n ] } \in t$ ] . then @xmath297 is determined by the external indices of @xmath247 .
there is nothing to prove for @xmath298k}\rangle \geq 4 $ ] . for @xmath298k}\rangle
< 4 $ ] we partitionate the sum over @xmath80 into @xmath299 , where each term yields the integrand ( [ an4 - 1plan ] ) as before in the case of determined indices ( [ graph - aa1 ] ) , and the sum over @xmath300 , which yields the desired factor in ( [ an4 - 1plan ] ) via ( [ est4 - 4 ] ) and the similarity @xmath301 $ ] of indices . as a subgraph of a planar graph , @xmath302 $ ] in @xmath246 , so that a possible @xmath303-summation can be transferred to @xmath82 . if @xmath296\mathfrak{o}[n ] } \in t'$ ] then in the same way as for ( [ graph - aa1 ] ) the summation splits into the four possibilities related to the pieces @xmath304}$ ] , @xmath305 and @xmath296k}$ ] , which yield the integrand ( [ an4 - 1plan ] ) via the induction hypotheses for the subgraphs and via ( [ est4 - 3 ] ) or ( [ est4 - 4 ] ) . the @xmath48-integration yields ( [ an4 - 1plan ] ) via ( [ ccb ] ) if the integrand is irrelevant , whereas we have to perform similar considerations as in ( [ 1pr ] ) if the integrand is relevant or marginal .
the discussion of graphs with two summation indices on the contracting propagator , such as in @xmath306 is similar .
note that the planarity requirement implies @xmath307 $ ] and @xmath308 $ ] .
4 . next , we look at self - contractions of the same vertex of a graph . among the examples discussed in @xcite
there are only two possibilities which can appear in subgraphs of planar graphs : @xmath309 there is nothing to prove for the left graph in ( [ graph - al2 ] ) . to verify the large-@xmath48 degree @xmath163 relative to the right graph , we partitionate the sum over @xmath80 into * @xmath310n}\rangle \leq 4 $ ] and @xmath311}\rangle \leq 4 $ ] , where each term yields ( [ an4 - 1plan ] ) via the induction hypothesis for the trajectories @xmath312n } \in t_1 $ ] and @xmath304}\in t_1 $ ] of the subgraph ( in the same way as for the examples with determined propagator indices ) , * @xmath310n}\rangle \leq 4 $ ] and @xmath313}\rangle \geq 5 $ ] , for which the induction hypothesis for @xmath312n}\notin t_1,t_1'$ ] and @xmath304}\in t_1'$ ] , together with @xmath314\approx n$ ] , gives a contribution of @xmath147 to the @xmath163-degree in ( [ polab ] ) , and * @xmath310n}\rangle \geq 5 $ ] , which via the induction hypothesis for @xmath312n}\in t_1'$ ] and @xmath304}\notin t_1,t_1'$ ] gives a contribution of @xmath147 to the @xmath163-degree in ( [ polab ] ) .
+ the case @xmath312\mathfrak{o}[n ] } \in t'$ ] is similar to discuss . at the end
we always arrive at the integrand ( [ an4 - 1plan ] ) . if it is irrelevant the integration from @xmath23 down to @xmath48 yields ( [ an4 - 1plan ] ) according to ( [ ccb ] ) .
if the integrand is marginal / relevant and @xmath94 is one - particle reducible , then the indices of the propagator contracting 1pi subgraphs are of the same order as the incoming and outgoing indices of the trajectories through the propagator ( otherwise the 1pi subgraphs are irrelevant ) . now a procedure similar to ( [ 1pr ] ) yields ( [ an4 - 1plan ] ) after integration from @xmath23 down to @xmath48 , too . if @xmath94 is 1pi and marginal or relevant , it is actually of the type [ defint4][defint1 ] of definition [ defint ] and will be discussed below .
+ [ contsamevertex ] 5 .
finally , there will be self - contractions of different vertices of a subgraph , such as in @xmath315 the vertices to contract have to be situated on the same ( distinguished ) boundary component , because the contraction of different boundary components increases the genus and for contractions of other boundary components the proof is immediate .
only the large-@xmath48 degree @xmath163 is questionable .
+ let @xmath258\mathfrak{o}[l]}\in t$ ] , with @xmath265\neq \mathfrak{o}[l]$ ] due to planarity . here , @xmath82 is regarded as a summation index . as before we split that sum over @xmath82 into a piece with @xmath316m}\rangle \leq 4 $ ] , which yields the @xmath163-degree of the integrand ( [ an4 - 1plan ] ) term by term via the induction hypothesis relative to @xmath258 m } , \overrightarrow{l\mathfrak{o}[l]}\in t_1 $ ] and ( [ est4 - 3 ] ) for the contracting propagator , and a piece with @xmath272m}\rangle \geq 5 $ ] , which gives ( [ an4 - 1plan ] ) via the induction hypothesis relative to @xmath258m}\in t_1'$ ] and @xmath259}\notin t_1,t_1'$ ] .
if @xmath258\mathfrak{o}[l]}\in t'$ ] the sum over @xmath266 $ ] with @xmath258\mathfrak{o}[l ] } \geq 5 $ ] is estimated by a finite number of combinations of @xmath273 with @xmath317m}\rangle , \langle \overrightarrow{ml}\rangle , \langle \overrightarrow{l\mathfrak{o}[l]}\rangle)\leq 4 $ ] , which yields the integrand ( [ an4 - 1plan ] ) via the induction hypothesis for @xmath318 and ( [ est4 - 3 ] ) , and the sum over index combinations * @xmath316m}\rangle \leq 4 $ ] , @xmath319 , @xmath320}\rangle \geq 5 $ ] * @xmath316m}\rangle \leq 4 $ ] , @xmath321 * @xmath316m}\rangle \geq 5 $ ] + which is controlled by the induction hypothesis relative to @xmath322 or ( [ est4 - 4 ] ) , together with the similarity of trajectory indices at those parts where the jumps is bounded by @xmath323 .
the case where @xmath296m_1 } \in t$ ] or @xmath296m_1 } \in t'$ ] is easier to treat .
we thus arrive in any case at the estimation ( [ an4 - 1plan ] ) for the integrand of @xmath94 , which leads to the same estimation ( [ an4 - 1plan ] ) for @xmath94 itself according to the considerations at the end of [ contsamevertex ] .
if @xmath94 is of type [ defint4][defint1 ] of definition [ defint ] we will treat it below .
+ the discussion of all other possible self - contractions as listed in @xcite is similar .
+ this finishes the part [ prp - sub - planar ] of the proof of proposition [ power - counting - prop ] .
+ 1ex 3 .
now we consider 1pi planar 4-leg graphs @xmath94 with constant index on each trajectory .
if the external indices are zero , we get ( [ hata4 - 0 ] ) directly from the general power - counting theorem @xcite , because the integration direction used there agrees with definition [ defint].[defint4 ] .
+ for non - zero external indices we decompose the difference ( [ hata4 ] ) according to ( [ deca4 ] ) into graphs with composite propagators ( [ comp - prop-0 ] ) bounded by ( [ qpl1 ] ) .
the composite propagators appear on one of the trajectories of @xmath94 , and as such already on the trajectory of a sequence of subgraphs of @xmath94 , starting with some minimal subgraph @xmath324 .
the composite propagator is the contracting propagator for @xmath324 .
now , the integrand of the minimal subgraph @xmath324 with composite propagator is bounded by a factor @xmath325 times the integrand of the would - be graph @xmath324 with ordinary propagator , where @xmath82 is the index at the trajectory under consideration .
+ if @xmath324 is irrelevant , the factor @xmath326 of the integrand survives according to ( [ ccb ] ) to the subgraph @xmath324 itself .
otherwise , if @xmath324 is relevant or marginal , it is decomposed according to [ defint4][defint1 ] of definition [ defint ] . here
, the last lines of ( [ inta4])([inta3 ] ) are independent of the external index @xmath82 so that in the difference relative to the composite propagator these last lines of ( [ inta4])([inta3 ] ) cancel identically .
there remains the first part of ( [ inta4])([inta3 ] ) , which is integrated from @xmath23 downward and which is irrelevant by induction .
thus , ( [ ccb ] ) applies in this case , too , saving the factor @xmath326 to @xmath324 in any case .
this factor thus appears in the integrand of the subgraph of @xmath94 next larger than @xmath324 . by iteration of the procedure
we obtain the additional factor @xmath327 in the integrand of the total graph @xmath94 with composite propagators , the @xmath48-degree of which being thus reduced by @xmath147 compared with the original graph @xmath94 .
since @xmath94 itself is a marginal graph according to the general power - counting behaviour ( [ an4 - 1 ] ) , the graph with composite propagator is irrelevant and according to definition [ defint].[defint4 ] to be integrated from @xmath23 down to @xmath48 . this explains ( [ hata4 ] ) .
+ 1ex plus 0.4ex 4 .
similarly , we conclude from the proof of ( [ an4 - 1plan ] ) that the integrands of graphs @xmath94 according to definition [ defint].[defint1 ] are marginal . in particular , we immediately confirm ( [ hata1 - 0 ] ) . for non - zero external indices
we decompose the difference ( [ hata1 ] ) according to ( [ deca2-all ] ) into graphs either with composite propagators ( [ comp - prop-0 ] ) bounded by ( [ qpl1 ] ) or with composite propagators ( [ comp - prop - p12])/([comp - prop - m12 ] ) bounded by ( [ qpl3 ] ) . in such a graph
there are apart from usual propagators with bound ( [ est4 - 0])/([est4 - 1])two propagators with @xmath328 in ( [ est4 - 3 ] ) and a composite propagator with bound ( [ qpl1 ] ) , or one propagator with @xmath328 in ( [ est4 - 3 ] ) and one composite propagator with bound ( [ qpl3 ] ) . in both cases we get a total factor @xmath329 compared with a general planar two - point graph ( [ an4 - 1 ] ) .
the detailed discussion of the subgraphs is similar as under [ prp - planar - a4 ] .
+ 1ex plus 0.4ex 5 .
finally , we have to discuss graphs @xmath94 according to definition [ defint].[defint0 ] .
we first consider the case that @xmath94 has constant index on each trajectory .
it is then clear from the proof of ( [ an4 - 1 ] ) that ( in particular ) at vanishing indices the graph @xmath94 is relevant , which is expressed by ( [ hata0 - 0 ] ) .
next , the difference ( [ hata0 - 1 ] ) of graphs can as in ( [ deca4 ] ) be written as a sum of graphs with one composite propagator ( [ comp - prop-0 ] ) , the bound of which is given by ( [ qpl1 ] ) .
after the treatment of subgraphs as described under [ prp - planar - a4 ] , the integrand of each term in the linear combination is marginal . according to definition [ defint].[defint0 ] we have to integrate these terms from @xmath16 up to @xmath48 which agrees with the procedure in @xcite and leads to ( [ hata0 - 1 ] ) . finally , according to ( [ deca1-all ] ) and ( [ deca1-b - all ] ) , the linear combination constituting the lhs of ( [ hata0 ] ) results in a linear combination of graphs with either one propagator ( [ comp - prop-1 ] ) with bound ( [ qpl2 ] ) , or with two propagators ( [ comp - prop-0 ] ) with bound ( [ qpl1 ] ) .
a similar discussion as under [ prp - planar - a4 ] then leads to ( [ hata0 ] ) .
+ the second case is when one index component jumps once on a trajectory and back .
according to the proof of ( [ an4 - 1plan ] ) the integrand of @xmath94 at vanishing external indices is marginal .
we regard it nevertheless as relevant using the inequality @xmath330 , where @xmath331 is some number kept constant in our renormalisation procedure .
we now obtain ( [ hata0 - 0 ] ) .
similarly , the integrand relative to the difference ( [ hata0 - 1 ] ) would be irrelevant , but is considered as marginal via the same trick . finally ,
the linear combination constituting the lhs of ( [ hata0 ] ) is according to ( [ deca1-all ] ) and ( [ deca1-c - all])([deca1-c - d ] ) a linear combination of graphs having either two propagators with @xmath328 in ( [ est4 - 3 ] ) and a composite propagator with bound ( [ qpl1 ] ) , or one propagator with @xmath328 in ( [ est4 - 3 ] ) and one composite propagator ( [ comp - prop - p12])/([comp - prop - m12 ] ) with bound ( [ qpl3 ] ) .
the discussion as before would lead to an increased large-@xmath48 degrees @xmath332 instead of @xmath333 in ( [ hata0 ] ) , which can be reduced to @xmath333 according to ( [ polab - range ] ) .
this finishes the proof of proposition [ power - counting - prop ] .
@xmath334 it is now important to realise @xcite that the estimations ( [ prp - a4])([an4 - 1 ] ) of proposition [ power - counting - prop ] do not make any reference to the initial scale @xmath23 .
therefore , the estimations ( [ prp - a4])([an4 - 1 ] ) , which give finite bounds for the interaction coefficients with finite external indices , also hold in the limit @xmath25 .
this is the renormalisation of the duality - covariant noncommutative @xmath0-model . in numerical computations
the limit @xmath25 is difficult to realise .
taking instead a large but finite @xmath23 , it is then important to estimate the error and the rate of convergence as @xmath23 approaches @xmath24 .
this type of estimations is the subject of the next section .
we finish this section with a remark on the freedom of normalisation conditions .
one of the most important steps in the proof is the integration procedure for the polchinski equation given in definition [ defint ] . for presentational reasons we have chosen the smallest possible
set of graphs to be integrated from @xmath16 upward .
this can easily be generalised .
we could admit in ( [ inta4 ] ) any planar 1pi four - point graphs for which the incoming index of each trajectory is equal to the outgoing index on that trajectory , but with arbitrary jump along the trajectory .
there is no change of the estimation ( [ hata4 ] ) , because ( according to proposition [ power - counting - prop].[prp - sub - planar ] ) @xmath335}$ ] is already irrelevant for these graphs , so is the difference in braces in ( [ inta4 ] ) .
moreover , integrating such an irrelevant graph according to the last line of ( [ inta4 ] ) from @xmath16 upward we obtain a bound @xmath336 $ ] , which agrees with ( [ hata4 - 0 ] ) , because @xmath337 is finite .
similarly , we can relax the conditions on the jump along the trajectory in ( [ inta1])([inta3 ] ) .
we would then define the @xmath149$]-functions in ( [ rhoa ] ) for that enlarged set of graphs @xmath94 . in a second generalisation
we could admit one - particle reducible graphs in [ defint4][defint1 ] of definition [ defint ] and even non - planar graphs with the same condition on the external indices as in [ defint4][defint1 ] of definition [ defint ] . since there is no difference in the power - counting behaviour between non - planar graphs and planar graphs with large jump , the discussion is as before .
however , the convergence theorem developed in the next section can not be adapted in an easy way to normalisation conditions involving non - planar graphs . in summary ,
the proposed generalisations constitute different normalisation conditions of the same duality - covariant @xmath0-model . passing from one normalisation to another one
is a finite re - normalisation .
the invariant characterisation of our model is its definition via four independent normalisation conditions for the @xmath61-functions so that at large scales the effective action approaches ( [ ct4 ] ) .
in this section we prove the _ convergence _ of the coefficients of the effective action in the limit @xmath25 , relative to the integration procedure given in definition [ defint ] .
this is a stronger result than the power - counting estimation of proposition [ power - counting - prop ] , which e.g. would be compatible with bounded oscillations .
additionally , we identify the rate of convergence of the interaction coefficients .
we have to control the @xmath23-dependence which enters the effective action via the integration procedure of definition [ defint ] .
there is an explicit dependence via the integration domain of irrelevant graphs and an implicit dependence through the normalisation ( [ initrho ] ) , which requires a carefully adapted @xmath23-dependence of @xmath338 . for fixed @xmath66 but variable @xmath23
we consider the identity @xmath339
-l[\phi,\lambda_r,\lambda_0'',\rho^0[\lambda_0 '' ] ] \equiv \int_{\lambda_0''}^{\lambda_0 ' } \frac{d\lambda_0}{\lambda_0 } \ , \big(\lambda_0 \frac{d}{d\lambda_0 } l[\phi,\lambda_r,\lambda_0,\rho^0[\lambda_0]]\big ) \nonumber \\ * & \qquad\qquad = \int_{\lambda_0''}^{\lambda_0 ' } \frac{d\lambda_0}{\lambda_0 } \ ; \big(\lambda_0 \frac{\partial l[\phi,\lambda_r,\lambda_0,\rho^0]}{\partial\lambda_0 } + \sum_{a=1}^4 \lambda_0 \frac{d \rho^0_a}{d \lambda_0 } \frac{\partial l[\phi,\lambda_r,\lambda_0,\rho^0]}{\partial\rho^0_a } \big)\;. \label{taut}\end{aligned}\ ] ] the model is defined by fixing the boundary condition for @xmath340 at @xmath16 , i.e. by keeping @xmath341 = \text{constant}$ ] : @xmath342 & = \frac{\partial \rho_b[\lambda_r,\lambda_0,\rho^0 ] } { \partial \lambda_0 } d\lambda_0 + \sum_{a=1}^4 \frac{\partial \rho_b[\lambda_r,\lambda_0,\rho^0 ] } { \partial \rho^0_a } \frac{d\rho^0_a}{d\lambda_0 } d\lambda_0\;.\end{aligned}\ ] ] assuming that we can invert the matrix @xmath343 } { \partial \rho^0_a}$ ] , which is possible in perturbation theory , we get @xmath344 } \frac{\partial \rho_b[\lambda_r,\lambda_0,\rho^0]}{\partial \lambda_0 } \;.
\label{rholam}\end{aligned}\ ] ] inserting ( [ rholam ] ) into ( [ taut ] ) we obtain @xmath339 - l[\phi,\lambda_r,\lambda_0'',\rho^0[\lambda_0 '' ] ] = \int_{\lambda_0''}^{\lambda_0 ' } \frac{d\lambda_0}{\lambda_0 } \ ; r[\phi,\lambda_r,\lambda_0,\rho^0[\lambda_0]]\ ; , \label{rlim}\end{aligned}\ ] ] with @xmath345 & : = \lambda_0 \frac{\partial l[\phi,\lambda,\lambda_0,\rho^0 ] } { \partial \lambda_0 } \nonumber \\ * & - \sum_{a , b=1}^4 \frac{\partial l[\phi,\lambda,\lambda_0,\rho^0 ] } { \partial \rho_a^0 } \frac{\partial \rho_a^0}{\partial \rho_b[\lambda,\lambda_0,\rho^0 ] } \lambda_0 \frac{\partial \rho_b[\lambda,\lambda_0,\rho^0 ] } { \partial \lambda_0}\;. \label{rdef}\end{aligned}\ ] ] following @xcite we differentiate ( [ rdef ] ) with respect to @xmath48 : @xmath346 we have omitted the dependencies for simplicity and made use of the fact that the derivatives with respect to @xmath347 commute . using ( [ poll ] ) , with @xmath348 , we compute the terms on the rhs of ( [ vv ] ) : @xmath349}{\partial \lambda } \big ) \nonumber \\ * & = \sum_{m , n , k , l } \frac{1}{2 } \lambda \frac{\partial \delta^k_{nm;lk}(\lambda)}{\partial \lambda } \bigg ( 2 \frac{\partial l[\phi,\lambda,\lambda_0,\rho^0 ] } { \partial \phi_{mn } } \frac{\partial}{\partial \phi_{kl } } \big ( \lambda_0 \frac{\partial}{\partial \lambda_0 } l[\phi,\lambda,\lambda_0,\rho^0]\big ) \nonumber \\ * & \hspace*{6em } - \frac{1}{(2\pi\theta)^2 } \big[\frac{\partial^2 } { \partial \phi_{mn}\,\partial \phi_{kl}}\big ( \lambda_0 \frac{\partial}{\partial \lambda_0 } l[\phi,\lambda,\lambda_0,\rho^0]\big)\big]_\phi
\bigg ) \nonumber \\ * & \equiv m\big[l,\lambda_0 \frac{\partial l}{\partial \lambda_0 } \big ] \;.
\label{v2}\end{aligned}\ ] ] similarly , we have @xmath350}{\partial \lambda } \big ) & = m\big[l,\frac{\partial l}{\partial \rho_a^0}\big]\;. \label{v2a}\end{aligned}\ ] ] for ( [ vv ] ) we also need the function @xmath351 , which is obtained from ( [ v2 ] ) by first expanding @xmath352 on the lhs according to ( [ lg4 ] ) and by further choosing the indices at the @xmath151-coefficients according to ( [ rhoa ] ) . applying these operations to the rhs of ( [ v2 ] ) , we obtain for @xmath353 or @xmath354 the expansions @xmath355 & = \sum_{n=2}^\infty \sum_{m_i , n_i\in \mathbb{n}^2 } \frac{1}{n ! } m_{m_1n_1;\dots;m_nn_n}[l , u ] \phi_{m_1n_1}\cdots \phi_{m_nn_n } \label{mu}\end{aligned}\ ] ] and the projections [ mua ] @xmath356&:= \sum_{\text{$\gamma$ as in def.~\ref{defint}.\ref{defint0 } } } m^\gamma_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[l , u]\ ; , \\ * m_2[l , u ] & : = \sum_{\text{$\gamma$ as in def.~\ref{defint}.\ref{defint0 } } } \big(m^\gamma_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0}}[l , u ] - m^\gamma_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[l , u]\big)\ ; , \\
m_3[m , u ] & : = \sum_{\text{$\gamma$ as in def.~\ref{defint}.\ref{defint1 } } } \big(- m^\gamma_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[l , u]\big)\ ; , \\ * m_4[l , u ] & : = \sum_{\text{$\gamma$ as in def.~\ref{defint}.\ref{defint4 } } } m^\gamma_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[l , u]\;.\end{aligned}\ ] ] since the graphs @xmath94 in ( [ mua ] ) are one - particle irreducible , only the third line of ( [ v2 ] ) can contribute , also the second line of ( [ v2 ] ) must be taken into account .
] to @xmath357 . using ( [ v2 ] ) , ( [ v2a ] ) and ( [ mua ] ) as well as the linearity of @xmath358 $ ] in the second argument we can rewrite ( [ vv ] ) as @xmath359 - \sum_{a=1}^4 \frac{\partial l}{\partial \rho_a } m_a[l , r ] \ ; , \label{vvv}\end{aligned}\ ] ] where the function @xmath360 & : = \sum_{b=1}^4 \frac{\partial l[\lambda,\lambda_0,\rho^0 ] } { \partial \rho_b^0 } \frac{\partial \rho_b^0}{\partial \rho_a[\lambda,\lambda_0,\rho^0 ] } \label{lr}\end{aligned}\ ] ] scales according to @xmath361 - \sum_{b=1}^4 \frac{\partial l}{\partial \rho_b } m_b\big[l , \frac{\partial l}{\partial \rho_a}\big]\ ; , \label{llr}\end{aligned}\ ] ] as a similar calculation shows .
next , we also expand ( [ rdef ] ) and ( [ lr ] ) as power series in the coupling constant : @xmath362 \nonumber \\ * & = \sum_{v = 0}^\infty \lambda^v \sum_{n=2}^{2v+4 } \frac{(2\pi\theta)^{\frac{n}{2}-2}}{n ! } \sum_{m_i , n_i } h^{a(v)}_{m_1n_1;\dots;m_nn_n}[\lambda,\lambda_0,\rho^0 ] \phi_{m_1n_1}\cdots \phi_{m_nn_n}\ ; , \\ & r[\phi,\lambda,\lambda_0,\rho^0 ] \nonumber \\ * & = \sum_{v = 1}^\infty \lambda^v \sum_{n=2}^{2v+2 } \frac{(2\pi\theta)^{\frac{n}{2}-2}}{n ! } \sum_{m_i , n_i } r^{(v)}_{m_1n_1;\dots;m_nn_n}[\lambda,\lambda_0,\rho^0 ] \phi_{m_1n_1}\cdots \phi_{m_nn_n}\;.\end{aligned}\ ] ] the differential equations ( [ llr ] ) and ( [ vvv ] ) can now with ( [ mu ] ) and ( [ mua ] ) be written as @xmath363 \nonumber \\ * & = \bigg\{\sum_{n_1=2}^n \sum_{v_1=1}^{v } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) a^{(v_1)}_{m_1n_1;\dots;m_{n_1 - 1}n_{n_1 - 1};mn}[\lambda ] h^{a(v - v_1)}_{m_{n_1}n_{n_1};\dots;m_{n}n_{n};kl}[\lambda ] \nonumber \\ * & \quad + \big(\binom{n}{n_1{-}1 } -1\big ) \text { permutations } \bigg\ } - \sum_{m , n , k , l } \frac{1}{2 } q_{nm;lk}(\lambda ) h^{a(v)}_{m_1n_1;\dots;m_{n}n_{n};mn;kl}[\lambda ] \nonumber \\ & - \sum_{v_1=0}^{v } h^{1(v - v_1)}_{m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \bigg\ { -\frac{1}{2 } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) h^{a(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint0 } } ] } \nonumber \\ & - \sum_{v_1=0}^{v } h^{2(v - v_1)}_{m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \nonumber \\*[-1ex ] & \qquad\qquad \times \bigg\ { { -}\frac{1}{2 } \sum_{m , n , k , l}\!\
! q_{nm;lk}(\lambda ) \big(h^{a(v_1)}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};mn;kl}[\lambda ] -h^{a(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda]\big ) \!\bigg\}_{[\text{def.~\ref{defint}.\ref{defint0 } } ] } \nonumber \\ & + \sum_{v_1=0}^{v } h^{3(v - v_1)}_{m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \bigg\ { -\frac{1}{2 } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) h^{a(v_1)}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint1 } } ] } \nonumber \\ & - \sum_{v_1=1}^{v } h^{4(v - v_1)}_{m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \bigg\ { -\frac{1}{2 } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) h^{a(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint4 } } ] } \ ; , \label{polb4 } \\ & \lambda \frac{\partial}{\partial \lambda } r^{(v)}_{m_1n_1;\dots;m_nn_n } [ \lambda,\lambda_0,\rho^0 ] \nonumber \\ * & = \bigg\{\sum_{n_1=2}^n \sum_{v_1=1}^{v-1 } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) a^{(v_1)}_{m_1n_1;\dots;m_{n_1 - 1}n_{n_1 - 1};mn}[\lambda ] r^{(v - v_1)}_{m_{n_1}n_{n_1};\dots;m_{n}n_{n};kl}[\lambda ] \nonumber \\ * & \quad + \big(\binom{n}{n_1{-}1 } -1\big ) \text { permutations } \bigg\ } - \sum_{m , n , k , l } \frac{1}{2 } q_{nm;lk}(\lambda ) r^{(v)}_{m_1n_1;\dots;m_{n}n_{n};mn;kl}[\lambda ] \nonumber \\ & - \sum_{v_1=1}^{v } h^{1(v - v_1)}_{m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \bigg\ { -\frac{1}{2 } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint0 } } ] } \nonumber \\ & - \sum_{v_1=1}^{v } h^{2(v - v_1)}_{m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \nonumber \\*[-1ex ] & \qquad\qquad \times \bigg\ { { -}\frac{1}{2 } \sum_{m , n , k , l}\ !
q_{nm;lk}(\lambda ) \big(r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};mn;kl}[\lambda ] -r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda]\big ) \bigg\}_{[\text{def.~\ref{defint}.\ref{defint0 } } ] } \nonumber \\ & + \sum_{v_1=1}^{v } h^{3(v - v_1)}_{m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \bigg\ { -\frac{1}{2 } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint1 } } ] } \nonumber \\ & - \sum_{v_1=1}^{v } h^{4(v - v_1)}_{m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \bigg\ { -\frac{1}{2 } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint4 } } ] } \;. \label{polv4}\end{aligned}\ ] ] we have used several times symmetry properties of the expansion coefficients and of the propagator and the fact that to the 1pi projections ( [ mua ] ) only the last line of ( [ mu ] ) can contribute . by @xmath364}$ ] we understand the restriction to @xmath365-graphs and @xmath366-graphs , respectively , which satisfy the index criteria on the trajectories as given in definition [ defint ] . the @xmath365-graphs will be constructed later in section [ sechinit ] .
the @xmath366-graphs are in their structure identical to the previously considered graphs for the @xmath151-functions , but have a different meaning .
see section [ graphr ] .
next , we derive the bounds for the @xmath365-functions . inserting ( [ ct4 ] ) into the definition ( [ lr ] )
we obtain immediately the initial condition at @xmath105 : @xmath367 = \delta_{n2 } \delta^{v0 } \delta_{n_1m_2}\delta_{n_2m_1}\ ; , \label{initrho1 } \\ & h^{2(v)}_{m_1n_1;\dots;m_nn_n}[\lambda_0,\lambda_0,\rho^0 ] = \delta_{n2 } \delta^{v0 } ( m_1 ^ 1{+}n_1 ^ 1{+}m_1 ^ 2{+}n_1 ^ 2 ) \delta_{n_1m_2}\delta_{n_2m_1}\ ; , \label{initrho2 } \\ & h^{3(v)}_{m_1n_1;\dots;m_nn_n}[\lambda_0,\lambda_0,\rho^0 ] \nonumber \\ * & = -\delta_{n2 } \delta^{v0 } \big ( \big(\sqrt{n_1 ^ 1m_1 ^ 1 } \delta_{n_1 ^ 1,m_2 ^ 1 + 1}\delta_{n_2 ^ 1 + 1,m_1 ^ 1 } + \sqrt{n_2 ^ 1m_2 ^ 1 } \delta_{n_1 ^
1,m_2 ^ 1 - 1}\delta_{n_2 ^ 1 - 1,m_1 ^ 1}\big ) \delta_{n_1 ^ 2,m_2 ^ 2}\delta_{n_2 ^ 2,m_1 ^ 2 } \nonumber \\ & + ( \sqrt{n_1 ^ 2m_1 ^ 2 } \delta_{n_1 ^ 2,m_2 ^ 2 + 1}\delta_{n_2 ^ 2 + 1,m_1 ^ 2 } + \sqrt{n_2 ^ 2m_2 ^ 2 }
\delta_{n_1 ^ 2,m_2 ^ 2 - 1}\delta_{n_2 ^ 2 - 1,m_1 ^ 2 } ) \delta_{n_1 ^ 1,m_2 ^ 1}\delta_{n_2 ^ 1,m_1 ^ 1}\big)\ ; , \label{initrho3 } \\ & h^{4(v)}_{m_1n_1;\dots;m_nn_n}[\lambda_0,\lambda_0,\rho^0 ] = \delta_{n4 } \delta^{v0 } \big ( \frac{1}{6 } \delta_{n_1m_2}\delta_{n_2m_3}\delta_{n_3m_4}\delta_{n_4m_1 } + 5\;\text{permutations}\,\big)\ , .
\label{initrho4}\end{aligned}\ ] ] we first compute @xmath368 $ ] for @xmath369 .
since there is no 6-point function at order 0 in @xmath76 , the differential equation ( [ polb4 ] ) reduces to @xmath370}{\partial \lambda } \nonumber \\ * & = \sum_{b=1}^3 \sum_{m , n , k , l , m',n',k',l ' } c_b^{m'n';k'l ' } h^{b(0)}_{m_1n_1;\dots;m_4n_4 } [ \lambda]\big ( q_{nm;lk}(\lambda ) h^{a(0)}_{m'n';k'l';mn;kl}[\lambda ] \big)\;,\end{aligned}\ ] ] for certain coefficients @xmath371 .
the solution is due to ( [ q4 ] ) given by @xmath372 = h^{a(0)}_{m_1n_1;\dots;m_4n_4 } [ \lambda_0 ] \nonumber \\ * & + \sum_{b=1}^3 \sum_{m , n , k , l , m',n',k',l ' } c_b^{m'n';k'l ' } h^{b(0)}_{m_1n_1;\dots;m_4n_4}[\lambda_0 ] \nonumber \\ * & \quad \times \big ( \big(\delta^k_{nm;lk}(\lambda ) - \delta^k_{nm;lk}(\lambda_0 ) \big ) h^{a(0)}_{m'n';k'l';mn;kl}[\lambda_0]\big ) \nonumber \\ * & + \sum_{b=1}^3 \sum_{m , n , k , l , m',n',k',l ' } c_b^{m'n';k'l ' } h^{b(0)}_{m_1n_1;\dots;m_4n_4}[\lambda_0 ] \nonumber \\ * & \quad \times \big ( \big(\delta^k_{nm;lk}(\lambda ) - \delta^k_{nm;lk}(\lambda_0 ) \big ) \nonumber \\ * & \qquad\qquad \times \sum_{b'=1}^3 \sum_{m'',n'',k'',l'',m''',n''',k''',l ' '' } c_{b'}^{m'''n''';k'''l ' '' } h^{b'(0)}_{m'n';k'l';mn;kl}[\lambda_0]\big ) \nonumber \\ * & \quad\quad \times \big ( \big(\delta^k_{n''m'';l''k''}(\lambda ) - \delta^k_{n''m'';l''k''}(\lambda_0 ) \big ) h^{a(0)}_{m'''n''';k'''l''';m''n'';k''l''}[\lambda_0]\big ) \nonumber \\ * & \quad + \dots\;.\end{aligned}\ ] ] with the initial conditions ( [ initrho1])([initrho3 ] ) we get @xmath373 \equiv 0 \qquad \text{for } a \in \{1,2,3\}\;. \label{h0 - 24}\end{aligned}\ ] ] inserting ( [ h0 - 24 ] ) into the rhs of ( [ polb4 ] ) we see that @xmath374 for @xmath375 and @xmath376 are constant , which means that the relations ( [ initrho1])([initrho4 ] ) hold actually at any value @xmath48 and not only at @xmath105 .
we need a graphical notation for the @xmath365-functions .
we represent the base functions ( [ initrho1])([initrho4 ] ) , valid for any @xmath48 , as follows : @xmath377 & = \quad \parbox{24\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = h01,scale=0.9,bb=71 666 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(10,8){\mbox{\scriptsize$1 $ } } \end{picture } } \label{h01 } \\[2ex ] h^{2(0)}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda ] & = \quad \parbox{24\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = h01,scale=0.9,bb=71 666 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(10,8){\mbox{\scriptsize$2 $ } } \end{picture } } \label{h02 } \\[2ex ] h^{3(0)}_{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda ] & = \quad \parbox{24\unitlength}{\begin{picture}(20,10 ) \put(0,3){\epsfig{file = h03,scale=0.9,bb=71 666 130 677 } } \put(2,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}$ } } \put(16,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(1,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}$ } } \put(16,8){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(10,8){\mbox{\scriptsize$3 ^ 1 $ } } \end{picture } } \label{h03 } \\[1ex ] h^{4(0)}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } ; { \genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2};{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda ] & = \frac{1}{6 } \quad \parbox{19\unitlength}{\begin{picture}(17,17 ) \put(0,0){\epsfig{file = h04,scale=0.9,bb=71 638 117 684 } } \put(-1,7.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{m^1}{m^2}$ } } \put(6,0){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{n^1}{n^2}$ } } \put(12,7.5){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{k^1}{k^2}$ } } \put(6,14){\mbox{\scriptsize${\genfrac{}{}{0pt}{}}{l^1}{l^2}$ } } \end{picture } } + \text { 5 permutations } \;.
\label{h04}\end{aligned}\ ] ] the special vertices stand for some sort of hole into which we can insert planar two- or four - point functions at vanishing external indices .
however , the graph remains connected at these holes , in particular , there is index conservation at the hole @xmath378 and a jump by @xmath379 or @xmath380 at the hole . by repeated contraction with @xmath151-graphs and self - contractions
we build out of ( [ h01])([h04 ] ) more complicated graphs with holes .
we use this method to compute @xmath381 $ ] . in this case , we need the planar and non - planar self - contractions of ( [ h04 ] ) : @xmath382 these contractions correspond ( with a factor @xmath383 ) to last term in the third line of ( [ polb4 ] ) , for @xmath384 and @xmath109 .
we also have to subtract ( again up to the factor @xmath383 ) the fourth to last lines of ( [ polb4 ] ) .
for instance , the fourth line amounts to insert the _ planar graphs _ of ( [ h041 ] ) with @xmath385 into ( [ h01 ] ) .
the total contribution to the rhs of ( [ polb4 ] ) corresponding to the first graph in ( [ h041 ] ) , with @xmath386 and @xmath387 , reads @xmath388 the second graph in the first line of ( [ h044 ] ) corresponds to the fourth line of ( [ polb4 ] ) .
the second line of ( [ h044 ] ) represents the fifth / sixth lines of ( [ polb4 ] ) , undoing the symmetry properties of the upper and lower component used in ( [ polb4 ] ) .
the difference of graphs corresponding to the @xmath389 component vanishes , because the value of the graph is independent of @xmath389 .
there is no planar contribution from the last two lines of ( [ polb4 ] ) . in total
, we get the projection ( [ qpl2 ] ) to the irrelevant part of the graph .
the same procedure leads to the irrelevant part of the second graph in ( [ h041 ] ) . with these considerations ,
the differential equation ( [ polb4 ] ) takes for @xmath109 and @xmath384 the form @xmath390 & = -\frac{1}{6 } \delta_{n_1m_2}\delta_{n_2m_1 } \big(\sum_{l\in \mathbb{n}^2 } \mathcal{q}^{(1)}_{n_1l;ln_1 } ( \lambda ) + \sum_{l\in \mathbb{n}^2 } \mathcal{q}^{(1)}_{n_2l;ln_2 } ( \lambda ) \big ) \nonumber \\ * & - \frac{1}{6 } q_{m_1n_1;m_2n_2}(\lambda)\;. \label{h0 - 42}\end{aligned}\ ] ] the first line comes from the planar graphs in ( [ h041 ] ) and the subtraction terms according to ( [ h044 ] ) , whereas the second line of ( [ h0 - 42 ] ) is obtained from the last ( non - planar ) graph in ( [ h041 ] ) . using the initial condition ( [ initrho4 ] ) at @xmath105 , the bounds ( [ est4 - 0 ] ) and ( [ qpl2 ] ) combined with the volume factor @xmath391 for the @xmath279-summation we get @xmath392| \leq \frac{c \big(\|m_1\|_\infty^2+\|n_1\|_\infty^2\big ) } { \omega \theta \lambda^2 } \ , \delta_{n_1m_2}\delta_{n_2m_1 } + \frac{c_0 } { \omega \theta \lambda^2 } \ , \delta_{n_1m_2}\delta_{n_2m_1 } \;.
\label{h0 - 42n}\end{aligned}\ ] ] it is extremely important here that the irrelevant projection @xmath393 and not the propagator @xmath394 itself appears in the first line of ( [ h0 - 42 ] ) .
the example suggests that similar cancellations of relevant and marginal parts appear in general , too .
thus , we expect all @xmath365-functions to be irrelevant .
this is indeed the case : [ prop - h ] let @xmath94 be a ribbon graph with holes having @xmath77 external legs , @xmath76 vertices , @xmath78 external vertices and segmentation index @xmath79 , which is drawn on a genus-@xmath71 riemann surface with @xmath72 boundary components . then , the contribution @xmath395 of @xmath94 to the expansion coefficient of the auxiliary function of a duality - covariant @xmath0-theory on @xmath27 in the matrix base is bounded as follows : 1 . [ h - planar - h4 ] for @xmath94 according to definition [ defint].[defint4 ] we have @xmath396 \big| \nonumber \\ * & \leq \big(\theta \lambda^2\big)^{-\delta^{1a } } p_{1-\delta^{v0}}^{4v-2 + 2\delta^{a4 } } \bigg[\frac{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } ; { \genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2};{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2 } \bigg]\big(\frac{1}{\omega}\big)^{3v-1+\delta^{a4}-v^e } \nonumber \\ * & \qquad \times p^{2v-1+\delta^{a4}}\big[\ln\frac{\lambda}{\lambda_r}\big ] \ ; , \label{h - h4 } \end{aligned}\ ] ] where all vertices on the trajectories contribute to @xmath78 .
[ h - planar - h0 ] for @xmath94 according to definition [ defint].[defint0 ] we have @xmath397 \big| \nonumber \\ * & \leq \big(\theta \lambda^2\big)^{1-\delta^{1a } } p^{4v+2\delta^{a4}}_{2-\delta_{v0}(2\delta^{a1}+\delta^{a2 } ) } \bigg[\frac{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2}\bigg ] \big(\frac{1}{\omega}\big)^{3v+\delta^{a4}-v^e}\ , p^{2v+\delta^{a4}}\big[\ln\frac{\lambda}{\lambda_r}\big ] \ ; , \label{h - h0 } \end{aligned}\ ] ] where all vertices on the trajectories contribute to @xmath78 .
[ h - planar - h1 ] for @xmath94 according to definition [ defint].[defint1 ] we have @xmath398 \big| \nonumber \\ * & \leq \big(\theta \lambda^2\big)^{1-\delta^{1a } } p^{4v+2\delta^{a4}}_{2-\delta_{v0 } } \bigg [ \frac{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2}\bigg ] \big(\frac{1}{\omega}\big)^{3v+\delta^{a4}-v^e } \ , p^{2v+\delta^{a4}}\big[\ln\frac{\lambda}{\lambda_r}\big ] \
; , \label{h - h1}\end{aligned}\ ] ] where all vertices on the trajectories contribute to @xmath78 .
[ h - sub - planar ] if @xmath94 is a subgraph of an 1pi planar graph with a selected set @xmath202 of trajectories on one distinguished boundary component and a second set @xmath205 of summed trajectories on that boundary component , we have @xmath399 \big| \nonumber \\ * & ~~ \leq \big(\theta \lambda^2\big)^{(2-\delta^{1a}-\frac{n}{2})+2(1-b ) } p^{4v+2 + 2\delta^{a4}-n}_{\big(2 t ' + \sum_{\overrightarrow{n_j\mathfrak{o}[n_j ] } \in t } \min(2,\frac{1}{2}\langle \overrightarrow{n_j\mathfrak{o}[n_j ] } \rangle ) \big ) } \bigg [ \frac { m_1n_1;\dots;m_nn_n}{\theta \lambda^2 } \bigg ] \nonumber \\ * & \qquad \times \big(\frac{1}{\omega}\big)^{3v-\frac{n}{2}+\delta^{a4 } + b - v^e-\iota+s+t ' } p^{2v+1+\delta^{a4}-\frac{n}{2 } } \big[\ln \frac{\lambda}{\lambda_r}\big]\;. \label{hn4 - 1plan}\end{aligned}\ ] ] the number of summations is now restricted by @xmath400 .
[ h - non - planar ] if @xmath94 is a non - planar graph or a graph with @xmath19 external legs , we have @xmath401 \big| \nonumber \\ * & \quad \leq \big(\theta \lambda^2\big)^{(2-\delta^{a1}-\frac{n}{2 } ) + 2(1-b-2 g ) } p^{4v+2 + 2\delta^{a4}-n}_0 \bigg [ \frac { m_1n_1;\dots;m_nn_n}{\theta \lambda^2 } \bigg ] \nonumber \\ * & \qquad \times \big(\frac{1}{\omega}\big)^{3v-\frac{n}{2}+\delta^{a4 } + b+2g - v^e-\iota+s } \,p^{2v+1+\delta^{a4 } -\frac{n}{2}}\big[\ln \frac{\lambda}{\lambda_r}\big]\;. \label{hn4 - 1}\end{aligned}\ ] ] the number of summations is now restricted by @xmath402 .
_ the proposition will be proven by induction upward in the number @xmath76 of vertices and for given @xmath76 downward in the number @xmath77 of external legs . 1 . taking ( [ polab - range ] ) into account , the estimations ( [ h - h4])([hn4 - 1plan ] ) are further bound by ( [ hn4 - 1 ] ) . in particular
, the inequality ( [ hn4 - 1 ] ) correctly reproduces the bounds for @xmath403 derived in section [ sechinit ] . by comparison with ( [ an4 - 1 ] )
, the estimation ( [ hn4 - 1 ] ) follows immediately for the @xmath365-_linear _ parts on the rhs of ( [ polb4 ] ) which contribute to the integrand of @xmath404 $ ] .
since planar two- and four - point functions are preliminarily excluded , the @xmath48-integration ( from @xmath23 down to @xmath48 ) confirms ( [ hn4 - 1 ] ) for those contributions which arise from @xmath365-linear terms on the rhs of ( [ polb4 ] ) that are non - planar or have @xmath19 external legs .
+ we now consider in the @xmath365-bilinear part on the rhs of ( [ polb4 ] ) the contributions of non - planar graphs or graphs with @xmath19 external legs .
we start with the fourth line in ( [ polb4 ] ) , with the first term being a non - planar @xmath365-function which ( apart from the number of vertices and the hole label @xmath162 ) has the same topological data as the total @xmath365-graph to estimate . from the induction hypothesis
it is clear that the term in braces @xmath160 is bounded by the planar unsummed version ( @xmath405 ) of ( [ hn4 - 1plan ] ) , with @xmath406 and @xmath407 , and with a reduction of the degree of the polynomial in @xmath239 by @xmath104 : + @xmath408 \big\}_{\text{def.~\ref{defint}.\ref{defint0 } } } \bigg| \nonumber \\ * & \qquad \leq \big(\theta \lambda^2\big)^{(1-\delta^{a1 } ) } \big(\frac{1}{\omega}\big)^{3v_1+\delta^{a4}-v^e_1 } \,p^{2v_1 - 1+\delta^{a4}}\big[\ln \frac{\lambda}{\lambda_r}\big]\ ; , \label{hn4 - 2a } \\ & \sum_{\mathcal{e}^s } \big|h^{1(v - v_1,v^e , b , g,\iota ) } _ { m_1n_1;\dots;m_{n}n_{n}}[\lambda]\big| \nonumber \\ * & \quad \leq \big(\theta \lambda^2\big)^{(1-\frac{n}{2 } ) + 2(1-b-2 g ) } p^{4(v - v_1)+2-n}_0 \big[\frac{m_1n_1;\dots;m_{n}n_{n}}{\theta \lambda^2 } \big ] \nonumber \\ * & \qquad \times \big(\frac{1}{\omega}\big)^{3(v - v_1)-\frac{n}{2 } + b+2g - v^e-\iota+s } \,p^{2(v - v_1)+1-\frac{n}{2}}\big[\ln \frac{\lambda}{\lambda_r}\big]\;. \label{hn4 - 2b}\end{aligned}\ ] ] + we can ignore the term @xmath409 $ ] , see ( [ polab ] ) , in ( [ hn4 - 2a ] ) because the external indices of that part are zero .
in the first step we exclude @xmath384 so that the sum over @xmath410 in ( [ polb4 ] ) starts due to ( [ initrho1])([initrho3 ] ) at @xmath411 . for @xmath412
there is a contribution to ( [ hn4 - 2b ] ) with @xmath109 only , where ( [ hn4 - 2a ] ) can be regarded as known by induction .
since the factor @xmath413 can safely be absorbed in the polynomial @xmath414 $ ] , the product of ( [ hn4 - 2a ] ) and ( [ hn4 - 2b ] ) confirms the bound ( [ hn4 - 1 ] ) for the integrand under consideration , preliminarily for @xmath415 . in the next step
we repeat the argumentation for @xmath384 , where ( [ hn4 - 2b ] ) , with @xmath416 , is known from the first step .
+ second , we consider the fifth / sixth lines in ( [ polb4 ] ) .
the difference of functions in braces @xmath160 involves graphs with constant index along the trajectories .
we have seen in section [ seccomposite ] that such a difference can be written as a sum of graphs each having a composite propagator ( [ qpl1 ] ) at a trajectory .
as such the @xmath417-degree of the part in braces @xmath160 is reduced$ ] introduced in ( [ polab ] ) , with @xmath418 in presence of a composite propagator ( [ qpl1 ] ) . the argument in the brackets of @xmath409 $ ] is the ratio of the maximal external index to the reference scale @xmath419 .
since the maximal index along the trajectory is @xmath104 , we can globally estimate in this case @xmath420 $ ] by a constant times @xmath421 .
] by @xmath104 compared with planar analogues of ( [ hn4 - 1 ] ) for @xmath109 .
the difference of functions in braces @xmath160 involves also graphs where the index along one of the trajectories jumps once by @xmath379 or @xmath380 and back . for these graphs
we conclude from ( [ est4 - 3 ] ) ( and the fact that the maximal index along the trajectory is @xmath147 ) that the @xmath417-degree of the part in braces @xmath160 is also reduced by @xmath104 : + @xmath422 -h^{a(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda]\big ) \big\}_{[\text{def.~\ref{defint}.\ref{defint0 } } ] } \bigg| \nonumber \\ * & \qquad \leq \big(\theta \lambda^2\big)^{(-\delta^{a1 } ) } \big(\frac{1}{\omega}\big)^{3v_1+\delta^{a4 } ( 1-v^e_1 ) } \,p^{2v_1 - 1+\delta^{a4}}\big[\ln \frac{\lambda}{\lambda_r}\big]\ ; , \label{hn4 - 3a } \\ & \sum_{\mathcal{e}^s } \big|h^{2(v - v_1,v^e , b , g,\iota)}_{m_1n_1;\dots;m_{n}n_{n } } [ \lambda ] \big| \nonumber \\ * & \quad \leq \big(\theta \lambda^2\big)^{(2-\frac{n}{2 } ) + 2(1-b-2 g ) } \ , p^{4(v - v_1)+2-n}_0 \big[\frac{m_1n_1;\dots;m_{n}n_{n}}{\theta \lambda^2 } \big ] \nonumber \\ * & \qquad \times \big(\frac{1}{\omega}\big)^{3(v - v_1)-\frac{n}{2}+1 + b+2g - v^e-\iota+s } \,p^{2(v - v_1)+1-\frac{n}{2}}\big[\ln \frac{\lambda}{\lambda_r}\big]\;. \label{hn4 - 3b}\end{aligned}\ ] ] + again we have to exclude @xmath384 in the first step , which then confirms the bound ( [ hn4 - 1 ] ) for the integrand under consideration . in the second step
we repeat the argumentation for @xmath384 . +
third , the discussion of the seventh line of ( [ polb4 ] ) is completely similar , because there the index on each trajectory jumps once by @xmath379 or @xmath380 .
this leads again to a reduction by @xmath104 of the @xmath417-degree of the part in braces @xmath160 compared with planar analogues of ( [ hn4 - 1 ] ) for @xmath109 .
+ finally , the part in braces in the last line of ( [ polb4 ] ) can be estimated by a planar @xmath99 version of ( [ hn4 - 1 ] ) , again with a reduction by @xmath104 of the degree of @xmath414 $ ] : + @xmath423 \big\}_{[\text{def.~\ref{defint}.\ref{defint4 } } ] } \bigg| \nonumber \\ * & \leq \big(\theta \lambda^2\big)^{-\delta^{a1 } } \big(\frac{1}{\omega}\big)^{3v_1 - 1+\delta^{a4}-v^e_1 } \
, p^{2v_1 - 2+\delta^{a4}}\big[\ln \frac{\lambda}{\lambda_r}\big]\ ; , \label{hn4 - 4a } \\ & \sum_{\mathcal{e}^s } \big|h^{4(v - v_1,b , g , v^e,\iota ) } _ { m_1n_1;\dots;m_{n}n_{n}}[\lambda ] \big|\nonumber \\ * & \leq \big(\theta \lambda^2\big)^{(2-\frac{n}{2 } ) + 2(1-b-2 g ) } \ , p^{4(v - v_1)+4-n}_0 \big[\frac{m_1n_1;\dots;m_{n}n_{n}}{\theta \lambda^2 } \big ] \nonumber \\ * & \qquad \times \big(\frac{1}{\omega}\big)^{3(v - v_1)-\frac{n}{2}+1 + b+2g - v^e-\iota+s } \,p^{2(v - v_1)+2 -\frac{n}{2}}\big[\ln \frac{\lambda}{\lambda_r}\big]\;. \label{hn4 - 4b}\end{aligned}\ ] ] + we confirm again the bound ( [ hn4 - 1 ] ) for the integrand under consideration .
+ since we have assumed that the total @xmath365-graph is non - planar or has @xmath19 external legs , the integrand ( [ hn4 - 1 ] ) is irrelevant so that we obtain after integration from @xmath23 down to @xmath48 ( and use of the initial conditions ( [ initrho1])([initrho4 ] ) ) the same bound ( [ hn4 - 1 ] ) for the graph , too .
+ 1ex 2 . according to section [ sechinit ] ,
the inequality ( [ hn4 - 1plan ] ) is correct for @xmath403 . by comparison with ( [ an4 - 1plan ] ) ,
the estimation ( [ hn4 - 1plan ] ) follows immediately for the @xmath365-_linear _ parts on the rhs of ( [ polb4 ] ) which contribute to the integrand of @xmath404 $ ] . excluding planar two- and four - point functions with constant index on the trajectory or with limited jump according to [ defint4][defint1 ] of definition [ defint ]
, the @xmath48-integration confirms ( [ hn4 - 1plan ] ) for those contributions which arise from @xmath365-linear terms on the rhs of ( [ polb4 ] ) that correspond to subgraphs of planar graphs ( subject to the above restrictions ) .
the proof of ( [ hn4 - 1plan ] ) for the @xmath365-bilinear terms in ( [ polb4 ] ) is completely analogous to the non - planar case .
we only have to replace ( [ hn4 - 2b ] ) , ( [ hn4 - 3b ] ) and ( [ hn4 - 4b ] ) by the adapted version of ( [ hn4 - 1plan ] )
. in particular , the distinguished trajectory with its subsets @xmath424 of indices comes exclusively from the ( [ hn4 - 1plan])-analogues of ( [ hn4 - 2b ] ) , ( [ hn4 - 3b ] ) and ( [ hn4 - 4b ] ) and not from the terms in braces in ( [ polb4 ] ) .
+ 1ex 3 .
we first consider @xmath415 .
then , according to ( [ initrho1])([initrho3 ] ) we need @xmath425 in order to have a non - vanishing contribution to ( [ h - h4 ] ) .
since according to definition [ defint].[defint4 ] the index along each trajectory of the ( planar ) graph @xmath94 is constant , we have @xmath426 = \frac{1}{6 } h^{a(v , v^e,1,0,0)}_{m_1m_2;m_2m_3;m_3m_4;m_4n_1}[\lambda ] + \text{5 permutations}\;.\end{aligned}\ ] ] then , using ( [ initrho1])([initrho4 ] ) and the fact that @xmath94 is 1pi , the differential equation ( [ polb4 ] ) reduces to @xmath427 \bigg)_{a\neq 4}\nonumber \\ * & = \bigg({-}\frac{1}{12 } \bigg\ { \sum_{m , n , k , l}\ !
q_{nm;lk}(\lambda ) \big ( h^{a(v , v^e , b,0,\iota)}_{m_1m_2;m_2m_3;m_3m_4;m_4n_1;mn;kl}[\lambda ] \nonumber \\ * & \hspace*{10em } - h^{a(v , v^e , b,0,\iota)}_{00;00;00;00;mn;kl}[\lambda ] \big ) \bigg\}_{\textup{[def.~\ref{defint}.\ref{defint4 } ] } } + \text{5 permutations } \bigg ) \nonumber \\ * & + \text{the $ 4^{\text{th}}$ to last lines of ( \ref{polb4 } ) with $ \displaystyle \sum_{v_1=0}^{v } \mapsto \sum_{v_1=1}^{v-1}$}\;. \label{hh4-a}\end{aligned}\ ] ] here , the term @xmath428 $ ] in the second line of ( [ hh4-a ] ) comes from the @xmath429-contribution of the last line in ( [ polb4 ] ) , together with ( [ initrho4 ] ) . in the same way as in section [ seccomposite ]
we conclude that the second line of ( [ hh4-a ] ) can be written as a linear combination of graphs having a composite propagator ( [ comp - prop-0 ] ) on one of the trajectories . as
such we have to replace the bound ( [ est4 - 1 ] ) relative to the contribution of an ordinary propagator by ( [ qpl1 ] ) . for the total graph
this amounts to multiply the corresponding estimation ( [ hn4 - 1plan ] ) of ordinary @xmath365-graphs with @xmath99 by a factor @xmath430 , which yields the subscript @xmath104 of the part @xmath431 $ ] of the integrand ( [ h - h4 ] ) , for the time being restricted to the second line of ( [ hh4-a ] ) .
since the resulting integrand is irrelevant , we also obtain ( [ h - h4 ] ) after @xmath48-integration from @xmath23 down to @xmath48 . clearly , this is the only contribution for @xmath432 so that ( [ h - h4 ] ) is proven for @xmath432 and @xmath415 .
+ in the second step we use this result to extend the proof to @xmath432 and @xmath384 .
now the differential equation ( [ polb4 ] ) reduces to the second line of ( [ hh4-a ] ) , with @xmath384 , and the fourth to sixth lines of ( [ polb4 ] ) with @xmath432 and @xmath416 .
there is no contribution from the seventh line of ( [ polb4 ] ) for @xmath416 , because the part in braces would be non - planar , which is excluded in definition [ defint].[defint1 ] . inserting ( [ initrho4 ] )
we obtain the composite propagator ( [ comp - prop-1 ] ) in the part in braces @xmath160 of the fifth line of ( [ polb4 ] ) .
together with ( [ h - h4 ] ) for @xmath432 and @xmath415 already proven we verify the integrand ( [ h - h4 ] ) for @xmath432 and @xmath384 .
after @xmath48-integration we thus obtain ( [ h - h4 ] ) for @xmath432 and any @xmath162 .
+ this allows us to use ( [ h - h4 ] ) as induction hypothesis for the remaining contributions in the last line of ( [ hh4-a ] ) .
this is similar to the procedure in [ h - non - planar ] , we only have to replace ( [ hn4 - 2b ] ) , ( [ hn4 - 3b ] ) and ( [ hn4 - 4b ] ) by the according parametrisation of ( [ h - h4 ] ) .
we thus prove ( [ h - h4 ] ) to all orders .
+ 1ex 4 .
we first consider @xmath415 .
then , according to ( [ initrho1])([initrho3 ] ) only terms with @xmath433 contribute to ( [ polb4 ] ) . using ( [ initrho1])([initrho4 ] ) and
the fact that @xmath94 is 1pi , the differential equation ( [ polb4 ] ) reduces to + [ hh0 ] @xmath434 \big)_{a\neq 4}\nonumber \\ * & = - \frac{1}{2 } \bigg\ { \sum_{m , n , k , l } q_{nm;lk}(\lambda ) \big ( h^{a(v , v^e , b,0,\iota)}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2};mn;kl}[\lambda ] - h^{a(v , v^e , b,0,\iota)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \nonumber \\ & \qquad\qquad - m^1 \big(h^{a(v , v^e , b,0,\iota ) } _ { { \genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};mn;kl}[\lambda ] -h^{a(v , v^e , b,0,\iota ) } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda]\big ) \nonumber \\ * & \qquad\qquad - n^1 \big(h^{a(v , v^e , b,0,\iota ) } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] -h^{a(v , v^e , b,0,\iota ) } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda]\big ) \nonumber \\ & \qquad\qquad -m^2 \big(h^{a(v , v^e , b,0,\iota ) } _ { { \genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};mn;kl}[\lambda ] -h^{a(v , v^e , b,0,\iota ) } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda]\big ) \nonumber \\ * & \qquad\qquad - n^2 \big(h^{a(v , v^e , b,0,\iota ) } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] -h^{a(v , v^e , b,0,\iota ) } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda]\big ) \big ) \bigg\}_{[\text{def.~\ref{defint}.\ref{defint0 } } ] } \label{hh0-a } \\ & - h^{4(0)}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda ] \bigg\
{ -\frac{1}{2 } \sum_{m , n , k , l } q_{nm;lk}(\lambda ) h^{a(v)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint4 } } ] } \label{hh0-b } \\ * & + \text{the $ 4^{\text{th}}$ to last lines of ( \ref{polb4 } ) with $ \displaystyle \sum_{v_1=0}^{v } \mapsto \sum_{v_1=1}^{v-1}$}\;. \label{hh0-c}\end{aligned}\ ] ] + if the graphs have constant indices along the trajectories , we conclude in the same way as in appendix [ appcompositeidentity ] that the part ( [ hh0-a ] ) can be written as a linear combination of graphs having either a composite propagator ( [ comp - prop-1 ] ) or two composite propagators ( [ comp - prop-0 ] ) on the trajectories . as such we have to replace the bound ( [ est4 - 1 ] ) relative to the contribution of an ordinary propagator by ( [ qpl2 ] ) or twice ( [ est4 - 1 ] ) by ( [ qpl1 ] ) . for the total graph
this amounts to multiply the corresponding estimation ( [ hn4 - 1plan ] ) of ordinary @xmath365-graphs with @xmath109 by a factor @xmath435 , which yields the subscript @xmath147 of the part @xmath436 $ ] of the integrand ( [ h - h0 ] ) , for the time being restricted to the part ( [ hh0-a ] ) . for graphs with index jump in definition [ defint].[defint0 ]
we obtain according to appendix [ appcompositeidentity ] the same improvement by @xmath437 .
next , the product of ( [ h0 - 42n ] ) with ( [ hn4 - 4a ] ) gives for ( [ hh0-b ] ) the same bound ( [ h - h0 ] ) for the integrand .
since the resulting integrand is irrelevant , we also obtain ( [ h - h0 ] ) after @xmath48-integration . clearly , this is the only contribution for @xmath432 so that ( [ h - h0 ] ) is proven for @xmath432 and @xmath415 .
+ in the second step we use this result to extend the proof to @xmath432 and @xmath384 . now
the differential equation ( [ polb4 ] ) reduces to the sum of ( [ hh0-a ] ) and ( [ hh0-b ] ) , with @xmath384 , and the fourth to sixth lines of ( [ polb4 ] ) with @xmath432 and @xmath416 .
there is again no contribution of the seventh line of ( [ polb4 ] ) for @xmath416 . inserting ( [ initrho4 ] )
we obtain the composite propagators ( [ comp - prop-1 ] ) in the fifth / sixth lines of ( [ polb4 ] ) , which together with ( [ h - h0 ] ) for @xmath432 and @xmath415 already proven verifies the integrand ( [ h - h0 ] ) for @xmath432 and @xmath384 .
after @xmath48-integration we thus obtain ( [ h - h0 ] ) for @xmath432 and any @xmath162 . +
this allows us to use ( [ h - h0 ] ) as induction hypothesis for the remaining contributions ( [ hh0-c ] ) for @xmath438 .
this is similar to the procedure in [ h - non - planar ] , we only have to replace ( [ hn4 - 2b ] ) , ( [ hn4 - 3b ] ) and ( [ hn4 - 4b ] ) by the according parametrisation of ( [ h - h0 ] ) .
we thus prove ( [ h - h0 ] ) to all orders .
+ 1ex 5 .
the proof of ( [ h - h1 ] ) is performed along the same lines as the proof of ( [ h - h4 ] ) and ( [ h - h0 ] ) .
there is one factor @xmath439 from @xmath440}\rangle + \langle \overrightarrow{n_2\mathfrak{o}[n_2]}\rangle = 2 $ ] in ( [ hn4 - 1plan ] ) and a second factor from the composite propagator ( [ qpl1 ] ) or ( [ qpl3 ] ) appearing according to appendix [ appcompositeidentity ] in the @xmath429-contribution to ( [ polb4 ] ) .
this finishes the proof of proposition [ prop - h ] . @xmath334
the estimations in propositions [ power - counting - prop ] and [ prop - h ] allow us to estimate the @xmath366-functions by integrating the differential equation ( [ polv4 ] ) .
again , the @xmath366-functions are expanded in terms of ribbon graphs .
let us look at @xmath366-ribbon graphs of the type described in definition [ defint].[defint4 ] .
since @xmath441\equiv \rho_4[\lambda,\lambda_0,\rho^0]$ ] , we can rewrite the expansion coefficients of ( [ rdef ] ) as follows : @xmath442\nonumber \\ * & = \lambda_0 \frac{\partial}{\partial \lambda_0 } \sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint4 } } } \big(a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } ; { \genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2};{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda,\lambda_0,\rho^0 ] \nonumber \\*[-1ex ] & \hspace*{11em } - a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda,\lambda_0,\rho^0]\big ) \nonumber \\ * & - \sum_{a , b=1}^4 \frac{\partial}{\partial \rho^0_a } \sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint4 } } } \big ( a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } ; { \genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2};{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda,\lambda_0,\rho^0 ] \nonumber \\*[-1ex ] & \hspace*{11em } - a^{(v)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad\qquad \times \frac{\partial \rho^0_a}{\partial \rho_b[\lambda,\lambda_0,\rho^0 ] } \lambda_0 \frac{\partial}{\partial \lambda_0 } \rho_b[\lambda,\lambda_0,\rho^0]\;. \label{rproj}\end{aligned}\ ] ] this means that ( by construction ) only the @xmath443-derivatives of the projection to the irrelevant part ( [ hata4 ] ) of the planar four - point function contributes to @xmath366 .
similarly , only the @xmath443-derivatives of the irrelevant parts ( [ hata0 ] ) and ( [ hata1 ] ) of the planar two - point function contribute to @xmath366 . according to the initial condition ( [ ct4 ] ) ,
these projections and the other functions given in definition [ defint].[defintnp ] vanish at @xmath105 _ independently _ of @xmath23 or @xmath338 : [ initr ] @xmath444 \nonumber \\ * & \hspace*{7em } - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_0,\lambda_0,\rho^0]\big ) \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defint4 } } } \nonumber \\ * & = \frac{\partial}{\partial \rho^0_a } \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } ; { \genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2};{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda_0,\lambda_0,\rho^0 ] \nonumber \\ * & \hspace*{7em } - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda_0,\lambda_0,\rho^0]\big ) \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defint4 } } } \ ; , \label{initr4 } \\ 0&= \lambda_0 \frac{\partial}{\partial \lambda_0 } \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0 ] \nonumber \\ * & \qquad - m^1 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad - n^1 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad - m^2 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad - n^2 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big)\big ) \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defint0 } } } \nonumber \\ & = \frac{\partial}{\partial \rho^0_a } \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0 ] \nonumber \\ * & \qquad - m^1 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad - n^1 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big ) \nonumber \\ & \qquad - m^2 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad - n^2 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big)\big ) \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defint0 } } } \ ; , \label{initr0 } \\ 0&= \lambda_0 \frac{\partial}{\partial \lambda_0 } \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda_0,\lambda_0,\rho^0 ] \nonumber \\ * & \qquad - \sqrt{(m^1{+}1)(n^1{+}1 ) } a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big ) \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defint1 } } } \nonumber \\ * & = \frac{\partial}{\partial \rho^0_a } \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda_0,\lambda_0,\rho^0 ] \nonumber \\ * & \qquad - \sqrt{(m^1{+}1)(n^1{+}1 ) } a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda_0,\lambda_0,\rho^0]\big ) \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defint1}}}\ ; , \label{initr1 } \\ 0&= \lambda_0 \frac{\partial}{\partial \lambda_0 } a^{(v , v^e , b , g,\iota)\gamma}_{m_1n_1;\dots;m_nn_n } [ \lambda_0,\lambda_0,\rho^0 ] \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defintnp } } } \nonumber \\ * & = \frac{\partial}{\partial \rho^0_a } a^{(v , v^e , b , g,\iota)\gamma}_{m_1n_1;\dots;m_nn_n } [ \lambda_0,\lambda_0,\rho^0 ] \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defintnp}}}\;. \label{initrnp}\end{aligned}\ ] ] the @xmath23-derivative at @xmath105 has to be considered with care : @xmath445 \nonumber \\ * & = \big(\lambda \frac{\partial}{\partial \lambda } a^{(v , v^e , b , g,\iota)\gamma}_{m_1n_1;\dots;m_nn_n } [ \lambda,\lambda_0,\rho^0 ] \big)_{\lambda = \lambda_0 } \!\ ! { + } \big(\lambda_0 \frac{\partial}{\partial \lambda_0 } a^{(v , v^e ,
b , g,\iota)\gamma}_{m_1n_1;\dots;m_nn_n } [ \lambda,\lambda_0,\rho^0 ] \big)_{\lambda = \lambda_0 } , \label{initrnp0}\end{aligned}\ ] ] and similarly for ( [ initr4])([initr1 ] ) . inserting ( [ rproj ] ) ,
( [ initr ] ) , ( [ initrnp0 ] ) and according formulae into the taylor expansion of ( [ rdef ] ) we thus have [ rn4 ] @xmath446 \nonumber \\ * & = - \sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint4 } } } \bigg(\lambda\frac{\partial}{\partial \lambda}\big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } ; { \genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2};{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda,\lambda_0,\rho^0 ] \nonumber \\ * & \hspace*{10em } - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda,\lambda_0,\rho^0]\big ) \bigg)_{\lambda=\lambda_0}\ , , \label{rn4 - 40 } \\ & \sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint0 } } } r^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda_0,\lambda_0,\rho^0 ] \nonumber \\ * & = - \sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint0 } } } \bigg(\lambda\frac{\partial}{\partial \lambda}\big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0 ] \nonumber \\ * & \qquad - m^1 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0 } } [ \lambda,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad - n^1 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad - m^2 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1 } } [ \lambda,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0]\big ) \nonumber \\ * & \qquad - n^2 \big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0 ] - a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0]\big)\big ) \bigg)_{\lambda=\lambda_0}\ ; , \\ & \sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint1 } } } r^{(v,2,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda_0,\lambda_0,\rho^0 ] \nonumber \\ * & = -\sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint1 } } } \bigg(\lambda\frac{\partial}{\partial \lambda}\big ( a^{(v , v^e,1,0,0)\gamma } _ { { \genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda,\lambda_0,\rho^0 ] \nonumber \\ * & \qquad - \sqrt{(m^1{+}1)(n^1{+}1 ) } a^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } } [ \lambda,\lambda_0,\rho^0]\big)\bigg)_{\lambda=\lambda_0}\ ; , \\ & r^{(v , v^e , b , g,\iota)\gamma}_{m_1n_1;\dots;m_nn_n } [ \lambda_0,\lambda_0,\rho^0 ] \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defintnp } } } \nonumber \\ * & = -\bigg(\lambda\frac{\partial}{\partial \lambda } a^{(v , v^e , b , g,\iota)\gamma}_{m_1n_1;\dots;m_nn_n } [ \lambda,\lambda_0,\rho^0 ] \big|_{\gamma \text { as in def.~\ref{defint}.\ref{defintnp } } } \bigg)_{\lambda=\lambda_0}\;.\end{aligned}\ ] ] in particular , @xmath447 \equiv 0\;. \label{rn4 - 40 - 0}\end{aligned}\ ] ] we first get ( [ rn4 - 40 - 0 ] ) at @xmath105 from ( [ rn4 - 40 ] ) . since the rhs of ( [ polv4 ] ) vanishes for @xmath432 and @xmath99 , we conclude ( [ rn4 - 40 - 0 ] ) for any @xmath48 .
[ r - prop ] let @xmath94 be an @xmath366-ribbon graph having @xmath77 external legs , @xmath76 vertices , @xmath78 external vertices and segmentation index @xmath79 , which is drawn on a genus-@xmath71 riemann surface with @xmath72 boundary components . then the contribution @xmath448 of @xmath94 to the expansion coefficient of the @xmath23-varied effective action describing a duality - covariant @xmath0-theory on @xmath27 in the matrix base is bounded as follows : 1 .
[ r - planar - r4 ] if @xmath94 is of the type described under [ defint4][defint1 ] of definition [ defint ] , we have @xmath449\big| \nonumber \\ * & \quad \qquad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) p^{4v-4}_1 \bigg [ \frac{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{k^1}{k^2 } ; { \genfrac{}{}{0pt}{}}{k^1}{k^2}{\genfrac{}{}{0pt}{}}{l^1}{l^2};{\genfrac{}{}{0pt}{}}{l^1}{l^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2 } \bigg ] \big(\frac{1}{\omega}\big)^{3v-2-v^e } \,p^{2v-2}\big[\ln\frac{\lambda_0}{\lambda_r}\big ] \ ; , \label{r4 } \\ & \sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint0 } } } \big|r^{(v , v^e,1,0,0)\gamma}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } [ \lambda,\lambda_0,\rho_0 ] \big| \nonumber \\ * & \quad \qquad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) ( \theta \lambda^2 ) p^{4v-2}_2 \bigg [ \frac{{\genfrac{}{}{0pt}{}}{m^1}{m^2}{\genfrac{}{}{0pt}{}}{n^1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2 } \bigg ] \big(\frac{1}{\omega}\big)^{3v-1-v^e } \,p^{2v-1}\big[\ln\frac{\lambda_0}{\lambda_r}\big ] \ ; , \label{r0 } \\ & \sum_{\gamma \text { as in def.~\ref{defint}.\ref{defint1 } } } \big| r^{(v , v^e,1,0,0)}_{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } ; { \genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2}}[\lambda,\lambda_0,\rho_0 ] \big| \nonumber \\ * & \quad \qquad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) ( \theta \lambda^2 ) p^{4v-2}_2 \bigg [ \frac{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2};{\genfrac{}{}{0pt}{}}{n^1}{n^2}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } { \theta \lambda^2 } \bigg ] \big(\frac{1}{\omega}\big)^{3v-1-v^e } \,p^{2v-1}\big[\ln\frac{\lambda_0}{\lambda_r}\big ] \ ; , \label{r1}\end{aligned}\ ] ] 2 .
[ r - sub - planar ] if @xmath94 is a subgraph of an 1pi planar graph with a selected set @xmath202 of trajectories on one distinguished boundary component and a second set @xmath205 of summed trajectories on that boundary component , we have @xmath450 \big| \nonumber \\ * & \quad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) \big(\theta \lambda^2\big)^{(2-\frac{n}{2})+2(1-b ) } \big(\frac{1}{\omega}\big)^{3v-\frac{n}{2}-1 + b+2g - v^e-\iota+s+t ' } \nonumber \\ * & \qquad \times p^{4v - n}_{\big(2t'+ \sum_{\overrightarrow{n_j\mathfrak{o}[n_j ] } \in t } \min(2,\frac{1}{2}\langle \overrightarrow{n_j\mathfrak{o}[n_j ] } \rangle)\big ) } \bigg [ \frac { m_1n_1;\dots;m_nn_n}{\theta \lambda^2 } \bigg ] p^{2v-\frac{n}{2}}\big[\ln \frac{\lambda_0}{\lambda_r}\big]\;. \label{rn4-sub}\end{aligned}\ ] ] 3 .
[ r - non - planar ] if @xmath94 is a non - planar graph , we have @xmath451 \big| \nonumber \\ * & \quad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) \big(\theta \lambda^2\big)^{(2-\frac{n}{2})+2(1-b-2 g ) } p^{4v - n}_0 \bigg [ \frac { m_1n_1;\dots;m_nn_n}{\theta \lambda^2 } \bigg ] \nonumber \\ * & \qquad \times \big(\frac{1}{\omega}\big)^{3v-\frac{n}{2}-1 + b+2g - v^e-\iota+s } \,p^{2v-\frac{n}{2}}\big[\ln \frac{\lambda_0}{\lambda_r}\big]\;. \label{rn4-np}\end{aligned}\ ] ] we have @xmath452 for @xmath453 or @xmath454 .
_ inserting the estimations of proposition [ power - counting - prop ] into ( [ rn4 ] ) we confirm proposition [ r - prop ] for @xmath105 , which serves as initial condition for the @xmath48-integration of ( [ polv4 ] ) .
this entails the polynomial in @xmath455 instead of @xmath239 appearing in propositions [ power - counting - prop ] and [ prop - h ] .
accordingly , when using propositions [ power - counting - prop ] and [ prop - h ] as the input for ( [ polv4 ] ) , we will further bound these estimations by replacing @xmath239 by @xmath456 . due to ( [ rn4 - 40 - 0 ] ) the rhs of ( [ polv4 ] ) vanishes for @xmath457 and for @xmath458 .
this means that the corresponding @xmath366-functions are constant in @xmath48 so that the proposition holds for @xmath459 $ ] , @xmath460 $ ] and @xmath461 $ ] .
since ( [ polv4 ] ) is a linear differential equation , the factor @xmath462 relative to the estimation of the @xmath151-functions of proposition [ power - counting - prop ] , first appearing in @xmath459 $ ] , @xmath460 $ ] and @xmath461 $ ] , survives to more complicated graphs , provided that none of the @xmath366-functions is relevant in @xmath48 . for graphs according to definition [ defint].[defintnp ] , the first two lines on the rhs of ( [ polv4 ] ) yield in the same way as in the proof of ( [ an4 - 1 ] ) the integrand ( [ rn4-np ] ) , with the degree of the polynomial in @xmath456 lowered by @xmath104 .
since under the given conditions an @xmath151-graph would be irrelevant , an @xmath366-graph with the additional factor @xmath462 is relevant or marginal .
thus , the @xmath48-integration of the first two lines on the rhs of ( [ polv4 ] ) can be estimated by the integrand and a factor @xmath463 $ ] , in agreement with ( [ rn4-np ] ) . in the same way we verify ( [ rn4-sub ] ) for the first two lines on the rhs of ( [ polv4 ] ) . in the remaining lines of ( [ polv4 ] ) we get by induction the following estimation : [ rn4-int ]
@xmath464 \bigg\}_{[\text{def.~\ref{defint}.\ref{defint0}}]}\bigg| \nonumber
\\ * & \qquad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) \big(\theta \lambda^2\big ) \big(\frac{1}{\omega}\big)^{3v-1-v^e } \,p^{2v-2}\big[\ln \frac{\lambda_0}{\lambda_r}\big]\ ; , \label{rn4-int - a } \\ & \bigg|\bigg\ { \sum_{m , n , k , l } q_{nm;lk}(\lambda ) \big(r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};mn;kl}[\lambda ] -r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda]\big ) \bigg\}_{[\text{def.~\ref{defint}.\ref{defint0 } } ] } \bigg| \nonumber \\ * & \qquad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) \big(\frac{1}{\omega}\big)^{3v-1-v^e } \,p^{2v-2}\big[\ln \frac{\lambda_0}{\lambda_r}\big]\ ; , \label{rn4-int - b } \\ & \bigg|\bigg\ { \sum_{m , n , k , l } q_{nm;lk}(\lambda ) r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint1}}]}\bigg| \nonumber \\ * & \qquad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) \big(\frac{1}{\omega}\big)^{3v-1-v^e } \,p^{2v-2}\big[\ln \frac{\lambda_0}{\lambda_r}\big]\ ; , \label{rn4-int - c } \\ & \bigg|\bigg\ { \sum_{m , n , k , l } q_{nm;lk}(\lambda ) r^{(v_1)}_{{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0 } ; { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};mn;kl}[\lambda ] \bigg\}_{[\text{def.~\ref{defint}.\ref{defint4}}]}\bigg| \nonumber \\ * & \qquad \leq \big(\frac{\lambda^2}{\lambda_0 ^ 2}\big ) \big(\frac{1}{\omega}\big)^{3v-2-v^e } \,p^{2v-3}\big[\ln \frac{\lambda_0}{\lambda_r}\big]\;. \label{rn4-int - d}\end{aligned}\ ] ] these estimations are obtained in a similar way as ( [ hn4 - 2a ] ) , ( [ hn4 - 3a ] ) and ( [ hn4 - 4a ] ) . in particular , the improvement by @xmath465 in ( [ rn4-int - b ] ) is due to the difference of graphs which according to section [ seccomposite ] yield a composite propagator ( [ comp - prop-0 ] ) . to obtain ( [ rn4-int - c ] ) we have to use ( [ rn4-sub ] ) with @xmath466 , which for the graphs under consideration is known by induction . multiplying ( [ rn4-int ] ) by versions of proposition [ prop - h ] according to ( [ polv4 ] ) , for @xmath467
, we obtain again ( [ rn4-np ] ) or ( [ rn4-sub ] ) , with the degree of the polynomial in @xmath455 lowered by @xmath104 , for the integrand
. then the @xmath48-integration proves ( [ rn4-np ] ) and ( [ rn4-sub ] ) . for graphs as in [ defint4][defint1 ] of definition [ defint ]
one shows in the same way as in the proof of [ h - planar - h4][h - planar - h1 ] of proposition [ prop - h ] that the last term in the third line of ( [ polv4 ] ) and the @xmath429-terms in the remaining lines project to the irrelevant part of these @xmath366-functions , i.e. lead to ( [ r4])([r1 ] ) .
this was already clear from ( [ rproj ] ) . for the remaining @xmath468-terms in the fourth to last lines of ( [ polv4 ] ) we obtain ( [ r4])([r1 ] ) from ( [ rn4-int ] ) and ( [ h - h4])([h - h1 ] ) .
this finishes the proof.@xmath334 we return now to the starting point of the entire estimation procedure the identity ( [ rlim ] ) .
we put @xmath66 in proposition [ r - prop ] and perform the @xmath23-integration in ( [ rlim ] ) : [ final - theorem ] the @xmath0-model on @xmath27 is ( order by order in the coupling constant ) renormalisable in the matrix base by adjusting the coefficients @xmath469 $ ] defined in ( [ rhoa0 ] ) and ( [ rhoa ] ) of the initial interaction ( [ ct4 ] ) to give ( [ initrho ] ) and by integrating the the polchinski equation according to definition [ defint ] .
the limit @xmath470 : = \lim_{\lambda_0 \to \infty } a^{(v , v^e , b , g,\iota)}_{m_1n_1;\dots;m_nn_n } [ \lambda_r,\lambda_0,\rho^0[\lambda_0]]$ ] of the expansion coefficients of the effective action @xmath471 $ ] , see ( [ lg4 ] ) , exists and satisfies @xmath472 - ( 2\pi\theta)^{\frac{n}{2}-2 } a^{(v , v^e , b , g,\iota)}_{m_1n_1;\dots;m_nn_n } [ \lambda_r,\lambda_0,\rho^0 ] \big| \nonumber \\ * & \quad \leq \frac{\lambda_r^{6-n}}{\lambda_0 ^ 2 } \big(\frac{1}{\omega\theta\lambda_r^2}\big)^{2(b+2g-1 ) } \nonumber \\ * & \qquad\times p^{4v - n}_0 \bigg [ \frac { m_1n_1;\dots;m_nn_n}{\theta \lambda_r^2 } \bigg ] \big(\frac{1}{\omega}\big)^{3v-\frac{n}{2}-v^e-\iota } \,p^{2v-\frac{n}{2}}\big[\ln \frac{\lambda_0}{\lambda_r}\big]\;. \label{lima4}\end{aligned}\ ] ] _ proof .
_ we insert proposition [ r - prop ] , taken at @xmath66 , into ( [ rlim ] )
. we also use ( [ polab - range ] ) in proposition [ r - prop].[r - planar - r4 ] .
now , the existence of the limit and its property ( [ lima4 ] ) are a consequence of cauchy s criterion .
note that @xmath473 = \frac{1}{x^2 } p^{\prime q}[\ln x]$ ] .
in this paper we have proven that the real @xmath0-model on ( euclidean ) noncommutative @xmath1 is renormalisable to all orders in perturbation theory .
the bare action of relevant and marginal couplings of the model is parametrised by four ( divergent ) quantities which require normalisation to the experimental data at a physical renormalisation scale .
the corresponding physical parameters which determine the model are the mass , the field amplitude ( to be normalised to @xmath104 ) , the coupling constant and ( in addition to the commutative version ) the frequency of an harmonic oscillator potential .
the appearance of the oscillator potential is not a bad trick but a true physical effect .
it is the self - consistent solution of the uv / ir - mixing problem found in the traditional noncommutative @xmath0-model in momentum space .
it implements the duality ( see also @xcite ) that _ noncommutativity relevant at short distances goes hand in hand with a modified structure of space relevant at large distances_. such a modified structure of space at very large distances seems to be in contradiction with experimental data .
but this is not true .
neither position space nor momentum space are the adapted frames to interpret the model .
an invariant characterisation of the model is the spectrum of the laplace - like operator which defines the free theory . due to the link to meixner polynomials ,
the spectrum is discrete . comparing ( [ b1 ] ) with ( [ b10 ] ) and ( [ b11 ] )
we see that the spectrum of the squared momentum variable has an equidistant spacing of @xmath474 .
thus , @xmath475 is the minimal ( non - vanishing ) momentum of the scalar field which is allowed in the noncommutative universe .
we can thus identify the parameter @xmath476 with the ratio of the planck length to the size of the ( finite ! ) universe .
thus , for typical momenta on earth , the discretisation is not visible .
however , there should be an observable effect at extremely huge scales .
indeed , there is some evidence of discrete momenta in the spectrum of the cosmic microwave background @xcite . ] .
of course , when we pass to a frame where the propagator becomes @xmath477 , with @xmath3 now being discrete , we also have to transform the interactions .
we thus have to shift the unitary matrices @xmath478 appearing in ( [ b1 ] ) from the kinetic matrix or the propagator into the vertex . the properties of that dressed ( physical ) vertex will be studied elsewhere .
another interesting exercise is the evaluation of the @xmath479-function of the duality - covariant @xmath0-model @xcite .
it turns out that the one - loop @xmath479-function for the coupling constant remains non - negative and and vanishes for the self - dual case @xmath480 .
moreover , the limit @xmath481 exists at the one - loop level .
this is related to the fact that the uv / ir - mixing in momentum space becomes problematic only at higher loop order . of particular interest
would be the limit @xmath482 .
in the developed approach , @xmath31 defines the reference size of an elementary cell in the moyal plane .
all dimensionful quantities , in particular the energy scale @xmath48 , are measured in units of ( appropriate powers of ) @xmath31 . in the final result of theorem [ final - theorem ] ,
these mass dimensions are restored .
then , we learn from ( [ lima4 ] ) that a finite @xmath31 regularises the non - planar graphs .
this means that for given @xmath23 and @xmath16 the limit @xmath482 can not be taken . on the other hand
, there could be a chance to let @xmath31 depend on @xmath23 in the same way as in the two - dimensional case @xcite the oscillator frequency @xmath157 was switched off with the limit @xmath25 .
however , this does not work .
the point is that taking in ( [ v2 ] ) instead of the @xmath23-derivative the @xmath31-derivative , there is now a contribution from the @xmath31-dependence of the propagator .
this leads in the analogue of the differential equation ( [ vvv ] ) to a term bilinear in @xmath352 .
looking at the proof of proposition [ r - prop ] , we see that this @xmath352-bilinear term will remove the factor @xmath26 .
thus , the limit @xmath482 is singular .
this is not surprising . in the limit @xmath482
the distinction between planar and non - planar graphs disappears ( which is immediately clear in momentum space ) .
then , non - planar two- and four - point functions should yield the same divergent values as their planar analogues . whereas the bare divergences in the planar sector are avoided by the mixed boundary conditions in [ defint4]-[defint1 ] of definition [ defint ] ,
the nave initial condition in definition [ defint].[defintnp ] for non - planar graphs leaves the bare divergences in the limit @xmath482 .
the next goal must be to generalise the renormalisation proof to gauge theories .
this requires probably a gauge - invariant extension of the harmonic osillator potential .
the result should be compared with string theory , because gauge theory on the moyal plane arises in the zero - slope limit of string theory in presence of a neveu - schwarz @xmath72-field @xcite .
as renormalisation requires an appropriate structure of the space at very large distances , the question arises whether the oscillator potential has a counterpart in string theory . in this respect
, it is tempting to relate the oscillator potential to the maximally supersymmetric pp - wave background metric of type iib string theory found in @xcite , @xmath483 for @xmath484 , which solves einstein s equations for an energy - momentum tensor relative to the 5-form field strength @xmath485 we are indebted to stefan schraml for providing us with the references to orthogonal polynomials , without which the completion of the proof would have been impossible . we had stimulating discussions with edwin langmann , vincent rivasseau and harold steinacker .
we are grateful to christoph kopper for indicating to us a way to reduce in our original power - counting estimation the polynomial in @xmath486 to a polynomial in @xmath487 , thus permitting immediately the limit @xmath25 .
we would like to thank the max - planck - institute for mathematics in the sciences ( especially eberhard zeidler ) , the erwin - schrdinger - institute and the institute for theoretical physics of the university of vienna for the generous support of our collaboration .
our goal is to diagonalise the ( four - dimensional ) kinetic matrix @xmath488 given in ( [ g4d ] ) , making use of the angular momentum conservation @xmath489 ( which is due to the @xmath490-symmetry of the action . for @xmath491
we thus look for a representation @xmath492 the sum over @xmath493 would be an integration for continuous eigenvalues @xmath494 . comparing this ansatz with ( [ g4d ] )
we obtain , eliminating @xmath495 in favour of @xmath496 , the recurrence relation @xmath497 to determine @xmath498 and @xmath496 .
we are interested in the case @xmath499 . in order to make contact with standard formulae we put @xmath500 we obtain after division by @xmath501 @xmath502 i.e. @xmath503
now we put @xmath504 and @xmath505 which yields the recursion relation for the meixner polynomials @xcite : @xmath506 the solution of ( [ lagmeix ] ) is @xmath507 we have to chose the upper sign , because the eigenvalues @xmath496 are positive .
we thus obtain @xmath508 the function @xmath509 is identified by comparison of ( [ b1a ] ) with the orthogonality relation of meixner polynomials @xcite , @xmath510 the result is @xmath511 the meixner polynomials can be represented by hypergeometric functions @xcite @xmath512 this shows that the matrices @xmath513 in ( [ b1 ] ) and ( [ b1a ] ) are symmetric in the lower indices .
now we return to the computation of the propagator , which is obtained by sandwiching the inverse eigenvalues @xmath514 between the unitary matrices @xmath515 . with ( [ b10 ] ) and the use of schwinger s trick @xmath516 we have for @xmath517 @xmath518 we use the following identity for hypergeometric functions , @xmath519 the identity ( [ fff ] )
is probably known , but because it is crucial for the solution of the free theory , we provide the proof in section [ id - hypergeom ] .
we insert the rhs of ( [ fff ] ) , expanded as a finite sum , into ( [ schwinger ] ) , where we also put @xmath520 : @xmath521 this formula tells us the important property @xmath522 i.e. all matrix elements of the propagator are positive and majorised by the massless matrix elements .
the representation ( [ schwinger1 ] ) seems to be the most convenient one for analytical estimations of the propagator .
the strategy would be to divide the integration domain into slices and to maximise the individual @xmath523-dependent terms of the integrand over the slice , followed by resummation @xcite .
the @xmath523-integration in ( [ schwinger1 ] ) leads according to @xcite again to a hypergeometric function : @xmath524 we have used @xcite to obtain the last line .
the form ( [ propalpha ] ) will be useful for the evaluation of special cases and of the asymptotic behaviour . in the main part , for presentational purposes ,
@xmath525 is eliminated in favour of @xmath526 and the summation variable @xmath527 is used .
the final result is given in ( [ noalpha ] ) .
for @xmath528 we can in few cases evaluate the sum over @xmath529 exactly .
first , for @xmath530 we also have @xmath531 .
if additionally @xmath532 we get @xmath533 one should notice here the exponential decay for @xmath499 .
it can be seen numerically that this is a general feature of the propagator : given @xmath59 and @xmath525 , the maximum of the propagator is attained at @xmath534 . moreover , the decay with @xmath535 is exponentially so that the sum @xmath536 converges .
we confirm this argumentation numerically in ( [ form4 - 2 ] ) .
it turns out numerically that the maximum of the propagator for indices restricted by @xmath537 is found in the subclass @xmath538 of propagators .
coincidently , the computation in case of @xmath539 simplifies considerably . if additionally @xmath540 we obtain a closed result : @xmath541 \dfrac{\sqrt{\pi}\theta}{4 \sqrt{m{+}\frac{3}{4 } } } \qquad & \text{for } \omega=0\;,~m \gg 1\;. \end{array}\right .
\label{propmmmm}\end{aligned}\ ] ] we see a crucial difference in the asymptotic behaviour for @xmath499 versus @xmath18 .
the slow decay with @xmath542 of the propagator is responsible for the non - renormalisability of the @xmath0-model in case of @xmath18 .
the numerical result ( [ form4 - 1 ] ) shows that the maximum of the propagator for indices restricted by @xmath543 is very close to the result ( [ propmmmm ] ) , for @xmath544 . for @xmath18
the maximum is exactly given by the @xmath545 line of ( [ propmmmm ] ) .
we consider various limiting cases of the propagator , making use of the asymptotic expansion ( stirling s formula ) of the @xmath546-function , @xmath547 this implies @xmath548 we rewrite the propagator ( [ propalpha ] ) in a manner where the large-@xmath525 behaviour is easier to discuss : @xmath549
we assume @xmath550 .
the term in braces @xmath160 in ( [ delta - as ] ) behaves like @xmath551 & \qquad + \mathcal{o}\big((\alpha^1{+}\alpha^2)^{-2}\big ) \big ) \nonumber \\ * & \times \big(1 + \frac{(m^1{-}u^1)(m^1{+}u^1{+}1)}{4\alpha^1 } + \frac{(l^1{-}u^1)(l^1{+}u^1{+}1)}{4\alpha^1 } + \mathcal{o}\big((\alpha^1)^{-2}\big ) \big ) \nonumber \\ * & \times \big(1 + \frac{(m^2{-}u^2)(m^2{+}u^2{+}1)}{4\alpha^2 } + \frac{(l^2{-}u^2)(l^2{+}u^2{+}1)}{4\alpha^1 } + \mathcal{o}\big((\alpha^2)^{-2}\big ) \big)\;. \label{est - brace-1}\end{aligned}\ ] ] we look for the maximum of the propagator under the condition @xmath552 . defining @xmath553 and @xmath554 , the dominating term in ( [ est - brace-1 ] )
is @xmath555 the maximum is attained at @xmath556 for @xmath557 and at @xmath558 for @xmath559 .
thus , the leading contribution to the propagator will come from the summation index @xmath560 .
next we evaluate the leading contribution of the hypergeometric function : @xmath561 assuming @xmath557 , we obtain from ( [ stirling ] ) , ( [ est - brace ] ) and ( [ est - f ] ) the following leading contribution to the propagator ( [ delta - as ] ) @xmath562 the numerator comes from @xmath563 for @xmath564 .
the estimation ( [ delta - ss ] ) is the explanation of ( [ est4 - 3 ] ) .
let us now look at propagators with @xmath565 and @xmath566 : @xmath567 this means @xmath568 the second and third line of ( [ delta-00 ] ) explains the estimation ( [ qpl1 ] ) . clearly , the next term in the expansion is of the order @xmath569 , which explains the estimation ( [ qpl2 ] ) . for @xmath570 and @xmath571
we have @xmath572 this yields @xmath573 which explains the estimation ( [ qpl3 ] ) .
similarly , we have @xmath574 which shows that the norm of ( [ comp - prop - rest ] ) is of the same order as ( [ qpl3 ] ) .
for terminating hypergeometric series ( @xmath575 ) we compute the sum in the last line of ( [ schwinger ] ) : @xmath576 in the step denoted by @xmath577 we have used @xcite .
all other transformations should be self - explaining .
we continue here the discussion of section [ seccomposite ] on composite propagators generated by differences of interaction coefficients . after having derived ( [ deca4 ] )
, we now have a look at ( [ inta2 ] ) .
since @xmath94 is one - particle irreducible , we get for a certain permutation @xmath171 ensuring the history of integrations the following linear combination : [ deca2-all ] @xmath578 } -\sqrt{(m^1{+}1)(n^1{+}1 ) } \widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } \nonumber \\ * & = \dots \left\ { \prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } } ( \lambda_{\pi_i } ) q_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}(k_{\pi_p}+1^+);k_{\pi_p } { \genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_p } ) \right .
\nonumber \\ * & \qquad \qquad \times \prod_{i={p+1}}^{a } q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) \prod_{j=1}^{q-1 } q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_j } ) \nonumber \\ * & \qquad \qquad \times q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_q } ; ( l_{\pi_q}+1^+){\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}}(\lambda_{\pi_q } ) \prod_{j={q+1}}^{b } q_{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2 } } ( \lambda_{\pi_j } ) \nonumber \\ * & - \sqrt{(m^1{+}1 ) ( n^1{+}1 ) } \prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) q_{{\genfrac{}{}{0pt}{}}{1}{0}(k_{\pi_p}+1^+);k_{\pi_p } { \genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_p } ) \ !
\prod_{i={p+1}}^{a } \!\ !
q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i } ) \nonumber \\ * & \qquad \qquad \times \left .
\prod_{j=1}^{q } q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j } ) q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_q } ; ( l_{\pi_q}+1^+){\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_q } ) \prod_{j={q+1}}^{b } q_{{\genfrac{}{}{0pt}{}}{1}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_j } ) \right\ } \nonumber \\ & = \dots \left\ { \bigg(\prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } } ( \lambda_{\pi_i } ) \right .
\nonumber \\*[-2ex ] & \qquad\qquad\qquad \times q_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}(k_{\pi_p}+1^+);k_{\pi_p } { \genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_p } ) \prod_{i={p+1}}^{a } q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) \nonumber \\*[-1ex ] & \qquad \quad - \sqrt{n^1{+}1 } \prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) q_{{\genfrac{}{}{0pt}{}}{1}{0}(k_{\pi_p}+1^+);k_{\pi_p } { \genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_p } ) \prod_{i={p+1}}^{a } q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i } ) \bigg ) \nonumber \\*[-1ex ] & \qquad \times \prod_{j=1}^{q-1 } q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_j } ) q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_q } ; ( l_{\pi_q}+1^+){\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}}(\lambda_{\pi_q } ) \ ! \prod_{j={q+1}}^{b } \!\ ! q_{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2 } } ( \lambda_{\pi_j } ) \label{deca2-a } \\[-1ex ] & + \sqrt{n^1{+}1 } \prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) q_{{\genfrac{}{}{0pt}{}}{1}{0}(k_{\pi_p}+1^+);k_{\pi_p } { \genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_p } ) \prod_{i={p+1}}^{a } q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i } ) \nonumber \\ * & \quad \times \bigg ( \prod_{j=1}^{q-1 } q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_j } ) q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_q } ; ( l_{\pi_q}+1^+){\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}}(\lambda_{\pi_q } ) \!\prod_{i={q+1}}^{b } \!\ ! q_{{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1 + 1}{m^2 } } ( \lambda_{\pi_j } ) \nonumber \\ * & \qquad -\sqrt{(m^1{+}1 ) } \left . \prod_{j=1}^{q } q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j } ) q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_q } ; ( l_{\pi_q}+1^+){\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_q } ) \!\!\prod_{j={q+1}}^{b } \!\ !
q_{{\genfrac{}{}{0pt}{}}{1}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_j } ) \bigg ) \right\ } , \label{deca2-b}\end{aligned}\ ] ] with @xmath579 .
we further analyse the the first three lines of ( [ deca2-a ] ) : @xmath580 & \quad \quad - \sqrt{n^1{+}1 } \prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) q_{{\genfrac{}{}{0pt}{}}{1}{0}(k_{\pi_p}+1^+);k_{\pi_p } { \genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_p } ) \prod_{i={p+1}}^{a } q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i } ) \bigg ) \nonumber \\[-1ex ] & = \bigg ( \big(\prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } } ( \lambda_{\pi_i } ) - \prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) \big ) \nonumber \\*[-1ex ] & \qquad\qquad \times q_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}(k_{\pi_p}+1^+);k_{\pi_p } { \genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_p } ) \prod_{i={p+1}}^{a } q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) \bigg ) \label{deca2-a - a } \\*[-1ex ] & + \prod_{i=1}^{p-1 } \ !
q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) \big ( \mathcal{q}^{(+\frac{1}{2})}_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}(k_{\pi_p}+1^+ ) ; k_{\pi_p } { \genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_p } ) \big ) \ !
\prod_{i={p+1}}^{a } \!\
! q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) \label{deca2-a - b } \\ * & + \sqrt{n^1{+}1 } \bigg ( \prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) q_{{\genfrac{}{}{0pt}{}}{1}{0}(k_{\pi_p}+1^+);k_{\pi_p } { \genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_p } ) \nonumber \\ * & \qquad\qquad \times \big(\prod_{i={p+1}}^{a } q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) - \prod_{i={p+1}}^{a } q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i } ) \big)\bigg)\;. \label{deca2-a - c}\end{aligned}\ ] ] according to ( [ deca4 ] ) , the two lines ( [ deca2-a - a ] ) and ( [ deca2-a - c ] ) yield graphs having one composite propagator ( [ comp - prop-0 ] ) , whereas the line ( [ deca2-a - b ] ) yields a graph having one composite propagator ) yields @xmath581 .
therefore , the prefactor @xmath582 in ( [ deca2-a - c ] ) combines actually to the ratio @xmath583 which is required for the ( [ polab])-term in proposition [ power - counting - prop].[prp - planar - a1].[fn - pre ] ] ( [ comp - prop - p12 ] ) . in total
, we get from ( [ deca2-all ] ) @xmath584 graphs with composite propagator ( [ comp - prop-0 ] ) or ( [ comp - prop - p12 ] ) .
the treatment of ( [ inta3 ] ) is similar .
second , we treat that contribution to ( [ inta1 ] ) which consists of graphs with constant index along the trajectories : [ deca1-all ] @xmath585 } \nonumber \\ * & \qquad -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } -m^1\big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\ * & \qquad -m^2\big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) -n^1\big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{1}{0};{\genfrac{}{}{0pt}{}}{1}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } \nonumber \\ * & \qquad -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) -n^2\big(\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{1};{\genfrac{}{}{0pt}{}}{0}{1}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda ' ] } -\widehat{\lambda'\frac{\partial}{\partial \lambda ' } a^{(v)\gamma } _ { { \genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0};{\genfrac{}{}{0pt}{}}{0}{0}{\genfrac{}{}{0pt}{}}{0}{0}}[\lambda']}\big ) \nonumber \\ & = \dots \bigg\ { \prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_j } ) \nonumber \\*[-1.5ex ] & \qquad\qquad\qquad - \prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i } ) \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j } ) \nonumber \\*[-0.5ex ] & \quad -\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i } ) \bigg(m^1 \big(\prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{1}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_j } ) - \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \nonumber \\*[-1ex ] & \qquad\qquad + m^2 \big(\prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{1}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{1 } } ( \lambda_{\pi_j } ) - \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big)\bigg ) \nonumber \\*[-0.5ex ] & \quad -\bigg(n^1 \big(\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) - \prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i})\big ) \nonumber \\*[-1ex ] & \qquad\qquad + n^2 \big(\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{1}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{1 } } ( \lambda_{\pi_i } ) - \prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i})\big)\bigg ) \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j } ) \bigg\ } \nonumber \\ & = \dots \bigg\ { \bigg(\big(\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) -\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i})\big ) \nonumber \\ * & \qquad\qquad \times \big(\prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_j } ) - \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big)\bigg ) \label{deca1-a } \\[-0.5ex ] & + \prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i } ) \bigg(\big(\prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_i } ) - \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \nonumber \\*[-1.3ex ] & \qquad\qquad - m^1 \big(\prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{1}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_j } ) - \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \nonumber \\*[-0.5ex ] & \qquad\qquad - m^2 \big(\prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{1}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{1 } } ( \lambda_{\pi_j } ) - \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big)\bigg ) \label{deca1-b } \\[-1ex ] & + \bigg ( \big(\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) -\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i})\big ) \nonumber \\*[-1ex ] & \qquad\qquad -n^1 \big(\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{1}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_i } ) - \prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i})\big ) \nonumber \\*[-1ex ] & \qquad\qquad - n^2 \big(\prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{1}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{1 } } ( \lambda_{\pi_i } ) - \prod_{i=1}^a q_{{\genfrac{}{}{0pt}{}}{0}{0}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_i})\big)\bigg ) \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j } ) \bigg\}. \label{deca1-c}\end{aligned}\ ] ] it is clear from ( [ deca4 ] ) that the part corresponding to ( [ deca1-a ] ) can be written as a sum of graphs containing ( at different trajectories ) two composite propagators @xmath586 and @xmath587 of type ( [ comp - prop-0 ] ) .
we further analyse ( [ deca1-b ] ) : [ deca1-b - all ] @xmath588 & \qquad - \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) - m^2 \big(\prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{1}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{1 } } ( \lambda_{\pi_j } ) - \prod_{j=1}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \nonumber \\[-1ex ] & = \mathcal{q}^{(1)}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_1};l_{\pi_1}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_1 } ) \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j } ) \label{deca1-b - a } \\*[-1ex ] & + \bigg ( \mathcal{q}^{(0)}_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_1};l_{\pi_1}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_1 } ) \big(\prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_j } ) - \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \nonumber \\*[-0.5ex ] & \qquad - m^1 \mathcal{q}^{(0)}_{{\genfrac{}{}{0pt}{}}{1}{0}l_{\pi_1};l_{\pi_1}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_1 } ) \big ( \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{1}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_j } ) - \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \nonumber \\*[-0.5ex ] & \qquad - m^2 \mathcal{q}^{(0)}_{{\genfrac{}{}{0pt}{}}{0}{1}l_{\pi_1};l_{\pi_1}{\genfrac{}{}{0pt}{}}{0}{1 } } ( \lambda_{\pi_1 } ) \big ( \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{1}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{1 } } ( \lambda_{\pi_j } ) - \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big)\bigg ) \label{deca1-b - b } \\[-1ex ] & + q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_1};l_{\pi_1}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_1 } ) \bigg ( \big(\prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{m^1}{m^2}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{m^1}{m^2 } } ( \lambda_{\pi_j } ) - \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \nonumber \\*[-1ex ] & \qquad - m^1 \big(\prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{1}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{1}{0 } } ( \lambda_{\pi_j } ) - \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \nonumber \\ * & \qquad - m^2 \big(\prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{1}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{1 } } ( \lambda_{\pi_j } ) - \prod_{j=2}^b q_{{\genfrac{}{}{0pt}{}}{0}{0}l_{\pi_j};l_{\pi_j}{\genfrac{}{}{0pt}{}}{0}{0 } } ( \lambda_{\pi_j})\big ) \bigg)\;. \label{deca1-b - c}\end{aligned}\ ] ] the part ( [ deca1-b - a ] ) gives rise to graphs with one propagator ( [ comp - prop-1 ] ) . due to ( [ deca4 ] ) the part ( [ deca1-b - b ] ) yields graphs with two propagators is according to ( [ qpl1 ] ) bounded by @xmath589 .
this means that the prefactors @xmath590 in ( [ deca1-b - b ] ) combine actually to the ratio @xmath591 which is required for the ( [ polab])-term in proposition [ power - counting - prop].[prp - planar - a0 ] . ]
( [ comp - prop-0 ] ) appearing on the same trajectory . finally , the part ( [ deca1-b - c ] ) has the same structure as the lhs of the equation , now starting with @xmath592 .
after iteration we obtain further graphs of the type ( [ deca1-b - a ] ) and ( [ deca1-b - b ] ) . finally , we look at that contribution to ( [ inta1 ] ) which consists of graphs where one index component jumps forward and backward in the @xmath593-component
. we can directly use the decomposition derived in ( [ deca1-all ] ) regarding , if the @xmath593-index jumps up , @xmath594 & \quad \mapsto \prod_{i=1}^{p-1 } q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i } ) q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_p};(k_{\pi_p}+1){\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2 } } ( \lambda_{\pi_p } ) \prod_{i = p+1}^{q-1 } q_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}k_{\pi_i } ; k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}}(\lambda_{\pi_i } ) \nonumber \\*[-1.5ex ] & \qquad\qquad \times q_{{\genfrac{}{}{0pt}{}}{n^1 + 1}{n^2}(k_{\pi_q}+1^+);k_{\pi_q}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_q } ) \prod_{i = q+1}^{a } q_{{\genfrac{}{}{0pt}{}}{n^1}{n^2}k_{\pi_i};k_{\pi_i}{\genfrac{}{}{0pt}{}}{n^1}{n^2 } } ( \lambda_{\pi_i})\;. \label{deca1-n}\end{aligned}\ ] ] this requires to process ( [ deca1-all ] ) slightly differently . the two parts ( [ deca1-a ] ) and ( [ deca1-b ] ) need no further discussion , as they lead to graphs having a composite propagator ( [ comp - prop-0 ] ) on the @xmath82-trajectory
. we write ( [ deca1-c ] ) as follows : [ deca1-c - all ] @xmath595 the part ( [ deca1-c - c ] ) leads according to ( [ deca4 ] ) and ( [ deca1-n ] ) to graphs either with composite propagators ( [ comp - prop-0 ] ) or with propagators @xmath596 inserting ( [ deca1-n ] ) into ( [ deca1-c - a ] ) we have [ deca1-c - a - all ] @xmath597 thus , we obtain ( recall also ( [ deca4 ] ) ) a linear combination of graphs either with composite propagator ( [ comp - prop-0 ] ) or with composite propagator ( [ comp - prop - p12 ] ) . in power - counting estimations , the prefactors @xmath582 combine according to footnote [ fn - pre ] to the required ratio with the scale @xmath598 . the part ( [ deca1-c - b ] ) is nothing but ( [ deca1-c - a ] ) with @xmath599 and @xmath600 .
if the index jumps down from @xmath593 to @xmath601 , then the graph with @xmath602 does not exist .
there is no change of the discussion of ( [ deca1-a ] ) and ( [ deca1-b ] ) , but now ( [ deca1-c ] ) becomes @xmath603 using the same steps as in ( [ deca1-c - a - all ] ) we obtain the desired representation through graphs either with composite propagator ( [ comp - prop-0 ] ) or with composite propagator ( [ comp - prop - p12 ] ) .
we show in appendix [ app : graph ] how the decomposition works in a concrete example . to make the considerations in section [ seccomposite ] and appendix [ appcompositeidentity ] about differences of graphs and composite propagators understandable , we look at the following example of a planar two - leg graph : @xmath604 according to proposition [ power - counting - prop ] it depends on the indices @xmath605 whether this graph is irrelevant , marginal , or relevant .
it depends on the history of contraction of subgraphs whether there are marginal subgraphs or not .
let us consider @xmath606 , @xmath607 , @xmath608 and @xmath609 and the history @xmath162-@xmath610-@xmath611-@xmath612-@xmath163 of contraction .
then , all resulting subgraphs are irrelevant and the total graph is marginal , which leads us to consider the following difference of graphs : @xmath613 it is important to understand that according to ( [ inta2 ] ) the indices at the external lines of the reference graph ( with zero - indices ) are adjusted to the external indices of the original ( leftmost ) graph : @xmath614\;.\end{aligned}\ ] ] thus , all graphs with composite propagators have the same index structure at the external legs .
when further contracting these graphs , the contracting propagator matches the external indices of the original graph .
the argumentation in the proof of proposition [ power - counting - prop].[prp - planar - a1 ] should be transparent now .
in particular , it becomes understandable why the difference ( [ brezel3 ] ) is irrelevant and can be integrated from @xmath23 down to @xmath48 . on the other hand ,
the reference graph to be integrated from @xmath16 up to @xmath48 becomes @xmath615 [ brezel4 ] @xmath619 the line ( [ brezel4a ] ) corresponds to the first graph in the braces @xmath160 of ( [ brezel - a - b ] ) for both graphs on the lhs of ( [ brezel4 ] ) .
these graphs are already irrelevant ) the composite propagator is according to ( [ qpl1 ] ) bounded by @xmath620 so that the combination with the prefactor @xmath621 leads to the ratio @xmath622 by which ( [ brezel4a ] ) is suppressed over the first graph on the lhs of ( [ brezel4 ] ) . ] so that no further decomposition is necessary . the second graph in the braces @xmath160 of ( [ brezel - a - b ] ) , inserted into the lhs of ( [ brezel4 ] ) , yields the line ( [ brezel4b ] ) .
finally , the last part of ( [ brezel - a - b ] ) leads to the line ( [ brezel4c ] ) .
let us also look at the relevant contribution @xmath623 , @xmath387 of the graph ( [ graph - at2 ] ) .
the history @xmath162-@xmath610-@xmath611-@xmath612-@xmath163 contains irrelevant subgraphs only : @xmath624 the line ( [ brezel0a ] ) corresponds to ( [ deca1-b - a ] ) , the line ( [ brezel0b ] ) to ( [ deca1-a ] ) and the line ( [ brezel0c ] ) to ( [ deca1-b - b ] ) .
if the history of contractions contains relevant or marginal subgraphs , we first have to decompose the subgraphs into the reference function with vanishing external indices and an irrelevant remainder .
for instance , the decomposition relative to the history @xmath162-@xmath163-@xmath610-@xmath611-@xmath612 would be @xmath625
for the power - counting theorem we need asymptotic formulae about the scaling behaviour of the cut - off propagator @xmath626 and certain index summations . we shall restrict ourselves to the case @xmath517 and deduce these formulae from the numerical evaluation of the propagator for a representative class of parameters and special choices of the parameters where we can compute the propagator exactly .
these formulae involve the cut - off propagator @xmath627 which is the restriction of @xmath628 to the support of the cut - off propagator @xmath629 appearing in the polchinski equation , with @xmath630 .
+ * formula 1 : * @xmath631 we demonstrate in figure [ fig - form4 - 1 ] for selected values of the parameters @xmath632 that @xmath633 at @xmath528 is asymptotically reproduced by @xmath634 .
s. minwalla , m. van raamsdonk and n. seiberg , `` noncommutative perturbative dynamics , '' jhep * 0002 * ( 2000 ) 020 [ arxiv : hep - th/9912072 ] .
i. chepelev and r. roiban , `` renormalization of quantum field theories on noncommutative @xmath639 .
i : scalars , '' jhep * 0005 * ( 2000 ) 037 [ arxiv : hep - th/9911098 ] . i. chepelev and r. roiban , `` convergence theorem for non - commutative feynman graphs and renormalization , '' jhep * 0103 * ( 2001 ) 001 [ arxiv : hep - th/0008090 ] .
e. langmann and r. j. szabo , `` duality in scalar field theory on noncommutative phase spaces , '' phys .
b * 533 * ( 2002 ) 168 [ arxiv : hep - th/0202039 ] .
v. gayral , j. m. gracia - bonda , b. iochum , t. schcker and j. c. vrilly , `` moyal planes are spectral triples , '' commun . math . phys . * 246 * ( 2004 ) 569 [ arxiv : hep - th/0307241 ] .
e. langmann , `` interacting fermions on noncommutative spaces : exactly solvable quantum field theories in 2n+1 dimensions , '' nucl .
b * 654 * ( 2003 ) 404 [ arxiv : hep - th/0205287 ] .
e. langmann , r. j. szabo and k. zarembo , `` exact solution of noncommutative field theory in background magnetic fields , '' phys .
b * 569 * ( 2003 ) 95 [ arxiv : hep - th/0303082 ] .
e. langmann , r. j. szabo and k. zarembo , `` exact solution of quantum field theory on noncommutative phase spaces , '' jhep * 0401 * ( 2004 ) 017 [ arxiv : hep - th/0308043 ] .
k. g. wilson and j. b. kogut , `` the renormalization group and the epsilon expansion , '' phys .
* 12 * ( 1974 ) 75 .
j. polchinski , `` renormalization and effective lagrangians , '' nucl .
b * 231 * ( 1984 ) 269 .
g. keller , c. kopper and m. salmhofer , `` perturbative renormalization and effective lagrangians in @xmath640 , '' helv .
acta * 65 * ( 1992 ) 32 .
h. grosse and r. wulkenhaar , `` power - counting theorem for non - local matrix models and renormalisation , '' arxiv : hep - th/0305066 , to appear in commun . math .
j. meixner , `` orthogonale polynomsysteme mit einer besonderen gestalt der erzeugenden funktion , '' j. london math .
* 9 * ( 1934 ) 6 .
h. grosse and r. wulkenhaar , `` renormalisation of @xmath0 theory on noncommutative @xmath641 in the matrix base , '' jhep * 0312 * ( 2003 ) 019 [ arxiv : hep - th/0307017 ] .
d. r. masson and j. repka , spectral theory of jacobi matrices in @xmath642 and the @xmath643 lie algebra , siam j. math .
22 * ( 1991 ) 1131 .
h. grosse and r. wulkenhaar , `` renormalisation of @xmath0 theory on noncommutative @xmath1 to all orders , '' arxiv : hep - th/0403232 , to appear in lett . math .
j. m. gracia - bonda and j. c. vrilly , `` algebras of distributions suitable for phase space quantum mechanics . 1 , '' j. math .
* 29 * ( 1988 ) 869 .
m. luminet , j. weeks , a. riazuelo , r. lehoucq and j. p. uzan , `` dodecahedral space topology as an explanation for weak wide - angle temperature correlations in the cosmic microwave background , '' nature * 425 * ( 2003 ) 593 [ arxiv : astro - ph/0310253 ] .
h. grosse & r. wulkenhaar , `` the @xmath479-function in duality - covariant noncommutative @xmath0-theory , '' eur .
j. c * 35 * ( 2004 ) 277282 [ hep - th/0402093 ] .
n. seiberg and e. witten , `` string theory and noncommutative geometry , '' jhep * 9909 * ( 1999 ) 032 [ arxiv : hep - th/9908142 ]
. m. blau , j. figueroa - ofarrill , c. hull and g. papadopoulos , `` a new maximally supersymmetric background of iib superstring theory , '' jhep * 0201 * , 047 ( 2002 ) [ arxiv : hep - th/0110242 ] .
i. s. gradshteyn and i. m. ryzhik , `` tables of series , produces , and integrals .
sixth edition , '' academic press , san diego ( 2000 ) .
|
we prove that the real four - dimensional euclidean noncommutative @xmath0-model is renormalisable to all orders in perturbation theory . compared with the commutative case
, the bare action of relevant and marginal couplings contains necessarily an additional term : an harmonic oscillator potential for the free scalar field action .
this entails a modified dispersion relation for the free theory , which becomes important at large distances ( uv / ir - entanglement ) .
the renormalisation proof relies on flow equations for the expansion coefficients of the effective action with respect to scalar fields written in the matrix base of the noncommutative @xmath1 .
the renormalisation flow depends on the topology of ribbon graphs and on the asymptotic and local behaviour of the propagator governed by orthogonal meixner polynomials .
|
galaxy collisions and mergers represent a key stage in the evolution of galaxies in the local universe and beyond . throughout cosmic time gravitational forces have assembled ever larger galactic systems from the collisions and mergers of smaller fragments . beginning with the theoretical @xcite sequence of merging galaxies , theoretical modeling of massive ,
merging galaxies now includes detailed physics , allowing us to make predictions about galaxy evolution over cosmic time .
the history of chemical enrichment is tied to both star formation and the dynamic redistribution of gas throughout the lifetime of a galaxy and is drastically modified by galaxy merger events @xcite .
theory predicts that major mergers encourage the formation of bars in the stellar and gas disks , which induce vigorous gas inflows as the gas looses angular momentum to the stellar component @xcite .
these inflows are thought to be responsible for fueling a massive central starburst and feeding agn and/or quasar activity @xcite . for a spiral galaxy with a preexisting metallicity gradient gas inflow flattens the gradient by diluting the higher abundance gas in the central regions with the lower abundance gas from the outer parts of the galaxy @xcite .
this flattening is compounded as the spiral arms are stretched by tidal effects @xcite .
here we investigate this process in nearby luminous infrared galaxies ( lirgs ) lirgs are excellent targets to inform the study of galaxy evolution .
although relatively rare , lirgs compose the larger part of the ir luminosity of the universe by z@xmath01 @xcite . as a group lirgs
span merger stages from isolated to post - merger and contain massive starbursts , agn , shocks and a variety of stellar population ages @xcite . in most cases
the starbursts and agn in lirgs are driven by ongoing mergers and the progression of these mergers is correlated with a rise in the ir luminosity , with massive merging lirgs eventually surpassing @xmath1 to become ultraluminous infrared galaxies ( ulirgs ) ( e.g. @xcite ) .
additionally , both merger activity and the incidence of u / lirgs increase at higher redshift @xcite .
it is clear that merging , ir - luminous systems played an important role during the peak of star formation and in the assembly of present - day massive galaxies .
this makes lirgs an ideal target to study the mixing and re - distribution of heavy elements in mergers to test the predictions of merger models .
@xcite found the first observational link between merging spiral galaxies and tidally induced gas flow in the depressed nuclear metallicities of close merging pairs .
other studies of mass - metallicity and luminosity - metallicity correlations find that merging systems tend to be underabundant for their size and brightness @xcite .
theoretical modeling of the metal distribution in merging systems reproduces the observed nuclear underabundance and predicts a flattening of a preexisting abundance gradient as the merger progresses @xcite .
recent multi - slit spectroscopy of hii regions in close - pair spiral galaxies confirmed that gas - phase metallicity gradients are indeed flatter in merging systems when compared to a control sample of isolated spirals @xcite .
in this paper we present an integral field spectroscopic ( ifs ) study of chemical abundances in 9 nearby lirgs .
our systems are a subset of a larger ifs sample of nearby u / lirgs : the wide field spectrograph ( wifes ) integral field unit ( ifu ) great observatory all - sky lirg survey ( goals ) sample ( wigs ) .
ifs data provides a wealth of useful information about nearby galaxies , allowing the observer to measure several aspects of the gas physics and stars over large areas . by using our ifs data to generate maps of emission line fluxes
, we can similarly create abundance maps and track metallicity gradient as a function of merger stage and sample later merger stages than previously considered .
we provide a summary of our sample , the observations and the data reduction in section 2 . in section 3
we discuss the analysis of our spectra and our metallicity determinations and calibrations .
we also present the resulting metallicity maps and gradients created from our ifu data in section 3 .
we analyze our observations and compare our measurements with previous observations of metallicities in merging systems in section 4 . in section 5
we discuss our observations in the context of recent merger models .
section 6 gives our conclusions . throughout this paper
we adopt the cosmological parameters h@xmath2=70.5 km s@xmath3mpc@xmath3 , @xmath4=0.73 , and @xmath5=0.27 , based on the five - year wmap results @xcite and consistent with the @xcite summary of the goals sample .
our targets are drawn from the great observatory all - sky lirg survey ( goals ) sample @xcite .
goals is a multi - wavelength survey of the brightest 60@xmath6 m extragalactic sources in the local universe ( @xmath7 ) with redshifts z @xmath8 0.088 and is a complete subset of the iras revised bright galaxy sample ( rbgs ) @xcite .
objects in goals cover the full range of nuclear spectral types and interaction stages and are excellent analogs for comparison with high - redshift galaxies . due to the location of wifes ,
our targets comprise a southern sample , with a declination limit of about + 15 .
the lirgs in our sample represent a variety of merger stages , nuclear separations , luminosities and physical processes .
the systems presented in this paper are those objects for which we could measure the extended gas - phase metallicities with sufficient spatial resolution , as the optical emission line gas in these systems is too compact and too extinguished to sufficiently measure a metallicity gradient with our ground - based data .
in addition , we rule out galaxies whose spectra are overwhelmingly dominated by non hii - region emission . in our sample
this is primarily widespread radiative shocks @xcite .
systems ruled out consist primarily of the post - merger objects in our sample .
for our sample , the data is insufficient for 10 of the 12 post - merger u / lirgs in our sample : follow - up on larger telescopes with ao and/or _ hubble space telescope _ would likely allow for a better investigation of post - merger targest .
we analyze a total of 11 galaxies in 9 systems , shown in table 1 .
we classify the merger stage of each system using the scheme adopted by @xcite , which is a slight modification of the merger stage classification outlined by @xcite based on the comparison of observations with the simulations of @xcite .
isolated systems ( iso ) show no signs of interaction and have no companion galaxies within 100 kpc in projected distance .
isolated galaxies may have extremely distant companions beyond 100 kpc , but they show no morphological indication of current or past major interaction . widely separated systems ( a ) show signs of an ongoing major merger including tidal tails and bridges , but companions are separated by a minimum projected distance of 10 kpc . closely interacting systems ( b ) show more advanced tidal structures and are separated by less than 10 kpc , but are not yet coalesced .
finally we group all coalesced mergers together ( cde ) , including diffuse , compact and post - merger systems .
all coalesced mergers are characterized by a single nucleus in ground - based optical and nir images , with an increasingly compact core as the final stages of the merger take place .
our sample has 2 isolated systems , 1 in stage a , 4 in stage b and 2 in stage cde. we compare our work to the previous samples of @xcite .
@xcite study a sample of 5 pairs of local luminous spiral galaxies from the optically selected galaxy pair sample of @xcite , each with separations of 15 - 25 kpc .
@xcite expand on this sample of interacting systems and also draw from the optically - selected @xcite catalog as well the infrared - selected samples of @xcite and @xcite for a total sample of 22 interacting galaxies in 9 pairs / groups , all classified in stage a according to the classification used in this paper .
@xcite also assemble a control sample of local , isolated spiral galaxies from the literature with properties similar to the spirals in their interacting galaxy sample .
our work overlaps with and extends the rupke et al .
sample to later merger stages .
the comparison samples consist primarily of gas - rich spirals which are accepted as the progenitors of lirgs @xcite .
the higher average l@xmath9 of our lirg sample is the result of intense star formation induced by the merger process - three of the interacting systems in the comparison sample are in fact lirgs themselves @xcite .
our systems are 1 to 2 magnitudes brighter in m@xmath10 , which generally correlates with stellar mass , as older late - type stars trace the majority of stellar mass with spectral emission peaking in the nir @xcite .
this enhancement is in part due to the merger process , as the total m@xmath10 of the later - stage , combined systems will reflect the total m@xmath10 of the progenitors . an additional enhancement in m@xmath10 also reflects a contribution from the intense , young starbursts in lirgs . both models and observations
show the possibility of a significant contribution to the nir flux ( from 30% to 60% ) from asymptotic and red giant branch stars @xcite .
thus we expect the m@xmath10 to be enhanced of order 1 to 2 magnitudes at approximately similar stellar mass indicating the masses of the systems in our sample are roughly consistent with the masses of the comparison sample .
this means m@xmath10 may not correlate well with stellar mass in lirgs , but the enhancement in m@xmath10 should still trace the merger process in the same way as l@xmath9 .
llcccccc f01053@xmath111746 & ic 1623 a / b & 11.71 & -24.61 & 63.4 & 34.8 & 1.05 & b + 08355@xmath114944 & - & 11.62 & -23.70 & ... & ... & 0.30 & cde + f10038@xmath113338 & eso 374-ig032 & 11.78 & ... & ... & ... & 0.76 & cde + f10257@xmath114339 & ngc 3256 & 11.64 & -24.78 & 83.2 & 48.7 & 3.31 & b + f13373@xmath120105 e / w & arp 240 & 11.62 & -24.96/-25.13 & 177.9/85.1 & 34.2/62.1 & 1.48/1.48 & a + f17222@xmath115953 & eso 138-g027 & 11.41 & -24.23 & 41.7 & 51.2 & 1.07 & iso + f18093@xmath115744 n / s & ic 4687/4689 & 11.62 & -24.47/-23.90 & 51.2/141.4 & 52.7/74.0 & 1.15/0.91 & b + f18341@xmath115732 & ic 4734 & 11.35 & -24.48 & 102.6 & 57.4 & 1.32 & iso + f19115@xmath112124 & eso 593-ig008 & 11.93 & -25.68 & ... & ... & 0.55 & b our data were taken using wifes at the mount stromlo and siding spring observatory ( sso ) 2.3 m telescope .
wifes is a new , dual beam , image - slicing ifu commissioned in may 2009 and described in detail by @xcite and @xcite .
our data consists of blue and red spectra with a resolution of r 3000 & r 7000 and wavelength coverage of @xmath03500 - 5800 & @xmath05500 - 7000
respectively .
the data were taken over 5 separate observing runs in july , august and september 2009 and in march and may 2010 . the data were reduced and flux calibrated using the wifes pipeline @xcite .
the pipeline uses iraf routines adapted from the gemini north s near - infrared integral field spectrometer ( nifs ) data reduction package .
a single wifes observation consists of 25 , 1wide , 38long slit spectra with contemporaneous sky spectra if the observation was taken in nod and shuffle mode .
all of the final combined data cubes in our sample consist of at least 2 observations of the same target and in some cases a mosaic of two or more pointings . where there were not sufficient observations to remove cosmic rays via median - combination
, cosmic ray removal was performed with the `` dcr '' routine @xcite . in this paper
we provide only a brief summary of the data reduction , the process is described in detail in @xcite .
individual observations are bias - subtracted using bias frames taken as near in time as possible to the observation frames to avoid temporal effects .
the observations were flat - fielded using quarts lamp flats and twilight sky flats .
the individual spectra are spatially calibrated using a thin wire combined with a continuum lamp .
cuar and near arc lamp spectra taken throughout each night are used to wavelength calibrate each observation .
the 25 resulting slitlet - spectra are then rectified into blue - arm and red - arm data cubes and sampled to a common wavelength scale .
telluric lines are then removed using observations of b - stars or featureless white dwarfs taken at similar airmass the same night as the data .
each data cube is then flux calibrated using observations of flux standard stars taken on the same night . if there were any non - photometric pointings of a target , these pointings were scaled and flux calibrated using photometric data of the same target .
we correct for the effects of atmospheric dispersion using the wifes pipeline .
each individual reduced data cube from observations of a single galaxy is then combined to form a final data cube sampled to a common spatial grid for analysis purposes .
the wifes detectors have 0.5pixels along the slit , so data taken in 2010 are binned on - chip by 2 pixels in the spatial direction in order to increase signal to noise and produce square spatial elements @xmath13 .
observations taken in 2009 are binned post - reduction to achieve similar results .
typical seeing achieved at sso during our observations is @xmath14 , with some variation , on par with the spaxel size for our data cubes .
we do not perform further binning ( e.g. voronoi tessellation ) prior to our data analysis .
combined data cubes were aligned astrometrically by comparing a pseudo r - band image generated using the red spectrum from each spectral pixel ( spaxel ) with either dss r - band or hst data where available . deprojected radii were calculated using inclination and position angle data from hyperleda @xcite , with the central spaxel defined using the wifes r - band images .
these data , as well as optical diameters are provided in table 1 . in the case of a few of our later - stage mergers we assume an uninclined disk when calculating deprojected radii - these have no pa or inclination given in table 1 .
this provides the most conservative estimate of a gradient in that system while avoiding ambiguities caused by the complex morphologies of late - stage major mergers .
we analyzed every spectrum using an automated fitting routine written in idl , uhspecfit , which is based on the code created for the work in @xcite and is also employed by and described with example fits in @xcite .
our routine fits and subtracts a stellar continuum from each spectrum using population synthesis models from @xcite and an idl routine which fits a linear combination of stellar templates to a galaxy spectrum using the method of @xcite .
lines in the resulting emission spectra are fit using a one or two - component gaussian , depending on the goodness of fit determined by the routine .
all of the emission lines are fit simultaneously using the same gaussian component or components .
both continuum and emission lines were fit using the mpfit package , which performs a least - squares analysis using the levenberg - marquardt algorithm @xcite .
all of the emission line fluxes used in this paper are subject to a minimum signal - to - noise cut of 5 .
to account for offset in flux between the red and the blue data cubes for each system , we compare the balmer decrement for h@xmath15 , which falls on the red arm , and h@xmath16 on the blue arm to the decrement between h@xmath16 and h@xmath17 which both fall on the blue arm .
the average scaling factors were generally within a few percent of unity , and were used to scale the red fluxes . due to the low flux and
thus limited spatial coverage in h@xmath17 , we did not use h@xmath16 vs. h@xmath17 to generate the extinction maps for our final analysis .
final extinction maps were created for each data cube using the h@xmath18/h@xmath19decrement and the extinction curve of @xcite .
these maps were then smoothed with a 2-d boxcar size of 3 spaxels to reduce noise in the maps .
the extinction map for each system is provided in the appendix .
we then deredden all of the measured emission line fluxes for a system using our derived extinction map .
we generate emission line ratios from the extinction - corrected line fluxes to diagnose the source of excitation for each spaxel in the systems we analyze .
emission line ratio maps are a convenient method of tracing the excitation mechanism within galaxies : region of high [ ] /h@xmath18or [ ] /h@xmath18are generally associated with non - hii region photoionization .
we subsequently classify each spectrum of every individual spaxel as hii - region like , composite or agn - like ( liner or seyfert ) using the [ ] /h@xmath18 , [ ] /h@xmath18and [ ] /h@xmath19ratios and the scheme of @xcite .
emission line ratio maps for each system are given in the appendix , and are shown without extinction correction applied .
to avoid contamination from non - hii region photoionization , we reject spaxels with emission line ratios that lie above the @xcite empirical pure star formatiion line in the [ ] /h@xmath18vs .
[ ] /h@xmath19diagram , as well as points that lie beyond the theoretical pure star formation line in the [ ] /h@xmath18vs .
[ ] /h@xmath19diagram of @xcite . to account for errors we allow points that fall within 0.1 dex to the right of the cutoffs , consistent with the work of @xcite .
we do not use the [ ] /h@xmath18vs .
[ ] /h@xmath19diagnostic due to the significantly lower s / n in [ ] for most of our systems .
the spectra that fall into the composite region and agn classes are contaminated primarily by shock excitation in our systems .
this is apparent in several of the emission line ratio maps , where off - nuclear regions show enhanced emission line ratios . in @xcite
we showed that these higher emission - line ratios are reproduced well by new slow - shock models and investigate the star - forming and shocked components separately using kinematic information from our wifes spectra . in the case of the galaxies iras f10257 - 4339 and iras f01053 - 1746 , we are able to separate the shocked component from the pure hii - region component kinematically using this method as outlined in @xcite . for these two systems
we apply a velocity dispersion cuts of 65 km / s and 90 km / s respectively as well as line - ratio classifications described above , leaving only pure hii - region emission line ratios for gas - phase metallicity calculation .
we calculate the oxygen abundance in each spaxel using several strong - line gas - phase metallicity calibrations .
we are unable to use the `` direct '' t@xmath20 method due to the low flux of the [ ] @xmath21 auroral line and the higher metallicities found in our objects .
we instead employ both diagnostics empirically calibrated against the t@xmath20 method and theoretical methods based on the strengths of other measurable emission - line fluxes , adopting the procedures outlined for each method in the summary of @xcite .
we measure relative abundances within a given diagnostic , avoiding the problems associated with absolute abundance calculations ( e.g. @xcite ) .
we employ the empirical calibration of @xcite ( pp04 ) , which uses the line ratios [ ] /h@xmath18and [ ] /h@xmath19 . using these line ratios
has the benefit of avoiding the necessity of extinction correction - pp04 developed their calibrations for use in higher redshift systems where proper measures of the reddening may not always be possible .
we also calculate metallicities with the theoretical method of @xcite ( kd02 ) , specifically we use the [ ] / [ ] line ratio which is insensitive to variations in ionization parameter .
previous work has shown that the kd02 diagnostic shows less rms dispersion than other diagnostics for this reason ( e.g. @xcite ) . when making comparisons with the @xcite and @xcite gradient studies , we convert abundances measured with the pp04 diagnostics into the kd02 diagnostic using the prescriptions of @xcite for consistency .
we also investigated the @xmath22 ( [ ] @xmath23+[]@xmath24)/h@xmath19 measures of abundance by employing the calibrations of @xcite ( kk04 ) , and @xcite ( m91 ) .
the kk04 method corrects for variations in ionization parameter , by using the [ ] / [ ] line ratio in an iterative fashion to calculate both an ionization parameter and a metallicity .
our results using the @xmath22 method are consistent with pp04 and kd02 with somewhat increased rms scatter and are thus excluded from further discussion in this paper .
we use the derived abundances to create maps of the metallicity for each system in our sample .
we then measure the metallicity gradient as a function of deprojected radius using the metallicity values calculated for each spaxel .
we calculate our gradients using an unweighted least - squares linear fit , with errors derived using monte - carlo methods , consistent with the work of @xcite .
only points which are below the hii - region cut are shown on the metallicity maps and gradient plots .
this is evident when comparing the line - ratio maps to the metallicity maps : regions of strong line ratios and/or low surface brightness in the line ratio maps are rejected and not shown on the metallicity maps .
iras f18341 - 5744 and iras
f13373 + 0105 w for instance are dominated by strong shock - like emission over large areas near the nuclei of those systems .
we calculate gradients in both dex / kpc and dex / r@xmath25 , where r@xmath25 is the effective optical radius , with the latter measure meant to account for differences in physical scale from system to system .
in general we are not able to extend our gradients to the same fractional isophotal radius as @xcite . in some of our systems
this is due to the spatially restricted , highly extinguished nature of the line - emitting gas , coupled in some cases with the limited total field of view of our wifes mosaics .
our monte - carlo error calculation takes this into account , and we also test the effect smaller spatial sampling may have on our gradient measures by taking systems with spatially well - sampled gradients recalculating new gradients at various limited spatial scales . the net effect when considering only the inner regions ( from 0.0 to 0.4 r / r@xmath25 )
is the calculation of a slightly steeper gradient than expected .
when other portions ( e.g. from 0.2 to 0.6 r / r@xmath25 ) of the gradient are sampled , the resulting gradient measured is within a few percent of the expected gradient .
these results indicate that our conclusions regarding gradient flattening remain relatively unaffected by spatial sampling .
we calculate metallicities for each spaxel using different methods in order to investigate any discrepancies in metallicity measurement introduced by errors in our extinction correction , flux measures and from ionization parameter variation .
due to the wide separation in wavelength of the diagnostic emission lines the kd02 calibration is most sensitive to errors in extinction correction while the pp04 diagnostic is essentially immune to reddening effects due to the close proximity of the line ratios employed .
on the other hand , the pp04 calibration should be the most sensitive to variations in ionization parameter , while the kd02 diagnostic is essentially insensitive to ionization parameter . in the case of very dusty lirgs ,
the pp04 diagnostic proves the most useful as it avoids using the very blue and thus very extinguished [ ] line .
indeed the pp04 diagnostic provides the least rms scatter from a straight line fit in our gradient measures , owing most likely to errors in our extinction correction which increase the scatter in o / h values calculated using kd02 .
the inability of the pp04 diagnostic to account for variations in ionization parameter , however , is a danger if the parameter varies in any systematic way , especially with radius .
we find a correlation between ionization parameter , as measured using the kk04 method , and the difference between metallicities measured using the pp04 diagnostic and the kd02 or kk04 diagnostic .
fortunately , the effect of varying ionization parameter is minor and does not affect the conclusions in this paper .
overall , the discrepancies between the metallicity calibrations in our analysis are consistent with the comprehensive study of @xcite .
it is also worthwhile to note that our gradients are derived using integrated spectra containing all of the light within each spaxel , including any diffuse emission not associated with hii regions .
this is in contrast to the targeted hii - region abundance gradients used in @xcite .
although we carefully remove non - photoionized spectra , we might still expect a diffuse contribution to line - ratios used in gradient calculation @xcite .
studies have shown , however , that abundances inferred from an integrated spectrum are consistent with hii - region abundances , regardless of the abundance calibration used @xcite .
@xcite in particular present highly resolved ifu spectroscopy of nearby ngc 628 and show that the abundance gradient derived from their total integrated spectra is very consistent with the gradient derived with spectra isolated from hii - regions .
@xcite and @xcite found a flattening in the metallicity gradients of widely separated pairs of galaxies .
their sample represents systems caught in the earliest stages of merging between first and second pericenter .
as the galaxies continue to merge , gas should continue to flow towards the individual galaxy nuclei , fueling the ongoing nuclear starburst and quenching star formation in the tidal remnants .
the nuclear metallicity would first see a depression due to the infalling gas , followed by a subsequent enrichment and further depression as the merger progresses to coalescence . the metallicity gradient is continuously flattened by the combined effects of infalling low - metal gas and tidal stretching of the spiral arms and outer portions of each system . in this section
we show that our results indicate the trend of flattening metallicity gradient may extend to later merger stages .
in the context of the @xcite merger scheme , the control and interacting samples of @xcite and @xcite are respectively isolated and widely - separated a stage systems , while the majority of the galaxies in our sample are close - pairs in the b stage , further along in the merger process .
our sample also has 3 late - stage coalesced systems ( cde ) .
it is not clear that all of the stage b lirgs in our sample will become true ulirgs given their morphology and l@xmath9 .
@xcite suggest that systems with mass ratios closer to unity are more likely to become true ulirgs during the later stages of the merger process
. this could account for systems such as iras f21330 - 3846 , which has a comparatively low total l@xmath9 despite the fact that it is a more advanced merger .
the gas fraction and dust mass of each system prior to the merger also strongly affect the total ir luminosity .
we plot metallicity gradient as a function of merger stage in figures 1 and 2 .
although we do not have a large control sample of isolated lirgs , the metallicity gradients of our 2 isolated lirg systems fall within the range of gradients of the control sample of local , isolated spirals from @xcite .
figure 1 shows that metallicity gradient flattens through each merger stage , with the three coalesced systems showing the flattest gradients .
the average gradient in our progressed mergers is similar to the average gradient measured in the widely separated pairs sample of @xcite .
figure 2 shows that our results are unchanged when when metallicity gradient is plotted as a function of optical radius and when it is plotted as a function of absolute deprojected radius in kpc .
interestingly , the galaxy iras f13373 + 0015 may be showing gradient - flattening effects due to both gas infall as well as the stretching of its spiral arms due to tidal effects from the merger as evidenced by an apparent kink in the metallicity gradient . unfortunately , the nuclear region is strongly contaminated by shock - like emission , so we excluded gas - phase metallicity measurements from our plot for these radii . if we ignore this contamination and calculate an abundance for the center - most region , the measured nuclear [ o / h ] does appear depressed with respect to the observed and calculated gradient , which would create a gradient that shows a slightly steeper portion between two flatter regions .
ccc m@xmath10 , pp04 , r@xmath25 & -0.49 & 0.2 + m@xmath10 , kd02 , r@xmath25 & -0.48 & 0.3 + l@xmath9 , pp04 , r@xmath25 & 0.56 & 0.02 + l@xmath9 , kd02 , r@xmath25 & 0.61 & 0.005 + m@xmath10 , pp04 , r@xmath26 & -0.60 & 0.008 + m@xmath10 , kd02 , r@xmath26 & -0.59 & 0.01 + l@xmath9 , pp04 , r@xmath26 & 0.65 & 0.001 + l@xmath9 , kd02 , r@xmath26 & 0.66 & 0.0006 we also plot both l@xmath9 and m@xmath27 , quantities which should be correlated with merger stage , as a function of metallicity gradient in figures 1 and 2 .
l@xmath9 traces star formation rate and is correlated with merger stage , while m@xmath27 is unaffected by the high extinctions encountered in lirgs and is a function of stellar mass ( though see sec .
values for l@xmath9 are taken from @xcite , which adjusts the original rbgs @xcite values to the 5-year wmap cosmological values .
m@xmath27 is derived using the distances from @xcite and the 2mass 20 mag / arcsec@xmath28 isophotal k - band magnitudes ( consistent with @xcite ) .
@xcite concluded there was no strong evidence of a correlation between gradient and either l@xmath9 or m@xmath27 , though they note that @xcite pointed to possible correlations when considering dex / kpc .
when our observations are combined with the work of @xcite , the case for a correlation between gradient and both l@xmath9 and m@xmath27 correlations is strengthened .
the spearman rank correlation coefficients indicate a moderately strong correlation in all cases , though l@xmath9 is more strongly correlated than m@xmath27 .
gradients calculated in dex / kpc are more tightly correlated than those measured in dex / r@xmath25 ( table 2 ) , consistent with @xcite .
increased scatter when considering optical radius may be due in part to the increasingly disturbed morphologies of merging systems . as a merger progresses
r@xmath25 represents a less consistently comparable value as tidal features and the coalescence of both galaxies alter this measure such that it is no longer consistent with r@xmath25 measured in pre - merger spiral galaxies . to further compare our sample to previous work
, we consider the nuclear abundances from our sample . in order to directly compare to @xcite
we also consider a luminosity - metallicity ( @xmath29 ) relation , rather than a mass - metallicity ( @xmath30 ) relation , as plotted in fig .
we consider the the near - ir l - z relation of @xcite for consistency with @xcite .
we adopt the same method as @xcite for nuclear abundance determination : we assign an abundance by extrapolating the calculated gradients to a fiducial radius of 0.1 r@xmath25 .
we also adopt the same offset in absolute metallicity with respect to the @xcite relation .
our nuclear metallicities are indeed lower than the expected l - z relation , again extending previous work on merging systems to higher nir luminosities .
this result is consistent with previous observations pairs @xcite as well as the merging galaxy models we consider in our discussion of gradient evolution with merger stage and properties .
our observations are a direct measure of the redistribution of gas within strongly interacting systems . the observational reality of metallicity gradients in non - interacting galaxies is well established @xcite .
observations of a change in an established abundance gradient and systematically lower nuclear metalllicities are clear indicators of gas flows and have been seen not only in strong interactions but also in bar structures ( e.g. @xcite ) . while @xcite demonstrated
the redistribution of gas traced by metallicity in barred galaxies , only recently have numerical simulations been used to address the detailed physical motions of gas and metals within interacting galaxies .
@xcite , @xcite and @xcite presented the first attempts at using hydrodynamical simulations to address the effects of major galaxy mergers on nuclear abundances and metallicity gradients .
the three separate investigations all involve mergers of massive spiral disk galaxies with 1:1 mass ratios . as noted by @xcite , these conditions are typical for the ulirg formation scenario and are thus a useful direct comparison .
the models of @xcite represent the simplest set of simulations analyzed for metallicity gradient evolution both in merger parameter space and in model sophistication .
their models use the methodology described in @xcite , but without any ongoing star formation and thus with no chemical enrichment .
this approach clearly captures the effect that low metallicity gas inflows will have on an evolving nuclear metallicity .
@xcite conclude that the nuclear metallicity can be substantially depressed by gas inflows and that metallicity gradient can be dramatically flattened following tidal tail formation .
they find , not surprisingly , that the depression in the nuclear metallicity is correlated with the gas mass that has migrated to the nuclear region .
@xcite employ simulations from the galmer database , described in @xcite .
their simulations include chemical enrichment due to ongoing star formation . as a result of including chemical enrichment , @xcite find that the nuclear metallicity changes non - monotonically as a function of the merger as chemical enrichment partially offsets the effect of low - metallicity gas inflows .
@xcite have studied merger induced nuclear metallicity evolution while considering star formation and chemical enrichment over a relatively wide range of merger parameter space .
the analysis in @xcite explicitly differentiated the effects of initial metallicity gradient re - distribution from ongoing metal enrichment on the nuclear metallicity evolution allowing for a basic template for merger nuclear metallicity evolution to be presented .
the simulations of @xcite included systems with varied initial mass and several merger orientations allowing for a more complete comparison to observations .
the interpretation of the data presented in this paper can benefit by considering a realistic theoretical model for the evolution of interacting galaxies . in principle , one could use numerical models to directly simulate the merging systems that have been presented in this paper . however , this would require accurately setting a large number of parameters for each merging system ( e.g. , the initial stellar mass of each galaxy , the initial gas fraction , the initial morphology , the merging orbital orientation , etc . ) .
fortunately there are certain aspects of major mergers , ( e.g. the formation of tidal features ) which are generically associated with strong tidal encounters and do not sensitively depend on the detailed characteristics ( e.g. the gas fraction ) of the merging galaxies . in the context of metallicity gradient evolution
, we can learn a lot about the expected evolution of a population of merging galaxies by exploring a limited number of carefully selected merger simulations .
our goal is to present a realistic theoretical model to augment the interpretation of our observational data . to achieve this
, we build a limited merger simulation suite that we can explore and understand in detail .
although this approach will leave certain areas of merger parameters space unexplored , we still expect the characteristics and tends present in the simulations to match that of the observations which allows us to physically probe the driving forces behind the observed evolving metallicity gradients presented in this paper .
we compare our observations to a suite of galaxy merger simulations ( @xcite ) carried out using gadget-2 @xcite .
our simulated merger suite consists of 32 simulations which are achieved by using 16 merger orientations ( orientations a - p in @xcite ) with 2 progenitor galaxies .
the two progenitor galaxies for our simulations both have equal total system mass , but with slightly different initial disk profiles .
both systems include a @xcite dark matter halo ( @xmath31 , @xmath32 kpc ) with rotationally supported exponential gas and stellar disks ( @xmath33 kpc , @xmath34 kpc ) .
the initial gas fraction of the system is 25% , which will decrease with time due to star formation .
a substantial discussion of the kind of models used for comparison in this paper as well as further analysis of theory derived from those models can be found in @xcite . here
, we briefly describe our methods for tracking the nuclear metallicity and metallicity gradient , which serves as the key component of our simulation analysis .
our simulations track a single metallicity scalar value that is allowed to increase as a result of chemical enrichment from star formation .
to compare our simulations to the observations , we focus only on the oxygen abundances , which we assume to be 30% of the total metal mass . because most oxygen is produced in type ii supernovae which return their mass and metals to the ism over relatively short ( i.e. @xmath35 year ) timescales , we assume instantaneous metal enrichment which we practically incorporate into our simulation by setting the rate of metal production proportional to the local star formation rate .
we impose metallicity properties on our galaxies at the beginning of the simulation such that they conform with observations of galaxies in the local universe .
namely , we initialize the nuclear metallicity of our galaxies onto the sloan digital sky survey derived mass - metallicity relation @xcite with the radial metallicity gradient following a decaying exponential profile with a characteristic length proportional to the disk scale length @xcite . using this model , the metallicity in our galaxy
is determined by the re - distribution of the initial metallicity gradient due to bulk movements of gas induced by the strong gravitational interaction and chemical enrichment associated with starburst driven star formation . at any time in the simulation
, the metallicity can be calculated for a region of space by finding the mass weighted average metallicity of all gas particles in the specified volume .
for example , we calculate the nuclear metallicity by finding the mass weighted average metallicity for all gas particles within a central 1kpc sphere of the galaxies center ( defined by the potential minimum ) . or , to determine the metallicity gradient , we calculate the average metallicity within a series of concentric spherical shells centered on the galaxies potential minimum , and find the best linear fit of the resulting radius - metallicity data points . for an instructive comparison to the observed systems presented in this paper , we track the nuclear separation , nuclear metallicity , metallicity gradient , and ir luminosity during the merger . using the data from all simulations , we
can then form an expected evolution and distribution of galaxy properties during the merger sequence .
the results are plotted in figures 4 , 5 & 6 . in figure 4 the merger model metallicity gradient
is plotted as a function of merger stage where the simulated mergers are classified using the same formalism as our observational analysis .
the solid black circles show the mean simulated gradient value of all merging systems at a given merger stage with the error bars representing the 1-@xmath36 standard deviations derived from the spread in gradient values for the simulated galaxies in that bin . by inspection
, we find that both the models and observations show a clear trend toward shallow metallicity gradients with increasing merger stage . in the models we find a strong change in the gradient that occurs immediately after first pericentric passage , when the formation of tidal features stretches the initial metallicity gradient and gas inflows begin to flood the nuclear region with low metallicity gas .
the gradient flattening continues as the merger progresses and by merger stage cde ( i.e. coalesced systems ) , both the models and observed systems have metallicity gradients consistent with being flat .
figure 5 shows the evolution of the metallicity gradient as a function of ir luminosity , projected separation , and nuclear metallicity .
l@xmath9 derived from the simulations is calculated based on the instantaneous star formation rate using the formulae of @xcite .
each panel shows the distribution of simulated galaxies as the background colored contours along with the observational data .
the colored contours represent the local density of simulated data points ( where each snapshot from each merger counts equally as one data point ) with the solid black lines denoting regions that contain 50 , 70 , 80 , and 90% of the simulated data points .
as discussed above , we find that the gradient becomes increasingly shallow as the merger progresses .
we also find the ir luminosity increases and the nuclear metallicity becomes slightly depressed .
this is not unexpected , as the same fundamental process is driving the evolution of all of these quantities .
strong tidal encounters that are associated with the merger sequence will form tidal features that will stretch or flatten the metallicity gradient .
the same tidal encounters are responsible for driving gas into the central region , which lowers the central metallicity and enhances starburst activity .
therefore , we expect that the association of flat metallicity gradients with enhanced ir luminosity and depressed nuclear metallicity is a robust byproduct of the merger process .
figures 5 and 6 show the metallicity gradient and nuclear metallicity against observationally projected separation .
again we find that the observations and merger models remain in general agreement , but the projected separation is not as useful for tracing merger progress .
we find the intermediate and late stage mergers are closely clustered in projected separation .
this is a result of the non - monotonic evolution of the nuclear separation with merger stage . statistically analyzing large surveys
should indeed show enhancements in the star formation rate , depressions in the nuclear metallicity , and a general flattening of the metallicity gradient that evolves with projected separation .
however , separating intermediate and late stage mergers can not effectively be carried out using only the projected separation information .
these conclusions will remain fundamentally unchanged if we applied a more extensive merger simulation suite based on the physical justification that the formation of tidal features and driving of gas to the galactic central region the physical engines behind the observed trends presented in this paper are events that are generically associated with the merger process .
the formation of tidal tails in particular is a well known result of a tidal encounter @xcite , which can depend on the merger orbital parameters ( e.g. * ? ? ?
if the tidal encounter is sufficiently weak because the perturbing galaxy is not massive enough or too distant , then we will not expect strong tidal features to form and we would not expect to see a dramatic flattening of the metallicity gradient .
however , in the regime of strong tidal encounters , we expect tidal tails to form and the metallicity gradient to flatten without any sensitive dependencies on , e.g. , varied gas fractions or the presence of a bulge .
the same arguments apply when considering the driving of gas into a galaxy s central region .
the tidal encounter sets up mis - aligned bars in the stellar and gaseous components allowing the gas to lose angular momentum and fall into the nuclear region @xcite .
additional considerations do exist , such as the stabilizing effect of stellar bulges against axis - symmetric perturbations during distant tidal encounters @xcite or the reduced torquing efficiency of very gas rich systems @xcite .
however , neither of these effects will prevent a strong gravitational interaction from driving gas inflows leading into final coalescence ( see , e.g. , * ? ? ?
* ; * ? ? ?
* for detailed justifications ) .
these comparisons provide a useful measure of the accuracy of merger models and strengthen the use of such models to make predictions about various aspects of galaxy mergers and their effect on galaxy evolution .
a more thorough discussion of the kind of models used for comparison in this paper as well as further analysis of theory derived from those models can be found in @xcite . when our new observations are combined with previous work
, there is also evidence of a correlation between metallicity gradient and both l@xmath9 and m@xmath10 , quantities that loosely trace merger progress .
assigning a precise merger stage is non - trivial . determining a rough estimate of the time elapsed since
first pericentric passage for instance is not possible when considering only the morphology and projected separation of a merging pair .
one solution could be the use of software like identikit @xcite to constrain merger parameters using observed velocity information .
a more reliable merger stage could be assigned once an accurate model is obtained , allowing for a more detailed study of the change in various physical quantities as a merger progresses .
there are other factors which hamper gradient measurements as mergers progress : @xcite note a marked increase in agn / liner influence on optical spectra as a function of l@xmath9 and merger stage , an effect also seen by @xcite .
this effect , coupled with the increasingly compact region of star formation as the remaining gas in a merger is concentrated entirely within smaller nuclear regions makes measurement of spatially resolved gas - phase metallicities difficult and prone to contamination from an agn .
in addition , increased contamination from radiative shocks would be expected in the most ir luminous systems , consistent with increasing gas inflows and the observed increase in supergalactic winds as a function of l@xmath9 @xcite . in our overall ifu sample of 27 u / lirgs , @xmath02/3 of the systems show a contribution from extended shock excitation , fractionally increasing as the merger progresses .
ifu observations of sufficiently high spatial and spectral resolution should be able to overcome these issues as long as there is of measurable star formation .
we have measured the gas - phase oxygen abundances in 9 nearby lirg systems using integral field spectroscopy .
we measured emission line fluxes for all of our ifs data cubes and used this information to create maps of excitation and metallicity measured with various strong - line metallicity diagnostics .
we calculated radial abundance gradients using our metallicity maps coupled with radii and deprojected radii calculated for each system .
our results agree with recent observations of gradient flattening in interacting pairs @xcite and indicate some evidence of flattening at later stages of the merger process , though further observations are needed to confirm this result .
our results are consistent with recent numerical models of metallicity and gas flows in major mergers @xcite .
we compare our observations with a new set of merger models used to track various quantities including metallicity gradient , nuclear metallicity , l@xmath9 and separation .
these new models track gradients and enrichment carefully throughout the merger process , including the effects of ongoing chemical enrichement for the first time .
our observations agree remarkably well with the models , adding observational support for theoretical predictions about gas flows in numerical merger models .
our comparison between model and observation also indicate the efficacy of metallicity gradient and l@xmath9 in tracking merger progress , while showing nuclear separation as less useful .
the latest stages of a major merger event prove the most difficult period to analyze and interpret using our observational methods and data . investigating the gradient evolution using numerical simulations of post - merger systems
requires fully cosmological simulations where accretion of prestine gas at late times can help reestablish an abundance gradient @xcite .
investigation of this phenomenon also requires a finer spatial scale and deeper observations than we have achieved with our current data set .
an increasing contribution from radiative shocks and agn to the emission line spectra of late stage mergers may inhibit these measurements , necessitating non emission - line abundance diagnostics .
we would like to thank the referee for his / her helpful comments , which helped clarify many points in this paper .
dopita , kewley and rich acknowledge arc support under discovery project dp0984657 . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . 98 natexlab#1#1 , m. 1977 , phd thesis , harvard university , cambridge , ma . ,
a. , garca - marn , m. , monreal - ibero , a. , colina , l. , arribas , s. , alfonso - garzn , j. , & labiano , a. 2009 , , 506 , 1541 , a. , garca - marn , m. , rodrguez zaurn , j. , monreal - ibero , a. , colina , l. , & arribas , s. 2010 , , 522 , a7 , a. , rieke , g. h. , rieke , m. j. , & scoville , n. z. 2002 , , 124 , 166 , l. , heckman , t. m. , & miley , g. k. 1990 , , 364 , 471 , l. , et al .
2009 , , 121 , 559 , h. 1966 , , 14 , 1 , j. e. 2004 , , 350 , 798 . 2011 , , 413 , 2860 , j. e. , & hernquist , l. 1992 , , 30 , 705 . 1996 , , 471 , 115 , j. e. , & hibbard , j. e. 2009 , , 137 , 3071 , e. j. , geller , m. j. , & kenyon , s. j. 2000 , , 530 , 660 , k. , fukugita , m. , ellis , r. s. , targett , t. a. , belli , s. , & kodama , t. 2009 , , 697 , 1369 , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 , i. v. , di matteo , p. , combes , f. , melchior , a .-
l . , & semelin , b. 2010 , , 518 , a61 + , j. , & giovanelli , r. 2002 , , 124 , 100 , l. , et al .
2009 , , 498 , 379 , e. , vogelsberger , m. , faucher - giguere , c .- a . , & hernquist , l. 2010 , , 725 , 353 , m. , hart , j. , mcgregor , p. , oates , p. , bloxham , g. , & jones , d. 2007 , , 310 , 255 , m. , et al .
2010 , , 95 , m. a. , pereira , m. , kewley , l. j. , & capaccioli , m. 2002 , , 143 , 47 , y. , & roy , j. 1999 , , 516 , 62 , s. l. , patton , d. r. , mendel , j. t. , & scudder , j. m. 2011 , , 1541 , s. l. , patton , d. r. , simard , l. , & mcconnachie , a. w. 2008 , , 135 , 1877 , i. f. , de carvalho , r. , contini , t. , & gal , r. r. 2004 , , 355 , 728 , d. , benz , w. , & kennicutt , r. 1994 , , 430 , l105 , r. m. , cervio , m. , martins , l. p. , leitherer , c. , & hauschildt , p. h. 2005 , , 357 , 945 , t. 2005 , , 360 , 322 , c. , & wilson , t. l. 1990 , , 229 , 431 , l. 1990 , , 356 , 359 , g. , et al . 2009 , , 180 , 225 , p. f. , cox , t. j. , younger , j. d. , & hernquist , l. 2009 , , 691 , 1168 , j. h. , et al .
2010 , , 715 , 572 , d. , ho , p. t. p. , yun , m. s. , matsushita , s. , peck , a. b. , & sakamoto , k. 2004 , , 616 , l63 , c. m. 2004 , phd thesis , university of hawaii , k. , sanders , d. b. , evans , a. s. , mazzarella , j. m. , armus , l. , & surace , j. a. 2009 , , 695 , l103 , g. , et al .
2003 , , 346 , 1055 , jr . ,
r. c. , et al .
2009 , , 703 , 1672 , l. j. , & dopita , m. a. 2002 , , 142 , 35 , l. j. , & ellison , s. l. 2008 , , 681 , 1183 , l. j. , geller , m. j. , & barton , e. j. 2006 , , 131 , 2004 , l. j. , groves , b. , kauffmann , g. , & heckman , t. 2006 , , 372 , 961 , l. j. , heisler , c. a. , dopita , m. a. , & lumsden , s. 2001 , , 132 , 37 , l. j. , rupke , d. , jabran zahid , h. , geller , m. j. , & barton , e. j. 2010 , , 721 , l48 , c. 2004 , , 347 , 740 , h. a. , & kewley , l. j. 2004 , , 617 , 240 , o. , & contini , t. 1999 , in iau symposium , vol . 193 , wolf - rayet phenomena in massive stars and starburst galaxies , ed .
k. a. van der hucht , g. koenigsberger , & p. r. j. eenens , 604+ , e. , et al .
2005 , , 632 , 169 , j. c. , salzer , j. j. , & melbourne , j. 2004 , , 616 , 752 , a. r. , dopita , m. a. , kewley , l. j. , harus , z. j. , nicholls , d. c. , scharwaechter , j. , & asiri , h. m. 2011 , mnras , .
r. , & esteban , c. 2010 , , 517 , a85 , c. , daddi , e. , renzini , a. , cimatti , a. , dickinson , m. , papovich , c. , pasquali , a. , & pirzkal , n. 2006 , , 652 , 85 , c. b. 2009 , in astronomical society of the pacific conference series , vol .
411 , astronomical society of the pacific conference series , ed .
d. a. bohlender , d. durand , & p. dowler , 251+ , p. , & roy , j. 1995 , , 445 , 161 , s. s. 1991 , , 380 , 140 , j. , et al .
2012 , arxiv e - prints , l. , lambas , d. g. , alonso , m. s. , & tissera , p. 2008
, , 386 , l82 , j. c. , & hernquist , l. 1996 , , 464 , 641 , a. , arribas , s. , colina , l. , rodriguez - zaurin , j. , alonso - herrero , a. , & garcia - marin , m. 2010 , arxiv e - prints , m. , di matteo , p. , lehnert , m. d. , combes , f. , & semelin , b. 2010 , , 518 , a56 + , m. , & lanon , a. 2002 , , 393 , 149 , j. , & kennicutt , jr .
, r. c. 2006 , , 164 , 81 . 2006 , , 651 , 155 , t. , jesseit , r. , & burkert , a. 2006 , , 372 , 839 , g. , petit , c. , prugniel , p. , theureau , g. , rousseau , j. , brouty , m. , dubois , p. , & cambrsy , l. 2003 , , 412 , 45 , m. s. , pogge , r. w. , & stanek , k. z. 2009 , , 695 , 259 , e. , & contini , t. 2009 , , 398 , 949 , m. , & pagel , b. e. j. 2004 , , 348 , l59 , w. 2004 , , 116 , 148 , j. a. , dopita , m. a. , kewley , l. j. , & rupke , d. s. n. 2010 , , 721 , 505 , j. a. , kewley , l. j. , & dopita , m. a. 2011 , arxiv e - prints , f. f. , daz , a. i. , kennicutt , r. c. , & snchez , s. f. 2011 , , 415 , 2439 , b. , & fischer , j. 2010 , , 712 , 318 , d. s. , veilleux , s. , & sanders , d. b. 2005 , , 160 , 115 , d. s. n. , kewley , l. j. , & barnes , j. e. 2010 , , 710 , l156 , d. s. n. , kewley , l. j. , & chien , l. 2010 , arxiv e - prints , d. s. n. , veilleux , s. , & baker , a. j. 2008 , , 674 , 172 , j. j. , lee , j. c. , melbourne , j. , hinz , j. l. , alonso - herrero , a. , & jangren , a. 2005 , , 624 , 661 , d. b. , mazzarella , j. m. , kim , d. , surace , j. a. , & soifer , b. t. 2003 , , 126 , 1607 , d. b. , & mirabel , i. f. 1996 , , 34 , 749 , d. b. , soifer , b. t. , elias , j. h. , madore , b. f. , matthews , k. , neugebauer , g. , & scoville , n. z. 1988 , , 325 , 74 , m. , michel - dansac , l. , & lambas , d. g. 2010 , , 514 , a57 + , v. 2005 , , 364 , 1105 , l. , norris , r. p. , chapman , j. m. , allen , d. a. , whiteoak , j. b. , & roy , a. l. 1992 , , 258 , 725 , j. a. , sanders , d. b. , & mazzarella , j. m. 2004 , , 127 , 3235 , a. 1977 , , 15 , 437 , a. , & toomre , j. 1972 , , 178 , 623 , p. , cox , t. j. , kewley , l. , & hernquist , l. 2011 , arxiv e - prints , c. a. , et al .
2004 , , 613 , 898 , p. , et al .
2008 , , 384 , 886 , l. , salzer , j. j. , haynes , m. p. , odonoghue , a. a. , & balonek , t. j. 1998 , , 116 , 2805 , s. , kim , d. , & sanders , d. b. 2002 , , 143 , 315 , s. , kim , d. , sanders , d. b. , mazzarella , j. m. , & soifer , b. t. 1995 , , 98 , 171 , m. b. , & edmunds , m. g. 1992 , , 259 , 121 , t. , kewley , l. j. , & sanders , d. b. 2010 , , 709 , 884 , h. j. , kewley , l. j. , & bresolin , f. 2011 , , 730 , 137 , d. , kennicutt , jr . , r. c. , & huchra , j. p. 1994
, , 420 , 87 lccccccc f01053@xmath111746 & @xmath37 & -@xmath38 & @xmath39 & -@xmath40 & @xmath41 & @xmath42 + 08355@xmath114944 & @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath47 & @xmath48 + f10038@xmath113338 & -@xmath49 & @xmath50 & -@xmath51 & @xmath52 & @xmath53 & @xmath54 + f10257@xmath114339 & -@xmath55 & -@xmath56 & -@xmath57 & -@xmath58 & @xmath59 & @xmath60 + f13373@xmath120105 w & -@xmath61 & -@xmath62 & -@xmath63 & -@xmath64 & @xmath65 & @xmath66 + f13373@xmath120105 e & @xmath67 & @xmath68 & @xmath69 & @xmath70 & @xmath71 & @xmath72 + f17222@xmath115953 & -@xmath73 & -@xmath74 & -@xmath75 & -@xmath76 & @xmath77 & @xmath78 + f18093@xmath115744 n & -@xmath79 & -@xmath80 & -@xmath81 & -@xmath82 & @xmath83 & @xmath84 + f18093@xmath115744 s & -@xmath85 & -@xmath86 & -@xmath87 & -@xmath88 & @xmath89 & @xmath90 + f18341@xmath115732 & -@xmath91 & -@xmath92 & -@xmath93 & -@xmath94 & @xmath95 & @xmath96 + f19115@xmath112124 & -@xmath97 & -@xmath98 & -@xmath99 & @xmath100 & @xmath101 & @xmath102
there are 2 isolated systems in our sample .
these systems have properties most consistent with the @xcite control sample , with well established metallicity gradients not yet disturbed by interactions .
both systems are lower luminosity lirgs , with l@xmath103l@xmath104 . _
iras f17222 - 5953 _ ( eso 138-g027).this system is more akin to a typical , non - interacting , strongly starbursting spiral galaxy in our sample .
it is in the vicinity of a few other galaxies , including the similarly bright eso 138-g026 , but even the nearest galaxy is at a projected distance of over 100 kpc and iras f17222 - 5953 is not yet interacting with any of these systems .
our ifu data applied to the [ ] /h@xmath18v [ ] /h@xmath19bpt diagram show a clean curve following the shape of the sdss sequence of local star forming galaxies ( e.g. @xcite ) , but with a slight apparent shift in total [ ] /h@xmath18 .
we interpret this shift as an overall nitrogen enhancement @xcite . _
iras f18341 - 5732 _ ( ic 4734 ) like iras f17222 - 5953 , our second isolated galaxy is in the vicinity of a few other luminous galaxies , but is not yet undergoing any interactions .
our nuclear spectra are dominated by a liner combined with an aging stellar population .
the strongest sights of star formation are where the bar in this galaxy meet the spiral arms evidenced by the two strong clumps of hii - region like points seen in the line - ratio and metallicity . h@xmath18imaging by @xcite shows further knots of star formation along the spiral arms and our nuclear spectra also show signs of an aging stellar population in the nucleus of ic 4734 . _
iras f13373 + 0105 w _ ( ngc 5257 ) although ngc 5257 and its equal mass partner ngc 5258 are still widely separated and retain much of their structure , they exhibit interaction features including tidal tails and a bridge between the two galaxies . the spiral arms show regions of strong star formation with some signatures of post - starburst populations away from the nucleus .
the nuclear regions of ngc 5257 are dominated by an older stellar population as the intense nuclear starburst associated with the later stages of major mergers has not yet begun . the extinction is higher in the nuclear regions and the measurable line - ratios place the nucleus in the composite region of the standard diagnostic diagrams , indicating possible liner activity - previous nuclear observations and integrated spectrophotometry are consistent with the higher nuclear extinction and overall line ratios we observe @xcite . _ iras f13373 + 0105 e _ ( ngc 5258 ) ngc 5258 is a near - twin to ngc 5257 in mass and luminosity . our spectra indicate higher extinction and a flatter gradient , though the extent of measurable hii - region metallicities is smaller than in ngc 5257 .
again similar to ngc 5257 the nuclear region of ngc 5258 is dominated by older stars and very little line - emission and there is some evidence of younger post - starburst populations away from the nucleus .
the strongest line emission is associated with the knot of star formation to the southwest of the nuclear region .
our sample has several closely interacting systems , we calculate gradients for 5 of them in this paper .
these systems span a range of l@xmath9 from typical lirg luminosities to nearly ulirg .
iras f01053 - 1746 , iras f10257 - 4339 and iras f19115 - 2124 are the most indicative of this class , while iras f18093 - 5744 n / s is a triple system with two widely separated spirals and a third , less massive closely interacting component .
although iras f01053 - 1746 and iras f10257 - 4339 are both composed of two individual galaxies , they have progressed along the merger sequence to the point where we are unable to effectively assign gas metallicities to either of the galaxies .
we thus treat both iras f01053 - 1746 and iras f10257 - 4339 as individual systems with the brightest optical nucleus defined as the zero point for a metallicity gradient . _
iras f01053 - 1746 _ ( ic 1623 ) this system contains two very closely interacting spiral galaxies .
it is kinematically very complex and exhibits evidence of widespread radiative shocks , which are analyzed in detail in @xcite .
the deprojected radii are plotted using the hyperleda values for the western , less obscured galaxy ic 1623 a. the eastern system is very intensely star - forming as seen in the infrared ( e.g. @xcite ) , but is so enshrouded in dust that our optical spectra do not trace star - forming gas . _
iras f10257 - 4339 _ ( ngc 3256 )
this advanced merger is quite nearby and well studied . as with iras f01053 - 1746 ,
one of the galaxies in iras f10257 - 4339 is very extinguished .
the second system and its nucleus are revealed at longer wavelengths , south of the main optical nucleus @xcite .
deprojected radii are calculated using the main optical nucleus .
the gas and tidal tails in this system extend quite far ( e.g. @xcite ) our ifu mosaic covers only the central 6 kpc , though this appears to be the physical extent of most of the ongoing star formation in this system .
ngc 3256 also shows evidence for widespread shocks in our data and is discussed alongside ic 1623 in @xcite .
_ iras f18093 - 5744 n _
( ic 4687 ) we classify this galaxy as a close merger ; in fact it is a member of a triplet .
ic 4687 is undergoing a close merger with the less massive starburst ic 4686 , classed as a wolf - rayet galaxy by @xcite .
ic 4687 itself is a is also in a wide merger with the equally massive spiral ic 4689 .
the archived hubble space telescope images of ic 4687 show a complex morphology tangled up with ic 4686 : gas and dust from ic 4687 appear to be obscuring the less massive system .
our ifu data covers the entirety of ic 4686/4687 and the metallicities we measure are consistent with the expected metallicties in the outskirts of ic 4687 as extrapolated from the gradient we present in this paper as well as a low - metallicity , flattened gradient in ic 4686 .
the kinematic information from our data also indicates that we are indeed seeing gas from both systems .
_ iras f18093 - 5744 s _ ( ic 4689 ) this spiral galaxy is slightly less massive and luminous than ic 4687 and less well - observed , though it is still intensely star forming @xcite .
it is less morphologically disturbed than the other two interacting galaxies ic 4686/4687 , though its gradient is quite flattened already according to our observations .
although it is widely separated from ic 4687 , we include it as part of the closely interacting system of iras f18093 - 5744 .
_ iras f19115 - 2124 _ ( eso 593-ig 008 ) @xcite analyze the morphology of this close pair in detail using adaptive optics k - band images . because of the complex structure we adopt the projected radii for gradient calculations and make no attempt to separate the pair into two separate gradients .
given the flat values for metallicity seen , however , adapting an inclination and/or two disks would only lead to an even shallower calculated slope in the case of this system .
our metallicity sample has two systems which are classified as coalesced mergers ( cde ) .
the most advanced mergers have complex orientations and morphologies which present a difficulty when calculating deprojected radii . as a conservative estimate we simply use the projected radii in our gradient calculations . _
iras 08355 - 4944 _ while hubble space telescope i - band images show remnant tidal tails extending nearly 20 kpc , only the central 5 kpc or so appears to harbor the intense ongoing star formation in this system .
our spectra are dominated by hii regions , with evidence in some regions of a blue - shifted component with low - velocity shock - dominated line ratios which could be associated with a galactic wind . _
iras f10038 - 3338 _ ( eso 374-ig032 ) this post - merger exhibits significant ongoing star formation in its southwestern tidal arm unlike the two other coalesced systems in our sample .
the total line emission in this region is much weaker than in the nucleus , inducing a large uncertainty in the extinction map and [ oii ] lines , creating the discrepant values seen in the metallicity gradient plots for this galaxy .
either gradient , however , is consistent with the overall flattening trend seen in all of our systems .
this system hosts an oh megamaser and has soft x - ray emission , all consistent with the advanced stage of merging and increasingly intense nuclear starburst @xcite .
our spectra also show evidence for extended off - nuclear shock emission dominating in areas where there is little to no evidence of ongoing star formation , consistent with the ifu observations of @xcite .
|
we present for the first time metallicity maps generated using data from the wide field spectrograph ( wifes ) on the anu 2.3 m of 9 luminous infrared galaxies ( lirgs ) and discuss the abundance gradients and distribution of metals in these systems .
we have carried out optical integral field spectroscopy ( ifs ) of several several lirgs in various merger phases to investigate the merger process . in a major merger of two spiral galaxies with preexisting disk abundance gradients
, the changing distribution of metals can be used as a tracer of gas flows in the merging system as low metallicity gas is transported from the outskirts of each galaxy to their nuclei .
we employ this fact to probe merger properties by using the emission lines in our ifs data to calculate the gas - phase metallicity in each system .
we create abundance maps and subsequently derive a metallicity gradient from each map .
we compare our measured gradients to merger stage as well as several possible tracers of merger progress and observed nuclear abundances .
we discuss our work in the context of previous abundance gradient observations and compare our results to new galaxy merger models which trace metallicity gradient .
our results agree with the observed flattening of metallicity gradients as a merger progresses .
we compare our results with new theoretical predictions that include chemical enrichment .
our data show remarkable agreement with these simulations .
|
in the presence of a large perpendicular magnetic field , coulomb interactions between electrons confined to a two - dimensional plane compete with disorder in determining the system s physical properties . in the clean limit interactions dominate and give rise to a wealth of exotic collective states , including both compressible and incompressible quantum liquids , various solid phases , and quantum nematic liquid crystals @xcite
. moreover , the relatively small spin zeeman energy in typical two - dimensional electron systems ( 2des ) can be so overwhelmed by these interactions that ground state spin configurations which defy simple pauli counting rules can be stabilized @xcite .
these unusual spin configurations have been detected and studied experimentally by various means , including conventional electrical transport , photo - luminescence and inelastic light scattering , nuclear magnetic resonance , etc . @xcite .
the effects of coulomb interactions on 2d electron systems at high magnetic field are most dramatically illustrated by the numerous fractional quantized hall effect states @xcite and the emergent gapless metallic phases of composite fermions ( cfs ) @xcite , all of which exist at specific fractional fillings of the landau levels created by the magnetic field .
nevertheless , these exotic phenomena actually represent relatively subtle variations in the strong electronic correlations that exist throughout the high field regime .
for example , experiments @xcite have revealed a suppression of the zero bias conductance for electrons tunneling perpendicularly into the 2des over a wide range of high magnetic fields .
this suppression of the tunneling conductance is observed regardless of whether the 2des is in a thermodynamically gapped or gapless phase , and extends over a range of voltages about the 2des fermi level @xcite .
these observations are by now well understood @xcite to reflect a coulomb - interaction - induced pseudo - gap in the tunneling density of states .
this pseudo - gap arises from the inability of the correlated 2des to rapidly relax the charge density defect created by the rapid injection of an electron by tunneling at low energies .
put another way , at low energies the ( @xmath0 + 1)-particle states created by tunneling are essentially orthogonal to the ( @xmath0 + 1)-particle eigenstates of the 2des . in this paper
we report experimental observations which indicate that the tunneling coulomb gap is sensitive to the spin configuration of the 2des , a possibility not considered in prior theoretical work @xcite .
our experiments consist of measurements of the current - voltage characteristics for electrons tunneling between parallel 2d electron systems in semiconductor double quantum wells in the presence of magnetic fields both perpendicular and parallel to the 2d planes .
we focus on the case of the half - filled lowest landau level ( in each 2d layer ) , both for simplicity and the existence of a well - developed theory @xcite for this situation .
we find that the tunneling coulomb gap increases and then saturates as the parallel field is applied in a manner consistent with the known transition from partial to complete spin polarization of the 2des .
the samples used in this experiment are gaas - based semiconductor heterostructures grown by molecular beam epitaxy .
each contains two 18 nm gaas quantum wells separated by a 10 nm al@xmath1ga@xmath2as barrier layer .
silicon delta - doping layers , positioned in the thick al@xmath3ga@xmath4as layers above and below the double quantum well , populate the lowest subband of each quantum well with a 2des of nominal density @xmath5 @xmath6 and low temperature mobility @xmath7 @xmath8/vs .
standard lithographic methods are used to pattern the 2des into a mesa structure ; for the data presented here this mesa consists of a 250 @xmath9 m square with 60 @xmath9m - wide arms extending from each side .
ohmic contacts ( niauge ) to the individual 2d layers @xcite are positioned at the ends of these arms .
these separate layer contacts enable direct measurements of the interlayer tunneling characteristics of the sample via conventional dc and low frequency ac methods .
the individual layer densities @xmath10 and @xmath11 in the central mesa region are controlled by electrostatic gates deposited on the top and back side of the thinned sample .
we focus here on the balanced case , @xmath12 , with @xmath13 varied from about @xmath14 to @xmath15 @xmath6 . via _ in situ _ tilting of the sample relative to an applied magnetic field @xmath16 , field components both perpendicular ( @xmath17 ) and parallel ( @xmath18 ) to the 2des plane could be applied . except where otherwise noted , the tunneling data presented here was obtained with @xmath17 adjusted to render the landau level filling factor @xmath19 in each layer @xcite .
the density @xmath13 was kept high enough that condensation into the total filling factor @xmath20 bilayer excitonic phase did not occur .
( color online ) ( a ) typical tunneling @xmath21 characteristic at @xmath22 .
data taken at @xmath23 mk with @xmath17=4.09 t and @xmath18=0 . @xmath24 and @xmath25 denote location of maximum tunneling , while @xmath26 is voltage where tunneling current has reached 0.02@xmath24 .
( b ) density dependence of @xmath25 at @xmath18=0 .
solid line is an unweighted least - squares fit of @xmath25 versus @xmath27 ; the fit extrapolates to @xmath28 mv at @xmath29.,width=297 ]
figure 1(a ) shows a typical tunneling current - voltage @xmath21 characteristic at @xmath22 . a pronounced suppression of the tunneling current @xmath30 around zero interlayer voltage @xmath31 is readily apparent .
this feature , a coulomb pseudogap , is the main focus of this paper .
away from @xmath32 the tunneling current rises and forms a broad peak . both the coulomb gap around @xmath32 and the broad peak at finite @xmath31 are due to the strong electron - electron interactions which dominate the physics of landau quantized 2d electron systems .
these features have received substantial theoretical scrutiny @xcite and are qualitatively well - understood .
we emphasize that these theoretical studies assumed the electron spins were fully polarized by the magnetic field @xcite .
figure 1(b ) presents the density dependence of the voltage @xmath25 at which the @xmath22 tunneling current reaches its maximum value .
as noted previously @xcite , if coulomb interactions within each 2des dominate the tunneling @xmath21 curve , one expects @xmath25 ( at fixed filling factor ) to be proportional to @xmath27 . as fig .
1(b ) shows , @xmath25 is linearly dependent on @xmath27 , but extrapolates to a @xmath33 value @xmath34 in the @xmath29 limit .
this negative value , @xmath28 mv for the data in fig .
1(b ) , represents the excitonic attraction , in the final state , of a tunneled electron and the hole it leaves behind . in agreement with earlier work , @xmath35 , with @xmath36 nm , the center - to - center spacing between the quantum wells @xcite . for a single layer 2des , adding an in - plane magnetic field @xmath18 to a pre - existing perpendicular field @xmath17 increases the spin zeeman energy ( @xmath37 , with @xmath38 the lande @xmath38-factor and @xmath39 the bohr magneton ) and couples to the finite thickness of the 2des thereby inducing mixing between landau levels and the subbands of the confinement potential . for electrons tunneling between two parallel 2dess separated by a distance
@xmath40 there is an additional effect arising from the lorentz force associated with the in - plane field ; a tunneling electron acquires a `` momentum boost '' @xmath41 , with @xmath42 . at @xmath17=0
this momentum boost can completely suppress the zero bias tunneling conductance if @xmath43 , with @xmath44 the fermi wavevector @xcite . at high @xmath17 , with the fermi level in the lowest landau level
, the momentum boost leads to an exponential suppression of the tunneling matrix element @xmath45 . ignoring all other effects of the in - plane field
, the tunneling current is expected to follow @xmath46 , with @xmath47 the magnetic length @xcite .
this @xmath18-induced suppression of the tunneling current at high @xmath17 is clearly displayed in fig .
the figure plots the maximum tunneling current @xmath24 in the broad peak above the coulomb gap , normalized by its value at @xmath18=0 , versus @xmath48 .
data for four different 2des densities , ranging from @xmath49 to @xmath50 @xmath6 are shown ; in all cases @xmath17 is set to produce @xmath22 in each 2des layer .
plotted in this way the various data sets collapse onto a single curve and the agreement with theory ( dashed line ) is excellent .
( color online ) effect of an in - plane magnetic field on the tunneling @xmath21 curve at @xmath22 .
( a ) maximum tunneling current versus @xmath51 at various densities .
dashed line is theory prediction described in text .
( b ) normalized @xmath52 tunneling @xmath21 curves at @xmath53 @xmath6 for @xmath18=0 and @xmath18=2.9 t. upward arrows near lower left show change in @xmath26 induced by the in - plane field .
data taken at @xmath23 mk.,width=297 ] if the momentum - boost was the only effect of the in - plane magnetic field then , aside from an amplitude scale factor , the basic @xmath21 curve would be independent of @xmath18 .
we find that this is clearly @xmath54 the case .
figure 2(b ) presents a typical example of how the normalized tunneling @xmath21 curve responds to the application of an in - plane field .
( for these @xmath52 , @xmath53 @xmath6 data , the perpendicular field is fixed at @xmath17= 3.51 t. ) roughly speaking , the entire @xmath21 curve expands to higher voltages as @xmath18 is applied . figures 3(a ) and 3(b ) demonstrate that the @xmath18-induced expansion of the @xmath21 curve is not a simple rigid shift to higher voltages . in fig .
3(a ) the voltage location , @xmath25 , of the peak tunneling current is shown to increase linearly with @xmath16 as the in - plane field is applied . over
the density range studied , we find that the rate of this increase is essentially constant .
in contrast , the voltage width of the region of strongly suppressed tunneling , @xmath55 the coulomb gap , behaves quite differently . to illustrate this , fig .
3(b ) plots @xmath26 , the voltage at which the tunnel current first reaches 2 percent of its subsequent maximum value , versus @xmath16 .
these data obtain from the same set of tunneling @xmath21 curves used to create fig .
in general , @xmath26 at first rises swiftly as @xmath18 is applied , but then quickly levels off . for the lowest density data ( solid red dots ; @xmath56 @xmath6 ) ,
the initial increase of @xmath26 is almost two - fold .
as the 2d density is increased , the net increase in @xmath26 declines until , at @xmath57 @xmath6 , little effect remains .
we emphasize that the precise definition of @xmath26 is not important here , so long as it corresponds to a voltage where the tunneling current is a small fraction of @xmath24 .
( color online ) effect of in - plane magnetic field on @xmath22 tunneling critical points , @xmath25 and @xmath26 , at three different densities , plotted versus total magnetic field @xmath16 . for each density ,
the perpendicular field @xmath17 is fixed while the in - plane field @xmath18 is varied .
the left - most data point of each set corresponds to @xmath18=0 ( tilt angle @xmath58 ) .
the arrows indicate our assignment of the `` knee '' in the @xmath26 @xmath59 @xmath16 data.,width=297 ] the relatively sharp `` knee '' in the dependence of @xmath26 on @xmath16 suggests that a qualtitative transition in the nature of the 2des at @xmath60 occurs as the in - plane field is applied . moreover , the transition is most prominent at low density , disappearing almost entirely at the highest densities investigated here .
these observations are at least consistent with a change in the spin configuration of the 2des , a possibility which we now consider .
there is by now copious evidence that the ground state of the 2des at @xmath60 is not fully spin polarized at low density @xcite .
for example , tracy _ et al . _
@xcite , using resistively - detected nuclear magnetic resonance ( rdnmr ) methods , observed a relatively sudden increase in the nuclear spin lattice relaxation time @xmath61 as the density of a 2des , maintained at @xmath60 , was increased .
this observation was readily explained by the disappearance of both up and down electronic spin states at the fermi level as the 2des transitions from partially to completely spin polarized .
similarly , li _ et al . _
@xcite , using both conventional transport and resistively - detected nuclear spin relaxation methods , showed that the @xmath22 spin transition could be driven either by increasing the 2des density or by adding an in - plane magnetic field .
the composite fermion ( cf ) theory of the @xmath22 state provides a simple way to understand this transition . in this theory the 2des at @xmath52 resembles a fermi gas at zero magnetic field , only the constituent fermions are cfs , electrons with two fictitious magnetic flux quanta attached .
in general , both up and down spin cfs are present at the fermi level , their relative populations determined by the comparison between the spin zeeman energy @xmath62 and the cf fermi energy @xmath63 .
the zeeman energy is presumed to be the same as for ordinary electrons , @xmath64 , with the @xmath38-factor that appropriate to the host crystal band structure ( @xmath65 for electrons in gaas ) .
in contrast , the cf fermi energy is unrelated to the band structure and , ideally , is determined only by electron - electron interactions .
dimensional arguments alone then require @xmath66 , with @xmath67 a universal , though only approximately known , constant @xcite . since @xmath68 while @xmath69 , it is clear that @xmath62 can be made to exceed @xmath63 either by increasing the density or by adding an in - plane magnetic field .
defining the normalized zeeman energy @xmath70 , the transition from partial to complete spin polarization at @xmath52 should occur at @xmath71 , independent of electron density .
various non - idealities of the 2des , @xmath72 the inescapable finite thickness of the 2d layer and interaction - driven landau level mixing effects , can be expected to disrupt this universality and produce sample - to - sample variations in the observed critical zeeman energy @xmath73 @xcite .
( color online ) solid red circles : critical normalized zeeman energy determined from `` knee '' in coulomb gap @xmath26 data @xmath59 2des density @xmath13 .
open circle : cf spin transition point found by tracy _
dashed line : theoretical estimate of park and jain @xcite.,width=297 ] using the `` knee '' observed in the @xmath26 @xmath59 @xmath16 data shown in fig .
3(b ) as an indicator of the cf spin transition , the deduced critical zeeman energy @xmath73 is plotted @xmath59 2des density @xmath13 at @xmath52 in fig .
4 . also shown is the @xmath60 spin transition point found by tracy _
@xcite in their rdnmr experiment .
while the agreement between the present experiment and this earlier one strongly supports the identification of the tunneling critical point with the cf spin transition , the near quantitative agreement may be fortuitous . both similar @xcite and somewhat larger @xcite values of @xmath73
have been observed in other experiments .
the dashed line is the theoretical estimate of park and jain @xcite . to support the identification of the `` knee '' in the @xmath26 @xmath59 @xmath16 data shown in fig .
3(b ) with the transition from partial to complete spin polarization in the 2des , the underlying physical mechanism relating these properties of the 2des needs to be determined .
as is already well - understood @xcite , the coulomb gap in the tunneling density of states arises from the electronic correlations created by coulomb interactions in the landau quantized 2des .
these correlations are obviously sensitive to the spin polarization of the 2des : parallel spin electrons avoid one another more strongly than anti - parallel electrons owing to the pauli principle .
a partially spin polarized 2des is in this sense less strongly correlated than a fully polarized one .
a more detailed picture emerges from consideration of the wavevector - dependent conductivity , @xmath74 , of the 2des .
the coulomb gap itself reflects the inability of the 2des to rapidly relax the charge defects created by tunneling .
an electron tunneling into ( or out of ) a landau quantized 2des creates a localized excess ( or deficit ) of charge . the rate at which these defects can relax to equilibrium is determined by the conductivity , with higher conductivity producing a smaller coulomb gap and lower conductivity a larger one .
this connection between the magnitude of the coulomb gap and the conductivity was made concrete by he , platzman , and halperin ( hph ) @xcite .
hence , we are led to ask how the spin polarization affects the conductivity . since at voltages of order
@xmath26 the charge defects created by tunneling are localized on the scale of several magnetic lengths @xmath75 , the disorder - dominated @xmath76=0 conductivity measured in an ordinary electrical transport measurement is not the relevant quantity .
instead , it is @xmath74 at @xmath77 that counts . in their study of the cf metal at @xmath60 , halperin ,
lee , and read @xcite calculated the conductivity for the fully spin polarized case : @xmath78 , where @xmath79 is the fermi wavevector of the spin polarized cf fermi sea .
this intriguing @xmath76-linear conductivity was confirmed experimentally in high frequency surface acoustic wave experiments by willett _
_ @xcite .
if instead of completely spin polarized , the cf fermi sea were completely unpolarized , the fermi wavevector is reduced to @xmath80 and the conductivity @xmath74 thereby _ increased _ @xcite by a factor of @xmath81 . between these extremes , where the 2des is partially spin polarized , we can expect the conductivity to steadily decrease with increasing spin polarization .
given the connection between conductivity and the width of the coulomb gap established by hph @xcite , this dependence qualitatively explains the behavior of @xmath26 @xmath59 @xmath16 shown in fig .
3(b ) and thus fortifies our association of the `` knee '' in the coulomb gap data with the transition to complete spin polarization .
there are , of course , issues that require further study . for example , as fig .
3(a ) demonstrates , @xmath25 , the voltage at which the maximum in the tunneling current occurs , does not exhibit any evidence of a transition similar to that displayed by @xmath26 .
we find this behavior unsurprising .
the voltage @xmath25 is the cross - over point between two regions of suppressed tunneling . at low energies
the slow dynamics of the correlated 2des produce the coulomb gap at the fermi level . at high energies , in the single particle cyclotron gap between landau levels
, there are no final states in which to tunnel . in between these extremes , at energies of order the net coulomb broadening @xmath82 of the landau level , the tunneling current has a maximum .
since @xmath83 depends only on the mean spacing @xmath84 between electrons ( @xmath85 at @xmath22 ) , no particular sensitivity to the spin state of these electrons is expected .
the modest @xcite linear increase of @xmath25 with @xmath16 shown in fig .
3(a ) nevertheless remains a puzzle .
one possibility is that it is due to a stiffening of the effective coulomb interaction between electrons due to a ` squeezing ' of their wavefunctions .
such squeezing can result from the subband / landau level mixings induced by the in - plane magnetic field . a more interesting question concerns the role of interlayer coulomb interactions .
that such interactions exist is made evident by the linear extrapolation to zero density of the @xmath25 @xmath59 @xmath27 data shown in fig .
the substantial negative intercept , @xmath86 mv , is believed to be due to a final - state excitonic attraction between a tunneled electron and the hole it leaves behind . at low energies , _
e.g. _ inside the coulomb gap , the charge defects created by tunneling are larger in lateral extent @xcite than they are at @xmath25 . while this reduces the magnitude of the excitonic attraction , some effect undoubtedly remains .
hph considered interlayer coulomb interactions and concluded that in addition to an excitonic ` down - shift ' of the peak in the tunnel current , the detailed shape of the @xmath21 curve deep in the gap was altered @xcite . to what extent these interactions change the way spin influences the coulomb gap is not presently known .
a more dramatic interlayer interaction effect is the condensation of the double layer 2des into an intrinsically bilayer collective state possessing spontaneous interlayer phase coherence @xcite .
this state , with its remarkable superfluid - like properties , occurs at the same total filling factor as studied here ( _ i.e. _ @xmath20 ) and _ is _ observable in the present samples , but only at densities lower than those explored here . while some hypothetical precursors of this transition might be influencing the present tunneling results , two considerations suggest otherwise
first , we have observed that a very similar increase of the coulomb gap is induced by an in - plane magnetic field when each 2des is at filling factor @xmath87 or @xmath88 ( @xmath55 at total filling factors @xmath89 and 1.1 , respectively ) .
thus , unlike the coherent bilayer @xmath90 state , the effects reported here exist over a relatively wide range of filling factors .
second , we have found that the in - plane field - induced increase in the coulomb gap persists to at least @xmath91 k. this is well above the temperature at which the bilayer @xmath90 interlayer coherent state has collapsed @xcite . in conclusion , we find that the coulomb gap which dominates the low energy tunneling between two parallel 2dess at high magnetic field can be acutely sensitive to the application of an in - plane magnetic field .
this sensitivity appears to coincide with the transition from partial to complete spin polarization which the in - plane field drives by increasing the spin zeeman energy .
the wavevector - dependent conductivity of the 2des provides a plausible link between the coulomb gap and the system s spin configuration .
the coulomb gap thus appears to offer a new perspective on the spin state of highly correlated 2d electrons .
it is a pleasure to acknowledge helpful discussions with h.r .
krishnamurthy , p.a .
lee , a.h .
macdonald , and especially s.h .
simon and a. stern .
this work was supported in part by the institute for quantum information and matter , an nsf physics frontiers center with support of the gordon and betty moore foundation through grant no .
the work at princeton university was funded by the gordon and betty moore foundation through grant gbmf 4420 , and by the national science foundation mrsec grant 1420541 . for a review ,
see _ perspectives on quantum hall effects _ , edited by s. das sarma and a. pinczuk ( wiley , new york , 1997 ) .
e. fradkin , s. a. kivelson , m. j. lawler , j. p. eisenstein , and a. p. mackenzie , annu .
rev . condens .
matter phys .
* 1 * , 153 ( 2010 ) .
this was first suggested by b. i. halperin in helv .
acta * 57 * , 75 ( 1983 ) . for early examples ,
for the case of the half - filled lowest landau level , of primary interest here , see refs .
d. c. tsui , h. l. stormer , and a. c. gossard , phys .
lett . * 48 * , 1559 ( 1982 ) .
j. k. jain , phys .
lett . * 63 * , 199 ( 1989 ) .
see also j. k. jain and r. k. kamila , in _
composite fermions _ , edited by o. heinonen ( world scientific , singapore , 1998 )
. b. i. halperin , patrick a. lee , nicholas read , phys . rev .
b * 47 * , 7312 ( 1993 ) .
r. c. ashoori , j. a. lebens , n. p. bigelow , and r. h. silsbee , phys .
lett . * 64 * , 681 ( 1990 ) .
eisenstein , l. n. pfeiffer , and k.w .
west , phys .
lett . * 68 * , 3904 ( 1992 ) . k. m. brown , n. turner , j. t. nicholls , e. h. linfield , m. pepper , d. a. ritchie , and g. a. c. jones , phys .
b * 50 * , 15465(r ) ( 1994 ) .
y. hatsugai , p. -a .
bares , and x. g. wen , phys .
lett . * 71 * , 424 ( 1993 ) .
song he , p. m. platzman , and b. i. halperin , phys .
. lett . * 71 * , 777 ( 1993 ) .
peter johannson and jari m. kinaret , phys .
* 71 * , 1435 ( 1993 ) .
a. l. efros and f. g. pikus , phys .
b * 48 * , 14694 ( 1993 ) . c. m. varma , a. i. larkin , and e. abrahams , phys.rev .
b * 49 * , 13999 ( 1994 ) .
rudolf haussmann , hiroyuki mori , and a. h. macdonald , phys .
lett . * 76 * , 979 ( 1996 ) .
l. s. levitov and a. v. shytov , jetp lett .
* 66 * , 214 ( 1997 ) .
j. p. eisenstein , l. n. pfeiffer , and k. w. west , appl .
* 57 * , 2324 ( 1990 ) .
strictly true only at zero interlayer voltage @xmath31 .
capacitive effects at non - zero @xmath31 disrupt this density balance , but the effects are small and of no relevance here .
except in ref .
@xcite where a classical calculation is reported .
j. p. eisenstein , l. n. pfeiffer , and k. w. west , phys .
* 74 * , 1419 ( 1995 ) .
j. p. eisenstein , t. j. gramila , l. n. pfeiffer , and k. w. west , phys .
b * 44 * , 6511 ( 1991 ) .
j. hu and a. h. macdonald , phys .
b * 46 * , 12554 ( 1992 ) .
i. v. kukushkin , k. von klitzing , and k. eberl , phys .
lett . * 82 * , 3665 ( 1999 ) .
a. e. dementyev , n. n. kuzma , p. khandelwal , s. e. barrett , l. n. pfeiffer , and k. w. west , phys .
lett . * 83 * , 5074 ( 1999 ) .
s. melinte , n. freytag , m. horvatic , c. berthier , l. p. levy , v. bayot , and m. shayegan , phys . rev . lett . * 84 * , 354 ( 2000 ) . i. b. spielman , l. a. tracy , j. p. eisenstein , l. n. pfeiffer , and k. w. west , phys . rev . lett . *
94 * , 076803 ( 2005 ) . n. kumada , k. muraki , k. hashimoto , and y. hirayama , phys .
94 * , 096802 ( 2005 ) .
irene dujovne , a. pinczuk , moonsoo kang , b. s. dennis , l. n. pfeiffer , and k. w. west , phys . rev
* 95 * , 056808 ( 2005 ) .
tracy , j.p .
pfeiffer , and k.w .
west , phys .
* 98 * , 086801 ( 2007 ) .
p. giudici , k. muraki , n. kumada , y. hirayama , and t. fujisawa , phys .
lett . * 100 * , 106803 ( 2008 ) .
y. q. li , v. umansky , k. von klitzing , and j. h. smet , phys .
* 102 * , 046803 ( 2009 ) and phys .
b * 86 * , 115421 ( 2012 ) .
a. d. k. finck , j.p .
eisenstein , l.n .
pfeiffer , and k.w .
west , phys .
lett . * 104 * , 016801 ( 2010 ) . for the fully spin polarized case @xmath92
, according to k. park and j. k. jain , phys .
80 * , 4237 ( 1998 ) .
for a recent study , see yang liu , s. hasdemir , a. wojs , j. k. jain , l. n. pfeiffer , k. w. west , k. w. baldwin , and m. shayegan , phys .
b * 90 * , 085301 ( 2014 ) .
r. l. willett , m. a. paalanen , r. r. ruel , k. w. west , l. n. pfeiffer , and d. j. bishop , phys .
lett . * 54 * , 112 ( 1990 ) .
s. h. simon and a. stern , private communication .
the relative increase in @xmath25 shown in fig .
3(a ) is about 30% . if the excitonic ` downshift ' ( discussed in the context of fig .
1(b ) in the text ) is removed from the data , the relative increase is only about 15% .
crudely speaking , the size @xmath93 of the defects is related to the interlayer voltage @xmath31 through @xmath94 .
hence , smaller @xmath31 implies larger @xmath93 .
for a review , see j.p .
eisenstein , annu .
rev . condens .
matter phys .
* 5 * , 159 ( 2014 ) .
a. r. champagne , j.p .
eisenstein , l.n .
pfeiffer , and k.w .
west , phys .
rev . lett . * 100 * , 096801 ( 2008 ) .
|
the coulomb gap observed in tunneling between parallel two - dimensional electron systems , each at half filling of the lowest landau level , is found to depend sensitively on the presence of an in - plane magnetic field . especially at low electron density ,
the width of the coulomb gap at first increases sharply with in - plane field , but then abruptly levels off .
this behavior appears to coincide with the known transition from partial to complete spin polarization of the half - filled lowest landau level .
the tunneling gap therefore opens a new window onto the spin configuration of two - dimensional electron systems at high magnetic field .
|
the concept of the stone - ech compactification of a semigroup has become one of central importance , and has been studied extensively . throughout this paper , we think of the stone - ech compactification of a discrete abelian semigroup @xmath0 as the set @xmath1 of ultrafilters on @xmath0 , where the point @xmath2 is identified with the principal ultrafilter @xmath3 , and the basic open sets are those of the form @xmath4 , for @xmath5 .
then these sets are actually clopen , and @xmath6 is really the closure in @xmath1 of the set @xmath7 , regarded as a subset of @xmath1 under the aforementioned identification of points in @xmath0 with principal ultrafilters . the semigroup operation @xmath8 on @xmath0 is also extended by the formula @xmath9 which turns @xmath1 into a right topological semigroup , meaning that for each @xmath10 the mapping @xmath11 is continuous ( note that the extended operation @xmath8 need not be commutative , and , even if @xmath0 is a group , elements @xmath12 do not necessarily have inverses ) .
the details of this construction ( as well as a lot more information , along with applications ) can be seen in @xcite . in this paper
, we will focus mainly on the case when @xmath0 is a group .
the lowercase roman letters @xmath13 are reserved for ultrafilters , while the uppercase roman letters @xmath14 , with or without subscripts , will always denote subsets of the abelian group at hand .
we will use the von neumann natural numbers , i.e. , a natural number @xmath15 is viewed as the set @xmath16 ( with @xmath17 equal to @xmath18 , the empty set ) ; and @xmath19 will denote the set of finite ordinals , i.e. the set of natural numbers along with zero ( thus the symbols @xmath20 and @xmath21 mean the same when applied to natural numbers , @xmath17 , and to @xmath19 itself ) .
the lowercase roman letters @xmath22 , with or without subscript , will be reserved to denote elements of @xmath19 .
the lowercase roman letters @xmath23 , with or without subscript , will stand for elements of @xmath24^{<\omega}$ ] , i.e. for finite subsets of @xmath19 .
the letters @xmath25 and @xmath26 , with or without subscripts , will in general be reserved for denoting ( finite or infinite ) subsets of @xmath19 .
given a subset @xmath27 , @xmath28^{<\omega}$ ] will denote the set of finite subsets of @xmath25 , and @xmath28^{\omega}$ ] denotes the set of infinite subsets of @xmath25 . of the groups that we study here ,
one of the most important ones is the circle group @xmath29 .
when dealing with this group , we will identify its elements ( which are cosets modulo @xmath30 ) with their unique representative @xmath31 satisfying @xmath32 .
therefore , when we refer to an element of @xmath33 as a real number in @xmath34 , we really mean the coset of that number modulo @xmath30 .
if @xmath0 is an abelian semigroup , we say that an ultrafilter @xmath10 is * strongly summable * if it has a base of @xmath35-sets , i.e. if for every @xmath36 there exists a sequence @xmath37 such that @xmath38 , where @xmath39^{<\omega}\setminus\{\varnothing\}\right\}\ ] ] denotes the _ set of finite sums of the sequence @xmath40_. note that if a strongly summable ultrafilter is principal , then it must actually be @xmath17 .
strongly summable ultrafilters on @xmath41 were first constructed , under @xmath42 , by neil hindman in @xcite , although at that time the terminology was still not in use .
their importance at first came from the fact that they are examples of idempotents in @xmath43 , but among idempotents they are special in that the largest subgroup of @xmath44 containing one of them as the identity is just a copy of @xmath30 .
more concretely , ( * ? ? ?
* th . 12.42 )
establishes that if @xmath45 is a strongly summable ultrafilter , and @xmath46 are such that @xmath47 , then @xmath48 . in @xcite ,
the authors generalize some results previously only known to hold for ultrafilters on @xmath43 or @xmath49 .
in particular , they proved there that every strongly summable ultrafilter @xmath50 on any abelian group @xmath0 is an idempotent ( ( * ? ? ? * th .
2.3 ) ) . and
4.6 ) states that if @xmath0 can be embedded in @xmath33 , then whenever @xmath51 are such that @xmath47 , it must be the case that @xmath52 .
it is possible to get a slightly stronger result if one strengthens the definition of strongly summable .
an ultrafilter @xmath10 is * sparse * if for every @xmath36 there exist two sequences @xmath37 , @xmath53 , where @xmath54 is a subsequence of @xmath40 such that @xmath55 is infinite , @xmath56 , and @xmath57 .
then obviously every sparse ultrafilter will be nonprincipal and strongly summable . and
( ( * ? ? ?
4.5 ) ) if @xmath0 can be embedded in @xmath33 and @xmath58 is sparse , then whenever @xmath59 are such that @xmath60 , it must be the case that @xmath52 . in @xcite ,
the authors investigate the different kinds of abelian semigroups on which every nonprincipal strongly summable ultrafilter must be sparse .
for example , every nonprincipal strongly summable ultrafilter @xmath45 must actually be sparse ( this follows from ( * ? ? ?
3.2 ) together with either ( * ? ? ?
* lemmas 12.20 , 12.32 ) or ( * ? ? ?
* lemmas 1a , 1c ) ) .
thus the above result about @xmath50 being expressible as a sum only trivially holds for all nonprincipal strongly summable ultrafilters on @xmath19 . more generally , ( * ? ? ?
4.2 ) establishes that if @xmath61 is a countable subsemigroup of @xmath33 , then every nonprincipal strongly summable ultrafilter on @xmath61 is sparse .
after , they build on this to prove a more general result .
let @xmath61 be a countable subsemigroup of @xmath62 and let @xmath50 be a nonprincipal strongly summable ultrafilter on @xmath61 . if @xmath63 then @xmath50 must be sparse ( here @xmath64 denotes the least @xmath65 such that @xmath66 is nonzero ) .
so , for example , this theorem , as well as the method for proving it , can not be applied if @xmath50 contains the set of @xmath67 all of whose nonzero entries equal @xmath68 .
this set is isomorphic to the countably infinite boolean group @xmath69 . while @xcite was still a preprint , it contained the question of whether it is consistent with @xmath70 that there exists a nonprincipal nonsparse strongly summable ultrafilter on @xmath69 .
this question is answered in the negative in section 2 , while section 3 gives a slight improvement of ( * ? ? ?
by the boolean group we mean the unique ( up to isomorphism ) countably infinite group all of whose nonidentity elements have order @xmath71 .
this group is usually thought of as the direct sum of countably many copies of @xmath72 .
however , we will think of it as the group whose underlying set is @xmath24^{<\omega}$ ] , equipped with the symmetric difference @xmath73 as the group operation . for @xmath74
, we can define the support of @xmath75 by @xmath76 so that the mapping @xmath77 is an isomorphism from @xmath69 onto @xmath24^{<\omega}$ ] . when dealing with @xmath35-sets on this group , we will talk about sets instead of sequences .
thus , if @xmath37 is a sequence of elements of @xmath24^{<\omega}$ ] , and @xmath78 is the range of that sequence , then instead of @xmath79 we will write @xmath80 , the set of `` finite symmetric differences '' , in order to emphasize that the elements of our group are sets and that their `` sum '' actually corresponds to taking symmetric differences , and using the fact that , even if the sequence @xmath40 is not injective , that does not alter the resulting @xmath35-set .
this means that , for example , if @xmath81 , and @xmath82 for @xmath83 , then @xmath84 , due to the fact that every element of our group at hand has order @xmath71 .
we will use the uppercase roman letters @xmath85 to denote infinite subsets of @xmath24^{<\omega}$ ] whenever we are interested in considering their sets of finite symmetric differences .
the main result of this section , and of this paper , is the following theorem .
[ teorema ] let @xmath50 be a nonprincipal strongly summable ultrafilter on @xmath24^{<\omega}$ ] .
then , @xmath50 is sparse . in order to prove this result ,
we need first of all a lemma which tells us that weakly summable ultrafilters in @xmath24^{<\omega}$ ] have a property that is somewhat analogous to that of extending the frchet filter . recall that an ultrafilter @xmath50 on an abelian semigroup @xmath0 is * weakly summable * if for every @xmath36 there is a sequence @xmath40 of elements of @xmath0 such that @xmath56 .
thus every strongly summable ultrafilter is weakly summable , and actually ( ( * ? ? ?
* th . 12.17 ) ) every idempotent ultrafilter on an arbitrary semigroup is weakly summable ( and in fact ( * ? ? ?
12.17 ) an ultrafilter is weakly summable if and only if it is a closure point in @xmath1 of the set of idempotents ) .
notice that a principal weakly summable ultrafilter must be idempotent , in particular if @xmath0 is a group then the only principal weakly summable ultrafilter is the one that corresponds to the identity element .
[ keylemma ] let @xmath50 be a weakly summable ultrafilter on @xmath24^{<\omega}$ ] .
then for any @xmath86 , there exists an @xmath36 such that @xmath87 .
if @xmath50 is principal , then @xmath88 will do .
otherwise , let @xmath89^{<\omega}\big|n\notin a\}$ ] and @xmath90^{<\omega}\setminus a_0=\{a\in[\omega]^{<\omega}\big|n\in a\}$ ] .
there is @xmath91 such that @xmath92 .
but @xmath93 can not equal @xmath94 , for otherwise , since @xmath50 is weakly summable , there would be an infinite set @xmath95^{<\omega}$ ] such that @xmath96 , so if @xmath97 are two distinct elements , we have that @xmath98 and @xmath99 , thus @xmath100 , a contradiction
. therefore @xmath101 , and certainly it is true that @xmath102 .
if @xmath50 is a weakly summable ultrafilter , then for any finite subset @xmath103 of @xmath19 , there is @xmath36 such that @xmath104 is disjoint from @xmath103 .
if @xmath50 is principal , then @xmath88 will do .
otherwise , for each @xmath105 , choose @xmath106 such that @xmath107 .
then @xmath108 , and certainly this set is as required .
originally , the author had a much more involved proof for the previous corollary , whith ideas similar to those of ( * ? ? ?
2.6 ) and ( * ? ? ?
4 ) , until he came up with the much simpler one that is presented above .
the fact that all elements of @xmath24^{<\omega}$ ] have order @xmath71 has some remarkable consequences , amongst which the following is relevant for our purposes .
[ fdoffd ] let @xmath95^{<\omega}$ ] .
then , @xmath109 .
the @xmath110 " part of the equality follows from the fact that @xmath111 , and holds in any ( semi)group . now let us illustrate the @xmath112 " part with the case when we add two finite sums .
thus let @xmath113^{<\omega}\setminus\{\varnothing\}$ ] be distinct , and notice that , since every element in our group at hand has order two , the following holds : @xmath114 and from this it is easy to conclude , by induction , the desired result .
now in order to prove our main result , namely theorem [ teorema ] , let @xmath50 be a nonprincipal strongly summable ultrafilter on @xmath24^{<\omega}$ ] .
we want to show that @xmath50 is sparse , thus pick @xmath36 , and pick @xmath115 such that @xmath116 .
we would like to find some sets @xmath117 such that @xmath118 , @xmath119 , @xmath120 and @xmath121 is infinite .
we will do so as follows .
[ claim ] it is possible to find a @xmath122 such that @xmath123 , @xmath120 , and such that there are infinitely many @xmath124 with @xmath125 .
. then claim [ claim ] guarantees that @xmath121 is infinite .
moreover @xmath120 , and now by lemma [ fdoffd ] we get that @xmath127 , and we are done . thus , the only thing that remains to be proved is claim [ claim ] .
consider the set @xmath128 which contains exactly those @xmath86 such that @xmath129 for infinitely many distinct @xmath124 .
then if this set is nonempty , say @xmath130 , we can use lemma [ keylemma ] to get a @xmath131 such that @xmath132 .
since @xmath50 is strongly summable , we can find a @xmath122 such that @xmath133 .
then @xmath134 , and for each @xmath124 containing @xmath15 ( and by assumption there are infinitely many such ) we have that @xmath125 , because otherwise we would have @xmath135 contradicting our choice of @xmath15 and @xmath136 .
the other case is when @xmath137 . in this case , let @xmath138 .
then @xmath25 is an infinite subset of @xmath19 , with the property that each @xmath139 is contained in only finitely many @xmath124 ; and we will construct by recursion a very special subset of @xmath25 .
start by letting @xmath140 , @xmath141 and @xmath142 .
then both @xmath143 are finite and nonempty ( although @xmath144 is a subset of @xmath24^{<\omega}$ ] , whilst @xmath145 is a subset of @xmath19 ) .
now recursively define @xmath146 @xmath147 , and @xmath148 .
then again @xmath149 is finite , nonempty , and disjoint from all previous @xmath150 .
also @xmath151 is finite and nonempty , although the @xmath152 need not be disjoint , and of course @xmath153 . now notice that @xmath154 is an infinite subset of @xmath115 , and if @xmath155 , then @xmath156 .
thus if we let @xmath157 , then for every @xmath158 , @xmath159 will be nonempty if and only if @xmath155 for some even index @xmath160 .
let @xmath161^{<\omega}\big|s\cap n=\varnothing\}$ ] and @xmath162^{<\omega}\setminus b_0 $ ] .
now notice that whenever @xmath155 for some @xmath163 , we must have that @xmath164 .
thus , if we let @xmath165 be such that @xmath166 , we will have that @xmath164 for all @xmath167 , and there are infinitely many such . now using the fact that @xmath50 is strongly summable , just pick @xmath122 such that @xmath168 .
i am thankful to juris steprns for pointing out an error in an earlier version of this proof , as well as to the anonymous referee for useful comments on it .
in this section we will investigate a necessary condition for the existence of a nonsparse strongly summable ultrafilter on some abelian group @xmath0 .
this represents some partial progress towards answering ( * ? ? ?
* question 4.12 ) , and sheds some light on what an answer to that question might look like .
let @xmath0 be any abelian group , @xmath61 a subsemigroup of @xmath0 , and @xmath10 an ultrafilter such that @xmath169 .
then @xmath170 will be an ultrafilter on @xmath61 , and it is easy to see that @xmath171 is a nonprincipal ultrafilter if and only if @xmath50 is .
it is also reasonably straightforward to see that @xmath171 is strongly summable if and only if @xmath50 is , and also that @xmath171 is sparse if and only if @xmath50 is , because @xmath36 if and only if @xmath172 . now
if @xmath0 is any infinite abelian group , and @xmath10 is a strongly summable ultrafilter , then by definition , there is a sequence @xmath37 such that @xmath173 . if we let @xmath61 denote the subsemigroup of @xmath0 generated by @xmath174 , then it must be the case that @xmath61 is countable ( and cancellative ) . since @xmath175 , then @xmath169 , and
thus @xmath176 will be strongly summable .
moreover the question of whether @xmath50 is sparse reduces to the question of whether @xmath177 is sparse .
thus , when investigating the possibility of a strongly summable ultrafilter being nonsparse on an arbitrary abelian group , we may as well focus our attention on strongly summable ultrafilters on countable cancellative abelian semigroups .
now if @xmath61 is a countable cancellative abelian semigroup , then it can be embedded in a countable abelian group @xmath178 ( just in the same way that @xmath179 can be embedded into @xmath180 , or @xmath181 into @xmath182 ) . and
it is a well - known result ( see , e.g. , ( * ? ? ?
24.1 ) , @xcite , or ( * ? ? ?
* th . 9.23 ) ) that every abelian group can be embedded in a divisible group ; moreover , each divisible group is a direct sum of copies of @xmath183 and of quasicyclic groups ( ( * ? ? ?
23.1 ) , @xcite , or ( * ? ? ?
* th . 9.14 ) ) . since @xmath183 , as well as
all quasicyclic groups , can certainly be embedded in @xmath33 , the conclusion is that every countable abelian group @xmath178 can be embedded in the direct sum of countably many circle groups @xmath184 .
( from now on , @xmath0 will denote that group ) .
thus we can think of @xmath61 as a subset of @xmath0 , and if @xmath185 is an ultrafilter , then by letting @xmath50 be the filter on @xmath0 generated by @xmath177 , we will actually get an ultrafilter .
moreover @xmath169 , and @xmath176 , so @xmath177 will be strongly summable if and only if @xmath50 is . and again , the question of whether @xmath177 is sparse reduces to the question of whether @xmath50 is sparse .
therefore , the whole investigation of whether there is a nonsparse strongly summable ultrafilter on some abelian group ( or abelian cancellative semigroup ) reduces to the question of whether there exists a nonsparse strongly summable ultrafilter on @xmath0 .
it is not hard to see that the set of elements of @xmath0 whose order is a power of two is exactly @xmath186 $ ] , where @xmath187=\left\{t\in\mathbb t\bigg|\left(\exists m , n\in\omega\right)\left(t=\frac{m}{2^n}\right)\right\}\ ] ] is the quasicyclic @xmath71-group ( also known as the prfer group of type @xmath188 ) . from now on we will focus on strongly summable ultrafilters on that group ( which we will keep denoting by @xmath178 ) .
we will also be using the groups @xmath189=\left\{t\in\mathbb t\bigg|\left(\exists m\in\omega\right)\left(t=\frac{m}{2^n}\right)\right\}\cong\mathbb z_{2^n},\ ] ] the isomorphism being given by @xmath190 ( whenever we refer to a number @xmath191 as an element of @xmath192 , we really mean its coset modulo @xmath160 ) . notice that if @xmath193 , then @xmath194\subseteq\mathbb t[2^m]$ ] , and @xmath195={\mathop{\bigcup}\limits}_{n<\omega}\mathbb t[2^n]$ ] . by ( * ? ? ?
3.2 ) , if @xmath50 is a nonprincipal ultrafilter , and for each @xmath36 there is a sequence @xmath40 satisfying strong uniqueness of finite sums such that @xmath38 , then @xmath50 is sparse .
we proceed by contraposition , so let us assume that there is @xmath86 such that @xmath202 , and , essentially without loss of generality , let us also assume that @xmath203 .
pick @xmath204 such that @xmath205 , where @xmath206 i.e. , thinking of @xmath207 as a number in @xmath208 written in binary notation , its first digits will be @xmath209 and then there will be an infinite string of zeroes and ones .
then @xmath210 if and only if the first such nonzero digit appears in a position that is congruent with @xmath93 modulo @xmath211 .
let @xmath212 , and let @xmath40 be a sequence of elements of @xmath178 satisfying @xmath213 .
note that if @xmath214 and for some @xmath215 , we have that @xmath216 and @xmath217 , then @xmath218 and so @xmath219 , which is impossible .
thus there is at most one @xmath220 in each interval @xmath221 ( the positions of the first nonzero digits of distinct @xmath220 are distinct ) , so we may assume that the sequence @xmath40 is arranged in such a way that @xmath222 implies @xmath223 ( @xmath40 is arranged in increasing order of its first nonzero digit in the @xmath15-th projection ) .
consequently , for each @xmath224 we have that @xmath225 and therefore @xmath226 .
this is easily seen to imply that for @xmath196^{<\omega}\setminus\{\varnothing\}$ ] and for @xmath227 , if @xmath228 then @xmath198 and @xmath229 . and of course this implies that for @xmath196^{<\omega}\setminus\{\varnothing\}$ ] and for @xmath227 , if @xmath230 then @xmath198 and @xmath229 .
the latter statement in turn easily implies that the sequence @xmath40 satisfies strong uniqueness of finite sums , hence , by ( * ? ? ?
3.2 ) , @xmath50 must be sparse .
enumerate the finite set @xmath231=\{g_0,\ldots , g_{k-1}\}$ ] and choose @xmath233 such that @xmath234 . since @xmath50 is weakly summable
, we can pick a sequence @xmath40 of elements of @xmath0 such that @xmath235 .
but then , for example , @xmath236 , thus @xmath237 and this implies that @xmath238 . in what follows we will use ( * ? ? ? * th .
4.5 ) , which says that if @xmath50 is a nonprincipal , strongly summable ultrafilter on a subsemigroup @xmath61 of @xmath184 , and if @xmath240 ( where @xmath64 denotes the least @xmath15 such that @xmath241 ) , then there exists an @xmath242 such that any sequence @xmath40 with @xmath243 satisfies strong uniqueness of finite sums ( and in particular @xmath50 is sparse ) .
now assume that @xmath50 is a nonprincipal , nonsparse strongly summable ultrafilter on @xmath178 .
notice that @xmath50 can not contain the set @xmath244 $ ] , because this set is a copy of @xmath69 and hence if @xmath50 contains it , that would induce a strongly summable ultrafilter @xmath177 on the boolean group , which by theorem [ teorema ] must be sparse and therefore @xmath50 will also be sparse . hence @xmath50 must contain the set @xmath245 . for @xmath246 ,
let @xmath247 denote the least @xmath15 such that @xmath248 .
consider the morphism @xmath250 given by @xmath251 , whose kernel is exactly @xmath252 $ ] .
since the latter is not an element of @xmath50 , then @xmath253 ( i.e. the image of @xmath50 under the continuous extension of @xmath254 to @xmath255 , which is given by @xmath256\in p\}$ ] ) is a nonprincipal ultrafilter .
moreover , since @xmath50 is strongly summable , by ( * ? ? ? *
lemma 4.4 ) , so is @xmath257 .
now notice that for @xmath258 , we have @xmath259 . thus @xmath257 contains the set @xmath260 , since its preimage under @xmath261 is exactly @xmath262 .
therefore by ( * ? ? ?
4.5 ) , there is a set @xmath263 such that whenever @xmath264 , @xmath265 must satisfy strong uniqueness of finite sums .
now for @xmath36 , we can pick a sequence @xmath40 such that @xmath266 $ ] .
then if we let @xmath54 be the sequence given by @xmath267 , we get that @xmath268\subseteq x$ ] , thus @xmath54 satisfies strong uniqueness of finite sums .
it is not hard to see that this implies that @xmath40 satisfies strong uniqueness of finite sums as well , thus @xmath50 has a basis of sets of the form @xmath79 for sequences @xmath40 satisfying strong uniqueness of finite sums .
therefore by ( * ? ? ?
3.2 ) , @xmath50 is sparse .
[ sucesiondezetasdosenes ] assume that there exists a nonsparse strongly summable ultrafilter @xmath50 on @xmath178 .
then there exists a sequence @xmath269 of natural numbers such that @xmath270\in p$ ] .
in particular , if there exists a ( nonprincipal ) nonsparse strongly summable ultrafilter on some abelian cancellative semigroup , then there exists one on @xmath271 , for some sequence @xmath272 .
let @xmath50 be a nonprincipal , nonsparse strongly summable ultrafilter on @xmath178 .
as was pointed out above , @xmath50 can not contain the set @xmath244 $ ] , hence @xmath273 .
moreover by lemma [ eluncuarto ] , @xmath274 .
now @xmath275 , where @xmath276 .
essentially without loss of generality , we can assume that @xmath277 .
now choose a sequence @xmath40 with @xmath278 , and for @xmath279 let @xmath280 ( so @xmath281 implies @xmath282 ) .
assume , by way of contradiction , that there are pairwise distinct @xmath284 .
let @xmath285 . for @xmath286 , we have that @xmath287 , because @xmath288 . on the other hand , @xmath289 thus @xmath290 , so @xmath291 and @xmath292 , which is a contradiction .
thus , by rearranging the sequence if necessary , we may assume that @xmath293 and @xmath294 implies that @xmath295 .
equivalently , @xmath295 implies that @xmath296 , where the inequality is strict if @xmath297 .
let @xmath301 . for @xmath286 , we since @xmath302 we have that @xmath287 .
on the other hand , @xmath282 while @xmath303 , so @xmath304 .
hence @xmath291 , now since @xmath305 , whe must have @xmath306 , which can only happen if @xmath300 .
if @xmath309 , then by claim [ alomas2 ] , we know that @xmath283 .
thus we can let @xmath310 and @xmath311 ( so that @xmath312 equals either @xmath160 or @xmath313 , @xmath314 , and @xmath315 ) . now claim [ iszeroonpreviousrho ] yields @xmath316 for @xmath317 , therefore @xmath318 which is finite .
now if @xmath319 , then claim [ alomas2 ] guarantees that there are only finitely many integers @xmath320 such that @xmath321 , so let @xmath160 be the greatest such integer if some exists , or @xmath322 otherwise ( equivalently @xmath323 where @xmath93 is the greatest integer less that @xmath65 for which @xmath324 , if such a @xmath93 exists , or @xmath322 otherwise ) .
thus we know that for @xmath325 , @xmath326 . therefore @xmath327 which is finite as well .
notice that , if @xmath328 is a finite subset of @xmath195 $ ] , then the subgroup of the latter generated by the former must be @xmath194 $ ] for suitable @xmath15 .
namely , if @xmath329 , where @xmath330 for @xmath331 , and @xmath332 , then @xmath328 generates the subgroup @xmath333 $ ] .
hence by claim [ finitesetofgenerators ] , for each @xmath279 we can choose @xmath334 such that the subgroup of @xmath195 $ ] generated by @xmath307 is @xmath335 $ ] .
in this way we construct the sequence @xmath269 of natural numbers which satisfies that @xmath336 $ ] .
recall that @xmath194\cong\mathbb z_{2^n}$ ] ; and that , if @xmath295 , then @xmath194\subseteq\mathbb t[2^m]$ ] . from this , it is not hard to see that the sequence @xmath272 from the previous theorem has to be unbounded . for
if that sequence was bounded , say by @xmath15 , then we would have that @xmath337\subseteq\bigoplus_{i<\omega}\mathbb t[2^n]$ ] .
thus if @xmath50 is the ultrafilter yielding @xmath272 , @xmath50 would contain @xmath338 $ ] , thus inducing a ( nonprincipal ) nonsparse strongly summable ultrafilter @xmath177 on its isomorphic copy @xmath339 .
but this can not happen , more generally , for every @xmath340 every strongly summable ultrafilter @xmath177 on @xmath341 is sparse .
the case when @xmath342 is just theorem [ teorema ] , and for @xmath343 , pick @xmath344 such that @xmath345 , where @xmath346 is the set consisting of those @xmath347 whose first nonzero coordinate equals @xmath160 .
it is then easy to see that , for @xmath36 , if @xmath40 is a sequence of elements of @xmath348 such that @xmath349 , then for distinct @xmath350 the indices of the first nonzero coordinates of @xmath351 and @xmath352 must be different ( otherwise the first nonzero coordinate of @xmath353 would be @xmath354 , so we would have @xmath355 , which is absurd ) .
this in turn implies that the sequence @xmath40 satisfies the strong uniqueness of finite sums , and thus by ( * ? ? ?
3.2 ) the desired conclusion follows .
therefore , since every sequence @xmath272 given by the theorem must be unbounded , one might be tempted to think that every such sequence should tend to infinity very quickly , but this is really not the case , as the following corollary shows .
i am thankful to andreas blass for pointing this out to me .
if @xmath50 is a nonprincipal nonsparse strongly summable ultrafilter on @xmath178 , then there is an injective homomorphism @xmath250 sending @xmath50 to an ultrafilter @xmath177 ( which must necessarily be also nonprincipal nonsparse strongly summable ) containing the set @xmath356 $ ] .
in particular , if there is a nonprincipal nonsparse strongly summable ultrafilter on some abelian cancellative semigroup , then there is one on @xmath357 .
given the sequence @xmath272 from theorem [ sucesiondezetasdosenes ] , create a new strictly increasing sequence @xmath358 by letting @xmath359 be the least @xmath160 such that @xmath360 , and recursively letting @xmath361 be the least @xmath160 such that @xmath362 .
then we can define the embedding @xmath250 by letting @xmath363 be the element of @xmath178 whose @xmath364-th coordinate is exactly the @xmath65-th coordinate of @xmath365 and whose @xmath160-th coordinate is zero whenever @xmath366 .
clearly @xmath261 is injective .
now notice that for @xmath367 $ ] , since by construction @xmath368 , we have for every @xmath279 that @xmath369\subseteq\mathbb t[2^{m_i}]$ ] , and of course for @xmath366 we have that @xmath370 $ ] . thus @xmath371 $ ] , hence @xmath372\right]\subseteq\bigoplus_{n<\omega}\mathbb t[2^n]$ ] and so the latter is an element of @xmath257 , and thus the result follows .
|
we investigate the possibility of the existence of nonsparse strongly summable ultrafilters on certain abelian groups .
in particular , we show that every strongly summable ultrafilter on the countably infinite boolean group is sparse .
this answers a question of hindman , steprns and strauss .
|
nuclear magnetic resonance ( nmr ) has turned out to be very useful for studying the superfluid phases of liquid @xmath0he .
the two superfluid phases a and b are distinguished in the nmr spectrum by different frequency shifts of the absorption peaks.@xcite in addition to these `` bulk '' peaks , one often observes additional `` satellite '' peaks .
these are caused by topological objects and textures that appear in the superfluid order parameter . especially in superfluid @xmath0he - a ,
several different objects have been identified based on the frequency shifts of the satellite peaks.@xcite the simplest of these are solitons .
they are domain - wall - like structures where a planar object separates two different but degenerate bulk states .
the satellite peaks in @xmath0he - a were first observed in measurements in the mid 1970s.@xcite soon after the theory of solitons in @xmath0he was developed by maki and kumar.@xcite their calculation gave a striking agreement with the measured frequency shifts of the satellite peaks at temperatures close to the superfluid transition temperature @xmath1 .
this initial success had the consequence that further studies of solitons went to other directions@xcite and , unfortunately , no more precise calculations were done .
a soliton has two basic structures , `` twist '' and `` splay , '' which correspond to the cases of a magnetic field ( @xmath2 mt ) perpendicular and parallel , respectively , to the plane of the soliton wall .
both these structures can be studied using small oscillations of the magnetization that are either transverse or longitudinal relative to the static field .
there are several points that can be improved in the previous calculations , given as follows .
( i ) the structure of the splay soliton was calculated only by using variational approximation .
( ii ) the calculations were limited to temperatures near @xmath1 .
( iii ) the effect of different parameter values was not studied .
( iv ) the amplitudes and widths of the satellite peaks as well as peaks of higher order were not studied .
( v ) dissipation was neglected .
it is just these points that we address in this paper .
an additional motivation is that the study of solitons opens the way to detailed understanding of more complicated topological objects such as vortices .
we find that , using dissipationless theory , the agreement between the theoretical and experimental frequency shifts is generally very good .
however , we find a puzzling disagreement in the transverse oscillation frequency of the splay soliton . this disagreement has remained unnoticed because no detailed comparison between theory and experiment has been published .
further , we find that the theory is particularly inflexible to explain this discrepancy away . taking into account dissipation , in particular spin diffusion ,
changes these conclusions . on one hand ,
the disagreement in the splay - soliton frequency is reduced . on the other hand
, we find considerable extra shift of resonance frequencies in high fields , which has not been reported experimentally .
we also point out that the longitudinal resonance of the splay soliton has not been studied experimentally .
measurement of these quantities would be important to test our understanding of the basic properties of superfluid @xmath0he .
we start in sec .
[ s.hyrdo ] with a short introduction to the hydrodynamic theory and nmr in @xmath0he - a . in sec .
[ s.eqstruc ] we solve numerically the structure of the splay soliton .
the frequency and the absorption of the principal satellite peak are determined in secs .
[ s.princnmr ] and [ s.intes ] ignoring dissipative effects . in sec .
[ s.intes ] we calculate the frequencies and absorption of the higher modes . in sec .
[ s.dissipation ] we take into account dissipation and calculate the effect of the spin diffusion and normal - superfluid relaxation on the absorption spectrum .
here we briefly present some main points of the hydrodynamic theory and nmr in @xmath0he - a .
the order parameter of superfluid @xmath0he - a is a @xmath3 tensor of the form@xcite @xmath4 where @xmath5 , @xmath6 and @xmath7 are unit vectors and @xmath8 .
it is conventional to define @xmath9 , which gives the axis of the orbital angular momentum of a cooper pair .
the unit vector @xmath5 defines the axis along which the spin of the cooper pair vanishes . in a static magnetic field
, the structure of a soliton can be determined by finding a local minimum for the free energy@xcite @xmath10 here @xmath11 comes from the dipole - dipole interaction between nuclear moments , @xmath12 @xmath13 from coupling to the external field @xmath14 , @xmath15 and @xmath16 from the gradient of the order parameter , @xmath17 ^ 2 .
\label{e.fgrad}\end{aligned}\ ] ] the gradient energy ( [ e.fgrad ] ) also includes the kinetic energy arising from the superfluid velocity @xmath18 , where @xmath19 is the mass of a @xmath0he atom .
however , in the following we limit to the case of zero superfluid velocity . the parameters appearing in the gradient energy have been calculated in the weak - coupling approximation by cross@xcite and drfle,@xcite the latter including more fermi - liquid parameters . for numerical values
see refs . and .
the characteristic scales are given by the dipole length @xmath20 and the dipole field @xmath21 .
we consider a static field @xmath22 .
we assume that the equilibrium @xmath5 , denoted by @xmath23 , lies in the plane perpendicular to @xmath24 : @xmath25 this situation is always achieved in large field @xmath26 , where @xmath5 is forced to the plane by @xmath27 ( [ e.fmag ] ) , but in some cases this happens in low fields as well .
minimization of the total energy ( [ e.fstatic ] ) gives for @xmath28 the equation @xmath29 where the operator @xmath30 is defined by @xmath31 in a dynamic magnetic state one has to include the spin magnetization @xmath32 as a new variable in addition to @xmath5 and @xmath33 .
the effective energy density has the form@xcite @xmath34 where @xmath35 is the gyromagnetic ratio and @xmath36 the susceptibility tensor .
this leads to the equations of motion [ e.leggett ] @xmath37 where @xmath38 and @xmath39 is the susceptibility in the normal state .
the motion of @xmath33 is strongly limited by viscosity and therefore we assume that @xmath33 is independent of time.@xcite equations ( [ e.leggett ] ) describe dissipationless dynamics .
the inclusion of dissipative terms is postponed to sec .
[ s.dissipation ] .
the field @xmath14 is the sum of the static field @xmath24 and a small radio - frequency field @xmath40 that oscillates at angular frequency @xmath41 . throughout this paper
we limit to study the linear response of @xmath42 and @xmath5 to @xmath40 .
we parametrize the deviation of @xmath5 with two parameters @xmath43 and @xmath44 , @xmath45 for @xmath42 we parametrize the deviation @xmath46 from the equilibrium @xmath47 by @xmath48 and circular components @xmath49 .
similar definitions are used for other vectors as well .
we linearize eqs.([e.leggett ] ) and assume the time dependence @xmath50 , etc . using the equilibrium condition ( [ e.dtheta ] )
we get [ e.setcons ] @xmath51 here @xmath52 is the larmor frequency . the potentials @xmath53 and @xmath54 are defined by @xmath55 in order to simplify eqs .
( [ e.setcons ] ) , let us consider the special case @xmath56 .
in this case the equations separate into independent blocks for longitudinal and transverse oscillations of the magnetization .
the resonance frequencies are determined by two independent schrdinger - type equations for @xmath43 and @xmath44 , @xmath57 the eigenvalues @xmath58 and @xmath59 of these equations are related to the resonance frequencies as @xmath60 here @xmath61 is the longitudinal resonance frequency of the a phase .
the corresponding eigenfunctions of eqs .
( [ e.schrod1 ] ) and ( [ e.schrod2 ] ) are denoted by @xmath62 and @xmath63 , respectively .
they are assumed to be normalized : @xmath64 and @xmath65 . because we have temporarily neglected dissipative processes ,
the power absorption @xmath66 consists of @xmath67-peaks , @xmath68 . instead of assuming @xmath56 , an alternative approach to eqs.([e.setcons ] )
is to study the high - field limit @xmath69 .
more precisely , one can calculate the resonance frequencies as a power series of @xmath70 and neglect terms of the order of @xmath71 and higher . in this approximation
all the three components @xmath72 , @xmath73 , and @xmath74 decouple . the eigenvalue equations and frequencies
are the same as above equations ( [ e.schrod1])-([e.freqperpgen ] ) except that eq.([e.freqperpgen ] ) is valid only to leading order in @xmath75 : @xmath76 . \label{e.freqperp}\end{aligned}\ ] ] for the rest of this section we assume the high - field limit @xmath69 . in the case of dipole locking , @xmath77 , the lowest bulk eigenvalues are @xmath78 . in this case
only the bulk eigenstate gives rise to absorption @xmath79 and @xmath80 , where the two modes @xmath81 and @xmath82 is the volume of the liquid . in the presence of dipole
unlocking also other eigenstates contribute to the absorption .
their intensities are given by@xcite @xmath83\right\vert^2 .
\label{e.intensity2}\end{aligned}\ ] ] here the @xmath84 s satisfy the sum rules @xmath85 \label{e.sumrule1b } , \\ \sum_{k=0}^{\infty } \alpha_{\parallel , k}^2 q_{\parallel , k } & = & \int d^3r\ , u_\parallel^2 \label{e.sumrule2a } , \\ \sum_{k=0}^{\infty } \alpha_{\perp , k}^2 q_{\perp , k } & = & \int d^3r\left\ { \left[(\hat{\bf l } \cdot \hat{\bf d}_0)^2 - l_z^2\right]^2 + ( \hat{\bf l}\times\hat{\bf d}_0)_z^2 ( \hat{\bf l } \cdot \hat{\bf d}_0)^2 \right\ } , \label{e.sumrule2b}\end{aligned}\ ] ] and so on . the sum rules can be derived using the orthogonality properties of the eigenfunctions . [ in eq .
( [ e.sumrule2b ] ) one also needs eq .
( [ e.dtheta ] ) . ]
the lowest - order rules ( [ e.sumrule0])-([e.sumrule1b ] ) apparently are equivalent to the sum rules presented by leggett.@xcite
the minimum of the dipole energy ( [ e.fdip ] ) can be achieved in two ways : either @xmath33 and @xmath5 are parallel or they are antiparallel .
this double degeneracy gives rise to domain walls or solitons .
on one side of the wall @xmath86 and on the other side @xmath87 . we will study cases where the static field @xmath24 is either perpendicular or parallel to the plane of the soliton wall .
the soliton structures in these cases are known to have composite twist and splay structures , respectively.@xcite the structure and the fundamental nmr frequency of a twist soliton have been solved analytically.@xcite here we concentrate on the case of a splay soliton , which has previously been studied only by variational methods.@xcite its structure is illustrated in fig .
[ splay ] . field ( light arrows ) has splay shape .
the structure is homogeneous in the @xmath88-@xmath89 plane of the soliton wall .
both @xmath5 and @xmath33 are perpendicular to the static field @xmath90.,width=302 ] in the cases of a twist or a splay soliton , both @xmath33 and @xmath5 are in the plane perpendicular to @xmath24 .
we parametrize @xmath91 and @xmath5 as above , eq . ( [ e.d0 ] ) . for a splay soliton
the angles @xmath28 and @xmath92 are taken as functions of @xmath93 only .
we use the boundary conditions @xmath94 and @xmath95 . substituting the vector fields ( [ e.d0 ] ) and ( [ e.ldvect ] ) into the total energy ( [ e.fstatic ] )
gives that the energy per unit area is @xmath96 \ .
\label{e.fcsplay0}\end{aligned}\ ] ] the energy functional ( [ e.fcsplay0 ] ) was discretized using equal intervals ( @xmath97 , @xmath98 ) , and the values of @xmath92 and @xmath28 at these discrete points were taken to be the minimizing variables .
the boundary conditions were taken into account by linear initial approximations @xmath99 and @xmath100 .
the minimization can be done by a simple relaxation method .
the resulting functions @xmath101 and @xmath102 at different temperatures are shown in fig .
[ angles ] . and @xmath102 for a splay soliton at temperatures @xmath103 .
the parameter values correspond to weak coupling at 3.4 mpa ( @xmath104 and @xmath105).,width=302 ]
we now apply the theory of sec . [ s.hyrdo ] to calculate nmr properties of solitons . for solitons @xmath106 , which decouples the longitudinal and transverse modes at arbitrary @xmath107 in eqs .
( [ e.setcons ] ) .
thus the eigenvalues @xmath59 are related to resonance frequencies as given in eq .
( [ e.freqperpgen ] ) at any field .
the main signature of solitons in the nmr spectrum comes from the lowest eigenvalue of schrdinger - like equations ( [ e.schrod1 ] ) and ( [ e.schrod2 ] ) .
this lowest frequency can be calculated , for example , using a variational formulation : @xmath108 this was discretized and the values of @xmath109 were taken to be the minimizing variables .
the form for the initial approximation used was @xmath110 and the boundary conditions @xmath111 were assumed at the end points .
the length @xmath112 was increased until its effect disappeared .
one solution is shown in fig .
[ wavefunc ] .
( solid line ) , the lowest eigenvalue @xmath113 , and the eigenfunction @xmath114 ( dashed line , vertical scale arbitrary ) for transverse resonance of a splay soliton at @xmath115 .
the absorption thickness @xmath116 gives the effective width of the wave function in nmr .
, width=302 ] in fig .
[ r2 ] we plot the temperature dependence of the lowest resonance frequencies as well as some experimental data . for completeness
we also include the analytical results for the twist soliton:@xcite @xmath117 ( [ e.freqll ] ) and @xmath113 ( [ e.freqperpgen ] ) as functions of temperature . splay - soliton transverse resonance : four upper solid lines and experimental points @xmath118 ( ref . ) , @xmath119 ( ref . ) , and @xmath120 ( ref .
splay - soliton longitudinal resonance : four lower solid lines .
twist - soliton transverse resonance : four upper dash - dotted lines and experimental points @xmath121 ( ref . ) and @xmath122 ( ref . ) .
twist - soliton longitudinal resonance : four lower dash - dotted lines and experimental points @xmath123 ( ref . ) and @xmath124 ( ref . ) . in each case
there are four theoretical lines , which in order of decreasing @xmath125 correspond to @xmath126 & weak coupling , @xmath126 & gap enhancement , @xmath127 & weak coupling , and @xmath127 & gap enhancement .
good agreement between theory and experiment is achieved with @xmath128 and weak coupling except for the high - temperature transverse splay data .
the effect of strong - coupling corrections is shown at @xmath129 where the lower ends of the bars correspond to @xmath130 and the upper ends to @xmath131 .
the experimental data is plotted with higher resolution in fig .
[ r2spindiff ] . ,
width=302 ] the reduced frequency shifts @xmath132 ( [ e.freqll ] ) and @xmath113 ( [ e.freqperpgen ] ) depend only on the ratios of the hydrodynamic coefficients @xmath133 , @xmath134 , @xmath135 , @xmath136 , and @xmath137 .
( note that the absolute magnitudes of @xmath138 s and @xmath139 define length and energy scales that do not affect @xmath132 or @xmath113 . ) in the weak - coupling approximation the ratios of @xmath138 s are functions of an infinite set of fermi - liquid parameters @xmath140 and @xmath141 , with @xmath142 , etc.@xcite here
we neglect all the coefficients with @xmath143 since they are unknown .
for @xmath144 we use the value by greywall@xcite at the melting pressure , @xmath145 . however , the dependence on @xmath144 is weak .
for example , the variation @xmath146 shows up only in the fourth decimal of @xmath125 at @xmath147 . in terms of pressure ,
the maximum difference in @xmath125 between 2.6 and 3.4 mpa is 1% at temperatures higher than @xmath148 .
the value @xmath105 was taken from ref . and
it is also consistent with ref . .
in order to see the effect of @xmath149 we also used @xmath150 . it can be seen in fig .
[ r2 ] that this shifts the resonance frequency up at temperatures below @xmath1 .
there are no quantitative calculations of strong - coupling effects in the a phase at a general temperature . in order to get some idea
how strong coupling could affect the soliton frequencies , we use a `` trivial strong - coupling '' model developed by serene and rainer@xcite for the b phase . in this model the weak - coupling energy gap is multiplied by a factor that depends on the temperature and on the jump of the specific heat @xmath151 at @xmath152 .
this dependence is tabulated in ref .. we adapt this model to the a phase by calculating the multiplying factor using the same table but substituting @xmath153 in place of @xmath154 , and using extrapolation when needed .
we take @xmath155 from measurements by greywall.@xcite it can be seen in fig .
[ r2 ] that the gap enhancement , which affects only intermediate temperatures , has a smaller effect than the change of @xmath149 . in the limiting case @xmath156 the reduced frequencies @xmath157 and @xmath158 are independent of any parameters appearing in the weak - coupling model , including also the gap enhancement .
this is a consequence of the ginzburg - landau expansion that gives to the parameters @xmath138 the ratios @xmath159 .
the only free parameter here is @xmath35 , which in the weak coupling ( with or without gap enhancement ) has the value @xmath160 .
this value is changed only when nontrivial strong - coupling corrections are included .
the `` weak - coupling - plus '' model by serene and rainer@xcite gives an estimate @xmath161 .
figure [ r2 ] shows the reduced frequencies corresponding to @xmath130 and @xmath131 .
we see that this variation changes the splay soliton @xmath113 less than 3% but for the twist soliton the effect is ten times larger .
there is rather good agreement between the experiment and the theory corresponding to weak coupling and @xmath105 .
equally good agreement is achieved with gap enhancement and @xmath162 .
the longitudinal twist data is a strong indication that the deviation from the weak - coupling value @xmath163 is small , as predicted by serene and rainer.@xcite the only major difference between theory and experiment exists in the transverse splay - soliton frequency at high temperatures .
it seems very difficult to improve the agreement in this case by any change of the parameters in the theory above since @xmath164 at @xmath1 is effectively fixed . on the experimental side ,
one possibility is that the field is not precisely in the plane of the soliton in the measurements .
this would add a small twist component to the splay soliton and thus shift up the frequency.@xcite another possibility is that the inclusion of relaxation mechanisms could shift the calculated resonance frequency , as will be discussed in sec .
[ s.dissipation ] .
at @xmath1 we find the eigenvalues @xmath165 and @xmath166 for @xmath163 .
these differ slightly from the variational results by maki and kumar@xcite that are 0.403 and 0.677 , respectively .
here we calculate the intensity of the principal soliton peak and analyze the absorption at other frequencies . for planar objects intensities ( [ e.intensity1 ] ) and ( [ e.intensity2 ] )
are most conveniently expressed in the form of an absorption thickness @xmath167 , which equals @xmath168 divided by the area of the planar object , @xmath169 for each mode @xmath170 .
the absorption thicknesses in the lowest eigenstate are plotted in fig .
[ abslen ] . ) of the principal soliton peak in four cases .
the parameter values correspond to weak coupling at 3.4 mpa ( @xmath104 and @xmath105 ) .
the numerical values at @xmath1 are @xmath171 , @xmath172 , @xmath173 , and @xmath174 .
, width=302 ] the absorption for a twist soliton can be calculated analytically .
substituting the eigenfunction @xmath175 into eqs .
( [ e.intensity1 ] ) and ( [ e.intensity2 ] ) and using @xmath176 ( refs . and ) we find @xmath177 here @xmath178 is the gamma function and @xmath179 in order to get all the resonance modes , we discretize the space as discussed above ( @xmath97 , @xmath98 , for twist soliton replace @xmath93 by @xmath89 ) .
this means that eqs .
( [ e.schrod1 ] ) and ( [ e.schrod2 ] ) turn into a matrix eigenvalue problem .
this can be solved by standard library routines for matrices of reasonable size .
this gives all the frequencies , including the fundamental one ( [ e.varform ] ) .
however , the higher modes extend over the whole interval , and it is important to fix boundary conditions for them . we require zero derivative of @xmath180 at @xmath181 .
this can be justified by considering a lattice of solitons .
the general eigenfunctions @xmath180 of eqs .
( [ e.schrod1 ] ) and ( [ e.schrod2 ] ) are of the bloch form , but only strictly periodic functions lead to nonzero absorption in eqs.([e.intensity1 ] ) and ( [ e.intensity2 ] ) .
the unit cell of a soliton lattice @xmath182 consists of two solitons located around @xmath183 and @xmath184 . the vectors @xmath33 and @xmath5 have the symmetries @xmath185 and @xmath186 .
this implies that the eigenfunctions can be classified as symmetric or antisymmetric ( with respect to @xmath187 ) .
the symmetric solutions have @xmath188 , and the antisymmetric solutions can be neglected since they do not contribute to the absorption . the results for the frequencies and intensities of the higher modes are shown in fig .
[ highmode ] . in the main frame whereas the inset also contains the points for @xmath189 . in both cases
the lattice spacing of the discretized lattice is @xmath190 .
other parameters are weak coupling at 3.4 mpa ( @xmath104 , @xmath105 ) and temperature @xmath191.,width=302 ] as expected , the higher modes depend on the size @xmath112 of the system used in the calculation .
the modes seem to appear as pairs above the bulk peak .
the bulk peak also seems to consist of two peaks that do not appear exactly at @xmath192 due to finite @xmath112 .
in fact , the finite @xmath112 corresponds to a periodic lattice of solitons , which should give rise to peaks that are analogous to the bragg peaks in the x - ray scattering from periodic solids .
such peaks are indeed seen experimentally in the case of a vortex sheet , where a nearly periodic arrangement is automatically generated.@xcite the observation of the peaks depends essentially on dissipative effects , which broadens the peaks
. this will be discussed in the following section .
there are two sources of dissipation that are important for solitons .
the nonequilibrium between normal and superfluid magnetizations is the main relaxation mechanism in homogeneous superfluid @xmath0he.@xcite in inhomogeneous situations the nonequilibrium between normal magnetizations at different locations causes relaxation via spin diffusion .
we study these two mechanisms below , the latter in the case where the inhomogeneity arises from the soliton .
we neglect extrinsic effects such as the inhomogeneity caused by nonuniform magnetic field .
we also do not consider dissipation caused by orbital motion . according to the estimation in ref .
, this is important only at temperatures very near @xmath1 , region which is vanishingly small on the scale of fig .
[ r2 ] .
our treatment of the dissipation is purely phenomenological . for normal - superfluid relaxation
we use the model developed by leggett and takagi , which is extensively described in ref . .
for the relaxation of spatially nonequilibrium magnetization we use simple diffusion equation .
this approach can be correct only when the mean free path of the quasiparticles is smaller than the thickness of the soliton , and therefore it necessarily fails at low temperatures , where the mean free path diverges .
our treatment also neglects the tensor character of the spin - diffusion coefficient , as well as the anisotropic energy gap with varying anisotropy axis @xmath193 , which leads to andreev reflection of the quasiparticles . in order to treat the relaxation ,
the spin polarization @xmath42 is divided into a superfluid ( pair ) part @xmath194 and a normal ( quasiparticle ) part @xmath195 .
both parts have separate equations of motion [ e.ltequ2]@xmath196 + \kappa \nabla^2 { \bf s}_q\label{e.ltequ2a } \\ \dot{\bf s}_p & = & \gamma{\bf s}_p \times \left ( { \bf b } - \mu_0\gamma \frac{f_0^a}{\chi_{0 } } { \bf s}_q \right ) - \frac{1}{\tau}\left [ ( 1-\lambda ) { \bf s}_p - \lambda { \bf s}_q \right ] -\hat{\bf d } \times \frac{\delta f}{\delta\hat{\bf d } } \\ \dot{\hat{\bf d } } & = & \gamma\hat{\bf d}\times \left [ { \bf b } - \mu_0\gamma\frac{f_0^a}{\chi_{0}}{\bf s}_q -\mu_0\gamma \left(\frac{f_0^a}{\chi_{0}}+\frac{1}{\lambda\chi_0}\right ) { \bf s}_p \right ] . \end{aligned}\ ] ] except the spin - diffusion term @xmath197 , these equations are the same as in ref . . here
@xmath198 is the susceptibility in the absence of fermi - liquid effects .
as above , @xmath39 is the susceptibility in the normal state , so that @xmath199 . the function @xmath200 is defined as the equilibrium fraction of the superfluid magnetization , and is given by eq . ( 4.23 ) in ref . .
the leggett - takagi relaxation time @xmath201 describes local relaxation of @xmath194 and @xmath202 to their equilibrium values .
the spin diffusion appears via term @xmath203 in the equation for @xmath204 ( [ e.ltequ2a ] ) . in the normal state
the spin - diffusion constant @xmath205 is related to the spin - diffusion time @xmath206
by@xcite @xmath207 where @xmath208 is the fermi velocity .
we use this parametrization also in the superfluid state . note that @xmath205 is defined as the spin - diffusion constant for @xmath202 , so that the effective diffusion constant for the total magnetization @xmath42 is @xmath209 .
we continue by writing the equations for variables @xmath5 , @xmath210 , and the deviation from local equilibrium @xmath211 similarly , as in sec
. [ s.hyrdo ] , we linearize the equations and assume harmonic time dependence .
we use the fact that @xmath212 in solitons . as a consequence ,
the transverse and longitudinal modes separate at any field @xmath213 .
differing from sec.[s.hyrdo ] we introduce dimensionless quantities by defining @xmath214 and @xmath215 . for transverse oscillations
we get the equations [ e.eigvgen]@xmath216s^{\pm } - { { \rm i}}\frac{\kappa}{\omega_0}\nabla^2\eta^{\pm } \mp\frac{\omega^2}{\omega_0 ^ 2}e^{\pm { { \rm i}}\theta } ( \mathcal{d}+u_\perp)d_z \mp \frac{b_\pm}{b_0 } \label{e.eigvgena } , \\ \frac{\omega}{\omega_0}\eta^{\pm } & = & \left(\pm \frac{1}{1+f_0^a}-\frac{{{\rm i}}}{\omega_0\tau } + { { \rm i}}\lambda\frac{\kappa}{\omega_0}\nabla^2\right)\eta^{\pm } -{{\rm i}}\lambda(1-\lambda)\frac{\kappa}{\omega_0}\nabla^2 s^{\pm } \nonumber \\ & & \mp(1-\lambda)\frac{\omega^2}{\omega_0 ^ 2}e^{\pm{{\rm i}}\theta } ( \mathcal{d}+u_\perp)d_z \label{e.eigvgenb } , \\
\frac{\omega}{\omega_0}d_z & = & \frac{1}{2}\left(s^{-}e^{{{\rm i}}\theta}-s^{+}e^{-{{\rm i}}\theta}\right ) + \frac{\chi}{2\lambda\chi_0}\left(\eta^{-}e^{{{\rm i}}\theta } - \eta^{+}e^{-{{\rm i}}\theta}\right ) \nonumber \\ & & + \frac{1}{2b_0 } \left(b_{+}e^{-{{\rm i}}\theta}-b_{-}e^{{{\rm i}}\theta}\right).\end{aligned}\ ] ] the full solution of this problem can be written as a five - component vector @xmath217 , where the @xmath218 s satisfy the homogeneous equation where @xmath219 .
the eigenvalues @xmath220 of the homogeneous equation are now complex valued , and the eigenvectors @xmath218 are not orthogonal to each other . in order to solve for the absorption spectrum
we find that the following adjoint eigenvalue problem:@xcite [ e.eigvgen2 ] @xmath221 \underline{s}^{\pm } -{{\rm i}}\lambda(1-\lambda)\frac{\kappa}{\omega_0}\nabla^2 \underline{\eta}^{\pm } \mp\frac{1}{2}e^{\mp{{\rm i}}\theta}\underline{d}_z \\ \frac{\omega}{\omega_0}\underline{\eta}^{\pm } & = & \left(\pm \frac{1}{1+f_0^a}-\frac{{{\rm i}}}{\omega_0\tau } + { { \rm i}}\lambda\frac{\kappa}{\omega_0}\nabla^2\right)\underline{\eta}^{\pm } - { { \rm i}}\frac{\kappa}{\omega_0}\nabla^2 \underline{s}^{\pm } \mp\frac{\chi}{2\lambda\chi_0 } e^{\mp{{\rm i}}\theta}\underline{d}_z \\ \frac{\omega}{\omega_0}\underline{d}_z & = & \frac{\omega^2}{\omega_0 ^ 2}(\mathcal{d}+u_\perp)\left [ e^{-{{\rm i}}\theta}\underline{s}^{- } - e^{{{\rm i}}\theta}\underline{s}^{+ } + ( 1-\lambda)(e^{-{{\rm i}}\theta}\underline{\eta}^{- } - e^{{{\rm i}}\theta}\underline{\eta}^{+ } ) \right]\end{aligned}\ ] ] has the same eigenvalues @xmath220 as the homogeneous version of problem ( [ e.eigvgen ] ) and that the eigenvectors @xmath222 are orthogonal to @xmath223 s : @xmath224 when deriving this we have assumed zero derivate for the wave functions far from the soliton by considering a lattice of solitons as was discussed in sec . [ s.intes ] . the power absorption can now be written as @xmath225 , \label{e.powerabs}\end{aligned}\ ] ] where the coefficients @xmath226 are given by @xmath227 .
\end{aligned}\ ] ] we observe that the eigenfunctions are not symmetric with respect to the center of the soliton .
this apparently is caused by the spin - diffusion term . in the hydrodynamic limit where @xmath228 one may solve @xmath229 from eq .
( [ e.eigvgen ] ) and to linear order in @xmath201 @xmath230 if we additionally ignore the term @xmath231 , the equations for the three - component vector @xmath232 read [ e.eigvhydr ] @xmath216s^{\pm } \mp\frac{\omega^2}{\omega_0 ^ 2}e^{\pm { { \rm i}}\theta}(\mathcal{d}+u_\perp)d_z \mp \frac{b_\pm}{b_0 } , \\
\frac{\omega}{\omega_0}d_z & = & \frac{1}{2}e^{{{\rm i}}\theta } \left[1-\frac{\tau\chi(1-\lambda)\kappa}{\chi_0 } \nabla^2\right ] s^{- } -\frac{1}{2}e^{-{{\rm i}}\theta } \left[1-\frac{\tau\chi(1-\lambda)\kappa}{\chi_0 } \nabla^2\right ] s^{+ } \nonumber \\ & -&{{\rm i}}\frac{\tau\chi\omega^2(1-\lambda)}{\chi_0\omega_0\lambda } ( \mathcal{d}+u_\perp)d_z + \frac{1}{2b_0 } \left(b_{+}e^{-{{\rm i}}\theta}-b_{-}e^{{{\rm
i}}\theta}\right ) .\end{aligned}\ ] ] the homogeneous adjoint problem for this is given by [ e.eigvhydr2 ] @xmath221\underline{s}^{\pm } \mp \frac{1}{2}\left(1-\frac{\tau\chi(1-\lambda)\kappa}{\chi_0 } \nabla^2\right ) \left(e^{\mp{{\rm i}}\theta } \underline{d}_z\right ) , \\ \frac{\omega}{\omega_0}\underline{d}_z & = & \frac{\omega^2}{\omega_0 ^ 2}(\mathcal{d}+u_\perp ) \left ( e^{-{{\rm i}}\theta } \underline{s}^{- } - e^{{{\rm i}}\theta } \underline{s}^{+}\right ) - { { \rm i}}\frac{\tau\chi\omega^2(1-\lambda)}{\chi_0\omega_0\lambda } ( \mathcal{d}+u_\perp ) \underline{d}_z.\end{aligned}\ ] ] the equation for the power absorption does not need any modifications .
one must be careful since the validity region of the hydrodynamic approximation is not very large . for temperatures near @xmath1 one must , for typical values of @xmath201 , have @xmath233 mt . we mainly used the hydrodynamic approximation for checking our calculations in the limit of @xmath234 . for longitudinal case , where @xmath235 , we also define dimensionless quantities by writing @xmath236 , @xmath237 , and @xmath238 .
the equations of motions are [ e.eigvpargen ] @xmath239 and the homogeneous adjoint problem for this is given by [ e.eigvpargen2 ] @xmath240 .\end{aligned}\ ] ] power absorption for longitudinal case is given by @xmath241 where the @xmath226 s are the coefficients of the full solution @xmath242 and given by @xmath243 similar to the transverse case one could write the hydrodynamic equations using only @xmath43 and @xmath244 . the numerical solution for these different eigenvalue problems is obtained by dividing the calculation length @xmath245 , for example , to 1000 points and approximating the spatial derivatives by differences .
the eigenvalues and vectors of the resulting sparse 5000@xmath1235000 ( or in the longitudinal case 3000@xmath1233000 ) matrix @xmath246 are then solved by matlab .
we make use of sparse matrices and calculate normally only @xmath247 lowest eigenvalues with @xmath248 > 0 $ ] that give the main contribution to the absorption spectrum .
the complex mode frequencies @xmath249 are related to the reduced frequency shifts @xmath250 as given by eqs .
( [ e.freqll ] ) and ( [ e.freqperpgen ] ) .
additional parameters appear in the calculation compared to the dissipationless case .
the most crucial ones are the relaxation times @xmath201 and @xmath206 .
the leggett - takagi time @xmath201 can be extracted from the width of the bulk peak .
measurements of longitudinal and transverse resonances are in good agreement.@xcite we have reanalyzed the data by gully _
et al_.@xcite including strong - coupling corrections as described in sec .
[ s.princnmr ] and get the fit @xmath251(1+f_0^a)10^{-7}$ ] s in the range of @xmath252@xmath253 .
we fix @xmath254 . for the spin - diffusion coefficient
@xmath205 we use the value @xmath255 mk@xmath256 from refs . and .
this corresponds to the spin - diffusion time @xmath257 mk@xmath258 s.
a few absorption spectra are plotted in figs .
[ spectrum][parallelsplayabs ] . , 0.95 , and 0.97 when @xmath259 , @xmath260 mt , @xmath251(1+f_0^a)10^{-7}$ ] s , and @xmath257 mk@xmath258 s. other parameters are obtained using weak coupling and pressure of 3.4 mpa with @xmath105 and @xmath254.,width=302 ] and @xmath261 mt ( solid line ) .
other parameters are the same as in fig .
[ spectrum ] .
the effect of the two dissipation mechanisms is demonstrated by setting @xmath262 ( dash - dotted line ) and @xmath263 ( dashed line).,width=302 ] ( solid line ) .
there is no dependence on the magnetic field , and other parameters are the same as in fig.[spectrum ] .
the effect of the two dissipation mechanisms is demonstrated by setting @xmath262 ( dash - dotted line ) and @xmath263 ( dashed line).,width=302 ] the effect of the normal - superfluid relaxation seems to be simple line broadening . in the transverse resonance
the linewidth measured on the @xmath113 scale is inversely proportional to the magnetic field @xmath213 for @xmath264 .
this is caused by @xmath265 term in eq .
( [ e.eigvgenb ] ) .
there is no field dependence in the longitudinal resonance since eqs .
( [ e.eigvpargen ] ) do not contain @xmath107 . in the longitudinal case
the linewidth on the @xmath125 scale has no strong temperature dependence and approaches a finite constant when @xmath156 . in the transverse case
the linewidth on the @xmath125 scale vanishes as @xmath156 .
all these characteristics are the same as predicted for the bulk peaks .
the effect of spin diffusion is more complicated .
in addition to broadening , it shifts the principal soliton peak to higher frequencies .
since the spin diffusion is due to inhomogeneous texture , its strength in the transverse case is obtained by comparing the term @xmath266 with the term @xmath267 in eq .
( [ e.eigvgena ] ) .
therefore , the effective diffusion constant @xmath268 is linearly proportional to the magnetic field . as a consequence also the reduced frequency @xmath113 of the satellite peak is field dependent . when @xmath269 mt the effect of diffusion is already quite large at @xmath191 and the soliton satellite peak is almost smeared out . in the longitudinal case
there is no field dependence and the effective diffusion constant reduces to @xmath270 . in both transverse and longitudinal cases the effective diffusion coefficients @xmath271 and @xmath272 diverge when @xmath156 , as @xmath273 . the relative contributions of normal - superfluid relaxation and spin diffusion are illustrated in figs .
[ spectrumhighh0 ] and [ parallelsplayabs ] , where both are separately shut off by setting @xmath262 or @xmath263 .
it can be seen that the broadening of the longitudinal satellite peak is mostly caused by normal - superfluid conversion at @xmath274 . in the transverse case at high field
the spin diffusion becomes the dominant relaxation mechanism , as well as approaching @xmath1 . in uniform order parameter
the width of the bulk peak comes solely from normal - superfluid relaxation .
our soliton lattice has many higher peaks in the absence of dissipation , but dissipation seems to remove them . note that the higher peaks are mostly suppressed by spin diffusion in the case of fig.[spectrumhighh0 ] but by normal - superfluid relaxation in the case of fig .
[ parallelsplayabs ] .
the relatively high field in fig . [ spectrumhighh0 ] has been chosen to allow a comparison to the spectrum of hakonen _
et al_.@xcite the shape of the twist - soliton spectra are very similar even if the temperature is quite low , where our modeling of the spin diffusion is under suspect .
further spectra at lower field are shown by parts _
et al_.@xcite the comparison to ref . is a different case . in this experiment
such a dense soliton lattice was created that no bulk peak was observable .
et al_.@xcite and gould and lee@xcite show soliton spectra measured by sweeping temperature . in all cases
a proper comparison to our calculations would require first estimation of the density of solitons and then correcting our calculations for that density , which we have not done .
what we can state that there seems to be no obvious contradictions between theory and experiment concerning the linewidth .
a summary of our results at @xmath275 mt is shown fig .
[ r2spindiff ] .
the solid lines give the frequency of maximum satellite absorption , the dash - dotted lines 75% , and dotted lines 50% of the maximum .
the dashed lines give resonance frequency in the absence of dissipation .
other parameters are @xmath276 , @xmath251(1+f_0^a)10^{-7}$ ] s , @xmath257 mk@xmath258 s , pressure 3.4 mpa ( @xmath104 , @xmath105 , and @xmath254 and weak coupling ) .
the experimental points are the same as in fig .
[ r2].,width=302 ] we see that the frequency of maximum absorption is shifted towards higher frequencies , especially at temperatures near @xmath1 .
this alleviates the disagreement found between dissipationless theory and the experiment by gould and lee@xcite in the transverse mode of splay soliton .
however , the experimental data does not show any sign of the divergence at @xmath1 predicted by our model .
the shift of the resonance frequency results from spin diffusion .
for the longitudinal mode it is field independent , but for the transverse mode it increases with increasing field . in high field
this shift can be substantial even at @xmath277 , see fig.[spectrumhighh0 ] .
this poses a problem since no field - dependent shift has been observed experimentally .
this is the case particularly for the transverse mode of twist soliton measured at @xmath278 mt ( @xmath122 ) ( ref . ) and 28 mt ( @xmath121 ) ( ref . ) .
similar evidence for transverse mode of splay soliton measured at @xmath279 mt ( @xmath118 ) ( ref . ) and at @xmath280 mt ( @xmath119 ) ( ref . )
is not so clear since the latter data is measured under different conditions , as discussed above .
the frequencies of soliton satellite peaks in the nmr spectrum are calculated at all temperatures .
the agreement of dissipationless theory with experiments is very good .
however , there is a small difference in the transverse - mode frequency of a splay soliton near @xmath1 .
this difference is partly explained by taking into account spin diffusion .
the spin diffusion shifts up the reduced frequencies @xmath125 at high fields , which has not been observed experimentally .
we hope that new experiments could clarify this problem .
we also point out the relatively narrow longitudinal line of the splay soliton , which has not been studied experimentally .
it would be of interest to extend the present calculations to vortex lines in @xmath0he - a .
this requires two - dimensional calculation and thus would be computationally much more demanding than the present one .
we thank pete sivonen for contribution in early stages of this work .
this research was supported by the academy of finland and the visl foundation .
o. avenel , m.e .
bernier , e.j .
varoquaux , and c. vibet , in _ proceedings of the 14th international conference on low temperature physics _ , edited .
m. krusius and m. vuorio ( north - holland , amsterdam , 1975 ) , vol . 5 , p. 429 .
leggett , j. phys .
c * 6 * , 3187 ( 1973 ) .
d. vollhardt and k. maki , phys rev .
b * 20 * , 963 ( 1979 ) .
. parts , v.m.h .
ruutu , j.h .
koivuniemi , m. krusius , e.v .
thuneberg , and g.e .
volovik , physica b * 210 * , 311 ( 1995 ) .
|
superfluid @xmath0he - a has domain - wall - like structures , which are called solitons .
we calculate numerically the structure of a splay soliton .
we study the effect of solitons on the nuclear - magnetic - resonance spectrum by calculating the frequency shifts and the amplitudes of the soliton peaks for both longitudinal and transverse oscillations of magnetization . the effect of dissipation caused by normal - superfluid conversion and spin diffusion is calculated .
the calculations are in good agreement with experiments , except a problem in the transverse resonance frequency of the splay soliton or in magnetic - field dependence of reduced resonance frequencies .
|
the objective of this paper is a critical review of the identification of light mesons with @xmath8 = @xmath7 and @xmath2 , particularly the separation of @xmath7 states into @xmath9 and @xmath4 .
that detail is important in its own right .
it will also be vital input information into the search for the @xmath7 glueball in radiative @xmath10 decays .
it is unrealistic that those data will be extensive enough to sort out the complex spectroscopy of @xmath11 states as well as glueballs .
section 2 reviews @xmath7 states .
they fall into a regular scheme except for @xmath5 , whose identification is presently weak .
section 3 reviews @xmath2 states and makes the case that @xmath5 may have been confused with @xmath6 . without an @xmath12 in that mass range ,
the @xmath2 spectrum is obviously incomplete , but @xmath6 fits in naturally .
section 4 draws conclusions and comments on improvements needed in present partial wave analyses to resolve this question ; these are straightforward and can be done with existing data and/or forthcoming bes iii data .
there are extensive data from crystal barrel on @xmath14 scattering in flight to 17 final states .
these concern purely neutral final states which can be classified into four families with isospin @xmath0 and 1 , and charge conjugation @xmath15 and @xmath16 .
we shall concentrate here on @xmath0 , @xmath15 where the data are most complete .
this is a formation experiment of the type @xmath17 .
these data are listed by the pdg under ` further states ' , requiring confirmation @xcite .
that is not possible for most states because other data come from production reactions of the form @xmath18 ( or n ) , where the exchanged meson is uncertain , leading to ambiguities in partial wave analysis ; also they have no polarisation data .
a combined analysis has been published @xcite of 8 sets of data with @xmath0 , @xmath19 on @xmath20 , @xmath21 , @xmath22 , and @xmath23 from crystal barrel at 9 beam momenta , plus measurements of differential cross sections and polarisation from the ps172 experiment @xcite and an earlier experiment of eisenhandler et al . at the cern ps @xcite .
the first set of @xmath24 data covers the momentum range 0.361.55 gev / c and the second covers 12.5 gev / c .
they agree accurately where they overlap .
the data of ps172 are particularly valuable because they extend down to a beam momentum of 360 mev / c ( a mass of 1910 mev ) in quite small steps of momentum ( 30100 mev / c ) and therefore cover in detail the lower side of the cluster of resonances around 2000 mev ; they also used a beam going through the detector and therefore cover centre of mass scattering angles to @xmath25 .
two further analyses were reported of crystal barrel data for @xmath26 @xcite and @xmath27 @xcite .
they find masses and widths consistent within errors with the combined analysis .
the polarisation data are very important . because @xmath28 and @xmath29 have orthogonal clebsch - gordan coefficients , they separate those partial waves accurately .
differential cross sections contain real parts of interferences within singlet and triplet sets of amplitudes ; polarisation data measure the imaginary part of interferences within the triplet set .
formulae are given in ref .
column 5 of table 1 gives the ratio @xmath30 of amplitudes , where @xmath31 are coupling constants for @xmath14 orbital angular momentum @xmath32 ; intensites depend on @xmath33 .
the phase sensitivity of the polarisation data improves greatly the accuracy of masses and widths .
the only singlet states with @xmath0 , @xmath34 are @xmath35 , @xmath36 and @xmath37 and these states are separated by their angular dependence .
the analysis relies on fitting with analytic functions of @xmath38 and assumes two towers of states in mass ranges 1910 - 2100 and 2200 - 2370 mev ; it includes blatt - weisskopf centrifugal barrier factors with a radius which optimises at @xmath39 fm .
the tails of @xmath40 and @xmath41 are included using masses and widths determined elsewhere .
it is fortunate that @xmath42 states near 2050 and 2300 mev are strong and accurately determined by their rapid angular dependence , and act as powerful interferometers to determine lower partial waves . starting from these partial waves and adding lower @xmath43 , the analysis finds a unique set of amplitudes ; only in two low partial waves with little or no angular dependence are there sizable errors in fitted masses and widths . recently , as a convenience , we have installed the relevant publications on the arxiv system , and give references in the bibliography .
further details and figures of data are given in a full length review @xcite .
there is a total of @xmath44 million fully reconstructed events .
the data and monte carlo sets are publicly available from the authors of this paper , subject to a joint publication of results .
a complete set of data is also available on @xmath14 annihilation at rest in liquid hydrogen and deuterium and in gas for both .
this makes one of the largest data sets available in meson spectroscopy . ccccccc state & mass & width & @xmath4 mixing & @xmath45 & observed + & @xmath46 & @xmath46 & angle ( deg ) & @xmath47 & channels + + @xmath48 [ 9 ] & @xmath49 & @xmath50 & 0 & 0 & @xmath51,@xmath52 + @xmath53 [ 9 ] & @xmath54 & @xmath55 & 0 & 0 & @xmath51,@xmath21,@xmath56 + @xmath57 [ 2,7 ] & @xmath58 & @xmath59 & 1.1 & @xmath60 & @xmath61,@xmath21,@xmath62,@xmath63 + @xmath64 [ 2,7 ] & @xmath65 & @xmath66 & 7.9 & @xmath67 & @xmath61,@xmath21,@xmath22 , @xmath68 + @xmath69 [ 2,7 ] & @xmath70 & @xmath71 & 7.5 & @xmath72 & @xmath61,@xmath21,@xmath22 , @xmath68 + @xmath73 [ 2,7 ] & @xmath74 & @xmath75 & -14.8 & @xmath76 & @xmath61,@xmath21,@xmath22 , @xmath68,@xmath63 + @xmath77 [ 10 ] & @xmath78 & @xmath79 & & & @xmath80 + @xmath77 [ 11 ] & @xmath81 & @xmath82 & & & @xmath21 + @xmath83 [ 12 ] & @xmath84 & @xmath85 & & & @xmath80 + @xmath86 [ 1 ] & @xmath87 & @xmath88 & & & @xmath21 , @xmath89 + @xmath90 [ 13 ] & @xmath91 & @xmath92 & & & @xmath93 , @xmath94 + @xmath95 [ 13 ] & @xmath96 & @xmath97 & & & @xmath98 , @xmath94 + table 1 lists @xmath0 , @xmath1 states . the top half of the table lists states which are dominantly @xmath3 and the lower half @xmath4 states .
masses and widths are from crystal barrel where available .
there is also evidence for a broad @xmath99 listed by the pdg as @xmath100 ; in crystal barrel data it appears in the @xmath21 channel with a mass of @xmath101 mev and a width of @xmath102 mev .
it is observed by other groups in @xmath61 , @xmath21 , @xmath52 , @xmath103 and @xmath104 .
it is a candidate for the @xmath105 glueball .
another possibility is that it is a dynamically generated state related to the opening of the strong @xmath52 and @xmath104 thresholds .
there is a publication concerning crystal barrel data in flight which determines the mixing angle between @xmath3 and @xmath106 using data on @xmath107 , @xmath21 and @xmath108 @xcite . to our knowledge , these are the only data making a clean identification of @xmath3 and @xmath4 and their mixing .
resonances @xmath109 are linear combinations : @xmath110 the quark content of @xmath111 and @xmath112 may be written in terms of the pseudoscalar mixing angle @xmath113 as latexmath:[\[\begin{aligned } amplitudes for decays of @xmath109 are given by @xmath115 where @xmath116 was taken as 0.6 and @xmath117 as 0.85 .
values of the @xmath118 mixing angle @xmath119 are listed in column 4 of table 1 .
errors are not listed in the original publication , but are quoted as typically @xmath120 .
each mixing angle of table 1 is consistent with zero within three standard devations , though from the overall @xmath121 there is a a definite indication that some small mixing with @xmath4 occurs ; that is to be expected from @xmath118 states across the mass range .
the conclusion is that none of the states listed in the first half of the table is dominantly @xmath4 . amongst recognised or possible
@xmath4 states , a small signal from @xmath77 can be detected in crystal barrel data for @xmath93 and @xmath21 via its interference with @xmath122 @xcite , @xcite .
the @xmath123 appears as a dip in the dalitz plot for @xmath124 , but with a much smaller branching ratio compared with the @xmath125 amplitude than in @xmath61 elastic scattering , see figs 2 and 4(a ) of ref .
. otherwise , recognised @xmath4 states are conspicuous by their absence in @xmath14 annihilation .
the @xmath90 and @xmath95 are observed by etkin et al . in @xmath93 and
@xmath94 @xcite .
the initial state for those data is @xmath61 , but no @xmath61 decays are observed , showing that any @xmath126 coupling in these states , hence @xmath3 component , must be small .
presently , the pdg lists @xmath69 under @xmath90 .
that is clearly inconsistent with the mixing angles of table 1 and needs to be corrected in pdg tables .
the @xmath73 is missing from the tables and needs to be included , since it is observed in five channels of data : @xmath61 , @xmath21 , @xmath22 , @xmath127 and @xmath128 ; dropping it from the @xmath61 channel alone increases @xmath121 by 2879 , which is highly significant .
figures 9(n ) and 9(o ) of that paper illustrate the effects of dropping them from the analysis and re - optimising all other components ; the changes near @xmath129 are very large , leaving no doubt of their significance . visual evidence for @xmath69 and @xmath73 are also displayed in the first analysis of @xmath130 data in ref .
16 of that paper displays the requirement for @xmath131_{l=1}$ ] , where @xmath32 is the angular momentum in the decay . in the final combined analysis of ref .
@xcite , log likelihood is worse by 468 if @xmath69 is omitted in the @xmath132 channel and by 1557 if @xmath73 is omitted ; here log likelihood is defined so that it increases by 0.5 for a one standard deviation change in each coupling constant .
there is no convincing evidence in crystal barrel data for @xmath86 , which is observed by other groups only in @xmath21 and @xmath93 .
the @xmath86 may be interpreted as the @xmath4 @xmath28 partner of @xmath57 .
their mass difference is similar to that between @xmath48 and @xmath77 .
the @xmath90 and @xmath95 are observed by etkin et al . in @xmath93 and @xmath133
s and d - waves . both @xmath90 and @xmath95
may be interpreted as the partner of the @xmath29 @xmath3 state at 2001 mev , with decays to @xmath94 s and d - waves ; the l=2 dependence makes the d - wave peak higher . however , the 150 mev mass gap between @xmath86 and @xmath90 is a little surprising compared with the 90 mev gap between @xmath57 and @xmath64 . under @xmath86 ,
the pdg lists anisovich 99k data on @xmath134 @xcite as presenting evidence that there is a state at @xmath135 mev consistent with the @xmath86 .
in fact , the paper presented a careful study of both the angular distribution , which is flat , and the energy dependence of production ( which is different for production of @xmath136 and 0 ) ; the conclusion from both sources is that the signal is due to @xmath137 , which is observed in many sets of data and unambiguously has @xmath138 .
it is the only state observed in crystal barrel data with a large mixing angle @xmath139 to @xmath4 .
it makes up @xmath140 of the @xmath141 intensity and @xmath142 of @xmath21 .
the branching ratio to @xmath143 is not well determined because of low statistics in this channel .
the best estimate of amplitude ratios is @xmath144 for an unmixed @xmath11 state , the ratio expected between @xmath145 and @xmath21 is @xmath146 .
a possible interpretation is that it is an @xmath4 state mixed with @xmath147 .
an alternative is that it is the second @xmath148 glueball predicted by morningstar and peardon in this general mass range @xcite .
a pointer in this direction is that it was first identified in mark iii data for @xmath149 @xcite .
a pure glueball would have a mixing angle of @xmath150 .
its strong coupling to @xmath14 is clearly anomalous .
we are able to check the partial wave analyses of @xmath151 done by hasan et al .
@xcite and oakden and pennington @xcite . because their analyses were limited to this channel alone ,
errors on mass and width are larger than from the full analyis by a factor @xmath152 .
this explains why their results have larger errors and fluctuations .
our experience is that , as a rule of thumb , each data set with large statistics reduces errors by a factor 2 because different sets of data have different sensitivity to details .
the pdg attributes the @xmath153 reported by hasan @xcite as @xmath86 .
that is a mistake .
it should be attributed to @xmath69 .
-11.5 mm -5.5 mm the outcome of this analysis is that states observed in the crystal barrel analysis for the four sets of quantum numbers @xmath0 , @xmath154 and @xmath155 , @xmath156 fall on to parallel trajectories displayed in figs . 1 and 3 of ref .
those with @xmath0 , @xmath15 are shown here in fig .
1 . they are particularly well identified because of the availability of the polarisation data .
some states are significantly displaced from straight line trajectories by thresholds .
a striking example is @xmath53 which coincides with the @xmath157 threshold and is displaced downwards from its isospin partner @xmath158 by @xmath159 mev .
the origin of this shift is a narrow cusp in the real part of the amplitude at the opening of any sharp threshold , as explained in ref .
@xcite .
the experimental data separate @xmath14 @xmath29 and @xmath28 states . these can be interpreted as @xmath29 and @xmath28 @xmath11 configurations for the following reasons .
a feature of the data is that @xmath160-states decay strongly to channels with high orbital angular momentum .
the origin of this effect is clearly a good overlap between initial and final state wave functions .
llanes - estrada et al .
point out a formal analogy with the frank - condon principle of molecular physics consistent with this observation @xcite .
in essence , this analogy provides a mechanism via which @xmath3 @xmath29 states couple preferentially to @xmath14 @xmath29 states and likewise @xmath28 @xmath3 states couple preferentially to @xmath14 @xmath28 states .
it is observed that @xmath29 states lie systematically higher in mass than @xmath28 by @xmath161 mev , see fig .
1 . for @xmath155 , @xmath156 ,
d - states lie roughly midway in mass .
these observations are consistent with stronger centrifugal barriers in @xmath160-states delaying the appearance of @xmath160-wave resonances to higher masses .
the spin - splitting between @xmath160-states is consistent within errors with tensor splitting , which is predicted to be dominantly from one - gluon exchange between @xmath11 @xcite .
the @xmath42 states have some admixtures of @xmath162 ; the @xmath163 state does not appear until @xmath164 mev @xcite .
the overall picture is the first appearance of the lowest - lying @xmath160-states at @xmath165 mev .
this is what is to be expected from regge trajectories for @xmath3 states .
there are other candidates for @xmath7 states listed by the pdg .
firstly , the @xmath167 has a simple explanation .
it is the @xmath168 decay mode of @xmath53 @xcite .
the latter state sits precisely at the @xmath169 threshold .
the @xmath167 has a line - shape which is well fitted by folding @xmath169 s - wave phase space with the line - shape of @xmath53 .
it is fitted like that by baker et al .
, including the dispersive term which originates from the opening of the @xmath169 threshold .
the square of the coupling constant to @xmath170 is 3 times that for @xmath169 by su(2 ) symmetry .
the result is that @xmath170 decays of @xmath53 also peak at @xmath171 mev , but the peak is broader than for @xmath169 when the widths of the two @xmath172 are folded in , see fig .
5(b ) of ref . @xcite .
the @xmath173 listed by the pdg has an explanation , illustrated in fig .
10 of ref .
@xcite . below the @xmath169 threshold ,
the breit - wigner denominator of @xmath53 needs to include an analytic continuation of the phase space factors below both @xmath170 and @xmath169 thresholds .
the analytic continuation causes a phase variation in @xmath61 and leads to an interference between the @xmath53 and @xmath48 .
this interference is very clear for the @xmath174 d - wave in crystal barrel data for @xmath14 at rest @xmath175 .
the effect is maximal at a mass of 1420 mev .
similar interferences in other channels listed under @xmath173 may be explained this way .
it was a good observation by the experimental groups before the existence of the @xmath53 was well known .
the @xmath176 listed by the pdg has a simple explanation . the peak observed in the data of etkin et al .
@xcite is at 2150 mev and agrees with the @xmath86 .
it is not necessary to have two separate @xmath99 states at 2010 and 2150 mev .
the partial wave analysis of etkin et al . used the k - matrix approach .
it is possible to have a k - matrix pole at the @xmath94 threshold , significantly displaced from the t - matrix pole at 2150 mev .
this leaves the @xmath5 .
it does not fit well on to the @xmath105 trajectories shown in fig .
1 ; it appears to be an ` extra ' state .
dudek @xcite has recently presented a lattice qcd calculation of hybrid masses and light mesons .
he predicts a lowest group of hybrids with @xmath178 , @xmath179 and @xmath180 .
his mass scale needs to be normalised against the well known @xmath181 and the @xmath182 @xmath183 states .
it then agrees quite well with the exotic @xmath184 ( actually at 1660 mev ) , the @xmath185 and the two ` extra ' @xmath180 states @xmath186 and @xmath187 , which do not fit into the @xmath36 trajectory of fig .
1(c ) above .
no @xmath7 hybrid is predicted in this mass range .
this prompts a careful re - examination of the data on which @xmath5 is based .
the @xmath5 is not well established .
also there is clear evidence for an @xmath9 @xmath188 state @xmath6 very close to this mass , distinct from @xmath189 .
we shall consider possible confusion between @xmath190 and 2 for this state . the primary data for @xmath5 come from the gams collaboration @xcite .
these are mostly on @xmath191 . in these data ,
the separation between @xmath192 and @xmath148 rests on the number of events observed below and above @xmath193 , where @xmath194 is defined in their paper . in practice
this is a rather fine distinction .
the argument is that @xmath99 events are enhanced at small @xmath195 and @xmath12 events are enhanced at large @xmath196 . that is counter - intuitive and
must depend strongly on the monte carlo of acceptance which is quoted but not shown
. it would be valuable if the compass collaboration could check these results in @xmath197 and in @xmath198 if that is possible .
data of costa et al .
@xcite are also quoted , but these refer to @xmath199 .
there are eight alternative solutions , which mostly contain some mild peaking near 1800 mev .
these could be explained by the @xmath83 of the l3 collaboration .
the pdg quotes also an @xmath200 fitted by longacre to data available in 1986 @xcite . at that time
, @xmath57 had not been discovered .
if it had been known at the time , there would undoubtedly have been some perturbation to longacre s analysis .
finally , the observation of a structure at 1799 mev in @xmath201 by cason et al . was not confirmed by prokoshkin et al @xcite .
cason et al . argue that their charge exchange data choose a unique solution from four ambiguous solutions to @xmath202 elastic scattering .
it is however somewhat puzzling that their solution does not contain any significant signal for @xmath122 , which ought to be conspicuous .
neither do they observe the @xmath53 which is conspicuous in the @xmath61 channel ; indeed , that is where it was first observed by the asterix collaboration @xcite .
the alternative assignment for the 1810 mev state is @xmath203 . here
, important data come from bes ii for @xmath10 decays
. there are six relevant sets of data .
the first two are for @xmath204 @xcite and @xmath205 @xcite .
a large @xmath189 signal is observed in @xmath206 .
in contrast , high statistics data on @xmath207 show no structure in @xmath24 at this mass .
those data set an upper limit of @xmath208 on the branching ratio @xmath209 of @xmath189 with @xmath210 confidence .
the third and fourth sets of data are for @xmath211 and @xmath212 @xcite . in the @xmath213 data
there is a definite @xmath188 @xmath61 peak , but at @xmath214 mev , visibly distinct from the mass of @xmath189 .
there is no definite evidence for a decay to @xmath93 , though a small amount can be fitted .
there is a factor 2225 discrepancy between the branching ratio @xmath215 for these data and the data for @xmath216 and @xmath217 , so this peak can not be due to @xmath189 .
the discrepancy points strongly to the existence of a second @xmath9 state distinct from @xmath6 .
-10 mm -3 mm a question is whether these data can alternatively be fitted by @xmath5 .
the observed decay angular distribution , shown here in fig .
2 , is flat within experimental errors in the region where the acceptance of the bes ii detector ( dotted histogram ) is uniform .
however , the acceptance falls rapidly at @xmath218 .
the bes publication comments that the @xmath5 can be produced with orbital angular momentum @xmath219 , 2 or 4 in the production step .
the @xmath219 component is likely to be dominant and gives a decay angular distribution proportional to the legendre polynomial @xmath220 , where @xmath221 is the decay angle of the @xmath222 from @xmath223 in the resonance rest frame . on resonance , the @xmath224 and @xmath223
are produced with momenta of 630 mev / c in the lab frame , and the centrifugal barrier reduces @xmath225 amplitudes by a factor @xmath226 and the @xmath227 amplitude by a factor @xmath228 , so the @xmath227 amplitude may safely be neglected . for @xmath225 , there are three combinations of @xmath225 with spin @xmath229 of the @xmath230 , making total spins @xmath231 , 1 and 2 .
of these , spin 1 makes two amplitudes proportional to @xmath232 and @xmath233 , but no amplitude proportional to @xmath234 .
the @xmath231 amplitude is proportional to @xmath235 .
the @xmath236 amplitude is formed from @xmath236 and the spin 2 of @xmath230 and contains a @xmath234 term .
the publication says : ` our experience elsewhere is that using four helicity amplitudes instead of two adds considerable flexibility to the fit .
we conclude that the state is most likely spin zero . '
.changes in log likelihood with a variety of @xmath230 amplitudes fitted to bes ii data on @xmath237 and @xmath238 . [ cols="^,^,^",options="header " , ] it is now worth amplifying this comment with numbers in table 2 from the analysis ; these may be understood in terms of the acceptance .
each amplitude is fitted freely in magnitude and phase .
angular correlations with the decay of the @xmath224 are included .
the fit to @xmath6 produced with both @xmath219 and 2 amplitudes is taken as a benchmark .
further entries in the table show changes in log likelihood ( defined so that a change of + 1 is better by one standard deviation for two degrees of freedom ) . the @xmath225 amplitude changes sign at @xmath239 . in entry
a , changes in log likelihood are shown for four single @xmath99 amplitudes .
entry b shows the best three pairs of @xmath99 amplitudes and entry c the best combination of 3 . in a
, the best fit with @xmath219 is considerably worse than the benchmark , but uses the @xmath235 dependence to produce a fit peaking at @xmath240 and dropping sharply at @xmath241 , though it gives a false peak near @xmath242 .
the fit with @xmath231 requires strong correlations between the production angle and decay angle and is considerably worse .
the fit with @xmath243 is bad because two of the contributions go to zero in the middle of the angular distribution . in b ,
the best fit is a combination of @xmath244 and @xmath245 with @xmath243 .
the second of these helps produce the box - shaped distribution of data on fig . 2 and is able to compensate @xmath234 to some extent near @xmath242 .
in c , three amplitudes can produce nearly as good a fit as @xmath6 .
there is considerably flexibility using three fitted phase angles ; random phases give much worse fits .
it is of course possible that @xmath5 is produced with a set of amplitudes which happen to agree with @xmath6 .
further bes iii data with considerably improved acceptance and higher statistics have a high chance of resolving the situation .
the bes ii collaboration also present data on @xmath246 @xcite .
there is a clear peak in @xmath247 at @xmath248 mev .
quantum numbers @xmath249 are favoured quite significantly over @xmath7 and @xmath250 .
the @xmath247 channel opens at 1802 mev .
the present data may be fitted within the sizable errors by folding the line shape of @xmath6 with @xmath247 phase space , using a reasonable form factor @xmath251 , where @xmath252 is the momentum in the @xmath247 channel in gev / c .
the pdg lists this state under @xmath253 , observed in @xmath254 @xcite @xcite .
however , as the pdg remarks , @xmath138 is not allowed for this final state .
the angular distribution of the photon for those data is consistent with @xmath255 , but might be accomodated with @xmath256 if the helicity ratio of the two possible @xmath256 amplitudes is just right . but
neither of these possibilities is consistent with the observed peak presently attributed to @xmath6 .
the @xmath257 decays to @xmath258 .
there is an important simplification that in radiative production of @xmath257 there are only three @xmath259 helicity amplitudes instead of five because helicity 0 is forbidden for the photon .
an analysis of the spin correlation between these two would identify @xmath43 of @xmath257 .
the spin of the @xmath224 is measured by @xmath260 , where @xmath261 and @xmath262 are momenta of the kaons from its decay in its rest frame ; the spin of the @xmath263 is normal to the decay plane of the @xmath263 in its rest frame . for spin 0 ,
the angular distribution of @xmath224 and @xmath263 decays is given by the dot product of these two vectors .
this is a delicate test of the spin of the @xmath257 .
formulae for other @xmath43 are given by zou and bugg @xcite .
the decay to @xmath258 is surprising ( okubo - zweig - iizuka violating ) .
it could arise from a glueball component mixed into @xmath257 .
two gluons couple to @xmath264 .
the cross - terms between @xmath265 and @xmath4 can generate @xmath247 .
the @xmath266 would fit naturally on to the @xmath2 trajectory of fig .
1 . it would not be surprising that there are two @xmath148 states close in mass ; a similar pair is @xmath77 and @xmath267 . there is earlier independent evidence for an @xmath268 in mark iii data for @xmath269 @xcite .
there is a further observation of a well defined @xmath2 signal in @xmath21 in crystal barrel data in flight at @xmath270 mev with width @xmath271 mev @xcite .
the mass is 4 standard deviations above @xmath189 and the width is 2 standard deviations higher .
this signal could come from a superposition of @xmath6 with @xmath189 , which clearly has a large @xmath4 component ( and/or glueball ) .
we have made a case that existing light mesons with @xmath259 fall into a regular pattern of @xmath3 @xmath28 and @xmath29 states except for @xmath5 .
there is the possibility that it has been confused with @xmath6 ; if not , there is a missing @xmath148 state on the trajectory of fig . 1 at a similar mass .
the pattern of @xmath189 and @xmath6 is like that of @xmath77 and @xmath53 .
we remark that the search for the @xmath7 glueball will require full use of existing identifications of @xmath11 and @xmath4 components with these quantum numbers .
we have pointed out some corrections to pdg tables .
we also remark that , in the long term , it would be possible to do further polarisation measurements in @xmath14 scattering in the beam momentum range from @xmath272 to 1940 mev / c at the forthcoming fair facility @xcite .
this was part of the proposed program at lear , but was cut short by the closure of that machine .
measurements of polarisation in @xmath130 and @xmath273 are realistic and would give information on interferences between singlet and triplet partial waves .
for @xmath155 , @xmath34 and @xmath0 , @xmath274 , there are presently no polarisation data . such data would improve vastly the identification of states with these quantum numbers .
99 k. nakamura _
et al . _ , j. phys .
g * 37 * ( 2010 ) 075021 .
et al . _ ,
b * 491 * ( 2000 ) 47 , arxiv : 1109.0883 .
a. hasan _ et al .
_ , nucl . phys .
b * 378 * ( 1992 ) 3 .
e. eisenhandler _ et al .
* b * 109 ( 1975 ) 98 .
et al . _ ,
b * 491 * ( 2000 ) 40 , arxiv : 1109.2092 .
et al . _ ,
b * 496 * ( 2000 ) 145 , arxiv : 1109.4008 .
et al . _ ,
a * 662 * ( 2000 ) 319 , arxiv : 1109.1188 .
bugg , phys . rep .
* 397 * ( 2004 ) 257 , arxiv : hep - ex/0411042 .
v.v . anisovich and a.v .
sarantsev , int .
phys . a * 24 * ( 2009 ) 2481 . c. amsler
_ et al . _ ,
b * 639 * ( 2006 ) 165 . c. amsler _ et al .
j. c * 23 * ( 2002 ) 29 . v. a. schegelsky _
et al . _ ,
( l3 collaboration ) eur . phys .
j. a * 27 * ( 2006 ) 207 .
et al . _ ,
b * 201 * ( 1988 ) 568 .
et al . _ ,
b * 449 * ( 1999 ) 154 , arxiv : 1109.4819 .
et al . _ , phys .
b * 323 * ( 1994 ) 233 .
a * 651 * ( 1999 ) 253 , arxiv : 1109.2086 .
morningstar and m. peardon , phys .
d * 60 * ( 1999 ) 034509 .
b * 353 * ( 1995 ) 378 .
a. hasan and d.v .
bugg , phys . lett .
b * 334 * ( 1994 ) 215 .
oakden and m.r .
pennington , nucl .
phys . a * 574 * ( 1994 ) 731 .
bugg , j. phys .
g * 35 * ( 2008 ) 075005 .
f. llanes - estrada _
_ , aip conf
. proc . * 1030 * , ( 2008 ) 171 , arxiv : 0803.0806 .
s.godfrey and n. isgur , phys .
d * 32 * ( 1985 ) 189 .
et al . _ , phys .
b * 467 * ( 1999 ) 147 , arxiv : 1109.2287 .
bugg , eur .
j. c * 52 * ( 2007 ) 55 , arxiv : hep - ex/0706.1341 .
d. alde _ et al .
b * 198 * ( 1987 ) 286 .
dudek , ( 2011 ) preprint , arxiv : 1106 .
. g. costa de beauregard _ et al .
b * 175 * ( 1980 ) 402 .
longacre , phys . lett .
b * 177 * ( 1986 ) 223 .
et al . _ ,
* 42 * ( 1997 ) 117
. m. may _ et al .
_ , ( asterix collaboration ) phys
b * 225 * ( 1989 ) 450 .
m. ablikim _
et al . _ , ( bes ii collaboration ) phys .
b * 603 * ( 2004 ) 138 , arxiv : hep - ex/0409007 .
m. ablikim _
et al . _ , ( bes ii collaboration ) phys .
b * 598 * ( 2004 ) 149 .
m. ablikim _
, ( bes ii collaboration ) phys . lett .
b * 607 * ( 2005 ) 243 , arxiv : hep - ex/0411001 .
m. ablikim _
et al . _ , ( bes ii collaboration ) phys .
* 96 * ( 2006 ) 162002 , arxiv : hep.ex/0602031 m. ablikim _
, ( bes ii collaboration ) phys . rev . lett . * 95 * ( 2005 ) 262001 , arxiv : hep - ex/0508025
. m. ablikim _
et al . _ , ( bes iii collaboration ) phys .
106 ( 2011 ) 072002 , arxiv : 01012.3510 .
zou and d.v .
bugg , eur .
j c * 16 * ( 2003 ) 537 .
b * 449 * ( 1999 ) 154 , arxiv : 1109.4819 .
bugg , j. phys . g. conf
. ser . 295 ( 2011 ) 012155 , arxiv : 1011.6237 .
|
the status of light @xmath0 , @xmath1 and @xmath2 mesons is discussed , particularly the separation of @xmath3 and @xmath4 states .
they fall into a simple scheme except for @xmath5 . a case
is made that this has been confused with the @xmath6 .
it should be possible to check this with existing or forthcoming data . * light @xmath7 and @xmath2 mesons *
|
@xmath1 doradus ( f0v ) is the brightest member of what appears to be a new class of pulsating variable stars . in recent years , a number of late a- and early f - type dwarfs have been found to be variable with periods of the order of one day ( see krisciunas & handler , 1995 for a list )
this is too long for membership of the @xmath7 scuti class , where the periods are just a few hours .
the period is , in many cases , compatible with that expected for rotation , which suggests that the variability may be due to the rotation of a spotted star ( mantegazza , poretti & zerbi 1994 ) .
however , most stars appear to be multiperiodic and require improbably large differential rotation .
the starspot model for @xmath1 dor has been examined by balona , krisciunas & cousins ( 1994b ) who find that large overlapping spots would be required to give the observed light and colour amplitudes .
the most likely explanation for the variability in these stars is nonradial pulsation ( nrp ) , since it is difficult to understand multiple periodicities in terms of the starspot model or as orbital motion .
multiple periods are typical of nrp , but one needs to prove that the physical parameters determined from an nrp model are realistic and that the other models are unable to explain the observations . in nrp , the light variability is due to the periodic variation in temperature and , to a smaller extent , the change in the shape of the star during pulsation .
pulsation with a period of about one day implies that the dominant restoring force is gravity ( @xmath8-modes ) , rather than pressure ( @xmath9-modes ) as in the case of the @xmath7 sct stars .
furthermore , because the stars show substantial light and radial velocity variations , the averaging effect over the visible hemisphere of the star must be quite small .
this implies that the spherical harmonic degree , @xmath10 , of the pulsation must be low , probably @xmath11 .
aerts & krisciunas ( 1996 ) have analysed 9 aur , a star belonging to this group , using photometry and the cross - correlation profile obtained with the coravel instrument .
they deduce that 9 aur is pulsating in two modes , both having spherical harmonic indices ( @xmath2 ) = ( 3 , @xmath12 ) . the mechanism which excites the pulsation
is not known at present .
the stars appear to be located at , or near , the red edge of the @xmath7 sct instability strip .
this indicates that the mechanism may be the same as for @xmath7 sct stars ( the ionization of h and he ) , but perhaps modified by convection
. it would certainly be interesting to examine the line profile and light variations of these stars .
these data may enable us to distinguish between nrp and rotational modulation of a spotted star .
nrp produces line profile and light variations which , in principle , allow the determination of the spherical harmonic degree , @xmath10 and the azimuthal number , @xmath13 .
this can be done by examining the variation of the moments of a line profile and calculating a discriminant .
the value of the discriminant as a function of angle of inclination is used to determine the most probable mode ( @xmath2 ) .
the method was developed by balona ( 1986a , 1986b , 1987 ) and extended by aerts ( 1993 ) .
the two periods of @xmath1 dor are stable and well determined from extensive photometric observations : @xmath14 d , @xmath15 d ( balona et al .
more recently , evidence has been found for a third period at @xmath16 d ( balona , krisciunas & cousins , 1994b ) .
( we call this period @xmath17 and not @xmath18 because @xmath18 was used in balona et al .
( 1994a ) to refer to a possible single period , later shown to be incorrect ) .
since the star is bright ( @xmath19 ) and well placed , it was chosen as one of the targets for the musicos 94 programme .
the aim was to obtain a large number of high signal - to - noise line profiles which , in conjunction with strmgren photometry , would decide which of the two models , starspots or nrp , is the correct one and to determine the spot geometry or pulsation modes .
another factor in choosing to observe this star is the rapid change in the spectrum which appears from time to time ( balona et al .
1994a ) . @xmath1
dor is an _ iras _ source with a probable circumstellar dust cloud ( aumann 1985 ) , which may be responsible for this effect . in this paper
we present details of the observations and reduction procedure and analyse the data in terms of the two models .
spectroscopic data were obtained from four sites during 1994 november : eso ( chile ) , mount stromlo ( australia ) , sutherland ( south africa ) and mount john ( new zealand ) .
the projected rotational velocity of @xmath1 dor is high ( @xmath20 km s@xmath21 ) , which means that it is impossible to find a totally unblended spectral line . the most suitable line is probably that of fe ii at @xmath225018.450 or possibly mg i at @xmath225183.619 .
instructions were given to ensure that the @xmath225018.450 fe ii line was observed whenever possible .
the time - coverage at the various sites is shown in table 1 .
.details of the musicos 94 campaign on @xmath1 dor .
the epoch of the julian day ( jd ) is 2449600.00 ; n is the number of spectra obtained .
[ cols="<,<,>,>,>",options="header " , ]
one of the most puzzling aspects of @xmath1 dor is the sudden appearance , from time to time , of isolated absorption or emission features ( balona et al .
careful inspection of all the spectra obtained during the musicos-94 campaign has failed to reveal such oddities .
the star appears to have a normal early - f spectrum , though broadened by rotation .
there is no doubt , however , that the light variations of the star can not be explained entirely by three periodic components .
there is a certain level of non - periodicity which is not yet understood ( balona et al .
1994a , balona , krisciunas & cousins 1994b ) .
the spectroscopic observations during musicos-94 show evidence for this effect on one night ( jd 2449664 ) , where the radial velocities are substantially smaller than expected .
our new photometric observations confirm the existence of a third oscillation , @xmath23 d@xmath21 , previously suspected by balona , krisciunas & cousins ( 1994b ) .
the existence of three frequencies implies substantial differential rotation on the basis of the starspot model .
this model also fails to account for the great differences which exist between the radial velocity to light amplitude ratios between the two main oscillations .
but the most serious objection to the starspot model is the fact that for @xmath4 radial velocity and light maximum coincide in phase .
one would expect a @xmath24 phase difference for a starspot .
the small radial velocity amplitude also excludes the possibility of a binary companion .
we conclude that nonradial pulsation is the only viable explanation for the observations . from the light curve and first four moments of the line profile of fe ii @xmath225018.450
, we show that @xmath1 dor has a high angle of inclination , probably @xmath6 and that the three modes may be identified as @xmath25 , @xmath26 with fair confidence .
the pulsation parameters indicate quite a large value @xmath27 km s@xmath21 for @xmath3 .
now the relative flux amplitude + @xmath28 + which gives @xmath29 for @xmath3 , implying that this mode produces a large temperature variation @xmath30 .
this may account for the significant phase dependence of the equivalent width for this mode , whereas the other modes show little or no phase dependence .
the fact that the amplitudes of the colour indices @xmath31 and @xmath32 are larger in this mode is also consistent with the large @xmath33 .
the fact that @xmath4 dominates the radial velocity variation , but that @xmath3 does not , though they have similar light amplitudes , is due to the fact that @xmath3 has a relatively high spherical harmonic degree @xmath34 , whereas @xmath4 has the lower value @xmath35 . the averaging effect in the latter case is much less , leading to a larger radial velocity amplitude . on the other hand ,
the much larger value of @xmath33 for @xmath3 compensates in the light curve .
it is interesting that all three modes seem to be sectorial retrograde modes .
the frequency of pulsation in the co - rotating frame , @xmath36 , is related to the frequency of pulsation in the observer s frame , @xmath37 , by : @xmath38 where @xmath39 is the frequency of rotation . if @xmath40 and @xmath41 , we find @xmath42 d@xmath21 ( @xmath43 d ) .
thus , in the co - rotating frame , the three oscillations have the following frequencies : @xmath44 , @xmath45 and @xmath46 d@xmath21 . @xmath3 and @xmath4 are substantially different and must have quite different radial orders in spite of the fact that they are nearly equal in the inertial frame . on the other hand , @xmath4 and @xmath5 , which are of the same degree , @xmath10 , probably have neighbouring radial orders .
this information could be used to extract seismological information , but the effect of rotation needs to be taken into account more accurately before this can done .
the success of mode identification in this star is encouraging .
the application of the method to other stars might well result in very interesting results , as in the case of 9 aur ( aerts & krisciunas 1996 ) , but other types of pulsating star may be equally interesting .
for example , an intensive campaign on a periodic be star ( and there are many of these which are bright ) should result in a definitive test of the nrp / starspot models ( balona 1996 ) .
all that would be required is an accurate determination of the radial velocity to light amplitude ratio .
jbh thanks o. k. petterson for assistance with midas reductions .
wal thanks dr s james , adfa , for computing support .
|
we present data from high - dispersion echelle spectra and simultaneous @xmath0 photometry for @xmath1 doradus .
these data were obtained from several sites during 1994 november as part of the musicos-94 campaign .
the star has two closely - spaced periods of about 0.75 d and is the brightest member of a new class of variable early f - type stars . a previously suspected third period , very close to the other two , is confirmed .
previous observations indicated that sudden changes could be expected in the spectrum , but none were found during the campaign .
the radial velocities rule out the possibility of a close companion .
the phasing between the radial velocity and light curve of the strongest periodic component rules out the starspot model .
the only viable mechanism for understanding the variability is nonradial pulsation .
we used the method of moments to identify the modes of pulsation of the three periodic components .
these appear to be sectorial retrograde modes with spherical harmonic degrees , ( @xmath2 ) , as follows : @xmath3 = ( 3,3 ) , @xmath4 = ( 1,1 ) and @xmath5 = ( 1,1 ) .
the angle of inclination of the star is found to be @xmath6 .
stars : pulsation - stars : variables - stars : individual : @xmath1 dor .
|
the physics at hera has proved the successes of ( resummed ) perturbative qcd in describing many observables accurately measured there : inclusive ones like the structure function @xmath0 , but also more exclusive ones , like the diffractive structure function @xmath1 , or heavy meson production .
the justification for relying on a perturbative development is that the deep - inelastic scattering process naturally provides a well - controlled hard scale given by the photon virtuality @xmath2 , which makes the effective strong coupling constant @xmath3 small enough . on the one hand ,
the physics of these observables is usually well described by the renormalization group evolution @xcite between a lower scale @xmath4 at which the proton parton densities are parametrized and the scale @xmath2 . on the other hand ,
the cross - section for the events selected with the requirement that a forward jet of transverse momentum @xmath5 of the order of @xmath2 be present in the final state is seemingly not described by a straightforward dglap evolution ( see ref.@xcite and references therein ) : as a matter of fact , these regge - like kinematics are expected to select the bfkl dynamics @xcite . in @xmath6 collisions at the tevatron ,
no hard scale is provided by the initial state .
however , it can be generated in the scattering and manifests itself in the final state in the form of a jet with a large transverse momentum .
events of this class are also accessible to a perturbative qcd interpretation . in this letter , we focus on high - energy onia ( massive @xmath7 states ) collisions , as a model for @xmath6 collisions at high energy when high - mass scales are selected by forward jets
. the inclusive cross - sections for onia collisions have been described using a dipole cascade modelling the rapidity evolution of the @xmath7 pairs before their interaction @xcite . here
we require that at least the first ( most forward ) gluon which goes to the final state has its transverse momentum larger than a scale @xmath8 ; this gluon becomes a forward - jet , and we interpret it as an effective colour - dipole distribution present at the time of the scattering . in section * 2 * , we detail the modelisation that we adopt for the forward jet . we show in section * 3 * how we can extract to double - leading logarithmic ( dll ) approximation , the dipole content of such an object .
section * 4 * contains our conclusions and outlook .
the onium - onium forward scattering amplitude involves the exchange of two gluons between the initial onia . at high - energy , one has to take into account the possibility of multiple splittings of these @xmath9-channel gluons .
this can be done either by ( @xmath10)factorizing @xcite a bfkl - like ladder between the bare onia , or equivalently by computing the two - gluon exchange diagram between the onia dressed by an arbitrary number of soft `` sea '' gluons .
the latter approach inspired the colour - dipole model of ref.@xcite .
the equivalence between these two methods was shown on different features of both pictures in ref.@xcite . in this section
, we shall derive in a @xmath9-channel picture the gluon density @xmath11 which is to be considered in an interacting `` onium+forward - jet '' system as the starting - point of a bfkl evolution .
the next section will be devoted to the interpretation of the obtained density as an effective primordial dipole density inside the forward jet .
we have to compute the `` @xmath12 '' amplitude , where the initial - state dipole of radius @xmath13 is part of an onium , and the final - state gluon has a transverse momentum @xmath14 . the modulus @xmath15 is larger than a given scale @xmath8 .
having in mind the fact that in a physical process the initial - state dipole will be part of a hadron instead of an onium , we will consider in the following that @xmath16 whenever needed for technical purpose .
we start with the ( virtual)gluon - dipole cross - section @xmath17 which defines the gluon density inside a dipole at lowest order in @xmath18
. it reads ( see ref.@xcite ) : @xmath19 it can be expressed as an inverse mellin - transform in the transverse plane : @xmath20 where @xmath21 is interpreted as the well - known dipole - gluon `` vertex '' in the mellin space : @xmath22 we then factorize the emission of a real gluon .
it can be computed directly in the high - energy limit by evaluating the relevant graphs ( one of them is pictured in fig.*[fig:1 ] * ) , but it is also convenient to see it as one step of the bfkl ladder .
the initial gluon density @xmath23 and the density @xmath11 after emission of a gluon are related through the formula : @xmath24 which is the lowest order ( in @xmath25 ) bfkl equation written in an unfolded form ( see for instance ref.@xcite ) .
the variable @xmath26 is proportional to @xmath27 , where @xmath28 and @xmath9 are the ordinary mandelstam variables for the reaction . in deep - inelastic scattering
, @xmath26 would stand for the bjorken variable .
let us write the mellin - transform of eq.([eq : mellin ] ) : @xmath29 where we used similar notations as in ref.@xcite , although we have not performed the mellin - transform with respect to @xmath26 .
the variable @xmath30 describes a path @xmath31\gamma_0\!-\!i\infty,\gamma_0\!+\!i\infty[$ ] in the complex plane , with @xmath32 .
inserting eqs.([eq:1])-([eq:4 ] ) into eq.([eq : hgsh ] ) , it follows that : @xmath33 the integration over @xmath34 can easily be performed switching to complex variables and using the well - known identity ( see @xcite and references therein ) : @xmath35 valid for @xmath36 , @xmath37 and @xmath38 .
the result reads : @xmath39 where the meijer - function @xmath40 ( which arguments will be abreviated in the following ) writes : @xmath41 the contour of integration @xmath42 being defined on fig.*2*. the function @xmath43 can be seen as the coefficient - function in the sense of ref.@xcite , for the onium+forward - jet system .
in ref.@xcite , a relation was established between the coefficient - function @xmath43 of a virtual photon and its corresponding squared wave - function @xmath44 on a dipole basis .
we shall make use of it in the present context of semi - exclusive factorization to extract the dipole content of the forward jet from the coefficient - function found above , at dll accuracy .
the double - logs should manifest themselves through factors like @xmath45 , where @xmath46 is a characteristic size for the final state .
the relationship between the squared wave - function and the coefficient - function reads : @xmath47 an intuitive way of understanding this relation could be the following : dividing the coefficient - function pictured in fig.*[fig:1 ] * by the factor @xmath48 amounts to getting rid of the vertex of the lowest gluon which in this picture is also a gluon - dipole vertex .
hence one ends up with the dipole content of the scattering object : @xmath49 let us explore the limit @xmath50 , in which the forward jet has a transverse momentum much larger than the characteristic scale of the initial onium .
the meijer - function @xmath40 can be approximated in a straightforward manner by picking the pole ( at @xmath51 ) which lies on the left of the integration path .
indeed , the contour @xmath42 can be deformed to @xmath52 ( see fig.*2 * ) since the integral converges on any path such that @xmath53 , where @xmath54 : @xmath55 the integral taken on the contour @xmath52 is subdominant by some power of @xmath56 with respect to the contribution of the double - pole at @xmath51 , and one writes : @xmath57 this approximation proves to be very good numerically even for relatively small @xmath58 .
then the squared dipole wave - function reads , in this approximation : @xmath59 we want to obtain an expression for the distribution of dipoles in the system in coordinate space , i.e. as a function of the transverse size @xmath60 .
it is given by the inverse - mellin transform of @xmath44 : @xmath61 we obtain the following result : @xmath62 the second term is not relevant in our approximation , since the limit of large @xmath58 selects the dll ; the terms beyond this approximation are not under control .
hence the interaction of the system formed by the initial dipole and the forward gluon can be viewed as a dipole - dipole interaction provided the system is described by the following dipole distribution : @xmath63 let us give an interpretation of the various factors in this distribution .
first , note that the dependence on the initial dipole size @xmath64 only appears in the factor @xmath65 .
this remarkable fact technically results from the combination of the term @xmath66 and the inverse - mellin transform of the @xmath67 functions in the inverse - mellin transform of eq.([eq : phigam ] ) .
physically , the overall factor @xmath68 can then be interpreted as the probability of finding a dipole of size @xmath60 inside a dipole of size @xmath64 .
indeed , in the dll approximation at lowest order in @xmath25 and assuming an available energy proportional to @xmath69 , the probability of finding a dipole of size @xmath70 between @xmath71 and @xmath72 inside a dipole of size @xmath73 reads @xcite : @xmath74 where the second approximate equality holds when @xmath75 . dividing out this factor in eq.([eq : dipolegluon ] ) leads to the universal dipole content of the gluon radiated into the final state .
we normalize the first moment of the obtained squared wave - function to unity has now the unusual dimension @xmath76 ^ 3 $ ] , but it enables a comparison to the photon squared wave - function @xmath77 , for which the @xmath78 moment diverges logarithmically . ]
( i.e. @xmath79 ) , and so we are led to : @xmath80 note that this squared wave - function for which we provide a derivation here is exactly the one ( integrated over the energy - share variable @xmath81 and properly normalized ) that was postulated in ref.@xcite . to see this , we only need to recall that the mellin transform of the function @xmath82 reads : @xmath83 the behaviour of @xmath84
is represented in fig.*3*. it is compared to the behaviour of the transverse virtual photon squared wave - function in @xmath7 pairs , integrated over the fraction of the photon longitudinal momentum @xmath81 carried by the quark .
this photon wave - function reads : @xmath85 where the same normalization @xmath86 has been enforced .
some remarks are in order .
the squared wave - function @xmath84 represents the effective distribution of dipoles resulting from a final - state gluon which has its transverse momentum larger than @xmath8 .
we note that it is slowly decreasing with @xmath60 ( @xmath87 ) , which means that dipoles of large size occur with a non - negligible probability .
the width of the distribution is of order @xmath88 , but since its behaviour at infinity is only powerlike , @xmath8 is not a clean cutoff .
it has an oscillatory behaviour , the oscillation length being of the order of @xmath88 .
however , @xmath84 takes also negative values , which indicates that it does not allow for a direct probabilistic interpretation like in the case of the @xmath89-pair distribution @xmath77 inside the photon .
using a straightforward model and a qcd calculation in the dll approximation , we have shown that a gluonic hard probe in the final - state can be characterized by a dipole - distribution @xmath90 .
this distribution decreases weakly with @xmath60 and takes negative values .
we confirm the ( integrated over @xmath81 ) result obtained in ref.@xcite .
the obtained distribution can now be used to compute processes of several topologies : for instance deep - inelasic scattering on a forward - jet at small-@xmath26 , @xmath91 interactions with two jets in the final - state separated by a large rapidity range . in any case
, the dipole formulation for these observables is of the type : @xmath92 where @xmath93 is the eigenvalue of the bfkl kernel , and @xmath94 the rapidity range between the two probes characterized by the dipole distributions @xmath84 and @xmath95 .
note that the simplest observable one can compute using this formula is the cross - section @xmath96 for the production of dijets of respective transverse momenta larger than @xmath97 and @xmath98 in hadronic collisions .
it reads : @xmath99 which is the mueller - navelet formula @xcite , modulo some normalization we did not keep precise track of . hence this little calculation provides a further check of our dipole distribution .
this study may deserve several further investigations and improvements .
first of all , we worked only up to terms beyond the dll approximation .
it would be useful to perform a more complete calculation , by computing to leading log-@xmath69 precision all the graphs of the type of the one in fig.*1 * which contribute , to see if the dipole factorization still holds . on the other hand
, it would be nice to supplement our indirect method of extracting the dipole content by a more direct calculation in the framework of the colour - dipole model : this would insure the full control of the leading log-@xmath69 , including a correct treatment of the virtual corrections .
although quite straightforward to formulate , the latter calculation exhibits many technical difficulties .
both these proposed improvements deserve more studies . +
* acknowledgements : * + i thank r. peschanski and h. navelet for many useful suggestions and a careful reading of the manuscript .
gribov and l.n .
lipatov , yad .
( 1972 ) 781 .
dokshitzer , sov .
jetp 46 ( 1977 ) 641 .
g. altarelli and g. parisi , nucl .
b126 ( 1977 ) 298 . b. ptter , ( 1999 ) , hep - ph/9909320 , talk given at the ringberg workshop _
new trends in hera physics 1999_. e.a .
kuraev , l.n .
lipatov and v.s .
fadin , sov .
jetp 44 ( 1976 ) 443 .
kuraev , l.n .
lipatov and v.s .
fadin , zh .
72 ( 1977 ) 377 .
i.i . balitsky and l.n .
lipatov , sov .
( 1978 ) 822 .
mueller , nucl .
b415 ( 1994 ) 373 .
s. catani , m. ciafaloni and f. hautmann , nucl .
b366 ( 1991 ) 135 .
z. chen and a.h .
mueller , nucl .
b451 ( 1995 ) 579 .
h. navelet and r. peschanski , nucl .
b507 ( 1997 ) 353 , hep - ph/9703238 . h. navelet and s. wallon , nucl .
phys . b522 ( 1998 ) 237 , hep - ph/9705296 .
s. munier and r. peschanski , nucl .
phys . b524 ( 1998 ) 377 , hep - ph/9802230 .
j. kwiecinski , a.d .
martin and p.j .
sutton , phys .
d52 ( 1995 ) 1445 , hep - ph/9503266 .
j.s . geronimo and h. navelet , ( 2000 ) , math - ph/0003019 , to appear .
r. peschanski , ( 1999 ) , hep - ph/9910377 , to appear .
mueller and h. navelet , nucl .
b282 ( 1987 ) 727 .
|
we show that a forward jet with large transverse momentum in an onium - onium collision is a hard probe which can be effectively characterized by a colour - dipole distribution at the time of the interaction .
the dipole distribution is computed , and compared to its counterpart for a virtual photon in the initial state .
we find that while in the photon case , the tail of large sizes is exponentially cut - off , it contributes sizeably in the forward - jet case , which signs the sensitivity of observables based on such events to the infrared region .
moreover , a direct probabilistic interpretation of the dipole distribution fails since it takes negative values in the large size region .
|
instructional physics laboratories are well - situated to promote student competence in physics practices , yet few studies focus on such environments especially at the upper - division level @xcite . recently , the american association of physics teachers ( aapt ) identified modeling as a major focus area for learning outcomes in undergraduate laboratory courses @xcite , consistent with model - based efforts in introductory instruction @xcite . over the last few years
, we have developed the experimental modeling framework ( emf ) to inform curricular transformation of upper - division laboratory courses @xcite .
the emf describes the process through which physicists iteratively revise system models , physical apparatus , and measurement tools in response to discrepant measurements and predictions .
the growing emphasis on modeling warrants continued application of the emf to a broad range of upper - division instructional laboratory contexts .
previous work demonstrated the utility of the emf as a tool for characterizing students model - based reasoning , focusing on an optical physics context @xcite . in the present work
, we use the emf in a similar capacity but a new context : troubleshooting a malfunctioning electronic circuit . here troubleshooting " refers to the iterative process through which one repairs a malfunctioning apparatus @xcite , a common experimental physics practice .
electronics courses are an ideal environment for studying troubleshooting , due in part to the simplicity of the models and systems with which students interact and the ease with which components can be replaced .
our focus on troubleshooting aligns with the ongoing transformation of an electronics course at the university of colorado boulder , informed by both national @xcite and local @xcite learning goals for upper - division laboratory courses . of course
, the emf is not the only useful lens through which to understand troubleshooting . in a companion article @xcite
, we analyze the same data presented here through a different , complementary lens : the socially mediated metacognition framework ( smmf ) @xcite .
together , the emf and smmf allow us to more fully explore students strategic navigation of an electronics task . in the present work
, we briefly highlight one area where a metacognitive perspective would provide additional insight into students troubleshooting process .
however , our main focus is on application of the emf to an electronics troubleshooting context .
in the the emf , models " are defined as abstract representations used to explain aspects of the real world and predict scientific phenomena @xcite .
models are embedded in known principles and concepts , but contain simplifications and assumptions that yield tractable mathematical , graphical , and other representations .
the emf divides systems into two parts : the _ physical system _ and the _ measurement system_. in the analysis herein , we focus on student engagement with only the _ physical system_. accordingly , we do not attend to the _ measurement system _ further .
a discussion of the interaction between _ physical _ and _ measurement systems _ can be found in ref .
@xcite .
the process of modeling " is dynamic and iterative , involving the following phases : _ model construction _ , the process of developing models of the system ; _ prediction _ , the use of a model to inform expectations about measurements ; _ comparison _ , the act of comparing measurements to predictions ; _ proposal _ , the act of proposing a potential explanation for , and/or solution to , discrepant measurements and predictions ; and _ revision _ , the process of making changes to either the apparatus or model in order to bring measurements and predictions into better alignment .
we operationalize _ comparison _ as utterances in service of answering the question , is the agreement between measurement and prediction good enough ? " similarly , _ proposal
_ includes statements that seek to answer the question , how can we get better agreement ? " a visualization of the modeling process is presented in fig .
[ fig : framework ] . experimental modeling framework .
the emf describes the iterative process of modeling the measurement and physical systems , comparing measurements to predictions , proposing explanations for discrepancies , and revising models and/or apparatuses . for brevity
, this figure is a simplified version of the full emf presented by zwickl et al . in ref
. @xcite . ]
similar to modeling , troubleshooting is also a dynamic and iterative process .
troubleshooting can be subdivided into four tasks @xcite : constructing the problem space , identifying fault systems , diagnosing faults , and generating and verifying solutions .
these tasks are similar to the modeling phases .
for example , construction of the problem space and identification of faults involve both _ model construction _ and _ comparison_. because troubleshooting involves fixing a malfunctioning apparatus , _ comparison _ typically involves comparing measurements of the actual , malfunctioning apparatus to expectations informed by models of an idealized , functioning apparatus .
diagnosis of faults and generation of solutions require _ proposal _ and _ revision_. because the goal of troubleshooting is to repair a malfunctioning system , _ revision _ is typically limited to making changes to the apparatus to bring its performance into better alignment with the idealized model .
troubleshooting further requires a strategic approach .
two examples of strategies present in our data include @xcite : _ trial and error _ , characterized by students arbitrarily focusing on any subsystem in which a fault may be present ; and _ split - half _ , characterized by students splitting the system into two subsystems and testing the midpoint in an attempt to isolate the fault in one of the two halves . in this work ,
we use the emf as a tool for understanding troubleshooting tasks and strategies .
inverting cascade amplifier . noninverting ( stage 1 ) and inverting ( stage 2 ) amplifiers were connected in series .
two faults were present in stage 2 : the resistor in position @xmath0 had a resistance of 100 @xmath1 rather than 1 k@xmath1 , and the op - amp was burned out in such a way that its output voltage was equal to the supply voltage regardless of the input voltage . ] to facilitate study of both the iterative and strategic aspects of troubleshooting an electronic circuit , we designed the inverting cascade amplifier shown in fig .
[ fig : diagram ] .
the cascade amplifier consists of two subsystems , or stages : a noninverting amplifer ( stage 1 ) , and an inverting amplifier ( stage 2 ) .
each stage amplifies its input voltage by a multiplicative factor called the gain , @xmath2 , which is determined by the resistor values . in a functional cascade amplifier ,
stage 1 would double the input voltage ( @xmath3 ) and stage 2 would both invert the input voltage and amplify it by a factor of ten ( @xmath4 ) . in the context of alternating current ( ac ) signals , inverting " is equivalent to shifting the phase of the signal by 180@xmath5 . because the two stages are connected in series ,
the overall gain of the cascade amplifier is the product of the gains of stages 1 and 2 : @xmath6 . to ensure that students engaged in more than one iteration of the troubleshooting tasks , we introduced two faults in the circuit ( fig .
[ fig : diagram ] ) .
first , the resistor @xmath0 had a value of 100 @xmath1 rather than the nominal value of 1 k@xmath1 , increasing the actual gain of stage 2 by an order of magnitude compared to the nominal gain .
second , we used a broken op - amp in stage 2 , which manifested in a direct current ( dc ) output voltage of @xmath7 v regardless of input .
both faults were localized in stage 2 so that the cascade amplifier consisted of both a functional subsystem ( stage 1 ) and a malfunctioning one ( stage 2 ) , making it possible for students to use the _ split - half _ strategy early on in their troubleshooting process .
we conducted think - aloud interviews in which pairs of students were asked to diagnose and repair a faulty inverting cascade amplifier .
students were given a schematic of the circuit , including an algebraic expression for the gain of the entire circuit , but not for either of the two stages .
we interviewed 4 pairs of physics majors from each of two institutions ( total of 16 students ) .
participants were paid volunteers enrolled in junior - level electronics courses at their respective institutions during fall 2014 .
interviews were conducted at the end of the fall term through the beginning of the following spring .
interviews , which lasted 2560 minutes , were videotaped and transcribed .
we used interaction analysis @xcite to investigate students interactions with each other and the apparatus , using the emf as an _ a priori _ analysis scheme .
we treated models and model - based reasoning as belonging to the shared interactional space ; hence we do not expect our results to be a perfect reflection of individual student reasoning .
two of the authors ( d.r.d.f . and k.l.v.d.b .
) collaboratively viewed and discussed entire interviews .
select excerpts were interpreted by the research team as a whole . to ensure that
our interpretations aligned with those of the broader community , some video excerpts were discussed with research groups and individual researchers outside of our research team .
we focus on a 90-second excerpt from one interview .
this excerpt was chosen for two reasons : first , it demonstrates connections between _ model construction _ , _ prediction _ , _ comparison _ , and _ proposal _ during students model - based reasoning ; and second , it provides an example of students engaging in socially mediated metacognition , which we discuss elsewhere @xcite . in this excerpt , a pair of students , s1 and s2 , troubleshoot the malfunctioning cascade amplifier ( fig .
[ fig : diagram ] ) .
the episode takes place about halfway through the task .
prior to this episode , the students computed the expected output of the circuit : given their ac input signal of amplitude 100 mv , the students articulated their expectation that the output of stage 2 should be an inverted ac signal with amplitude 2 v. due to the details of their measurement apparatus ( the students chose a setting that filters out dc signals ) , the students measured the output of stage 2 to be a noninverted ac signal with a 10-mv amplitude . in response to this discrepancy
, the students decided to measure the output of stage 1 .
this decision is consistent with the _ split - half _ troubleshooting strategy .
the only expectation the students previously articulated about the stage 1 output is that it should be different from ground . after articulating this expectation ,
the students measured the output of stage 1 to be a noninverted ac signal of amplitude 200 mv .
the transcript begins immediately after this measurement : + + r>**lp7.5 cm 1&s1 : & so , it s doubled v - in , but it s not inverted it . + 2 & & and it should nt be inver should be + 3&s2 : & is that an inverting amplifier ?
+ 4&s1 : & no , it s not .
+ 5 & & an inverting amplifier is connected to the + 6 & & v - in is connected to the negative terminal , right ?
+ 7&s2 : & yeah ,
+ 8&s1 : & so it should nt be inverted .
+ + here s1 stated that the effect of stage 1 was to double the input voltage but to leave its phase unchanged ( line 1 ) .
he then attempted to reconcile his findings with his expectations .
however , even though he started to make a prediction , he interrupted himself and did not complete his statement ( line 2 ) . s2 then immediately interrupted s1 : to ask whether stage 1 is an inverting amplifier ( line 3 ) .
s1 responded by claiming that the first stage is not an inverting amplifier ( line 4 ) and justifying this claim by describing a characteristic feature of inverting amplifiers ( lines 56 ) .
s1 then asked for , and received , confirmation from s2 about his line of reasoning ( lines 67 ) . after receiving this affirmation ,
s1 articulated his expectation about the phase of the stage 1 output : so it [ the output of stage 1 ] should nt be inverted " ( line 8) .
this utterance is an example of both _
prediction _ and _ comparison _ : s1 articulated a model - informed expectation about the phase of the stage 1 output in the context of determining whether the observed output was what the students should " be seeing .
the students then went on to discuss the second stage : + r>**lp7.5 cm 9&s1 : & so this one _ ( points to schematic ) _
+ 10&s2 : & well , neither of them are inverting .
+ 11 & & oh , yes . _ ( points to schematic ) _ + 12 & & this one is inverting .
+ 13&s1 : & the second one is inverting .
+ in lines 9 and 12 , this one " is the second stage , as evidenced by line 13 . in this exchange ,
s1 and s2 correctly identified stage 2 as an inverting amplifier .
together , lines 313 provide an example of _ model construction _ through which the students constructed a model of the inverting cascade amplifier as comprising two distinct amplifier subsystems , one noninverting and the other inverting .
the students previously had a model of the circuit that was sufficient to predict the amplitude and phase of the output of the entire circuit , but not of the midpoint .
while this preliminary model was made more sophisticated as a result of the _ model construction _ in lines 313 , this is not an example of _ revision _ because the changes were not made in response to a discrepancy between a prediction and a measurement .
in fact , we argue that the lack of a sufficiently sophisticated model prevented s1 from articulating a prediction in line 2 . only after constructing a model of the circuit as being comprised of a noninverting subsystem ( lines 37 ) was s1 able to make a prediction about the expected output of stage 1 ( line 8) , thus facilitating the _ split - half _ strategy .
this connection between _ model construction _ and _ prediction _ is consistent with the emf .
the students then focused the output of stage 2 : + r>**lp7.5 cm 14&s2 : & but our v - out right now is nt inverting .
+ & & _ ( points to oscilloscope ) _ + 15 & &
so that probably means that these positive plus + 16 & & and minus terminals on the second one are just + 17 & & mixed up . _ ( points to schematic ) _ + +
line 14 is an example of _ comparison _ : s2 compared the previously measured output of the circuit to his expectation that the output should be inverted .
a prediction is not explicitly articulated in the excerpt ; however , one was articulated earlier in the interview .
lines 1517 , on the other hand , are an example of _ proposal _ : to explain the discrepancy , s2 proposed that the input terminals of the op - amp in stage 2 were connected incorrectly .
the flow from _ comparison _ to _ proposal _ is also consistent with the emf .
the students then discussed s2 s proposal further : + r>**lp7.5 cm 18&s1 : & why ?
+ 19&s2 : & because it s not inverting .
+ 20 & & so this is an inverting amplifier so they just mixed + 21 & & up the plus and minus . _ ( points to schematic ) _ + 22&s1 : & but this one s not doing anything at all .
+ & & _ ( points to schematic ) _ + 23 & & the way this is drawn here is inverting .
+ 24&s2 : & yeah .
but on here_(points to circuit ) _
+ 25&s1 : & on here it s not _ ( leans over circuit ) _
+ 26 & & there s no output at all .
+ 27 & & i mean there s this tiny _ ( points to oscilloscope ) _ + 28&s2 : & what do you mean ? _ ( points to oscilloscope ) _ + 29 & & yeah , theres + 30&s1 : & i guess , but that s like + 31&s2 : & how much well ,
how big is it ?
+ 32&s1 : & it s tiny .
it s like ten millivolts .
+ 33&s2 : & oh .
well , okay . + in lines 1833 , the students engaged in a cyclic interaction as a mutual attempt to understand one another s thinking
. s2 identified that he was concerned with explaining why the output was not inverted ( line 19 ) .
s1 followed by offering his own interpretation of the problem : the output of stage 2 was negligibly small ( lines 2632 ) .
each student asked the other to clarify their reasoning ( lines 18 and 28 ) .
ultimately , s2 accepted s1 s reasoning ( line 33 ) .
student interactions like those presented in lines 1833 are common in our data .
however , such interactions are not well - captured by the emf because they do not constitute a modeling phase , but instead represent students negotiations about whether and when to move from one phase to the next . in this case , the negotiation culminated in a decision _ not _ to follow up on s2 s idea about mixed up " inputs ( lines 1517 ) . a complementary framework to the emf
is needed to fully understand this interaction . in a companion study
, we investigate this interaction from a metacognitive perspective @xcite .
s2 then stepped back to consolidate his thoughts : + r>**lp7.5 cm 34 & s2 : & we have a good output for the first op - amp , + 35 & & so we are going to have + 36 & & the problem is in the second one .
+ + line 34 is another example of _ comparison _ : s2 reiterated that , regarding the output of stage 1 , the students measurements and expectations were in good enough agreement with one another .
in this case , good enough " meant good enough to eliminate stage 1 as a potential source of fault . "
while there is no explicit comparison being made in lines 3536 , we nevertheless infer that s2 made an appraisal for the second stage as well .
the lack of good enough agreement between measurements and expectations in stage 2 led s2 to conclude that the problem is in the second one " ( line 36 ) . from a troubleshooting perspective
, line 36 represents the culmination of the _ split - half _ strategy , which enabled the students to narrow their search for faults to a single subsystem of the cascade amplifier .
we applied the emf to an activity in which a pair of students troubleshot a malfunctioning electronic circuit , demonstrating that students engage in model - based reasoning throughout the troubleshooting process .
in particular , we discussed the interaction of various emf phases during a troubleshooting cycle in which students employed the _ split - half _ strategy .
we further identified an example of student reasoning that is beyond the scope of the emf : collaborative decisions about how to navigate between phases .
elsewhere , we discuss a metacognitive framework for describing this process @xcite .
in addition to its utility in characterizing model - based reasoning , the emf is a useful tool for facilitating curricular transformation of laboratory courses . along these lines
, our work will inform explicit instruction and assessment of troubleshooting skills in electronics courses .
the authors acknowledge a. little , b. zwickl , b. wilcox , g. quan , j. stanley , and the colorado and maryland per groups for useful input on study design and interpretation of results .
this work was supported by nsf grants due-1323101 , due-1323426 , due-1245313 , and due-0962805 .
13ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop _ _ ( , , ) @noop `` , '' ( ) \doibase http://dx.doi.org/10.1119/1.2983148 [ * * , ( ) ] in @noop _ _ , vol .
( ) \doibase http://dx.doi.org/10.1119/1.4875924 [ * * , ( ) ] \doibase http://dx.doi.org/10.1119/1.4768890 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.3680077 [ * * , ( ) ] @noop link:\doibase 10.1007/s10648 - 006 - 9001 - 8 [ * * , ( ) ]
link:\doibase 10.1518/001872000779656570 [ * * , ( ) ] @noop \doibase 10.1023/a:1016209010120 [ * * , ( ) ] @noop * * , ( )
|
troubleshooting systems is integral to experimental physics in both research and instructional laboratory settings .
the recently adopted aapt lab guidelines identify troubleshooting as an important learning outcome of the undergraduate laboratory curriculum .
we investigate students model - based reasoning on a troubleshooting task using data collected in think - aloud interviews during which pairs of students attempted to diagnose and repair a malfunctioning circuit .
our analysis scheme is informed by the experimental modeling framework , which describes physicists use of mathematical and conceptual models when reasoning about experimental systems . we show that this framework is a useful lens through which to characterize the troubleshooting process .
|
in this work , we consider the quark and gluon parton distribution functions ( pdfs ) in nuclei at small values of bjorken @xmath0 and their reduction as compared with the incoherent sum of the nucleon pdfs because of the phenomenon of nuclear shadowing .
most of experimental information on nuclear pdfs comes from inclusive deep inelastic scattering ( dis ) with nuclear targets which measures the nuclear structure function @xmath2 . for @xmath3 , @xmath4 , which is called nuclear shadowing [
@xmath5 is the isoscalar nucleon structure function and @xmath6 is the number of nucleons ] .
because of the factorization theorem for dis ( for a review , see ref .
@xcite ) , which relates @xmath2 to nuclear parton distributions @xmath7
( @xmath8 is the parton flavor ) , nuclear shadowing is also present in nuclear pdfs , @xmath9 for @xmath3 , where @xmath10 is the pdf of the free nucleon .
this finds evidence in the results of the global fits that extract nuclear pdfs from various data on hard scattering with nuclei @xcite .
nuclear pdfs at small @xmath0 play an important role in the phenomenology of hard scattering with nuclei .
their knowledge is required for the evaluation and interpretation of hard phenomena in proton - nucleus and nucleus - nucleus collisions at relativistic heavy ion collider ( rhic ) and the large hadron collider ( lhc ) , in real photon - nucleus interactions in ultraperipheral collisions at the lhc @xcite , and in lepton - nucleus scattering at the future electron - ion collider ( eic ) @xcite .
in addition , nuclear pdfs at small @xmath0 are needed for the quantitative estimation of the onset of saturation in ultra high energy interactions with nuclei , which can be studied at the lhc and the eic .
a comparison of the results of the global fits for nuclear pdfs obtained by various groups @xcite shows significant discrepancies in the predictions for nuclear pdfs at small @xmath0 ( uncertainties of individuals fits at small @xmath0 are also very large @xcite ) .
the main reason for this is that the global fits are predominantly based on fixed - target data that do not cover the small-@xmath0 region ( by requiring that @xmath11 gev@xmath12 is sufficient for the applicability of the factorization theorem , one limits @xmath13 ) .
in addition , the gluon nuclear pdf is determined indirectly from the scaling violations using the very limited data .
therefore , the extrapolation of the obtained nuclear pdfs to the low values of bjorken @xmath0 that will be probed at the lhc and the eic is essentially uncontrolled . an alternative to the global fits
is provided by the approaches that attempt to predict nuclear shadowing for nuclear pdfs using the high - energy dynamics of the strong interactions .
we use the so - called leading twist theory of nuclear shadowing @xcite .
it combines the technique used by gribov to derive nuclear shadowing for the total hadron - deuteron cross section at high energies @xcite and the qcd factorization theorems for inclusive @xcite and diffractive @xcite dis .
the numerical predictions employ the results of the leading twist qcd analyses of hard diffraction in lepton - proton dis at hera @xcite .
although in the leading twist approach the hard probe interacts with one parton of the nucleus , in the target rest frame , nuclear shadowing appears as the effect of multiple interactions of the projectile ( virtual photon ) with several ( all ) nucleons of the target .
the interaction with @xmath14 nucleons is related in a model - independent way to the diffractive pdfs of the nucleon .
the account of the interaction with @xmath15 nucleons is model - dependent and sensitive to the underlying dynamics of the hard diffraction .
the recent hera data @xcite revealed that the energy dependence of the hard diffraction in dis ( dependence on the light - cone fraction @xmath16 ) is close to that of the soft processes .
this indicates that the hard diffraction in dis is dominated by large - size hadron - like configurations in the photon wave function .
this observation allows us to improve the treatment of the contribution to nuclear shadowing coming from the interactions with @xmath17 nucleons as compared to the simplified quasi - eikonal approximation used in our earlier papers @xcite by taking into account the presence of both point - like and hadron - like configurations in the virtual photon .
the goal of the present letter is to present a new , improved treatment of such multiple interactions and to update predictions for nuclear shadowing in nuclear pdfs .
we emphasize that the presence of the small - size ( point - like ) configurations and , in general , configurations of different transverse sizes that interact with different cross sections ( we call such configurations color or cross section fluctuations ) is much more important for the virtual photon than for hadronic projectiles .
our letter is organized as follows . in sec .
[ sec : lt_theory ] , we briefly review the leading twist theory of nuclear shadowing and present a new formalism for the treatment of the multiple interactions of the virtual photon with the nucleons of the nuclear target , which is based on the concept of color ( cross section ) fluctuations . in sec .
[ sec : cfa ] , we show that the complete treatment of color fluctuations can be well approximated by the so - called color fluctuation approximation . in sec . [
sec : predictions ] , we present updated predictions for the effect of nuclear shadowing in nuclear parton distributions .
we focus on the predictions for heavy nuclei at small @xmath0 since modifications of the predictions for light nuclei ( where the @xmath14 term dominates ) are rather small .
our results are summarized in sec .
[ sec : summary ] .
the phenomenon of nuclear shadowing is fairly well - understood : in the target rest frame , nuclear shadowing arises as the result of multiple interactions of the projectile ( virtual photon ) with several nucleons of the nuclear target .
the number of the interactions increases with decreasing bjorken @xmath0 , which is a result of the space - picture of the strong interactions , see e.g. , ref . @xcite . at sufficiently high energies ( small bjorken @xmath0 )
, the virtual photon can interact with all the nucleons of the target that are located in the photon s path .
the graphs that contribute to the nuclear structure function @xmath2 are presented in fig .
[ fig : master1 ] , where we used the optical theorem to relate the imaginary part of the @xmath18 forward scattering amplitude to the nuclear structure function . in the figure , graphs @xmath19 , @xmath20 , and @xmath21 correspond to the interaction with one , two , and three nucleons of the nuclear target , respectively .
note that the graphs for the interaction with four or more nucleons are not shown , but assumed . graphs @xmath20 , @xmath21 and higher scattering terms are responsible for nuclear shadowing in @xmath2 .
the contribution of graph @xmath19 , which is conveniently denoted @xmath22 , is @xmath23 where @xmath5 is the isospin - averaged structure function of the nucleon . in eq .
( [ eq : m2 ] ) , we neglected the deviation from the many - nucleon approximation for the description of nuclei and the fermi motion effect , which are numerically unimportant at small @xmath0 .
the calculation of the contribution of graph @xmath20 , @xmath24 , is fairly straightforward , but lengthy @xcite .
the detailed derivation , including the effect of the real part of the diffractive amplitude , is given in ref .
here we present the final result for @xmath24 , @xmath25 where @xmath26 is the nucleon diffractive structure function measured in hard @xmath27 inclusive diffraction ; @xmath28 is the nuclear density ; @xmath29 is the ratio of the real to imaginary parts of the @xmath27 diffractive amplitude ; @xmath30 gev@xmath31 is the slope of the @xmath32 dependence of the diffractive @xmath33 cross section ; @xmath16 is the light - cone fraction of the nucleon momentum carried by the pomeron ( see the discussion below ) ; @xmath34 is the nucleon mass . the nuclear density @xmath28 depends on the transverse coordinate ( impact parameter ) , @xmath35 , and the longitudinal coordinates , @xmath36 and @xmath37 , of the interacting nucleons .
the ordering @xmath38 follows from the space - time evolution of the scattering process .
the @xmath39 factor accounts for the excitation of the intermediate diffractive state denoted by @xmath40 in fig .
[ fig : master1 ] . at high energies ,
the @xmath41 interaction that leads to nuclear shadowing is diffractive in character .
this is represented by the zigzag lines in fig .
[ fig : master1 ] .
it is convenient to think of the zigzag lines as depicting effective pomeron exchanges . in this case
, @xmath16 represents the light - cone fraction of the nucleon momentum carried by the pomeron , @xmath42 , where @xmath43 is the invariant mass of the diffractive state @xmath40 , and @xmath44 is the invariant @xmath33 energy . the lower limit of integration over @xmath16 in eq .
( [ eq : m11 ] ) corresponds to @xmath45 ; the upper limit is determined by the typical cut on @xmath43 , @xmath46 , which arises because of the nuclear form factor .
note that eq .
( [ eq : m11 ] ) is valid independently of the validity of the leading twist approximation for the hard diffraction .
one of the key features of the leading twist theory of nuclear shadowing @xcite is the possibility to predict nuclear shadowing at the level of parton distributions . using the qcd factorization theorems for inclusive dis and hard diffraction in dis
, one can replace the observable structure functions by the corresponding parton distribution .
this is shown in fig .
[ fig : master1b_2009 ] , which represents the multiple scattering series for the quark distribution in nuclei .
a similar graphical representation can also be given for the gluon distribution using a hard probe directly coupled to gluons .
the contribution of graph @xmath19 in fig .
[ fig : master1b_2009 ] , which we denote @xmath47 , is readily obtained from eq .
( [ eq : m2 ] ) , @xmath48 the contribution of graph @xmath20 is obtained from eq .
( [ eq : m11 ] ) , @xmath49 where @xmath50 is the diffractive parton distribution of flavor @xmath8 in a proton . according to the factorization theorem
@xcite , @xmath50 is a leading - twist distribution , whose @xmath51 evolution is given by the dglap equations .
this is supported by the analyses of diffraction at hera @xcite .
therefore , the contribution of @xmath52 to nuclear shadowing is also a leading twist function , which gives the name to the present approach the leading twist theory of nuclear shadowing .
the derivation of the expressions for @xmath53 and @xmath54 is general and model - independent : the only simplifying approximations are the neglect of nucleon correlations in the nuclear wave function and of the @xmath32 dependence of the elementary diffractive @xmath55 amplitude .
graph @xmath20 in figs .
[ fig : master1 ] and [ fig : master1b_2009 ] approximates well nuclear shadowing in the low nuclear density limit , when the interaction with only two nucleons is important .
as one decreases @xmath0 , graph @xmath21 and higher rescattering terms also become progressively important . to evaluate their contribution
, one needs to model the interaction of the intermediate state @xmath40 with the nucleons of the target .
our approach is based on the high - energy formalism of cross section fluctuations @xcite , which provides a good description of the total hadron - nucleus cross sections and , which is far less trivial , of the coherent inelastic diffraction in hadron - nucleus scattering ; for a review and references , see ref .
@xcite . in this formalism , the wave function of a fast projectile ( virtual photon ) is expanded in terms of eigenstates of the scattering operator , @xmath56 .
each eigenstate interacts with target nucleons with a certain cross section @xmath57 .
the probability for the incoming virtual photon to fluctuate in a given eigenstate is given by the distribution @xmath58 .
we explicitly show the dependence of @xmath58 on the parton flavor @xmath8 as a reminder that dis probes a particular parton distribution of the target .
the entire series of multiple interactions shown in fig . [ fig : master1b_2009 ] can be summed by the standard glauber formalism generalized to include cross section fluctuations , see , e.g. , ref .
@xcite . assuming that
@xmath59 such that the multiple interactions can be exponentiated , we obtain @xmath60 where @xmath61 ; @xmath62 denotes the integration over @xmath57 with the weight @xmath58 .
the interaction with @xmath63 nucleons probes the @xmath63th moment of the distribution @xmath58 , @xmath64 .
equation ( [ eq : fluct1 ] ) is valid at small @xmath0 ( high energies ) , when the effect of the finite coherence length ( the coherence length is proportional to the lifetime of the fluctuations @xmath56 ) is unimportant . in this case , the @xmath65 factor in eq .
( [ eq : m12 ] ) can be set to unity .
the contribution to nuclear shadowing from the interaction with @xmath14 nucleons is related in a model - independent way to the diffractive pdfs of the nucleon .
this means that @xmath66 is proportional to the diffractive parton distribution @xcite , @xmath67 therefore , eq . ( [ eq : fluct1 ] )
can be written as @xmath68 in order to cast eq .
( [ eq : fluct2 ] ) into the more standard form @xcite , we reintroduce the dependence on the longitudinal coordinates @xmath36 and @xmath37 , use the definition of @xmath69 from eq .
( [ eq : m17 ] ) , and identically rewrite eq .
( [ eq : fluct2 ] ) in the following form , @xmath70 finally , we restore the effect of the finite coherence length by reintroducing the @xmath39 factor , replace @xmath71 by @xmath72 to have the correct number of the nucleon pairs , and obtain our general expression for the nuclear parton distribution modified by nuclear shadowing , @xmath73 the evaluation of nuclear shadowing in the leading twist theory of nuclear shadowing , and eq .
( [ eq : fluct2_c ] ) in particular , does not take into account the possible ultra high - energy branching of the diffractive exchange which would couple to different nucleons of the target ( the so - called triple pomeron fan or enhanced reggeon diagrams ) @xcite . using the model for the interaction with @xmath15 nucleons which takes into account such diagrams , nuclear shadowing in nuclear pdfs
was predicted in refs .
@xcite . the general form of the distribution @xmath58 that enters eq .
( [ eq : fluct2_c ] ) is not known .
however , one can still infer the properties of @xmath58 that determine the strength of nuclear shadowing . for the virtual photon
, @xmath58 is very broad and includes the states @xmath56 that correspond to both small and large cross sections @xmath57 @xcite .
the fluctuations with small cross sections constitute the perturbative contribution to the photon - nucleon cross section ; the fluctuations with large cross sections correspond to the hadronic component of the virtual photon . in practice ,
@xmath74 is dominated by the hadronic - size configurations .
this expectation is based on the qcd aligned jet model @xcite , and agrees well with the final analyses of the hera data on hard diffraction which find that @xmath75 @xcite , which is practically the same as in soft processes , @xmath76 @xcite . hence , the diffractive state @xmath40 in figs .
[ fig : master1 ] and [ fig : master1b_2009 ] is dominated by the large-@xmath57 hadron - like fluctuations . the key feature of eq .
( [ eq : fluct2_c ] ) is that it separates the contributions of the small and large cross sections [ this was the main purpose of rewriting eq .
( [ eq : fluct1 ] ) in the form of eq .
( [ eq : fluct2 ] ) which led to eq .
( [ eq : fluct2_c ] ) ] .
while the fluctuations with large cross sections contribute to all moments @xmath77 , the fluctuations with small cross sections contribute significantly only to @xmath78 and @xmath74 , i.e. , to the @xmath79 term and the double scattering term proportional to @xmath50 .
therefore , since the @xmath80 term in eq .
( [ eq : fluct2_c ] ) probes the higher moments of @xmath58 , @xmath81 with @xmath82 , it can be evaluated with the distribution @xmath58 , which neglects the small-@xmath57 perturbative contribution and uses only the information on cross section fluctuations from soft hadron - hadron scattering . in particular , we assume that the relevant @xmath58 is equal to the distribution over cross section fluctuations for the pion .
the @xmath80 term in eq .
( [ eq : fluct2_c ] ) can be identically expanded in terms of @xmath83 with @xmath82 .
we have just explained that the required distribution @xmath58 is dominated by soft hadron - like fluctuations . for such fluctuations ,
the dispersion of @xmath58 does not lead to significant modifications of higher moments of @xmath58 , and it is a good approximation to use @xmath84 for all @xmath85 , which we shall call the _ color fluctuation approximation_. therefore , the @xmath80 term in eq .
( [ eq : fluct2_c ] ) is expressed in terms of a single cross section , @xmath86 , @xmath87 applying the color fluctuation approximation to eq .
( [ eq : fluct2_c ] ) , we obtain our final expression for the nuclear parton distributions modified by nuclear shadowing , @xmath88 in the treatment of multiple rescatterings in the leading twist theory of nuclear shadowing in ref .
@xcite , one used the so - called quasi - eikonal approximation , which prescribes the use of @xmath89 in eq .
( [ eq : fluct5 ] ) . while the quasi - eikonal and color fluctuation approximations give identical results for the interaction with two nucleons of the nuclear target , the color fluctuation approximation provides a more accurate treatment of the interaction with three and more nucleons . in particular , the interaction with three nucleons is treated exactly in the formalism of cross section ( color ) fluctuations . as follows from the definition and modeling ( see below ) of the effective rescattering cross section @xmath86 , the color fluctuation approximation represents the scenario corresponding to the lower limit on nuclear shadowing within the framework of the leading twist nuclear shadowing . to model the distribution @xmath58
, we assume that @xmath90 , where @xmath91 is the distribution over cross sections for the pion .
this assumption attempts to capture the observation which we explained above that for the higher rescattering contributions to nuclear shadowing , only large - size fluctuations of the virtual photon wave function are important .
this is a somewhat extreme assumption which results in the smallest nuclear shadowing ( see fig . [
fig : lt_predictions ] below ) .
the distribution @xmath91 is conveniently parameterized in the following form @xcite : @xmath92 the parameters @xmath93 , @xmath94 and @xmath95 are constrained by following requirements : @xmath96 where @xmath97 is the total pion - nucleon cross section ; @xmath98 is the parameter characterizing the dispersion of the distribution @xmath91 . both @xmath97 and @xmath98
depend on @xmath99 . in our numerical analysis
, we used the donnachie - landshoff parameterization for @xmath97 @xcite : @xmath100 note that in our calculations , we effectively use only the first term in eq .
( [ eq : sigma_pion ] ) , see also fig .
[ fig : sigma3_2009 ] .
the parameter @xmath98 decreases with increasing energy @xcite , which means that cross section fluctuations decrease with increasing energy .
for the pion projectile , @xmath101 at the pion energy of @xmath102 gev @xcite , which corresponds to @xmath103 gev@xmath12 . at the cdf energy of @xmath104 gev@xmath12 , @xmath105 , where the factor @xmath106 is @xmath98 for the proton at the cdf energy @xcite and the factor @xmath107
reflects the constituent quark counting @xcite . assuming a simple linear interpolation between the two energies , we arrive at the following model for @xmath98 : @xmath108 where @xmath109 gev@xmath12 and @xmath110 gev@xmath12 .
equations ( [ eq : psigma_moments ] ) , ( [ eq : sigma_pion ] ) and ( [ eq : omega_sigma ] ) fully determine @xmath91 and @xmath111 and their energy ( bjorken @xmath0 ) dependence . in order to keep track of our modeling @xmath90 , it is also convenient to introduce the notation @xmath112
. figure [ fig : sigma3_2009 ] presents @xmath112 and @xmath69 as functions of bjorken @xmath0 at fixed @xmath113 gev@xmath12 ( @xmath114 ) .
the left panel corresponds to the @xmath115-quark ; the right panel corresponds to gluons .
( note that @xmath111 is flavor - independent in our model . )
the fact that @xmath116 , which is equivalent to @xmath117 , is a general property of the distribution @xmath58 .
equation ( [ eq : fluct5 ] ) allows one to calculate nuclear pdfs modified by nuclear shadowing .
the key inputs for this calculation are the diffractive parton distributions @xmath50 and the slope of the diffractive @xmath33 cross section @xmath118 .
the current experimental uncertainties of @xmath50 and @xmath118 lead to an uncertainty in the predictions of nuclear shadowing which is much larger than the uncertainty associated with the use of the color fluctuation approximation instead of the complete treatment of color fluctuations .
it is also important to note that @xmath111 depends weakly on energy ( bjorken @xmath0 ) .
hence , measuring nuclear shadowing with one nucleus at e.g. , @xmath119 , will further improve our predictions for all @xmath0 and @xmath6 .
figure [ fig : lt_predictions ] presents the ratio of the nuclear to nucleon parton distributions , @xmath120 $ ] , as a function of bjorken @xmath0 at the input scale @xmath113 gev@xmath12 .
the solid curves correspond to the color fluctuation approximation , eq .
( [ eq : fluct5 ] ) , and @xmath121 , eqs .
( [ eq : psigma_moments ] ) , ( [ eq : sigma_pion ] ) and ( [ eq : omega_sigma ] ) .
the dotted curves are obtained using the quasi - eikonal approximation with @xmath122 in eq .
( [ eq : fluct5 ] ) .
the two left panels correspond to @xmath115-quarks ; the two right panels correspond to gluons .
note that we added the effect of antishadowing for the gluon distribution for @xmath123 using the method described in ref .
the panels in the top row are for @xmath124ca ; the bottom panels are for @xmath125pb . as one can see from fig .
[ fig : lt_predictions ] , the color fluctuation approximation corresponds to the smaller nuclear shadowing at the input scale @xmath113 gev@xmath12 than the quasi - eikonal approximation .
for instance , at @xmath126 and for @xmath124ca , @xmath127 and @xmath128 ( the subscripts indicate the color fluctuation and quasi - eikonal approximations , respectively ) . at @xmath126 and for @xmath125pb , @xmath129 and @xmath130 .
we described the modifications of the nuclear pdfs at small @xmath0 at the input scale @xmath113 gev@xmath12 .
predictions for nuclear pdfs at higher scales @xmath131 are obtained using the standard dglap evolution equations with the input given by eq .
( [ eq : fluct5 ] ) .
as one increases @xmath51 , at a given @xmath0 , the difference between the predictions of the color fluctuation and quasi - eikonal approximations reduces : in the dglap evolution , the pdfs at small @xmath0 and large @xmath51 are obtained from the region of larger @xmath0 and smaller @xmath51 , where nuclear shadowing is smaller .
one should also keep in mind that while we present our predictions for nuclear pdfs all the way down to @xmath132 , our results for @xmath133 should be considered only as guiding ones since , for this kinematic region , various effects beyond the leading twist dglap equation should start becoming important .
the leading twist theory of nuclear shadowing allows one to predict nuclear parton distributions modified by nuclear shadowing .
nuclear shadowing arises as the effect of the interaction of a hard probe with a parton simultaneously belonging to several nucleons of the nuclear target .
while the leading twist theory of nuclear shadowing gives unambiguous predictions for nuclear shadowing in the case when the interaction with only two nucleons is important , the interaction with @xmath15 nucleons requires a more detailed knowledge of the dynamics of the hard diffractive processes .
we propose a new approach to the treatment of such multiple interactions using the formalism of cross section ( color ) fluctuations which allows us to take into account the presence of both point - like and average hadronic - size configurations in the virtual photon wave function . for practical applications ,
we propose a new approximation the color fluctuation approximation which approximates well the complete treatment of color fluctuations .
this approximation improves the treatment of the multiple interactions in the quasi - eikonal approximation , which works well in soft processes , where color fluctuations are smaller and have a different structure than in dis . using the developed framework
, we present updated predictions for the effect of nuclear shadowing in nuclear parton distributions of heavy nuclei at small @xmath0 . authored by jefferson science associates , llc under u.s .
doe contract no .
de - ac05 - 06or23177 .
government retains a non - exclusive , paid - up , irrevocable , world - wide license to publish or reproduce this manuscript for u.s .
government purposes . supported by doe grant under contract de - fg02 - 93er40771 .
r. brock _ et al . _
[ cteq collaboration ] , rev .
phys . * 67 * , 157 ( 1995 ) .
k. j. eskola , v. j. kolhinen and p. v. ruuskanen , nucl .
b * 535 * , 351 ( 1998 ) .
k. j. eskola , v. j. kolhinen and c. a. salgado , eur .
j. c * 9 * , 61 ( 1999 ) .
k. j. eskola , h. honkanen , v. j. kolhinen and c. a. salgado , phys .
b * 532 * , 222 ( 2002 ) .
k. j. eskola , h. paukkunen and c. a. salgado , jhep * 0807 * , 102 ( 2008 ) .
k. j. eskola , h. paukkunen and c. a. salgado , jhep * 0904 * , 065 ( 2009 ) .
m. hirai , s. kumano and m. miyama , phys .
d * 64 * , 034003 ( 2001 ) .
m. hirai , s. kumano and t. h. nagai , phys .
c * 70 * , 044905 ( 2004 ) .
m. hirai , s. kumano and t. h. nagai , phys .
c * 76 * , 065207 ( 2007 ) .
d. de florian and r. sassot , phys .
d * 69 * , 074028 ( 2004 ) .
k. hencken _ et al .
_ , phys .
rept . * 458 * , 1 ( 2008 ) .
a. deshpande , r. milner , r. venugopalan and w. vogelsang , ann .
nucl . part .
* 55 * , 165 ( 2005 ) . c. aidala
_ , physics opportunities with e+a collisions at an electron ion collider , white paper prepared for the nsac lrp 2007 , april 4 , 2007 , available at http://www.eic.bnl.gov/. j. c. collins , phys . rev .
d * 57 * , 3051 ( 1998 ) [ erratum - ibid .
d * 61 * , 019902 ( 2000 ) ] .
_ et al . _
[ h1 collaboration ] , eur .
j. c * 48 * , 715 ( 2006 ) .
a. aktas _ et al . _
[ h1 collaboration ] , eur .
j. c * 48 * , 749 ( 2006 ) .
s. chekanov [ zeus collaboration ] , nucl .
b * 800 * , 1 ( 2008 ) .
l. frankfurt , v. guzey and m. strikman , phys .
d * 71 * , 054001 ( 2005 ) .
v. n. gribov , arxiv : hep - ph/0006158 . v. n. gribov , sov .
jetp * 30 * , 709 ( 1970 ) [ zh . eksp .
teor . fiz .
* 57 * , 1306 ( 1969 ) ] . m. l. good and w. d. walker , phys .
rev . * 120 * , 1857 ( 1960 ) .
b. blaettel , g. baym , l. l. frankfurt , h. heiselberg and m. strikman , phys .
d * 47 * , 2761 ( 1993 ) .
l. frankfurt , v. guzey and m. strikman , j. phys .
g * 27 * , r23 ( 2001 ) .
a. schwimmer , nucl .
b * 94 * , 445 ( 1975 ) .
n. armesto , a. capella , a. b. kaidalov , j. lopez - albacete and c. a. salgado , eur .
j. c * 29 * , 531 ( 2003 ) .
k. tywoniuk , i. arsene , l. bravina , a. kaidalov and e. zabrodin , phys .
b * 657 * , 170 ( 2007 ) .
l. frankfurt , a. radyushkin and m. strikman , phys .
d * 55 * , 98 ( 1997 ) .
l. frankfurt , v. guzey and m. strikman , phys .
d * 58 * , 094039 ( 1998 ) .
h. abramowicz , l. frankfurt and m. strikman , econf * c940808 * , 033 ( 1994 ) [ surveys high energ .
phys . * 11 * , 51 ( 1997 ) ] .
a. donnachie and p. v. landshoff , phys .
b * 296 * , 227 ( 1992 ) .
|
the leading twist theory of nuclear shadowing predicts the shadowing correction to nuclear parton distributions at small @xmath0 by connecting it to the leading twist hard diffraction in electron - nucleon scattering .
the uncertainties of the predictions are related to the shadowing effects resulting from the interaction of the hard probe with @xmath1 nucleons .
we argue that the pattern of hard diffraction observed at hera allows one to reduce these uncertainties .
we develop a new approach to the treatment of these multiple interactions , which is based on the concept of the color fluctuations and accounts for the presence of both point - like and hadron - like configurations in the virtual photon wave function . using the developed framework
, we update our predictions for the leading twist nuclear shadowing in nuclear parton distributions of heavy nuclei at small @xmath0 .
|
experimental measurements of heavy - light semileptonic decays , combined with theoretical input , can be used to extract the cabibbo - kobayashi - maskawa ( ckm ) matrix elements @xmath3 , @xmath4 , @xmath5 , and @xmath6 .
the determination of these matrix elements provides constraints on the ckm unitarity triangle and thus tests the standard model .
conversely , @xmath5 and @xmath6 are known with high precision , and can be used to test the lattice techniques involved in calculating the decay rates of @xmath7 mesons .
recent progress reports of current @xmath7 meson semileptonic decay calculations are presented in refs .
the most recent study where systematic errors are taken into account was presented in ref .
@xcite . in this paper
we study the efficiency of two stochastic propagator based alternatives to the sequential propagator method used in all of the above calculations .
the calculation we focus on is the semileptonic decay of a heavy - light pseudoscalar meson ( @xmath8 ) to a light - light pseudoscalar meson ( @xmath9 ) . for processes
@xmath10 the differential decay rate can be parameterized as a product of ckm matrix elements , perturbatively known quantities , and poorly known non - perturbative quantities , @xmath11 where @xmath12 is the squared difference between the initial and final state four - momenta and @xmath13 .
the greatest source of uncertainty in the theoretical calculation of the decay rate is due to the non - perturbative interactions parameterized by the form factor @xmath14 .
these interactions appear in the hadronic matrix element @xmath15 , where @xmath16 is a weak flavor - changing vector current and @xmath17 is the charm heavy - quark and @xmath18 is the @xmath19 or @xmath20 daughter - quark spinor .
the matrix element can be parameterized as a linear combination of the vector , @xmath21 , and scalar , @xmath22 , form factors , @xmath23 on the lattice the matrix element is extracted from three - point functions of the following form , @xmath24 where @xmath25 is the spectator - antiquark spinor , @xmath26 is the propagator for quark flavor @xmath27 , and @xmath28 is the time - slice at which the sink is fixed .
the three - point function is shown schematically in fig .
[ fig : c3 ] . in the limit of large time separations eq .
( [ eq : c3 ] ) has the form , @xmath29 the wave - function amplitudes , @xmath30 and @xmath31 , and energies can be removed by forming ratios with or performing simultaneous fits to the @xmath32 and @xmath33 meson two - point functions , @xmath34 ) .
point - to - all propagators are used for the daughter - quark and spectator - antiquark .
the heavy - quark propagator is constructed either by sequential or by stochastic all - to - all propagator techniques .
[ fig : c3 ] ] currently , the approach of most lattice calculations of semileptonic decay rates of @xmath7 mesons involves constructing the three - point functions using a so - called sequential ( or extended ) propagator , the sequential propagator method @xcite . as is shown in eq .
( [ eq : c3 ] ) , this method uses two light - quark , point - to - all propagators @xmath35 and @xmath36 , which may or may not be mass degenerate .
it is routinely the case that these computationally expensive ( expensive relative to the heavy - quark propagator ) light - quark propagators have been previously generated to study other lattice quantities , and then stored .
for example , the pseudoscalar pion mass is required to guide chiral extrapolations of many lattice quantities . to construct the desired three - point functions ,
the light - quark propagators are combined via the sequential propagator method with the heavy - quark propagator generated on - the - fly " .
the point - to - all light - quark propagators connect the source , at @xmath37 , to the current at @xmath38 and sink at @xmath39 .
a point source is sufficient because of translational invariance .
the heavy - quark propagator , @xmath40^{\dagger } \gamma_5 $ ] in eq .
( [ eq : c3 ] ) , must then connect all lattice sites at the vector current time - slice to all lattice sites at the sink time - slice .
this is achieved by solving the heavy - quark action on the spectator - antiquark propagator sink at a given time - slice @xmath28 , with a momentum and @xmath41 matrix insertion .
although inverting the heavy - quark action with a @xmath42 corresponding to the charm quark mass is inexpensive , relative to the calculation of the light - quark propagators , this approach requires an additional 12 heavy - quark solves for every spectator - quark , @xmath7 meson momentum , and/or @xmath7 meson smearing .
if several combinations of these parameters are wanted , the cost of calculating the heavy - quark propagators becomes significant .
in particular , at small lattice spacings the momenta are more finely grained , and therefore more three - point functions will have an appreciable signal , requiring many more heavy - quark solves if all data are to be utilized . another approach is the one - end method .
this method combines sequential propagator techniques with stochastic propagators , through the so - called one - end trick " as described for two - point functions in refs .
the method has been used successfully in form factor calculations in light - light systems @xcite .
this method provides an additional volume average and momentum projection which may greatly reduce the gauge noise in the correlators .
however , additional propagators are required for every source and sink momentum combination , suggesting it is computationally expensive relative to the sequential propagator method .
nevertheless , there is the possibility that the three - point function noise is so greatly reduced that very few momentum points would be required in the @xmath43 extrapolation and a gain in the overall efficiency is realized . a more flexible method than the previous two that we also investigate in this paper uses an all - to - all @xcite , stochastically generated , propagator to connect the sink to the current .
we will refer to this approach as the stochastic propagator method . with this method all momenta and
sink smearings are in principle available using only a single spin - color inversion .
in addition , no new heavy - quark inversions would be required to calculate the three - point function using multiple ( previously generated ) spectator - quark propagators on the same configuration , as might be useful for chiral extrapolations in the valence light - quark mass .
there is also the interesting possibility that correlators could be generated with multiple current and sink time slices with no additional inversions .
as we will show , the stochastic noise is significant and must be reduced by averaging over multiple noise vectors and partitioning @xcite . in particular , time partitioning is required for efficient reduction of noise , even when using other variance reduction techniques such as the hopping parameter acceleration ( hpa ) @xcite . in the following ,
we study and minimize the stochastic noise appearing in the methods using variance reduction techniques . comparing the errors and costs of the three methods in sec .
[ comps ] , we determine the one - end method to be inefficient but the stochastic propagator method to be promising .
we discuss when the stochastic propagator method is more efficient than the sequential propagator method , and finally apply the method to determine @xmath2 on a single ensemble .
this study is an extension of our preliminary results presented in ref .
.simulation parameters for the 2 qcdsf ensembles @xmath44 and @xmath45 used .
the scale was set from @xmath46 fm , which was obtained by chirally extrapolating the nucleon mass times @xmath47 , @xmath48 , to the physical point at @xmath49 @xcite .
[ tab : params ] [ cols="^,^,^",options="header " , ] as can be seen in the calculation above we chose a very expensive one - end procedure to perform this calculation , the one - end sink method ; however , even the one - end source method would be much costlier than the sequential or stochastic propagator method .
we mention again that the one - end sink method was chosen to minimize the error , and so the cheaper one - end source method would be even noisier
. the one - end methods at these quark masses and volumes are evidently not close to being competitive with the other two approaches , and we drop these methods from further consideration .
+ in this section we compare the two more promising methods , as suggested by the plots and rough cost calculation from the last subsection .
we perform a direct comparison of the sequential and stochastic propagator methods to determine when it is advantageous to use the stochastic propagator method .
+ as a first step in an effort to understand the behavior of the noise present in both methods we examine the effects of smearing . @xmath50 with two different smearings
is shown in figs .
[ smrcomp : a][smrcomp : b ] for the two ensembles .
we remind the reader that in this setting , in the limit @xmath51 , the stochastic method will yield exactly the same result as the sequential method . the results from 100 configurations on ensemble @xmath44
are presented in fig .
[ smrcomp : a ] , and for 220 configurations on ensemble @xmath52 in fig .
[ smrcomp : b ] .
the behavior of the gauge noise can be directly seen from the sequential propagator method correlators .
it decreases slightly with distance from the @xmath7 meson sink regardless of smearing .
the contribution of the stochastic error can be determined by comparing with this purely gauge error : for all smearings and both ensembles it is negligible compared to the gauge noise near the stochastic source at the @xmath7 meson sink , but grows mildly with distance from the source .
the effects of smearing on the gauge and stochastic noises are both mild , and smearing clearly helps to produce cleaner plateaus .
we now compare the total errors of both methods using the five statistically cleanest combinations of correlators with finite momenta on ensemble @xmath45 , where each combination of correlators corresponds to a particular @xmath53 and @xmath14 .
when averaging over so many equivalent correlators it is more straightforward to extract the form factors from the appropriate combination of matrix elements and kinematic factors , and then perform the average .
the energies needed for this construction are determined from bootstrap ensembles of two - point fits .
these are then combined with the corresponding bootstrap ensembles of the three - point correlators .
we assume only three @xmath7 meson momenta are desired ( on these lattices , @xmath7 mesons with momentum @xmath54 have significantly distorted dispersion relations ) : @xmath55 we average over all rotationally equivalent correlators to improve the statistics ; however , there are many correlators with rotationally equivalent sink momenta which are not available to the sequential propagator method but are essentially free for the stochastic propagator method .
these additional momenta are utilized in the stochastic case .
we also consider the same momentum range for the pion ( on these lattices and valence quark masses correlators with @xmath56 are too noisy to produce a clear signal ) .
both methods have the same rotationally equivalent source momenta available . figs .
[ fig : a][fig : e ] are representative of the range of behavior of the relative errors in our data set .
the number of correlators averaged over is shown below each figure for both methods .
the sequential propagator method uses @xmath57 heavy - quark inversions to generate the correlators .
the stochastic propagator method uses @xmath58 . as suggested in figs .
[ fig : err ] and [ smrcomp : a][smrcomp : b ] , less vectors could be used without a significant increase of the error . however , as seen in the cost estimation in the previous section of a realistic calculation , choosing @xmath59 is still cheaper ( although modestly ) than the sequential method . if we performed a naive fixed cost comparison with no assumptions about what constitutes a realistic data set , only 9 spin diluted vectors ( requiring 36 heavy - quark solves ) could be used , and the stochastic contribution to the total error in the stochastic propagator method would increase approximately by a factor of 1.6 . as the stochastic error is a small percentage of the total error ( the worst case appears in fig .
[ smrcomp : b ] in time slices 1 - 10 , where the stochastic error accounts for @xmath60 of the total error ) the stochastic propagator method would still be competitive with the sequential propagator method for most data points .
+ + + we have examined all correlators with an appreciable signal and draw the conclusion that with the modest number of 24 full time and spin partitioned stochastic estimates the error in the stochastic correlators is at worst comparable to the error in the sequential correlators , while at best the additional data available result in smaller total errors for the stochastic propagator method .
it is also clear that the stochastic propagator method is much more flexible , because additional sink smearings , sink interpolating fields , and spectator quark masses do not add significantly to the cost of the calculation . in the next section we use
the stochastic propagator method generated correlators to perform a @xmath43 extrapolation .
in order to connect our results to observables of phenomenological interest we perform the matching and @xmath1 improvement of the vector current .
the matching calculation takes the form , @xmath61\,,\ ] ] where @xmath62 is the tensor current with @xmath63 $ ] .
the matching factor @xmath64 is known non - perturbatively @xcite , while the coefficient of the improvement term , @xmath65 , is known to one - loop in perturbation theory @xcite .
we have examined the effects of @xmath1 improvement and they do not change the behavior of the noise .
it is interesting to note that a similar relation to that used in ref .
@xcite , where @xmath53 can be directly related to the scalar current , holds to the order at which we are working .
the observation the authors make in ref .
@xcite is that the continuum ward identity , @xmath66 is valid on the lattice up to a multiplicative matching factor . for the hisq action @xcite , which they use
, they obtain , @xmath67 this relation holds because the right hand side is renormalization group invariant when the same action is used for both valence quarks . due to the anti - symmetry of the tensor operator
this relationship is also valid when using wilson - type quarks , and allows us to calculate @xmath53 : @xmath68 where @xmath69 .
we extract @xmath70 using this approach as well as using the conventional approach with the vector current . as in the comparisons shown in fig .
[ fig : ae ] , to extract the form factors we first combine the appropriate three - point correlators , which are proportional to the vector matrix element , using the same bootstrap samples for all rotationally equivalent components , so that the resulting correlators in time are proportional to the form factors : @xmath71 we then form the ratio from the bootstrap ensembles of our two - point and three - point functions , @xmath72 and remove @xmath73 and @xmath74 using our two - point fit results . it can be seen in fig .
[ fig : ae ] that the plateaus in the first two correlators are not clear .
the lack of clear plateaus requires us to include excited states in our fits . to do this we determine the energy differences between the first excited state and ground state from two - state fits to the two - point functions . using the energy differences we can perform a three parameter fit to extract the form factors , @xmath75 and
obtain reasonable @xmath76 values .
although plateau fits were suitable for the remaining correlators we found additional time - slices and wider fitting windows could be included using eq .
( [ plat ] ) for all correlators , and use this fit model for all form factor results presented here . renormalized and @xmath77
improved lattice results using the stochastic sink method for @xmath14 and @xmath53 calculated on the @xmath78 ensemble are presented in fig .
the form factors are plotted versus @xmath79 , where @xmath80 gev @xcite .
530 configurations with 24 stochastic vectors and 4 different time sources on every lattice were used .
we also present results , in fig .
[ ffs24 ] , for the same calculation using the relation eq .
( [ f0tos ] ) .
this is advantageous because no matching coefficients are required , removing a source of systematic uncertainty .
the two approaches to constructing the @xmath22 correlators are in all cases comparable within statistical errors ; however , we find the statistical errors of the scalar current based data to be slightly larger . the @xmath43 extrapolation is performed using the beirevi and kaidalov parameterization @xcite , @xmath81 with @xmath82 .
@xmath83 is the mass of the heavy - light vector meson and @xmath84 , @xmath85 , @xmath86 are fit parameters .
reasonable @xmath76 were obtained using @xmath87 or @xmath88 , with @xmath89 and @xmath90 , respectively .
the errors are statistical only .
of course to extract physical results from any of these methods we would need to perform a chiral and lattice spacing extrapolation .
we note that these results are in agreement with the value @xmath91 , obtained in ref .
@xcite , the most recent lattice calculation where systematic errors are estimated .
we have studied two stochastic based methods , the one - end method and stochastic propagator method , for calculating three - point functions in heavy - light semileptonic decays .
the efficiency of these methods was optimized by minimizing the statistical noise , while taking into account the cost incurred by the different set - ups and variance reduction techniques .
we found that the noise was reduced when placing the stochastic sources at the sink rather than at the current insertion of the three - point functions .
this enables us to vary the position of the current but requires that the distance between source and sink is fixed , very similar to the sequential propagator method .
the one - end method naively appears promising because of the additional volume average it induces , which also has the effect of projecting out the desired momentum at the source .
this creates the possibility of reducing the statistical errors , but at a high computational cost . in our analysis
this method clearly fails to outperform the other methods , suggesting it is a poor alternative in heavy - light three - point functions .
the stochastic propagator method is competitive with the sequential propagator method , and in many cases superior , due to the additional momenta freely available .
the stochastic propagator method is also more versatile , since correlators with different sink smearings , sink interpolating operators , and spectator quark masses can be computed at negligible cost .
this flexibility would allow calculations with the variational method and/or partial quenching to be performed with greater efficiency .
this should further improve with finer lattice spacing , where the momenta are more finely grained .
we have used the stochastic propagator method to calculate the semileptonic form factors at many different values of @xmath92 s , using both the vector current and scalar current .
our data are parameterized well by the beirevi and kaidalov expressions .
the results for the @xmath43 extrapolation , @xmath89 and @xmath90 , calculated from the vector and scalar current respectively , are comparable to previous determinations . in the future
we will extend this calculation to additional quark masses and lattice spacings , taking advantage of the method s flexibility to improve plateaus and reduce statistical errors .
the chroma software suite @xcite was used extensively in this work .
the simulations were run on the athene cluster of the university of regensburg .
our work is supported by the dfg sonderforschungsbereich / transregio 55 and by the research executive agency of the european union under grant pitn - ga-2009 - 238353 ( itn strongnet ) .
sara collins acknowledges support from the claussen - simon - foundation ( stifterband fr die deutsche wissenschaft ) .
99 j. a. bailey _
[ fermilab lattice and milc collaborations ] , pos * lat2009 * , 250 ( 2009 ) , e - print arxiv:0912.0214 [ hep - lat ] . h. na , c. t. h. davies , e. follana , p. lepage and j. shigemitsu , pos * lat2009 * , 247 ( 2009 ) , e - print arxiv:0910.3919 [ hep - lat ] .
s. di vita , v. lubicz , c. tarantino , b. haas , f. mescia and s. simula [ etm collaboration ] , pos * lat2009 * , 257 ( 2009 ) , e - print arxiv:0910.4845 [ hep - lat ] . c. aubin
_ et al . _
[ fermilab lattice , milc and hpqcd collaborations ] , phys .
rev . lett . * 94 * , 011601 ( 2005 ) , e - print arxiv : hep - ph/0408306 . g. martinelli and c. t. sachrajda , nucl .
b * 316 * , 355 ( 1989 ) .
m. foster and c. michael [ ukqcd collaboration ] , phys .
d * 59 * , 074503 ( 1999 ) , e - print arxiv : hep - lat/9810021 . c. mcneile and c. michael [ ukqcd collaboration ] , phys .
b * 556 * , 177 ( 2003 ) , e - print arxiv : hep - lat/0212020 . s. simula [ etmc collaboration ] , pos * lat2007 * , 371 ( 2007 ) , e - print arxiv:0710.0097 [ hep - lat ] .
a. boyle _ et al .
_ [ rbc and ukqcd collaborations ] , jhep * 0807 * , 112 ( 2008 ) , e - print arxiv:0804.3971 [ hep - lat ] . k. bitar , a. d. kennedy , r. horsley , s. meyer and p. rossi , nucl .
b * 313 * , 348 ( 1989 ) .
s. bernardson , p. mccarty and c. thron , comput .
commun . * 78 * , 256 ( 1993 ) .
j. viehoff _ et al . _
[ t@xmath94l collaboration ] , nucl .
* 63 * , 269 ( 1998 ) , e - print arxiv : hep - lat/9710050 .
w. wilcox , e - print arxiv : hep - lat/9911013 . j. foley , k. jimmy juge , a. ocais , m. peardon , s. m. ryan and j. i. skullerud , comput .
commun . * 172 * , 145 ( 2005 ) , e - print arxiv : hep - lat/0505023 . c. ehmann and g. s. bali , pos * lattice2008 * , 114 ( 2008 ) , e - print arxiv:0903.2947 [ hep - lat ] . c. thron , s. j. dong , k. f. liu and h. p.
ying , phys .
d * 57 * , 1642 ( 1998 ) , e - print arxiv : hep - lat/9707001 . c. michael ,
m. s. foster and c. mcneile [ ukqcd collaboration ] , nucl .
suppl . * 83 * , 185 ( 2000 ) , e - print arxiv : hep - lat/9909036 . g. s. bali , t. dssel , t. lippert , h. neff , z. prkacin and k. schilling , nucl . phys .
* 140 * , 609 ( 2005 ) , e - print arxiv : hep - lat/0409137 . g. s. bali , h. neff , t. dssel , t. lippert and k. schilling [ sesam collaboration ] , phys . rev .
d * 71 * , 114513 ( 2005 ) , e - print arxiv : hep - lat/0505012 .
r. evans , g. bali and s. collins , pos * lat2009 * , 244 ( 2009 ) , e - print arxiv:0911.3254 [ hep - lat ] .
m. gckeler _ et al . _ [ qcdsf and ukqcd collaborations ] , nucl .
b * 812 * , 205 ( 2009 ) , e - print arxiv:0810.3762 [ hep - lat ] .
a. ali khan _ et al .
_ [ qcdsf and ukqcd collaborations ] , nucl .
b * 689 * , 175 ( 2004 ) , e - print arxiv : hep - lat/0312030 .
s. gsken , u. lw , k. h. mtter , r. sommer , a. patel and k. schilling , phys .
b * 227 * , 266 ( 1989 ) .
c. michael , nucl .
b * 259 * , 58 ( 1985 ) .
m. lscher and u. wolff , nucl .
b * 339 * , 222 ( 1990 ) .
t. bakeyev , m. gckeler , r. horsley , d. pleiter , p. e. l. rakow , g. schierholz and h. stben [ qcdsf and ukqcd collaborations ] , phys .
b * 580 * , 197 ( 2004 ) , e - print arxiv : hep - lat/0305014 . m. gckeler _ et al .
_ [ qcdsf collaboration ] , e - print arxiv:1003.5756 [ hep - lat ] . s. sint and p. weisz , nucl . phys .
b * 502 * , 251 ( 1997 ) , e - print arxiv : hep - lat/9704001 .
e. follana _ et al .
_ [ hpqcd and ukqcd collaborations ] , phys .
d * 75 * , 054502 ( 2007 ) , e-=print arxiv : hep - lat/0610092 .
d. becirevic and a. b. kaidalov , phys .
b * 478 * , 417 ( 2000 ) , e - print arxiv : hep - ph/9904490 .
r. g. edwards and b. joo [ scidac , lhpc and ukqcd collaborations ] , nucl . phys .
* 140 * , 832 ( 2005 ) , e - print arxiv : hep - lat/0409003 .
|
we investigate the computational efficiency of two stochastic based alternatives to the sequential propagator method used in lattice qcd calculations of heavy - light semileptonic form factors . in the first method , we replace the sequential propagator , which couples the calculation of two of the three propagators required for the calculation , with a stochastic propagator so that the calculations of all three propagators are independent .
this method is more flexible than the sequential propagator method but introduces stochastic noise .
we study the noise to determine when this method becomes competitive with the sequential propagator method , and find that for any practical calculation it is competitive with or superior to the sequential propagator method .
we also examine a second stochastic method , the so - called one - end trick " , concluding it is relatively inefficient in this context .
the investigation is carried out on two gauge field ensembles , using the non - perturbatively improved wilson - sheikholeslami - wohlert action with @xmath0 mass - degenerate sea quarks .
the two ensembles have similar lattice spacings but different sea quark masses .
we use the first stochastic method to extract @xmath1-improved , matched lattice results for the semileptonic form factors on the ensemble with lighter sea quarks , extracting @xmath2 .
|
the unification of gravity and quantum theory is one of the great challenges of physics .
the last years were dominated by attempts to reach this goal by rather radical new concepts , as is exemplified by the string theory and loop quantum gravity .
then , one of the main points treated in the current literature is the close relation between the quantum structure of the spacetime and its discretization at particular scale . in the pioneering works of the last century ,
the concept of fundamental scale was associated to the minimal length or , geometrically speaking , to the metric ( e.g.through the line element describing the spacetime ) .
actually , contrarily to these prior investigations , arguments favoring the use of the area as a fundamental entity were appearing in recent years .
this fact was motivated strongly by the relation area / entropy of the black hole on one side , and the theoretical structure of theories such as loop quantum gravity ( spin networks ) , dynamical triangulations and however , the string theory on the other side .
two questions motivated by the gravity - quantum unification immediately arise .
one of them without answer until today : is the exact behaviour at macroscopic regime of the quantum gravity thermal or not ?
the another question is : there exists a consistent formulation where the length is the minimum fundamental entity and that the spectrum of such theory meets the correct limits ?
( e.g. correct spacing between levels for n small and for large n ) .
we will demonstrate through this letter that such consistent description certainly exists , and it can contemplate the classical ( thermal ) or not , behaviour of the spectrum . our claim is based in some previous research of one of us trying to give an unambiguous quantum mechanical description of a particle in a general spacetime .
because the introduction of supersymmetry provide new approach to several physical problems of interest , in [ 1 ] a particular interesting riemannian @xmath3 superspace was introduced .
the main feature of this superspace that makes it especially important , is that the corresponding supermetric , which is the basic ingredient of a volkov - pashnev particle action [ 8 ] , is _ invertible and non - degenerate _ , that is , of @xmath4 type in the casalbuoni s classification [ 10 ] as shown in [ 2,3,5 ] , the non - degeneracy of the supermetrics ( and therefore of the corresponding superspaces ) leads to important consequences in the description of physical systems [ 5 ] .
in particular , notorious geometrical and topological effects on the quantum states , namely , _ consistent mechanisms of localization and confinement _ , due purely to the supergeometrical character of the lagrangian . also _ an alternative to the randall - sundrum ( rs ) model without extra bosonic coordinates _
, can be consistently formulated in terms of such non degenerated superspace approach , eliminating the problems that the rs - like models present at the quantum level [ 1,2,3 ] . and from the probabilistic point of view was recently demonstrated in [ 4 ] that , using the probability current as the probability density , the quantum counterpart of the fisher s metric can be exactly implemented being all the relevant quantum operators exactly constructed in a manner that was already inferred in 1988 on a quite general basis by caianiello . in this letter
( strongly motivated by the above results ) we will show , after brief description of the superspace and the emergent spacetime of ref.[1,2,3 ] , that as a result of quantization of this supersystem[2,3 ] , a discrete metric associated with coherent states ( in obedience to a poisson distribution ) , is immediately obtained without prescription of discretization . this discrete solution , that represents an emergent metric , is described by a physical coherent state belonging to the metaplectic group @xmath0 with a poissonian distribution at lower n ( number basis ) restoring the classical thermal continuum behaviour at large @xmath5 number basis@xmath6 , or leading to non - classical radiation states , as is conjectured in a quite general form by the dynamic bekenstein - mukhanov effect .
the results that we present here are absolutely without black / hole entropy arguments given a priory .
finally a discussion linking our results with the black hole entropy and spectrum are giving and some perspectives and future directions of research suggested .
the model introduced in [ 1,8 ] , represents a free particle in a superspace with coordinates @xmath7 .
it is described by the lagrangian density @xmath8 where @xmath9 , and the dot indicates derivative with respect to the parameter @xmath10 , as usual . in coordinates ,
the line element of the superspace reads , @xmath11 is important to note that the quantization was exactly performed by a new method introduced by one of us in [ 2,3 ] given the correct physical and mathematical interpretation to the square root hamiltonian .
the method is based on two fundamental points : first , introducing a modification of lanczos technique [ 2 ] that permits to pick the correct phase space of the problem without modify the form of the relevant quantum geometrical operators ( i.e.the particular form of the hamiltonian or lagrangian remains as square root ) . and second , using the underlying covering group of @xmath12 ( that is the metaplectic group ) to give a quantum meaning to the radical operator ( lagrangian or hamiltonian ) . with these ingredients
the problem is schematically reduced to @xmath13 were @xmath14 @xmath15 are the momenta corresponding to the supercoordinates . without lose generality and for simplicity , the ` squared ' solution with three compactified dimensions ( @xmath16 spin fixed ) is [ 1,3,5 ] @xmath17 where the initial values of the metric components are given by @xmath18{c}a\\ a^{\dagger}\end{array } \right ) _ { ab}|\psi(0)\rangle,\ ] ] or , explicitly , @xmath19 it worth mention here that these components were obtained in the simplest case in [ 9 ] .
the bosonic and spinorial parts of the exponent in the superfield solution ( [ cssol ] ) are , respectively , @xmath20{rcl}a(t ) & = & -\left ( \frac{m}{|{\mathbf{a}}|}\right ) ^{2}t^{2}+c_{1}t+c_{2},\\ \xi\varrho\left ( t\right ) & = & \xi\left ( \phi_{\alpha}(t)+\bar{\chi}_{\dot{\alpha}}(t)\right ) \\ & = & \theta^{\alpha}\left ( \overset{\circ}{\phi}_{\alpha}\cos\left ( \omega t/2\right ) + \frac{2}{\omega}z_{\alpha}\right ) -\overline{\theta}^{\overset{\cdot}{\alpha}}\left ( -\overset{\circ}{\overline{\phi}}_{\overset{\cdot}{\alpha}}\sin\left ( \omega t/2\right ) -\frac{2}{\omega } \overline{z}_{\overset{.}{\alpha}}\right ) \\ & = & \theta^{\alpha}\overset{\circ}{\phi}_{\alpha}\cos\left ( \omega t/2\right ) + \bar{\theta}^{\overset{\cdot}{\alpha}}\overset{\circ}{\overline{\phi}}_{\overset{\cdot}{\alpha}}\sin\left ( \omega t/2\right ) + 4|\mathbf{a}|re(\theta z ) , \end{array } \label{expo}\ ] ] where @xmath21 are constant spinors , @xmath22 and the constant @xmath23 , due to the obvious physical reasons and the chiral restoration limit of the superfield solution [ 1,3,5 ] .
now we will see how the discrete spacetime structure naturally arise from the model , . expanding on a number basis , as usual @xmath24 we have @xmath25 then @xmath26@xmath27{c}a\\ a^{\dagger}\end{array } \right ) _ { ab}|n\rangle=\left ( \begin{array } [ c]{c}\langle m|n-1\rangle\sqrt{n}\\ \langle m|n+1\rangle\sqrt{n+1}\end{array } \right ) _ { ab}=\left ( \begin{array } [ c]{c}\delta_{m , n-1}\sqrt{m}\\ \delta_{m , n+1}\sqrt{m+1}\end{array } \right ) _ { ab } \tag{7}\ ] ] it follows@xmath28{c}\delta_{m , n-1}\sqrt{m}\\ \delta_{m , n+1}\sqrt{m+1}\end{array } \right ) _ { ab}\langle n|\psi(0)\rangle\]]@xmath29{c}1\\ 0 \end{array } \right ) _ { ab}+\sum_{m}\sqrt{n+1}\langle\psi(0)|n+1\rangle\langle n|\psi(0)\rangle\left ( \begin{array } [ c]{c}0\\ 1 \end{array } \right ) _ { ab}\ ] ] from the equation above we see that the only clear sense for it is due the decomposition of @xmath30 into the basic states of the metapletic representation @xmath31 where the constants @xmath32 and @xmath33 are arbitrary and they control the classical behavior of the spectrum at the macroscopic level .
we , without lose generality in this part of the discussion , take @xmath34 such that @xmath35 but we will return to this important point later .
this , in fact , is the effect of the decomposition of the so(2,1 ) group in two irreducible representations of the metaplectic group mp(2 ) : spanning even and odd n respectively .
the important feature of the state@xmath36 is that is invariant ( if @xmath34 ) to the action of operators @xmath37 and @xmath38 .
this fact is because in the metaplectic representation the general behaviour of these states are : @xmath39 and @xmath40(we will not enter in more details here ) .
is easily checked from the poissonian distribution for the coherent states : @xmath41 obeying @xmath42 @xmath43 that it differs with the individual distributions coming from each one of the two irreducible representations of the metaplectic group mp(2 ) ( spanning even and odd n respectively ) : . @xmath44
although the different form between above equations , the limit n@xmath45 is the same for the sum of the two distributions coming from the mp@xmath46 irreducible representations ( irreps ) . and
for the so(2,1 ) representation as it should be . having this in mind
, the specific form of @xmath47 was given in [ 1,2,3 ] and are @xmath48 where in the parameter @xmath49 all the possible remaining @xmath50 ( odd ) dependence is stored .
then , we arrive to following result @xmath51 \left ( \begin{array } [ c]{c}1\\ 0 \end{array } \right ) _ { ab}+\right .
\nonumber\\ + \left .
\left [ p_{\alpha_{+}^{\ast}}(2m)\cdot2m+p_{\alpha_{-}^{\ast}}(2m+1)\cdot(2m+1)\right ] \left ( \begin{array } [ c]{c}0\\ 1 \end{array } \right ) _ { ab}\right\ } \tag{11}\ ] ] this expression is the core of our discussion : it shows explicitly the discrete structure of the spacetime as the fundamental basis for a consistent quantum field theory of gravity . by the other hand , when we reach the limit n@xmath52 the metric solution goes to the continuum due : @xmath53 = \underset{n=0}{\overset{\infty}{\sum } } \left [ p_{\alpha_{+}}(2m+2)\cdot\left ( 2m+2\right ) + p_{\alpha_{-}}(2m+1)\cdot(2m+1)\right ] = \alpha e^{-|\alpha|}\left ( \cosh(\alpha ) + \sinh(\alpha)\right ) = \alpha$ ] and similarly for the lower part ( spinor down ) of above equation @xmath54 = \alpha^{\ast}.$ ] consequently , when the number of levels increase the metric solution goes to the continuum `` manifold '' general relativistic behaviour:@xmath55{c}1\\ 0 \end{array } \right ) _ { ab}+\alpha^{\ast}\left ( \begin{array } [ c]{c}0\\ 1 \end{array } \right ) _ { ab}\right\ } = f\left ( t\right ) \langle\psi(0)|\left ( \begin{array } [ c]{c}a\\ a^{\dagger}\end{array } \right ) _ { ab}|\psi(0)\rangle\tag{12}\ ] ] as expected .
is not difficult to see that for @xmath56 the metric solution takes its minimal value @xmath57{c}1\\ 0 \end{array } \right ) _ { ab}+p_{\alpha_{-}^{\ast}}(1)\left ( \begin{array } [ c]{c}0\\ 1 \end{array } \right ) _ { ab}\right ] \nonumber\\ & = \frac{f\left ( t\right ) } { 2}e^{-\left\vert \alpha\right\vert } \left [ \alpha\left ( \begin{array } [ c]{c}1\\ 0 \end{array } \right ) _ { ab}+\alpha^{\ast}\left ( \begin{array } [ c]{c}0\\ 1 \end{array } \right ) _ { ab}\right ] \tag{13}\ ] ] this evidently defines the minimal length due the metric axioms in a riemannian manifold .
this point will be not analyzed fully here , but in principle ( due the existence of discrete poincare subgroups of this supermetric ) fundamental symmetries as the lorentz one , can be preserved at this level of discretization .
some of interested issues related to this problem in general were given in ref . [ 11 ] .
as is well known , the black hole entropy @xmath58 where is the horizon area and @xmath59 is the planck length was first found by bekenstein and hawking [ 18 ] using thermodynamic arguments of preservation of the first and second laws of thermodynamics . an information theory proof was also found by bekenstein in which can be treated as the measure of `` inaccessibility '' of the information of an external observer on an actual internal configuration of the black hole realized in a given state ( described by values of mass , charge , and angular momentum ) `` .
the first controversial thing that immediately appears from the point of view of statistical mechanics ( in which the entropy is the mean logarithm of the density matrix ) is that the entropy of a black hole is proportional to its surface area . about this issue , bekenstein propose a model of quantization of the horizon area in the section with the suggestive title ' ' demystifying black hole s entropy proportionality to area " in ref.[19 ] . resuming the proposal ,
the horizon is formed by patches of equal area @xmath60 ( however , which are added one after another at a time ) .
their standard size is important and makes them all equivalent .
the horizon can be regarded as having many degrees of freedom , one per each patch , due it is made from equivalent patches all with the same number @xmath61 of quantum states .
consequently , the total number of quantum states of the horizon is @xmath62 and the statistical ( boltzmann ) entropy associated with the horizon is @xmath63 : putting @xmath64 bekenstein arrived to the expected thermodynamical black hole formula .
as was pointed out in [ 20 ] , bekenstein do nt gives account that putting @xmath65 into the original black hole entropy formula we obtain the poisson expression for the total number of states , @xmath66 being this precisely _ the link _ with the structure of the emergent coherent state metric . considering the similar poissonian expression for the number of states from @xmath67 , namely @xmath68 , the relation between the coherent state eigenvalue @xmath69 corresponding to the solution ( 11 ) and the above equation is clear:@xmath70 this expression relates the space of phase of the coherent state solution @xmath67 and the @xmath71 through @xmath72the @xmath73 and @xmath74 the linear behaviour of area and entropy with respect to@xmath69 is given in figure 1 ( this important issue will be focused by the authors in a separate work ) .
recently was discussed the fact if the black hole hawking s radiation is finally thermal or it can be quantically affected , as suggested again by bekenstein and mukhanov [ 13 ] . due the interplay between the area of the black hole surface and the black hole mass , it is likely to be quantized as well .
then , the mass of the black hole decreases when radiation is emitted due the quantum jump from one quantized value of the mass ( energy ) to a lower quantized value . in consequence ( because radiation is emitted at quantized frequencies , corresponding to the differences between energy levels ) quantum gravity implies a discretized emission spectrum for the black hole radiation . as is well known from the comments of the ( lqg ) loop quantum gravity community [ 12,14,16,17 ]
, the spectral lines can be very dense in macroscopic regimes leading physically no contradiction with hawking s prediction of a continuous thermal spectrum .
was also verified that the behavior of the spectrum is ansatz - dependent at macroscopic regimes : if we pick ( as bekenstein and mukhanov does ) the simplest ansatz for the quantization of the area -that the area is quantized in multiple integers of an elementary area @xmath75 , then the emitted spectrum turns out to be macroscopically discrete , this effect as the kinematical bekenstein - mukhanov effect . by the other hand , into the loop quantum gravity context
, the celebrated thermal spectrum is reached because the density of levels increase in parallel with the number of levels .
then , is possible to circumvented the theoretical dilemma ? : from the point of view of lqg the spectrum is always continuous at macroscopic regime , and from the viewpoint of mukhanov and bekenstein the spectrum may be macroscopically discrete ( but ansatz - dependent finally ) . from the point of view of our model
we can give an affirmative answer to this question : if now we suppose simply that @xmath76 in the state solution ( 8,11 ) , that is one possibility in our proposal due the arbitrariness of the constants a and b , we have @xmath77 then we can not reach the thermal ( hawking ) spectrum at the macroscopic level .
this fact is clear because we need exact balance between the superposition of the two irreducible representations of the metaplectic group as clearly given by expressions ( 11,9 ) .
this will leads , as a result , non classical states of radiation in the sense of [ 21 ] as can be easily seen putting , for example , @xmath33 @xmath78 zero:@xmath79 \left ( \begin{array } [ c]{c}1\\ 0 \end{array } \right ) _ { ab } \tag{16}\ ] ] notice that only the up spinor part survives and the classical ( thermal ) limit is not reached , even in the continuous limit were the number of levels increases accordingly to@xmath80{c}1\\ 0 \end{array } \right ) _ { ab}=af\left ( t\right ) \langle\psi(0)|\left ( \begin{array } [ c]{c}a\\ 0 \end{array } \right ) _ { ab}|\psi(0)\rangle\tag{17}\ ] ] in such a case @xmath81 @xmath82 , the spectrum will takes only even(odd ) levels becoming evidently non thermal .
then , if @xmath34 the kinematical bekenstein - mukhanov effect disappears and the thermal hawking spectrum is reached at the continuum classical gravity level ( the poissonian behaviour of the distribution is complete ) . otherwise , with @xmath76 , the spectrum belongs to a non classical one and the quantum properties of the gravity are macroscopically manifest .
trough this paper we have been shown that a n=1 superspace equipped with a nondegenerate and invertible supermetric where the unconstrained quantization was exactly performed by new methods based on coherent states and respecting the form of the hamiltonian ( modified lanczos technique ) , a discrete structure of spacetime naturally emerges without any prescription of discretization ( in sharp contrast of the other attempts in the literature ) .
due the metaplectic representation ( double covering of the @xmath12 ) of the coherent state solution representing the emergent spacetime , the crossover from the quantum to the macroscopical regime ( classical or not ) is natural and consistent .
this important fact permits us to conciliate the two apparently different pictures of the macroscopical quantum gravity regime given by the lqg claims supporting the hawking ( thermal ) spectrum ) and the dynamical bekenstein - mukhanov effect that point out that quantum ( non thermal ) imprints can survive at macroscopical regime . despite the simplicity of the model introduced here ,
we have been obtained physically and geometrically , an amount of important answers with respect to a consistent quantum gravity formulation .
the main properties that any consistent formulation of quantum gravity must have , in the light of the results presented , are : \1 ) emergent nature of the spacetime .
\2 ) independence of the discretization method .
\3 ) consistent suitable transition to the macroscopic ( classical , semiclassical , etc . ) regime .
\4 ) total and absolute independence of particular solutions or other arguments involving particular geometries ( e.g. black - hole / area and the entropy ) .
\5 ) solutions , arguments involving particular geometries , etc . of the previous point , must be reached by the quantum gravity theory but not depending them at the fundamental level .
djcl is very grateful to the jinr directorate and the bltp for his hospitality and financial support .
[ 1 ] diego julio cirilo - lombardo ; _ the geometrical properties of riemannian superspaces , exact solutions and the mechanism of localization_. physics letters * b * 661,(2008 ) 186 - 191 .
[ 3 ] diego julio cirilo - lombardo ; _ non - compact groups , coherent states , relativistic wave equations and the harmonic oscillator ii : physical and geometrical considerations .
_ found phys 39 ( 2009 ) 373396 . [ 5 ] diego julio cirilo - lombardo ; _ geometrical properties of riemannian superspaces , observables and physical states _ ,
the european physical journal c - particles and fields , ( 2012 ) , volume 72 , number 7 , 2079 [ 7 ] c rovelli , l smolin , _ knot theory and quantum gravity _
phys rev lett 61 ( 1988)1155 , _ loop space representation of quantum general relativity _
, nucl phys * * b**331 ( 1990 ) 80 . for various perspectives on loop quantum
|
we show that the metric ( line element ) is the first geometrical object to be associated to a discrete ( quantum ) structure of the spacetime without necessity of black hole - entropy - area arguments , in sharp contrast with other attempts in the literature . to this end ,
an emergent metric solution obtained previously in [ physics letters b 661 , 186 - 191 ( 2008 ) ] from a particular non - degenerate riemmanian superspace is introduced .
this emergent metric is described by a physical coherent state belonging to the metaplectic group @xmath0 with a poissonian distribution at lower @xmath1 ( number basis ) restoring the classical thermal continuum behaviour at large @xmath1 ( @xmath2 , or leading to non - classical radiation states , as is conjectured in a quite general basis by mean the bekenstein - mukhanov effect .
group - dependent conditions that control the behavior of the macroscopic regime spectrum ( thermal or not ) , as the relationship with the problem of area / entropy of the black hole are presented and discussed .
|
lie - algebraic and lie group theoretic methods have played a significant role in finding exact solutions to the schrdinger equation in quantum mechanics . in the classical applications
, the lie group appears as a symmetry group of the hamiltonian operator , and the associated representation theory provides an algebraic means for computing the spectrum .
of particular importance are the exactly solvable problems , such as the harmonic oscillator or the hydrogen atom , whose point spectrum can be completely determined using purely algebraic methods . the concept of a _
spectrum generating algebra _ dates back to a paper by goshen and lipkin @xcite , and was later rediscovered in the context of high energy physics by two different groups @xcite .
the study of spectrum generating algebras received much impetus in the subsequent years ( see the review @xcite ) and it was soon applied also in the field of molecular dynamics by iachello , levine , alhassid , grsey and their collaborators ( see the book @xcite for a survey of theory and applications ) .
most of the applications of spectrum generating algebras concerned _ exactly solvable _
hamiltonians whose spectrum could be completely determined by algebraic means . an intermediate class of spectral problems
are those for which only a finite part of the point spectrum can be calculated by algebraic methods , but possibly not the whole spectrum .
the usual example of these type of problems is the sextic oscillator in which the potential is an even polynomial of degree six whose coefficients depend on a parameter @xmath1 . for each positive integer value of @xmath1
the hamiltonian is shown to preserve an @xmath2-dimensional subspace of @xmath3 .
it is clear from sturm - liouville theory that the hamiltonian with the sextic potential has an infinite number of bound states , but only @xmath4 of them belong to the so called _ algebraic _ sector .
it was soon realized that the sextic oscillator was just the first example of a large class of systems having this property , and the works of shifman , turbiner , ushveridze @xcite and their collaborators initiated the study of the mathematical properties and physical applications of this new class of spectral problems , which they named _ quasi - exactly solvable_. however , even in the simplest case of one spatial dimension , there is no way to ascertain whether a given potential is quasi - exactly solvable or not , so two methods were proposed to construct large families of quasi - exactly solvable problems : the lie - algebraic method @xcite and the analytic method @xcite .
the idea underlying the lie - algebraic method is to use results from the representation theory of lie algebras to ensure that a given hamiltonian preserves a certain finite dimensional subspace of functions . in one dimension ,
the only algebra of first order differential operators with finite dimensional is @xmath0 , whose generators : @xmath5 preserve the @xmath2-dimensional linear space of polynomials in the @xmath6 variable of degree less than or equal to @xmath1 : @xmath7 the most general second order differential operator that preserves @xmath8 can be written as a quadratic combination of the generators ( [ eq : sl2generators ] ) of @xmath0 @xmath9 where the indexes run over the values @xmath10 .
the operator @xmath11 is said to be _ lie algebraic _ and @xmath0 is often referred to as a hidden symmetry algebra .
the operator @xmath11 does not have in general the form of a schrdinger operator , but it can be transformed into a schrdinger operator by a change of variables and a conjugation by a non - vanishing function .
such a transformation always exists in one spatial dimension , but in more dimensions the equivalence problem remains open @xcite .
this has been a long standing obstacle to classify multidimensional quasi - exactly solvable hamiltonians , where only a few families are known @xcite .
one important fact lies at the core of the lie - algebraic method : it is ensured by burnside s classical theorem that every differential operator which leaves the space @xmath12 invariant belongs to the enveloping algebra @xmath13 , since @xmath14 is an irreducible module for the @xmath0 action .
this is probably the reason for the relative success of the lie - algebraic constructions in the context of quasi - exact solvability , up to the point that the terms _ lie - algebraic _ and _ quasi - exactly solvable _ are often used as synonims in the literature .
however , there are many other finite dimensional polynomial spaces which are not irreducible modules for the @xmath0 action , and in these cases there might be non - lie algebraic differential operators which leave the space invariant .
the first exploration of this type was done by post and turbiner @xcite , who classified all second order differential operators which preserve a polynomial space generated by monomials . in their work
they did not use lie algebras but other considerations based on the grading of the operators and basis elements .
it was later realized that these differential operators can also be transformed into schrdinger form thereby providing new examples of quasi - exactly solvable operators which are not lie algebraic . in this
_ direct approach _ to quasi - exact solvability @xcite more general polynomial spaces are considered and the set of differential operators that preserve them are investigated without any reference to lie algebras .
it was later shown that exactly solvable potentials exist which are not lie algebraic , but can be obtained from a lie algebraic potentials via a state - adding darboux transformation @xcite .
further research along this line suggests to regard a darboux or susy transformation not just as a transformation on the potential and eigenfunctions , but as an algebraic deformation of an infinite polynomial flag @xcite .
non lie - algebraic potentials appear also in the recent work of gonzlez - lpez and tanaka in the context of supersymmetry @xcite .
in contrast with the lie algebraic method , the analytic method has received less attention over the last decades .
however recent results suggest that it is a more suitable approach to encompass this new type of quasi - exactly solvable systems which do not fit in the lie algebraic scheme,@xcite . in this paper
we would like to illustrate the relevance of studying quasi - exact solvability beyond the lie algebraic approach by providing two suitably chosen examples : in section [ sec : more ] we show how more energy levels can be obtained from a lie algebraic potential which can not be obtained in the @xmath0 approach . in section [ sec : nolie ] we show an example of a potential which is quasi - exactly solvable but not lie algebraic .
in this section we expose the first reason to study quasi - exactly solvable problems beyond the lie - algebraic approach .
we present a potential which is known to admit an @xmath0 algebraization , and therefore allows for some of its energy levels to be calculated algebraically , all belonging to the even sector . for this same potential , we show that a different algebraization exists in which the algebraic hamiltonian can not be expressed as an element of the enveloping algebra of @xmath0 . this new algebraization allows to calculate all the previous levels , plus some extra ones .
consider the following schrdinger operator @xmath15 where @xmath16 is a real parameter and where @xmath1 is a positive integer .
this potential is known to be lie - algebraic and it appears for instance in the classification performed by gonzlez - lpez , kamran and olver in @xcite .
the @xmath0 algebraization of this potential is achieved by the transformation @xmath17 with the following choice of gauge factor and change of coordinate @xmath18 up to an additive constant , the transformed operator @xmath19 becomes @xmath20 which is easily seen to preserve the space @xmath8 defined in ( [ pnz ] ) .
the explicit quadratic combination of @xmath19 in terms of @xmath0 generators is ( again up to an additive constant ) @xmath21 as a consequence of this @xmath0 algebraization we can calculate @xmath22 energy levels of the hamiltonian , by diagonalizing the corresponding matrix of the restricted action of @xmath19 to @xmath23 .
the @xmath2 algebraic eigenfunctions have the form @xmath24 where @xmath25 is one of the @xmath2 polynomial eigenfunctions that @xmath19 is ensured to have .
all the eigenfunctions obtained via the lie algebraic method correspond to the even sector . up to this point , all these results are well known .
however , the _ same _ hamiltonian remarkably admits a _ different _
algebraization characterized by the following gauge factor and change of variables : @xmath26 the transformed operator @xmath27 becomes @xmath28 where an additive constant has been dropped .
it is straightforward to realise that due to the rational coefficients in the previous expression , the operator @xmath29 does not belong to the enveloping algebra of @xmath0 .
nevertheless it preserves a @xmath30-dimensional subspace generated by polynomials in the variable @xmath31 that we shall denote as @xmath32 and define as @xmath33 this polynomial subspace is an example of an _ exceptional polynomial module _ ,
the name exceptional being due to the fact that the space of second order operators that leave it invariant have a rich structure @xcite .
an explicit basis of @xmath32 can be given in the following manner @xmath34 this second algebraization allows us to compute @xmath30 algebraic eigenfunctions of the schrdinger operator ( [ eq : sch ] ) , which will be of the form @xmath35 where @xmath36 is one of the @xmath30 polynomial eigenfunctions that @xmath29 is ensured to have .
we observe that these eigenfunctions can be both odd and even , as opposed to the @xmath0 algebraic ones , which were only even .
we thus obtain @xmath4 algebraic eigenfunctions via the @xmath0 algebraization and @xmath30 algebraic eigenfunctions via the exceptional module algebraization .
the situation is depicted schematically in figure [ fig0 ] .
c it is maybe worth to explore the relation between the two different algebraizations .
it is known in the theory of lie algebraic problems that under linear fractional transformations the lie - algebraic character is preserved @xcite .
more specifically , a change of variable by a linear fractional transformation , @xmath37 together with an appropriate gauge transformation preserve the space @xmath23 since @xmath38 moreover , if @xmath19 is lie - algebraic and preserves @xmath12 , then @xmath39 is also lie - algebraic and preserves @xmath40 .
we see thus that projective transformations do not change the @xmath0 character , in fact they amount to a linear transformation of the generators of @xmath0 @xcite .
however , from expressions ( [ mu1]),([change1]),([mu2 ] ) and ( [ change2 ] ) it follows that the relation between the two algebraic operators @xmath19 and @xmath29 in the previous example is given by @xmath41 which is not a projective transformation .
we have performed an explicit computation to show a few of the algebraic eigenfunctions of the hamiltonian ( [ eq : sch ] ) .
for illustrative purposes it suffices to fix the integer parameter @xmath1 to some low value .
for the values @xmath42 and @xmath43 the potential ( [ eq : sch ] ) has the shape of a double - well potential shown in figure [ fig1 ] . for @xmath42 the operator @xmath19 in ( [ eq :
t ] ) satisfies @xmath44 which allows to compute four algebraic eigenfunctions via the @xmath0 algebraization .
the action of @xmath19 relative to the canonical basis of @xmath45 is given by the following @xmath46 matrix : @xmath47 0.1 cm for @xmath43 we have calculated four algebraic eigenfunctions of ( [ eq : sch ] ) which have the form @xmath48 where @xmath49 is one of the four eigen - polynomials of @xmath19 and the gauge factor @xmath50 is given by ( [ mu1 ] ) .
these four eigenfunctions have been plotted in figure [ fig1 ] along with their corresponding energies , which have been normalized so that the energy of the ground state is zero . from the second algebraization we know that the schrdinger operator @xmath51 is conjugate via the change of variables ( [ change2 ] ) and gauge factor ( [ mu2 ] ) to the operator @xmath52 which preserves an exceptional polynomial module .
for @xmath42 this space is spanned by @xmath53 and the matrix of the action of @xmath29 relative to this basis is 0.1 cm @xmath54 0.1 cm for the same value @xmath43 we have calculated six algebraic eigenfunctions of the hamiltonian ( [ eq : sch ] ) which have the form @xmath55 where @xmath56 is one of the six eigen - polynomials of @xmath29 and the gauge factor @xmath57 is given by ( [ mu2 ] ) .
these six algebraic eigenfunctions along with their energies are shown in figure [ fig2 ] .
the first thing to note is that we obtain all the even eigenfunctions from the @xmath0 algebraization plus two extra odd eigenfunctions corresponding to the first and third excited states .
however , the eigenfunction corresponding to the fifth excited state is not present in the algebraic sector . for arbitrary @xmath1
it seems that there is always a gap in the algebraic sector just below the highest energy in the exceptional module algebraization , @xcite .
although all evidences show that this is true in the general case , there is yet no proof of this result .
in the previous section we have seen how the exceptional module algebraization can provide more energy levels from a lie - algebraic potential . in this section
we will show that some schrdinger operators only admit an exceptional module algebraization and not the traditional @xmath0 one .
we can therefore construct new quasi - exactly solvable potentials on the line , which are not in the classifications of qes potentials in @xcite , since those classifications deal only with potentials that admit an @xmath0 algebraization . in this section
we show one first simple example of this phenomenon , postponing the full classification of these new potentials to a future publication,@xcite .
consider the following schrdinger operator : @xmath58 with potential @xmath59a\big ) x^2 + 4 \frac{x^2 - 1}{(x^2 + 1)^2}.\ ] ] if the last rational term were absent , this potential would be the well known quasi - exactly solvable sextic potential discussed for instance in @xcite .
the potential is always even , so its eigenfunctions will have well defined parity .
however , as it happens with the sextic , only even or odd eigenfunctions but not both appear in the algebraic sector . this does not exclude in principle that other algebraizations exist in which both even and odd eigenfunctions are obtained .
in fact , the choice of @xmath60 gives a potential with algebraic even eigenfunctions while @xmath61 corresponds to potentials with odd algebraic eigenfunctions .
for arbitrary values of @xmath62 and @xmath63 the hamiltonian does not preserve any finite dimensional subspace , but for the following values : @xmath64 where @xmath65 is an arbitrary real parameter , the hamiltonian @xmath51 admits an exceptional module algebraization .
more specificallly , the transformation @xmath66 with @xmath67 transforms @xmath51 into the algebraic operator @xmath19 given by @xmath68 up to an additive constant .
the operators @xmath69 , @xmath70 and @xmath71 are given by @xmath72 and every one of them leaves invariant the exceptional module @xmath73 given by @xmath74 the sextic potential ( [ eq : sextic ] ) has eigenfunctions in @xmath3 provided that @xmath16 , or @xmath75 and @xmath76 .
this last case corresponds in fact to an exactly solvable potential , which can be obtained by a susy transformation from the harmonic oscillator @xcite .
the algebraic operator preserves in this last case an infinite flag of exceptional polynomial subspaces . we show some explicit eigenfunctions of this potential corresponding to the even sector by setting @xmath77 and @xmath60 .
the matrix of the action of @xmath19 with respect to the basis ( [ eq : basis ] ) of @xmath78 is @xmath79 for @xmath80 the four polynomial eigenfunctions of @xmath19 are approximately @xmath81where polynomial @xmath25 has @xmath82 zeros .
the corresponding energies are : @xmath83 therefore the four even eigenfunctions of ( [ eq : sextic ] ) coming from the exceptional module algebraization are : @xmath84 where the gauge factor @xmath50 is given by ( [ musextic ] ) with @xmath60 . for this choice of parameters @xmath65 , @xmath85 and @xmath1 the potential , gauge factor and eigenfunctions have been plotted in figure [ fig3 ] .
we observe once again that we obtain eigenfunctions with zero , two , four and eight zeros , but not one with six zeros . for arbitrary @xmath1 the eigenfunction with @xmath86 zeros would be missing form the algebraic sector .
the calculations for the odd case are similar : setting @xmath61 in ( [ eq : sextic ] ) we have calculated @xmath77 odd eigenfunctions which are shown in figure [ fig4 ] .
by means of two examples we provide further arguments to motivate the study of quasi - exactly solvable problems beyond the lie algebraic approach . even in the simplest case of one dimensional quantum mechanical problems the classifications performed in @xcite
do not cover all quasi - exactly solvable potentials .
since lie algebraic potentials are only a subclass of the potentials with partial algebraization of their spectrum , it would be desirable to find a precise mathematical characterization of the latter .
this problem is related to the so called _ generalized bochner problem _ @xcite , that of finding the most general differential operator that preserves a general finite dimensional space of polynomials .
all the new quasi - exactly solvable potentials obtained recently are based on the exceptional modules , but other polynomial subspaces might exist which do not have a monomial basis and yet have a rich structure . in conclusion , the class of hamiltonians with partial algebraization of their spectrum is larger than it was previously thought , and the new results call for new theoretical developments on this field .
the authors would like to thank prof .
pogosyan whose observations lead to some of the results of this paper .
the research of dgu is supported in part by the ramn y cajal program of the ministerio de ciencia y tecnologa and by the dgi under grant fis2005 - 00752 .
the research of nk and rm is supported in part by the nserc grants rgpin 105490 - 2004 and rgpin-228057 - 2004 , respectively .
a. bohm and y. ne eman , _ dynamical groups and spectrum generating algebras _ in : dynamical groups and spectrum generating algebras , a. bohm , y. ne eman and a.o .
barut , eds . ,
world scientific , singapore , 1988 .
pp . 3 - 68 .
a. turbiner , _ lie - algebras and linear operators with invariant subspaces _ in `` lie algebras , cohomology , and new applications to quantum mechanics '' , n. karmran and p. j. olver eds .
, 263310 , contemp .
, * 160 * , ams , 1994 .
|
we present evidence to suggest that the study of one dimensional quasi - exactly solvable ( qes ) models in quantum mechanics should be extended beyond the usual @xmath0 approach .
the motivation is twofold : we first show that certain quasi - exactly solvable potentials constructed with the @xmath0 lie algebraic method allow for a new larger portion of the spectrum to be obtained algebraically .
this is done via another algebraization in which the algebraic hamiltonian can not be expressed as a polynomial in the generators of @xmath0 .
we then show an example of a new quasi - exactly solvable potential which can not be obtained within the lie - algebraic approach .
|
ground - state ( @xmath4 , @xmath5 ) hydroxyl ( oh ) masers were first discovered towards several galactic hiiregions ( e.g. w3(oh ) ) by @xcite and @xcite . since their discovery , extensive work on ground - state oh masers
has shown that they are commonly associated with regions of high - mass star formation ( hmsf ) ( @xcite ; @xcite ) and these masers are predominantly strong in main - line transitions ( @xmath6 , @xmath7 ) ( i.e. 1665mhz and 1667mhz ; @xcite ) .
@xcite compiled a catalog of @xmath8375 ground - state oh maser sites associated with hmsf from the literature , representing an up to date compilation of this category of sources .
ground - state oh masers are also detected towards circumstellar envelopes of evolved giant and supergaint stars , such as miras and oh / ir stars ( @xcite ; @xcite , @xcite , @xcite ) , and are typically dominated by emission from the 1612 mhz satellite line ( i.e. @xmath9 , 1612mhz ; @xcite ) .
the maser emission in the 1720mhz satellite - line transition ( @xmath10 ) , is usually associated with supernova remnants ( snrs ) ( @xcite ) and traces the interaction between snrs and surrounding dense molecular clouds .
ground - state oh masers are also found associated with planetary nebulae ( pn ; @xcite ) , comets @xcite and the centres of active galaxies @xcite .
previous work on ground - state oh masers has targeted regions likely to have maser emission , such as infrared astronomical satellite ( _ iras _ ) point sources with infrared ( ir ) colors indicative of high - mass young stellar objects ( ysos ) @xcite , circumstellar envelopes of evolved stars with h@xmath11oand / or sio masers @xcite and snrs @xcite .
about 30 years ago , a single - dish blind survey of oh main - line transitions was carried out toward a thin strip of the southern galactic plane between galactic longitudes 233@xmath12through the galactic center to 60@xmath12and galactic latitudes within @xmath130.5@xmath12(@xcite ; @xcite ; @xcite ; @xcite ) .
@xcite also carried out a systematic survey close to the galactic plane with the australia telescope compact array ( atca ) in the main - line transitions .
however , these observations have favoured hmsf regions because only main - line transitions were observed . @xcite ( @xcite , @xcite and @xcite ) used the atca and very large array ( vla ) to survey the 1612mhz oh line towards a large portion of the galactic plane , between galactic longitudes of 315@xmath12and 45@xmath12and galactic latitudes of @xmath143@xmath12and @xmath153@xmath12 .
the survey of @xcite observed only the 1612mhz oh line and was thus heavily favoured towards detecting evolved star oh masers .
therefore , these previous surveys suffer from biases ( e.g. towards hmsf regions or evolved star sites ) and thus the full population of each type of astrophysical object mentioned above may not be comprehensively studied .
although it is clear that ground - state oh masers are detected towards different types of astrophysical objects ( hmsf regions , the envelopes of evolved stars , snrs ) , the proportion of masers associated with each kind of object remains unknown .
studying the proportion will yield a better understanding of the ground - state oh maser populations and characteristics of their associated astrophysical objects . in order to reduce biases caused by targeted surveys or surveys only observing specific ground - state oh transitions , the southern parkes large - area survey in hydroxyl ( splash )
was carried out with simultaneous observations of all four ground - state transitions .
splash surveyed 176 square degrees of the southern galactic plane and galactic centre ( @xcite ) , i.e. between galactic longitudes of 332@xmath12and 10@xmath12and galactic latitudes of @xmath142@xmath12and @xmath152@xmath12(152 square degrees ) , plus an extra region around the galactic centre , where coverage was extended up to galactic latitudes @xmath156@xmath12for galactic longitudes 358@xmath12to 4@xmath12(24 square degrees ) .
unsurprisingly , the survey has detected about 600 sites exhibiting oh maser emission .
the observations were conducted with the australia telescope national facility ( atnf ) parkes 64-m telescope and achieved a mean rms ( root - mean - square ) point - source sensitivity of @xmath865 mjy in each 0.18 kms@xmath16channel in maser - optimised cubes , but were limited by the spatial resolution ( about 13@xmath17 ) .
this spatial resolution is insufficient to reliably identify the astrophysical objects associated with the maser emission .
therefore , higher spatial resolution observations are needed , which is the motivation for this work .
the majority of oh masers detected in splash are in the 1612mhz satellite - line transition and show a double - horned spectral profile .
these masers are evolved star oh masers , which also occasionally exhibit detections in the 1665 and/or 1667mhz main - line transitions .
oh masers usually have many spectral features in the main - line transitions .
each spectral feature corresponds to one maser spot on the sky , which is usually considered to arise in a single , well - defined position @xcite .
star formation oh masers preferentially trace a later stage of the hmsf process , compared to methanol masers and water masers ( e.g. @xcite ; @xcite ; @xcite ) .
as mentioned above , oh masers in the 1720mhz transition are quite rare and are commonly associated with shocked gas , which can be detected in snrs ( @xcite ) , hmsf regions ( @xcite ; @xcite ) , occasionally asymptotic giant branch ( agb ) stars @xcite and pne @xcite .
this paper presents accurate positions of ground - state oh masers , which are obtained from atca observations , in the pilot region of splash between galactic longitudes of 334@xmath12and 344@xmath12and galactic latitudes of @xmath142@xmath12and @xmath152@xmath12 .
observations were conducted with the atca from 2013 october 24 to 2013 october 29 and 2015 january 27 , using the 6a configuration . at the frequency of the ground - state
oh masers , the 6a array results in a synthesised beam between 4.07@xmath18 and 5.50@xmath19 .
observational pointing centres were determined based on the parkes oh maser detections introduced in @xcite . for masers within one field of view , we chose the pointing centre between them ; if the masers were previously observed by the mapping the galactic magnetic field through oh masers ( magmo ; @xcite ) project , we obtained their accurate positions directly from the magmo project ( green , j. private communication ) .
we do not include the magmo data in the results presented here and only use them to study the detection statistics .
each region was typically observed with five 6 min snapshot observations for a total on - source integration time of 30 min .
primary flux calibration was performed with the standard flux density calibrator pks b1934@xmath14638 and bandpass calibration with pks b0823@xmath14500 .
phase calibrators were chosen to be within 7@xmath12of each target and the phase calibrators were monitored about every 20 min .
the compact array broadband backend was used to collect data , using the cfb 1m-0.5k mode with 16 zoom bands , each with 2048 channels over 1mhz giving 0.5 khz channel spacings , with all four polarisation products .
an analysis of the polarization properties of the masers , including the identification of zeeman pairs , will be presented in a forthcoming paper .
the setup included 3 concatenated zoom bands at each of the 1612 and 1720mhz maser transitions , 5 concatenated zoom bands covering the 1665 and 1667mhz transitions and 3 concatenated zoom bands at the transition .
the channel separation is 0.09 kms@xmath16 .
edge channels were masked out during the data reduction process . since
our observations were targeted towards single - dish maser detections , we know the velocity range of most of the maser emission expected in each source . the velocity coverage over which maser emission was searched for each transition was approximately @xmath14250 to + 210 kms@xmath16for 1612mhz , @xmath14210 to + 140 kms@xmath16for 1665/7mhz and @xmath14180 to + 250 kms@xmath16for 1720mhz in the local standard of rest ( lsr ) reference frame .
the mean rms sensitivity of the atca observations is @xmath865 mjy in each 0.09 kms@xmath16channel .
the sensitivity of the atca observations is approximately the same as the original splash survey , but with a factor of two higher velocity resolution .
the miriadtask * uvlin * was adopted to subtract the continuum emission .
the data were then reduced using miriadstandard data reduction routines to produce cleaned and restored data cubes , which were also corrected for the primary beam response ( @xcite ) .
the cubes were then searched ( over the area of the primary beam ) for maser emission . in order to manage the large volume of data and search masers within a reasonable amount of time , the following method ( similar to @xcite )
was used to search for masers : 1 .
the full data cube was binned in ten velocity channels ( after binning , the channel width becomes 0.9 kms@xmath16 ) to maximise the sensitivity to maser spots that have a typical oh maser line - width ( based on the parkes splash spectra ; @xcite ) .
a peak - intensity map ( in miriada moment @xmath142 map ) was formed from the whole binned cube .
the peak - intensity map was visually inspected for bright spots , which are the signatures of masers .
any maser candidate found in the peak - intensity maps was then verified in the full data cube .
our observations resulted in sparse sampling of the uv - plane , which , when combined with very bright maser spots , resulted in side - lobe artefacts in the peak intensity maps .
these can be distinguished from real maser spots by visually checking their appearance in the full data cube .
a real maser spot will typically be seen as the brightest feature in multiple contiguous channels .
5 . for channel ranges
excluding the maser emission , binned peak - intensity maps were created to check if there were other weak masers .
the above manual method of searching for and characterising features is at least as accurate as an automated method ( @xcite , @xcite ) . after identifying maser spots ,
their positions are determined in the following steps .
an integrated intensity map ( in miriada moment 0 map ) is created with the unbinned channels over which the maser spot is detected .
this integrated intensity map has the highest signal - to - noise ratio for that maser spot and therefore exhibits the most accurate position .
if two maser spots are overlapping in velocity , we identify their positions by only integrating over velocities where the maser spot in question dominates the flux density .
the miriadtask * imfit * is used to fit the integrated intensity map and obtain both the position and relative uncertainty in the position of the maser spot .
the peak velocity and peak flux density of a maser spot are determined from the spectrum , extracted at the position of the maser spot using miriadtask * uvspec*. the integrated flux density is also calculated using the spectrum of the maser spot ( see section [ images ] for more details ) .
we used a systematic approach to determine the velocity range for each maser spot : in order to improve the s / n of the spectra , we averaged the data by binning five channels to give a velocity resolution of about 0.45 kms@xmath16 .
we calculated the rms noise ( 1-@xmath20 ) of the binned data and identify the velocity range as those channels with flux density in excess of 3-@xmath20 .
in addition , we excluded any part of the spectrum with noise . where spectra appear to show two overlapping velocity components with a trough between them
, we classify each peak as a separate spot if the difference in flux density between the peak of the weaker line and the lowest point of the trough is greater than the 1-@xmath20 noise on the data .
this method includes most emission from maser spots , but at the same time does not include weak emission from channels such that the noise may dominate in that velocity range
. however , this method is not perfect and so the derived velocity ranges and integrated intensities are intended to be used as a guide .
this is particularly the case when two maser spots are blended in both velocity and position .
the absolute uncertainty in maser positions depends on the phase noise during the observations , which is related to the distance between the phase calibrator and the target region and the accuracy of the known locations of the phase calibrators , the locations of the antennas , as well as the atmospheric conditions .
a typical value of 1@xmath21 for the absolute positional uncertainty is adopted .
this number is based on previous oh maser surveys with the atca by @xcite , who found absolute positional uncertainties of oh masers were better than 1@xmath21 .
the relative positional uncertainty between maser spots detected in the same field of view can be more accurate than the absolute uncertainty .
the 1-@xmath20 relative positional uncertainty is approximately @xmath22 ( @xcite ) , where @xmath23 is the convolved size of the source ( i.e. , the beam size in the case of point sources , such as masers ) .
this allows us to compare the distributions of maser spots within a site on smaller scales than the absolute uncertainty , but does not allow us to compare the relative positions of maser spots with other data sets ( e.g. ir survey : glimpse , wise ; @xcite , @xcite , @xcite ; as described below ) to better than 1@xmath21 .
given that there are very few detections of 1720mhz masers , we have used that band to investigate the continuum emission associated with star formation oh masers .
we concatenated 3 zoom bands at this frequency which resulted in a bandwidth of 2mhz .
the data were reduced using standard techniques for atca continuum data , resulting in a typical rms noise of 5 mjy .
the results are detailed in section [ starformation ] .
in the splash pilot region , of the 196 pointing centres identified in the parkes observations of @xcite ( @xcite ) , 175 are detected with maser emission from our dataset and the magmo dataset .
we did not detect any maser emission toward 21 positions , which are discussed in more detail in section [ non - detection ] . from our dataset and the magmo dataset
, we detect 700 maser spots .
the strongest flux density is 114 jy and the weakest is 0.17 jy .
maser spots have been grouped into maser sites based on their separations .
the size of maser sites is discussed in section [ size ] .
we identify a total of 215 maser sites , 111 of which are newly detected . among these 215 maser sites ,
162 maser sites are from our observations and 53 maser sites are from the magmo data .
details of these 53 magmo maser sites will be published later by the magmo team .
we include these 53 magmo maser sites for statistical purposes only .
of these 215 maser sites , 154 show maser emission at 1612mhz , 61 have maser emission at 1665mhz , 57 show maser emission at 1667mhz and 9 show maser emission at 1720mhz .
the 162 maser sites , which are obtained from our observations , are detailed in table [ maintable ] . in column 1
, we assign a name for each maser spot , based on the galactic coordinates of their accurate positions and the frequency ( 1612 , 1665 , 1667 , 1720mhz ) plus a letter to identify the maser spots within the same maser site . for each frequency , letters are assigned sequentially , based on their peak velocities ( from low to high ) .
columns 2 and 3 show accurate positions ( r.a . and decl . ) for each maser spot , column 4 is the peak flux density and column 5 is the integrated flux density .
columns 6 , 7 and 8 are peak , minimum and maximum velocities , respectively . columns 9 , 10 and 11 show the uncertainties in position , i.e. uncertainties in minor axis , major axis and position angle .
column 12 shows the astrophysical identification of each maser site , which is discussed in section [ identification ] .
the final column indicates whether the maser site is a new detection .
there are 44 maser sites which only have one maser spot , while the remaining 171 sites have more than one . the richest maser site ( [email protected] )
has 26 maser spots in total at 1612 , 1665 and 1667mhz and the second richest maser site ( [email protected] from magmo ) exhibits 21 maser spots at 1665 and 1667mhz .
lccccrrrrrrll g333.942 + 0.387 - 1612a&16:21:01.143&-49:26:14.21&0.42&0.53&-72.9&-74.4&-71.7&0.38&0.59&-12.4&u&n + + g334.097 + 0.185 - 1612a&16:22:34.371&-49:28:12.98&0.78&0.87&-25.8&-26.7&-24.0&0.32&0.49&-12.4&es - sev&s97 + g334.097 + 0.185 - 1612b&16:22:34.365&-49:28:12.79&0.21&0.19&1.2&1.0&2.3&0.56&0.88&-12.4&es - sev&s97 + g334.097 + 0.185 - 1612c&16:22:34.350&-49:28:12.60&0.41&0.58&3.9&2.8&5.1&0.34&0.53&-12.4&es - sev&s97 + + g334.132 + 0.384 - 1720a&16:21:51.466&-49:18:17.11&0.36&0.31&-45.4&-45.7&-44.4&0.84&1.30&-12.4&u&n + + [email protected] - 1612a&16:26:41.198&-49:54:27.31&1.07&1.51&-59.5&-60.2&-57.0&0.18&0.28&-13.0&es - che&n + [email protected] - 1612b&16:26:41.229&-49:54:27.45&0.46&0.67&-21.0&-24.8&-20.2&0.38&0.58&-13.0&es - che&n + + g334.279 + 1.354 - 1612a&16:18:20.651&-48:30:43.48&1.12&1.61&-168.9&-171.2&-167.6&0.25&0.39&-9.0&es - li1&l91 + g334.279 + 1.354 - 1612b&16:18:20.635&-48:30:43.32&0.83&0.93&-148.1&-150.3&-147.1&0.32&0.50&-9.0&es - li1&l91 + + [email protected] - 1612a&16:26:20.937&-49:46:30.13&0.30&0.82&-61.4&-62.0&-55.7&0.44&0.68&-13.0&es - vis&n + [email protected] - 1612b&16:26:20.948&-49:46:29.22&0.24&0.29&-29.6&-31.1&-29.3&0.65&0.99&-13.0&es - vis&n + + [email protected] - 1612a&16:25:24.933&-49:31:09.11&1.29&3.92&-126.3&-127.7&-120.0&0.23&0.35&-9.0&es - hop&s97 + [email protected] - 1612b&16:25:24.939&-49:31:09.25&1.14&3.47&-87.2&-92.3&-84.2&0.27&0.42&-9.0&es - hop&s97 + + g334.458 + 0.877 - 1612a&16:21:08.406&-48:43:34.80&1.24&1.35&-129.2&-130.0&-127.8&0.24&0.38&-5.7&es - sev&s97 + g334.457 + 0.877 - 1612b&16:21:08.404&-48:43:34.91&1.37&1.35&-98.1&-99.6&-97.4&0.23&0.38&-5.7&es - sev&s97 + g334.458 + 0.877 - 1667a&16:21:08.383&-48:43:33.43&0.28&0.24&-97.4&-98.8&-97.1&0.77&1.22&-5.7&es - sev&s97 + + g334.498 + 0.380 - 1612a&16:23:27.200&-49:02:55.48&0.37&0.53&-6.3&-7.2&-4.5&0.45&0.72&-5.5&es - vis&n + g334.498 + 0.380 - 1612b&16:23:27.178&-49:02:55.38&0.29&0.23&27.1&26.4&27.7&0.66&1.46&-5.5&es - vis&n + + g334.577 + 1.958 - 1612a&16:17:05.722&-47:52:18.49&0.40&0.53&-90.4&-91.0&-88.8&0.49&0.77&-1.5&u&n + + g334.710 + 1.561 - 1612a&16:19:19.354&-48:03:44.17&2.53&4.17&-59.4&-60.3&-55.8&0.14&0.21&-1.6&es - sev&s97 + g334.710 + 1.561 - 1612b&16:19:19.357&-48:03:44.23&1.71&3.48&-27.1&-33.1&-26.3&0.20&0.31&-1.6&es - sev&s97 + g334.710 + 1.561 - 1667a&16:19:19.347&-48:03:44.15&0.34&0.21&-26.3&-27.0&-25.7&0.61&0.97&-1.6&es - sev&s97 + + g334.726 + 1.616 - 1612a&16:19:09.556&-48:00:42.92&3.27&6.76&35.2&34.1&41.4&0.08&0.12&-1.6&es - sev&s97 + g334.726 + 1.616 - 1612b&16:19:09.553&-48:00:43.09&0.94&1.45&58.2&56.8&60.0&0.27&0.42&-1.6&es - sev&s97 + g334.726 + 1.616 - 1667a&16:19:09.591&-48:00:42.39&0.25&0.42&59.0&57.3&60.3&0.48&0.75&-1.6&es - sev&s97 + + [email protected] - 1612a&16:33:01.008&-50:10:55.93&0.29&0.39&-7.2&-8.4&-5.2&0.45&0.72&2.0&es - li1&l91 + [email protected] - 1612b&16:33:00.986&-50:10:55.78&0.30&0.46&19.5&17.5&20.7&0.54&0.85&2.0&es - li1&l91 + [email protected] - 1612c&16:33:00.999&-50:10:55.87&0.88&1.13&22.6&21.2&23.4&0.28&0.44&2.0&es - li1&l91 + [email protected] - 1612d&16:33:00.994&-50:10:55.71&1.49&2.30&24.6&23.9&26.6&0.19&0.29&2.0&es - li1&l91 + + [email protected] - 1612a&16:35:14.930&-50:24:16.12&0.40&0.55&-70.0&-70.7&-68.0&0.44&0.82&-2.6&es - sev&s97 + [email protected] - 1612b&16:35:14.977&-50:24:16.59&0.35&0.34&-33.6&-34.9&-33.1&0.40&0.62&-2.6&es - sev&s97 + + g334.898 + 0.520 - 1612a&16:24:33.445&-48:39:54.91&3.16&3.84&-72.1&-73.0&-70.7&0.17&0.27&1.9&es - sev&s97 + g334.898 + 0.520 - 1612b&16:24:33.442&-48:39:54.89&1.62&2.40&-69.8&-70.2&-66.2&0.30&0.47&1.9&es - sev&s97 + g334.898 + 0.520 - 1612c&16:24:33.443&-48:39:54.83&3.73&5.92&-43.8&-49.8&-42.6&0.16&0.25&1.9&es - sev&s97 + + [email protected] - 1612a&16:30:48.759&-49:34:03.00&0.25&0.16&-145.4&-146.2&-145.3&0.90&1.30&-25.2&u&n + + g335.000 + 0.687 - 1612a&16:24:16.053&-48:28:29.39&0.46&0.44&-125.1&-125.6&-123.8&0.43&0.69&1.9&es - vis&n + g335.001 + 0.687 - 1612b&16:24:16.015&-48:28:28.42&0.34&0.32&-100.6&-102.0&-100.2&0.45&0.71&1.9&es - vis&n + + g335.136 + 0.196 - 1612a&16:26:57.829&-48:43:14.92&0.40&0.30&18.8&18.1&19.4&0.38&0.58&0.3&u&n + g335.136 + 0.196 - 1612b&16:26:57.764&-48:43:14.91&0.26&0.35&21.5&19.9&22.6&0.45&0.70&0.3&u&n + + g335.238 + 1.283 - 1612a&16:22:43.581&-47:53:13.76&0.32&0.28&31.9&30.8&32.1&0.44&0.71&7.7&es - vis&n + g335.238 + 1.283 - 1612b&16:22:43.525&-47:53:13.35&0.22&0.21&48.2&47.6&48.9&0.81&1.31&7.7&es - vis&n + + g335.476 + 1.826 - 1612a&16:21:26.738&-47:20:04.58&0.40&0.32&-114.9&-115.2&-113.4&0.38&0.61&8.9&es - vis&n + g335.476 + 1.826 - 1612b&16:21:26.718&-47:20:05.27&0.23&0.22&-86.1&-87.1&-85.3&0.60&0.99&8.9&es - vis&n + + g335.715 + 0.991 - 1612a&16:25:57.606&-47:45:05.98&5.20&8.23&-41.9&-43.3&-37.4&0.13&0.22&-3.3&es - sev&s97 + g335.715 + 0.991 - 1612b&16:25:57.606&-47:45:05.98&11.22&19.63&-7.9&-13.8&-6.1&0.10&0.16&-3.3&es - sev&s97 + g335.715 + 0.991 - 1667a&16:25:57.646&-47:45:06.74&0.31&0.32&-43.2&-43.9&-42.2&0.48&0.80&-3.3&es - sev&s97 + g335.715 + 0.991 - 1667b&16:25:57.617&-47:45:06.39&0.34&0.63&-8.6&-10.6&-7.5&0.36&0.62&-3.3&es - sev&s97 + g335.715 + 0.991 - 1667c&16:25:57.617&-47:45:06.18&0.37&0.38&-6.4&-7.1&-5.3&0.56&0.95&-3.3&es - sev&s97 + + g335.751 + 1.454 - 1612a&16:24:08.966&-47:24:06.65&0.41&0.94&-71.4&-72.3&-68.7&0.55&0.84&-12.6&es - vis&n + g335.751 + 1.454 - 1612b&16:24:08.998&-47:24:07.14&1.09&2.26&-39.2&-45.5&-38.3&0.22&0.34&-12.6&es - vis&n + + g335.789 + 0.174 - 1665a&16:29:47.309&-48:15:51.77&1.02&0.57&-54.0&-54.3&-53.4&0.32&0.47&-13.6&sf - mmb&c98 + g335.789 + 0.174 - 1665b&16:29:47.339&-48:15:52.19&0.82&1.01&-51.3&-52.9&-50.3&0.42&0.63&-13.6&sf - mmb&c98 + g335.789 + 0.174 - 1667a&16:29:47.337&-48:15:51.97&5.07&3.13&-52.3&-53.1&-51.4&0.12&0.18&-16.9&sf - mmb&c98 + + g335.832 + 1.434 - 1612a&16:24:33.962&-47:21:30.20&5.18&7.35&-97.1&-99.1&-95.0&0.09&0.14&-12.6&es - rms&s97 + g335.832 + 1.434 - 1612b&16:24:33.943&-47:21:30.48&0.52&1.06&-94.2&-94.5&-90.9&0.37&0.59&-12.6&es - rms&s97 + g335.832 + 1.434 - 1612c&16:24:33.951&-47:21:30.37&2.55&3.16&-69.7&-78.2&-69.6&0.19&0.29&-12.6&es - rms&s97 + g335.832 + 1.434 - 1612d&16:24:33.957&-47:21:30.42&2.75&3.40&-68.7&-69.1&-67.3&0.15&0.23&-12.6&es - rms&s97 + g335.832 + 1.434 - 1612e&16:24:33.950&-47:21:30.31&0.43&1.62&-64.6&-66.9&-61.0&0.39&0.60&-12.6&es - rms&s97 + g335.832 + 1.434 - 1665a&16:24:33.955&-47:21:30.22&0.36&1.84&-87.6&-92.0&-82.3&0.35&0.53&-12.6&es - rms&s97 + g335.832 + 1.434 - 1665b&16:24:33.937&-47:21:30.61&0.45&2.35&-79.1&-81.9&-70.0&0.33&0.51&-12.6&es - rms&s97 + + [email protected] - 1612a&16:35:10.812&-48:48:50.24&0.35&0.38&-159.1&-160.0&-158.2&0.40&0.61&-9.8&es - vis&n + [email protected] - 1612b&16:35:10.825&-48:48:49.93&0.46&0.80&-134.7&-137.3&-133.2&0.38&0.58&-9.8&es - vis&n + + g336.064 + 0.003 - 1612a&16:31:40.078&-48:10:53.47&0.54&1.34&-85.4&-86.9&-82.4&0.32&0.49&-10.6&es - vis&s97 + g336.064 + 0.003 - 1612b&16:31:40.081&-48:10:53.42&0.98&1.74&-50.9&-53.3&-49.3&0.24&0.37&-10.6&es - vis&s97 + + [email protected] - 1612a&16:36:32.522&-48:54:37.95&2.13&1.37&-90.6&-91.4&-90.1&0.17&0.26&-9.8&es - rms&s97 + [email protected] - 1612b&16:36:32.529&-48:54:37.95&2.45&3.94&-88.9&-89.6&-86.9&0.13&0.20&-9.8&es - rms&s97 + [email protected] - 1612c&16:36:32.492&-48:54:37.84&2.27&4.27&-85.1&-86.4&-83.3&0.13&0.20&-9.8&es - rms&s97 + [email protected] - 1612d&16:36:32.496&-48:54:37.79&1.51&2.28&-81.6&-82.8&-80.1&0.19&0.29&-9.8&es - rms&s97 + [email protected] - 1612e&16:36:32.508&-48:54:37.63&0.93&2.54&-78.0&-79.6&-74.2&0.26&0.40&-9.8&es - rms&s97 + [email protected] - 1612f&16:36:32.531&-48:54:37.91&3.49&4.12&-59.6&-62.8&-58.8&0.10&0.15&-9.8&es - rms&s97 + [email protected] - 1612g&16:36:32.532&-48:54:37.96&3.23&2.51&-57.5&-58.3&-57.4&0.08&0.12&-9.8&es - rms&s97 + [email protected] - 1612h&16:36:32.526&-48:54:37.97&4.61&3.53&-56.1&-56.9&-56.0&0.06&0.08&-9.8&es - rms&s97 + [email protected] - 1612i&16:36:32.529&-48:54:37.96&13.12&20.16&-53.9&-55.6&-53.3&0.03&0.05&-9.8&es - rms&s97 + [email protected] - 1612j&16:36:32.529&-48:54:37.97&16.94&25.77&-52.5&-52.8&-47.9&0.04&0.07&-9.8&es - rms&s97 + [email protected] - 1665a&16:36:32.524&-48:54:38.04&3.89&1.70&-90.8&-91.2&-90.3&0.11&0.17&-9.8&es - rms&s97 + [email protected] - 1665b&16:36:32.523&-48:54:37.97&4.01&2.36&-88.8&-89.9&-88.1&0.13&0.19&-9.8&es - rms&s97 + [email protected] - 1665c&16:36:32.522&-48:54:38.03&1.47&0.66&-87.6&-87.7&-86.8&0.20&0.32&-9.8&es - rms&s97 + [email protected] - 1665d&16:36:32.551&-48:54:38.04&0.44&0.83&-84.8&-86.4&-82.9&0.34&0.52&-9.8&es - rms&s97 + [email protected] - 1665e&16:36:32.507&-48:54:37.40&0.23&0.11&-69.5&-69.7&-68.8&0.63&0.99&-9.8&es - rms&s97 + [email protected] - 1665f&16:36:32.515&-48:54:37.58&0.22&0.32&-66.5&-67.9&-65.7&0.44&0.67&-9.8&es - rms&s97 + [email protected] - 1665g&16:36:32.549&-48:54:38.33&0.29&0.24&-64.6&-64.8&-63.5&0.54&0.81&-9.8&es - rms&s97 + [email protected] - 1665h&16:36:32.573&-48:54:38.50&0.23&0.17&-59.9&-60.4&-59.1&0.64&1.02&-9.8&es - rms&s97 + [email protected] - 1665i&16:36:32.562&-48:54:38.60&0.21&0.22&-57.9&-58.7&-56.9&0.71&1.12&-9.8&es - rms&s97 + [email protected] - 1665j&16:36:32.515&-48:54:38.11&0.82&1.31&-52.8&-54.7&-51.7&0.31&0.47&-9.8&es - rms&s97 + [email protected] - 1665k&16:36:32.539&-48:54:37.86&0.62&0.60&-50.1&-51.2&-49.9&0.27&0.71&-9.8&es - rms&s97 + [email protected] - 1665l&16:36:32.520&-48:54:37.91&1.18&0.81&-49.0&-49.5&-48.2&0.21&0.31&-9.8&es - rms&s97 + [email protected] - 1667a&16:36:32.526&-48:54:37.97&3.87&2.19&-90.6&-91.4&-89.6&0.10&0.16&-9.8&es - rms&s97 + [email protected] - 1667b&16:36:32.538&-48:54:37.31&0.26&0.23&-88.9&-89.2&-87.4&0.59&0.90&-9.8&es - rms&s97 + [email protected] - 1667c&16:36:32.450&-48:54:38.22&0.25&0.13&-85.1&-85.7&-84.8&0.91&1.36&-9.8&es - rms&s97 + [email protected] - 1667d&16:36:32.582&-48:54:38.28&0.29&0.37&-58.4&-60.2&-58.0&0.47&0.71&-9.8&es - rms&s97 + + [email protected] - 1612a&16:35:42.641&-48:45:19.31&0.28&0.61&-66.0&-66.9&-62.9&0.43&0.66&-9.8&es - vis&n + [email protected] - 1612b&16:35:42.638&-48:45:19.34&0.38&0.94&-26.1&-29.2&-25.2&0.33&0.52&-9.8&es - vis&n + + g336.277 + 0.614 - 1612a&16:29:53.196&-47:36:26.06&0.58&0.58&-110.4&-111.2&-109.0&0.40&0.63&-6.2&es - vis&n + g336.277 + 0.615 - 1612b&16:29:53.187&-47:36:25.82&0.29&0.31&-78.1&-79.0&-77.2&0.61&0.96&-6.2&es - vis&n + + g336.313 + 1.378 - 1612a&16:26:46.634&-47:03:12.12&1.93&2.74&-137.7&-138.6&-132.7&0.19&0.29&-6.0&es - sev&s97 + g336.313 + 1.378 - 1612b&16:26:46.629&-47:03:12.20&1.92&2.98&-105.7&-110.9&-104.6&0.19&0.29&-6.0&es - sev&s97 + + g336.573 + 0.701 - 1612a&16:30:42.863&-47:19:58.17&0.31&0.30&-67.2&-68.0&-66.2&0.36&0.56&-2.5&es - vis&n + g336.572 + 0.701 - 1612b&16:30:42.856&-47:19:58.59&0.20&0.19&-37.9&-38.9&-37.6&0.59&0.96&-2.5&es - vis&n + + g336.615 + 1.307 - 1612a&16:28:18.369&-46:53:07.78&0.23&0.14&-74.9&-75.6&-74.7&0.65&1.04&-6.4&es - vis&n + g336.615 + 1.307 - 1612b&16:28:18.386&-46:53:07.54&0.30&0.27&-45.2&-46.0&-44.7&0.40&0.64&-6.4&es - vis&n + + [email protected] - 1612a&16:37:06.581&-48:13:42.54&0.39&0.47&-45.8&-46.3&-44.1&0.38&0.58&-9.6&pn - usc&n + [email protected] - 1667a&16:37:06.567&-48:13:42.32&0.37&0.45&-45.8&-46.4&-44.7&0.39&0.63&-4.6&pn - usc&n + [email protected] - 1720a&16:37:06.588&-48:13:42.39&0.95&0.89&-61.0&-61.6&-59.9&0.29&0.47&-4.6&pn - usc&n + [email protected] - 1720b&16:37:06.628&-48:13:42.43&0.34&0.28&-55.8&-56.1&-54.8&0.59&0.92&-4.6&pn - usc&n + [email protected] - 1720c&16:37:06.535&-48:13:42.35&0.30&0.29&-49.0&-50.1&-48.4&0.48&0.78&-4.6&pn - usc&n + [email protected] - 1720d&16:37:06.634&-48:13:42.46&0.43&0.71&-24.5&-26.3&-23.3&0.34&0.55&-4.6&pn - usc&n + [email protected] - 1720e&16:37:06.636&-48:13:41.44&0.29&0.23&-22.6&-22.9&-21.6&0.41&0.64&-4.6&pn - usc&n + + g336.701 + 0.924 - 1612a&16:30:16.805&-47:05:10.50&0.54&0.52&-56.4&-57.1&-55.8&0.42&0.63&-2.5&es - vis&n + g336.701 + 0.925 - 1612b&16:30:16.783&-47:05:09.56&0.29&0.30&-37.2&-38.5&-36.7&0.47&0.74&-2.5&es - vis&n + + g336.739 + 0.522 - 1612a&16:32:09.599&-47:20:04.21&0.79&1.21&-58.7&-59.6&-56.4&0.30&0.48&2.4&es - vis&n + g336.739 + 0.522 - 1612b&16:32:09.598&-47:20:04.44&0.40&0.64&-18.7&-21.0&-17.8&0.35&0.56&2.4&es - vis&n + + g336.817 + 0.300 - 1612a&16:33:26.010&-47:25:42.63&1.36&4.60&35.8&34.0&42.1&0.28&0.44&-2.7&es - sev&s97 + g336.817 + 0.300 - 1612b&16:33:26.010&-47:25:42.59&1.02&3.26&59.6&54.8&62.5&0.32&0.50&-2.7&es - sev&s97 + + [email protected] - 1612a&16:39:36.472&-48:17:02.98&0.73&1.56&-131.2&-131.9&-126.9&0.38&0.58&-1.7&es - vis&n + [email protected] - 1612b&16:39:36.475&-48:17:03.07&1.74&2.07&-101.4&-105.1&-100.6&0.19&0.30&-1.7&es - vis&n + + [email protected] - 1612a&16:38:47.950&-48:00:07.49&0.35&0.10&-122.1&-122.2&-121.8&0.51&0.79&1.3&u&n + [email protected] - 1665a&16:38:47.917&-48:00:09.39&0.32&0.37&-123.0&-123.6&-121.9&0.41&0.67&1.3&u&n + + [email protected] - 1612a&16:43:36.938&-48:40:11.67&0.61&1.17&-180.3&-181.2&-177.6&0.28&0.43&0.6&es - rms&s97 + [email protected] - 1612b&16:43:36.932&-48:40:11.40&0.77&1.69&-148.3&-151.7&-147.1&0.29&0.45&0.6&es - rms&s97 + + [email protected] - 1612a&16:40:55.824&-48:13:59.82&1.94&13.67&-90.5&-96.5&-80.1&0.19&0.29&-1.7&es - bai&s97 + + g337.201 + 0.114 - 1665a&16:35:46.541&-47:16:16.91&2.08&1.53&-64.9&-66.2&-64.4&0.22&0.36&1.3&sf - mmb&n + g337.201 + 0.114 - 1665b&16:35:46.493&-47:16:17.31&0.34&0.19&-63.4&-64.0&-63.1&0.62&0.94&1.3&sf - mmb&n + g337.201 + 0.114 - 1667a&16:35:46.534&-47:16:16.22&0.44&0.21&-60.7&-61.1&-60.2&0.33&0.50&1.3&sf - mmb&n + + g337.231 + 0.044 - 1612a&16:36:11.815&-47:17:47.91&0.67&1.81&-26.3&-28.8&-22.5&0.32&0.51&1.3&es - vis&n + g337.230 + 0.044 - 1612b&16:36:11.796&-47:17:48.93&0.27&0.33&23.3&21.1&23.8&0.78&0.93&1.3&es - vis&n + + g337.241 + 0.146 - 1612a&16:35:47.671&-47:13:14.21&0.30&0.36&-25.9&-27.0&-25.2&0.87&1.31&1.3&es - vis&n + g337.241 + 0.146 - 1612b&16:35:47.719&-47:13:14.04&0.29&0.15&1.7&1.2&2.1&0.49&0.82&1.3&es - vis&n + + [email protected] - 1665a&16:36:56.314&-47:22:27.46&0.65&1.01&-69.3&-71.0&-67.9&0.36&0.56&1.3&sf - mmb&c98 + [email protected] - 1665b&16:36:56.294&-47:22:27.45&0.35&0.21&-65.9&-66.2&-65.3&0.54&0.88&1.3&sf - mmb&c98 + [email protected] - 1665c&16:36:56.309&-47:22:27.32&0.45&0.67&-62.0&-64.4&-61.3&0.49&0.76&1.3&sf - mmb&c98 + [email protected] - 1667a&16:36:56.339&-47:22:26.90&0.33&0.24&-70.0&-70.7&-69.4&0.69&1.07&1.3&sf - mmb&c98 + + [email protected] - 1612a&16:37:38.987&-47:23:48.89&0.69&1.33&-151.8&-153.3&-149.7&0.33&0.50&-13.5&u&s97 + + [email protected] - 1612a&16:37:29.216&-47:19:31.72&2.08&3.38&-43.7&-44.8&-42.1&0.35&0.53&-13.7&es - sev&s97 + [email protected] - 1612b&16:37:29.201&-47:19:31.58&1.38&4.55&-40.8&-41.6&-34.4&0.44&0.68&-13.7&es - sev&s97 + [email protected] - 1612c&16:37:29.201&-47:19:31.55&2.29&11.90&-12.6&-28.0&-10.8&0.33&0.49&-13.7&es - sev&s97 + + g337.423 + 0.762 - 1612a&16:33:51.426&-46:40:12.75&0.37&0.58&-71.1&-72.1&-69.0&0.32&0.54&-5.5&es - vis&n + g337.423 + 0.762 - 1612b&16:33:51.427&-46:40:12.68&0.77&0.69&-46.7&-48.1&-45.8&0.37&0.62&-5.5&es - vis&n + + g337.539 + 0.130 - 1612a&16:37:02.571&-47:00:39.53&8.44&26.26&-143.8&-147.4&-139.7&0.13&0.20&-13.5&es - sev&s97 + g337.539 + 0.130 - 1612b&16:37:02.575&-47:00:39.51&1.08&0.78&-138.9&-139.3&-138.4&0.34&0.52&-13.5&es - sev&s97 + g337.539 + 0.130 - 1612c&16:37:02.577&-47:00:39.54&1.10&1.82&-136.9&-137.9&-135.2&0.26&0.40&-13.5&es - sev&s97 + g337.539 + 0.130 - 1612d&16:37:02.593&-47:00:39.79&0.74&1.26&-133.9&-134.7&-130.6&0.29&0.43&-13.5&es - sev&s97 + g337.539 + 0.130 - 1612e&16:37:02.575&-47:00:39.54&1.77&6.02&-97.0&-105.7&-96.1&0.22&0.33&-13.5&es - sev&s97 + g337.539 + 0.130 - 1612f&16:37:02.571&-47:00:39.59&3.41&5.23&-94.2&-95.7&-93.4&0.17&0.25&-13.5&es - sev&s97 + g337.539 + 0.130 - 1612g&16:37:02.577&-47:00:39.64&3.47&5.68&-92.4&-92.9&-89.8&0.16&0.25&-13.5&es - sev&s97 + g337.539 + 0.130 - 1665a&16:37:02.594&-47:00:40.19&0.28&0.28&-144.8&-145.0&-143.2&0.51&0.78&-13.7&es - sev&s97 + g337.539 + 0.130 - 1667a&16:37:02.586&-47:00:39.20&0.39&1.19&-93.0&-98.5&-90.6&0.41&0.62&-13.9&es - sev&s97 + + [email protected] - 1665a&16:38:09.478&-47:05:00.46&0.47&0.50&-58.7&-59.7&-57.9&0.47&0.71&-13.5&sf - mmb&c98 + [email protected] - 1665b&16:38:09.522&-47:05:00.04&1.04&0.88&-42.7&-43.4&-42.1&0.33&0.50&-13.5&sf - mmb&c98 + [email protected] - 1665c&16:38:09.506&-47:04:59.87&0.98&0.61&-41.5&-41.7&-40.4&0.29&0.44&-13.5&sf - mmb&c98 + [email protected] - 1720a&16:38:09.540&-47:05:00.40&0.61&0.60&-39.9&-41.4&-39.3&0.40&0.62&-13.5&sf - mmb&c98 + + [email protected] - 1665a&16:38:29.657&-47:00:35.36&14.08&15.35&-50.7&-55.4&-50.6&0.42&0.64&-11.1&sf - mmb&c98 + [email protected] - 1665b&16:38:29.675&-47:00:35.53&21.31&29.41&-49.2&-50.1&-46.2&0.37&0.59&-11.1&sf - mmb&c98 + [email protected] - 1667a&16:38:29.671&-47:00:35.99&0.55&0.91&-60.7&-61.3&-58.2&0.45&0.69&-10.6&sf - mmb&c98 + [email protected] - 1667b&16:38:29.636&-47:00:36.27&0.38&0.42&-56.9&-57.3&-55.6&0.61&0.93&-10.6&sf - mmb&c98 + [email protected] - 1667c&16:38:29.649&-47:00:35.52&0.80&0.67&-53.3&-54.7&-52.9&0.52&0.80&-10.6&sf - mmb&c98 + [email protected] - 1667d&16:38:29.669&-47:00:35.27&5.92&13.11&-49.6&-52.5&-46.8&0.27&0.41&-10.6&sf - mmb&c98 + + [email protected] - 1612a&16:45:39.532&-48:03:39.18&0.35&0.50&-98.5&-99.1&-96.4&0.55&0.81&-14.9&es - vis&n + [email protected] - 1612b&16:45:39.523&-48:03:38.69&0.52&0.64&-63.1&-64.6&-62.3&0.28&0.42&-14.9&es - vis&n + + g337.719 + 0.350 - 1612a&16:36:47.649&-46:43:46.72&0.61&0.37&-56.3&-57.5&-56.1&0.38&0.57&-14.7&es - rms&s97 + g337.719 + 0.350 - 1612b&16:36:47.644&-46:43:46.98&2.20&4.77&-54.0&-55.6&-47.5&0.19&0.29&-14.7&es - rms&s97 + g337.719 + 0.350 - 1612c&16:36:47.648&-46:43:46.93&1.69&5.79&-24.1&-31.6&-21.6&0.21&0.31&-14.7&es - rms&s97 + g337.719 + 0.350 - 1667a&16:36:47.659&-46:43:47.05&0.73&0.97&-21.8&-23.8&-20.7&0.31&0.47&-14.3&es - rms&s97 + + g337.782 + 1.394 - 1612a&16:32:35.294&-45:58:37.84&0.53&0.69&-33.9&-35.7&-32.5&0.35&0.54&-10.4&u&n + + g337.800 + 0.737 - 1612a&16:35:26.810&-46:24:32.25&0.55&0.62&-139.4&-139.9&-137.6&0.36&0.57&-9.4&es - vis&n + g337.800 + 0.738 - 1612b&16:35:26.790&-46:24:31.98&0.26&0.38&-109.6&-112.7&-109.0&0.66&1.02&-9.4&es - vis&n + + [email protected] - 1720a&16:38:52.189&-46:56:15.96&1.99&3.38&-44.9&-46.6&-43.2&0.20&0.30&-11.1&sn - cas&c04 + + [email protected] - 1612a&16:39:47.870&-47:04:06.46&0.49&0.57&-122.9&-123.5&-121.7&0.68&1.04&-11.1&es - li2&l89 + [email protected] - 1612b&16:39:47.894&-47:04:06.39&0.68&2.47&-118.7&-121.2&-114.9&0.51&0.77&-11.1&es - li2&l89 + [email protected] - 1612c&16:39:47.872&-47:04:06.60&0.97&1.34&-96.8&-99.4&-96.7&0.43&0.67&-11.1&es - li2&l89 + [email protected] - 1612d&16:39:47.875&-47:04:06.55&1.35&2.03&-95.1&-96.2&-94.0&0.31&0.47&-11.1&es - li2&l89 + + g337.860 + 0.271 - 1612a&16:37:41.509&-46:40:40.17&2.92&2.69&-83.4&-85.1&-82.9&0.16&0.24&-14.7&es - hop&s97 + g337.861 + 0.271 - 1612b&16:37:41.535&-46:40:40.24&2.34&1.76&-82.2&-82.4&-81.5&0.14&0.21&-14.7&es - hop&s97 + g337.861 + 0.271 - 1612c&16:37:41.542&-46:40:40.17&2.51&1.83&-80.7&-81.1&-80.2&0.15&0.22&-14.7&es - hop&s97 + g337.861 + 0.271 - 1612d&16:37:41.556&-46:40:40.00&3.07&3.09&-78.9&-79.7&-77.0&0.16&0.25&-14.7&es - hop&s97 + g337.860 + 0.271 - 1612e&16:37:41.514&-46:40:40.60&1.12&1.04&-41.3&-42.0&-40.2&0.29&0.43&-14.7&es - hop&s97 + g337.860 + 0.271 - 1612f&16:37:41.510&-46:40:41.10&0.45&0.54&-39.2&-39.7&-37.5&0.34&0.53&-14.7&es - hop&s97 + g337.860 + 0.271 - 1665a&16:37:41.520&-46:40:40.91&0.61&0.39&-57.9&-58.3&-57.0&0.34&0.50&-14.7&es - hop&s97 + g337.860 + 0.271 - 1667a&16:37:41.508&-46:40:40.69&1.35&2.79&-80.6&-83.1&-76.9&0.22&0.32&-14.3&es - hop&s97 + g337.860 + 0.271 - 1667b&16:37:41.513&-46:40:40.82&0.79&1.57&-75.0&-76.5&-73.0&0.28&0.43&-14.3&es - hop&s97 + g337.860 + 0.271 - 1667c&16:37:41.529&-46:40:40.76&0.79&1.20&-71.0&-72.5&-69.0&0.32&0.48&-14.3&es - hop&s97 + g337.860 + 0.271 - 1667d&16:37:41.506&-46:40:41.49&0.56&0.56&-67.7&-68.6&-66.4&0.51&0.78&-14.3&es - hop&s97 + g337.860 + 0.271 - 1667e&16:37:41.519&-46:40:40.86&1.19&1.26&-59.2&-61.1&-58.9&0.28&0.43&-14.3&es - hop&s97 + g337.860 + 0.271 - 1667f&16:37:41.519&-46:40:40.82&0.87&0.81&-58.2&-58.5&-57.2&0.29&0.44&-14.3&es - hop&s97 + g337.860 + 0.271 - 1667g&16:37:41.547&-46:40:41.11&0.56&0.35&-56.4&-56.7&-55.8&0.37&0.56&-14.3&es - hop&s97 + g337.860 + 0.271 - 1667h&16:37:41.516&-46:40:40.95&1.66&2.18&-54.0&-55.4&-51.9&0.20&0.31&-14.3&es - hop&s97 + + g337.877 + 0.820 - 1612a&16:35:23.558&-46:17:46.52&0.29&0.40&-115.3&-116.7&-114.0&0.50&0.79&-9.4&u&n + + [email protected] - 1612a&16:47:34.787&-48:01:20.45&0.51&0.61&-80.8&-81.4&-79.2&0.29&0.44&-11.5&es - vis&n + [email protected] - 1612b&16:47:34.793&-48:01:19.19&0.27&0.38&-51.0&-54.6&-50.6&0.36&0.56&-11.5&es - vis&n + + g337.989 + 1.134 - 1612a&16:34:29.831&-46:00:04.71&0.60&0.79&-115.1&-116.6&-113.9&0.32&0.51&-6.1&es - rms&s97 + g337.989 + 1.134 - 1612b&16:34:29.834&-46:00:04.61&0.29&0.34&-112.3&-113.4&-111.6&0.54&0.90&-6.1&es - rms&s97 + g337.989 + 1.134 - 1612c&16:34:29.822&-46:00:04.46&0.47&0.58&-84.0&-87.1&-83.9&0.31&0.50&-6.1&es - rms&s97 + g337.989 + 1.134 - 1612d&16:34:29.815&-46:00:04.38&0.40&0.35&-83.1&-83.4&-82.1&0.53&0.84&-6.1&es - rms&s97 + + [email protected] - 1612a&16:39:55.769&-46:50:29.96&7.21&17.27&-88.6&-91.7&-87.2&0.39&0.60&-11.1&es - hop&s97 + [email protected] - 1612b&16:39:55.764&-46:50:29.99&3.73&5.40&-86.6&-86.7&-84.9&0.36&0.56&-11.1&es - hop&s97 + [email protected] - 1612c&16:39:55.760&-46:50:29.87&1.93&5.71&-83.9&-84.4&-78.5&0.28&0.42&-11.1&es - hop&s97 + [email protected] - 1612d&16:39:55.766&-46:50:29.93&2.63&7.10&-49.6&-60.3&-49.0&0.31&0.49&-11.1&es - hop&s97 + [email protected] - 1612e&16:39:55.764&-46:50:30.02&2.71&5.84&-46.4&-48.5&-44.9&0.31&0.48&-11.1&es - hop&s97 + + [email protected] - 1667a&16:40:04.006&-46:51:18.60&0.53&2.25&-66.9&-72.7&-64.8&0.40&0.61&-10.6&sf - rms&n + [email protected] - 1667b&16:40:03.998&-46:51:18.06&0.67&3.38&-59.9&-64.3&-56.9&0.29&0.44&-10.6&sf - rms&n + [email protected] - 1667c&16:40:04.009&-46:51:18.11&0.78&4.80&-48.5&-56.4&-43.3&0.36&0.55&-10.6&sf - rms&n + [email protected] - 1667d&16:40:04.024&-46:51:17.85&0.35&0.83&-35.4&-38.4&-34.9&0.44&0.69&-10.6&sf - rms&n + [email protected] - 1667e&16:40:04.032&-46:51:18.92&0.40&1.61&-28.8&-34.5&-26.2&0.55&0.84&-10.6&sf - rms&n + + [email protected] - 1612a&16:47:39.049&-47:55:35.42&0.65&0.60&-98.8&-99.1&-97.3&0.39&0.60&-11.5&es - rms&n + [email protected] - 1612b&16:47:39.052&-47:55:35.28&0.43&0.56&-69.1&-70.1&-67.8&0.38&0.60&-11.5&es - rms&n + + g338.232 + 1.204 - 1612a&16:35:08.506&-45:46:30.17&0.27&0.29&-150.2&-151.1&-149.3&0.47&0.75&-6.1&es - vis&n + g338.232 + 1.205 - 1612b&16:35:08.460&-45:46:29.69&0.31&0.27&-131.1&-131.6&-130.2&0.58&0.96&-6.1&es - vis&n + + g338.513 + 1.503 - 1612a&16:34:58.165&-45:21:58.09&0.31&0.34&-72.8&-73.8&-71.6&0.45&0.75&-5.3&u&n + + g338.514 + 1.458 - 1612a&16:35:09.747&-45:23:44.29&0.25&0.15&14.2&13.4&14.7&0.83&1.41&-5.3&es - rms&n + + g338.519 + 0.072 - 1612a&16:41:06.159&-46:19:06.32&1.73&3.24&-144.5&-146.1&-143.9&0.33&0.53&-0.2&es - sev&s97 + g338.519 + 0.072 - 1612b&16:41:06.158&-46:19:06.37&1.45&1.39&-143.0&-143.4&-142.1&0.32&0.52&-0.2&es - sev&s97 + g338.519 + 0.072 - 1612c&16:41:06.177&-46:19:06.83&0.57&0.75&-141.4&-141.6&-138.9&0.41&0.61&-0.2&es - sev&s97 + g338.519 + 0.072 - 1612d&16:41:06.168&-46:19:06.34&3.33&5.54&-114.6&-122.1&-113.9&0.33&0.51&-0.2&es - sev&s97 + + g338.567 + 0.110 - 1665a&16:41:06.948&-46:15:27.90&0.28&0.10&-76.6&-76.7&-76.2&0.66&1.08&-0.2&sf - mmb&n + g338.567 + 0.110 - 1665b&16:41:07.023&-46:15:27.88&1.15&0.68&-75.0&-75.6&-74.6&0.32&0.51&-0.2&sf - mmb&n + g338.567 + 0.110 - 1667a&16:41:07.012&-46:15:27.39&0.86&0.41&-74.6&-74.9&-74.0&0.43&0.70&-0.2&sf - mmb&n + + g338.660 + 1.290 - 1612a&16:36:25.708&-45:24:03.96&0.24&2.12&-114.5&-128.4&-106.1&0.36&0.61&-0.2&es - rms&n + g338.660 + 1.290 - 1612b&16:36:25.729&-45:24:04.53&0.24&1.82&-89.4&-95.2&-73.4&0.44&0.83&-0.2&es - rms&n + g338.660 + 1.290 - 1665a&16:36:25.725&-45:24:04.09&0.34&2.19&-109.5&-117.7&-104.1&0.26&0.42&-4.9&es - rms&n + g338.660 + 1.290 - 1665b&16:36:25.745&-45:24:03.90&0.37&3.30&-92.5&-101.0&-81.7&0.26&0.41&-4.9&es - rms&n + + [email protected] - 1665a&16:42:23.973&-46:17:59.70&8.20&5.11&-21.6&-22.4&-21.1&0.31&0.49&-0.2&sf - hop&c98 + [email protected] - 1665b&16:42:23.971&-46:17:59.89&6.61&4.62&-19.9&-20.6&-18.9&0.30&0.48&-0.2&sf - hop&c98 + [email protected] - 1665c&16:42:24.002&-46:18:00.44&0.45&0.13&-16.9&-17.1&-16.7&0.71&1.12&-0.2&sf - hop&c98 + [email protected] - 1665d&16:42:23.983&-46:17:59.71&0.41&0.22&-14.9&-15.4&-14.5&0.55&0.88&-0.2&sf - hop&c98 + + g338.699 + 1.346 - 1612a&16:36:20.545&-45:20:07.38&0.45&0.40&4.4&3.8&5.6&0.42&0.69&0.6&es - vis&n + g338.698 + 1.346 - 1612b&16:36:20.508&-45:20:08.22&0.28&0.40&32.0&30.5&34.2&0.43&0.70&0.6&es - vis&n + + [email protected] - 1612a&16:49:07.868&-47:14:41.07&0.66&1.13&-72.4&-73.6&-70.5&0.38&0.60&-4.4&es - sev&s97 + [email protected] - 1612b&16:49:07.846&-47:14:41.32&0.54&1.03&-46.2&-49.1&-44.6&0.42&0.66&-4.4&es - sev&s97 + + g338.829 + 0.162 - 1612a&16:41:53.693&-46:01:36.58&0.83&2.82&-34.9&-36.3&-28.6&0.43&0.70&-0.2&es - vis&n + g338.829 + 0.162 - 1612b&16:41:53.709&-46:01:36.67&0.55&1.29&12.9&10.1&15.0&0.37&0.60&-0.2&es - vis&n + + [email protected] - 1612a&16:46:22.274&-46:38:09.93&0.31&0.39&-120.1&-120.5&-118.3&1.37&4.50&-1.2&es - vis&n + [email protected] - 1612b&16:46:22.354&-46:38:11.27&0.41&0.29&-92.2&-92.8&-91.5&0.54&0.86&-1.2&es - vis&n + + [email protected] - 1665a&16:43:08.236&-46:09:13.53&0.39&0.23&-41.9&-42.2&-41.3&0.55&0.87&-0.2&sf - mmb&c98 + [email protected] - 1665b&16:43:08.239&-46:09:13.74&0.62&0.31&-39.4&-40.0&-39.1&0.47&0.83&-0.2&sf - mmb&c98 + [email protected] - 1665c&16:43:08.241&-46:09:13.20&1.15&0.69&-38.2&-38.6&-37.8&0.34&0.55&-0.2&sf - mmb&c98 + [email protected] - 1665d&16:43:08.237&-46:09:13.06&1.49&0.59&-37.2&-37.3&-36.9&0.27&0.43&-0.2&sf - mmb&c98 + [email protected] - 1665e&16:43:08.229&-46:09:13.18&1.90&2.03&-36.3&-36.4&-34.7&0.23&0.37&-0.2&sf - mmb&c98 + [email protected] - 1665f&16:43:08.232&-46:09:13.46&1.82&1.05&-33.8&-34.3&-33.4&0.25&0.40&-0.2&sf - mmb&c98 + [email protected] - 1667a&16:43:08.233&-46:09:13.32&0.61&0.91&-37.8&-39.3&-35.8&0.38&0.62&-0.2&sf - mmb&c98 + [email protected] - 1667b&16:43:08.220&-46:09:13.15&1.13&0.76&-33.6&-34.9&-33.2&0.36&0.58&-0.2&sf - mmb&c98 + + g338.902 + 0.015 - 1612a&16:42:48.704&-46:04:06.75&0.33&0.56&-80.8&-82.6&-78.0&0.46&0.74&-0.2&u&n + + [email protected] - 1612a&16:46:21.375&-46:30:19.49&1.12&1.62&-143.0&-144.6&-140.5&0.25&0.39&-1.2&es - sev&s97 + [email protected] - 1612b&16:46:21.343&-46:30:19.49&1.29&2.60&-115.6&-121.0&-114.6&0.22&0.36&-1.2&es - sev&s97 + + [email protected] - 1665a&16:44:48.948&-46:10:13.59&0.38&0.24&-120.4&-121.0&-120.1&0.58&0.96&-5.7&sf - mmb&c98 + [email protected] - 1665b&16:44:48.962&-46:10:13.23&0.34&0.28&-110.5&-110.8&-109.5&0.45&0.76&-5.7&sf - mmb&c98 + [email protected] - 1667a&16:44:48.971&-46:10:14.24&0.48&0.25&-121.1&-121.5&-120.6&0.31&0.53&-5.5&sf - mmb&c98 + + g339.092 + 0.244 - 1612a&16:42:32.071&-45:46:29.68&1.60&2.82&-22.4&-23.9&-19.9&0.23&0.36&1.6&es - vis&n + g339.092 + 0.244 - 1612b&16:42:32.066&-45:46:29.37&0.41&0.82&-18.7&-19.4&-15.3&0.34&0.54&1.6&es - vis&n + g339.092 + 0.244 - 1612c&16:42:32.090&-45:46:29.50&0.29&0.47&5.5&3.3&6.9&0.56&0.90&1.6&es - vis&n + g339.092 + 0.244 - 1612d&16:42:32.055&-45:46:29.60&0.38&0.59&8.4&7.4&10.6&0.59&0.93&1.6&es - vis&n + + g339.156 + 0.786 - 1612a&16:40:27.070&-45:22:05.75&0.43&0.48&-116.4&-117.3&-115.5&0.48&0.69&-16.7&es - vis&n + g339.156 + 0.786 - 1612b&16:40:27.080&-45:22:05.54&0.37&0.38&-84.3&-86.0&-83.7&0.50&0.71&-16.7&es - vis&n + + g339.282 + 0.136 - 1665a&16:43:43.080&-45:42:08.71&0.54&1.22&-71.7&-73.3&-68.9&0.24&0.37&1.6&sf - mmb&c98 + + g339.284 + 2.051 - 1612a&16:35:37.311&-44:25:38.47&2.63&3.00&61.2&59.4&63.0&0.13&0.19&-13.6&es - sev&s97 + g339.284 + 2.051 - 1612b&16:35:37.321&-44:25:38.54&1.92&2.72&83.7&80.3&84.8&0.18&0.28&-13.6&es - sev&s97 + + g339.294 + 0.139 - 1667a&16:43:44.937&-45:41:28.54&0.99&0.57&-73.3&-73.9&-72.6&0.31&0.48&1.1&sf - mmb&n + + [email protected] - 1612a&16:48:39.519&-46:19:13.88&0.35&0.64&-83.7&-85.3&-81.7&0.44&0.65&-14.7&es - vis&n + [email protected] - 1612b&16:48:39.484&-46:19:13.88&0.42&0.45&-59.7&-61.2&-59.0&0.55&0.94&-14.7&es - vis&n + + g339.380 + 0.593 - 1612a&16:42:07.068&-45:19:42.56&0.30&0.79&-58.6&-60.6&-56.1&0.42&0.65&-12.3&es - vis&n + g339.380 + 0.593 - 1612b&16:42:07.047&-45:19:42.38&0.35&1.09&-28.2&-32.5&-27.1&0.27&0.41&-12.3&es - vis&n + + [email protected] - 1612a&16:48:35.626&-46:08:55.21&0.41&0.38&-129.4&-130.2&-128.4&0.37&0.54&-14.7&es - vis&n + [email protected] - 1612b&16:48:35.647&-46:08:54.76&0.36&0.40&-99.8&-101.6&-98.9&0.47&0.74&-14.7&es - vis&n + + g339.559 + 0.237 - 1612a&16:44:19.018&-45:25:37.61&0.64&1.35&-186.6&-187.4&-184.3&0.32&0.48&-13.9&es - vis&n + g339.559 + 0.237 - 1612b&16:44:19.029&-45:25:38.42&0.32&0.71&-162.4&-165.2&-161.1&0.51&0.77&-13.9&es - vis&n + + g339.587 + 1.638 - 1612a&16:38:28.619&-44:28:48.89&0.17&0.19&-81.8&-82.9&-81.1&0.66&1.03&-9.4&es - vis&n + g339.587 + 1.638 - 1612b&16:38:28.661&-44:28:49.60&0.27&0.14&-67.4&-68.8&-67.0&0.89&1.40&-9.4&es - vis&n + + [email protected] - 1612a&16:49:25.339&-45:46:35.12&0.30&0.31&-129.8&-130.6&-128.8&0.39&0.61&-7.8&u&n + + [email protected] - 1665a&16:49:07.978&-45:37:58.50&0.81&0.70&-98.0&-98.8&-97.0&0.37&0.61&-7.2&sf - mmb&n + [email protected] - 1665b&16:49:07.957&-45:37:58.92&1.05&0.63&-95.8&-96.6&-95.3&0.31&0.49&-7.2&sf - mmb&n + [email protected] - 1667a&16:49:08.010&-45:37:58.30&0.42&0.24&-99.8&-100.2&-99.4&0.53&0.68&-7.4&sf - mmb&n + [email protected] - 1667b&16:49:07.965&-45:37:58.75&0.26&0.13&-96.9&-97.6&-96.7&0.66&1.05&-7.4&sf - mmb&n + [email protected] - 1667c&16:49:07.976&-45:37:58.36&0.51&0.28&-96.1&-96.3&-95.4&0.45&0.71&-7.4&sf - mmb&n + + [email protected] - 1612a&16:47:42.749&-45:23:25.61&1.57&2.72&-43.9&-45.1&-40.1&0.27&0.42&-4.7&es - sev&s97 + [email protected] - 1612b&16:47:42.755&-45:23:25.53&2.49&3.47&-7.0&-12.9&-6.1&0.34&0.53&-4.7&es - sev&s97 + + [email protected] - 1612a&16:48:46.334&-45:31:51.34&0.40&0.32&-89.7&-90.5&-89.2&0.41&0.67&-4.7&sf - mmb&n + [email protected] - 1665a&16:48:46.261&-45:31:51.29&0.71&0.66&-97.1&-99.1&-96.4&0.31&0.50&-4.7&sf - mmb&n + [email protected] - 1665b&16:48:46.308&-45:31:50.62&0.84&0.70&-92.4&-93.4&-91.6&0.36&0.56&-4.7&sf - mmb&n + [email protected] - 1665c&16:48:46.321&-45:31:50.85&4.63&3.30&-90.0&-91.2&-89.4&0.26&0.41&-4.7&sf - mmb&n + [email protected] - 1665d&16:48:46.315&-45:31:50.83&1.16&1.27&-88.5&-89.0&-87.2&0.29&0.46&-4.7&sf - mmb&n + [email protected] - 1667a&16:48:46.351&-45:31:50.70&0.26&0.33&-91.1&-92.6&-90.4&0.66&1.04&-4.7&sf - mmb&n + [email protected] - 1667b&16:48:46.305&-45:31:50.71&0.49&0.49&-88.6&-90.0&-87.8&0.34&0.53&-4.7&sf - mmb&n + [email protected] - 1667c&16:48:46.335&-45:31:50.66&0.42&0.14&-87.0&-87.4&-86.9&0.51&0.80&-4.7&sf - mmb&n + [email protected] - 1667d&16:48:46.343&-45:31:50.39&0.73&0.52&-82.2&-83.0&-81.7&0.24&0.38&-4.7&sf - mmb&n + [email protected] - 1667e&16:48:46.344&-45:31:50.62&0.20&0.17&-79.8&-80.8&-79.5&0.53&0.84&-4.7&sf - mmb&n + + [email protected] - 1612a&16:49:10.967&-45:34:19.82&1.31&3.93&-79.3&-82.5&-77.5&0.37&0.70&-7.8&es - jon&s97 + [email protected] - 1612b&16:49:10.967&-45:34:19.71&1.02&6.28&-76.4&-77.0&-50.7&0.23&0.37&-7.8&es - jon&s97 + [email protected] - 1612c&16:49:10.970&-45:34:19.81&2.54&6.15&-40.0&-50.3&-38.9&0.23&0.36&-7.8&es - jon&s97 + [email protected] - 1612d&16:49:10.965&-45:34:19.84&2.34&5.80&-37.1&-38.4&-35.3&0.37&0.70&-7.8&es - jon&s97 + [email protected] - 1612e&16:49:10.965&-45:34:19.97&1.40&1.50&-34.3&-34.8&-32.6&0.29&0.45&-7.8&es - jon&s97 + + [email protected] - 1667a&16:47:31.796&-45:16:22.65&0.46&0.48&-10.5&-11.4&-9.2&0.39&0.62&-4.7&es - sev&s97 + + [email protected] - 1665a&16:48:13.869&-45:21:44.74&1.10&1.56&-57.9&-59.1&-56.4&0.47&0.75&-4.7&sf - mmb&c98 + [email protected] - 1665b&16:48:13.891&-45:21:44.69&64.42&50.36&-53.6&-56.0&-52.5&0.06&0.09&-4.7&sf - mmb&c98 + [email protected] - 1665c&16:48:13.863&-45:21:45.43&9.91&13.04&-51.3&-52.0&-49.9&0.08&0.13&-4.7&sf - mmb&c98 + [email protected] - 1665d&16:48:13.884&-45:21:45.30&6.31&6.08&-48.4&-49.4&-46.8&0.13&0.21&-4.7&sf - mmb&c98 + [email protected] - 1667a&16:48:13.878&-45:21:45.30&1.38&1.99&-69.7&-71.5&-69.4&0.31&0.49&-4.7&sf - mmb&c98 + [email protected] - 1667b&16:48:13.874&-45:21:45.33&1.10&1.51&-68.6&-68.9&-66.7&0.36&0.58&-4.7&sf - mmb&c98 + [email protected] - 1667c&16:48:13.884&-45:21:44.98&0.74&1.44&-58.0&-61.0&-57.5&0.55&0.91&-4.7&sf - mmb&c98 + [email protected] - 1667d&16:48:13.842&-45:21:44.51&3.77&4.92&-56.1&-57.1&-55.3&0.11&0.18&-4.7&sf - mmb&c98 + [email protected] - 1667e&16:48:13.852&-45:21:44.92&4.17&2.48&-54.3&-54.9&-54.0&0.11&0.18&-4.7&sf - mmb&c98 + [email protected] - 1667f&16:48:13.849&-45:21:45.60&3.87&1.61&-53.3&-53.5&-53.1&0.13&0.20&-4.7&sf - mmb&c98 + [email protected] - 1667g&16:48:13.800&-45:21:45.67&4.21&2.50&-52.5&-52.7&-51.8&0.15&0.24&-4.7&sf - mmb&c98 + [email protected] - 1667h&16:48:13.865&-45:21:45.38&3.22&2.28&-50.9&-51.4&-50.0&0.20&0.33&-4.7&sf - mmb&c98 + [email protected] - 1667i&16:48:13.871&-45:21:45.36&4.33&5.59&-48.9&-49.6&-45.2&0.16&0.26&-4.7&sf - mmb&c98 + [email protected] - 1667j&16:48:13.998&-45:21:44.24&0.30&0.25&-43.9&-44.8&-43.5&1.44&2.63&-4.7&sf - mmb&c98 + + g340.091 + 1.485 - 1612a&16:40:59.420&-44:12:19.51&0.23&0.20&-34.6&-35.8&-34.4&0.78&1.26&-9.5&u&n + + [email protected] - 1612a&16:49:22.391&-45:24:46.61&2.19&3.21&-101.8&-102.8&-96.9&0.29&0.46&-4.7&es - sev&s97 + [email protected] - 1612b&16:49:22.379&-45:24:46.64&1.76&2.88&-71.9&-75.5&-71.0&0.34&0.54&-4.7&es - sev&s97 + + [email protected] - 1612a&16:48:04.860&-45:05:22.79&0.82&0.57&-122.0&-122.6&-121.3&0.92&1.48&-4.7&u&n + + [email protected] - 1612a&16:53:45.207&-45:53:23.06&0.29&0.56&31.6&29.4&33.0&0.36&0.59&-3.6&es - rms&n + [email protected] - 1612b&16:53:45.175&-45:53:22.55&0.38&0.54&58.9&56.6&59.8&0.58&0.96&-3.6&es - rms&n + + g340.267 + 0.825 - 1612a&16:44:25.258&-44:30:23.78&0.37&0.51&-81.2&-82.8&-80.5&0.34&0.56&-3.0&es - gre&n + g340.267 + 0.825 - 1612b&16:44:25.294&-44:30:23.88&0.25&0.30&-65.9&-67.4&-65.1&0.59&0.99&-3.0&es - gre&n + + g340.299 + 0.115 - 1612a&16:47:34.271&-44:56:38.93&0.38&0.43&-88.5&-89.5&-87.7&0.48&0.78&-4.6&es - vis&n + g340.299 + 0.115 - 1612b&16:47:34.298&-44:56:38.94&0.22&0.30&-51.4&-52.3&-50.0&0.41&0.66&-4.6&es - vis&n + + g340.361 + 1.171 - 1612a&16:43:18.026&-44:12:30.71&0.68&1.02&-183.4&-183.9&-179.9&0.36&0.58&0.3&es - rms&n + g340.361 + 1.171 - 1612b&16:43:18.039&-44:12:30.80&1.29&1.51&-154.1&-156.2&-153.1&0.21&0.33&0.3&es - rms&n + + [email protected] - 1612a&16:48:36.730&-44:56:19.86&6.89&15.85&-168.0&-168.9&-161.2&0.10&0.16&-4.0&es - rms&s97 + [email protected] - 1612b&16:48:36.745&-44:56:19.47&0.43&0.57&-160.2&-160.8&-158.5&0.44&0.70&-4.0&es - rms&s97 + [email protected] - 1612c&16:48:36.771&-44:56:19.74&0.31&0.41&-157.6&-158.0&-155.8&0.41&0.66&-4.0&es - rms&s97 + [email protected] - 1612d&16:48:36.728&-44:56:19.87&5.28&14.90&-134.8&-142.6&-132.2&0.09&0.15&-4.0&es - rms&s97 + [email protected] - 1667a&16:48:36.699&-44:56:20.44&0.44&0.23&-168.9&-169.3&-168.5&0.55&0.89&-0.1&es - rms&s97 + [email protected] - 1667b&16:48:36.749&-44:56:19.93&0.41&0.63&-166.7&-168.0&-165.4&0.38&0.62&-0.1&es - rms&s97 + [email protected] - 1667c&16:48:36.763&-44:56:19.51&0.29&0.46&-135.1&-137.7&-134.2&0.49&0.77&-0.1&es - rms&s97 + + g340.505 + 0.316 - 1612a&16:47:27.669&-44:39:24.19&0.60&0.85&-121.6&-122.2&-119.5&0.34&0.56&-0.9&es - rms&s97 + g340.505 + 0.316 - 1612b&16:47:27.703&-44:39:24.55&0.42&0.59&-89.3&-92.2&-89.1&0.25&0.40&-0.9&es - rms&s97 + + g340.539 + 0.766 - 1612a&16:45:39.747&-44:20:20.46&0.30&0.48&-47.8&-49.2&-46.1&0.51&0.85&0.8&es - rms&n + g340.539 + 0.766 - 1612b&16:45:39.750&-44:20:20.08&1.71&2.57&-14.0&-17.0&-12.9&0.20&0.32&0.8&es - rms&n + g340.539 + 0.766 - 1667a&16:45:39.744&-44:20:20.36&0.29&0.32&-48.9&-49.6&-47.0&0.68&1.14&1.2&es - rms&n + + g340.747 + 1.442 - 1612a&16:43:34.227&-43:44:25.46&0.73&0.65&-21.3&-22.2&-19.9&0.34&0.58&-5.3&es - vis&n + g340.747 + 1.442 - 1612b&16:43:34.296&-43:44:24.91&0.26&0.21&7.0&6.4&7.8&0.56&0.96&-5.3&es - vis&n + + [email protected] - 1612a&16:57:39.603&-45:47:21.62&0.29&0.26&21.9&20.9&22.7&0.57&0.79&-18.5&es - vis&n + [email protected] - 1612b&16:57:39.637&-45:47:21.53&0.34&0.25&46.7&45.4&47.2&0.37&0.52&-18.5&es - vis&n + + [email protected] - 1665a&16:50:14.822&-44:42:26.37&0.69&0.51&-108.6&-109.1&-107.4&0.42&0.64&-13.8&sf - mmb&c98 + [email protected] - 1665b&16:50:14.804&-44:42:26.52&0.71&0.41&-106.6&-106.9&-106.1&0.36&0.55&-13.8&sf - mmb&c98 + [email protected] - 1665c&16:50:14.775&-44:42:26.16&0.62&0.33&-105.3&-105.6&-104.7&0.40&0.60&-13.8&sf - mmb&c98 + [email protected] - 1665d&16:50:14.800&-44:42:25.97&4.26&2.58&-103.5&-104.3&-103.4&0.36&0.55&-13.8&sf - mmb&c98 + [email protected] - 1665e&16:50:14.807&-44:42:26.38&3.90&2.61&-102.5&-103.0&-102.1&0.30&0.46&-13.8&sf - mmb&c98 + [email protected] - 1665f&16:50:14.802&-44:42:26.45&7.01&5.40&-101.6&-101.7&-98.6&0.32&0.48&-13.8&sf - mmb&c98 + [email protected] - 1665g&16:50:14.806&-44:42:26.43&0.72&0.67&-89.6&-91.1&-88.5&0.45&0.67&-13.8&sf - mmb&c98 + [email protected] - 1667a&16:50:14.810&-44:42:26.52&0.25&0.20&-104.2&-104.9&-103.5&0.37&0.57&-13.8&sf - mmb&c98 + [email protected] - 1667b&16:50:14.809&-44:42:26.57&3.32&4.59&-102.0&-103.1&-99.2&0.36&0.55&-13.8&sf - mmb&c98 + [email protected] - 1667c&16:50:14.799&-44:42:26.32&0.70&0.66&-92.1&-92.6&-91.3&0.37&0.56&-13.8&sf - mmb&c98 + [email protected] - 1667d&16:50:14.792&-44:42:25.84&0.43&0.25&-90.8&-90.8&-89.5&0.45&0.70&-13.8&sf - mmb&c98 + [email protected] - 1720a&16:50:14.795&-44:42:25.88&0.46&0.61&-105.5&-106.7&-103.7&0.40&0.60&-14.9&sf - mmb&c98 + [email protected] - 1720b&16:50:14.912&-44:42:26.97&0.26&0.19&-100.3&-100.7&-99.5&0.63&0.94&-14.9&sf - mmb&c98 + + [email protected] - 1612a&16:49:55.349&-44:37:42.11&2.55&2.71&-144.4&-145.4&-141.3&0.20&0.31&-13.8&es - sev&s97 + [email protected] - 1612b&16:49:55.349&-44:37:42.13&1.90&2.58&-115.7&-119.5&-115.0&0.27&0.41&-13.8&es - sev&s97 + + g340.847 + 0.144 - 1612a&16:49:26.221&-44:30:21.22&1.15&2.25&-86.7&-87.7&-81.9&0.33&0.50&-13.8&es - sev&s97 + g340.847 + 0.144 - 1612b&16:49:26.224&-44:30:21.20&0.60&1.11&-61.2&-64.6&-60.1&0.46&0.70&-13.8&es - sev&s97 + + [email protected] - 1612a&16:59:14.333&-45:50:31.45&0.82&0.94&-51.6&-52.7&-50.5&0.34&0.50&-14.6&es - sev&s97 + [email protected] - 1612b&16:59:14.380&-45:50:31.54&0.53&0.86&-23.3&-25.0&-21.9&0.38&0.57&-14.6&es - sev&s97 + + g340.987 + 0.831 - 1612a&16:47:00.487&-43:57:20.86&1.02&0.94&-59.9&-60.6&-59.3&0.33&0.56&-8.2&es - sev&s97 + g340.987 + 0.831 - 1612b&16:47:00.486&-43:57:21.18&0.67&0.87&-58.7&-58.8&-55.7&0.34&0.55&-8.2&es - sev&s97 + g340.987 + 0.831 - 1612c&16:47:00.507&-43:57:20.92&0.54&1.39&-41.9&-44.7&-39.3&0.29&0.47&-8.2&es - sev&s97 + + [email protected] - 1612a&16:50:59.912&-44:32:28.27&0.40&0.91&-82.3&-84.2&-79.2&0.37&0.59&-11.5&es - sev&s97 + [email protected] - 1612b&16:50:59.884&-44:32:27.41&0.30&0.52&-41.5&-43.3&-39.7&0.32&0.49&-11.5&es - sev&s97 + + [email protected] - 1612a&17:00:30.707&-45:49:44.09&0.89&1.30&-132.5&-133.6&-130.4&0.39&0.60&-14.6&es - rms&s97 + [email protected] - 1612b&17:00:30.722&-45:49:44.11&1.39&1.41&-111.1&-114.0&-110.4&0.29&0.43&-14.6&es - rms&s97 + + [email protected] - 1665a&16:55:37.821&-45:06:11.74&0.32&0.28&3.4&2.0&4.2&0.31&0.47&-9.1&es - mcd&n + [email protected] - 1665b&16:55:37.853&-45:06:11.32&1.46&2.11&5.5&4.7&8.6&0.20&0.31&-9.1&es - mcd&n + [email protected] - 1667a&16:55:37.871&-45:06:11.41&0.80&1.44&2.3&0.9&4.0&0.32&0.50&-9.5&es - mcd&n + [email protected] - 1667b&16:55:37.854&-45:06:11.20&1.47&3.41&6.8&4.4&9.7&0.22&0.34&-9.5&es - mcd&n + + [email protected] - 1612a&16:59:22.628&-45:36:08.78&1.72&1.46&-63.7&-65.8&-63.6&0.21&0.32&-10.3&es - sev&s97 + [email protected] - 1612b&16:59:22.631&-45:36:08.75&1.63&2.09&-62.8&-63.1&-59.5&0.28&0.42&-10.3&es - sev&s97 + [email protected] - 1612c&16:59:22.593&-45:36:07.78&0.28&0.15&-39.2&-39.5&-38.6&0.50&0.76&-10.3&es - sev&s97 + [email protected] - 1612d&16:59:22.623&-45:36:08.69&0.96&1.71&-31.3&-35.0&-29.5&0.29&0.45&-10.3&es - sev&s97 + [email protected] - 1665a&16:59:22.598&-45:36:08.44&0.29&0.17&-39.0&-40.0&-38.7&0.57&0.93&-7.6&es - sev&s97 + [email protected] - 1665b&16:59:22.624&-45:36:08.64&0.51&0.73&-36.8&-38.3&-36.1&0.41&0.67&-7.6&es - sev&s97 + [email protected] - 1665c&16:59:22.613&-45:36:08.92&0.38&0.46&-34.7&-35.6&-33.9&0.36&0.60&-7.6&es - sev&s97 + [email protected] - 1667a&16:59:22.595&-45:36:08.62&0.48&0.51&-37.2&-38.9&-36.7&0.33&0.53&-7.1&es - sev&s97 + [email protected] - 1667b&16:59:22.622&-45:36:08.68&0.55&1.23&-35.4&-36.3&-31.9&0.30&0.47&-7.1&es - sev&s97 + + [email protected] - 1612a&16:51:02.039&-44:23:29.66&1.83&3.31&-61.3&-62.4&-56.0&0.19&0.29&-11.5&es - sev&s97 + [email protected] - 1612b&16:51:02.041&-44:23:29.47&1.67&4.33&-25.0&-31.0&-23.8&0.25&0.40&-11.5&es - sev&s97 + + [email protected] - 1665a&16:52:17.850&-44:26:52.90&0.66&0.22&-41.8&-42.0&-41.6&0.34&0.53&-5.8&sf - mmb&c98 + [email protected] - 1665b&16:52:17.856&-44:26:52.52&1.49&1.34&-40.1&-41.1&-39.8&0.32&0.50&-5.8&sf - mmb&c98 + [email protected] - 1665c&16:52:17.842&-44:26:52.06&0.49&0.32&-39.1&-39.3&-38.5&0.35&0.56&-5.8&sf - mmb&c98 + [email protected] - 1665d&16:52:17.849&-44:26:52.48&3.81&3.40&-37.3&-38.0&-35.4&0.27&0.43&-5.8&sf - mmb&c98 + [email protected] - 1667a&16:52:17.833&-44:26:53.07&0.52&0.25&-41.7&-42.2&-41.3&0.48&0.76&-5.8&sf - mmb&c98 + [email protected] - 1667b&16:52:17.847&-44:26:52.21&0.72&0.50&-40.0&-40.9&-39.6&0.38&0.60&-5.8&sf - mmb&c98 + [email protected] - 1667c&16:52:17.868&-44:26:52.85&0.70&0.96&-37.8&-39.1&-36.5&0.36&0.57&-5.8&sf - mmb&c98 + + [email protected] - 1612a&16:54:42.705&-44:43:26.90&1.76&1.93&11.6&10.5&14.5&0.29&0.51&-1.2&es - rms&s97 + [email protected] - 1612b&16:54:42.700&-44:43:26.63&1.04&1.09&28.5&26.4&29.1&0.35&0.62&-1.2&es - rms&s97 + + g341.276 + 0.062 - 1665a&16:51:19.426&-44:13:44.31&0.30&0.19&-74.0&-74.6&-73.8&0.77&1.18&-11.5&sf - mmb&c98 + g341.276 + 0.062 - 1665b&16:51:19.441&-44:13:44.21&1.58&1.10&-72.8&-73.3&-71.8&0.33&0.52&-11.5&sf - mmb&c98 + + g341.285 + 0.118 - 1612a&16:51:06.780&-44:11:09.77&4.08&6.02&-32.2&-32.9&-29.3&0.29&0.45&-11.5&es - sev&s97 + g341.285 + 0.118 - 1612b&16:51:06.776&-44:11:10.33&0.61&1.17&-27.7&-28.8&-24.7&0.55&0.86&-11.5&es - sev&s97 + g341.285 + 0.118 - 1612c&16:51:06.782&-44:11:09.74&3.64&6.96&-3.4&-9.7&-2.5&0.28&0.43&-11.5&es - sev&s97 + + g341.319 + 0.821 - 1612a&16:48:14.294&-43:42:33.88&0.19&0.25&-102.7&-103.3&-101.1&0.67&1.13&-5.9&es - vis&n + g341.319 + 0.821 - 1612b&16:48:14.300&-43:42:33.40&0.26&0.20&-80.6&-81.1&-79.8&0.48&0.78&-5.9&es - vis&n + + [email protected] - 1612a&16:53:33.678&-44:21:45.38&0.44&0.91&-44.0&-47.9&-42.9&0.37&0.61&-5.8&es - vis&n + [email protected] - 1612b&16:53:33.715&-44:21:44.91&0.44&0.75&-11.3&-13.4&-10.2&0.33&0.52&-5.8&es - vis&n + + [email protected] - 1612a&16:52:54.184&-44:09:48.40&0.30&0.49&-66.1&-67.9&-65.2&0.44&0.70&-5.8&es - vis&n + [email protected] - 1612b&16:52:54.150&-44:09:47.25&0.28&0.48&-32.9&-33.8&-30.7&0.83&1.40&-5.8&es - vis&n + + [email protected] - 1612a&16:57:30.590&-44:38:12.39&0.55&0.59&-180.3&-181.3&-179.0&0.42&0.76&-0.7&es - vis&n + [email protected] - 1612b&16:57:30.653&-44:38:11.10&0.37&0.30&-153.4&-154.0&-152.7&0.46&0.82&-0.7&es - vis&n + + g341.681 + 0.264 - 1612a&16:51:53.641&-43:47:16.47&0.45&2.16&54.0&47.2&58.1&0.34&0.57&1.9&u&n + + g341.850 + 0.215 - 1612a&16:52:41.547&-43:41:17.84&0.53&0.88&-83.0&-84.1&-80.5&0.35&0.56&-2.0&es - vis&n + g341.850 + 0.215 - 1612b&16:52:41.562&-43:41:18.16&0.29&0.46&-79.1&-80.0&-76.9&0.32&0.51&-2.0&es - vis&n + g341.850 + 0.215 - 1612c&16:52:41.554&-43:41:18.03&0.65&1.82&-43.9&-49.6&-42.4&0.37&0.62&-2.0&es - vis&n + + g341.938 + 0.773 - 1612a&16:50:37.781&-43:15:54.15&0.80&1.08&-173.8&-174.6&-171.9&0.37&0.59&2.1&es - rms&s97 + g341.938 + 0.773 - 1612b&16:50:37.791&-43:15:54.22&0.68&1.06&-147.9&-150.1&-147.0&0.30&0.52&2.1&es - rms&s97 + + g341.993 + 1.402 - 1612a&16:48:10.907&-42:49:09.52&0.53&0.99&-88.0&-89.5&-85.9&0.35&0.63&-2.4&es - sev&s97 + g341.993 + 1.402 - 1612b&16:48:10.903&-42:49:10.34&0.30&0.19&-85.2&-85.4&-84.1&0.48&0.83&-2.4&es - sev&s97 + g341.993 + 1.402 - 1612c&16:48:10.908&-42:49:09.60&1.14&2.73&-48.1&-51.8&-47.3&0.26&0.47&-2.4&es - sev&s97 + + g342.004 + 0.251 - 1612a&16:53:04.885&-43:32:47.07&1.59&3.70&-52.0&-53.7&-47.8&0.28&0.45&-2.0&es - hop&s97 + g342.004 + 0.251 - 1612b&16:53:04.891&-43:32:47.06&0.68&1.45&-40.6&-42.8&-37.8&0.40&0.63&-2.0&es - hop&s97 + g342.004 + 0.251 - 1612c&16:53:04.876&-43:32:47.43&0.47&0.29&-33.6&-34.2&-32.8&0.31&0.49&-2.0&es - hop&s97 + + [email protected] - 1612a&16:55:18.552&-43:47:55.39&0.67&0.58&-99.3&-99.8&-98.4&0.92&0.60&-82.9&es - vis&n + [email protected] - 1612b&16:55:18.602&-43:47:55.29&0.39&0.48&-71.9&-73.4&-71.2&1.04&0.65&-82.9&es - vis&n + + [email protected] - 1612a&16:55:32.799&-43:48:53.29&0.51&0.66&-120.3&-121.6&-119.3&0.81&0.56&-82.4&es - vis&n + [email protected] - 1612b&16:55:32.766&-43:48:54.35&0.25&0.23&-87.2&-88.0&-86.2&0.76&0.53&-82.4&es - vis&n + + g342.094 + 0.978 - 1612a&16:50:18.946&-43:00:51.12&1.88&3.38&-69.3&-70.2&-66.2&0.32&0.38&26.7&es - sev&s97 + g342.094 + 0.978 - 1612b&16:50:18.941&-43:00:51.10&0.45&1.19&-39.7&-43.4&-38.9&0.65&0.78&26.7&es - sev&s97 + g342.094 + 0.978 - 1612c&16:50:18.947&-43:00:51.29&1.53&2.57&-35.9&-38.4&-34.4&0.44&0.52&26.7&es - sev&s97 + + g342.183 + 0.351 - 1612a&16:53:16.744&-43:20:39.32&0.34&0.19&-121.6&-122.2&-121.3&0.91&1.07&29.5&es - sev&s97 + g342.183 + 0.351 - 1612b&16:53:16.690&-43:20:38.78&0.27&0.16&-120.3&-120.9&-120.0&0.53&0.65&29.5&es - sev&s97 + g342.183 + 0.351 - 1612c&16:53:16.727&-43:20:39.62&0.44&0.35&-94.8&-95.4&-94.1&0.53&0.64&29.5&es - sev&s97 + g342.183 + 0.351 - 1612d&16:53:16.753&-43:20:39.41&0.57&0.39&-93.1&-93.6&-92.7&0.47&0.57&29.5&es - sev&s97 + + g342.222 + 0.041 - 1612a&16:54:44.146&-43:30:35.05&0.22&0.33&-52.5&-53.9&-51.7&1.14&0.75&-82.9&u&n + + [email protected] - 1612a&16:55:49.861&-43:39:31.12&0.24&0.17&-95.6&-95.7&-94.3&0.81&0.52&-82.9&es - vis&n + [email protected] - 1612b&16:55:49.767&-43:39:32.22&0.32&0.28&-61.9&-62.5&-60.7&1.14&0.75&-82.9&es - vis&n + + [email protected] - 1612a&17:01:38.078&-44:28:55.62&0.29&0.37&-128.0&-128.9&-126.7&0.54&0.67&21.1&es - sev&s97 + [email protected] - 1612b&17:01:38.079&-44:28:55.06&0.30&0.48&-91.5&-93.0&-90.3&0.41&0.50&21.1&es - sev&s97 + + g342.304 + 1.366 - 1612a&16:49:25.341&-42:36:17.70&0.46&0.43&54.0&53.4&55.2&1.07&2.25&18.3&es - vis&n + g342.303 + 1.366 - 1612b&16:49:25.310&-42:36:17.65&0.31&0.31&79.6&78.4&80.2&1.22&3.16&18.3&es - vis&n + + g342.369 + 0.140 - 1612a&16:54:49.428&-43:20:00.93&0.50&0.22&-9.0&-9.7&-8.9&0.62&1.03&-16.1&sf - mmb&n + g342.369 + 0.140 - 1612b&16:54:49.477&-43:20:00.02&1.64&1.15&-7.8&-8.4&-7.0&0.38&0.61&-16.1&sf - mmb&n + + [email protected] - 1612a&17:04:16.172&-44:12:24.50&1.40&1.82&-161.5&-162.6&-159.9&0.42&0.77&23.1&es - sev&s97 + [email protected] - 1612b&17:04:16.175&-44:12:23.92&0.60&0.58&-146.9&-148.1&-146.3&0.75&1.35&23.1&es - sev&s97 + + [email protected] - 1612a&16:57:11.744&-43:09:35.98&2.82&2.07&-126.8&-127.9&-126.5&0.36&0.64&23.1&es - rms&s97 + [email protected] - 1612b&16:57:11.748&-43:09:36.39&2.02&2.03&-125.8&-126.1&-124.7&0.37&0.68&23.1&es - rms&s97 + [email protected] - 1612c&16:57:11.757&-43:09:36.23&0.85&1.23&-123.9&-124.3&-119.3&0.73&1.35&23.1&es - rms&s97 + [email protected] - 1612d&16:57:11.776&-43:09:36.11&0.88&3.21&-90.4&-99.3&-90.2&0.85&1.55&23.1&es - rms&s97 + [email protected] - 1612e&16:57:11.760&-43:09:35.90&1.57&3.68&-88.5&-89.7&-85.7&0.51&0.93&23.1&es - rms&s97 + [email protected] - 1667a&16:57:11.741&-43:09:36.11&1.52&6.86&-127.1&-130.5&-118.6&0.53&0.96&23.1&es - rms&s97 + [email protected] - 1667b&16:57:11.753&-43:09:35.97&1.51&6.68&-87.0&-97.6&-84.4&0.55&1.01&23.1&es - rms&s97 + + [email protected] - 1612a&16:57:52.393&-43:05:42.29&0.67&0.52&-23.2&-23.9&-21.7&0.64&1.20&23.1&u&n + + [email protected] - 1612a&16:59:02.738&-43:11:09.05&0.36&0.72&-83.6&-84.9&-81.7&0.83&1.42&24.3&u&n + + g343.001 + 0.910 - 1612a&16:53:44.113&-42:21:26.15&1.78&3.68&-120.1&-122.9&-118.4&0.61&0.98&27.1&es - sev&s97 + g343.001 + 0.910 - 1612b&16:53:44.127&-42:21:26.31&0.64&1.54&-117.1&-117.9&-114.3&0.70&1.14&27.1&es - sev&s97 + g343.001 + 0.910 - 1612c&16:53:44.152&-42:21:26.06&0.38&0.74&-112.4&-113.9&-110.2&1.01&1.65&27.1&es - sev&s97 + g343.001 + 0.910 - 1612d&16:53:44.097&-42:21:26.45&0.31&0.60&-84.2&-87.1&-83.0&0.80&1.28&27.1&es - sev&s97 + g343.001 + 0.910 - 1612e&16:53:44.111&-42:21:26.30&4.17&8.26&-78.4&-82.5&-76.2&0.30&0.49&27.1&es - sev&s97 + + [email protected] - 1612a&16:58:16.617&-42:52:38.72&24.67&20.45&-27.8&-29.9&-26.7&0.20&0.37&23.4&sf - vis&n + + [email protected] - 1665a&16:58:17.208&-42:52:08.16&113.80&77.13&-33.8&-36.8&-32.4&0.11&0.20&23.4&sf - hop&c98 + [email protected] - 1665b&16:58:17.209&-42:52:08.11&40.21&21.80&-31.6&-32.0&-31.3&0.07&0.14&23.4&sf - hop&c98 + [email protected] - 1665c&16:58:17.212&-42:52:08.00&47.85&28.28&-30.5&-30.8&-28.5&0.10&0.20&23.4&sf - hop&c98 + [email protected] - 1667a&16:58:17.091&-42:52:06.52&0.59&0.37&-35.8&-36.6&-35.3&0.83&1.49&23.5&sf - hop&c98 + [email protected] - 1667b&16:58:17.219&-42:52:08.60&1.02&0.72&-32.7&-33.5&-32.2&0.51&0.91&23.5&sf - hop&c98 + [email protected] - 1667c&16:58:17.186&-42:52:08.26&0.99&0.75&-31.0&-31.8&-29.6&0.68&1.22&23.5&sf - hop&c98 + + [email protected] - 1612a&16:59:25.937&-42:57:06.86&0.54&0.81&-84.0&-84.8&-81.2&0.75&1.35&23.4&u&n + + g343.339 + 0.774 - 1612a&16:55:27.519&-42:10:48.53&0.40&0.31&-130.6&-131.2&-129.9&1.25&2.09&27.1&es - vis&n + g343.339 + 0.774 - 1612b&16:55:27.585&-42:10:48.04&0.34&0.18&-102.7&-103.1&-101.7&1.57&2.78&27.1&es - vis&n + + g343.368 + 0.244 - 1612a&16:57:47.564&-42:29:20.44&0.26&0.45&-101.5&-101.9&-99.2&1.50&1.27&-50.2&es - sev&s97 + g343.368 + 0.244 - 1612b&16:57:47.544&-42:29:20.61&0.68&1.04&-72.8&-74.2&-71.5&1.02&0.87&-50.2&es - sev&s97 + + [email protected] - 1612a&17:04:40.638&-43:27:59.88&0.26&0.38&-72.3&-72.8&-70.1&1.01&0.83&-57.3&es - vis&s97 + [email protected] - 1612b&17:04:40.584&-43:27:59.49&0.24&0.16&-51.8&-52.3&-51.0&1.30&1.08&-57.3&es - vis&s97 + [email protected] - 1612c&17:04:40.660&-43:28:00.54&0.47&0.75&-48.9&-50.5&-47.4&0.69&0.56&-57.3&es - vis&s97 + [email protected] - 1612d&17:04:40.667&-43:28:00.13&0.17&0.11&-46.0&-46.4&-45.5&1.34&1.09&-57.3&es - vis&s97 + [email protected] - 1667a&17:04:40.571&-43:27:59.59&0.24&0.11&-68.0&-68.4&-67.6&1.08&0.90&-57.3&es - vis&s97 + + g343.384 + 1.309 - 1612a&16:53:22.576&-41:48:29.90&2.56&3.49&-64.5&-65.5&-62.4&0.26&0.45&20.7&es - sev&s97 + g343.384 + 1.309 - 1612b&16:53:22.564&-41:48:29.30&0.32&0.29&-61.4&-61.9&-60.6&0.68&1.18&20.7&es - sev&s97 + g343.384 + 1.309 - 1612c&16:53:22.565&-41:48:30.01&0.95&0.55&-57.2&-57.8&-56.9&0.51&0.94&20.7&es - sev&s97 + g343.384 + 1.309 - 1612d&16:53:22.589&-41:48:29.72&1.31&1.05&-55.8&-56.5&-55.1&0.36&0.63&20.7&es - sev&s97 + g343.384 + 1.309 - 1612e&16:53:22.575&-41:48:29.83&0.63&1.60&-52.9&-54.6&-48.3&0.64&1.19&20.7&es - sev&s97 + g343.384 + 1.309 - 1612f&16:53:22.603&-41:48:29.53&0.84&1.09&-25.9&-29.7&-25.6&0.60&1.06&20.7&es - sev&s97 + g343.384 + 1.309 - 1612g&16:53:22.579&-41:48:29.91&2.91&1.97&-24.5&-25.1&-24.3&0.21&0.37&20.7&es - sev&s97 + g343.384 + 1.309 - 1612h&16:53:22.587&-41:48:29.78&4.15&3.42&-22.9&-23.8&-22.9&0.16&0.28&20.7&es - sev&s97 + g343.384 + 1.309 - 1612i&16:53:22.584&-41:48:29.75&18.58&25.47&-20.8&-22.4&-18.8&0.09&0.16&20.7&es - sev&s97 + g343.384 + 1.309 - 1665a&16:53:22.578&-41:48:29.58&0.47&1.47&-54.0&-56.8&-51.1&0.60&1.05&20.7&es - sev&s97 + g343.384 + 1.309 - 1667a&16:53:22.606&-41:48:29.71&0.56&0.95&-64.9&-66.2&-62.7&0.58&1.02&20.8&es - sev&s97 + + [email protected] - 1612a&17:02:06.074&-42:59:26.86&0.45&1.18&-9.7&-11.1&-5.7&1.33&1.02&-54.3&es - vis&n + [email protected] - 1612b&17:02:06.081&-42:59:26.90&0.43&0.95&16.1&12.5&17.0&1.09&0.84&-54.3&es - vis&n + + g343.543 + 0.345 - 1612a&16:57:57.369&-42:17:19.87&0.29&0.38&17.6&17.1&19.3&0.93&0.79&-50.2&u&n + + g343.562 + 0.144 - 1612a&16:58:52.268&-42:23:56.20&0.29&0.18&7.2&7.1&8.4&1.20&1.02&-50.2&es - vis&n + g343.562 + 0.144 - 1612b&16:58:52.278&-42:23:57.68&0.32&0.23&44.2&43.4&44.7&1.12&0.97&-50.2&es - vis&n + + [email protected] - 1612a&17:02:02.315&-42:50:07.58&0.32&0.22&-81.5&-82.4&-81.2&1.79&1.43&-54.3&u&n + + g343.721 + 1.003 - 1612a&16:55:47.413&-41:44:21.48&0.58&0.56&-20.9&-22.0&-19.7&1.02&1.91&-13.1&u&n + + [email protected] - 1665a&17:00:49.781&-42:26:08.71&0.38&0.20&-33.0&-33.3&-32.4&1.31&1.22&-48.1&sf - mmb&n + [email protected] - 1665b&17:00:50.009&-42:26:10.37&0.40&0.20&-29.0&-29.3&-28.5&1.10&1.02&-48.1&sf - mmb&n + + g343.929 + 0.125 - 1665a&17:00:10.938&-42:07:20.75&0.42&0.42&10.5&10.0&11.7&1.21&1.17&-46.2&sf - mmb&c98 + g343.929 + 0.125 - 1665b&17:00:10.928&-42:07:20.33&0.50&0.46&12.4&12.2&13.9&1.02&0.99&-46.2&sf - mmb&c98 + g343.929 + 0.125 - 1665c&17:00:10.873&-42:07:20.12&0.40&0.46&15.6&15.3&17.0&1.37&1.33&-46.2&sf - mmb&c98 + + g343.966 + 0.815 - 1667a&16:57:23.688&-41:39:57.48&0.31&0.19&-90.9&-91.8&-90.5&1.83&1.94&-43.5&u&n + g343.966 + 0.815 - 1667b&16:57:23.616&-41:39:56.79&0.34&0.27&-54.0&-55.3&-53.6&1.25&1.31&-43.5&u&n + + g344.101 + 1.794 - 1612a&16:53:46.983&-40:56:48.11&0.59&1.20&1.6&1.1&4.7&0.92&1.55&-28.7&es - sev&s97 + g344.101 + 1.794 - 1612b&16:53:46.994&-40:56:48.13&0.53&0.78&30.4&27.9&31.0&0.79&1.33&-28.7&es - sev&s97 + + [email protected] - 1612a&17:04:29.934&-42:24:30.49&0.76&1.17&-67.5&-69.1&-66.4&0.55&0.95&-16.9&u&n + + [email protected] - 1665a&17:04:07.800&-42:18:40.01&10.55&6.67&-31.3&-31.8&-28.3&0.18&0.31&-16.9&sf - mmb&c98 + [email protected] - 1665b&17:04:07.797&-42:18:40.48&0.58&0.29&-26.3&-27.0&-26.1&0.80&1.41&-16.9&sf - mmb&c98 + [email protected] - 1665c&17:04:07.811&-42:18:39.39&0.48&0.20&-23.2&-23.5&-22.6&0.80&1.41&-16.9&sf - mmb&c98 + [email protected] - 1665d&17:04:07.888&-42:18:39.17&0.52&0.34&-21.7&-22.2&-21.3&0.74&1.31&-16.9&sf - mmb&c98 + [email protected] - 1665e&17:04:07.894&-42:18:39.41&0.48&0.22&-18.8&-19.1&-18.2&0.95&1.73&-16.9&sf - mmb&c98 + [email protected] - 1667a&17:04:07.797&-42:18:39.84&22.64&15.38&-31.2&-32.4&-29.8&0.12&0.24&-16.9&sf - mmb&c98 + [email protected] - 1667b&17:04:07.829&-42:18:40.14&0.57&0.30&-29.1&-29.4&-28.5&0.64&1.17&-16.9&sf - mmb&c98 + [email protected] - 1667c&17:04:07.908&-42:18:40.31&0.51&0.26&-20.9&-21.5&-20.6&0.82&1.54&-16.9&sf - mmb&c98 + for each maser site , we present the spectrum or spectra , maser spots overlaid on the ir maps and their relative positional error ellipses for each maser spot ( e.g. figure [ figs1 ] , similar to figure 5 in @xcite ) . the upper panel shows the flux density ( constructed from stokes i ) plotted against the radial velocity with respect to the lsr for the unbinned spectrum , or spectra ( depending on the number of transitions detected ) , for each maser site .
the maser transition is usually labelled in the upper - right corner of the spectrum .
the shaded velocity range shows the emission channels for each maser spot , which is labelled with letters at the top .
the integrated flux density of each maser spot is the area under the spectral line curve in the shaded channels .
note that the shaded channels were inspected in the full data cube and only include the emission which is easily identified at a single location
. thus , some spectra have features which may appear as peaks and lie outside the shaded regions .
in such circumstances , we have identified them as either noise spikes or emission from nearby ( but unrelated ) strong masers ( for individual sources , see details in section [ individual ] ) .
therefore , the shaded channels should be considered as a guide to represent the velocity range over which we believe there is real emission originating from this maser site .
three spectra ( figures [ figs_bad1 ] , [ figs_bad2 ] and [ figs_bad3 ] ) have zero values in some velocity ranges due to flagged channels . only the 162 maser sites from our observations are shown here .
the remaining 53 magmo maser sites will be published later by the magmo team .
the bottom panel is a 6@xmath17 @xmath24 3@xmath17 glimpse three - color image , which plots band 1 for blue , 2 for green and 4 for red , with band wavelengths of 3.6 , 4.5 and 8.0 @xmath25 , respectively .
note that for masers in the galactic latitude range of @xmath26@xmath12 , the three - color image is made from wise data , based on band 1 for blue , 2 for green and 3 for red , with band wavelengths of 3.4 , 4.6 and 12.2 @xmath25 , respectively .
wise data are used in this galactic latitude range because the region is outside the glimpse survey area .
the image is centred on the maser site .
all maser spots that were detected within this field of view are shown .
1612mhz maser spots are presented with 3-pointed stars , 1665mhz maser spots with plus symbols , 1667mhz maser spots with cross symbols and 1720mhz maser spots with triangles . occasionally , there are two or more maser sites in the image , which can be used to see the relative positions of all maser sites in the same field of view .
a scale bar of 30@xmath21 is shown in the lower - left corner of this panel .
a white box located in the centre of the panel shows the field of view in the middle - left panel .
the middle - left panel presents the zoomed - in region around the maser site .
this region is a square 21.6@xmath21 on each side .
the background is the same three - color image as the bottom panel .
the masers at different frequencies are shown as different symbols as described above .
this figure only shows maser spots associated with single maser site .
thus , when comparing with the bottom panel , we can identify which maser spots are associated with this maser site .
a scale bar of length 1@xmath21 is shown in the lower - left corner of this panel , indicating the absolute positional uncertainty of the maser sites . in the centre of this panel ,
there is a white box , which shows the region of the zoomed area presented in the middle - right panel .
the size of the middle - right panel is chosen such that all masers , together with the full extent of their error ellipses , will fit within the panel .
thus , different maser sites have different white box sizes in the middle - left panel .
the middle - right panel is a zoomed - in region showing the positions and relative error ellipses of all maser spots for this maser site .
each maser spot is represented both with a colored ellipse and a colored symbol , with different symbols for different transitions as described above .
the symbols are filled with colors and have black borders in order to be seen easily .
the position of the ellipse is the fitted position of that maser spot . the major axis , the minor axis and the position angle of the ellipse represents the relative positional uncertainty of the maser spot , which is obtained with the miriadtask * imfit * in section [ observation ] . at the bottom of this panel
, there is a velocity color bar .
the color of the ellipses and symbols shows the peak velocity of the maser spot .
a scale bar is shown in the lower - left corner of this panel .
note that there is no absolute coordinate presented in this panel because the absolute position uncertainty is only about 1@xmath21 .
the relative offsets of maser spot positions shown in this panel are typically on much smaller scales .
a blind survey of oh masers in four ground - state transitions over a large proportion of galactic plane allows us to study their associations and the proportion of different associations in an unbiased way . even though splash is an untargeted survey ,
it is still sensitivity limited .
this factor may affect the detection statistics of evolved star and star formation masers , but it is still important to classify the masers according to which kind of astrophysical objects they are associated with .
we can identify some maser sites with reliable tracers of star formation , such as class ii methanol maser sites ( @xcite ) .
the double - horned profile in the 1612mhz transition , combined with a bright ir stellar - like source in the glimpse or wise map as in figure [ figs1 ] , is a reliable indicator of an evolved star .
however , for other maser sites , there may not be very clear evidence about their associations .
therefore , we adopt a variety of methods to identify each maser site and assign them to different categories , i.e. an evolved star site ( including pne ) , a star formation site , a snr site or an unknown site .
note that although we expect the identifications to be correct for each maser site , it is possible that a small number of our classifications are incorrect . in table
[ maintable ] , the last column is our identification for each maser site and reasons for this identification .
the process of identifying the associations is in the following steps : 1 .
an oh maser site was classified as a star formation site if its position is close to a class ii methanol maser site .
we compared our oh masers to the methanol multibeam survey detections for this region ( mmb ; @xcite ) in galactic coordinates .
the absolute positional accuracy of the mmb survey is about 0.4@xmath21 . when there is a 0.001@xmath12difference in one coordinate ( galactic longitude or galactic latitude ) , it is equivalent to an angular separation of 3.6@xmath21 .
we chose 3.6@xmath21 as the maximum separation for ease of use .
we find oh maser sites and methanol maser sites are either closely associated ( the separation is less than 3.6@xmath21 ) or well separated ( the separation is larger than 12.9@xmath21 ) .
remaining maser sites were correlated with the h@xmath27o southern galactic plane survey ( hops ; @xcite ) identifications , which classified 22ghz water masers carefully and reliably .
the absolute positional accuracy of hops is about 1 .
the largest distance for an association is chosen to be 3.6@xmath21 , which is the same value used in mmb data .
the comparison result also shows that oh maser sites and water maser sites are either closely associated ( the separation is smaller than 3.6@xmath21 ) or well separated ( the separation is larger than 12.9@xmath21 ) .
remaining maser sites were compared with the red msx source ( rms ; @xcite ) catalogue .
an association threshold of 9.5@xmath21 was adopted , corresponding to the astrometry limits of the rms survey ( @xcite ) .
the oh maser site was assigned to the rms identification if the separation was smaller than this value .
the remaining sources were identified in the literature with simbad .
some sources with rms identifications ( e.g. pn and snr ) were also searched in simbad to confirm the identification or revise the identification .
the associations could also be identified according to their ir properties or ir properties combined with spectral features .
the ir images of all maser sites were checked to see if their identifications from the previous steps were correct .
we also tried to classify some sources which did not have a clear identification based on steps 1@xmath144 .
when masers have double - horned profiles in the 1612mhz transition and are associated with bright ir stellar - like sources in the glimpse or wise images , we classified them as evolved star oh masers .
if masers were associated with extended green objects ( egos ; @xcite ) in the ir image , they were categorised as star formation
oh masers , because egos trace the energetic shocked gas in the interstellar medium , which is the signature of star formation ( e.g. @xcite ) . with this step
, we could identify many evolved star oh masers and check identifications from above steps , e.g. hops and rms , to obtain a correct identification . 6 .
any oh masers which could not be reliably identified using the preceding steps were assigned as unknown sites . from both our dataset and the magmo dataset
, we identify 122 oh maser sites ( 57 percent ) associated with the circumstellar envelopes of evolved stars ( one of which is a pn ) , 64 oh maser sites ( 30 percent ) associated with sfrs , two maser sites towards snrs and 27 unknown maser sites ( 13 percent ) .
compared to the literature ( e.g. @xcite ; @xcite ) , about half of the evolved star sites ( 60/122 ) are new detections , 39 percent of the star formation sites ( 25/64 ) are new detections and almost all of the unknown maser sites ( 26/27 ) are new detections .
the identification result above demonstrates that the majority of oh masers originate from the envelopes of evolved stars .
we will discuss the transition overlap of the evolved star category and star formation category separately in more detail in section [ overlap ] .
the only pn site ( [email protected] ; in the evolved star category ) has 1612 , 1667 and 1720mhz oh masers and shows variability in the 1720mhz transition at two epochs ( see @xcite for more details ) .
the two snr sites ( [email protected] , @xcite , from magmo ; [email protected] , @xcite , from our dataset ) show single - peak 1720mhz oh masers , which trace the interaction of the snr and surrounding molecular clouds .
the unknown oh maser sites usually have one maser spot at 1612mhz and are associated with a bright ir stellar - like source in the glimpse or wise image ( 20/27 ) .
these oh masers may come from the circumstellar envelopes of evolved stars where only one maser spot at 1612mhz can be seen .
in order to classify maser spots into maser sites , and thereby determine the upper limit of oh maser site sizes , we have studied the angular separation between each maser spot and its nearest neighbour .
figure [ dis_spot ] shows that the nearest neighbours are generally located within 2@xmath21 of one another .
this figure further shows that few spots are separated by values in the range between 2@xmath21 and 4@xmath21 ( a total of five ) .
given that figure [ dis_spot ] plots the separation to the nearest spot ( and therefore could be considered a lower limit to site size ) we have adopted a site size upper limit of 4@xmath21 .
we note that this is the same upper limit adopted for the water masers detected in the hops survey ( @xcite ) . after establishing an upper limit , the true size of the maser sites
can be determined by taking the maximum separation between oh maser spots within that site . in a few cases , we found maser spots spread over more than 4@xmath21 , but upon inspection were clearly two clusters of spots that were well separated and
could therefore be easily classified into two separate sites .
figure [ size_all ] shows the size of 171
oh maser sites ( red ) and indicates that all of our oh maser sites are smaller than 3.1@xmath21 , and only eight sources show spreads of equal to or greater than 2@xmath21 ( equating to 95 percent of sources ) .
we therefore conclude that the majority of oh masers show extents of less than 2@xmath21 ( equivalent to a linear size of about 0.05 pc at a distance of 5 kpc ) which is consistent with previous investigations ( e.g. @xcite )
. for comparison , figure [ size_all ] also shows the site sizes of the 470 22ghz water maser sites ( blue ) from hops .
this shows that the distribution of water maser sizes peaks at a smaller size than oh maser sites , but has a longer tail with more water maser sites showing larger sizes . the result of a kolmogorov - smirnov ( k - s ) test shows that there is a @xmath28 percent chance these two distributions are drawn from the same underlying population .
the significant difference in distributions of oh maser sites compared to water maser sites can be accounted for by the fact that very small water maser sites are dominated by evolved star masers ( @xcite ) , which are found in the inner envelopes compared to oh masers that are associated with the outer envelopes ( @xcite ) , causing there to be comparatively few very small oh masers . at the other end of the scale ,
there are more large water maser sites which are dominated by star formation sources ( @xcite ) that often trace larger - scale outflows emanating from their powering sources , which may cause the wider range of water maser site sizes .
figure [ size_es_sf ] shows oh maser site sizes for sources associated with evolved stars ( 117 maser sites ) and star formation ( 51 maser sites ) , respectively .
it can be seen from the plot that evolved star sites are all smaller than 1.6@xmath21 , while the 12 oh maser sites that show greater extents are all associated with star formation , and , further , that the distribution of the evolved star sites peaks at smaller sizes than the star formation sites .
the significance of these differences is confirmed with a k - s test of the two distributions , showing that there is a @xmath29 percent chance they are drawn from the same underlying population .
it is perhaps not surprising that evolved star oh maser sites generally tend to be smaller because they are located in a circumstellar shell , typically on size scales of 80 au ( @xcite ) , whereas star formation oh maser sites are typically distributed over 3000 au ( @xcite ) .
@xcite also concluded that 22ghz water maser sites associated with evolved stars have smaller angular sizes than those sites associated with star formation .
percent chance these two distributions are drawn from the same underlying population.,scaledwidth=40.0% ] many of the oh maser sites show emission from only a single transition , however , some sites show two or three transitions .
the upper panel of figure [ vennfig ] shows the transition overlap of evolved star oh maser sites .
of the 122 evolved star maser sites , 101 evolved star sites ( 83 percent ) only exhibit 1612mhz emission . among
these 101 evolved star sites , 97 sites have double - horned spectral profiles in their spectra , three have a single peak ( [email protected] , [email protected] from our dataset and [email protected] from magmo ) and one has three peaks ( g342.004 + 0.251 from our dataset ) .
two evolved star sites ( [email protected] from our dataset and [email protected] from magmo ) only show a single peak in the 1667mhz transition .
one evolved star site only has 1665 and 1667mhz masers ( [email protected] from our dataset ) with two maser spots at each transition .
one pn site ( [email protected] ; @xcite ) exhibits the only 1720mhz oh maser among evolved stars , and it is accompanied by maser emission at 1612 and 1667mhz .
1612 and 1667mhz oh masers show the largest overlap : 84 percent of 1667mhz oh masers ( 16/19 ) have a 1612mhz counterpart .
main - line transitions have the second largest overlap with 75 percent of 1665mhz oh masers ( 6/8 ) associated with 1667mhz oh masers .
the bottom panel of figure [ vennfig ] is venn diagram illustrating the overlap between the four transitions in star formation regions .
70 percent of 1612mhz oh masers ( 7/10 ) are solitary , i.e. not associated with other ground - state oh transitions .
33 percent of 1665mhz oh masers ( 17/52 ) are solitary .
main - line oh masers show the largest overlap : 32 sites exhibit both 1665 and 1667mhz oh masers .
62 percent of 1665mhz oh masers ( 32/52 ) have a 1667mhz counterpart , which is higher than the value reported in ( * ? ? ?
* percent in the splash pilot region ) .
@xcite detected 38 star formation oh masers in the splash pilot region and we detect 37 of these .
the undetected oh maser ( [email protected] ) only showed weak ( @xmath80.4 jy ) emission at 1665mhz in @xcite and is not detected here likely due to variability . for the re - detections of @xcite ,
72 percent of 1665mhz oh masers ( 26/36 ) have a 1667mhz counterpart .
this is partly caused by the broader association radius we used ( 3.1@xmath21 versus 1@xmath21 adopted by caswell , 3.1@xmath21 is the largest size of star formation oh maser sites from our results ) .
moreover , it is possible that the @xcite data have some biases in their 1667mhz observations .
since his observations were focused on the 1665mhz line , the limited bandwidth of the observations meant that the 1667mhz line fell outside the range of the observations in some cases . to combat this , @xcite adopted previous observations where available , meaning that source variability may influence these associations ( since the observations of the two lines would necessarily be separated in time ) . with a larger number of masers from the full splash region ( that have been observed simultaneously ) ,
we aim to accurately determine the overlap for both evolved star and star formation maser sites . for 118 evolved star sites ( excluding the pn site ) with 1612mhz emission , we classified them according to their integrated flux densities of blue- and red - shifted components ( i@xmath30 and i@xmath31 ) .
since blue- and red - shifted components come from two sides of the circumstellar envelope , we can estimate the number of photons emitting from each side using the integrated flux density .
if the ratio of i@xmath30 and i@xmath31 is between 0.5 and 2 , we classify the source as symmetric .
if the ratio of i@xmath30 and i@xmath31 is larger than 2 or smaller than 0.5 , we classify them as asymmetric . three sources ( [email protected] from our dataset , [email protected] and [email protected] from magmo ) are not included because the velocity range of our observations or the magmo observations does not cover the expected velocity of either the blue- or red - shifted component .
two sources ( [email protected] and [email protected] from our dataset ) are classified as extremely asymmetric sources because both of them only have one spectral component in the spectrum .
-[22 ] color distributions of 83 symmetric ( blue ) and 29 asymmetric ( red ) evolved star sites showing 1612mhz transition .
a k - s test shows that there is a 10 percent chance that these two distributions are from the same underlying population.,scaledwidth=40.0% ] of 86 symmetric ( blue ) and 27 asymmetric ( red ) evolved star sites showing 1612mhz transition .
a k - s test shows that there is a 34 percent chance that these two distributions are from the same underlying population.,scaledwidth=40.0% ] there are 86 ( out of 118 , 73 percent ) symmetric sources and 29 ( out of 118 , 25 percent ) asymmetric sources .
we compared the mid - ir properties of these two samples , because emission from circumstellar dust can become the dominant source of emission over radiation from the photosphere for wavelengths longer than 5 @xmath25(@xcite ) .
83 ( 97 percent ) symmetric sources have wise point source counterparts and 29 ( 100 percent ) asymmetric sources have wise point source counterparts .
we find that the [ 12]-[22 ] color ( 12 @xmath25and 22 @xmath25 ) of these two samples has different distributions , shown in figure [ hist_1222 ] .
this is confirmed with a k - s test showing that there is a 10 percent chance that these two distributions are from the same underlying population .
symmetric sources show redder colors than asymmetric sources , probably indicating that their envelopes are thicker and thus we see colder temperatures .
a thicker envelope could indicate that the symmetric sources are at an earlier evolutionary stage than asymmetric ones , and/or that the symmetric sources have more massive precursors and thus , more massive envelopes .
regarding stellar mass , @xcite found oh / ir stars with higher main - sequence masses ( @xmath32 kms@xmath16 , @xmath33 is the velocity separation between most blue- and red - shifted maser components in the 1612mhz transition ) are clearly redder than oh / ir stars with lower main - sequence masses ( @xmath34 kms@xmath16 ) .
we also studied the velocity separation of symmetric sources and asymmetric sources , shown in figure [ vel_sep ] .
a k - s test shows that there is a 34 percent chance that these two distributions are from the same underlying population , indicating that there is not significant difference in velocity separation ( and thus , stellar mass ) between symmetric and asymmetric sources . in principle , these results seem to suggest that age is a more determinant factor in the color difference between the symmetric and asymmetric oh sources .
however , there could be some biases that need to be considered , such as the particular ratio between i@xmath30 and i@xmath31 we chose to separate these two samples .
moreover , it is possible that the detection of an underlying difference in main - sequence masses requires a sample with a larger size .
further investigation will be carried out with the full splash sample .
we compared star formation oh maser sites in the splash pilot region to 6.7ghz methanol maser sites from the mmb survey and 22ghz water maser sites from hops .
we chose the overlap region of these three surveys , i.e. between galactic longitudes of @xmath0 and @xmath1 and galactic latitudes of @xmath35 and @xmath36 .
there are 52 oh maser sites , 112 methanol maser sites and 67 water maser sites in this region .
the association of these three species of masers is shown in figure [ vennsf ] .
we used the same criteria as described in section [ identification ] to identify whether two different maser species are associated .
oh maser sites have the largest overlap with methanol maser sites ( 38/52 ; 73 percent ) . as for the oh masers , water masers
also have the largest overlap with methanol masers ( 38/67 ; 57 percent ) .
half of methanol maser sites ( 56/112 ) are solitary , i.e. not associated with any oh and water masers . the fraction of solitary methanol maser sites is higher than solitary oh maser sites ( 21 percent ) and solitary water maser sites ( 39 percent ) . a sensitive 22ghz water maser survey towards the mmb masers found that @xmath848 percent of 6.7ghz methanol masers have a 22ghz water maser counterpart ( @xcite ; @xcite ) , which is higher than our result ( 38/112 , 34 percent ) .
moreover , @xcite made a targeted search for 22ghz water masers toward star formation oh maser sites and concluded about 79 percent of oh masers sites show coincident water maser emission .
this value is much higher than our result , which is about 44 percent ( 23/56 ) .
the @xcite and @xcite water maser surveys have comparable sensitivities with the typical rms noise about 0.1 jy , whereas hops has a typical rms noise of around 1 jy .
therefore , association results between oh / methanol and water masers will be biased by the lower sensitivity of hops .
however , it still provides useful insights on the association between oh and 6.7ghz methanol masers .
we will investigate this overlap with the larger dataset from the full splash survey region . and @xmath1 and galactic latitudes of @xmath35 and @xmath36.,scaledwidth=40.0% ] in the splash pilot region , 64 maser sites are classified as star formation sites .
we also checked whether these oh maser sites are associated with continuum sources at 1.7ghz .
we find six of them ( 6/64 ; @xmath89 percent ) are associated with continuum sources .
note that we only used the 2mhz bandwidth from 3 concatenated zoom bands for 1720mhz to check for continuum sources .
a previous study of oh masers in star formation regions shows that @xmath838 percent of oh maser sites have compact 3 cm continuum sources ( at 8.2ghz and 9.2ghz ) within 2 ( @xcite ) . for 6.7ghz methanol masers , @xmath820 percent of methanol masers
are associated with continuum sources ( @xcite ) .
the low fraction of continuum sources we observe is likely to be due to the low frequency ( 1.7ghz ) and limited sensitivity of the continuum observations .
since the turnover frequency of ultracompact hiiregions is typically @xmath37ghz ( @xcite ) , ultracompact hiiregions are likely to be optically thick at 1.7ghz and the flux density drops off quickly ( @xmath38 , where @xmath39 is the flux density and @xmath40 is the frequency ) .
table [ nondets ] lists the positions at which we did not detect any maser emission in the atca observations . from their parkes
spectra , we find all of them are weak ( weaker than @xmath80.4 jy ) .
nine of them showed very weak features in the parkes spectra that we considered as possible masers .
but all of these nine could also be consistent with noise . since we make no detections of these with the atca data , we consider the original parkes identifications as spurious .
one of them ( [email protected] ) appears to be diffuse emission , showing line ratios expected for quasi - thermal emission .
the remaining eleven positions appear to be real detections with eight showing emission from the 1612mhz transition and three exhibiting 1665mhz transition . among these eleven sources , six of them
have two peaks and five of them only show one peak in the parkes spectra .
the approximate time difference between the parkes observations and the current atca observations is about 1.5 years .
previous work showed that ground - state oh masers could be variable on short ( minutes - hours ) and long ( days - years ) time scales ( e.g. @xcite ) .
given the higher sensitivity of the atca observations , when compared to the previous parkes observations , we surmise that these masers may be variable and can not be detected because of low flux density during the atca observations .
.list of positions which exhibit maser emission in @xcite but were not detected in our observations . [ cols="^,^,^ " , ] notes : s may be spurious detections ; v may be variable ; d diffuse emission , not a maser . the non - detection in half of these regions ( 11/21 )
is thought to be due to intrinsic variability , making the masers undetectable in our observations .
* g333.942 + 0.387 , [email protected] , + [email protected] , g337.877 + 0.820 , + g338.902 + 0.015 , [email protected] , + g342.222 + 0.041 , [email protected] , + [email protected] , [email protected] , + g343.543 + 0.345 , [email protected] + and [email protected] . *
these maser sites only have one oh maser spot in the 1612mhz transition . in the glimpse three - color images , they are associated with bright ir stellar - like sources , as shown in figure [ g333.942 ] .
we studied their ir properties by finding their counterparts in the glimpse point source catalogue . for the sources with ir point source counterparts
, we found that the magnitude of the 4.5 @xmath25band is brighter than 7.8 , which indicates that these sources are `` obscured '' agb star candidates with very high mass - loss rates according to @xcite .
however , in the absence of other confirmation of their nature , we identify them as `` unknown sites '' , although they very possibly originate from the circumstellar envelope of evolved stars .
* g334.132 + 0.384 .
* this maser site only has one 1720mhz maser spot .
there is no clear identification for this position in the literature , thus it is identified as an unknown site . from the glimpse three - color image , we find it is located in an extended emission region .
it possibly belongs to the variety noted by @xcite , i.e. associated with star formation without main - line emission .
this source is resolved on 6 km baselines , but clearly detected on shorter baselines .
it was detected with splash parkes observations but only in the 1720mhz line .
we did not detect any emission or absorption towards the other three lines with the atca .
thus , we concluded it is likely to be a maser detection
. * g334.577 + 1.958 , g337.782 + 1.394 , + g338.513 + 1.503 and g340.091 + 1.458 . *
these maser sites only have one maser spot at 1612mhz .
there are no clear identifications for them in the literature , thus their associations are unknown . in the wise three - color images , they are located in the same position as bright ir stellar - like objects .
* g335.136 + 0.196 .
* this unknown maser site has two 1612mhz maser spots .
there is no clear identification for this position in the literature . in the glimpse three - color image , it is associated with a bright stellar - like source . * g335.832 + 1.434 .
* this maser site is an oh / ir star site , which belongs to the evolved star category .
figure [ g335.832 ] shows the line - widths of the 1665mhz maser spots are quite broad : about 10 kms@xmath16for each spot .
but since this maser is detected on even the longest baselines and is not detected in the other lines , it is unlikely to be diffuse . in the wise three - color image , it is associated with a very red star ( bright at 12 @xmath25 ) .
* [email protected] .
* this maser site is an oh / ir star site with 26 maser spots at 1612 , 1665 and 1667mhz , shown in figure [ g336.075 ] .
this is the richest maser site in the splash pilot region .
the 1612mhz maser spots are stronger in the velocity range @xmath1462 kms@xmath16to @xmath1448 kms@xmath16 , while 1665 and 1667mhz maser spots are stronger at the blue - shifted velocity range of @xmath1492 kms@xmath16to @xmath1488 kms@xmath16 .
this result indicates that 1665 and 1667mhz masers favour a similar physical environment .
the glimpse three - color image appears saturated .
* [email protected] . *
the unassociated emission in the 1612mhz spectrum is from a nearby source , [email protected] .
* [email protected] .
* this maser site is a pn , which also has 22ghz water masers at the velocity of @xmath4141.8 kms@xmath16and @xmath4143.9 kms@xmath16@xcite and radio continuum emission @xcite .
as shown in figure [ g336.644 ] , we detected one 1612mhz oh maser spot at @xmath1445 kms@xmath16 , one 1667mhz oh maser spot at @xmath1445 kms@xmath16and five 1720mhz maser spots , which have the double - horned profile with the central velocity @xmath4143 kms@xmath16 .
this is only the second known pn showing 1720mhz oh masers after k 3@xmath1435 and the only evolved stellar object with 1720mhz oh masers as the strongest transition , compared to the other three transitions of ground - state oh . we also observed it with the atca about one and half years later and found variability in the 1720mhz transition .
the magnetic fields are also measured from the zeeman splitting of the 1720mhz maser spots , which suggests that the 1720mhz masers are formed in a magnetised environment .
further details of this source can be found in @xcite .
* [email protected] .
* this maser site is classified as an unknown site . as shown in figure [ g337.001 ] , it has two maser spots at velocity of @xmath41122 kms@xmath16 in the 1612 and 1665mhz transitions . in the glimpse three - color image ,
these two masers are associated with a very bright ir stellar - like source , thus it may originate from an evolved star site .
* [email protected] .
* this maser site is an evolved star site , which appears to have evolved into the fast wind stage @xcite .
this maser has been observed as a single spectral feature by @xcite , which is consistent with our observations . in the wise three - color image ,
the maser site is bright at 12 @xmath25(very red ) .
* g337.241 + 0.146 . *
the unassociated emission in the 1612mhz spectrum is from a nearby source , [email protected] . *
* this maser site is classified as a snr site @xcite .
it contains one 1720mhz oh maser spot , which traces the interaction between the snr with surrounding molecular clouds .
there are no obvious features in the glimpse three - color image .
* g337.860 + 0.271 .
* this maser site is an evolved star site , exhibiting 15 maser spots at 1612 , 1665 and 1667mhz transitions , shown in figure [ g337.860 ] .
the velocity of the 1665mhz maser spot is close to the central velocity of the double - horned profile in the 1612mhz transition , while the 1667mhz maser spots spread a broad velocity range across the 1612mhz velocity range . in the glimpse three - color image ,
these masers are associated with a very bright ir stellar - like source .
the unassociated emission in the 1612mhz spectrum is from g337.719 + 0.350 and the unassociated emission in the 1665 and 1667mhz spectra is from g337.997 + 0.136 ( the magmo dataset ) .
* g338.660 + 1.290 .
* this maser site is identified as an evolved star site .
it has 1612 and 1665mhz oh maser spots , as shown in figure [ g338.660 ] .
the line - width of these maser spots is quite broad with @xmath820 kms@xmath16for the 1612mhz transition and @xmath816 kms@xmath16for the 1665mhz transition .
these two masers are detected even on the longest baselines and are not detected in the other lines , thus they are unlikely to be diffuse oh emission . in the wise three - color image ,
the star is very red ( bright at 12 @xmath25 ) . *
[email protected] . *
the unassociated emission in the 1612mhz spectrum is from a nearby source , [email protected] . *
g339.282 + 0.136 .
* this maser site only exhibits one maser spot at the 1665mhz transition .
it is also associated with 6.7ghz methanol masers and is identified as a star formation site .
the line - width of this maser spot is @xmath84 kms@xmath16 . in the glimpse three - color image , it is associated with a very faint ego .
@xcite reported the variability of this 1665mhz maser , with the highest peak decreasing from 1.5 to 0.7 jy between 2004 and 2005 .
the 1667mhz oh maser they detected was listed in our maser site g339.294 + 0.139 .
* g339.294 + 0.139 .
* this maser site only has one maser spot at 1667mhz transition .
it is also associated with 6.7ghz methanol masers and is thus identified as a star formation site .
the line - width of this maser spot is quite narrow , @xmath80.5 kms@xmath16 . in the glimpse
three - color image , it is associated with an ego .
@xcite detected high linear polarization of the 1667mhz maser at @xmath1473.5 kms@xmath16(@xmath42 percent ) .
* [email protected] . *
the unassociated emission in the 1612mhz spectrum is from a nearby source , [email protected] .
* g340.043 - 0.092 . *
this source is an evolved star . from @xcite observations
, it has the typical double - horned profile at 1612mhz , which spreads from @xmath1446.8 to @xmath1411.8 kms@xmath16 . in our observation
, we did not detect the 1612mhz oh maser and only detected one red - shifted maser spot at 1667mhz , which is @xmath81.5 kms@xmath16width .
* [email protected] .
* this maser site is identified as an unknown site .
it only contains one 1612mhz oh maser with line - width of @xmath81 kms@xmath16 . in the glimpse three - color image , it is located in the 12 @xmath25extended emission ( red background ) and is not associated with any stellar - like objects .
it might be associated with star formation .
* [email protected] .
* this maser site is associated with a variable star of mira cet type , which is well studied ( @xcite ) .
it only exhibits 1665 and 1667mhz oh masers . in the glimpse three - color image
, the pixels are saturated .
* [email protected] .
* this maser site is associated with an evolved star .
it not only shows the typical double - horned profile at 1612mhz , but also a maser spot between the velocity of the center star and the red - shifted maser spot ( named as the middle maser spot ) , shown in figure [ g341.102 ] .
the five maser spots in 1665 and 1667mhz transitions locate between the middle and the red - shifted 1612mhz maser spots .
* g341.681 + 0.264 .
* this maser site is identified as an unknown site .
it only exhibits a very broad 1612mhz oh maser spot with line - width of @xmath812 kms@xmath16 , shown in figure [ g341.681 ] .
this emission is detected on even the longest baselines and is not detected in the other lines , thus it is unlikely to be diffuse oh emission . in the glimpse three - color image , it is associated with a faint stellar - like object , which may be an ego or an evolved star . * g342.004 + 0.251 . *
this maser site is identified as an evolved star site . in hops , water masers
were also observed towards this site with the velocity range from @xmath1445 kms@xmath16to @xmath1420 kms@xmath16 . in our observations , there are three peaks in the 1612mhz spectrum and the velocity range is from @xmath1453 kms@xmath16to @xmath1432 kms@xmath16 .
the blue - shifted component of the 1612mhz oh maser is more blue - shifted than the water masers and its red - shifted component is absent
. * g342.369 + 0.140 . *
this maser site is a star formation site , which also exhibits 6.7ghz methanol masers and 22ghz water masers .
it only contains two 1612mhz oh maser spots and is associated with very faint egos in the glimpse three - color image .
* [email protected] . *
the unassociated emission in the 1612mhz spectrum is from a nearby source , [email protected] .
* [email protected] .
* this maser site exhibits one strong maser spot at 1612mhz . in the glimpse three - color image ,
this source is associated with extended dark red emission ( 12 @xmath25 ) and we classify it as star formation oh maser site .
the glimpse image also shows the nearby star formation oh maser site associated with egos .
* g343.721 + 1.003 .
* this sources has similar properties as [email protected] in the 1612mhz spectrum and the glimpse three - color map .
it might be associated with star formation .
* g343.929 + 0.125 .
* this maser site is a star formation site , which also has 6.7ghz methanol masers .
it has three weak maser spots in the 1665mhz transition and is associated with egos in the glimpse three - color map .
@xcite reported the variability of the strong 1665mhz left - hand circular polarization component at 12 kms@xmath16and the absence of the 1667mhz maser .
* g343.966 + 0.815 .
* this maser site is identified as an unknown maser site .
it shows a double - horned profile in the 1667mhz transition and is associated with an ir stellar - like source in the glimpse three - color map .
it may be associated with an evolved star .
in this paper , we present accurate positions for ground - state oh masers in the splash pilot region , between galactic longitudes of @xmath0 and @xmath1 and galactic latitudes of @xmath2 and @xmath3 .
we also identify the associated astrophysical objects for these maser sites .
we obtain a total of 215 maser sites , which exhibit maser emission at one , two or three transitions .
more than half of these 215 maser sites ( 111/215 ) are new detections . of these 215 maser sites ,
154 sites show maser emission at 1612mhz , 61 and 57 sites emit maser emission at 1665mhz and 1667mhz , and 9 sites exhibit the rare 1720mhz oh masers .
we identify 122 maser sites associated with evolved stars .
these maser sites usually have the characteristic double - horned profile at 1612mhz , sometimes accompanied by 1665 and/or 1667mhz oh masers . in the glimpse or
wise three - color images , these maser sites are associated with bright ir stellar - like sources .
one pn site [email protected] from the evolved star category is detected with oh masers at 1612 , 1667 and 1720mhz and may represent a very peculiar phase of the pn evolution .
64 maser sites are classified as being associated with star formation .
these sites commonly have several strong maser spots at main - line transitions and occasionally also exhibit 1612 or 1720mhz oh masers . in the ir images
, they tend to be associated with egos , which are signatures of star formation .
two 1720mhz maser sites are associated with snrs and trace the interaction of the snr with the surrounding giant molecular cloud .
27 maser sites are categorised as unknown because of the lack of information from literature and ir images .
we find the size of most oh maser sites ( 98 percent ) are smaller than 2.5@xmath21 according to their accurate positions .
we also find the angular size of oh maser sites associated with evolved stars are smaller than that of star formation sites , which is consistent with 22ghz water masers in hops . for evolved star sites , oh masers in the 1612mhz transition
are dominant ( 98 percent ) . for star formation sites ,
main - line transitions have the largest overlap , i.e. 62 percent of 1665mhz oh masers are associated with a 1667mhz oh maser .
we classify evolved star sites showing the 1612mhz transition according to integrated flux densities of blue- and red - shifted components .
we find that symmetric sources are redder than asymmetric sources in the wise [ 12]-[22 ] colors
. this may be mainly due to a younger age of symmetric sources , compared to asymmetric sources .
we study associations of star formation oh maser sites with 6.7ghz methanol masers from the mmb survey and 22ghz water masers from hops .
we find that oh masers have the largest overlap with methanol masers ( 73 percent ) , which may be due to the high sensitivity of the methanol maser survey .
we also find six of 64 star formation oh maser sites in the splash pilot region are associated with continuum sources at 1.7ghz , which is lower than the fraction at 9ghz .
this lower value is likely due to the frequency we observe , at which hiiregions may be optically thick , or may be caused by the low sensitivity of our observations .
we did not detect any maser emission in 21 fields . from their parkes
spectra , we find they tend to have simpler and weaker profiles . we conclude that about half of non - detections are the result of intrinsic variability .
the australia telescope compact array is part of the australia telescope which is funded by the commonwealth of australia for operation as a national facility managed by csiro .
this research has made use of : nasa s astrophysics data system abstract service ; and the simbad data base , operated at cds , strasbourg , france .
this work is based in part on observations made with the spitzer space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with national aeronautics and space administration ( nasa ) .
this publication also makes use of data products from the wide - field infrared survey explorer , which is a joint project of the university of california institute of technology , funded by nasa .
h.q . would thank the china scholarship council ( csc ) support .
slb is a recipient of a loral - unesco for women in science fellowship .
jfg is partially supported by mineco ( spain ) grant aya2014 - 57369-c3 - 3 ( co - funded by feder ) .
|
we report on high spatial resolution observations , using the australia telescope compact array ( atca ) , of ground - state oh masers .
these observations were carried out toward 196 pointing centres previously identified in the southern parkes large - area survey in hydroxyl ( splash ) pilot region , between galactic longitudes of @xmath0 and @xmath1 and galactic latitudes of @xmath2 and @xmath3 .
supplementing our data with data from the magmo ( mapping the galactic magnetic field through oh masers ) survey , we find maser emission towards 175 of the 196 target fields .
we conclude that about half of the 21 non - detections were due to intrinsic variability . due to the superior sensitivity of the follow - up atca observations , and the ability to resolve nearby sources into separate sites
, we have identified 215 oh maser sites towards the 175 fields with detections . among these 215 oh maser sites ,
111 are new detections . after comparing the positions of these 215 maser sites to the literature
, we identify 122 ( 57 per cent ) sites associated with evolved stars ( one of which is a planetary nebula ) , 64 ( 30 per cent ) with star formation , two sites with supernova remnants and 27 ( 13 per cent ) of unknown origin .
the infrared colors of evolved star sites with symmetric maser profiles tend to be redder than those of evolved star sites with asymmetric maser profiles , which may indicate that symmetric sources are generally at an earlier evolutionary stage .
|
atomic physics tests of the standard model @xcite play a very special role because of the small momentum transfers involved .
comparisons between their results and high energy data are highly sensitive to radiative corrections and thus to extensions of the standard model @xcite . with the percent precision reached in the cs experiments described in @xcite , the effect of radiative corrections
is of the order of the experimental accuracy .
if a system is found for which a 0.1% accuracy can be reached the experimental results would allow most interesting and far reaching conclusions ( see e.g. @xcite ) . for the atoms and experimental setups studied so far this seems unluckily to be out of reach , which motivates the search for significantly different alternatives .
the possibility we want to discuss is the use of highly charged heavy ions , which can be produced and stored in great variety at , e.g. , gesellschaft fr schwerionenforschung in darmstadt , germany .
we discussed already some time ago the prospects for inducing a two - photon transition in helium - like uranium @xcite . in this contribution
we extend our studies to systems with up to five electrons and we adopt the ingenious ideas proposed by botz , bru , and nachtmann @xcite especially suited for the investigation of parity - violating effects in storage rings .
the starting point for all such experiments is that , due to the parity - violating exchange of neutral @xmath0 bosons between nucleus and electrons , every electron state is mixed with states of opposite parity . in first order perturbation theory the coefficient @xmath1 of this admixture
is given by @xmath2 where @xmath3 denotes fermi s constant , @xmath4 the weinberg angle , @xmath5 the neutron number , @xmath0 the proton number , and @xmath6 the nuclear density normalized to @xmath0 . from this formula
we see why heavy ions with few electrons left in inner shells are good candidates for investigating parity non - conservation effects : the admixture coefficient @xmath1 is very large ( typically orders of magnitude larger than for usual , neutral atoms ) due to the big overlap between the nucleus and the electron states .
the other factor that can make @xmath1 large is the energy difference between the two mixing electronic states @xmath7 and @xmath8 that ought to be very small .
therefore we are especially interested in level crossings of electron states with the same spin but opposite parity .
it was pointed out in @xcite that equation ( [ eins1 ] ) has to be modified by radiative corrections , the weak charge @xmath9 included in ( [ eins1 ] ) changes according to @xmath10 here @xmath11 and @xmath12 are constants that arise from the radiative corrections mentioned above .
the crucial point is , that they depend on the masses of the particles involved in the radiative processes , especially the top quark and the higgs boson .
as it seems now that there is evidence for the top quark to exist , it should be from a theoretical point of view possible to determine from @xmath11 and @xmath12 the value of the mass of the higgs boson that makes the standard model renormalizable , thus giving important guidance to identify this particle in high energy experiments . in section 2
, we will discuss uranium with two to five electrons as a model for other heavy ions reaching from gold to plutonium .
in section 3 we will discuss the possibility of level crossing in compound heavy ions and finally in section 4 we will investigate the possibility of polarization rotations in heavy ions .
as to an experiment with heavy ions with few inner shell electrons we have to give a criterion under which we can judge feasibility of such an experiment .
as such a criterion we should compare the @xmath1 values of the systems regarded here with the @xmath1 value of the helium - like uranium system discussed in @xcite , i.e. , @xmath13 , when taking the energy difference to @xmath14 ev .
even this relatively high value of @xmath1 left the proposed experiment beyond the scope of experimental feasibility for the set up discussed there .
we get a second restriction by the following consideration .
if , for example , the interesting electron states are excited during the stripping process of the ion in a stripping foil , then for any realistic experiment the experimental set up should be placed a little distance behind this foil , lets say one meter .
then the lifetime of these excited states should be long enough to survive this one meter of flight .
taking into account a time dilation factor of about 5 for an ion accelerated to 5 gev per nucleon , the lifetime should be larger than @xmath15 s. this would be an optimal value , but also a lifetime of @xmath16 s would probably do , corresponding to a distance of 10 cm .
we furthermore consider only the lowest lying electron states that offer a possibility for a parity - violation experiment . as
the parity admixture is proportional to the overlap of the electron states in question with the nucleus , this admixture should become very large for low lying states , if the energies are sufficiently degenerate .
we should state that these criterions do not rule out all imaginable experiments .
it could be possible for example to store ions in an ion trap and to generate the excited state by a laser beam , maybe by a laser that still has to be invented or that will be available in a few years , then the question of the lifetime of the electron states may be superfluous
. also there could be other electronic configurations in the ions studied here or in ions having a few more electrons , with suitable properties .
the relativistic corrections can lead to very rich structures , with level crossings and metastable states , which have just begun to be explored ( see for example @xcite ) and can lead to increased sensitivity to e.g. , electric quadrupole hyperfine interaction @xcite .
the systems we are interested in are highly charged heavy ions for which two states with equal angular momentum @xmath17 , but opposite parity have similar energy .
we have investigated the binding energies of the lower lying levels of ions with two to five electrons , to identify the most promising candidates . when not available from the literature , energies were calculated with the multiconfiguration dirac - fock program ( mcdf ) published by grant et _
@xcite , which provide relativistic correction as well as one - electron qed corrections and approximate , although inaccurate , many - body qed corrections . for helium - like systems however
, we can use very precise mcdf or relativistic configuration - interaction ( rci ) calculations including correlation and qed effects . for each electron configuration
we show as an example the results for uranium .
there are no noticeable qualitative differences for other heavy ions down to gold as is graphically shown for the interesting electron states , except for the two - electron @xmath18 case for which two crossings at @xmath19 and @xmath20 occur . as in this section
we do only exploratory work , we do not claim a precision much better than a few ev , except for two - electron systems .
lifetimes are calculated in the single ls configuration approximation .
we take the inverse of the main transition probability to be the lifetime of the respective state , neglecting hereby other contributions of lower order .
the parity admixture coefficient in this second section is determined only for the main electron state and therefore also gives only the order of magnitude .
extensive calculations of two - electron ions binding energies have appeared in the literature over the past 10 years @xcite . in figure
[ fig : he_cross ] , we plot the @xmath21 energy difference as calculated in @xcite and @xcite .
the first one is an all - order relativistic many - body perturbation theory ( rmbpt ) calculation , which uses ref .
@xcite two - body qed corrections .
the second calculation is a mcdf calculation done along the line of @xcite , which uses the welton model for two - body self - energy corrections , experimental nuclear size when available and includes finite - nuclear size correction to the self - energy @xcite .
the energy separation between @xmath22 and @xmath23 is plotted in fig.[fig : he_cross ] as a function of z. in order to show how this level crossing happens we show in detail the contributions to the energy separation at @xmath24 and 92 in table [ tab : hecross ] .
it should be noted that this crossing mostly involves the interplay between magnetic energy and qed correction contributions . with this new energy determination the parity admixture @xmath25 @xcite
would be enhanced by a factor of 3 . for the experimental set up discussed in @xcite with the detection of a laser - induced two - photon transition
, the laser intensity required would still be unrealistically large , of order @xmath26 w/@xmath27 ( presently only lasers up to an intensity of @xmath28 w/@xmath29 w/@xmath27 exist . ) .
the main problem in this context is that the heavy ions are only available in the form of a rapid ion beam and that the only possibility to excite the interesting electron states is by means of the stripping process .
one hope to improve the situation is to study different isotopes to see if one could still reduce the energy difference .
fig.[fig : he_iso ] shows that by chosing suitable isotopes the degeneracy can be improved .
only the coulomb energy is modified due to the change in nuclear radius . for uranium
the energy separation does cancel between isotope 233 and 234 , within the present calculation .
one should keep in mind , however , that the present calculation as well as the one in @xcite are not precise enough for finding exactly at which atomic number and for which isotope the crossing occurs .
the main uncertainty is in the self - energy screening . in table [
tab : hecross ] the self - energy screening is evaluated with the welton model@xcite , which has been proven to be rather accurate @xcite , but which is not ab - initio . in @xcite drake
s screening calculations , which are more adapted to low @xmath0 are used .
if one uses ab - initio qed calculations @xcite , one gets a larger screening .
however @xcite did not include relaxation , which seems to be sizable for the @xmath22 state . for uranium the welton model with relaxation
gives 4.29 ev , while the result from @xcite is only 1.08 ev .
it has been shown on other systems that the welton model should not be wrong by more than 10% for this atomic number , while it can be good to 1% at lower z @xcite .
one should note also that higher order radiative corrections ( of order @xmath30 , i.e. , of order @xmath31 with respect to the one electron self - energy ) , and qed corrections to the two - photon exchange diagrams @xcite have not been evaluated .
both corrections could be as large as 0.5 ev .
the position of the crossing point as well as the smallest energy which can be obtained is thus very uncertain .
also it should be remembered that if the energy separation is too small it may be difficult to find a laser to excite the two - photon transition . the characteristic feature of the lithium - like uranium ( cf . table [ tab : lilike ] )
is the fact , that already the ground state and the first excited state fulfill the main conditions of a parity - violation experiment , i.e. , they have the same angular momentum and opposite parity . moreover
the lifetime of the first excited state lies in the range of @xmath16 seconds .
very sophisticated calculation of the ionization energies in lithium - like uranium including discussion on nuclear effects can be found in @xcite .
complete calculations with relativistic correlation energy and radiative corrections for lower atomic numbers can be found in refs .
@xcite unluckily , between these two energy states there is a wide energy gap that reduces the magnitude of the parity admixture , which is in rough approximation about @xmath32 .
we shall discuss a scheme of detecting parity - violation in lithium - like atoms in section 3 .
[ fig : lith ] shows that the @xmath0-dependence of the energy difference of the first two electron states is nearly linear for atomic numbers ranging from @xmath33 , such that no element can be found for which the situation would be substantially different .
the beryllium - like ions case is comparable to the lithium - like case .
the first two electron levels are in principle suitable for parity admixture experiments .
the lifetime of the first excited state is very large and depends crucially on the spin of the nucleus @xcite . in the case of an even - even nucleus , e.g. , uranium 238 the lifetime is dominated by a two - photon @xmath34 transition which is in general very slow ( @xmath35 for @xmath36@xcite ) , and can therefore be treated to be infinity in comparison with the lifetime of the next higher levels . in the case of uranium 235
the nucleus has an angular momentum of 7/2 , and due to hyperfine mixing of electron orbitals the lifetime is severely reduced to @xmath37 s @xcite . as a model for beryllium - like heavy ions
we tabulate the energy and lifetime of the lower level of beryllium - like uranium in table [ tab : belike ] . in order to get a reasonable precision both the ground state and the @xmath38
are calculated as the lower and intermediate level of the @xmath39 configuration set , because intrashell correlation is very large in that case .
as in the lithium - like case the energy gap between the mixing levels is large , leading to a parity admixture of about @xmath40 .
for a five - electron system we examine again uranium ions ( cf . table [ tab : bolike ] ) .
the first two electronic levels are in principle usable for a parity - violation experiment , but the comparatively short lifetime of the first excited state and the small admixture of only @xmath41 make this system completely unattractive .
we shall discuss in the following therefore mainly lithium - like ions .
no level crossing was found for @xmath42 in any of the three , four and five - electron systems .
in this section we study super - heavy lithium - like ions .
it is interesting to check , how the situation would change if @xmath0 would be increased beyond the existing periodic system .
such high-@xmath0 systems can be formed for a short time in heavy - ion collisions . here
we treat the high-@xmath0 system as an ordinary atom with the charge @xmath43 being just the sum of its components .
while the energy difference @xmath44 is nearly linearly increasing in the range from @xmath45 to @xmath46 it again decreases in the higher @xmath0 region and has a crossing point at @xmath47 .
this effect is due to the relativistic contraction of the @xmath48 wave function which dominates over all other contributions for very large @xmath0 . for further increasing @xmath0 the @xmath48 state
, being below the @xmath49 state , reaches the the negative energy continuum @xcite .
we used desclaux s code to evaluate a number of systems for @xmath50 , with self - consistent magnetic interaction @xcite , vacuum polarization of order @xmath51 , @xmath52 and @xmath53 , self - energy extrapolated from mohr s values and corrected for finite nuclear size . for this to be valid , however we had to limit ourselves to @xmath54 .
it happens that the interesting region lies well inside this boundary . from table [
tab : supheav ] , one can see how for such high @xmath0 values the two interesting lithium - like states cross around the united charge number @xmath55 .
we analized only symmetric collision systems , which are parity even , provided their charge states are equal .
this section follows the analysis given in @xcite by g. w. botz , d. bru and o. nachtmann . we follow here their notations .
the energies , lifetimes , stark and parity admixture coefficients were calculated with the multi - configuration dirac fock package from @xcite . to make this paper self contained
let us shortly repeat some of the basic arguments of @xcite .
the atomic system we are interested in is a lithium - like ion that has a nonzero nuclear angular momentum . for simplicity
we take the nuclear angular momentum @xmath56 and look at the first four electron states ( cf . fig .
[ fig : hyp ] ) .
the situation for ions with other nuclear angular momentum is completely the same except that other numbers for the total angular momentum @xmath57 have to be inserted ( the formalism could also be applied to the boron - like case where we look at boron - like uranium 235 that has @xmath58 . ) .
the experimental situation in which we like to place this system is shown in fig .
[ fig : exp ] .
the lithium - like ion moves in the 1-direction of our coordinate system .
this ion is moving through alternating electric fields of width @xmath59 , at a distance of @xmath60 .
the electric fields point in the positive and negative 3-direction .
the moving ion sees a magnetic field due to the boost , but as this field is even under parity transformation we can neglect it .
the arrangement has still one symmetry operation @xmath61 under which it is invariant and this is a combination of parity transformation and rotation about @xmath62 around the 2-axis .
together , this gives a reflection with respect to the 1 - 3 plane .
@xmath63 @xmath64 it is clear that the angular momentum states @xmath65 are in general no eigenstates of this operation . but from @xmath66 it is easily seen that states with @xmath67 still are eigenstates of the reflection symmetry operator , and for simplicity we will constraint our considerations to those states .
this reflection symmetry is destroyed by the weak interaction of the electron with the nucleus which adds to the atomic hamiltonian the terms @xmath68 here @xmath69 runs over all quarks , @xmath3 is fermi s constant and @xmath70 denotes the neutral current coupling constants for the quark flavour @xmath69 or the electron @xmath71 , respectively .
both terms together have no defined parity and consequently no defined quantum number according to the reflection symmetry operation @xmath61 . on its flight
the ion stays for the time @xmath72 in the stark field and during the time @xmath73 outside of it .
following essentially the notation of @xcite we get for the transition amplitude during the time @xmath72 , in the case that there is no change in angular momentum @xmath74 in this formula we take @xmath75 to be the energy of the @xmath76 hyperfine states , perturbed by the parity - violating weak interaction denoted by the hat over the @xmath77 .
@xmath78 is the electric stark field @xmath79 multiplied with @xmath71 and the bohr radius : @xmath80 @xmath81 is the energy difference of the two electron states of opposite parity considered in fig .
[ fig : hyp ] and @xmath82 the decay constant of the @xmath83 state mentioned above . here
the hyperfine splitting is neglected because of its relative smallness .
the @xmath84 s are perturbative constants that give the admixtures due to the quadratic stark effect . in the case
that there is a transition between the angular momentum states the amplitude is proportional to the applied electric field i.e. , e.g. @xmath85 the total transition amplitude for an ion flying through one capacitor and the subsequent free drift length is given by @xmath86 for an experimental set up with @xmath87 capacitors the amplitude for the @xmath88 symmetry violating transition @xmath89 is : @xmath90 the basic idea is now to make the absolute value of such a transition amplitude large . to this end , with the definitions given before one can express first @xmath91 \right .
\nonumber \\ \nonumber \\ & & \left .
\quad - \frac{1}{2 } ( \kappa_{0,0 } - \kappa_{1,0 } ) \left ( \frac{\sqrt 3 { \cal f } } { l } \right)^2 \gamma t_1 \right \ } \quad.\end{aligned}\ ] ] here @xmath92 denotes the energy difference due to hyperfine splitting of the 2s electron orbitals .
this very expression can be made real by a suitable choice of the length of the free drift space so that the condition @xmath93 holds .
we will come back to this later . with the above choice of @xmath94 we can get for the absolute value of the amplitude @xmath95 the expression
@xmath96 here we have assumed @xmath97 .
now the aim is to maximize @xmath98 which is the same as to minimize @xmath99 .
this quantity @xmath99 plays an important role in this connection because as is shown in @xcite @xmath99 is a measure for the polarization rotation of the ion flying through the capacitor arrangement as at @xmath100 there is no component of angular momentum @xmath57 parallel to the direction of flight .
@xmath101 for definiteness we discuss the case of a pair of states with @xmath102 and @xmath103 .
we abbreviate @xmath104 and use as independent variables @xmath105 and @xmath87 .
we get up to factors independent of @xmath87 and @xmath105 @xmath106 let us assume @xmath87 to be large , then @xmath99 is antiproportional to the number of capacitors @xmath87 .
we now treat @xmath87 as a fixed number and then look for the minimum of @xmath99 as a function of @xmath105 . as @xmath97
the formal minimum of @xmath99 is obtained for @xmath107 such that in the vicinity of the minimum one has @xmath108 at the minimum the quantity @xmath78 , essentially the electric field @xmath79 , is determined by @xmath109 we shall discuss below that this optimal situation can not be reached for the ions considered here .
the derivation of these equations has been done for a pair of atomic states @xmath110 .
but there is no principal difference for other combinations like @xmath111 , which is considered here for boron - like uranium . while the formulae are just the same as derived in @xcite the quantities involved are quantitatively very different .
various large factors appear both in favor and in disfavor of the heavy - ion system and there is no simple way to estimate the relative size of the effect .
we shall present the numerical results for @xmath112 in table [ table : agl ] .
it will turn out that also some light ions might be of interest .
therefore we also add to table [ table : agl ] the results for the three lithium - like systems @xmath113 , @xmath114 and @xmath115 .
their atomic properties are shown in tables [ be9 ] , [ bo11 ] and [ ca13 ] . the atomic properties of @xmath116 are shown in tables [ tab:235u ] , [ meur235 ] and [ hfur235 ] .
for the calculation of the @xmath84 coefficients we use perturbation theory : @xmath117 here @xmath118 denotes the state with the quantum numbers @xmath119 and @xmath120 the other admixing states . solving this for the @xmath121 and using the wigner eckart , 6j- and 9j- theorems one gets @xmath122 here @xmath123 .
the point is now that the kappa coefficients only deviate by the small energy differences that are due to the hyperfine splitting . in table
[ table : kappa ] we show the @xmath84 values for @xmath113 , @xmath114 , @xmath115 and @xmath124 . together with the numerical values of the hyperfine splitting and the stark matrix elements , which are given in tables [ be9][hfur235 ]
, we can calculate the interesting expressions for the polarization rotation effects .
let us first start in the same way as in @xcite and analize the situation for the minimal @xmath99 .
it will turn out , that this assumption would imply unrealistically large electrical fields resulting from : @xmath125 \frac{1}{k } \qquad ( n = 0)\quad.\ ] ] from ( [ 4.12 ] ) we get the requirement for the individual effects to add .
@xmath126 = 2 \pi n \quad.\ ] ] this implies that the deviation @xmath127 in @xmath94 should be smaller than @xmath128 \quad.\ ] ] for the time @xmath72 , which gives the length of the capacitor we are required to take @xmath129 , but there are no other constraints . to make the required electric field small one has to chose @xmath72 large [ see eq . ( [ 4.24 ] ) below ] , so we take @xmath130
. finally from the relation @xmath131 \frac{1}{k}\ ] ] one can well calculate the electric field .
the terms in @xmath132 $ ] give always the necessary factors for the translation into si units .
the resulting numbers are given in table [ tab : txq ] .
here we always set @xmath133 .
the values for other @xmath87 can easily be determined from the formulas above .
note that @xmath87 has to be chosen very large and that the electric field @xmath79 is for the choice @xmath130 or for any choice @xmath134 independent of @xmath87 .
table [ tab : txq ] shows the results for @xmath113 , @xmath114 , @xmath115 and @xmath124 .
the values for the electric field @xmath79 are so unrealistically large that such an experiment can not be realized .
the reason for the large values of @xmath79 is the fact that in atoms with more than one electron the energy difference between the @xmath48 and the @xmath49 states is orders of magnitude larger than for hydrogen - like atoms because the @xmath135 degeneracy is removed by the electron - electron interaction .
we now proceed in the opposite direction .
we take a realistic field @xmath79 and other realistic values @xmath136 we then calculate @xmath137 \quad.\ ] ] as @xmath105 is very small we approximate @xmath138 \quad.\end{aligned}\ ] ] in this way we get the values of table [ tab : realistic ] .
these values must be compared to that obtained in @xcite for hydrogen @xmath139 .
in principle it is obvious that heavy ions with few inner shell electrons offer a possibility to test the effects of parity admixture .
this admixture is in heavy ions orders of magnitude larger than in neutral atoms .
the ideal case is a parity - violation effect to a sizable extent without applying any of the elaborated methods used in the cesium experiment @xcite .
then , the only chance to get measurable parity admixtures is to find a pair of energy states near the ground state with equal angular momentum but opposite parity that is nearly degenerated with respect to its energy .
unfortunately there is no such pair of orbitals in uranium with two to five electrons except for the already known degeneracy in helium - like uranium .
as the electron levels do only change very slowly with @xmath0 the same is true for the neighboring heavy ions.the next step will consequently be a very detailed analysis of the degeneracy in helium - like heavy ions including nuclear and isotopic effects because here a level crossing must exist .
level crossing also exists for compound nuclear reactions but here the lifetime of the compound nucleus is too short to allow for atomic physics experiments . looking for parity - violating spin rotations opened another perspective .
we showed , however , that the net effect ( value of @xmath140 ) for heavy ions is about thirty times weaker than for hydrogen .
lrrrrrrrr & @xmath141 & & & @xmath142 & & + & @xmath143&@xmath22 & diff & @xmath143&@xmath22 & diff + coulomb & -68868.56 & -68861.61 & -6.948 & -165518.05 & -165487.55 & -30.50 + magnetic & 38.30 & 17.12 & 21.17 & 151.30 & 66.36 & 84.91 + retardation & -3.26 & 1.30 & -4.56 & -10.09 & 5.56 & -15.65 + mass pol & -0.029 & 0.00 & -0.03 & -0.04 & 0.00 & -0.04 + correlation & -0.39 & -0.59 & 0.20 & -1.02 & -1.18 & 0.16 + 1e- self - energ & 82.66 & 95.16 & -12.50 & 364.88 & 420.68 & -55.80 + 2e- self - energ & -0.18 & -1.24 & 1.06 & -1.15 & -5.44 & 4.29 + uehling & -13.52 & -15.18 & 1.67 & -96.13 & -108.71 & 12.59 + & -0.02 & -0.10 & 0.087 & -0.28 & -0.89 & 0.61 + wichman & kroll & 0.38 & 0.42 & -0.04 & 4.75 & 5.28 & -0.53 + kallen & sabry & -0.11 & -0.12 & 0.01 & -0.73 & -0.83 & 0.09 + nuclear pol . & & & & -1.10 & -1.28 & 0.18 + total energy & -68764.71 & -68764.83 & 0.11 & -165107.70 & -165108.00 & 0.30 + ccccc main conf.&parity & energy ( ev ) & lifetime ( s ) + @xmath144 & @xmath145 & @xmath146 & @xmath147 + @xmath148 & @xmath149 & @xmath150 & @xmath151 + @xmath152 & @xmath149 & @xmath153 & @xmath154 + @xmath155 & @xmath145 & @xmath156 & @xmath157 + @xmath158 & @xmath149 & @xmath159 & @xmath160 +
|
we discuss the prospects for parity non - conservation experiments with highly charged heavy ions .
energy levels and parity mixing for heavy ions with two to five electrons are calculated .
we investigate two - photon - transitions and the possibility to observe interference effects between weak - matrix elements and stark matrix elements for periodic electric field configurations . plus 12pt minus 1pt plus 1pt pacs numbers : 32.30.-r , 35.10.wb , 32.90.+a uftp preprint 408/1996
|
the study of @xmath7 correlation functions ( cf ) ( along with the study of other two - particle correlations such as @xmath8 , @xmath9 ) allows one to obtain the information about the character of evolution of the matter formed in relativistic nuclear collisions , in particular about the character of collective flow .
another great possibility which is open for a researcher in baryon-(anti)baryon correlation analysis is the study of strong interaction between the particles of different sorts with final state interaction ( fsi ) correlation technique @xcite . since lhc produces copiously hadrons of different species , including multi - strange , charmed and beauty ones , the advantage of this approach is the ability to analyze the interactions even in exotic particle pairs , hardly achieved by other means .
fitting the experimental correlation function with some analytical formula , e. g. lednick - lyuboshitz model @xcite , allows one to extract the quantities describing both the interaction in corresponding pairs and the particle emission region size .
of course , if the source function @xmath10 , describing the spatial structure of the pair emission , is known , an extraction of unknown interaction parameters becomes more reliable . the corresponding spatial structure can be obtained from realistic collision model that simulates the evolution of the system formed in high energy heavy ion collisions . in the recent paper @xcite the hybrid variant of the hydrokinetic model ( hkm ) @xcite
was used for this purpose .
the choice of hkm is highly reliable since it is known to provide a successful simultaneous description of a wide class of bulk observables in nuclear collision experiments at rhic and lhc @xcite .
the model also reproduces well @xcite the source functions for pion and kaon pairs in au+au collisions at the top rhic energy @xcite , including non - gaussian tails observed in certain experimental source function projections . in @xcite in order to extract unknown @xmath11 scattering length ,
the corresponding experimental correlation function for 10% most central au+au collisions at top rhic energy , measured by star collaboration @xcite , was fitted with the lednick - lyuboshitz formula using the effective source radius @xmath12 extracted from the hkm source function .
it was also found that @xmath12 values obtained in hkm for baryon - baryon and baryon - antibaryon correlations are expectedly close , while in the star experimental analysis @xcite , where the source radii were considered as free fit parameters , the extracted @xmath13 radius was @xmath14 times smaller than the @xmath7 one .
one can assume that this apparent difference is due to neglect of residual correlations in @xcite .
such correlations can exist between secondary protons and lambdas if their parents were correlated ( or the parent of one particle was correlated with another particle in the pair ) . for taking into account the residual correlation contribution to the baryon - antibaryon cf , a modified analytical formula was introduced in @xcite . as a result ,
a good description of the experimental @xmath13 correlation function is obtained and the @xmath13 spin - averaged scattering length is extracted from the corresponding fit . in the present paper we are going to apply the method developed in @xcite for description of the rhic data to predict the @xmath7 and @xmath13 correlation functions in the 5% most central pb+pb collisions at the lhc energy @xmath4 tev , utilizing the results of ref .
@xcite as a starting point .
the experimental correlation functions presented in @xcite are purity corrected .
they are obtained from the measured ones as @xmath15 where @xmath16 and @xmath17 are the measured and the corrected cf respectively , and @xmath18 is pair purity .
the latter is defined as the fraction of pairs consisting of primary , correctly identified particles . as well as in
@xcite we model the purity corrected lhc baryon - baryon correlation function with lednick - lyuboshitz analytical formula @xcite : @xmath19,\end{aligned}\ ] ] where @xmath20 and @xmath21 .
this formula is derived starting from the basic equation @xmath22 , where the wave function @xmath23 represents the stationary solution of the scattering problem with the opposite sign of the vector @xmath24 .
the angle brackets mean averaging over the total spin @xmath25 and the distribution of the relative distances @xmath26 .
since typically the source radius can be considered much larger than the range of the strong interaction potential , @xmath23 can be approximated at small @xmath27 by the s - wave solution in the outer region : @xmath28 the scattering amplitude @xmath29 here is taken in the effective range approximation @xmath30 where @xmath31 is the scattering length and @xmath32 is the effective radius for a given total spin @xmath33 or @xmath34 . the singlet and triplet weights @xmath35 for unpolarized particles ( supposing the polarization @xmath36 ) are in @xmath37 and @xmath38 correspondingly .
as for the baryon - antibaryon case , following @xcite , we fit the experimentally measured correlation function @xmath39 with the following analytical expression @xmath40 where @xmath18 is purity or the fraction of correctly identified pairs consisting of primary particles , @xmath41 is `` true '' correlation function approximated by eq . ( [ ll ] )
, @xmath42 is the fraction of secondary particles which are residually correlated , @xmath43 and @xmath44 is the residual correlation contribution .
the latter is taken in the gaussian form @xcite @xmath45 where @xmath46 is the annihilation ( wide ) dip amplitude and @xmath47 is the dip inverse width . since @xmath48 and @xmath49 enter ( [ cfunc ] ) only as a product @xmath50 , the latter is treated as a single parameter @xmath51 at fitting .
the source radii @xmath12 in both @xmath5 and @xmath11 cases are extracted from the gaussian fit @xmath52 to the angle averaged source function , @xmath53 calculated in the hydrokinetic model @xcite .
the latter simulates the evolution of the matter in relativistic nuclear collision as consisting of two stages the hydrodynamic expansion of the matter being in local thermal and chemical equilibrium and gradual system decoupling , beginning when the equilibrium is lost .
the first stage is described within ideal hydrodynamics and for the second one the hydrokinetic approach is utilized , based on the boltzmann equations in the integral form , with switching to urqmd cascade at a space - like hypersurface . in current study hkm is taken in its simplified hybrid form @xcite with sudden switch from hydrodynamic evolution to the cascade at the hadronization hypersurface defined by the isotherm @xmath54 mev .
the model output consists of generated particle momenta and coordinates , that are further used to build different observables .
the considered angle - averaged source function histograms are filled using the following procedure ( here @xmath55 is the particle spatial separation in the pair rest frame ) @xmath56}{\sum_{n=1}^{n_{\mathrm{ev}}}\sum_{i_1^n , i_2^n}1}\ ] ] here @xmath57 and @xmath58 are the pair rest frame @xmath59-coordinates of particles 1 and 2 produced in the @xmath60th event , @xmath61 is the @xmath59-coordinate of the @xmath62-th histogram bin center , the function @xmath63 if @xmath64 and @xmath65 otherwise , and @xmath66 is the size of the histogram bin .
the initial conditions ( ic ) for hkm calculations simulating the considered case of 5% most central pb+pb collisions at the lhc energy @xmath4 tev are described in detail in @xcite .
we assume longitudinal boost invariance , so that the ic are specified in the plane transverse to the beam axis only .
the transverse energy density profile @xmath67 at the starting time @xmath68 fm/@xmath69 corresponds to monte carlo glauber model and is calculated in glissando code @xcite , where the overall scale factor @xmath70 , being the maximal initial energy density , is fixed basing on the experimental mean charged particle multiplicity .
the initial transverse flow in present calculations is absent . in fig .
[ sfav ] one can see the angle averaged @xmath5 source function @xmath71 obtained in hydrokinetic model for considered lhc collisions together with the gaussian fit to it .
the source radius value extracted from this fit is @xmath72 fm , which is about 1.15 times larger than for the rhic case .
the @xmath11 source function fitting results in the same source radius value . angle averaged source function calculated in hkm ( markers ) and the gaussian fit to it ( line ) .
the calculations correspond to 5@xmath73 most central pb+pb collisions at lhc energy @xmath4 tev , in pseudorapidity range @xmath74 .
proton @xmath75 range is @xmath76 gev/@xmath69 , and for lambdas @xmath77 gev/@xmath69 . the extracted @xmath72 fm . [ sfav],scaledwidth=88.0% ]
having obtained @xmath12 from hkm and fixing @xmath31 and @xmath32 according to the paper @xcite for baryon - baryon case , we can model the corresponding correlation function ( see fig .
[ plamlhc ] ) .
as compared to rhic , the lhc @xmath5 cf is slightly more narrow ( the two corresponding gaussian widths differ by a factor of @xmath78 ) and has lower intercept of about 1.7 . correlation function in the lhc pb+pb collisions at @xmath4 tev , @xmath79 , @xmath74 , with @xmath76 gev/@xmath69 for protons and @xmath77 gev/@xmath69 for lambdas ( red line ) .
the lhc source radius value calculated in hkm is @xmath72 fm .
the lednick - lyuboshitz fit to the top rhic energy correlation function , corresponding to the star experiment @xcite , with @xmath80 fm extracted from the hkm source function is presented for comparison ( blue line ) .
[ plamlhc],scaledwidth=90.0% ] to build the corresponding baryon - antibaryon correlation function one should determine the values of parameters entering ( [ cfunc ] ) and ( [ g ] ) for the lhc case . as well as for rhic we assume @xmath81 and @xmath82 . the source radius is again fixed from hkm calculation , @xmath72 fm . as for the real and imaginary parts of the scattering length , @xmath83 and @xmath84
, they characterize antiproton - lambda strong interaction and hence are not changed when switching from rhic to lhc .
so we can use the values extracted from the fit to rhic @xmath11 correlation function shown in fig .
6 of the paper @xcite , where the gaussian parametrization ( [ g ] ) for the residual correlation contribution @xmath44 is applied , @xmath85 fm and @xmath86 fm .
the parameter @xmath87 describes the strength of residual correlations and the fraction of residually correlated non - primary particles .
so , this parameter depends on particle interaction kinematics .
thus , it also can be taken the same as for rhic , @xmath88 .
as for the @xmath89 parameter , being some effective size associated with residual correlations ( @xmath90 fm ) , one can suppose that for lhc it will be larger than for rhic , approximately proportionally to the source radii ratio @xmath91 , i. e. @xmath92 .
such an assumption gives @xmath93 fm .
changing by such a close to unity factor . ]
note , that since the mentioned parameter values have errors , the lhc correlation function can not be predicted exactly , but only up to uncertainties caused by these errors in the fit parameters .
the uncertainty in the predicted correlation function is calculated as @xmath94 , where @xmath95 are the 4 fit parameters , @xmath83 , @xmath84 , @xmath51 and @xmath89 , @xmath96 are the corresponding parameter standard errors , and @xmath97 are the covariances of parameters @xmath95 and @xmath98 .
the determination of purity @xmath99 at the lhc is not so explicit , since it depends not only on the fraction of secondary pairs , coming from resonance decays , but also on the experimental setup , or more precisely , on the fraction of misidentified pairs . to clarify this issue
we have calculated the fractions of pairs made by particles having different origination ( primary or coming from certain decays ) in the hydrokinetic model for rhic and lhc collisions .
the results obtained in both cases are quite close and are presented in table i. comparing these fractions with those in table iii for rhic @xcite , one can see that experimental fraction of @xmath100 pairs @xmath101 is about 2.5 times lower than in hkm , likely due to misidentification problem , which takes place in the experiment .
in hkm simulations , on the contrary , all the produced particles are correctly identified , that leads to such a difference between corresponding fraction values . however , since hkm is a realistic model that describes both rhic and lhc bulk observables well , basing on its results one can conclude , that true purities , understood as the fractions of primary pairs , at rhic and lhc should be quite similar .
.the fractions of @xmath5 pairs , primary and coming from different decays calculated in hkm .
these fractions are quite similar for rhic and lhc cases . [ cols="^,^",options="header " , ] for purity uncorrected @xmath0 correlation function .
the hkm source radius for lhc is @xmath72 fm .
the purity @xmath99 is the same as for rhic case @xcite .
the scattering length real and imaginary parts , @xmath83 and @xmath84 , are taken from the fit to rhic cf that corresponds to fig . 6 from @xcite , where hkm source radius @xmath102 fm and the gaussian parametrization ( [ g ] ) for the residual correlation contribution @xmath44 are utilized .
for the lhc fit the @xmath44 parameter @xmath51 coincides with that for rhic , while parameter @xmath89 is scaled by the factor @xmath103 .
the lhc fit is determined up to errors in parameters @xmath83 , @xmath84 , @xmath51 and @xmath89 , that is illustrated by the band around the lhc curve .
[ aplamlhc1],scaledwidth=90.0% ] , but lhc purity @xmath99 is scaled as compared to rhic according to primary @xmath5 pairs fraction in hkm simulations , where there is no misidentification problem .
[ aplamlhc2],scaledwidth=90.0% ] . [ aplamlhc3],scaledwidth=90.0% ] in such a situation
we demonstrate three different plots for the lhc @xmath0 correlation function ( see fig .
[ aplamlhc1][aplamlhc3 ] ) .
the first one demonstrates the model cf with @xmath104 the same as for rhic @xcite , @xmath105 .
this function is again more narrow than for rhic . in fig .
[ aplamlhc2 ] one can see our prediction for the lhc correlation function with @xmath106 , where the factor 2.5 corresponds to the ratio of corresponding primary pairs fractions in hkm and in the star experiment . as for fig .
[ aplamlhc3 ] , it shows the purity and residual correlation corrected cfs for lhc and rhic .
they are expressed ( as it follows from eq .
( [ cfunc ] ) ) through the uncorrected ones as @xmath107 .
the latter `` true '' function can be easily compared with the experimental result ( corrected in the same way for purity and residual correlations ) , since it does not depend on the fraction of misidentified particles in the concrete experiment . as compared to rhic ,
the lhc curve apparently has smaller amplitude and width .
the first predictions for @xmath5 and @xmath11 correlation functions , calculated within lednick - lyuboshitz and hydrokinetic ( hkm ) models and accounting for the residual correlation effect , are presented for the 5% most central pb+pb lhc collisions at the energy @xmath4 tev .
the functions behavior is predicted based on the results previously obtained for au+au collisions at the top rhic energy .
the pair purities ( the fractions of pairs consisting of primary particles ) , calculated in hkm for rhic and lhc are very close and are about 2.5 times larger than the experimental primary pairs fraction in the star collaboration experiment at the rhic .
this difference is most probably because of particle misidentification problem existing in the experiment .
that is why we present results for purity uncorrected @xmath11 correlation functions in two variants : with purity @xmath99 similar to that in the experiment at rhic and with rhic purity scaled by a factor of 2.5 , that corresponds to hkm purity .
the measured lhc correlation function should lie somewhere between these two limiting possibilities . depending on the concrete experimental setup and the related fraction of misidentified particles , the real curve can be closer either to one or to another variant .
we also demonstrate the purity and residual correlation corrected baryon - antibaryon cf which should not depend on the experiment details and thus can be easily compared with experimental result .
this function for lhc has smaller amplitude than for rhic .
the authors are grateful to iurii karpenko for his assistance with computer code .
the research was carried out within the scope of the eurea : european ultra relativistic energies agreement ( european research group : `` heavy ions at ultrarelativistic energies '' ) and is supported by the national academy of sciences of ukraine ( agreement mvc1 - 2015 ) .
r. lednick , v. l. lyuboshitz , yad . fiz . *
35 * , 1316 ( 1981 ) [ sov .
. phys . * 35 * , 770 ( 1982 ) ] ; in proceedings of the international workshop on particle correlations and interferometry in nuclear collisions ( corinne 90 ) , nantes , france , 1990 , edited by d. ardouin ( world scientific , singapore , 1990 ) , pp . 4254 ; r. lednicky , j. phys .
g : nucl . part .
* 35 * ( 2008 ) 125109 .
|
the theoretical analysis of @xmath0 correlation function in 10% most central au+au collisions at rhic energy @xmath1 gev shows that the contribution of residual correlations is the necessary factor to obtain a satisfactory description of the experimental data .
a neglecting of the residual correlation effect , leads to unrealistically low source radius , about 2 times smaller than the corresponding value for @xmath2 case , when one fits the experimental correlation function within lednick - lyuboshitz analytical model .
recently an approach accounting effectively for residual correlations for the baryon - antibaryon correlation function was proposed , and a good rhic data description was reached with the source radius extracted from the hydrokinetic model ( hkm ) . the @xmath3 scattering length , as well as the parameters characterizing the residual correlation effect annihilation dip amplitude and its inverse width were extracted from the corresponding fit . in this paper
we use these extracted values and simulated in hkm source functions for pb+pb collisions at the lhc energy @xmath4 tev to predict the corresponding @xmath5 and @xmath6 correlation functions .
keywords : _ final state interaction , baryons , lead - lead collisions , lhc , residual correlations _
|
since markowitz ( 1952 ) formulated the idea of diversification of investments , the mean - variance approach has been widely used in practice in asset allocation and portfolio management , despite many sophisticated models proposed in literature . on the other hand , some drawbacks of the standard markowitz approach are reported in literature ( michaud ( 1989 ) ) .
one of the critical weaknesses of the classical mean - variance analysis is its lack of robustness . since the classical estimators of the mean and the covariance matrix , which are inputs in the optimization procedure , are very sensitive to the presence of gross errors or atypical events in data , the weights of the resulted portfolio , which are outputs of this procedure , can be drastically affected by these atypical data
this fact was proved by perret - gentil and victoria - feser ( 2005 ) by using the influence function approach .
also , some other recent papers underline this idea and show that the large or small values of asset returns can have an abnormally large influence on the estimations leading to portfolio that are far to be optimal ( grossi and laurini ( 2011 ) ) . in order to remove this drawback and to construct portfolios
not overly affected by deviations of the data from the assumed model , many methods have been proposed in literature . for an overview on the methods used for robust portfolio optimization we refer to fabozzi et al .
( 2010 ) . among the methods which improve the stability of portfolio weights by using robust estimators of the mean and covariance , we recall those proposed by vaz - de melo and camara ( 2005 ) which use m - estimators , perret - gentil and victoria - feser ( 2005 ) which use the translated biweight s - estimator , welsch and zhou ( 2007 ) which use minimum covariance determinant estimator and winsorization , demiguel and nogales ( 2009 ) which use both m- and s - estimators , ferrari and paterlini ( 2010 ) which use maximum l@xmath0-likelihood estimators .
these contributions have the merit to consider the role of robust estimation for improving the mean - variance portfolios .
on the other hand , it is known that traditional robust estimators suffer dramatic looses in efficiency compared with the maximum likelihood estimator .
therefore , a trade - off between robustness and efficiency should be carefully analyzed .
our contribution to robust portfolio optimization is developed within a minimum pseudodistance framework .
we can say that the minimum pseudodistance methods for estimation take part to the same cathegory with minimum divergence methods .
the minimum divergence estimators are defined by minimizing some appropriate divergence between an assumed model and the true model underlying the data .
depending on the choice of the divergence , minimum divergence estimators can afford considerable robustness at minimal loss of efficiency .
however , the classical approaches based on divergence minimization require nonparametric density estimation , which can be problematic in multi - dimensional settings .
some proposals to avoid the nonparametric density estimation in minimum divergence estimation have been made by basu et al .
( 1998 ) and broniatowski and keziou ( 2009 ) and robustness properties of their estimators have been studied by toma and leoni - aubin ( 2010 ) , toma and broniatowski ( 2011 ) . in this paper
we consider estimators of location and covariance obtained by minimizing a family of pseudodistances .
these estimators have the advantages of not requiring any prior smoothing and conciliate robustness with high efficiency , usually requiring distinct techniques .
the minimum pseudodistance estimators have been introduced by broniatowski et al .
( 2012 ) and consist in minimization of an empirical version of a pseudodistance between the assumed model and the true model underlying the data .
this method can be applied to any parametric model , but in the present paper we focus on the multivariate normal location - scale model .
the behavior of the estimator depends on a tuning positive parameter @xmath1 which controls the trade - off between robustness and efficiency .
when the data are consistent with normality and @xmath2 , the estimation method corresponds to the maximum likelihood method ( mle ) which is known to have full asymptotic efficiency at the model .
when @xmath3 , the estimator gains robustness , while keeping high efficiency .
the new minimum pseudodistance estimators can be easily used in place of the classical mean and covariance matrix estimators , thereby providing robust and efficient mean - variance optimized portfolios .
the outline of the paper is as follows : in section 2 , we shortly describe the markowitz s mean - variance model whose inputs are estimations of location and covariance of asset returns . the minimum pseudodistance estimators of location and covariance are introduced in section 3 . here
we prove theoretical properties of these estimators , such as the affine equivariance and b - robustness .
we also determine the asymptotic covariance matrices of the estimators and discuss the asymptotic relative efficiency .
the estimators of the portfolio weights together with their properties are presented in section 4 . in sections 5 and 6 , a monte carlo simulation study and then an application on real data show the advantages of the new approach .
we consider a portfolio formed by @xmath4 financial assets .
the returns of the assets are characterized by the random vector @xmath5 , where @xmath6 denotes the random variable associated to the return of the asset @xmath7 , @xmath8 .
let @xmath9 be the vector of weights associated to the portfolio , where @xmath10 represents the proportion of the investor s capital invested in the asset @xmath7 .
the total return of the portfolio is given by the random variable @xmath11 supposing that the random vector @xmath12 follows a multivariate normal distribution @xmath13 , where @xmath14 is the vector containing the mean returns of the assets and @xmath15 is the covariance matrix of the returns of the assets , the mean of the portfolio return can be written as @xmath16 and the portfolio variance as @xmath17 the markowitz approach for optimal portfolio selection consists in solving the following optimization problem .
for a given investor s risk aversion @xmath18 , the mean - variance optimization selects the portfolio @xmath19 , solution of @xmath20 with the constraint @xmath21 , @xmath22 being the @xmath23 vector of ones .
the set of optimal portfolios for all possible values of the risk aversion parameter @xmath24 defines the mean - variance efficient frontier .
the solution of the above optimization problem is explicit and the optimal portfolio weights , for a fixed value of @xmath24 , are given by @xmath25 where @xmath26 this is the case when short selling is allowed .
when short selling is not allowed , we have a supplementary constraint in the optimization problem , namely all the weights @xmath10 are positive .
when the true parameters @xmath14 and @xmath15 and the portfolio weights are all known , then we have the true efficient frontier .
an estimated efficient frontier can be obtained by using estimators of the mean and covariance matrix . throughout this paper
we denote by @xmath27 and @xmath28 the estimators of the parameters @xmath14 and @xmath15 , and by @xmath29 the estimator of the optimal portfolio weights , as resulting with ( [ f1 ] ) @xmath30.\ ] ] the mean and the covariance matrix of the returns are in practice estimated by their sample counterparts , i.e. the maximum likelihood estimators under the multivariate normal model .
it is known that , under normality , the maximum likelihood estimators are the most efficient . however , in the presence of outlying observations , the asymptotic bias of these estimators can be arbitrarily large and this bias is induced to the corresponding optimal portfolio weights .
for this reason , @xmath14 and @xmath15 should be robustly estimated .
in the following , for the robust estimation of the parameters @xmath14 and @xmath15 we consider minimum pseudodistance estimators . for two probability measures @xmath31 and @xmath32 , admitting densities @xmath33 , respectively @xmath34 with respect to the lebesgue measure @xmath24 , the pseudodistances that we consider are defined through @xmath35 for @xmath3 and satisfy the limit relation @xmath36 note that @xmath37 coincides with the modified kullback - leibler divergence .
let @xmath38 be a parametric model with parameter space @xmath39 and assume that every probability measure @xmath40 in @xmath38 has a density @xmath41 with respect to the lebesgue measure .
the family of minimum pseudodistance estimators of the unknown parameter @xmath42 is obtained by replacing the hypothetical probability measure @xmath43 in the pseudodistances @xmath44 by the empirical measure @xmath45 pertaining to the sample and then minimizing @xmath46 with respect to @xmath47 on the parameter space .
let @xmath48 be a sample on @xmath49 and denote by @xmath50 the parameter of interest .
a minimum pseudodistance estimator @xmath51 of @xmath47 is defined by @xmath52 which can be written equivalently as @xmath53{ll}% \arg\sup_{\theta}\frac{1}{tc_{\alpha}(\theta)}\sum_{i=1}^{t}p_{\theta}^{\alpha}(x^{i } ) & \text { \ \ \ } \mbox{if}\ \alpha>0\medskip\\ \arg\sup_{\theta}\frac { { 1 } } { { t}}\sum_{i=1}^{t}\ln p_{{\theta}}(x^{i } ) & \text { \ \ \ } % \mbox{if}\text { } \alpha=0 \end{array } \right.\ ] ] where @xmath41 is the @xmath4-variate normal density @xmath54 and @xmath55 .
note that , the choice @xmath56 leads to the definition of the classical mle . throughout the paper we will also use the notation @xmath57^{1/2}$ ] .
a simple calculation shows that @xmath58 and then , for @xmath3 the minimum pseudodistance estimator ( [ mpe ] ) can be expressed as @xmath59 the location and dispersion estimators defined above are affine equivariant .
more precisely , if @xmath61 and @xmath62 are estimators corresponding to a sample @xmath63 , then @xmath64 for any @xmath65 nonsingular matrix @xmath66 and any @xmath67 . indeed , let @xmath66 be a @xmath65 nonsingular matrix , @xmath67 and @xmath68 with @xmath69 .
the estimators @xmath70 and @xmath71 are solutions of the system obtained from ( [ sys1 ] ) and ( [ sys2 ] ) by replacing @xmath72 with @xmath73 .
then , by replacing @xmath73 with @xmath74 , we find @xmath75 and @xmath76 . hence , ( [ ae1 ] ) and ( [ ae2 ] ) hold .
a fundamental tool used for studying statistical robustness is the influence function .
the influence function is useful to determine analytically and numerically the stability properties of an estimator in case of model misspecification . recall that
, a map @xmath77 defined on a set of probability measures and parameter space valued is a statistical functional corresponding to an estimator @xmath78 of the parameter @xmath47 , if @xmath79 , where @xmath45 is the empirical measure associated to the sample .
as it is known , the influence function of @xmath77 at @xmath40 is defined by @xmath80 where @xmath81 @xmath82 , @xmath83 being the dirac measure putting all mass at @xmath84 . whenever the influence function is bounded with respect to @xmath84 the corresponding estimator
is called robust .
the statistical functionals associated to the minimum pseudodistance estimators of @xmath14 and @xmath15 are @xmath85 and @xmath86 defined by the solutions of the system @xmath87dp(x)=0.\end{aligned}\ ] ] this system can be rewritten under the form @xmath88 dp(x)=0\label{f4}\end{aligned}\ ] ] where @xmath89 arslan , o. ( 2004 ) .
convergence behavior of an iterative reweighting algorithm to compute multivariate m - estimates for location and scatter .
_ journal of statistical planning and inference , 118 _ , 115 - 128 .
ferrari , d. , & paterlini , s. ( 2010 ) .
efficient and robust estimation for financial returns : an approach based on @xmath34-entropy .
materiali di discussione , no .
623 , universita degli studi di modena e reggio emilia .
jaupi , l. , & saporta , g. ( 1993 ) . using the influence function in robust principal component analysis . in s. morgenthaler , e. ronchetti , & w.a .
stahel ( eds . ) , _ new directions in statistical data analysis and robustness _ ( pp .
147 - 156 ) .
verlag : birchauser .
|
the presence of outliers in financial asset returns is a frequently occuring phenomenon and may lead to unreliable mean - variance optimized portfolios .
this fact is due to the unbounded influence that outliers can have on the mean returns and covariance estimators that are inputs in the optimization procedure . in the present paper
we consider new robust estimators of location and covariance obtained by minimizing an empirical version of a pseudodistance between the assumed model and the true model underlying the data .
we prove statistical properties of the new mean and covariance matrix estimators , such as affine equivariance , b - robustness and efficiency .
these estimators can be easily used in place of the classical estimators , thereby providing robust optimized portfolios . a monte carlo simulation study and an application to real data
show the advantages of the proposed approach .
robustness and sensitivity analysis , portfolio optimization
|
it has been established at lower energies that the @xmath0s created in nuclear collisions are suppressed when the system size and the centrality of the collision increase @xcite . for heavy enough nuclei
the suppression exceeds the _ normal _ nuclear absorption observed with light nuclei .
several mechanisms , including color screening in the quark gluon plasma , have been proposed to explain this _ abnormal _ suppression observed at low energy .
measurements performed at higher energy are therefore crucial to constrain the models and discriminate between them .
moreover they may uncover new mechanisms such as quark recombination .
the phenix experiment measures @xmath2 decay at mid rapidity ( @xmath3 ) and @xmath4 decay at forward rapidity ( @xmath5 $ ] ) .
electrons are identified using rich detectors and by matching their momentum measured in drift chambers with the energy deposited in electromagnetic calorimeters .
muons are selected using a thick absorber located close to the interaction point , tracked using cathode strip chambers and triggered on using a succession of iarocci tube planes and steel walls .
the results presented here correspond to 240 @xmath6b@xmath7 au+au collisions collected in 2004 , and 3.1 nb@xmath7 cu+cu collisions in 2005 , both at @xmath1 gev . for each centrality , rapidity or transverse momentum bin ,
the @xmath0 yield is obtained by counting the number of unlike - sign dilepton pairs in a mass window centered on the @xmath0 mass after the background has been subtracted using either an event mixing technique or the mass distribution of the like - sign pairs .
it is corrected by the efficiency and acceptance of the spectrometer as well as the trigger efficiency and normalized to the number of recorded collisions .
this yield enters the numerator of the nuclear modification factor which is the ratio between the measured yield in a+b collisions over the expected yield when assuming binary scaling from p+p collisions .
the @xmath0 yield measured in p+p collisions is taken from phenix 2003 published results @xcite .
systematic errors have been assigned to each contribution .
for au+au collisions , the dominant systematic error comes from the background subtraction , due to the poor signal over background ratio , especially for central collisions .
figures [ fig_vogt ] and [ fig_theo ] represent the @xmath0 nuclear modification factor as a function of the number of participant for au+au and cu+cu collisions both at mid an forward rapidity together with the published measurement from d+au @xcite .
a suppression of about a factor 3 is observed for the most central collisions . on figure [ fig_vogt ] ,
the data are compared to r. vogt prediction @xcite , assuming 3 mb nuclear absorption on top of eks98 shadowing .
the prediction miss the most central points .
moreover , from @xcite it appears that 3 mb is an upper bound for the nuclear absorption cross - section in d+au collisions .
@xmath0 nuclear modification factor as a function of the number of participant , compared to r. vogt predictions for the normal nuclear absorption .
vertical bars are statistical errors , brackets are point to point systematics and boxes are global systematics.,width=283 ] on figure [ fig_theo ] the same data are compared to two sets of models .
they all involve additional final state interactions and reproduce both the cern sps results @xcite and the phenix low statistics results from 2002 @xcite .
the predictions plotted on the left panel ( @xcite ) overestimate the @xmath0 suppression .
the predictions plotted on the right panel show better agreement with the data or even underestimate the suppression .
the predictions on the right involve either quark recombination mechanisms @xcite or detailed @xmath0 transport in medium @xcite .
note however that these predictions differ in the way the cold nuclear absorption is accounted for , in the p+p @xmath0 production cross - section used for normalization and in the open charm production cross - section entering recombination mechanism .
[ cols="^,^ " , ]
phenix measured @xmath0 in au+au and cu+cu collisions at @xmath1 gev at both mid and forward rapidity . the dependence of the @xmath0 nuclear modification factor on the number of participants allows to discriminate between some of the models capable of reproducing the suppression seen at lower energy .
the @xmath0 rapidity spectra show no obvious variation in shape .
interpretation of the @xmath0 mean square transverse momentum is unclear due to large errors . 9 na50 collaboration , m.c .
abreu et al , eur .
j. c * 39 * , 335 - 345 ( 2005 ) .
phenix collaboration , s.s .
adler et al , nucl - ex/0507032 .
r. vogt , nucl - th/0507027 .
phenix collaboration , s.s .
adler et al , phys .
c * 69 * , 014901 ( 2004 ) .
a. capella , e.g. ferreiro , hep - ph/0505032 .
l. grandchamp , r. rapp , g. e. brown , phys .
lett . * 92 * , 212301 , ( 2004 ) + ( hep - ph/0306077 ) .
a.p . kostyuk , m.i .
gorenstein , horst stoecker , w. greiner , phys .
c * 68 * , 041902 ( 2003 ) ( hep - ph/0305277 ) .
a. andronic , p. braun - munzinger , k. redlich , j. stachel , phys .
b * 571 * , 36 - 44 ( 2003 ) ( nucl - th/0303036 ) .
bratkovskaya , a.p .
kostyuk , w. cassing , horst stoecker , phys .
c * 69 * , 054903 ( 2004 ) ( nucl - th/0402042 ) .
xiang - lei zhu , peng - fei zhuang , nu xu , phys .
b * 607 * : 107 - 114 ( 2005 ) + ( nucl - th/0411093 ) .
thews , m.l .
mangano , nucl - th/0505055 .
|
heavy quarkonia production is predicted to be sensitive to the formation of the quark gluon plasma in relativistic heavy ion collisions via competing mechanisms such as color screening and/or quark recombination . during 2004 and 2005
rhic data taking periods , the phenix collaboration has measured @xmath0 decay into lepton pairs both at mid and forward rapidity in au+au and cu+cu collisions at @xmath1 gev .
we present the @xmath0 nuclear modification factor and the @xmath0 mean square transverse momentum as a function of the collision centrality for both systems , as well as the rapidity dependence of the @xmath0 yield for different centrality classes .
it is compared to different theoretical predictions .
all au+au and cu+cu results shown here are preliminary .
|
the hidden order problem in uru@xmath0si@xmath0 is an unanswered question in the field of strongly correlated electron materials .
although it is studied since several decades there is still no consensus about how this new phase forms .
understanding the hidden order phase formation is not only an intellectual problem , it will also advance concepts for designing quantum materials with new exotic properties .
many hidden order scenarios are based on the assumption of certain ground state symmetries and the present study addresses this aspect . a novel spectroscopic technique , non - resonant inelastic x - ray scattering ( nixs ) , that has become available thanks to high brilliance synchrotrons , allows to measure directly in a bulk sensitive experiment the symmetry of the 5@xmath1 ground state wave function in uru@xmath0si@xmath0 . in heavy fermion rare earth or actinide compounds , the @xmath1 electrons are well localized at high temperatures , but as temperature is lowered hybridization with conduction electrons becomes increasingly effective , resulting in a more itinerant @xmath1-electron character at low temperatures .
these hybridized @xmath1 electrons form narrow bands and have large effective masses .
quasiparticle interaction effects in these narrow bands are responsible for the many exciting phenomena present in heavy fermion compounds : multipolar order@xcite , unconventional superconductivity@xcite or quantum criticality@xcite .
the hidden order phase in / is one example of the exotic low temperature phases found in this material class .
/ is a tetragonal heavy fermion compound that undergoes two phase transitions , the nonmagnetic _ hidden order _ ( @xmath7 ) transition at @xmath8=17.5k that goes along with an appreciable loss of entropy , and a superconducting one at about 1.5k@xcite . below the @xmath7 transition small ordered magnetic moments
were observed in the earlier studies , but turned out later to belong to a parasitic minority phase . with applied pressure ( [email protected] )
the @xmath7 order is replaced by an antiferromagnetic phase with large ordered moments ( so - called @xmath10-phase)@xcite .
the order parameter of the @xmath7 phase has been subject of intense investigations since more than 30 years , but so far it remained hidden which has been the inspiration for its name . this second order transition into an electronically ordered state involves a reconstruction of the fermi surface@xcite and a change of quasiparticle scattering rate@xcite .
the fermi surfaces of the @xmath7 and high pressure @xmath10 phase are very similar@xcite . in / three energy scales
have been identified : a hybridization gap of @xmath11@xmath1213mev [ 150k]@xcite that opens below 27k , another gap that opens in the @xmath7 phase with @[email protected] [ 50k ] in the charge@xcite as well as spin channel@xcite and a resonance mode that appears in the @xmath7 gap at @xmath141.6mev [ 18k ] , also in both channels@xcite .
furthermore with entering the @xmath7 phase the breaking of the fourfold rotational symmetry has been reported from torque experiments@xcite and high resolution x - ray diffraction on high quality crystals@xcite . for a more detailed experimental and theoretical survey of physical properties of
/ we refer to the review article by mydosh and oppeneer@xcite . in intermetallic actinide compounds the valence state
is often intermediate , and indeed , @xmath15-edge sum rules@xcite , life time reduced l - edge absorption@xcite and soft photoelectron spectroscopy measurements@xcite find a valence between 3@xmath16 and 4@xmath16 for /.
it is an itinerant system , and yet , electron correlations on the u atom will reduce the charge fluctuations and favor also a particular local irreducible representation@xcite . in this respect
it is suggestive to assume that the u@xmath17@xmath18 configuration will give the dominant contribution which is in line with first - principle dmft calculations@xcite .
the question is now which of the u@xmath17(@xmath18 ) states build up the itinerant state and lead to the formation of the @xmath7 .
the present work presents the asymmetry of the inelastic x - ray scattering function s(@xmath19,@xmath20 ) as measured in a bulk sensitive , non - resonant inelastic x - ray scattering experiment ( nixs ) and gives direct and quantitative information on the 5@xmath1 symmetry in uru@xmath0si@xmath0 .
dependence ( left ) and energy dependence ( right ) of the scattering function s(@xmath19,@xmath20 ) at the u @xmath21-edge for dipole ( k=1 ) , octupole ( k=3 ) , and dotriacontapole ( k=5 ) scattering orders .
the gray vertical line marks the @xmath22-range of the experiment .
note : features above @xmath12106ev appear unrealistically narrow since the proximity of continuum states is not accounted for . ] to address the local 5@xmath1 degrees of freedom of uru@xmath0si@xmath0 we will make use of the crystal - electric field ( cef ) description of the u@xmath17@xmath18 configuration in @xmath23 symmetry .
the cef splits the nine fold degenerate @xmath24=4 multiplet into five singlets and two doublets which can be written in the following way when using the @xmath3 representation .
@xmath25 here the values @xmath26 and @xmath27 define the mixing of states that have equal irreducible representation , that is the singlet states @xmath28 and doublet states @xmath29 .
the phase relations between the @xmath3 states are defined such that the operator @xmath30 is non - negative .
note , that @xmath31(90/)=-@xmath32(0/ ) and @xmath33(90/)=@xmath34(0/ ) and , depending on the mixing angles @xmath27 and @xmath26 , the cef states correspond to pure @xmath3 states ( @xmath31(90/)@xmath35@xmath4+@xmath5 , @xmath6@xmath35@xmath4-@xmath5 , @xmath32(90/)@xmath35@xmath36 , @xmath34(90/)@xmath35@xmath37 , and @xmath33(90/)@xmath35@xmath38 ) .
determining cef excitations and their symmetry in intermetallic u compounds is by no means trivial since the 5@xmath1 electrons are more itinerant than e.g. the 4@xmath1 electrons in the rare earth series , and the classical tool inelastic neutron scattering fails to observe sharp cef excitations@xcite due to dispersive effects and the large intrinsic widths that goes along with itinerant states .
nevertheless there have been many experimental and also theoretical attempts to determine the symmetries of the 5@xmath1 ground state and low lying electronic states in / , and in literature a wide spectrum of different scenarios can be found .
the anisotropy of the static susceptibility is well described with a @xmath31 singlet ground state , a @xmath6 as a first excited state and the next states above 15mev [ 170k]@xcite .
analyses of elastic constant measurements find similar results@xcite . also kiss and fazekas@xcite , hanzawa@xcite and kusunose _ _ et al.__@xcite favour a @xmath31 , the model of kiss and fazekas being also compatible with a @xmath32 singlet ground state@xcite , but they all propose different first excited states from their theoretical considerations .
haule and kotliar@xcite also propose two low lying singlet states , a @xmath6 singlet ground state and a @xmath32 as first excited state , a scenario that is compatible with the interpretation of polarized raman studies that find a resonance at 1.6mev in the @xmath39 channel in the @xmath7 phase@xcite @xcite .
thermodynamic measurements by santini and amoretti@xcite and resonant x - ray scattering data by nagao and igarashi@xcite are interpreted in terms of a @xmath40-singlet ground state with the @xmath31 as first excited state or alternatively with a @xmath34 ground state@xcite .
another elastic constant study by kuwahara _ _ et al.__@xcite yields a @xmath41 as lowest state .
@xmath34 and @xmath33 doublets as ground states are concluded by thermodynamic studies of diluted /@xcite and theoretical considerations by ohkawa and shimizu@xcite and chandra et al.@xcite .
finally @xmath42-edge x - ray absorption measurements by wray _ _
et al.__@xcite favour the @xmath34 and sugiyama _ et al .
_ the @xmath33 doublet@xcite as ground state .
-edge for @xmath43=9.6@xmath44 and corresponding calculations for 5@xmath454@xmath18@xmath465@xmath474@xmath48 .
( a ) simulation of s(@xmath19,@xmath20 ) of u crystal - field states for @xmath24=4 in @xmath23 symmetry for the two directions @xmath49@xmath50[100 ] ( blue ) and [ 001 ] ( red ) .
the insets show the corresponding electron densities ( see section _
charge density plots _ in the appendix ) .
( b ) nixs data for momentum transfers @xmath49@xmath50[100 ] ( blue ) and [ 001 ] ( red ) at t=25k .
( c ) dichroism at 25k in % defined as difference i@xmath51@xmath52}$]-i@xmath51@xmath53}$ ] relative to peakhight r@xmath54 as defined in the isotropic data ( see fig.3 ) , data ( black dots ) and calculations ( green lines ) for the crystal - field states with the correct sign of dichroism . here
the data points have been convoluted with a gaussian of 0.5ev fwhm . ]
si@xmath0 at the u @xmath21 edge ( 5@xmath454@xmath18@xmath465@xmath474@xmath48 ) for t@xmath5525k . for details
see text . ]
there is clearly room for clarification
. hence we aim at determining the symmetries of the ground state and low lying states in / using a spectroscopic method that directly probes the u 5@xmath1 shell .
we performed a core - level _ non - resonant _ inelastic x - ray scattering experiment ( nixs ) at the u @xmath21 edges ( @xmath56@xmath46@xmath57 ) with hard x - rays ( @xmath1210kev ) and large momentum transfers ( @[email protected]@xmath44 ) .
nixs is a photon - in - photon - out technique that was used in the recent past on single crystals for determining the wave functions of cerium based systems@xcite . in nixs
the direction dependence of the momentum transfer * @xmath19 * is used in analogy to the linear polarization dependence in an x - ray absorption spectroscopy ( xas ) experiment ( see e.g. ref.@xcite and also @xcite ) and accordingly _ multipole selection rules _ give access to the ground state symmetry ( dipole for xas ) .
the higher multipoles that contribute significantly to the scattering function s(@xmath19,@xmath20 ) at large momentum transfers contain more information than dipole so that e.g. asymmetries with higher than twofold rotational symmetry can be detected@xcite .
in addition , at the u @xmath21-edge these excitations are significantly narrower than the dipole signal which is strongly broadened due to the proximity of continuum states@xcite .
most importantly , it should be mentioned that a nixs experiment does not involve an intermediate state so that the quantitative modeling is as straightforward as for xas and the use of hard x - rays makes the signal truly bulk sensitive in contrast to a soft xas or soft rixs experiment .
@xmath50[100 ] ( blue dots ) and @xmath49@xmath50[001 ] ( red dots ) . for better comparison , the @xmath58 dependent data are underlaid with the difference in spectral weight of the 5k data .
( b ) dichroism i@xmath51@xmath52}$]-i@xmath51@xmath53}$ ] for all temperatures , convoluted with a gaussian of 0.5ev fwhm .
( c ) nixs data and simulations ( 5@xmath454@xmath18@xmath465@xmath474@xmath48 ) for in - plane momenta @xmath49 parallel to [ 100 ] and for @xmath49 turned towards [ 010 ] by 22.5/ and 45/ , i.e. @xmath49@xmath50[100 ] ( blue ) , @xmath49@xmath50[520 ] ( dark green ) and @xmath49@xmath50[110 ] ( light green ) for all cef states . ]
several nixs studies , also on uranium compounds , show experimentally and theoretically how the multiplet excitations develop with increasing momentum transfer@xcite .
however , for convenience of the reader we recapitulate briefly the principle of nixs : when working at large momentum transfers the expansion of the transition operator exp(i@xmath19@xmath59 ) in spherical harmonics can not be truncated after the first term , thus giving rise to excitations due to quadru- , octupole and higher order contribution in s(*@xmath19*,@xmath20 ) .
figure[fig1 ] shows the three non vanishing contributions to s(@xmath19,@xmath20 ) calculated for the u @xmath21-edge ; the radial part as function of momentum transfer in ( a ) and the isotropic spectra in ( b ) , each for the dipole and higher multipole contributions .
the excitations due to scattering from higher multipoles contribute substantially to the total intensity already for momentum transfers of @[email protected]@xmath44 . by performing such an experiment on a single crystal and measuring the
* q*-direction dependence will give s(@xmath19,@xmath20 ) .
this then can be used for the cef analysis where each state will have a specific direction dependence .
figure[fig2](a ) shows the simulation of s(@xmath19,@xmath20 ) of the @xmath56@xmath46@xmath57 transition ( @xmath21-edge ) for the nine states of the @xmath24=4 ground state multiplet .
contributions from other valence configurations are neglected . for more detailed information about the simulation we refer to the section _
methods_. the spectra corresponding to the respective cef states are calculated for the two directions * * @xmath49**@xmath50[100 ] and * * @xmath49**@xmath50[001 ] and some of them exhibit a strong direction dependence . here
@xmath26 and @xmath27 are chosen such that the anisotropies are maximum , i.e. for the extreme cases of pure @xmath3 states ( see definition of cef states ) .
the insets in fig.[fig2](a ) show the respective two electron 5@xmath1 charge densities .
the charge densities of the pure states in fig.[fig2](a ) that appear rotational invariant do show lobes for @xmath26 and @xmath27@xmath600 or 90/ ( see fig.[figs1 ] and fig.[figs2 ] in appendix ) .
the nixs experiment was performed at large momentum transfers ( see experimental set - up in section _ methods _ ) so that the signal is dominated by higher order scattering ( _ beyond dipole _ ) .
data were taken below and above the @xmath7 transition at 5k and 25k and with successively rising temperature up to 300k .
all data shown are background corrected .
before discussing the direction dependence , we first show in fig.[fig4 ] the isotropic data ( see section _ isotropic spectra _ in the appendix for construction of isotropic spectra ) together with a simulation using the ionic 5@xmath18 configuration for the u. we can clearly observe a very good agreement , thereby establishing that the spectrum is dominated by the atomic multiplet structure .
this is important since this validates a posteriori the use of local probes ( such as core level nixs ) to test models utilizing also local irreducible representations .
perhaps more surprising is that a single ionic configuration can reproduce the spectrum so well despite the known covalency of the u. however , it has been explained by , for example , gunnarrson and schnhammer @xcite as well as degroot @xcite , that @xmath61 and @xmath62 core - level xas ( nixs ) for 4@xmath1/5@xmath1 and 3@xmath63 compounds , respectively , highlights the spectral weight of the energetically lowest lying ( and major ) configuration at the expense of those of the higher lying configurations , making the technique extremely powerful for determining the symmetry of the ground state ( see _ spectroscopy _ in the appendix ) .
we now discuss the direction dependence of the data measured above the @xmath7 transition since they are not affected by any possible impact of the @xmath7 . in fig.[fig2](b )
the nixs data of / at 25k are shown for the in - plane direction * * @xmath49**@xmath50[100 ] ( blue dots ) and out - of - plane direction * * @xmath49**@xmath50[001 ] ( red dots ) .
the error bars reflect the statistical error .
there is a large anisotropy that can be directly compared with our simulations .
a more detailed comparison of data and simulations excludes immediately the @xmath28(@xmath26 ) states with strong @xmath3=@xmath36 contributions , the @xmath40 and @xmath41 singlets with @xmath3=@xmath64 and @xmath3=@xmath65 as well as the @xmath29(@xmath27 ) states with strong @xmath3=@xmath66 weight
. actually , only singlet states with majority @xmath67 and @xmath5 or a doublet with majority @xmath68 show the correct direction dependence , i.e. red over blue ( see fig.[fig2](a ) ) . to be more quantitative
we compare the measured dichroism of about 35% ( see fig.[fig2](c ) ) with the simulated dichroism of the / wave functions in question . here
the dichroism is defined as the difference of the intensities for * * @xmath49**@xmath50[100 ] and * * @xmath49**@xmath50[001 ] relative to the peak height r@xmath54 with r@xmath54 being the intensity difference of pre - edge at @xmath1290ev and peak height at @xmath1296ev of the isotropic spectrum ( see fig.4 , described in section _ appendix _ ) .
we find that @xmath31(90/ ) ( or @xmath32(0/ ) ) and also @xmath6 reproduce the size of the anisotropy quite well , although their dichroism is with 43% or 50% slightly larger than the measured value .
a @xmath31 state of majority @xmath3=@xmath4 and @xmath5 symmetry , but with some @xmath3=@xmath36 ( @xmath31(70/ or 110/ ) ) would produce a slightly smaller dichroism of about 36% ( see fig.2(c ) ) .
the @xmath34(90/ ) ( or @xmath33(0/ ) ) , i.e. the doublet states with the highest amount of @xmath3=@xmath68 , do not yield sufficient dichroism : the dichroism of 9% is by a factor four too small and would decrease further or even change sign with increasing amount of @xmath66 .
( in the following we skip writing out the @xmath69 alternative state because of @xmath70(90/)=-@xmath69(0/ ) , i=1 or 5 . )
figure[fig3](a ) shows the @xmath71 asymmetry of the scattering function for all temperatures .
also here the error bars reflect the statistical error .
we find that within the error bars the 5k and 25k are identical .
we further find that there is no change with temperature up to 300k as is demonstrated by plotting the dichroism for all temperatures in fig.[fig3](b ) .
the boltzmann population with temperature of any state other than the @xmath31(70/-90/-110/ ) and @xmath6 state will change the direction dependence of the scattering ( compare fig.[fig2](a ) ) .
hence we conclude from the absence of any changes in the spectra up to 300k that the ground state consists mainly of the @xmath31(70/-90/-110/ ) or the @xmath6 singlet , or that one of the two singlets forms the ground state with the respective other state close in energy .
we can further estimate from the impact of thermal occupation that the states with weak dichroism like the @xmath34(90/ ) , @xmath40 and @xmath41 must be higher than 150k ( 13mev ) while states with stronger opposite anisotropy must be even higher in energy .
figure[fig3](c ) shows data taken within the plane , for * * @xmath49**@xmath50[100 ] and for two directions 22.5/ and 45/ towards [ 010 ] as well as the respective simulations for all cef states . neither below nor above the @xmath7 order transition
we can resolve any anisotropy within the statistical error bar .
this is not in contradiction with our previous findings that either one of the two singlet states @xmath31(70/-90/-110/ ) and @xmath6 forms the ground state since the asymmetries expected from simulations are rather small and most likely covered by statistics of this low count experiment .
the in - plane data even confirm the out - of - plane data when comparing the measured and simulated shape of the spectra : for example , the peak at 94ev is clearly smaller than the peak at 97ev for the simulated @xmath31(90/ ) and @xmath6 spectra ( see gray lines in fig.[fig3](c ) ) in agreement with the data while the two peaks are about the same for all other states .
for the in - plane simulation for different values of @xmath26 and @xmath27 we refer to fig.[fig6 ] ( see appendix ) .
our results of a ground state that consists mainly of @xmath31(70/-90/-110/ ) and/or @xmath6 agree well with the description of the anisotropy of the static susceptibility@xcite and the analysis of the temperature dependence of the elastic constants@xcite which are well described with a @xmath31 of majority @xmath3=+4 and -4 , a @xmath6 as first excited state and another state above 150k .
it also confirms dmft calculations that finds these two singlet states as low lying states close in energy ( see ref.@xcite and also in supplementary of kung _ _ et al.__@xcite ) but the experiment yields the additional information that the @xmath3=+4 and -4 in the @xmath2 is dominating .
we further would like to stress that linear polarized xas data at the u @xmath21 edge@xcite also agree with our findings in the sense that both , the nixs _ and _ xas dichroism , rule out the @xmath32(90/ ) ( or @xmath31(0/ ) ) , the @xmath40 or the @xmath41 as possible ground states and find no temperature dependence across the @xmath7 transition . the smaller direction dependence that lets the authors of ref.@xcite
assign the @xmath34 doublet as ground state might be due to the higher surface sensitivity of the xas experiment .
a pure @xmath31(70/-90/-110/ ) or @xmath6 or both close in energy does confront us with the dilemma that neither would break the @xmath72 in - plane symmetry as suggested by the torque@xcite , high precision x - ray@xcite , and elastoresistance@xcite results nor would an ordering out of a singlet state yield sufficient loss of entropy across the @xmath7 transition . to allow for a rank-5 @xmath73 @xmath7 parameter as in the fully microscopic itinerant approach@xcite
the inclusion of the twofold degenerate cef state of @xmath74-type is a necessity in the present more localized picture .
also the interpretation of the resonance intensity in the main a@xmath75 and other raman channels in terms of a staggered chirality density wave requires a mixing of @xmath2 and @xmath6 singlet states that support a hexadecapole - type hidden order parameter .
the model of kung _ et al .
_ contains both dominant a@xmath75 and subdominant b@xmath76 symmetry parts .
the latter involves higher energy cef states and couples to the lattice leading to a secondary orthorhombic distortion that leaves only twofold symmetry .
this causes a leakage of resonance intensity into forbidden channels @xcite .
our experiment shows that the cef components are mainly of the @xmath31(90/ ) or @xmath6 singlet type but we did not observe the theoretically maximum possible dichroism ( see fig.[fig2](c ) ) so that the data allow the presence of some other symmetry .
the mixing of an irreducible representation , other than @xmath31(70/-90/-110/ ) or @xmath6 , into the ground state can not rely on boltzmann occupation since that would have been observable in the temperature dependence of the nixs data .
however , a kondo - type mechanism where an @xmath48 kramers doublet hybridizes with the crystal - electric field manifold of the energetically more favorable @xmath77 configuration is feasible for constructing a ground state with different @xmath18 cef characters . here
@xmath78 denotes an electron in the host conduction band .
we recall that a hybridization gap of 13mev opens up below 27k@xcite which is also seen in the dmft calculation that includes these kondo processes@xcite .
the stabilization energy of this kondoesque wave function should be of the same order as the hybridization gap and the contributing cef configurations should also be within this energy range .
the bulk sensitive , u @xmath21 non - resonant inelastic x - ray scattering data of uru@xmath0si@xmath0 exhibit the atomic multiplet structure of the @xmath18 configuration .
the huge out - of - plane anisotropy shows that the symmetry of the ground state consists mainly of the @xmath31(90/ ) or @xmath6 singlet states in the u@xmath17(@xmath18 ) configuration and/or that these two states are close in energy .
the data do not exhibit any temperature dependence , neither across the @xmath7 phase transition nor in the temperature interval up to 300k , the latter setting constraints to the proximity of next higher excited states .
scenarios for constructing a ground state that is a superposition of different irreducible representations without relying on boltzmann statistics are discussed .
a high - quality single crystal of uru@xmath0si@xmath0 was grown with the traveling zone method in the two - mirror furnace in amsterdam under high purity ( 6n ) argon atmosphere .
the crystal was checked and oriented with x - ray laue diffraction for its single - crystalline nature .
the oriented crystal was cut using the spark erosion method after which the relevant surfaces [ ( 100 ) , ( 110 ) , and ( 001 ) ] were polished .
a bar - shaped piece of the single crystal was characterized by resistance measurements .
the scattering function s(@xmath19,@xmath20 ) was measured in a non - resonant inelastic x - ray scattering experiment ( nixs ) at the beamline id20 at esrf .
two monochromators [ si(111 ) and si ( 311 ) ] set the incident energy to 9690 ev and the scattered intensity was analyzed by one column of three si(660 ) crystal analyzers at an in - plane scattering angle of 2@xmath79=153/ and detected in a _ maxipix _
2d detector with an overall energy resolution of about 0.8ev .
this setting corresponds to a momentum transfer of @xmath43=9.6@xmath44 . the crystals with ( 100 ) , ( 110 ) , and ( 001 ) surfaces allowed realizing * * @xmath49**@xmath50[100],[110 ] , and [ 001 ] in specular geometry and also other directions when going off specular .
it turned out that specular geometry ( same path for photon in as for photon out ) is not necessary since @xmath49@xmath50[110 ] measured specular on the ( 110 ) crystal and 45/ off specular on the ( 100 ) crystal gave the same result .
for cooling the samples were mounted in a he flow cryostat .
the elastic line was measured before each setting to determine the zero energy transfer and exact instrumental resolution for each analyzer .
the spectra of the 45/ edges were then normalized to their pre - edge intensity .
scans over a wide energy range were taken in order to correct for the compton scattering and some minor constant background .
the compton background was fitted to a gaussian and then subtracted from the data .
why does @xmath42-edge xas or nixs resemble the @xmath18 multiplet structure and why is @xmath42-edge xas or nixs sensitive to the symmetry ?
the reason for this is that the energy order of the local configurations ( in a configuration interaction picture ) of the ground state problem and of the nixs / xas core - hole final state problem is identical .
hence , the spectral weights of the other local configurations are strongly suppressed due to quantum mechanical interference effects .
this is e.g. well explained by o. gunnarsson and k. schnhammer @xcite fig.7 for the case of ce @xmath80 xas and by de groot @xcite pages 549 and 550 for the case of 3@xmath63 transition metal @xmath81 xas . a quote from the latter : _ xps is sensitive to the charge transfer effects ........ while xas is sensitive to the symmetry of the ground state with its characteristic multiplet .
_ it is important that the relevant intra- and inter - shell coulomb interactions are of similar size in order to have the same order of configuration energies in the ground state problem as well as core - hole final state problem : this is true for the ce @xmath80 ( 3@xmath63 , 4@xmath1 ) , u @xmath21 ( 5@xmath63 , 5@xmath1 ) , transition metal @xmath81 ( 2@xmath82,3@xmath63 ) edges , but not for e.g. the ce @xmath81 ( 2@xmath82,5@xmath63 with 4@xmath1 as spectator ) or u @xmath81 ( 2@xmath82,6@xmath63 with 5@xmath1 as spectator ) since here the coulomb interaction of 5@xmath63 - 4@xmath1 or 6@xmath63 - 5@xmath1 is negligible in comparison with the 2@xmath82 - 4@xmath1 and 4@xmath1 - 4@xmath1 or 2@xmath82 - 5@xmath1 and 4@xmath1 - 5@xmath1 , respectively . for further reading
we refer to ref.@xcite and @xcite .
the simulations include / coupling and coulomb interaction and are based on an ionic model with a u@xmath175@xmath18 configuration .
the atomic values are calculated with the cowan code@xcite but the slater integrals for coulomb interactions are reduced by a constant factor to account for the screening of the moments in the solid .
the 5@xmath15@xmath1 and 5@xmath635@xmath1 reduction was adjusted to about 50% to match the experimental energy spread of the multiplet signal of the isotropic data in fig.s1 ( for construction of isotropic spectrum see below ) .
the ratio of multipole contributions was slightly adjusted by varying @xmath83@xcite . in the simulations the actual value for @xmath83 was slightly larger than according to the experimental scattering triangle because the radial part of the wave functions that enter the calculations are based on the atomic values .
for all finite values of / coupling and coulomb interaction the @xmath24=4 multiplet forms the hund s rule ground state .
the relative contributions of different angular momenta @xmath84=3,4,5 depend on the ratio of / coupling and coulomb interaction and are 1% , 14% , and 85% for our reduction factors , respectively . within the @xmath24=4 basis
we create the local eigenstates ( @xmath2 to @xmath85 ) restricted to the @xmath18 configuration by combining different states of @xmath3 considering the constraints by group theory .
the hamiltonian includes the local spin - orbit coupling and multipolar coulomb interaction , which are much larger ( up to 1 rydberg ) than the final - state core - hole lifetime ( order of 1ev ) , but neglects the effects of crystal - field , covalent bonding and band - formation of the crystal , which will be smaller or of the same order of magnitude as the core - hole lifetime .
the calculations are performed using the _ quanty _ code@xcite . to account for instrumental resolution , lifetime effects , and interference effects with the continuum
the multiplet lines are broadened with a gaussian ( fwhm=0.8ev ) , a lorentzian ( fwhm=1.3ev ) and a mahan - type line shape ( with an asymmetry factor 0.18 and an energy width of the continuum of 1000ev ) in order to mimic the asymmetry due to the itinerancy .
the charge densities are calculated for two electrons .
the surface thus tells how much charge can be found in a specific direction . for a system with one electron
this would be a plot of the orbital that is occupied . for a system that can be represented by a single slater determinant it shows a `` sum '' of the different orbitals occupied .
the color is related to the spin density with up as red , down as blue and zero as gray . for uru@xmath0si@xmath0 often @xmath86 coupling
is assumed whereby @xmath84 @xmath87 @xmath24 and @xmath3 are all good quantum numbers . doing so results in density plots with much more features than in the present manuscript .
this assumption basically is equivalent to saying that @xmath88 , @xmath89 , and @xmath90 slater integrals are infinitely larger than the spin - orbit coupling .
that approximation is not valid and spin - orbit coupling mixes states @xmath91 with states @xmath84 + 1 @xmath87 - 1 or @xmath84 - 1 @xmath87 + 1 .
this mixing is included in our calculations .
furthermore we know that both the multipole part of the coulomb interaction as well as the spin - orbit interaction are not ( not really ) screened in a solid .
in other words the u 5@xmath1 shell in uru@xmath0si@xmath0 is in - between @xmath91 and @xmath92 coupling and was taken in account .
the isotropic spectrum is given by the trace of the conductivity tensor .
for dipole transitions ( k=1 ) this tensor can be written as a 3x3 matrix with two independent diagonal elements in @xmath23 symmetry .
however , for higher multipoles the conductivity tensor has also a higher dimension , i.e. 7x7 for octupole ( k=3 ) and 11x11 for dotriacontapole ( k=5 ) .
here we obtain the _ experimental _ isotropic spectrum containing these three relevant conductivity tensors by combining 10 independently measured directions .
the _ calculated _ isotropic spectrum is obtained by averaging over all cef states .
the red line in fig.[fig4 ] is the simulation of the isotropic data after optimizing the respective parameters .
* out - of - plane anisotropy : * figures1(a ) and ( b ) show the out - of - plane anisotropy in the u @xmath21-edge nixs spectra of the @xmath28(@xmath26 ) and @xmath29(@xmath27 ) wave functions for values of @xmath26 and @xmath27 between 0 and 90/. the insets show the respective charge densities . please note that -@xmath31(90/)=@xmath32(0/ ) and @xmath32(90/)=@xmath31(0/ ) .
the same holds for the @xmath29 , i.e. -@xmath34(90/)=@xmath33(0/ ) and @xmath33(90/)=@xmath34(0/ ) .
* in - plane anisotropy : * figures2(a ) and
( b ) show the in - plane anisotropy in the u @xmath21-edge nixs spectra of the @xmath28(@xmath26 ) and @xmath29(@xmath27 ) wave functions for values of @xmath26 and @xmath27 between 0 and 90/. the insets show the respective charge densities . according to simulations none of the states
exhibits a noteworthy in - plane anisotropy in the scattering intensity although the charge densities appear anisotropic and although there are multipole contributions in s(@xmath19,@xmath20 ) .
this can be understood when considering that the scattering process at the u@xmath94-edge is from 5@xmath63@xmath465@xmath1 , i.e. we scatter from a state with a complicated shaped charge distribution to another one with a non - simple shape . for momenta @xmath49 parallel to [ 100 ] and
for @xmath49 turned towards [ 010 ] by 22.5/ and 45/ , i.e. @xmath49@xmath50[100 ] ( blue ) , @xmath49@xmath50[520 ] ( dark green ) and @xmath49@xmath50[110 ] ( light green ) for the two states ( a ) @xmath28(@xmath26 ) and ( b ) @xmath29(@xmath27 ) for different @xmath3 admixtures expressed in terms of @xmath26 and @xmath27 . ]
we thank areem nikitin for for characterizing the sample by transport .
m.s . and a.s .
benefited from support of the german funding agency dfg ( project 600575 ) .
we further acknowledge esrf for provision of synchrotron radiation facilities ( proposal hc1533 and hc2252 ) .
65ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1103/revmodphys.81.807 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.81.1551 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.79.1015 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.55.2727 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.56.185 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.116402 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.111.127002 [ * * , ( ) ] link:\doibase 10.1038/ncomms5326m3 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.105.216409 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1073/pnas.1005892107 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.43.12809 [ * * , ( ) ] \doibase doi:10.1038/nphys522 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.113.266405 [ * * , ( ) ] link:\doibase 10.1126/science.1259729 [ * * , ( ) ] link:\doibase 10.1143/jpsj.79.064719 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1038/ncomms5188 m3 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.83.1301 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.033103 [ * * , ( ) ] link:\doibase 10.1103/physrevb.94.045121 [ * * , ( ) ] link:\doibase 10.7566/jpsj.85.062001 [ * * , ( ) ] , link:\doibase 10.1103/physrevb.68.052508 [ * * , ( ) ] link:\doibase 10.1038/nphys1392 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.58.1467 [ * * , ( ) ] link:\doibase 10.1103/physrevb.35.5260 [ * * , ( ) ] link:\doibase 10.1103/physrevb.88.195150 [ * * , ( ) ] link:\doibase 10.7566/jpsj.82.013601 [ * * , ( ) ] link:\doibase 10.1103/physrevb.71.054415 [ * * , ( ) ] link:\doibase 10.1143/jpsj.81.114713 [ * * , ( ) ] link:\doibase 10.1143/jpsj.80.084702 [ * * , ( ) ] @noop link:\doibase 10.1103/physrevlett.73.1027 [ * * , ( ) ] link:\doibase 10.1143/jpsj.74.765 [ * * , ( ) ] link:\doibase 10.1143/jpsj.66.3251 [ * * , ( ) ] link:\doibase 10.1143/jpsj.63.736 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1038/nature11820 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.114.236401 [ * * , ( ) ] link:\doibase 10.1143/jpsj.68.3394 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.046401 [ * * , ( ) ] link:\doibase 10.1103/physrevb.91.201108 [ * * , ( ) ] link:\doibase 10.1038/srep17937 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.066405 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.84.075134 [ * * , ( ) ] link:\doibase 10.1103/physrevb.72.045136 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/revmodphys.82.847 [ * * , ( ) ] link:\doibase 10.1103/physrevb.81.193104 [ * * , ( ) ] link:\doibase 10.1103/physrevb.81.195104 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.077401 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.28.4315 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1038/ncomms7425 [ * * , ( ) ] link:\doibase 10.1038/nphys2330n1 [ * * , ( ) ] link:\doibase 10.1080/14786435.2013.861615 [ * * , ( ) ] @noop ( ) link:\doibase 10.1103/physrevb.85.165113 [ * * , ( ) ] link:\doibase 10.1088/1742 - 6596/712/1/012001 [ * * , ( ) ]
|
the second - order phase transition into a hidden order phase in uru@xmath0si@xmath0 goes along with an order parameter which is still a mystery , despite 30 years of research .
however , it is understood that the symmetry of the order parameter must be related to the symmetry of the low lying local electronic @xmath1-states . here
we present results of a novel spectroscopy , namely core - level non - resonant inelastic x - ray scattering ( nixs ) .
this method allows for the measurement of local high - multipole excitations and it is bulk sensitive .
the observed anisotropy of the scattering function unambiguously shows that the 5@xmath1 ground state wave function is composed mainly , but essentially not purely , of the @xmath2 with majority @xmath3=@xmath4+@xmath5 and/or @xmath6 singlet states .
|
this research used resources of the national energy research scientific computing center ( nersc ) , which is supported by the office of science of the united states department of energy under contract no .
de - ac02 - 05ch11231 , and the oak ridge leadership computing facility located in the national center for computational sciences at oak ridge national laboratory , which is supported by the office of science of the united states department of energy under contract no .
de - ac05 - 00or22725 .
these resources were obtained through the advanced scientific computing research ( ascr ) leadership computing challenge ( alcc ) .
this work is supported by the laboratory directed research and development program at sandia national laboratories .
sandia national laboratories is a multi - program laboratory managed and operated by sandia corporation , a wholly owned subsidiary of lockheed martin corporation , for the u.s .
department of energy s national nuclear security administration under contract de - ac04 - 94al85000 .
36 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , , , , * * , ( ) . , , , , , , * * , ( ) . , , , ,
* * , ( ) . ,
* * , ( ) . , , , , * * , ( ) . , , , * * , ( ) .
, , , , * * , ( ) . , , , , , , , * * , ( ) . , , , , , , * * , ( ) . , , , * * , ( ) . , , , , , * * , ( ) . , , , , , , , , * * , ( ) . , , , , ,
* * , ( ) . , , , , , , * * , ( ) .
, , , , * * , ( ) . , , , , * * , ( ) . , , , , , , ,
* * , ( ) . , , , , , , * * , ( ) .
, , , , * * , ( ) . , , , ,
* * , ( ) . , , , * * , ( ) . ,
* * , ( ) . , * * , ( ) . , , , * * , ( ) . , , , , * * , ( ) . ,
* * , ( ) . , , , , * * , ( ) . ,
* * , ( ) . .
, , , * * , ( ) . , * * , ( ) .
in this supplementary material , we include voronoi constructions of nps in the top layer for the other 2 systems that we have studied .
they underwent various evaporation schemes , as is detailed in the captions of each figure .
figure [ voronoi_np ] is for a system with a square cross section ( @xmath40 ) .
figure [ voronoi_solcon ] is for another system that had a similar area , but the sizes of the liquid / vapor interface were adjusted to accommodate an integer number of nps in both @xmath18 and @xmath19 directions ( @xmath102 and @xmath103 ) , respectively .
|
while evaporating solvent is a widely used technique to assemble nano - sized objects into desired superstructures , there has been limited work on how the assembled structures are affected by the physical aspects of the process .
we present large scale molecular dynamics simulations of the evaporation - induced assembly of nanoparticles suspended in a liquid that evaporates in a controlled fashion .
the quality of the nanoparticle crystal formed just below the liquid / vapor interface is found to be better at relatively slower evaporation rates , as less defects and grain boundaries appear .
this trend is understood as the result of the competition between the accumulation and diffusion times of nanoparticles at the liquid / vapor interface . when the former is smaller , nanoparticles are deposited so fast at the interface that they do not have sufficient time to arrange through diffusion , which leads to the prevalence of defects and grain boundaries .
our results have important implications in understanding assembly of nanoparticles and colloids in non - equilibrium liquid environments .
* i. introduction * large , defect - free crystalline arrays of nanoparticles ( nps ) are critical for many technologically important ultra thin film materials including filters , sensors , optical devices , and magnetic storage media.@xcite one common method for dispersing nanoparticles is to suspend them in solution , spread the suspension on a surface , and then evaporate the solvent.@xcite this technique has been used to assemble nps into desired structures such as nanoclusters , rings , wires , stripes , films , and superlattices .
@xcite the factors that influence the assembled structure include the evaporation kinetics of the solvent , the flow properties and concentration of nps in the solution , the np - np interactions , the interactions between nps and the liquid / vapor interface , and the wetting / dewetting behavior of the suspension on the solid surface.@xcite one ubiquitous example is the `` coffee - ring '' stain left behind after a particle - laden drop has evaporated .
a simple and beautiful physical picture of this class of phenomena was first clarified by deegan et al.@xcite when the drop dries , a flow of liquid from the interior of the drop to its edge emerges to replenish the liquid that has evaporated from the region around the pinned contact line .
this flow convects particles to the drop s edge , where they form ring - like depositions .
the model was tested recently by chen et al . on the molecular scale with molecular dynamics ( md ) simulations.@xcite recently , marin et al . demonstrated an order - to - disorder transition of the particle organization in coffee - ring stains.@xcite in the early stage of evaporation , the deposition speed of particles at the contact line is low and they have time to arrange into an ordered superlattice by brownian diffusion . however , at the end of evaporation , because of the increase of the flux velocity toward the contact line , particles are deposited at a high speed and jammed into a quenched disordered phase as the result of rush - hour " traffic .
the experiment clearly showed the importance of deposition / accumulation and diffusive time scales on controlling the particle ordering during solvent evaporation .
bigioni et al . showed earlier that particle accumulation and diffusion times determine the assembly behavior and growth kinetics of monolayers of gold nps with diameters @xmath0 nm at a liquid / vapor interface.@xcite they found that in order to induce np assembly at the interface , the evaporation rate of the solvent has to be rapid and an attractive interaction between the nps and interface is required .
the latter was induced by the excess of dodecanethiol in a toluene solution where gold nps were suspended .
after the assembly is initiated , the growth kinetics of each crystalline domain can be exponential , sub - exponential , or linear in time , depending on the ratio of a diffusive length scale to the domain size . by tuning the concentration of dodecanethiol , bigioni et al .
were able to make long - range ordered arrays of gold nps over macroscopic areas.@xcite the work of of bigioni et al
. showed that evaporation generally needs to be fast so that the accumulation of nps at the interface passes certain critical density , leading to the nucleation and growth of 2-dimensional ( 2d ) crystals .
however , if evaporation is too fast , the crystalline quality may diminish because nps do not have sufficient time to arrange via diffusion , but instead quenched disordered structures will emerge , similar to the rush - hour " effect observed in the coffee - ring stains .
this trend was actually observed earlier in an experiment of i m and park using larger polystyrene colloids with diameters @xmath1 nm,@xcite where the evaporation rate was tuned by varying the temperature .
they found that there exists an optimum evaporation temperature that yields particle arrays of the highest quality at the water / air interface . when the evaporation temperature is higher than the optimum one , the accumulation rate of colloids at the interface is too large , leading to defects in the array .
it is thus interesting to determine if the effect of evaporation rate on particle assembly persists to smaller nps in the fast - evaporation regime , which nicely fits the accessible length and time scales of md simulations .
this was the motivation of our work reported here . to anticipate the final results
, we will show that there exits an optimum rate , determined by the diffusive times of nps along the interface , at which the 2d crystal of nps formed near the interface has the highest quality even without the annealing " step .
the remainder of this paper is organized as follows .
the simulation methods are briefly introduced in sec .
. then sec .
iii is devoted to results and discussion .
conclusions are included in sec .
iv . *
ii .
simulation methods * we modeled the liquid solvent and its vapor with atoms interacting through a standard lennard - jones ( lj ) potential , @xmath2 $ ] , where @xmath3 is the distance between two atoms , @xmath4 the unit of energy , and @xmath5 the diameter of atoms .
the potential was truncated and shifted to 0 at @xmath6 .
the mass of each lj atom is @xmath7 and a time unit @xmath8 can be defined as @xmath9 .
the nps of diameter @xmath10 were assumed to consist of a uniform distribution of atoms interacting with a lj potential .
for spherical particles , their mutual interaction can be determined analytically by integrating over all the interacting lj atom pairs between the two particles.@xcite the resulting potential depends on @xmath11 , the inter - np distance , and a hamaker constant @xmath12 that characterizes the interaction strength .
we used a standard value @xmath13 and a cut - off @xmath14 .
this led to a purely repulsive np - np interaction , which physically corresponds to adding a short surfactant coating on nps to avoid flocculation.@xcite the interaction between lj atoms and nps was determined similarly by integrating the interaction between a single lj atom and those within a np .
we set the corresponding hamaker constant @xmath15 , which with a cut - off of @xmath16 resulted in nps that were fully solvated and dispersed in the lj solvent.@xcite the simulation cell was a rectangular box of dimensions @xmath17 .
the liquid / vapor interface was parallel to the @xmath18-@xmath19 plane , in which periodic boundary conditions were employed .
in the @xmath20 direction , the lj atoms and nps were confined by two flat walls at @xmath21 and @xmath22 , respectively .
the depth of the liquid was at least @xmath23 and the vapor was at least @xmath24 thick before evaporation .
the resulting liquid and vapor densities are @xmath25 and @xmath26 , respectively .
note that lj monomers have a vapor density higher than most real fluids ( fig .
[ assembly](a ) ) .
this implies that the vaporization rate of lj liquids is quite large and they can evaporate very fast.@xcite to provide a rough mapping of the evaporation rates ( number fluxes of solvent atoms ) in our simulations to experimental values , we take @xmath27 nm and @xmath28 ps , the slowest evaporation rate we can simulate with md is at the order of @xmath29 , corresponding to @xmath30 .
this is about an order of magnitude higher than the maximum rate at which water evaporates , which is not surprising since water has an unusually high boiling temperature and surface tension compared with the lj liquids studied here .
all simulations were performed using lammps.@xcite the equations of motion were integrated using a velocity - verlet algorithm with a time step @xmath31 . during the equilibration ,
the temperature @xmath32 was held at @xmath33 by a langevin thermostat weakly coupled to all lj atoms with a damping constant @xmath34 .
once the liquid / vapor interface was equilibrated , the langevin thermostat was removed except for those liquid atoms within @xmath35 of the lower wall at @xmath21 .
we confirmed that the evaporation occurring at the interface was not affected by the thermostat.@xcite * iii .
results and discussion * * a. effect of evaporation rate *
most of our simulations started with 668 nps dispersed in a liquid that contained about 7.1 million lj atoms,@xcite as shown in fig .
[ assembly](a ) .
the volume fraction of nps was 0.205 , which was based on the diameter of bare nps .
if we took into account the effect of a layer of solvent of thickness @xmath36 that was effectively attached to nps,@xcite the effective volume fraction of nps would be even higher , roughly by a factor of 1.16 .
we have studied systems of which the liquid / vapor interface has various aspect ratios .
since the nps were spherical and with a short - range isotropic interaction , their ordering at the interface was expected to be hexagonal .
thus to reduce the finite size effects , we focused on a system with a cross - section that was commensurate with hexagonal packing , i.e. , @xmath37 ( figs .
[ assembly ] and [ voronoi_main ] ) . @xmath38 and @xmath39 were chosen so that the lattice spacing in a perfectly ordered hexagonal array would be the same as that found after some of the solvent was evaporated in a system that had a square cross - section with @xmath40 .
however , this latter system has a larger interfacial area than the previous one . to reduce the potential influence of areal size , we studied another system of which @xmath39 was increased to @xmath41 while @xmath38 was kept at @xmath42 , so that not only the interfacial area was close to that of the square interface , but also the aspect ratio of the interface favored a hexagonal packing with a lattice spacing found in the previous two systems .
our results indicated that the size and aspect ratio of the interface had negligible effects on the particle packing .
below we mainly discuss the results for the first system and include the results of other two systems in supplementary material.@xcite the side view of the np distribution in the equilibrium solution is shown in fig .
[ assembly](e ) , where the dashed line indicates the location of the liquid / vapor interface .
nps are almost uniformly distributed in the solvent .
note that during all our simulations , including the evaporation runs described below , all nps were fully solvated in the liquid solvent .
nps within @xmath10 of the interface were grouped together and referred as the top layer . at equilibrium
their areal coverage is about 45% , which is below the critical coverage 70% at the 2d hexatic phase transition.@xcite as shown in fig .
[ assembly](i ) , the layer is clearly disordered .
to quantitatively characterize the in - plane packing geometry , we have defined an order parameter @xmath43 for each np . a voronoi construction was performed and those nps with 6 neighbors were identified since in a close 2d packing each particle
was expected to have 6 nearest neighbors forming a hexagon . the angle @xmath44 corresponding to the @xmath45-th side of the hexagon was found and @xmath43 was computed as @xmath46 .
we found that @xmath43 was typically in the range of @xmath47 to @xmath48 .
so we normalized @xmath43 by @xmath48 and the normalized @xmath49 was thus in the range of @xmath47 to @xmath50 .
@xmath51 indicates a perfect hexagonal packing , while @xmath52 indicates situations far from a hexagonal lattice .
the voronoi construction also showed that all other nps had either 5 or 7 neighbors and they were assigned @xmath53 to indicate that their packing configuration was not hexagonal . in fig .
[ assembly](i)-(l ) , nps are colored according to their values of @xmath49 ( nps with @xmath54 are colored red and with @xmath55 are colored blue ) . as shown in fig .
[ assembly](i ) , in an equilibrated solution nps near the interface are randomly distributed without any ordering , which is also clear from the voronoi construction shown in fig .
[ voronoi_main](a ) .
evaporation of the solvent was modeled by introducing a deletion zone from @xmath56 to @xmath57 .
all or a certain number of vapor atoms in the deletion zone were removed every @xmath58 . the former mimicked evaporation into a vacuum which proceeded at the maximum rate , while the latter simulated evaporation at a fixed rate @xmath59 , namely the rate of removal of solvent atoms out of the simulation box .
results in figs .
[ assembly](b ) , ( f ) , and ( j ) are for a system of which the solvent had evaporated at @xmath60 since @xmath61 . during evaporation ,
the liquid film where nps were dispersed shrunk , and nps started to accumulate below the liquid / vapor interface that impinged from above . in this case
the vapor density was only slightly reduced compared with its equilibrium value ( fig .
[ assembly](b ) ) throughout the whole simulation and the interface always remained close to equilibrium .
as a result , the rate @xmath59 was roughly the same as the actual outward flux of solvent atoms near the liquid / vapor interface or in the vapor .
the side view of the np distribution in fig .
[ assembly](f ) shows that nps formed layers where their concentration had peaks . at @xmath62 , 52% of the solvent ( @xmath63 million atoms )
had evaporated , and the volume fraction of nps increased to 0.35 .
at least 3 layers of nps were formed as shown in fig .
[ assembly](f ) , with 171 nps in the top layer .
figure [ assembly](j ) shows the in - layer structure of these nps , from which a hexagonal packing is clearly visible .
this is also evident from the corresponding voronoi construction ( fig .
[ voronoi_main](b ) ) .
note that the number of layers induced by evaporation depends on the volume fraction of nps and the interfacial area . in many experiments , np solutions
are quite dilute and only one layer of nps forms , but more layers were also experimentally observed,@xcite as in our simulations .
the structures in the np layers below the top one were essentially similar but with a diminishing crystalline quality as the distance from the interface increased .
two consecutive layers were roughly in - registry as in fcc or hcp crystals , but we had too few layers to distinguish the two crystal structures . for clarity ,
we focus on the in - plane structure in the top np layer below . the results in figs .
[ assembly](c ) , ( g ) , and ( k ) are for a system with the same starting state , but with the solvent evaporating into a vacuum since @xmath61 . here
the vapor was quickly depleted ( fig .
[ assembly](c ) ) and the evaporation at the liquid / vapor interface proceeded at the maximum rate , which decreased with time and eventually reached a plateau value around @xmath64 ( see fig.[evap_rate ] ) .
the actual flux of solvent atoms at the interface was lower than @xmath59 for the first @xmath65 during which the vapor depletion occurred , and then the two were almost identical .
it only took @xmath66 to evaporate roughly the same amount ( 52% ) of solvent as in the previous case with a fixed rate .
the layering of nps was obvious from the side view ( fig .
[ assembly](g ) ) .
however , each layer was composed of several 2d grains separated by grain boundaries , where the packing deviated significantly from a hexagonal lattice as indicated by colors based on the values of @xmath49 ( fig .
[ assembly](k ) ) and the corresponding voronoi construction ( fig .
[ voronoi_main](c ) ) for nps in the top layer . comparing figs . [ assembly](j ) and ( k ) ( or figs . [ voronoi_main](b ) and ( c ) ) shows that the slower evaporation led to a np packing with a better crystalline order .
the same trend was also found for the other two systems evaporating at various rates ( see figs .
s1(a ) and ( b ) , and figs . s2(a ) and ( b ) in supplementary material).@xcite this effect of evaporation rate @xmath59 on the assembly quality can be understood on the basis of two time scales : one describes how fast nps accumulate at the interface , and another is determined by the diffusivity of nps along the interface . in the steady state
, the liquid / vapor interface recedes almost uniformly , with a receding velocity @xmath67 that is related to @xmath59 through @xmath68 , where @xmath69 is the liquid density . using a model discussed by bigioni _ et .
al._,@xcite a time scale @xmath70 can be defined to describe the speed of np accumulation at the interface , where @xmath71 is the np concentration in the solution .
after nps reach the top layer , they primarily diffuse along the interface . we denote the diffusive time scale as @xmath72 , which can be estimated as @xmath73 with @xmath74 as the diffusion coefficient . for @xmath10 and @xmath75,@xcite
the diffusive time is @xmath76 .
nps accumulate faster than they diffuse in the limit @xmath77 , where the growth of 2d crystals is induced primarily by np addition .
as it is unlikely that crystalline grains in different spatial regions have orientations that match , this leads to the occurrence of grain boundaries when grains meet .
however , in the opposite limit @xmath78 , nps are added to the top layer slowly and small crystal regions can act as seeds which then grow larger through np diffusion . in this
case defects and grain boundaries are less likely to appear .
the np layer also has more time to relax into a uniform hexagonal lattice even if imperfections in the packing occur . for the system shown in figs .
[ assembly](k ) and [ voronoi_main](c ) , the solvent evaporated fast and the interface velocity @xmath67 was about @xmath79 , which leads to @xmath80 .
the value of @xmath81 has a range since the np concentration @xmath71 increased as the solvent evaporated . for figs .
[ assembly](j ) and [ voronoi_main](b ) where @xmath59 was fixed at @xmath82 , @xmath67 was about @xmath83 , which gives @xmath84 . as expected , @xmath85 in the fast evaporation ( into a vacuum ) , while @xmath86 in the slow evaporation ( at a small fixed rate ) .
since @xmath81 and @xmath59 are related through @xmath87 , by equating @xmath81 to @xmath72 one can estimate a critical rate @xmath88 .
when @xmath89 , an almost defect - free hexagonal np array is expected ; while for @xmath90 , grain boundaries and point defects will appear in the 2d crystal induced by evaporation .
for our simulations , @xmath91 .
when the solvent evaporated into a vacuum , the evaporation rate was always larger than @xmath92 , leading to np layers with defects and grain boundaries .
however , when @xmath59 was reduced and fixed at a value below @xmath92 , the assembly quality in np layers was greatly improved .
furthermore , from experiments it is already known that if the evaporation is too slow , then the density of nps at the interface never reaches the critical nucleation density since nps have abundant time to diffuse back into the bulk of the solution , and assembly is not initiated in the first place.@xcite combined with this observation , our simulations indicate that there exits an optimum evaporation rate @xmath93 at which the np arrays formed near the liquid / vapor interface are of the highest quality . though it is very difficult to precisely calculate @xmath93 from simulations ,
our results indicate that @xmath93 should be about an oder of magnitude smaller than @xmath92 since only a few defects and grain boundaries were observed at @xmath94 ( fig .
[ voronoi_main](b ) and fig .
s1(a ) in supplementary material),@xcite while more were seen at @xmath95 ( fig .
s2(a ) in supplementary material).@xcite therefore , a reasonable estimate is @xmath96 .
* b. effect of annealing * since the volume fraction of nps was only 0.35 in fig .
[ assembly](c ) and still below the critical fraction of hard - sphere crystallization ( @xmath97),@xcite we expected that the layering structure in fig . [ assembly](g ) would be destroyed when the evaporation was stopped and the system was allowed to relax .
however , this was not the case on md time scales , and the layering became even more dramatic after the evaporation ceased at @xmath98 and the system had re - requilibrated , as shown in fig .
[ assembly](d ) and ( h ) at @xmath99 .
the robustness of layering was partly due to the fact that some solvent evaporated to fill the vapor - depleted region above the liquid and to re - establish the liquid / vapor equilibrium .
this increased the volume fraction of nps in the solution to 0.38 .
furthermore , melting requires the diffusion of nps in the direction perpendicular to the layers , which was hindered by the ordered structure in each layer and the commensurability between layers .
our results thus indicate that the system was dominated by non - equilibrium dynamics with long relaxation and equilibration times , much larger than those achievable with md simulations .
the organization of nps in the top layer was greatly improved after relaxation .
the grain boundaries in this layer completely disappeared after approximately @xmath100 and nps formed an almost perfect hexagonal lattice , as shown in fig .
[ assembly](l ) and fig .
[ voronoi_main](d ) ( also see figs .
s1(c ) and s2(c ) in supplementary material).@xcite this finding is consistent with an experimental observation that alternative evaporation and relaxation produce 2d crystals of nps with less defects.@xcite it reveals the importance of relaxation that allows imperfections in the np packing to heal via structural re - arrangement , which only gradually proceeds through the diffusion of nps .
the improvement of the crystalline quality was also observed for the other layers below the top one after annealing , though over the time scale of simulations they were still not completely ordered . for nps with diameters at the nanometer scale and without mutual attractions as studied here , their diffusion is relatively fast and thus the time scale of structural re - arrangement is accessible with md .
however , it can be imagined that for larger nps or colloidal particles in the micrometer range and with attractive inter - particle interactions , diffusion becomes much slower and it will take much longer time to remove defects and grain boundaries .
this is one of the reasons that they are frequently observed in assembly experiments.@xcite * c. effect on evaporation rate of assembled nanoparticle layers * when nps crystallize below the liquid / vapor interface , they slow down the evaporation , since liquid atoms need to circumvent nps to reach the interface and then evaporate .
this blockage effect on evaporation is shown in fig .
[ evap_rate](a ) , where the rate @xmath59 is plotted against time for a neat solvent and the np solution that both evaporated into a vacuum . at earlier times , nps were dilute near the interface and liquid atoms
essentially did not feel their presence .
therefore @xmath59 was roughly the same for the two systems . at later times , since nps started to form a dense layer below the interface , @xmath59 for the np solution became smaller than that for the pure solvent .
the factor of reduction was 25% when @xmath59 reached the plateau values in the two systems at @xmath101 .
however , the reduction in @xmath59 is not as dramatic as what a naive estimate of the areal reduction of the liquid / vapor interface would indicate .
the reason is that nps still sit below the interface in our simulations and their top sides only barely touched the liquid / vapor transition zone . as illustrated in fig .
[ evap_rate](b ) , only liquid atoms attached to the surface around the north pole of nps were affected in terms of evaporation . at several @xmath5 away from this region ( outside the box enclosing the north pole of the immersed np in fig .
[ evap_rate](b ) ) , the distance from the interface to np surface quickly increases beyond a few diameters of lj atoms , and the evaporation of the liquid there was not influenced by the presence of nps .
however , if the np layer straddled the liquid / vapor interface ( see the straddling np in fig .
[ evap_rate](b ) ) , we would expect @xmath59 to decrease by an order of magnitude compared with the case of a neat solvent . in this case
, the interfacial area is indeed reduced by nps .
the evaporation can not occur in the area covered by nps and will slow down significantly .
vs. time for a np solution ( solid red line ) and a neat solvent ( dotted green line ) , showing the blockage effect of nps on evaporation .
( b ) a schematic illustration of how the evaporation is affected by nanoparticles immersed just below or straddling the liquid / vapor interface : only the evaporation of the solvent in the area enclosed by the black boxes is blocked.,width=312 ] * iv .
conclusions * to summarize , we have studied the assembly of nps in a solution induced by solvent evaporation with md simulations .
our results showed that nps formed layers which started from below the liquid / vapor interface .
the structure in each layer found in simulations is similar to that observed in many experiments,@xcite i.e. , a close - packed hexagonal 2d crystal .
interestingly , we found that the assembly quality was better when the evaporation rate was relatively slower .
this trend was understood with a simple physical picture based on two time scales .
the time scale associated with np accumulation in the top layer , which is controlled by the evaporation rate , needs to be larger than the diffusion time of nps along the interface in order to form a defect - free 2d crystal .
otherwise , growth of crystals at different locations leads to grains with different orientations , and grain boundaries and point defects appear when grains meet . since it is known that evaporation needs to fast enough to initiate interfacial np assembly in the first place ,
our simulations indicate that there exits an optimum evaporation rate at which the packing quality of np arrays at the liquid / vapor interface is the best .
if evaporation proceeds at a higher rate , a relaxation process , which can have a large time scale depending on the particle size and interaction , is needed to anneal " the crystal and remove the imperfections in the packing .
our simulations also showed the blockage effect on evaporation because of the presence of the dense np layer near the interface .
the effective area of the interface was reduced and evaporation slowed down compared with a neat solvent . however , in our simulations nps were still immersed in the solvent and only the solvent atoms in the region around the top of nps were actually affected in terms of evaporation and therefore only a minor reduction in the evaporation rate was observed .
|
dwarf galaxies ( @xmath5 - 18 ) are the most common objects in the nearby universe ( ferguson & binggeli 1994 ) .
their importance resides in the fact that they represent in cold dark matter models the building blocks of hierarchical galaxy formation ( e.g. white & rees 1978 , white & frenk 1991 ) .
their study is thus fundamental for constraining models of galaxy formation and evolution .
+ observations of dwarf galaxies , necessarily limited to the nearby universe , revealed however a more complex origin than that predicted by models , making this class of galaxies even more interesting . among dwarf galaxies , dwarf ellipticals ( de ) and the less luminous dwarf spheroidals ( dsph )
are more common than star forming i m and bcds ( ferguson & binggeli 1994 ) .
firstly thought as the low - luminosity extension of bright ellipticals , several observational evidences indicate that they form an independent class of objects ( bender et al .
dwarf and giant ellipticals have sersic light profiles of index @xmath6 , with @xmath7 progressively increasing with luminosity ( graham & guzman 2003 ; gavazzi et al .
the colour magnitude relations measured using resolved stars in local group dwarfs spheroidals ( mateo 1998 ; grebel 1999 ) , the optical ( conselice et al .
2003a ) and uv ( boselli et al .
2005a ) integrated color magnitude relations , the subsolar [ @xmath8/fe ] ratios ( van zee et al .
2004a ) and the uv to near - ir spectral energy distribution of de in the virgo cluster ( gavazzi et al .
2002a ) , however , all indicate a more gradual star formation history with recent episodes of activity in these low - luminosity quiescent systems than in massive ellipticals .
this constitutes the first evidence against their very old origin .
a further disagreement with model predictions is that , although more frequent than bright galaxies , their number density is still significantly smaller than predicted by hierarchical models of galaxy formation for the field luminosity function ( kauffmann et al .
1993 ; cole et al . 1994
; somerville & primack 1999 , nagashima et al . 2005 ) or for the local group ( klypin et al . 1999 ; bullock et al . 2000 ) .
+ the strong similarities in the structural properties of star forming and quiescent dwarf galaxies in the nearby universe , namely their similar optical morphology and light profiles ( both roughly exponentials ) , suggested that dwarf ellipticals might result from gas removal and subsequent suppression of star formation in gas rich dwarf galaxies .
gas removal might result either from its blowout due to the kinetic energy injected into the ism by supernova explosion following a strong burst of star formation ( dekel & silk 1986 , vader 1986 , yoshii & arimoto 1987 ) , gas exhaustion through subsequent episodes of star formation ( davies & phillipps 1988 ) , or from external perturbations induced by the hostile environment in which galaxies evolve .
external perturbations include tidally induced mass loss in high speed encounters ( moore et al . 1998 ) , tidal stirring ( mayer et al .
2001a , 2001b ) and ram - pressure stripping induced by nearby companions ( lin & faber 1983 ) or by the hot intergalactic medium in massive clusters ( van zee et al .
+ several observational evidences favor the environmental scenario against the gas blowout due to supernova explosions .
the clearest indication that the environment plays an important role in the formation and evolution of dwarfs is the morphology segregation effect ( dressler 1980 ) which extends to low - luminosity systems ( binggeli et al . 1988 ; 1990 ; ferguson & binggeli 1994 ) .
furthermore it has been recently claimed that the removal of the ism through supernova winds is quite difficult in low - luminosity , dark matter dominated systems ( mac low & ferrara 1999 ; ferrara & tolstoy 2000 ; silich & tenorio - tagle 1998 , 2001 ) . the discrete star formation history of several dwarf galaxies in the local group , as deduced by the analysis of their colour magnitude relation ( mateo 1998 ; grebel 1999 ) , is a further indication that dwarf spheroidals can retain their ism through several episodes of star formation . +
another observational evidence favoring the transformation of star forming galaxies into quiescent dwarf ellipticals is the presence of rotationally supported ( pedraz et al .
2002 ; geha et al . 2003 ; van zee et al .
2004b ) and/or hi gas rich ( conselice et al . 2003b ; van zee et al .
2004b ) de in the virgo cluster .
recent studies based on sdss imaging and spectroscopic data have shown that @xmath9 50 % of the bright end of the de galaxy population in the virgo cluster is characterized by disk features such as spiral arms or bars , this fraction decreasing down to @xmath9 5% at lower luminosities ( lisker et al .
2006a ; see also graham et al .
meanwhile @xmath9 15% of the bright de in virgo have blue centers revealing a recent activity of star formation . from a statistical point of view
, the line - of - sight velocity distribution of de inside clusters is similar to that of late - type galaxies suggesting a recent infall ( binggeli et al .
1993 ; conselice et al .
+ not all observational evidences , however , are consistent with the transformation of dwarf star forming galaxies into dwarf ellipticals under the effect of the environment . to reproduce the color magnitude relation of dwarf ellipticals ,
it has been claimed that magellanic irregulars should have faded @xmath9 1.5 magnitudes in the b band thus reaching surface brightnesses weaker than @xmath10 @xmath11 25 mag arcsec@xmath12 , values significantly smaller than the observed ones ( bothun et al .
most of the bright dwarf ellipticals in the virgo cluster have a bright nucleus ( binggeli et al .
1985 ; ferguson & binggeli 1994 ) of small size , as shown by hst observations ( @xmath9 4 pc , cte et al .
2006 ) while dwarf irregulars do not . while both dwarf irregulars and dwarf ellipticals follow different metallicity - luminosity relations , dwarf spheroidals are more metal - rich than dwarf irregulars at the same optical luminosity ( grebel et al .
2003 and references therein ) .
the flattening distribution of non nucleated de is similar to that of late - type spirals , i m and bcd ( binggeli & popescu 1995 ) .
this class of objects , however , is significantly less round than nucleated systems .
although they exist , the fraction of objects belonging to the intermediate dirr / de transition class is too small .
furthermore a simple transformation of i m galaxies recently infalling in the cluster into de does not seem to reproduce the observed difference in the cluster and field luminosity functions ( conselice 2002 ) .
other strong constraints against the transformation of star forming into quiescent dwarfs are given by the studies of globular clusters , taken as a probe of the early phases of galaxy formation .
the specific frequency of globular clusters in dwarf ellipticals , in fact , is significantly higher than that of star forming galaxies .
this result has been interpreted as an evidence for a different formation scenario for the two galaxy populations ( miller et al .
1998 ; strader et al .
+ with the aim of explaining these evidences , slightly different evolutionary scenarios have been proposed : cluster dwarf ellipticals might have been formed from higher surface brightness bcd galaxies ( bothun et al . 1986 ) , or from tidally induced mass loss in multiple high speed encounters of massive galaxies ( galaxy harassment ; mastropietro et al .
alternatively , lisker et al .
( 2006a ; 2006b ; 2007 ) proposed that the cluster dwarf elliptical galaxy population is composed of different subcategories of objects , of which not all have been formed from gas stripped star forming dwarfs . as here emphasized the issue is still hotly debated .
+ a few years ago we started collecting multifrequency data covering the whole electromagnetic spectrum for galaxies in nearby clusters in order to study the effects of the environment on galaxy evolution . up to now our research
was primarily focused on the bright end of the luminosity function .
our interest covered the present and past star formation activity ( gavazzi et al .
1991 , 1998 , 2002b , 2006a ) , the atomic ( gavazzi et al .
2005b , 2006b ) and molecular gas content ( boselli 1994 ; boselli et al .
1994 , 1997a , 2002 ) , the radio continuum and ir properties of cluster galaxies ( gavazzi & boselli 1999 ; gavazzi et al .
the results of our analysis , combined with those obtained by other teams , are summarized in a recent review article ( boselli & gavazzi 2006 ) .
+ the aim of the present paper is to extend the multifrequency analysis to the low - luminosity end of the luminosity function with the purpose of studying the possible transformation of star forming objects into dwarf ellipticals .
this exercise is here done using a complete sample of galaxies in the virgo cluster .
the novelty of this work compared to previous investigations is twofold : the analyzed sample has an unprecedented multifrequency spectral coverage from the uv to near - ir imaging data .
furthermore this unique dataset is compared to the predictions of multizone chemo - spectrophotometric models of galaxy evolution here adapted to take into account the perturbations induced by the cluster environment . the type of perturbation simulated by the models are ram - pressure stripping ( gunn & gott 1972 ) and starvation ( larson et al . 1980 ) .
the combination of multifrequency data with these models on the galaxies ngc 4569 ( boselli et al .
2006 ) and ngc 4438 ( boselli et al .
2005b ) indeed indicated how powerful this method is for studying and constraining the evolution of cluster galaxies .
+ our previous investigations have emphasized that the most important parameter governing the evolution is the total mass as traced by the near - ir h band luminosity ( gavazzi et al .
1996 ; 2002a ; boselli et al .
2001 ) . below a certain mass ( @xmath0 @xmath1 10@xmath2 l@xmath3 )
we identify the sequence of dwarf galaxies that we subdivide into quiescent and star forming disregarding their detailed morphology .
+ a major uncertainty in the interpretation of the multifrequency data , and in particular of those at short wavelength , is the extinction correction , which however is expected to be minor in low - metallicity star forming dwarf systems ( buat & xu 1996 ) and probably negligible in dwarf ellipticals .
for these reasons we decided to limit the present analysis to the cluster dwarf galaxy population , leaving the discussion relative to massive galaxies to a future communication . with the aim of driving the reader
s attention to the scientific results of this work , the presentation of the dataset and the general description of the models are given in appendix .
the analysis presented in this work is based on an optically selected sample of galaxies in the virgo cluster ( 12h @xmath1 r.a.(2000 ) @xmath113h ; 0@xmath13 @xmath1 dec @xmath1 18@xmath13 ) extracted from the virgo cluster catalogue ( vcc ) of binggeli et al .
( 1985 ) , with @xmath14 @xmath118 mag ( that for a distance of 17 mpc gives @xmath15 @xmath16 -13.15 ) which corresponds to its completeness limit .
+ although the present analysis is focused on the dwarf star forming ( scd - im - bcd ) and quiescent ( de - ds0 ) galaxy populations , that we define as @xmath0 @xmath1 10@xmath2 l@xmath3 ( which roughly corresponds to @xmath15 @xmath17 -18 mag ) , we also include brighter objects selecting all galaxies with @xmath14 @xmath1 18 mag .
this are all bona - fide virgo cluster members , whose distances have been assigned following the subcluster membership criteria of gavazzi et al@xmath18 ( 1999 ) which is based on combined position and redshift data ( now available for 83% of the objects ) .
the adopted distance for each subgroup is : 17 mpc for cluster a , the north and east clouds and the southern extension , 23 mpc for cluster b and 32 mpc for the w and m clouds .
the selected sample includes a total of 1010 galaxies , 445 of which are classified as de or ds0 , 215 as scd - sd , i m or bcd and 45 as de / im or ? in the vcc . because of their intrinsic color , the lowest h band luminosities ( thus the lowest total dynamical masses )
are reached only in star forming dwarfs . + model predictions are compared to uv to near - ir imaging and spectroscopic data collected thanks to our own observations ( optical , near - ir and h@xmath8 imaging , optical integrated spectroscopy ) or from different ground based ( sdss ) and space ( galex ) missions as well as from multifrequency data available in the literature .
the detailed description of the used dataset is given in the appendix .
the evolution of galaxies is traced using the multi - zone chemical and spectro - photometric models of boissier & prantzos ( 2000 ) , updated with an empirically - determined star formation law ( boissier et al@xmath18 2003 ) relating the star formation rate to the total - gas surface densities , and modified to simulate the effects induced by the interaction with the cluster igm . in the starvation scenario ( larson et al .
1980 , balogh et al .
2000 , treu et al . 2003 ) , the cluster acts on large scales by removing any extended gaseous halo surrounding the galaxy , preventing further infall of such gas onto the disk .
the galaxy then become anemic simply because it exhausts the gas reservoir through on - going star formation .
starvation has been simulated by stopping infall in the model .
+ the ism of galaxies crossing the cluster with velocities of @xmath9 1000 km s@xmath19 can be removed by the ram pressure exerted by the hot and dense igm ( gunn & gott 1972 ) .
gas removal induces a quenching of the star formation activity ( boselli & gavazzi 2006 ) .
the ram pressure stripping event is simulated by assuming a gas - loss rate inversely proportional to the potential of the galaxy , with an efficiency depending on the igm gas density radial profile of the virgo cluster given by vollmer et al .
the details of the models are given in the appendix .
here we present their results .
the total gas content , the activity of star formation and the metallicity of dwarf galaxies are barely affected by starvation unless the interaction started long time agoo , as shown for a model galaxy of rotational velocity @xmath20=55 km s@xmath19 and spin parameter @xmath21 = 0.05 ( fig . [ figure1 ] to [ figure5 ] ) .
hi - deficiencies of @xmath9 0.8 , similar to the average value of virgo galaxies within the virial radius of the cluster , can be obtained only if starvation started more than 6 gyr ago ( see fig .
[ figure1 ] ) .
+ gas removal induces a mild decrease of the star formation activity ( fig .
[ figure1 ] ) .
for this reason galaxies suffering starvation for a long time are redder than similar unperturbed objects , in particular in those color indices tracing significantly different stellar populations ( @xmath22 , @xmath23 , @xmath24 ; fig .
[ figure2 ] ) .
the @xmath25 color index is barely affected , because starvation acts on longer timescales than the lifetimes of the stellar populations emitting in the two uv filters ( @xmath4 10@xmath26 years ) , and does not totally eliminate these populations as significant amount of star formation , although reduced , still takes place .
+ gas metallicity increases more than in the unperturbed case since newly formed metals are diluted in smaller amounts of gas , and since the metallicity is not reduced by the infall of pristine gas ( fig . [ figure4 ] ) .
+ the effects of starvation on the structural properties of galaxies ( provided that the interaction started a long time ago ) are major .
the decrease of the surface brightness is expected to be @xmath9 1 mag in the h band to @xmath9 3 mag in the fuv band for an interaction started 6 gyr ago ( fig .
[ figure5 ] )
. the effect would be significantly smaller for a recent interaction .
+ the effects of starvation on the effective radius are minor since starvation is a global effect , and thus does not depend on radius .
nevertheless , an effect on the effective radius is visible , especially at short wavelengths .
the reason is the radial dependence of infall ( galaxies are formed inside - out in our model ) .
when the starvation starts , infall has occurred in the inner zones , where relatively large amount of gas can be found ( and thus high level of star formation and uv surface brightness ) .
this gas is progressively depleted by star formation , leading to a decrease of the inner surface brightness while the outer parts , where the star formation is low because of the low amount of gas present , dim at a lower rate .
the net effect is an increase in the effective radius .
the effects of a ram - pressure stripping event started 1 gyr ago ( @xmath27=1 gyr ) on the evolution of total gas content , star formation activity , stellar populations ( uv to near - ir color indices ) , metallicity and structural ( effective radius and surface brightness ) parameters of dwarf galaxies as predicted by our models are also shown in fig .
[ figure1 ] to [ figure5 ] .
because of the shallow potential well of dwarf galaxies , most of their gas is efficiently removed on very short time scales ( @xmath9 150 myr ) , leading to the formation of extremely gas deficient objects with hi - deficiencies @xmath9 2 for @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 , or @xmath9 1.2 for @xmath28 = 0.4 m@xmath29 kpc@xmath12 yr@xmath19 ( @xmath28 is the parameter that quantifies the efficiency of the process of ram pressure stripping ; see the appendix b ) .
+ as a consequence of this lack of gas , star formation abruptly decreases by a factor of @xmath9 10 to @xmath9 100 and continues at a very low level afterward . in the @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 model ,
the gas content progressively increases after the interaction as stars of small masses reach their end of life point and return more gas to the interstellar medium ( we do not use the instantaneous recycling approximation ) .
the star formation rate is fed by this recycled gas , and increases as well , staying nevertheless at low levels . in models with a lower ram pressure efficiency ,
the recycled gas represents a small amount with respect to the one left after the interaction . as a result ,
this effect is not visible : the gas and the star formation rate are about constant after the event .
+ stopping the star formation makes galaxies redder . since the decrease of the star formation activity is rapid , the younger stellar populations , those emitting in the fuv and nuv filters , are the most affected : the @xmath23 color index , for instance , increases by @xmath9 4.5 magnitudes in 1 gyr , while @xmath24 by only 1 magnitude .
+ as for the starvation scenario , the gas metallicity index significantly increases with time .
the increase of the gas metallicity is higher during a ram pressure stripping event since virtually all the gas has been removed , while some gas is left in the disk during a starvation event .
there is thus less dilution for a ram pressure event than in the occurrence of starvation . +
stopping the star formation drastically changes the structural properties of galaxies .
stripped galaxies have lower surface brightnesses and are slightly smaller with respect to unperturbed objects when observed at short wavelengths . at long wavelengths , where the emission is dominated by the evolved stellar populations , the effects are almost negligible ( fig .
[ figure5 ] ) . , the effective radius presents large variations in its value after ram pressure started .
this is due to the fact that the galaxy is left with very low amount of gas , and the effective radius becomes sensitive to two opposite effects . in a first phase ,
the dominant effect is that the inner disk is depleted through star formation at a larger rate than the outer disk , leading to an increase of @xmath30 ( like in the starvation case ) . in a second phase
, the inner disk is partially replenished with recycled gas ( ejected at the end of the life of long - lived stars ) , leading this time to a decrease of @xmath30 . ] + an accurate comparison between the effects induced by ram - pressure stripping and starvation on dwarf galaxy properties indicates that the former interaction is more efficient than the latter in removing gas and suppressing any star formation activity in galaxies on short time scales . to be efficient , starvation should have started several gyr ago .
recent ram pressure stripping events , however , did not have enough time to perturb the old stellar component emitting in the near - ir . + the effects of multiple encounters on the gas content and star formation activity of a dwarf , star forming galaxy of spin parameter @xmath21 = 0.05 and rotational velocity @xmath20 = 55 km s@xmath19 are shown in fig . [ multiple ] . the comparison of fig .
[ multiple ] with fig .
[ figure1 ] ( single stripping event ) shows how multiple interactions make galaxies even poorer of gas and , as a consequence , redder in colors , than similar objects that underwent a single ram pressure stripping event . for low efficiencies ,
successive crossings are almost additive ( some gas left on the first crossing is removed on the next one ) .
for large efficiencies , the second crossing , occurring 1.7 gyr after the first one , mostly remove any recycled gas , quenching a possible increase of star formation . in both cases , multiple crossing galaxies
end up as redder , poorer in gas objects .
a summary of the predicted perturbations induced by ram pressure stripping and starvation on the physical and structural properties of dwarf galaxies are given in table [ tabmod ] .
the multifrequency dataset available for dwarf galaxies in the virgo cluster ( see appendix ) allows us to compare model predictions with observations .
ram - pressure stripping models are able to reproduce the hi - deficient 15% of molecular gas ( boselli et al .
2002 ) and 30% of helium , thus @xmath31 @xmath9 @xmath32/1.4 . ]
galaxy population while starvation does not ( fig .
[ def ] ) . as previously remarked in sect .
3.2 , ram - pressure stripping can produce hi - deficient massive galaxies ( @xmath0 @xmath9 10@xmath33 l@xmath3 ) such as those observed in virgo . for a massive galaxy with @xmath20 = 220
km s@xmath19 and spin parameter @xmath21=0.05 , a ram pressure stripping event started 1 gyr ago with @xmath28 = 0.4 m@xmath29 kpc@xmath12 yr@xmath19 leads to an hi - deficiency of @xmath9 0.8 , similar to the average hi - deficiency of massive spiral galaxies in the center of the virgo cluster ( gavazzi et al 2005b ; boselli & gavazzi 2006 ) . more extreme hi - deficient objects of the cluster ( hi - deficiency @xmath9 1.4 )
can be produced by assuming @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 .
ngc 4569 can be taken as a typical object of this second category ( boselli et al .
+ we notice however that the hi emission of the ram - pressure stripped dwarf galaxy population can be hardly detected by the recent hi surveys such as alfalfa ( giovanelli et al .
2005 ) and ages ( auld et al .
2006 ) , able to reach only galaxies with hi gas masses @xmath34 10@xmath35 m@xmath29 ( alfalfa ) or slightly smaller ( ages ) at the distance of virgo ( 17 mpc ) ( dashed line in fig .
[ def ] ) .
galaxies less luminous than @xmath0 @xmath9 10@xmath36 l@xmath3 would be detected only in the case stripping is still under way . since the gas stripping event is short in time ( @xmath9 150 myr )
, we expect that these objects are extremely rare .
figure [ colmag ] shows several color - magnitude relations for galaxies in the virgo cluster .
we easily recognize the well known red and blue sequences relative to the quiescent and star forming galaxy populations ( scodeggio et al .
2002 ; gil de paz et al .
2007 ; see also visvanathan & sandage 1977 ; bower et al .
1992 for early - type galaxies and tully et al .
1982 , gavazzi et al . 1996 for late - type galaxies ) and a relatively large number of objects in between these two sequences ( see sec . 5 ) .
we remind that the reddest quiescent dwarfs might be undetected in the uv bands because of the sensitivity of galex .
model predictions clearly show that only a ram - pressure stripping event ( and not starvation ) is able to significantly modify the colors of galaxies .
as expected the effects are more important at short than at long wavelengths , affecting more colors as e.g. @xmath23 , @xmath22 than e.g. @xmath24 .
the colour index of bright spirals ( @xmath0 @xmath34 10@xmath37 l@xmath3 ) can increase by up to @xmath9 1 mag in @xmath25 and 2 mag in @xmath23 or @xmath22 @xmath9 500 myr after the interaction , making galaxies as red as the most deficient spirals located in between the red and blue sequences in the color - magnitude diagram .
ngc 4569 might be one of these objects .
+ the effect on dwarfs is more important , making star forming irregulars as red as dwarf ellipticals of similar luminosity in the case of an efficient gas stripping event or just after they crossed twice the cluster core ( @xmath9 2 gyr ) , as shown in fig .
[ colmag_2 ] ( see also fig . [ sed_ref ] ) . + figure [ metal ] shows the relationship between the gas 12 @xmath38 log(o / h ) index and the total h band luminosity of galaxies in the virgo cluster .
given the large uncertainty in both models ( it is well known that metal yields are uncertain by at least a factor 2 , see e.g. prantzos 2000 , and chemical evolution models are somewhat degenerate due to uncertainties on e.g. the initial mass function or star formation history ) and observables ( 0.3 dex ) , it is not surprising that models for the unperturbed galaxies are slightly shifted with respect to the data .
we should thus consider only relative values .
it is however clear that all starvation and ram pressure models predict both an increase of the gas metallicity and a decrease of the total h band luminosity consistent with the dispersion in the data .
this might explain the increase of gas metallicity observed in virgo ( vilchez 1995 ) and hydra ( duc et al .
2001 ) cluster galaxies .
the structural properties of quiescent and star forming dwarf galaxies are compared to model predictions in fig .
[ c31 ] and [ binggelilh ] .
the near - ir light concentration index @xmath39 ( defined as the ratio of the radii including 75 and 25% of the total light ) , is a quantitative indicator of the shape of the light profile in galaxies which reflects the @xmath7 exponent of the sersic law . in practice
it can be used as an indicator of the presence of bulges , thus to quantitatively compare the radial profiles of star forming and quiescent dwarfs , as shown in fig .
[ c31 ] . for luminosities
@xmath0 @xmath4 10@xmath40l@xmath3 the quiescent galaxy population is undersampled . in the 10@xmath41 @xmath1 @xmath0 @xmath4 10@xmath42 l@xmath3
both galaxy populations are dominated by objects with near - ir concentration indices @xmath43 @xmath4 3 ( 55% and 78% for early and late - type respectively ) , typical of exponential profiles .
the frequency of dwarfs with pronounced central cusps ( @xmath43 @xmath44 3 ) increases with luminosity , with an overall frequency comparable in the quiescent ( 64% ) and star forming ( 42% ) galaxy populations in the luminosity range 10@xmath42 @xmath1 @xmath0 @xmath4 10@xmath2 l@xmath3 . values of @xmath43 @xmath44 3 are more frequent in nucleated des ( large filled dots ) , and in high surface brightness star forming galaxies ( spirals , black open squares and green crosses , and bcds , blue open triangles ) .
we remind that ram pressure stripping events are not expected to deeply modify the relative weight of bulges in perturbed galaxies ( boselli et al .
+ for the star forming galaxies , the h band effective surface brightness and the luminosity are correlated , with a relatively large scatter , with higher disk surface brightnesses in more massive galaxies .
the unperturbed models predict a similar correlation , but with a small shift with respect to the observations ( which show higher surface brightnesses than model predictions ) .
these discrepancies could be explained by a combination of several effects .
first , as we noted above , the models are computed for an average spin , but surface brightness is sensitive to the spin parameter , thus we do expect a large scatter in such a plot .
second , small galaxies may have intrinsic irregularities ( linked to sporadic episodes of star formation ) affecting the effective surface brightness , which are not included in the models .
finally , the models only consider pure disks , which is the case only at low luminosities . if we make a further separation of the star forming dwarfs into different morphological classes ( see fig .
[ binggelilh ] ) we can see that the lowest surface brightnesses are reached only by magellanic irregulars ( sm - im ) . the h band effective surface brightnesses of the de are similar to those of the star forming bcd and spirals ( sa - sd ) of similar luminosity .
we notice however that the sample of de with available near - ir imaging is slightly biased towards high surface brightness , nucleated objects : among the 15 de galaxies with the lowest values of @xmath45 , only 20% have a measure of @xmath46 , while among b band high surface brightness objects 69% have measured @xmath46 .
+ the relationship between the b band effective surface brightness and the h band luminosity of de and scd - im - bcd is significantly dispersed , but consistent with the model predictions .
the effective surface brightnesses of quiescent and star forming systems are more similar in the b band than in the h band .
we notice however that dwarf ellipticals have effective surface brightnesses slightly higher than magellanic irregulars and smaller than late - type spirals and bcd of similar luminosity .
quiescent systems resulting from the quenching of the star formation activity of magellanic irregulars , whose existence is predicted by our models , are excluded from our sample just because they have b band surface brightnesses below the detection limit of the vcc . + the variations induced by the interactions are more evident in the uv bands . in the nuv band , for instance ,
star forming galaxies ( of all types ) have @xmath9 2 - 3 magnitudes brighter surface brightnesses than quiescent dwarfs , as predicted by models .
the most important variations of the surface brightness of perturbed galaxies predicted by models are in the fuv band ( up to 4 magnitudes ) .
the low galex sensitivity , however , prevents us to detect most of the de that , as expected , have @xmath47 @xmath44 26.5 mag arcsec@xmath12 .
+ + we can thus conclude that the observed spectrophotometrical and structural properties of dwarf ellipticals are consistent with those of star forming dwarfs once their activity has been stopped after a ram pressure gas stripping event able to remove their total gas content
. high surface brightness star forming objects should be the progenitors of the dwarf ellipticals , while magellanic irregulars the progenitors of quiescent objects of low surface brightness not detected in the vcc .
the analysis of fig . [ colmag ]
( see previous section ) reveals the existence of several objects having spectro - photometric properties in between that of star forming late - type galaxies and quiescent spheroidals .
the uv selection , which is strongly biased towards star forming objects , favors the detection of such blue , still active dwarf ellipticals .
figure [ colmag2 ] shows the same fuv - h color - magnitude relation given in fig .
[ colmag ] but using different colour symbols ( only for early - types ) to indicate still star forming galaxies ( [email protected].@xmath48@xmath44 2 , red ) , post - starburst galaxies which recently ended their activity as indicated by their strong balmer h@xmath49 absorption line ( [email protected].@xmath50@xmath44 2.8 , blue ) , hi detected objects ( green ) and rotationally supported spheroidals ( @xmath51/@[email protected] , magenta)/@xmath52 is defined as : @xmath51/@xmath52=@xmath53^{1/2}$ ] , where @xmath54 is the ellipticity of the galaxy ] .
we can photometrically define the transitional class as being formed by those objects of luminosity log @xmath0 @xmath1 9.6 l@xmath3 with a 3@xmath1 @xmath23 @xmath1 6 .
the early type galaxies in our sample satisfying this criteria are listed in table [ tabtrans ] .
figure [ colmag2 ] and table [ tabtrans ] clearly show that most of the low luminosity early - type galaxies with blue colors are characterized by an ongoing or recent star formation activity ( strong emission or absorption balmer lines ; see fig .
[ sed ] ) .
two of them have been detected in hi , while none have rotational velocity measurements .
+ figure [ sed_ref ] shows how the uv to near - ir spectral energy distribution of a star forming dwarf galaxy of @xmath21 = 0.05 and rotational velocity @xmath20 = 55 km s@xmath19 changes with time after a ram pressure stripping event characterized by an efficiency @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 . for comparison
the model sed are plotted along with the observed sed of the bcd galaxy vcc 24 and the dwarf ellipticals m32 and ngc 205 .
the sed of an unperturbed galaxy is similar to that of star forming dwarf systems such as the bcd vcc 24 , while it becomes redder soon after the interaction , reaching the values of the dwarf elliptical ngc 205 after 1.3 gyr .
two interactions are necessary to get colors as red as those of the compact dwarf elliptical m32 .
+ during the transition phase , the fuv - h colour gets redder while the h band magnitude is slightly reduced with respect to the reference case ( fig .
[ sed_fit ] ) .
we can use these properties to estimate the rotational velocity and lookback time to the interaction .
all the galaxies are close to the 40 - 55 - 70 km s@xmath19 range for which we computed models .
we thus interpolated ( and extrapolated above 70 km s@xmath19 ) in 1 km s@xmath19 bins our models .
[ sed_fit ] shows the position of each of these models ( for the 1.2 m@xmath29 kpc@xmath12 efficiency ) in the @xmath23 vs log @xmath0 diagram .
the models at the bottom of the diagram are unperturbed .
each point along the lines for different velocities correspond to various value of the stripping look - back time ( @xmath27=0 , 0.1 , 0.2 , 0.3 ... ) .
velocities and look back times for each transitional object can easily be found by a simple least - square test ( see table [ tabtrans ] and fig .
[ sed ] ) .
models indicate that all transitional objects recently ( @xmath1 650 myr ) interacted with the cluster igm . it has been shown that some dwarf elliptical galaxies have inverted color gradients , with bluer colors in the center ( lisker et al .
2006b and references therein ) .
indeed six out of the sixteen galaxies listed in table [ tabtrans ] have been defined as `` blue center '' by lisker et al .
( 2006b ) .
+ the observed inversion of the color gradient in the center of dwarf galaxies is obtained in our multizone models in a ram pressure stripping scenario ( see fig . [ profiles ] ) for recent interaction of relatively weak efficiency ( 0.1 m@xmath29 kpc@xmath12 ) , and has the same origin than that observed in massive spirals such as ngc 4569 ( boselli et al .
blue colors in the center of these dwarf galaxies are expected because ram pressure stripping is more efficient in the outer disk , where all the gas is removed , while some gas needed to feed star formation might be retained in the central part where the galaxy potential is deeper . for more efficient interactions ,
the contribution of the recycled gas becomes important whenever the interaction started long ago in the center of galaxies , where the stellar surface density is high and we obtain again blue centers .
we should note that if galaxies have deep potential wells because of a central mass condensation ( not taken into account in our models ) the gas in the center should be harder to remove , easily making blue cores .
given the shape of the potential well , whose deepness increases with mass ( catinella et al .
2006 ) , we expect that the inversion of the color gradient is more frequent and radially extended in the brightest dwarfs , which are more likely to retain some gas , as indeed observed ( lisker et al .
the lookback time to the ram pressure stripping event , combined with the age of the different stellar populations sampled in the various filters , make the observed colour inversion more pronounced at short wavelengths , as shown in figure [ profiles ] .
we can also add that such an inversion of the colour gradient can not be reproduced in a starvation scenario where gas removal is not a radial but rather a global effect .
+ for those galaxies with available sdss images we reconstructed their @xmath55 colour gradient ( see fig . [
sed ] ) using sdss images smoothed to the galex resolution .
twelve out of the 14 galaxies with available data have indeed blue colors in their center ( fig . [ sed ] and
table [ tabtrans ] ) , as expected if these galaxies recently underwent a ram pressure stripping event .
we remark that the observed blueing of the inner gradient happens in the inner 1 - 2 kpc , as predicted by models ( fig .
[ profiles ] ) . +
seven of the transitional class galaxies ( vcc 327 , 450 , 597 , 710 , 1175 , 1617 and 1855 ) have been imaged in h@xmath8 with the san pedro martir 2 m telescope ( gavazzi et al .
all galaxies ( except vcc 327 , undetected in h@xmath8 imaging ) show a similar pattern , with the star formation activity always limited to the central region and not extended to the whole disk , once again as predicted by our model , as shown in fig .
[ ha ] for vcc 710 .
the analysis done so far clearly indicates that some dwarf ellipticals might result from the rapid decrease of the star formation activity of low luminosity , late type galaxies after most of their gas was lost because of their strong interaction with the hostile cluster environment .
the comparison between model predictions and observations favors ram pressure stripping events with respect to galaxy starvation .
models in fact predict that low luminosity spirals , i m and bcd suffering starvation since 6 gyr are still star forming galaxies with blue uv colors , relatively strong balmer emission lines ( [email protected].@xmath48 @xmath9 10 ) and have an hi gas content ( @xmath9 10@xmath26 m@xmath29 ) easily detectable by hi surveys such as alfalfa and ages .
these objects would thus still be classified as star forming dwarf galaxies , and not dwarf ellipticals .
+ we should first notice that our simple representation of a ram pressure stripping event is totally consistent with the high resolution hydrodynamical simulations in cluster dwarf galaxies of mori & burkert ( 2000 ) murakami & babul ( 1999 ) and marcolini et al .
( 2003 ) , which predict the stripping of the whole gas reservoir on short timescales
. + as mentioned in the introduction , however , several observational evidences , namely 1 ) the different structural properties of magellanic irregulars and dwarf ellipticals , 2 ) the presence of nuclei in de , 3 ) the higher specific frequency of globular clusters and 4 ) the higher stellar metallicity of quiescent systems with respect to star forming dwarfs , 5 ) their different velocity distribution and 6 ) the different shape of the field and cluster luminosity functions have been often indicated as major limitations to the formation of cluster dwarf ellipticals through the gas stripping of star forming systems .
we see in this section how these difficulties can be overcome .
in general we can say that the present work gives a natural explication to the presence of rotationally supported ( pedraz et al .
2002 ; geha et al . 2003 ; van zee et al .
2004b ) , gas rich ( conselice et al . 2003b
; van zee et al .
2004b ) star forming dwarf ellipticals .
the presence in some dwarf ellipticals of spiral arms observed by lisker et al .
( 2006a ) is a further evidence of the rotating / spiral origin of these objects .
the grand design spiral arms of these de , apparently different from the flocculent open arms of very late - type spirals , might be driven by disk instabilities produced by the displacement of the gas over the stellar disk during the stripping event ( elmegreen et al .
2002 ) . consistently with observations , which have shown that the frequency of de with grand design spiral arms increases with luminosity ( lisker et al .
2006a ) , the perturbations induced by the gas displacement are expected to be important only in those objects where the potential is sufficient to retain some gas , while totally absent in the lowest luminosity systems where the whole gas content is instantaneously removed .
+ the comparison of model predictions with observations suggests that on average the variation of the structural parameters ( surface brightnesses , effective radii and concentration indices ) of high surface brightness low - luminosity star forming galaxies such as bcd or late - type spirals ( mostly scd - sd ) at various wavelengths expected after a ram pressure stripping event might indeed reach the values observed in de .
consistently with bothun et al .
( 1986 ) we confirm that the sampled de have near - ir and optical surface brightness higher than those of magellanic irregulars ( sm - im ) and thus can not result from this galaxy population .
our models , however , predict the existence in the virgo cluster of an extremely low surface brightness quiescent galaxy population ( @xmath45 @xmath34 25 mag arcsec@xmath12 ) resulting from the quenching of the star formation activity of magellanic irregulars .
such a population has been indeed observed by sabatini et al .
+ our model predictions can not be easily compared to the results of lisker et al .
( 2006a , b ; 2007 ) and binggeli & popescu ( 1995 ) on the flattening distribution of star forming and quiescent dwarfs since models do not take into account the 3-d evolution of the disks .
we can argue that the lack of supply of young stars with low velocity dispersion because of the suppression of the star formation activity causes the disk to heat up , dumping spiral waves on timescales of a few revolutions ( sellwood & calberg 1984 ; fuchs & von linden 1998 ; elmegreen et al .
2002 ) in particular in low luminosity systems where the rotational component is not always dominant . in low luminosity rotating systems
( 70 - 130 km s@xmath19 ) it has been observed that the disk scale height increases by @xmath9 a factor of 2 in @xmath9 3 gyr ( seth et al .
we could thus expect that the flattening distribution is age dependent . in the lack of external gravitational perturbations ,
the rotational velocity is conserved while @xmath52 increases once the gas is removed .
gravitational perturbations induced by the interaction with other galaxies or with the cluster potential would heat up the system and disperse the angular momentum , thus speeding up the decrease of @xmath56 ( wozniak 2007 , private communication ) .
we would thus expect rounder shapes and more relaxed distributions in those objects where the star formation stopped long time ago .
we remark that , although still not virialized , nucleated objects are more centrally clustered than non nucleated systems ( lisker et al .
2007 ) .
the angular resolution of our model is too poor ( @xmath9 1 kpc ) to make any prediction on the formation process and evolution of the nuclei observed in dwarf ellipticals . as a general remark we can say that recent acs / hst high resolution observations of nuclei of early - type galaxies in the virgo cluster revealed that they have statistical ( frequencies ) and physical ( sizes , luminosities , colors ) properties similar to those of the nuclear stellar clusters found in late - type galaxies ( cte et al . 2006 ) , consistently with our evolutionary picture . hst observations of de show color gradients getting bluer towards their nuclei witnessing a younger stellar population than that of the outer disk ( lotz et al .
2004 ; cte et al .
furthermore the presence of a nucleus is more frequent in massive de than in low luminosity objects ( sandage et al .
1985 ; ferguson & binggeli 1994 ) .
+ the rapid evolution expected after the galaxy - icm interaction which abruptly stops the star formation is able to change the total luminosity of the perturbed galaxy .
the typical progenitor of a given dwarf elliptical that did not undergo such a violent truncation of its star forming activity today would be much brighter than its potential progeny even in the near - ir ( see for instance fig .
[ colmag ] ) . if we make the reasonable assumption that globular clusters were formed in the early phase of galaxy formation ( thus well before any possible recent ram pressure stripping event ) and regulated by the primordial dark matter condensation , the specific frequency of globular cluster should thus be determined by normalizing the total number of globular clusters to the absolute magnitude of the unperturbed galaxy which is always brighter than its gas stripped counterpart .
being the specific frequency of globular clusters defined as @xmath57 @xmath11 @xmath58 @xmath59 10@xmath60 ( where @xmath58 is the number of globular cluster ; strader et al .
2006 ) , we can see that @xmath57 increases with time after a ram pressure stripping event just because the absolute magnitude @xmath61 of the perturbed system does not increase with time as that of its unperturbed counterpart ( see fig .
the specific frequency of globular clusters in de galaxies ranges in between @xmath9 1 and @xmath9 20 with @xmath9 a double distribution peaked at @xmath57 @xmath9 2 and @xmath57 @xmath9 10 ( strader et al .
these values are consistent with a scenario where de have been formed through a ram pressure gas stripping event of low luminosity late type galaxies ( whose specific frequency is @xmath57 @xmath4 1 ; miller et al .
1998 ) occurred @xmath4 4 gyr ( @xmath4 1 gyr for de with @xmath57 @xmath9 2 ) .
+ grebel et al .
( 2003 ) observed an increase of the metallicity measured in the old stellar populations in dwarf spheroidals with respect to dwarf irregulars of similar luminosity .
it is equivalent to say that the spheroidals are fainter than irregulars for the same stellar metallicity .
this is naturally obtained in our models : let s consider a `` common ancestor '' , giving rise to a low luminosity , star forming galaxy if it evolves in isolation , and to a de if its star formation activity is suppressed after an interaction .
the metallicity in the old population will be the same ( the one of the common ancestor ) .
the luminosity of the star forming object will be the one of the unperturbed model , while the one of the quiescent system will be the one of the model after an interaction .
[ gc ] shows that the difference in v band luminosity ( partly due to the fading of the stellar population , but also to the fact that less stars formed in the latter than in the former case ) is of @xmath9 1 magnitude after 1 gyr .
this luminosity difference for the same metallicity corresponds well to that observed by grebel et al .
( 2003 ) between the star forming and quiescent dwarf galaxy populations .
+ the velocity distribution of the low luminosity ( @xmath0 @xmath1 10@xmath37 l@xmath62 ) , high surface brightness ( @xmath46@xmath1 22 mag arcsec@xmath12 ) star forming galaxies ( see table [ tabvel ] ) , i.e. of those objects that our models indicate as the probable progenitors of the observed dwarf ellipticals , is strongly non gaussian , with a pronounced wing at high velocity ( see fig .
[ vel ] ) , witnessing infall .
if limited to the virgo a subcluster , thus to the region associated to the hot gas emitting in x - ray where ram pressure is active ( gavazzi et al .
1999 ) , the probability that both nucleated and non nucleated dwarf ellipticals are driven by the same parent population as low luminosity , high surface brightness star forming galaxies is relatively important ( @xmath63 @xmath9 40% ) , as indicated by kolmogorov - smirnov tests ( see table [ tabks ] ) .
although dynamical considerations are consistent with the infall of the whole dwarf elliptical galaxy population , it is however difficult to determine when this happened : the velocity dispersion inside the cluster of their progenitor star forming , low - luminosity galaxy population should not change very much over a hubble time as a result of dynamical friction , relaxation or energy equipartition as shown by conselice et al .
+ the scenario where all virgo cluster early type dwarfs have been created after the gas removal of low - luminosity late type galaxies recently infalled into the cluster is however too extreme since quiescent dwarf galaxies are present also in the field ( pasquali et al . 2005 ) .
the contribution to the faint end of the field luminosity function of de , however , is less important than that of blue , star forming objects ( blanton et al . 2005 ) , it is thus possible that the fraction of dwarf galaxies recently perturbed by the cluster environment is very high .
we also remark that the slope of the virgo cluster luminosity function ( @xmath8 @xmath9 -1.4 ; sandage et al .
1985 ) is comparable to the most recent determinations of the field luminosity function based on sdss data , @xmath8 @xmath9 -1.4 , -1.5 ( blanton et al . 2005 ) .
steeper slopes in other clusters luminosity functions have been however determined by popesso et al .
( 2006 ) . + the question now is to see how many virgo cluster dwarf ellipticals might have been formed after ram pressure stripping of low - luminosity star forming systems . to answer this fundamental question
we should know both the infalling rate of low luminosity , star forming galaxies in cluster and the time scale for a galaxy to totally stop its activity .
our models give us an idea of the different time scales for gas removal and suppression of the star formation and of the subsequent evolution of the different stellar populations inhabiting these galaxies .
figure [ clock ] shows how , on relatively short time scales , the hi deficiency parameter , on two different indicators tracing the current ( h@xmath8 emission line ; kennicutt 1998 ; boselli et al .
2001 ) and past ( h@xmath49 absorption line emission , which might be present whenever star formation is still active ] ; poggianti & barbaro 1997 ; poggianti et al .
2001a , 2001b ; thomas et al .
2004 ) star formation activity and the @xmath23 color index are expected to evolve after a ram pressure stripping event for galaxies with @xmath0 @xmath1 10@xmath2 l@xmath62 .
gas removal is extremely efficient in such low - mass objects , making the hi mass decrease by @xmath9 two orders of magnitude on very short time scales , i.e. @xmath9 150 myr ( see table [ tabstat ] ) . the lack of gas causes , on similar time scales , a rapid decrease of the [email protected].@xmath48 ( we remind that the h@xmath8 emission of a galaxy is due to the gas ionized by o - b stars of ages @xmath1 10@xmath64 years , kennicutt 1998 ) .
the interaction is so efficient that both indicators are already strongly perturbed even before the galaxy reaches the center of the cluster ( lookback time to the interaction=0 in fig .
[ clock ] ) .
the [email protected].@xmath50 and the @xmath23 color index , on the other hand , evolve more gradually in time : we can say that on average the equivalent width of the h@xmath49 absorption line drops down to @xmath1 2.8 after @xmath9 0.5 - 0.8 gyr , while @xmath23 becomes redder by 5 mag after @xmath9 0.8 gyr independently from the adopted ram pressure stripping model .
the transition in between the blue sequence of low - luminosity star forming galaxies and the red sequence of quiescent dwarf ellipticals is thus a very rapid event , it is thus not surprising that only a few objects populate the intermediate region .
+ table [ tabstat ] indicates that the fraction of low - luminosity early type galaxies in the absolute magnitude range -13.15@xmath44 @xmath15 @xmath44 -17.5 ( this upper limit roughly corresponds to @xmath0 @xmath1 10@xmath2 l@xmath3 ) with a residual star formation activity ( [email protected].@xmath48 @xmath44 2 ) , still rich in hi gas , with a post starburst activity ( [email protected].@xmath50@xmath44 2.8 ) or with blue colors ( @xmath23 @xmath1 5 ) is important .
these values indicate that @xmath9 10 - 16 % of the quiescent dwarfs might result from star forming systems which underwent an interaction with the cluster medium less than 150 myr ago , while up to @xmath9 40 % in the last @xmath9 500 - 800 myr . and
h@xmath49 data are limited to the central 3 - 4 arcsec of the observed galaxies , and are thus not directly comparable to the model predictions which are relative to the whole galaxy .
integrated spectroscopy however is available only for a small fraction of the quiescent dwarfs . ]
these values can be used to infer a dwarf galaxy infall rate of @xmath9 300 objects per gyr ( see tab . [ tabstat ] ) , a relatively important rate if we consider the total number of dwarf members of virgo ( @xmath9 650 , see sect .
this rate , however , is consistent with that determined by adami et al .
( 2005 ) for coma scaled to the mass of virgo : 1 - 4 @xmath59 10@xmath65 l@xmath29 gyr@xmath19 if a mass to light ratio of 10 is adopted .
+ these values consistently indicate that the whole virgo cluster dwarf quiescent galaxy population might have been formed by the suppression of star formation in low luminosity , gas rich systems recently entered into the cluster if an infall rate similar to the present one lasted @xmath9 2 gyr ( thus equivalent to @xmath66=0.16 in a @xmath67 = 70 km s@xmath19 mpc@xmath19 , @xmath68=0.3 and @xmath69=0.7 cosmology ) .
this result is consistent with the most recent analysis of the color magnitude relation of clusters at different redshift which all indicate a significant decrease of the fraction of low luminosity objects on the red sequence with increasing @xmath66 ( de lucia et al . 2004 ; 2007 ) from @xmath66=0.24 ( smail et al .
it has also been shown that the fraction of galaxies in between the red and the blue sequence of the color magnitude relation ( which might be considered as the analogues of our transient population ) in high density environment at @xmath66=0.7 from cosmos increases with the decrease of the galaxy luminosity ( cassata et al .
2007 ) , as predicted by our model .
our results are also consistent with the analysis of nelan et al .
( 2005 ) on 93 nearby clusters based on the age-@xmath52 relation which indicates that the quiescent dwarf galaxy population only recently joined the red sequence .
we can conclude that models and observations are consistent with an evolution of star forming , low - luminosity late type galaxies recently accreted in virgo into quiescent dwarfs because of the ram pressure gas stripping and the subsequent stopping of their star formation activity .
for consistency with surface brightness measurements , we show that high surface brightness star forming dwarf galaxies ( low luminosity spirals and bcd ) might be at the origin of the optically selected dwarf ellipticals ( both normal and nucleated ) analyzed in this work , although we expect that the low surface brightness magellanic irregulars ( sm and i m ) once stripped of their gas reservoir , produce quiescent dwarfs with surface brightnesses below the detection limit of the vcc , as those observed in virgo by sabatini et al .
the process of transformation is extremely rapid and efficient , since it works on all dwarfs and last on average less than 150 myr . on
longer time scales galaxies get structural and spectrophotometric properties similar to that of dwarf ellipticals . the whole star forming dwarf galaxy population dominating the faint end of the field luminosity function ( blanton et al .
2005 ) , if accreted , can be totally transformed by the cluster environment into dwarf ellipticals on time scales as short as 2 gyr and thus be at the origin of the morphology segregation observed also at low luminosities .
+ this interesting result is of fundamental importance even in a cosmological context because it shows that the majority of dwarf galaxies ( if not all of them ) are `` young '' even in clusters , and not old as expected in a hierarchical galaxy formation scenario ( see also nelan et al .
for comparison with de lucia et al ( 2006 ) , we give in table [ tabage ] the lookback time when 50% and 80% of the stars were formed for a model galaxy of @xmath20 @xmath11 55 km s@xmath19 for a ( or two ) ram pressure stripping event of efficiency @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 . although the values given in table [ tabage ] are not directly comparable to those given by de lucia et al .
( 2006 ) since limited to relatively high mass objects ( 2.5 10@xmath70 m@xmath29 , while our model galaxy is of only 2.4 10@xmath26 m@xmath29 ) , the mean age of the stellar population of low luminosity , quiescent galaxies is significantly younger than that predicted by hierarchical models of galaxy formation ( lookback times @xmath9 10 and 8.5 gyr for 50% and 80% of the stellar population ) .
+ despite their morphological type , the star formation activity of dwarf systems has been abruptly interrupted by the interaction with the environment in quiescent systems .
this conclusion can be extended to the local group , where the study of the stellar color magnitude relation of dwarf spheroidal systems revealed that the star formation activity , although in an episodic manner , lasted for several gyr ( mateo 1998 ; grebel 1999 ) .
we can add that the only cluster galaxy population likely to be issued by major merging events , as those predicted by hierarchical models of galaxy formation , is that of massive ellipticals , whose origin is probably very remote ( @xmath66 @xmath34 2 - 3 , dressler 2004 ; treu 2004 ; nolan 2004 ; franx 2004 ; renzini 2006 ) .
indeed this is the only virgo cluster galaxy population with a virialized velocity distribution ( conselice et al .
if we consider clusters of galaxies as those regions where merging events were more frequent at early epochs since the big bang just because characterized by an high galaxy density , we can conclude that the hierarchical formation scenario was the principal driver of galaxy evolution only in massive objects at very early epochs .
all observational evidences are consistent with a secular evolution afterward .
we wish to thank c. adami , e. athanassoula , c. balkowski , v. buat , g. comte , p.a .
duc , g. hensler , j. lequeux , h. wozniak for precious comments and suggestions .
we are grateful to the anonymous referee for his constructive suggestions extremely useful in the preparation of the final version of the manuscript .
galex ( galaxy evolution explorer ) is a nasa small explorer , launched in april 2003 .
we gratefully acknowledge nasa s support for construction , operation , and science analysis for the galex mission , developed in cooperation with the centre national detudes spatiales of france and the korean ministry of science and technology .
we wish to thank the galex soda team for their help in the data reduction .
this research has made use of the gold mine database .
@xmath71 notes : * = six gyr is the lookback - time for starvation in our models . for a lookback - time @xmath1 6 gyr
the effects induced by starvation are less important , while are more important for a lookback - time @xmath44 6 gyr .
@xmath72 column 1 : vcc name , from the virgo cluster catalogue of binggeli et al .
( 1985 ) . + column 2 : morphological classification , from binggeli et al . ( 1985 ) . + column 3 : presence of a blue center in @xmath73 from lisker et al .
( 2006b ) ( bc ) and/or @xmath55 from this work ( bc ) .
1 stands for galaxies not observed by the sdss .
+ column 4 : h band luminosity , in solar units .
+ column 5 : @xmath23 color index , in ab magnitudes . + column 6 : the equivalent width of the h@xmath8 emission line , in , from ( in order of preference ) imaging ( gavazzi et al .
2006a ) and integrated spectroscopy(gavazzi et al .
nuclear ( sdss , dr5 ) spectroscopic data are given in parenthesis .
+ column 7 : the equivalent width of h@xmath49 underlying absorption line , in , from ( in order of preference ) integrated ( gavazzi et al .
2004 ) or nuclear ( sdss , dr5 ) spectroscopy . + column 8 : the logarithm of the hi mass , in solar units , from gavazzi et al .
( 2005b ) .
+ column 9 and 10 : lookback time to the interaction and rotational velocity of the best fitting model for a ram pressure stripping event with @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 .
lookback times of 0 myr are for galaxies now at the peak ( maximum ) of their interaction .
+ @xmath74 notes : column 1 : sfd stands for star forming dwarfs , low - luminosity ( @xmath0@xmath110@xmath37 l@xmath3 ) , high surface brightness ( @xmath46@xmath1 22 mag arcsec@xmath12 ) late - type galaxies .
+ columns 6 and 11 : @xmath63 is the probability that galaxies follow a normal distribution .
+ @xmath75 sfd : star forming dwarfs are all low - luminosity ( @xmath0@xmath110@xmath37 l@xmath3 ) , high surface brightness ( @xmath46@xmath1 22 mag arcsec@xmath12 ) late - type galaxies .
+ @xmath76 notes : column 4 gives the fraction ( total number / observed ) of the observed early - type virgo galaxies with -13.15@xmath44 @xmath15 @xmath44 -17.5 satisfying the condition , where the total number of virgo galaxies satisfying these criteria are 498 . + * * : 127 is the total number of galaxies with an h@xmath49 absorption line @xmath34 2.8 ( 89 ) or h@xmath49 in emission ( 38 ) ; if we consider also those with h@xmath8 in emission ( 50 ) , for which the h@xmath49 absorption line is generally underestimated , the total number of objects increases to 148 ( 46% ) . + * * * : given the detection limit of galex , the value relative to the color index @xmath23 is a lower limit .
+ @xmath77 note : models are for a galaxy @xmath20 @xmath11 55 km s@xmath19 for a ( or two ) ram pressure stripping event of efficiency @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 + column 2 : lookback time to the stripping event
the dataset used in the present analysis is composed of imaging and spectroscopic data covering the whole uv to ir spectral range .
the virgo cluster region was observed in spring 2004 as part of the all imaging survey ( ais ) and of the deeper nearby galaxy survey ( ngs ) carried out by the galaxy evolution explorer ( galex ) in the two uv bands fuv ( @xmath78 ) and nuv ( @xmath79 ) .
this public dataset ( ir 1.0 release ) , now available on mast , has been combined with data extracted from several open time observations obtained at a similar sensitivity than the ngs .
details of the galex instrument and data characteristics can be found in martin et al@xmath18 ( 2005 ) and morrissey et al@xmath18 ( 2005 ) .
+ whenever available , we extracted fluxes from the ngs and open time images , obtained with an average integration time of @xmath9 1500 sec .
these data are complete to @xmath80 @xmath9 21.5 in the nuv and fuv . elsewhere
uv fluxes have been extracted from the shallower ais images ( @xmath9 70 sq .
degrees ) , obtained with an average integration time of @xmath9 100 sec , complete to @xmath80 @xmath9 20 in both the fuv and nuv bands .
+ b and v frames are available for most of the analyzed galaxies thanks to our own observations ( gavazzi & boselli 1996 ; gavazzi et al . 2001 ; 2005a and boselli et al .
h ( 1.65 @xmath81 m ) and k ( 2.1 @xmath81 m ) band frames have been obtained during a near - ir imaging survey of the virgo cluster ( boselli et al . 1997b ; 2000 ; gavazzi et al . 2000b , 2001 ) .
sdss imaging data , taken from the most recent dr5 ( release 5 , adelman - mccarthy et al .
2007 ) , have been also used for the determination of the radial nuv-@xmath82 color gradients .
+ fluxes were obtained by integrating all images within elliptical annuli of increasing diameter up to the optical b band 25 mag arcsec@xmath12 isophotal radii .
independent measurements of the same galaxies obtained in different exposures give consistent photometric results within 10% in the nuv and 15% in the fuv in the ais , and about a factor of two better for bright ( nuv @xmath416 ) galaxies .
we thus estimate that the overall uncertainty in the uv photometry is on average a factor of @xmath9 2 better in the ngs or open time images than in the ais especially for the faintest objects .
optical and near - ir data have on average a photometric accuracy of @xmath9 10% .
the h and b band structural parameters effective radii @xmath83 ( radius containing half of the total light ) and surface brightnesses @xmath84 ( mean surface brightness within @xmath83 ) have been measured as described in gavazzi et al .
( 2000a ) . given the poor resolution of the galex data ( fwhm of the psf @xmath9 5 arcsec for a pixel size of 1.5 arcsec ) , these structural parameters are available only for the brightest ( and bluest in fuv ) dwarf galaxies in virgo .
+ spectral line indices ( [email protected].@xmath48 , [email protected].@xmath50 ) are available for a large fraction of the analyzed galaxies .
most of the selected objects have been observed as part of a spectroscopic survey of the virgo cluster ( gavazzi et al .
2004 ) still under way .
these data have been extracted from long slit integrated spectroscopy obtained by drifting the slit over the whole galaxy disc , as in kennicutt ( 1992 ) , thus providing values representative of the whole object .
these are intermediate ( @xmath85 1000 ) resolution spectra in the range ( @xmath86 ) .
low resolution spectroscopy of the central region of several galaxies is also available from the sdss ( release 5 ) . to minimize aperture effects ,
sdss spectroscopy has been taken only for axisymmetric quiescent galaxies ( e , de , ds0 ) .
a few other spectroscopic data for dwarf ellipticals have been taken from geha et al .
( 2003 ) and michielsen et al .
line absorption indices ( [email protected].@xmath50 ) have been measured in the lick system , although their exact calibration is still under way .
+ hi data have been taken from the compilation of gavazzi et al .
( 2005b ) , conselice et al .
( 2003b ) and ( giovanelli et al .
the hi - deficiency parameter ( @xmath87 ) is defined as the logarithmic difference between the average hi mass of a reference sample of isolated galaxies of similar type and linear dimension and the hi mass actually observed in individual objects : @xmath87 = log@xmath88 - log@xmath89 . according to haynes & giovanelli ( 1984 ) , log@xmath88 = @xmath90 + @xmath91log@xmath92 , where @xmath90 and @xmath91 are weak functions of the hubble type and @xmath93 is the linear diameter of the galaxy ( see gavazzi et al .
since we make the hypothesis that dwarf ellipticals are gas - stripped low - luminosity , star forming objects , their hi deficiency parameter has been determined assuming as reference isolated star forming dwarf galaxies ( @xmath90 = 7.00 and @xmath91 = 1.88 , from gavazzi et al .
galaxies with an hi - deficiency parameter larger than 0.3 are considered as deficient in hi gas .
+ rotational velocities for dwarf elliptical galaxies have been taken from geha et al .
( 2003 ) , van zee et al .
( 2004b ) , pedraz et al .
( 2002 ) , thomas et al .
( 2006 ) and simiens & prugniel ( 1997a , 1997b , 1998 , 2002 ) .
+ uv to near - ir imaging data have been corrected for galactic and internal extinction as described in boselli et al .
internal extinction corrections have been applied only to late - type galaxies .
owing to the high galactic latitude of virgo , galactic extinction corrections are very small ( @xmath94 @xmath4 0.05 ) .
+ most of the imaging and spectroscopic data used in the present analysis are available in electronic format on the goldmine database ( http://goldmine.mib.infn.it ; gavazzi et al .
to trace the evolution of late - type galaxies of different luminosity , we have used the multi - zone chemo - spectrophotometric models of boissier & prantzos ( 2000 ) , updated with an empirically - determined star formation law ( boissier et al@xmath18 2003 ) relating the star formation rate to the total - gas surface densities ( @xmath95 , @xmath96 ) : @xmath97 where @xmath98 is the rotation velocity at radius @xmath99 .
this is a variant of the traditional `` schmidt law '' with two parameters : the index @xmath7 and an efficiency of the star formation activity @xmath8 .
the values of @xmath8 and @xmath7 are taken as equal respectively to 2.63 10@xmath100 and 1.48 , as determined from the gaseous and h@xmath8 profiles of 16 nearby galaxies ( boissier et al@xmath18 2003 ) , where @xmath95 is expressed in m@xmath29pc@xmath12gyr@xmath19 , @xmath96 in m@xmath29pc@xmath12 , @xmath98 in km s@xmath19 and @xmath99 in kpc .
the resulting models are extremely similar to those presented in boissier & prantzos ( 2000 ) and show the same global trends .
we consider the star formation law as fixed and we keep the same mass accretion ( infall ) histories as in boissier & prantzos ( 2000 ) , based on the assumption that before the interaction with the cluster , galaxies were `` normal '' spirals .
the free parameters in this grid of models are the spin parameter , @xmath21 and the rotational velocity , @xmath20 .
these two parameters are theoretical quantities , although @xmath20 should be similar to the asymptotic value of the rotation curve at large radii .
the spin parameter is a dimensionless measure of the specific angular momentum ( defined in e.g. mo , mao & white 1998 ) .
its value in spirals ranges typically between @xmath9 0.02 for relatively compact galaxies to @xmath9 0.09 for low surface brightness galaxies ( boissier & prantzos 2000 ) .
the models of boissier & prantzos ( 2000 ) contains scaling relationships ( the total mass varies as @xmath101 , the scale - length as @xmath102 ) .
star formation histories depend on the infall timescales , which are a function of @xmath20 in these models , so that roughly speaking , @xmath20 controls the stellar mass accumulated during the history of the galaxy , and @xmath21 its radial distribution . + in order to avoid the multiplication of parameters , we decided to fix the spin parameter to an average value of 0.05 and investigate global trends linked to the mass -or the velocity of galaxies- or to the interaction itself .
other values of the spin parameter should mostly create some scatter around these trends . from e.g.
boissier & prantzos ( 2000 ) , it can be seen that trends are mostly driven by mass ( see their figure 8) .
various spin parameters create a modest dispersion with respect to the average one ( e.g. about a few tenths of magnitude for the b - k colour index , a few tenths of dex for stellar and gaseous masses ... ) .
the only quantities which are clearly much more affected by the spin parameter than the velocity are the surface brightnesses .
the central surface brightness in the b band ( or @xmath103 band , see boissier 2000 ) can change by a few mag arcsec@xmath12 for a given circular velocity when using various realistic spin parameters . to sample the whole dynamic range covered by dwarf galaxies we focused our analysis to the velocity range 40 - 100 km s@xmath19 , although we tested several other models for larger velocities ( for comparison with bright spirals ) .
we remind that the model , which has a resolution of @xmath9 1 kpc , does not include any bulge or nuclear component , corresponding well to pure exponential disks such as these dwarf star forming galaxies ( gavazzi et al .
2000a ) . on the other hand ,
the models were developed and extensively tested for relative massive galaxies ( rotation velocities above 80 km s@xmath19 ) . in this work ,
we are interested mostly in the 40 - 100 km s@xmath19 range , and low velocity models were simply extrapolated from the massive ones . by this , we mean that we assumed that the star formation law , initial mass function and other ingredients used for massive galaxies also apply to the low mass case .
especially , the models include a dependence of infall time scales on the circular velocity ( more massive galaxies form the bulk of their stars earlier than lower mass galaxies ) .
this assumption was found to be in agreement with observations in normal spirals , above @xmath9 80 km / s ( boissier & prantzos , 2000 ) , and is here generalized to lower masses .
these considerations call for some caution in the interpretation of our results . especially , low mass galaxies are very small and the models do not take into account the likely radial mixing on scales smaller than @xmath9 1 kpc , making them extremely uncertain with respect to e.g. metallicity gradients .
nevertheless , we show in this paper that our extrapolated models reproduce reasonably well the properties of normal late type low velocity galaxies ( magnitudes , colors ... ) .
in addition , errors due to model uncertainties are likely to be systematic effects : they should affect less differences between two models than absolute values .
thus , we believe that the differences obtained when ram - pressure is introduced with respect to the unperturbed case can be trusted .
the models of boissier & prantzos ( 2000 ) provide total luminosities and colors in all bands as well as the total gas content .
they do not compute the nebular emission , but we estimated the h@xmath8 emission by using the number of ionizing photons predicted by version 5 of starburst 99 ( vazquez & leitherer 2005 ) for a single generation of stars distributed on the kroupa et al .
( 1993 ) initial mass function ( as used in our models ) , convolving it with our star formation history , and converting the result into our h@xmath8 flux following appendix a of gavazzi et al .
( 2002a ) .
+ the models also produce low resolution theoretical spectra from which are used to compute the lick h@xmath49 index .
the model spectral resolution is actually quite poor ( 20 in the visible ) but we checked that the evolution of this index in the case of a burst is consistent with the one presented in e.g. tantalo & chiosi ( 2004 ) .
+ the evolution of the metallicity is computed in the models taking into account the yields , finite lifetimes of stars , infall of pristine gas ( boissier & prantzos 1999 ) .
the spectra are computed taking into account this metallicity and the star formation history .
given the agreement between model predictions and observations for different samples of unperturbed galaxies ( boissier & prantzos 1999 ; 2000 ; prantzos & boissier 2000 ; boissier et al .
2001 ) , we will consider them as the reference models for the unperturbed case .
as for ngc 4569 ( boselli et al .
2006 ) we study two different interaction scenarios , starvation and ram - pressure stripping . in the starvation scenario ( larson et al . 1980 , balogh et al .
2000 , treu et al . 2003 ) , the cluster acts on large scales by removing any extended gaseous halo surrounding the galaxy , preventing further infall of such gas onto the disk .
the galaxy then become anemic simply because it exhausts the gas reservoir through on - going star formation .
+ infall is a necessary assumption in models of the chemical evolution of the milky way to account for the g - dwarf metallicity distribution ( tinsley 1980 ) and is supported by some chemo - dynamical models ( samland et al .
as the disk galaxy models were obtained through a generalization of the milky way model , infall is present in all our models .
it is a schematic way to describe the growth of any galaxy from a proto - galactic clump in the distant past to a present - day galaxy .
infall time scales in the models were chosen to reproduce the properties of present day normal galaxies ( boissier & prantzos 2000 , boissier et al .
this includes a dependency on the rotational velocity such that infall in low mass galaxies is shifted towards later time : this mimics the fact that massive galaxies are formed early on and that low mass galaxies are still forming at the current time , a well known fact ( gavazzi et al .
1996 , 2002 ; boselli et al .
2001 ) nowadays called `` downsizing '' . because of this effect , low mass galaxies , in the absence of interaction , are accreting large amounts of gas and are still actively forming stars . in the event of starvation ,
the galaxy stops accreting gas , and continue forming stars from the reservoir already present in the disk .
this proceed at a lower rate than in the unperturbed case , with a declining activity in time as the consumed gas is not replaced with fresh gas . stopping infall ( in order to mimic starvation ) at a given time is straightforward to include in the model .
we shall call @xmath104 the elapsed time since the infall termination ( look back time ) .
in addition to the starvation scenario we can study the effect of ram - pressure gas stripping . as in boselli et al .
( 2006 ) , we adopt the plausible scenario of vollmer et al .
( 2001 ) explicitly tailored to virgo : i.e. , the galaxies being modeled have crossed the dense igm only once ( in its simplest version ) , on elliptical orbits .
the ram pressure exerted by the igm on the galaxy ism varies in time following a gaussian profile , whose peak at ( @xmath105=@xmath27 ) is when the galaxy is crossing the dense cluster core at high velocity ( @xmath105 and @xmath27 are look - back times , where the present epoch corresponds to @xmath105=0 ) .
the gaussian has a width @xmath106 = 9 @xmath59 10@xmath64 years ( see fig@xmath18 3 of vollmer et al .
we make the hypothesis that the gas at each radius is removed at a rate that is directly proportional to the galaxy gas column density @xmath96 and inversely proportional to the potential of the galaxy , measured by the total ( baryonic ) local density @xmath107 ( provided by the model ) .
the gas - loss rate adopted is then equal to @xmath108 @xmath109 , with the efficiency @xmath108 following a gaussian having a maximum @xmath28 at the time @xmath27 , chosen to mimic the variation of the ram pressure suggested by vollmer et al .
this very simple , but physically - motivated prescription should allow us to model the gas removal from galaxies using only two free parameters ( @xmath27 and @xmath28 ) to age - date and measure the magnitude of this effect .
to further constrain the models we assume two different values of @xmath28 : the first is @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 , the value best reproducing the radial profiles of ngc 4569 ( boselli et al . , 2006 ) , the second one is one third of it , @xmath28 = 0.4 m@xmath29 kpc@xmath12 yr@xmath19 .
this second value is suggested by the following considerations : besides the potential of the galaxy , which is an intrinsic property of each object , the efficiency of ram - pressure ( @xmath63 ) stripping depends on the density of the cluster intergalactic medium ( @xmath110 ) , on the crossing velocity of the galaxy within the cluster ( @xmath111 ) and on its orientation with respect to its motion ( boselli & gavazzi 2006 ) : @xmath63 = @xmath110 @xmath112 .
statistically speaking , the line of sight velocity dispersion of star forming galaxies in the cluster is 1150 km s@xmath19 , and drops to 766 km s@xmath19 if limited to low luminosity star forming objects in the virgo a subgroup from where the x ray emission comes from .
these values can be compared to the line of sight velocity of ngc 4569 with respect to the cluster center , which is 1374 km s@xmath19 . to test other realistic models we thus decided to use an @xmath28 lower by a factor of @xmath9 ( 1374/766)@xmath113 = 3.2 than in ngc 4569 .
@xmath28 = 0.4 m@xmath29 kpc@xmath12 yr@xmath19 can thus be considered as an average value of relatively weak interactions while @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 is a realistic value representative of more extreme interactions .
as we discuss in the main body of the paper , @xmath28 = 0.4 m@xmath29 kpc@xmath12 yr@xmath19 leads to hi - deficiency parameters of @xmath9 0.8 for massive galaxies , consistent with the average hi - deficiency observed within one virial radius in virgo , while @xmath28 = 1.2 m@xmath29 kpc@xmath12 yr@xmath19 to hi - deficiencies of @xmath9 1.5 , close to the highest values found in virgo ( gavazzi et al .
2005b ; boselli & gavazzi 2006 ) .
in massive galaxies hi deficiencies as high as @xmath9 2 have been observed ( gavazzi et al . 2005 ) , and are probably the result of stripping events more efficient than those simulated .
these extreme hi deficiencies can also be the result of subsequent crossings of the cluster center , that should happen on average every @xmath9 1.7 gyr ( boselli & gavazzi 2006 ) .
multiple crossing can be reproduced by our models just considering that the average crossing time of star forming galaxies in the virgo cluster is 1.7 gyr ( boselli & gavazzi 2006 ) .
+ we make the further assumption that no extra star formation is induced during the interaction .
this assumption is reasonable since ram pressure stripping models of fujita ( 1998 ) and fujita & nagashima ( 1999 ) indicate that on short time scales ( @xmath9 10@xmath26 yr ) the star formation activity of galaxies can increase by up to a factor of 70% at most in high density , rich clusters such as coma , but it decreases to values lower than before the interaction whenever the density of the igm is relatively low as in the core of virgo .
indeed we do not have any evidence of a statistically significant increase of the star formation activity in galaxies that recently underwent a ram - pressure stripping event ( iglesias - paramo et al .
2004 ) . given the rapidity of the gas stripping process ( 150 myr , see sect .
5.2 ) in these dwarf systems , a possible mild increase of their star formation activity does not have enough time to produce significant modification in their spectrophotometric and structural properties .
|
we study the evolution of dwarf ( @xmath0 @xmath1 10@xmath2 l@xmath3 ) star forming and quiescent galaxies in the virgo cluster by comparing their uv to radio centimetric properties to the predictions of multizone chemo - spectrophotometric models of galaxy evolution especially tuned to take into account the perturbations induced by the interaction with the cluster intergalactic medium .
our models simulate one or multiple ram pressure stripping events and galaxy starvation .
models predict that all star forming dwarf galaxies entering the cluster for the first time loose most , if not all , of their atomic gas content , quenching on short time scales ( @xmath4 150 myr ) their activity of star formation .
these dwarf galaxies soon become red and quiescent , gas metal - rich objects with spectrophotometric and structural properties similar to those of dwarf ellipticals .
young , low luminosity , high surface brightness star forming galaxies such as late - type spirals and bcds are probably the progenitors of relatively massive dwarf ellipticals , while it is likely that low surface brightness magellanic irregulars evolve into very low surface brightness quiescent objects hardly detectable in ground based imaging surveys .
the small number of dwarf galaxies with physical properties intermediate between those of star forming and quiescent systems is consistent with a rapid ( @xmath1 1 gyr ) transitional phase between the two dwarf galaxies populations .
these results , combined with statistical considerations , are consistent with the idea that most of the dwarf ellipticals dominating the faint end of the virgo luminosity function were initially star forming systems , accreted by the cluster and stripped of their gas by one or subsequent ram pressure stripping events .
|
analysis in finite fields is a useful subject because it interacts with other mathematical fields . in addition
, the finite field case serves as a typical model for the euclidean case and possesses structural advantages which enable us to relate our problems to other well - studied problems in number theory , arithmetic combinatorics , or algebraic geometry . for these reasons , problems in euclidean harmonic analysis
have been recently reformulated and studied in the finite field setting .
for example , see @xcite and references therein . in this paper
we investigate @xmath0 estimates of averaging operators over algebraic varieties given by a system of homogeneous polynomials in finite fields . for euclidean averaging problems
, we refer readers to @xcite and @xcite .
we begin with notation and definitions for averaging problems in finite fields .
let @xmath1 be a @xmath2-dimensional vector space over a finite field @xmath3 with @xmath4 elements . throughout this paper ,
we assume that the characteristic of @xmath3 is sufficiently large .
we denote by @xmath5 the counting measure on the space @xmath1 . the pair @xmath6 is named as a function space .
we now consider a frequency space , denoted by the pair @xmath7 , where @xmath8 and @xmath9 denote the dual space of @xmath1 and the normalized counting measure on @xmath8 , respectively .
since @xmath1 is isomorphic to @xmath8 as an abstract group , we identify @xmath1 with @xmath10 .
for instance , we write @xmath11 for @xmath7 .
this convention helps us to avoid complicated notation appearing in doing some computations .
we shorten both @xmath6 and @xmath11 as just @xmath1 if there is no risk of confusion between the function space @xmath6 and the frequency space @xmath11 .
let @xmath12 be an algebraic variety in the frequency space @xmath11 .
we endow @xmath12 with a normalized surface measure , denoted by @xmath13 , which can be defined by the relation @xmath14 where @xmath15 and @xmath16 denotes the cardinality of @xmath12 .
notice that we can replace @xmath17 by @xmath18 , where @xmath19 indicates the characteristic function on @xmath12
. then the convolution function of @xmath20 and @xmath13 is defined on @xmath11 : @xmath21 in the finite field setting , the averaging problem is to determine @xmath22 such that @xmath23 where @xmath24 is independent of the function @xmath20 and the size of the underlying finite field . [
def : ppr ] we use @xmath25 to indicate that inequality holds . as an analogue of averaging problems in euclidean space , this problem has first been addressed by carbery , stones and wright @xcite .
they mainly investigated the @xmath0 estimates of the averaging operator over a @xmath26-dimensional variety given by a vector - valued polynomial @xmath27 .
in particular , carbery , stones , and wright @xcite consider a variety @xmath28 , which is given by the range of @xmath29 defined by @xmath30 for @xmath31 observe that the generalized parabolic variety @xmath32 can be written by @xmath33 where @xmath34 for @xmath35 .
namely , the variety @xmath32 is exactly the collection of the common solutions of the @xmath36 equations : @xmath37 for @xmath35 .
it is clear that @xmath38 for all @xmath39 , because @xmath40 are uniquely determined whenever we choose @xmath41 .
applying the weil theorem @xcite , the aforementioned authors @xcite have obtained the sharp fourier decay estimates on the variety @xmath32 and , as a consequence , they give the complete solution of the averaging problem over the variety @xmath32 . before we present the result of @xcite , we need to introduce one more notation : [ def : conhul ] for points @xmath42 of the euclidean plane , we use @xmath43 to denote their _
convex hull_. we use definitions [ def : ppr ] and [ def : conhul ] to formulate our main results , in which @xmath25 is related to belonging the point @xmath44 to certain convex polygon .
we also denote @xmath45 it is shown in @xcite that @xmath46 however , if the variety @xmath32 is replaced by the homogeneous variety @xmath47 defined as @xmath48 where @xmath49 for @xmath35 , then the averaging problem over @xmath47 becomes much harder .
there are two main reasons why it is difficult to find sharp @xmath0 averaging estimates over @xmath47 .
first , it is not clear to find the size of @xmath47 .
second , the computation of the fourier decay estimate on @xmath47 is not easy , in part because it can not be obtained by simply applying the weil theorem @xcite .
moreover , it may be possible that the fourier decay on @xmath47 is slower than that on @xmath32 , because the homogeneous variety @xmath47 contains lots of lines which could be key factors to make @xmath47 flat .
these reasons suggest that the @xmath0 averaging estimates over @xmath32 maybe be much better than those over @xmath47 . in some cases ,
it is true but is not always true in the finite field setting .
indeed , we show here that if @xmath50 , then @xmath32 and @xmath47 yield the same @xmath0 averaging estimates .
the @xmath0 averaging estimates over @xmath47 depend on the maximal dimension of subspaces lying in the variety @xmath47 .
let us denote by @xmath51 the normalized surface measure on @xmath47 .
for a moment , let us assume that @xmath52 for @xmath53 , which in fact follows from proposition [ pro2 ] and lemma [ lem : lw ] below .
we recall that for any real @xmath54 and @xmath12 , @xmath55 or @xmath56 means that there exists @xmath24 independent of @xmath4 such that @xmath57 , and @xmath58 is used to indicate that @xmath55 and @xmath59 . throughout the paper ,
the implied constants may depend on degrees and the number of variables of the polynomials defining algebraic varieties under consideration , in particular on the integer parameters @xmath60 .
suppose that the following averaging estimate over @xmath47 holds true for @xmath22 : @xmath61 then taking @xmath62 as a test function , it follows that @xmath63 where @xmath64 if @xmath65 and @xmath66 otherwise .
in fact , this necessary condition for @xmath25 has been observed by the authors in @xcite who have also remarked that if @xmath47 contains an @xmath67-dimensional subspace @xmath68 with @xmath69 , then the necessary condition can be improved as @xmath70 where @xmath71 hence , to find a more precise necessary condition for @xmath25 , we need to observe the maximal dimension @xmath67 of the subspaces @xmath68 lying in the homogeneous variety @xmath47 . since @xmath67 is always an integer ,
if @xmath26 is odd , then it is impossible that @xmath47 contains a @xmath72-dimensional subspace @xmath68 . in this case
, if @xmath73 is a square number , then it may happen that @xmath74 ( namely @xmath75 ) .
however , if @xmath26 is even , then @xmath67 can be taken as an integer @xmath72 . combining these observations with and ,
we are lead to the following conjecture ( see figure [ fig ] ) .
[ conj ] for each @xmath53 , let @xmath47 be the homogeneous variety defined as .
* if @xmath26 is even , we have @xmath76 * if @xmath26 is odd and @xmath73 is a square , then @xmath77 where @xmath78 when @xmath79 is odd and @xmath80 , it is observed in @xcite that the conjecture holds true .
in the case when @xmath81 is even and @xmath82 , the conjecture has recently been established in @xcite .
namely , the averaging problem over @xmath83 has been completely solved where @xmath84 and @xmath73 is a square . however , there are no known results on the conjecture for @xmath85 .
our first result below says that conjecture [ conj ] is true for any integer @xmath79 and @xmath50 ( see figure [ fig ] ) . for each @xmath86 ,
let @xmath51 be the normalized surface measure on the homogeneous variety @xmath87 given in .
[ main1 ] if @xmath79 is an integer and @xmath50 , then , assuming that the characteristic of @xmath3 is sufficiently large , @xmath88 [ rem1 ] as mentioned before , this statement has only been known in @xcite for @xmath89 and @xmath50 .
notice from theorem [ main1 ] that the optimal averaging result for the homogeneous variety @xmath90 is exactly the same as that in for the general parabolic variety @xmath91 defined in .
let @xmath92 be the affine @xmath2-space @xmath93 , where @xmath94 denotes the algebraic closure of the finite filed @xmath3 with @xmath4 elements . for each @xmath53 ,
let us consider the algebraic variety @xmath95 where @xmath96 , @xmath97 , are the homogeneous polynomials defined as in .
one interesting point is that the smoothness of @xmath98 depends on the dimension @xmath2 of @xmath92 .
indeed , we see from proposition [ pro3 ] below that the variety @xmath98 is smooth away from the origin if and
only if @xmath99 . in the case when @xmath98 for @xmath100 is smooth away from the origin , we are able to obtain certain @xmath0 averaging estimates on @xmath47 ( see figure [ fig ] ) . next , we state our averaging results over @xmath47 for @xmath100 .
[ main2 ] if @xmath101 , then , assuming that the characteristic of @xmath3 is sufficiently large , @xmath102 the result of theorem [ main2 ] is far from the conjectured averaging result , but in general it gives a sharp @xmath0 estimate for @xmath103 .
is even , @xmath104 is region of @xmath44 of conjecture [ conj ] .
theorem [ main1 ] covers the conjectured region for @xmath50 .
furthermore , @xmath105 indicates the conjectured region in the case when @xmath100 is odd and @xmath73 is a square number .
the region @xmath106 is corresponding to the result of theorem [ main2].,scaledwidth=80.0% ] our work has been mainly motivated by the exponential sum estimates on abstractly given homogeneous varieties due to authors in @xcite . to prove our main results , we first derive a useful result about averaging on general homogeneous varieties with abstract algebraic structures .
then our main results follow by applying it to our variety @xmath47 . to do this ,
we make the following three key observations on @xmath107 for @xmath53 .
[ pro1 ] suppose that the characteristic of @xmath3 is sufficiently large .
then , for every @xmath108 , the algebraic variety @xmath109 is not contained in any hyperplane in @xmath92 .
[ pro2 ] suppose that the characteristic of @xmath3 is sufficiently large .
then for every @xmath108 , the algebraic variety @xmath109 is absolutely irreducible and @xmath110 .
that is , propositions [ pro1 ] and [ pro2 ] assert that @xmath47 is a complete intersection .
[ pro3 ] suppose that the characteristic of @xmath3 is sufficiently large .
then for every @xmath108 , the algebraic variety @xmath109 is smooth away from the origin if and only if @xmath99 .
furthermore , the smoothness condition on @xmath90 is not necessary in completing the proof of theorem [ main1 ] .
therefore , the conclusion of theorem [ main1 ] holds true for any @xmath2 and @xmath50 , and conjecture [ conj ] for @xmath50 is established . on the contrary , we use the smooth condition on @xmath47 for @xmath100 in proving theorem [ main2 ] .
thus , the condition that @xmath99 is imposed to the statement of theorem [ main2 ] .
however , conjecture [ conj ] proposes that such a smooth condition on @xmath47 may not be important in determining @xmath0 averaging estimates over @xmath47 .
we hope that experts in both harmonic analysis and algebraic geometry may be able to shed insight on the conjecture . in the remaining parts of this paper ,
we concentrate on proving theorem [ main1 ] and theorem [ main2 ] which are our main results . instead of proving directly main theorems , we derive them by means of working on more general homogeneous varieties with specific geometric structures . to this end , in section [ newsec2 ] we collect facts about the multiplicative character sums and the existence of a primitive prime divisor of a family of shifted monomials . in particular ,
we make use of a polynomial analogue of the zsigmondy theorem which is due to flatters and ward @xcite . section [ sec2 ] is devoted to setting up notation and basic concepts essential in defining abstract varieties in algebraic geometry . in section [ sec3 ]
, we derive a result for averaging problems over general homogeneous varieties , where we adapt the standard analysis technique in @xcite together with the results on exponential sums in @xcite , see lemma [ lem : homvar ] below .
in fact , this result generalizes our main results related to @xmath47 . in section [ sec4 ] , we show that lemma [ lem : homvar ] applies to the variety @xmath47 and complete the proofs of our main results , that is , theorems [ main1 ] and [ main2 ] .
we note that our main tool are bounds of exponential sums along algebraic varieties , which we interpret as results about the decay of fourier coefficients .
we need the following simple observation , which immediately follows from the taylor formula ( which applies if the characteristic is large enough ) .
[ endlem1 ] for any fixed integer @xmath111 , if the characteristic of @xmath3 is sufficiently large then for any @xmath112 , the polynomial @xmath113 $ ] has no multiple roots .
[ endlem15 ] for any fixed integer @xmath111 , if the characteristic of @xmath3 is sufficiently large then the polynomial @xmath114 $ ] has at least one root which is not a root of the polynomials @xmath115 $ ] , @xmath116 . by a result of flatters and ward (
* theorem 2.6 ) , if the characteristic of @xmath3 is large enough then @xmath117 has an irreducible factor @xmath118 $ ] that does not divide any of the polynomials @xmath119 $ ] , @xmath120 . in particular , @xmath121 is relatively prime with @xmath122 and thus is a divisor of @xmath123 furthermore , @xmath121 is relatively prime to @xmath124 which concludes the proof .
we also need the following result due to wan ( * ? ? ?
* corollary 2.3 ) that follows almost instantly from the weil bound in the form given in ( * ? ? ? * theorem 11.23 ) .
[ endlem2 ] let @xmath125 be @xmath126 monic pairwise prime polynomials in @xmath127 $ ]
. denote by @xmath128 nontrivial multiplicative characters of @xmath3 with order @xmath129 , respectively .
if for some @xmath130 , the polynomial @xmath131 is not of the form @xmath132 with @xmath118 $ ] , then we have @xmath133 [ endlem3 ] for any fixed integer @xmath111 , if the characteristic of @xmath3 is sufficiently large , then for any multiplicative characters @xmath134 , @xmath135 , among which at least one is nontrivial , we have @xmath136 after ignoring all trivial characters , it suffices to prove that for some positive integer @xmath137 , we have @xmath138 where @xmath139 and @xmath140 denote nontrivial multiplicative characters of @xmath3 .
factoring the polynomials @xmath141 , @xmath142 , into irreducible factors over @xmath143 and using the multiplicativity , we see from lemma [ endlem1 ] that @xmath144 for some multiplicative characters @xmath145 and monic pairwise prime polynomials @xmath146 , @xmath147 .
furthermore , by lemma [ endlem15 ] we have @xmath148 for at least one @xmath149 , and thus @xmath150 is a nontrivial character . now using lemma [ endlem2 ] we complete the proof .
in this section , we review known facts on general varieties generated by a system of @xmath151homogeneous polynomials in @xmath152 $ ] .
we begin by setting up notation .
let @xmath153 be an integer .
assume we are given @xmath151homogeneous polynomials in @xmath2 variables over @xmath143 of degree at least two each , which we write as @xmath154 , \quad \deg f_j \ge 2 , \qquad j = 1 , \ldots , s,\ ] ] where @xmath155 .
now , define the closed algebraic set @xmath156 let @xmath157 be the collection of points in @xmath158 with coordinates in @xmath159 @xmath160 we also use the standard notation @xmath161 for the @xmath162-dimensional projective space over @xmath94 , which can be considered as the collection of all one dimensional subspaces of the vector space @xmath92 . for @xmath163\in \mathbb p^{d-1}$ ] and a polynomial @xmath164 $ ] , recall that @xmath165 means that @xmath166 for all @xmath167 .
like the algebraic subset @xmath158 of the affine space @xmath92 , we define the projective algebraic set @xmath168 let us recall an affine cone over a projective subset in @xmath161
. denote by @xmath169 the projection map defined by @xmath170.\ ] ] then the affine cone over @xmath171 is defined by @xmath172 notice that @xmath158 is the affine cone over the projective variety @xmath173 .
[ def1 ] we say that a homogeneous variety @xmath174 defined as in is a * complete intersection * if the following two conditions hold : * @xmath175 is an affine cone over a projective variety @xmath173 which is not contained in a hyperplane , * @xmath158 is an absolutely irreducible variety of dimension @xmath176 ( or @xmath177 ) .
[ def2 ] we say that a homogeneous variety @xmath174 defined as in is * smooth * if @xmath158 is smooth away from the origin . it is observed in @xcite that if @xmath157 is a smooth homogeneous variety , which is a complete intersection , then @xmath178 can have at most isolated singularities in @xmath161 where @xmath179 for @xmath180 , where @xmath181 denotes the inner products of the vectors @xmath182 and @xmath183 . from this observation , they state and prove the following exponential sum estimates on @xmath157 ( see ( * ? ? ? * theorem 1 ) ) .
[ igor ] let @xmath184 be defined as in .
suppose that @xmath157 is a * smooth * homogeneous variety , which is a complete intersection , and the characteristic of @xmath3 is sufficiently large .
then we have for all @xmath185 @xmath186 where @xmath187 denotes a nontrivial additive character of @xmath3 . here , we point out that the proof of lemma [ igor ] for @xmath188 was given by the authors in @xcite without using the smoothness assumption on @xmath157 . therefore , the smoothness condition on @xmath157 can be relaxed for @xmath188 .
indeed , the following bound follows immediately from a result of cochrane ( * ? ? ?
* theorem 4.3.5 ) .
[ igor1 ] if @xmath184 is a homogeneous variety which is a complete intersection given by of dimension @xmath189 , where @xmath158 is given by , then @xmath190 for all @xmath191
. the following estimate on the cardinality of @xmath157 due to chatzidakis , van den dries and macintyre ( * ? ? ?
* proposition 3.3 ) gives an extension of the result of lang and weil @xcite .
[ lem : lw ] suppose that @xmath192 is an algebraic variety with @xmath193 absolutely irreducible components and of dimension @xmath194 defined by polynomials over @xmath143 and let @xmath195 .
then @xmath196 it is clear from lemma [ lem : lw ] that @xmath197 if @xmath157 is a homogeneous variety , given by which is a complete intersection in @xmath1 .
now , we endow a homogeneous variety @xmath157 with the normalized surface measure @xmath198 . recall that if @xmath199 , then @xmath200 the following decay estimates of the fourier coefficients @xmath201 on @xmath157 follow immediately from lemma [ igor1 ] and [ igor ] .
[ maindecay ]
let @xmath198 be the normalized surface measure on the homogeneous variety @xmath184 given by , which is a complete intersection .
if the characteristic of @xmath3 is sufficiently large , then : * if @xmath188 , then we have @xmath202 for all @xmath191 . * if @xmath157 is smooth and @xmath203 , then @xmath204 for all @xmath191 .
first we need the following general result which can be obtained by adapting the arguments in @xcite . for the sake of completeness
, we provide the proof in full detail .
[ formula ] let @xmath198 be the normalized surface measure on an affine homogeneous variety @xmath174 given by , which is a complete intersection . if @xmath205 for all @xmath191 and for some fixed @xmath206 , then @xmath207 holds with @xmath208 we must show that @xmath209 for all function @xmath210 with the above values of @xmath211 and @xmath212 .
define a function @xmath213 on @xmath214 by @xmath215 .
observe that @xmath216 for @xmath217 , where for a function @xmath218 on @xmath214 we define @xmath219 and , as before , @xmath181 denotes the inner products of @xmath182 and @xmath183 .
since @xmath206 and @xmath9 is the normalized counting measure on @xmath1 , it follows from young s inequality ( see @xcite ) that @xmath220 thus , it suffices to prove that @xmath221 for all functions @xmath222 .
notice that can be obtained by interpolating @xmath223 and @xmath224 it remains to prove and . by the definition of @xmath213 and the assumption
that @xmath225 for @xmath226 , we see that @xmath227 therefore , follows by applying the plancherel theorem ( see @xcite ) : @xmath228 to prove , notice that @xmath197 , because @xmath157 is a complete intersection . from young s inequality and the observation that @xmath229 , we obtain : @xmath230 thus , the proof of lemma [ formula ] is complete . as a direct application of the fourier decay estimates in lemma [ maindecay ]
, we can now derive averaging results related to general homogeneous variety @xmath157 , which is a complete intersection .
applying lemma [ formula ] with lemma [ maindecay ] yields the result below .
[ lem : homvar ] let @xmath198 be the normalized surface measure on a homogeneous variety @xmath231 , given by , which is a complete intersection .
if the characteristic of @xmath3 is sufficiently large , then : * if @xmath188 , then @xmath232 * if @xmath157 is a smooth and @xmath203 , then @xmath233 to prove * ( i ) * , let us assume that @xmath188 .
then @xmath234 , because @xmath157 is a complete intersection .
now , suppose that @xmath25 .
in particular , we see that @xmath235 it therefore follows that @xmath236 by duality , we also have @xmath237 where @xmath238 denote the hlder conjugates of @xmath211 and @xmath212 , respectively . in conclusion ,
@xmath239 conversely , we now assume that the inclusion holds . if @xmath240 , then it is clear that @xmath25 , because both @xmath198 and @xmath241 have total mass @xmath242 . by the interpolation theorem , it therefore suffices to prove that @xmath243 holds . since @xmath188 , applying lemma [ formula ] with lemma [ maindecay ] * ( i ) * yields the above property , and the the proof of lemma [ lem : homvar ] * ( i ) * is complete . in order to prove * ( ii ) * , it is enough to show that @xmath244 holds
. however , this follows immediately by using lemma [ formula ] together with lemma [ maindecay ] * ( ii)*. even if lemma [ lem : homvar ] provides us of powerful averaging results on general homogeneous varieties , applying it in practice may not be simple , because it contains certain abstract hypotheses .
in this section , we complete the proofs of theorem [ main1 ] and theorem [ main2 ] which are considered as main theorems in this paper .
we complete the proofs by showing that lemma [ lem : homvar ] is a general version of both theorem [ main1 ] and theorem [ main2 ] . to do this , we begin by recalling from that for each @xmath53 , our homogeneous variety @xmath47 is exactly the common solutions in @xmath1 of a system of the @xmath245 equations @xmath246 where @xmath247 .
let @xmath248 which is the number of homogeneous equations @xmath96 defining @xmath47 .
then it is clear that @xmath98 is an affine cone over its corresponding projective variety determined by @xmath126-homogeneous polynomials @xmath96 .
thus , if we are able to show that the conclusions of propositions [ pro1 ] and [ pro2 ] hold for @xmath50 then theorem [ main1 ] follows from lemma [ lem : homvar ] * ( i)*. furthermore , if all propositions [ pro1][pro3 ] hold true for @xmath100 , then theorem [ main2 ] follows from lemma [ lem : homvar ] * ( ii ) * where the smoothness condition on @xmath98 is essential . in summary , to prove both theorem [ main1 ] and theorem [ main2 ] , it suffices to justify propositions [ pro1][pro3 ] .
since any hyperplane in @xmath92 is a subspace with dimension @xmath249 , it suffices to prove that there exists @xmath2 linearly independent points @xmath250 .
now fix @xmath251 . for each @xmath252 , choose a @xmath253 such that @xmath254 and @xmath255 . since @xmath256 is an algebraic closure and
the characteristic of @xmath3 is sufficiently large , the @xmath257 always exists .
denote by @xmath258 the @xmath259 identity matrix .
we also define @xmath260 as the @xmath261 matrix whose all entries are @xmath242 . also define the following @xmath262 matrix @xmath263 and the @xmath264 matrix @xmath265 : @xmath266 , \\ & d_{(d - k)\times ( d - k ) } = \left[\begin{matrix } \beta_{k+1 } & 1 & \cdots & 1 & 1\\ 1 & \beta_{k+2}&1 & \cdots & 1\\ \vdots&\vdots & \vdots&\vdots&\vdots \\ 1 & \cdots&1 & \beta_{d-1}&1\\ 1 & 1&\cdots & 1&\beta_d \end{matrix}\right ] . \end{split}\ ] ] now consider the @xmath267 matrix @xmath268 defined by @xmath269 = \left [ \begin{matrix } i_{k\times k } & 1_{k\times ( d - k)}\\ c_{(d - k)\times k } & d_{(d - k)\times(d- k ) } \end{matrix}\right].\ ] ] note that all @xmath270 are solutions of a system of equations .
hence , it follows that @xmath271 for any @xmath251 .
moreover , since @xmath272 for all @xmath35 , it follows from simple gauss elimination that the rank of the matrix @xmath273 is exactly @xmath2 , which completes the proof of proposition [ pro1 ] .
recall from that @xmath274 is given by a system of @xmath245 homogeneous equations .
for each @xmath35 , define an algebraic set @xmath275 where @xmath96 is defined by . by the definition of @xmath98
, it follows that @xmath276 we need the following claim . [ claim1 ] for each @xmath277 , we have @xmath278 and there exists @xmath279 such that @xmath280 the first part is trivial . for the second part of this claim , fix @xmath277 and let @xmath281 with @xmath282 .
now , choose an @xmath283 whose coordinates satisfy that @xmath284 then it is straightforward to check that @xmath285 and the result follows . to compute the dimension of @xmath98 ,
we apply the following result , see ( * ? ? ?
* page 55 ) for a proof .
[ dimlem ] let @xmath286 be an irreducible algebraic set , and let @xmath287 $ ] be a nonconstant polynomial which does not vanish identically on @xmath288 .
in addition , let us define @xmath289 .
if @xmath290 , then we have @xmath291 we are ready to prove proposition [ pro2 ] .
it is not hard to see that @xmath292 , @xmath86 , is absolutely irreducible , because it is the same as the absolute irreducibility of the polynomial @xmath293 assume @xmath294 . clearly , we see @xmath295 .
write @xmath296 we see that we should have @xmath297 and so @xmath298 , which is easy to rule out for @xmath299 ( for example , by specializing @xmath300 . since @xmath292 , @xmath86 , is absolutely irreducible , it follows from the affine jacobian criterion that @xmath301 ( see lemma [ jac ] below ) .
notice that this completes the proof of proposition [ pro2 ] in the case when @xmath80 with @xmath302 .
thus , we may assume that @xmath303 .
observe by induction that proposition [ pro2 ] is a direct result from the following statement .
[ irreducible ] assume that the characteristic of @xmath3 is sufficiently large .
let @xmath304 with @xmath305 .
suppose that @xmath306 is absolutely irreducible with dimension @xmath307 .
then @xmath308 is also absolutely irreducible with dimension @xmath309 . from lemma [ claim1 ] and lemma [ dimlem ] ,
it is clear that @xmath310 thus , it remains to prove that @xmath308 is absolutely irreducible .
assume that @xmath308 has @xmath193 absolutely irreducible components in @xmath94 .
by and lemma [ lem : lw ] to show that @xmath311 , it is enough to prove that @xmath312 where @xmath313 . notice that @xmath314 is the number of common solutions in @xmath1 of the following equations @xmath315 for each @xmath316 , define @xmath317 since @xmath318 are free variables and @xmath319 depend only on @xmath320 , we can write @xmath321 in order to prove , it therefore suffices to show that @xmath322 for each @xmath323 , let @xmath324 and denote by @xmath134 the multiplicative character of order @xmath325 .
then , from the orthogonality of multiplicative characters it follows that @xmath326 see ( * ? ? ?
* section 3.1 ) .
hence , the left hand side of is written by @xmath327 when @xmath328 , the sum over @xmath320 is @xmath329 , where we use the usual convention that @xmath330 for the trivial multiplicative character @xmath331 .
thus , to establish , it is enough to prove that for each @xmath332 with @xmath333 , @xmath334 now , we define the sets @xmath335 ( of ` bad ' and ` good ' vectors @xmath336 ) by @xmath337 and @xmath338 for each fixed @xmath339 , define @xmath340 note that @xmath341 for @xmath342 and all @xmath323 . since @xmath343
, it suffices to prove that for each @xmath332 with @xmath333 , @xmath344 from the definition of @xmath134 and the fact that @xmath345 , notice that @xmath346 is a nontrivial character for some @xmath347 .
if @xmath346 is a trivial character , then the term @xmath348 can be replaced by 1 .
thus , it suffices to prove under the assumption that all @xmath346 are not - trivial characters .
we now consider the cases @xmath50 and @xmath349 separately . *
if @xmath50 , we must show that @xmath350 for all @xmath345 with @xmath333 . recall that if @xmath351 , then @xmath352 , thus @xmath353 and recalling lemma [ endlem3 ] we obtain the desired estimate . * if @xmath354 , then it is easy to show that if the characteristic of @xmath3 is sufficiently large , then for all but @xmath355 choices of @xmath356 , the polynomials @xmath357 , \qquad j=1 , \ldots , n+1\ ] ] have no pairwise common roots .
indeed , assume @xmath100 .
let @xmath358 with @xmath359 .
notice that if @xmath360 and @xmath361 have a common root then @xmath362 .
for our expressions for @xmath106 and @xmath363 in @xmath364 ( assuming that @xmath354 ) one can easily show that this leads to a nontrivial equation and thus has @xmath355 solutions .
for such @xmath365 , the inner sum over @xmath366 in is trivially estimated as @xmath4 , and for the remaining choices of @xmath365 , we apply lemmas [ endlem1 ] and [ endlem2 ] to estimate the inner sum over @xmath366 in . in conclusion ,
the left hand side of is bounded by @xmath367 .
this establishes for every @xmath299 and concludes the proof .
as we have mentioned , lemma [ irreducible ] implies proposition [ pro2 ] .
the proof is based on the affine jacobian criterion below ( see ( * ? ? ? * proposition 4.4.8 ) ) .
[ jac]let @xmath286 be an irreducible algebraic set given by a system of @xmath126-polynomial equations @xmath37 , @xmath368 .
suppose that @xmath369 .
then @xmath288 is smooth at @xmath370 if and only if the rank of the @xmath371 jacobian matrix satisfies @xmath372_{s\times d } \ge d-\dim\overline{v}.\ ] ] now , since @xmath36 is the number of the polynomials @xmath96 in defining @xmath98 which is absolutely irreducible with dimension @xmath26 by proposition [ pro2 ] , we see from lemma [ jac ] that @xmath98 is smooth away from the origin if and only if @xmath373_{(d - k)\times d}=d - k\ ] ] for all @xmath374 . since we have assumed that the characteristic of @xmath3 is sufficiently large , it is clear from the gauss elimination that @xmath373_{(d - k)\times d}= \rank j_{k , d}(x),\ ] ] where @xmath375 denotes the @xmath376 matrix given by the concatenation @xmath377\ ] ] of the wandermonde @xmath378_{(d - k)\times k},\ ] ] and the diagonal matrix @xmath379.\ ] ] in order to complete the proof of proposition [ pro3 ] , it therefore suffices to prove the following two statements : * if @xmath380 and @xmath302 , then there exists @xmath381 with @xmath382 ; * if @xmath99 and @xmath302 , then @xmath383 for all @xmath384 . first , let us prove * ( a1)*. suppose that @xmath302 is an even integer . for each @xmath385 , choose an @xmath386 with @xmath387 , and define @xmath388 letting @xmath389 , it is easy to check that @xmath390 .
since @xmath380 , the matrix @xmath375 has at least four rows , and its second row and fourth row are exactly same .
thus , the rank of @xmath375 must be less than @xmath36 .
next , assume that @xmath100 is an odd integer . for each @xmath385 ,
select a @xmath391 with @xmath392 , and define @xmath393 taking @xmath394 , we also see that @xmath390 , and the second row and the fourth row of @xmath375 are same . thus , the rank of @xmath375 is less than @xmath36 , which completes the proof of the statement @xmath395 .
now , we prove the statement * ( a2)*. if @xmath396 and @xmath397 , then * ( a2 ) * is clearly true , because @xmath398 \neq \left [ \begin{matrix}0&0\cdots & 0 & 0\end{matrix}\right]\ ] ] for @xmath399 .
assume that @xmath400 and @xmath397 .
we must show that @xmath401=2\ ] ] for all @xmath381 , where @xmath402 notice that if @xmath403 , then @xmath404 for some @xmath405 . without loss of generality
, we therefore assume that @xmath406 .
letting @xmath407 for @xmath408 , it is enough to show that @xmath409=2,\ ] ] where @xmath410 satisfies @xmath411 notice that if @xmath412 , then holds , because @xmath413=2.\ ] ] if @xmath414 or @xmath242 for all @xmath415 , then we see from that @xmath416 and so there is nothing to prove .
on the other hand , if @xmath417 for some @xmath418
then @xmath419=u_j(u_j-1 ) \neq 0 \quad\text{or}\quad \rank \left[\begin{array}{cc } 1 & u_j\\ 1 & u_j^2\end{array}\right]=2.\ ] ] thus is also true and we completes the proof of the statement * ( a2 ) * in the case when @xmath400 and @xmath420 .
finally let us prove the statement * ( a2 ) * when @xmath421 and @xmath302 .
following the previous arguments , our task is to show that @xmath422=3,\ ] ] where @xmath410 satisfies @xmath423 * case 1 * : suppose that @xmath414 or 1 for all @xmath424 .
then it follows from that @xmath425 , which implies that @xmath426\neq 0\ ] ] hence @xmath427=3.\ ] ] thus also follows .
* case 2 * : suppose that @xmath428 with @xmath429 for some @xmath430 .
then it follows that @xmath431=u_iu_j(u_i-1)(u_j-1)(u_j - u_i)\neq 0.\ ] ] thus @xmath432=3,\ ] ] which implies .
* case 3 * : suppose that @xmath433 for some @xmath424 , and @xmath434 if @xmath435 for @xmath436 .
let @xmath437 then @xmath438 and @xmath439 are integers .
thus is same as @xmath440 where @xmath441 .
we now claim that either @xmath442 or @xmath443 . to see this , assume that @xmath444 . then from we see that @xmath445 .
this implies that @xmath446 .
thus we conclude that @xmath447 , because @xmath433 and @xmath448 ( assuming that the characteristic of @xmath3 is sufficiently large ) .
however , since @xmath449 , it is impossible that @xmath447 ( again assuming that the characteristic of @xmath3 is sufficiently large ) and the claim is justified .
if @xmath442 , then follows by the observation that @xmath450=-u_{d-2}u_j^2(u_j-1)\neq 0.\ ] ] on the other hand , if @xmath443 , then also follows by the observation that @xmath451=-u_{d}^3u_j(u_j-1)\neq 0.\ ] ] by * case 1 * , * case 2 * and * case 3 * , we establish the statement * ( a2 ) * in the case when @xmath421 and @xmath302 .
this concludes the proof of proposition [ pro3 ] .
the authors are grateful to anthony flatters and tom ward for useful discussions and in particular for the idea of the proof of lemma [ endlem15 ] .
the first author was supported by the research grant of the chungbuk national university in 2012 and basic science research program through the national research foundation of korea funded by the ministry of education , science and technology ( 2012010487 ) .
the third author was supported by the research grant of the australian research council ( dp130100237 ) .
|
in this paper we study the mapping properties of the averaging operator over a variety given by a system of homogeneous equations over a finite field . we obtain optimal results on the averaging problems over two dimensional varieties whose elements are common solutions of diagonal homogeneous equations . the proof is based on a careful study of algebraic and geometric properties of such varieties . in particular
, we show that they are not contained in any hyperplane and are complete intersections .
we also address partial results on averaging problems over arbitrary dimensional homogeneous varieties which are smooth away from the origin .
|
in this paper we deal with a semi - infinite ( @xmath0 ) homogeneous and isotropic viscoelastic medium of density @xmath1 , which is assumed to be at rest for @xmath2 . for @xmath3 , the accessible extremum of the body ( defined by @xmath4 )
is subjected to a perturbation denoted by @xmath5 . here , we are interested in the solution of the wave equation for the response function of the material @xmath6 . in the framework of a linear theory , this problem can be analyzed by working in either the creep or the relaxation representation .
indeed , it is known that a linear viscoelastic body is featured by either the creep compliance ( the strain response to a step input stress ) or the relaxation modulus ( the stress response to a step input of strain ) , for further details see @xcite .
we fourther assume that the dynamics of the material is described in terms of a given bessel model of order @xmath7 ( see @xcite ) .
for such a model one has that the dynamical equation for the response function @xmath8 , in the creep representation , is given by @xmath9 \ ,
\frac{\partial ^2 r}{\partial t ^2 } = c^2 \ , \frac{\partial ^2 r}{\partial x ^2 } \ , , \quad \nu > -1 \ , , \ ] ] where @xmath10 represents the laplace convolution , @xmath11 is the _ wave - front velocity _ and @xmath12 is the _ rate of creep function _ for the model .
if we denote by @xmath13 the creep compliance , and with @xmath14 , then we have that @xmath15 and @xmath16 in particular , for our model the rate of creep , in the laplace domain , is provided by ( see @xcite ) @xmath17 so that the rate of creep function , in the time domain , is given by @xmath18 where @xmath19 s are the modified bessel functions of the first kind , @xmath20 is the @xmath21th positive real zero of the bessel function of the first kind @xmath22 .
it is worth remarking that the rate of creep is a completely monotonic function with a _ weak power law singularity _ as @xmath23 . given this brief introduction to the bessel models of linear viscoelasticity , we have that the aim of this paper is to compute the response @xmath8 of the material for @xmath24 as @xmath25 and @xmath26 .
first of all , let us summarize an important technique that allows us to compute the asymptotic wave - front expansions for _ singular _ viscoelastic models ( models that exibit a non - analytic creep compliance featured by power law singularities as @xmath23 ) .
such a procedure was first introduced by buchen and mainardi in @xcite .
the creep representation of the equation of motion ( [ eq - motion ] ) , in the laplace domain , is given by @xmath27 { \widetilde{r } } ( s , x ) = 0 \ , , \ ] ] where @xmath28 ^{1/2 } \ , .\ ] ] here @xmath29 is the creep compliance of a given viscoelastic model in the laplace domain . from the boundary condition we can easily infer the formal solution of eq .
( [ eq - motion - laplace ] ) , @xmath30 \ , .\ ] ] for a singular model it is known ( see @xcite ) that @xmath31 will have an expansion in terms of decreasing powers of @xmath32 , as @xmath32 goes to infinity .
thus , one could write this expansion as @xmath33 now , it is important to define two quantities whose relevance will appear clear later in this discussion . then , let us denote @xmath34 the sum of the first @xmath35 ( @xmath36 ) terms of eq .
( [ eq - mu - as ] ) that depend on positive powers of @xmath32 , @xmath37 moreover , we denote with @xmath38 the remainder of the series ( [ eq - mu - as ] ) .
therefore , taking profit of these definitions , we can rewrite the solution in eq .
( [ eq - sol - laplace ] ) , for @xmath39 , as follows @xmath40 \ , { \widetilde{r } } ( s , x ) \ , , \ ] ] with @xmath41 \ , .\ ] ] now , from the equation of motion in the laplace domain ( [ eq - motion - laplace ] ) , it is clear that the function @xmath42 is required to satisfy the following differential equation @xmath43
\right\ } \ , { \widetilde{r } } ( s , x ) = 0 \ , , \ ] ] together with the condition @xmath44 if we now expand @xmath31 and @xmath34 in powers of @xmath32 , as @xmath39 , in eq .
( [ eq - l ] ) and then divide the whole equation for the term containing the highest power of @xmath32 , we get a new differential operator @xmath45 .
this is nothing but the rescaled version of @xmath46 and it is featured by the following asymptotic expansion @xmath47 where @xmath48 and where the coefficients @xmath49 , @xmath50 and @xmath51 , together with the exponents @xmath52 , are determined by the asymptotic expansion of @xmath31 as @xmath39 .
now , one can easily observe that @xmath42 can be express as a ( negative ) power series @xmath53 for some functions @xmath54 .
moreover , we also have that the operator @xmath45 matches the conditions of the friedlander - keller theorem ( see @xcite ) .
therefore , given this last remark for @xmath45 one can prove ( see @xcite ) that the functions @xmath54 are actually given by @xmath55 where the coefficients @xmath56 are computed by means of the recursive system @xmath57 \qquad 1 \leq \ell \leq k \
a_{k , \ell } = 0 \qquad \ell > k \ , .
\end{cases}\ ] ] moreover , the function @xmath58 is deduced by the condition @xmath59 and the coefficients @xmath60 are such that @xmath61 with the @xmath62 s are positive integers . as discussed in @xcite and extended and perfected in a quite general fashion in @xcite , if we now plug what we have found so far for @xmath42 into eq .
( [ eq - sol - laplace - as ] ) , we have that @xmath63 \ , . \end{aligned}\ ] ] then , defining @xmath64 \ , , \ ] ] we get we have that @xmath65 . ] @xmath66 \ , \sum _ { k=0 } ^\infty v_k ( x ) \ , { \widetilde{\phi } } _ k ( s , x ) \ , .\ ] ] for a mathematical discussion of the functions @xmath67 we invite the interested reader to refer to @xcite .
then , inverting this last result back to the time domain ( see @xcite ) we get the asymptotic expansion for @xmath8 as @xmath68 , @xmath69 where the functions @xmath70 are computed by means of eq . ( [ eq - poly ] ) and ( [ eq - system - a ] ) and the functions @xmath71 are the laplace inverse of the functions @xmath67 in eq .
( [ eq - phi - laplace ] ) .
clearly , this is just a summary of the technique .
the interested reader can find a complete description of this algorithm in @xcite .
in this section we discuss the problem of the propagation of transient pulses within a semi - infinite bessel medium .
such a problem has a very well known formal solution in the laplace domain ( see @xcite ) , however inverting back to the time domain is often very difficult even for simple viscoelastic models . here
we present an asymptotic expansion of the solution of the propagation problem in the neighborhood of the pulse onset . to do this ,
we take profit of the buchen - mainardi algorithm , discussed and extended in section [ sec-2 ] . for sake of simplicity , we consider two inputs : @xmath72 ( or , in the laplace domain , @xmath73 ) and an inpulsive input @xmath74 ( or , in the laplace domain , @xmath75 ) , where @xmath76 and @xmath77 are respectively the heaviside and dirac generalized functions .
this is done without loss of generality , indeed the general response to an arbitrary input @xmath5 can be obtained by convolution . in order to further simplify our discussion , given
that for a general bessel body we have that @xmath78 , let us also assume @xmath79 . then , as discussed in @xcite , for a bessel body we have that @xmath80 hence , @xmath81^{1/2 } \sim s \ , [ 1 + 2 \ , ( \nu + 1 ) \ , s^{-1/2}]^{1/2 } \sim\\ & \sim s \ , \left [ 1 + ( \nu + 1 ) \ , s^{-1/2 } - \frac{(\nu + 1)^2}{2 } \ , s^{-1 } + \sum _ { n = 3 } ^\infty \binom { 1/2 } { n } \ , 2^{n } \ , ( \nu + 1)^n \ , s^{-n/2 } \right ] \sim\\ & \sim s + ( \nu + 1 ) \ , s^{1/2 } - \frac{(\nu + 1)^2}{2 } + \sum _ { n = 3 } ^\infty \binom { 1/2 }
, 2^{n } \ , ( \nu + 1)^n \ , s^{1-n/2 } \ , , \end{split}\ ] ] as @xmath39 , where we have also taken profit of the taylor expansion @xmath82 following the procedure presented in section [ sec-2 ] , we set @xmath83 now , from eq .
( [ eq - mu ] ) and eq .
( [ eq - mu+ ] ) we have @xmath84 \sim s^2 + 2 \ , ( \nu + 1 ) \ , s^{3/2 } \ , , \\
_ + ^2 ( s ) & \!\!=\!\ ! & \left ( s + ( \nu + 1 ) \ , s^{1/2 } - \frac{(\nu + 1)^2}{2 } \right ) ^2 = \\ & \!\!=\!\ ! & s^2 + 2 \ , ( \nu + 1 ) \ ,
s^{3/2 } - ( \nu + 1)^3 \ , s^{1/2 } + \frac{(\nu + 1)^4}{4 } \ , , \notag\end{aligned}\ ] ] from which we get @xmath85 therefore , the differential operator @xmath45 , defined in eq .
( [ eq - l ] ) , for a bessel medium reduces to @xmath86 \ , \frac{\partial}{\partial x}\\ & - \bigg [ ( \nu + 1)^3 \ , s^{1/2 } - \frac{(\nu + 1)^4}{4 } \bigg]\bigg\ } \ , , \end{split}\ ] ] as @xmath39 .
if we rescale this operator by dividing the previous expression by the term containing the highest power of @xmath32 , i.e. @xmath87 , we get @xmath88 \ , \frac{d}{dx } \\ & + \bigg [ \frac{(\nu + 1)^3}{2 } \frac{1}{s^{1/2 } } - \frac{(\nu + 1)^4}{8 } \frac{1}{s } \bigg]\bigg\ } \ , .
\end{split}\ ] ] then , grouping the differential operators in the previous expression in terms of powers of @xmath32 , we get @xmath89 \frac{1}{s^{1/2 } } \\ & + \bigg [ - \frac{1}{2 } \frac{d^{2}}{dx^{2 } } - \frac{(\nu + 1)^2}{2 } \frac{d}{dx } - \frac{(\nu + 1)^4}{8 } \bigg ] \frac{1}{s } \ ,
\bigg\ } \ , .
\end{split}\ ] ] therefore ,
the rescaled operator can be re - expressed as @xmath90 where , @xmath91 thus , comparing the rescaled operator with the first hypothesis of the theorem by friedlander and keller , @xmath92 we infer that for a bessel solid @xmath93 , @xmath94 , @xmath95 , @xmath96 and @xmath97 , @xmath98 . moreover , the coefficients @xmath60 can be deduced by the system @xmath99 where @xmath100 is the number of terms in the asymptotic expansion of @xmath45 .
indeed , in our case we have that the only non - vanishing terms that appear in the expansion of the operator @xmath45 feature the following values of @xmath52 , @xmath101 therefore , form eq .
( [ eq - lambdak ] ) we have @xmath102 from which we can infer that the coefficients @xmath60 in the buchen - mainardi algorithm for the scenario of our concern are given by @xmath103 , with @xmath104 .
finally , form the condition @xmath59 , we infer that @xmath105 .
now , taking profit of the system in eq .
( [ eq - system - a ] ) , we can compute the values of the coefficients @xmath56 for a generic bessel medium .
in particular , given the previous discussion we have that the system in eq .
( [ eq - system - a ] ) can be rewritten as @xmath106 in particular , the second line can be rewritten , dropping the sum and the delta , as @xmath107 now , accordingly with the discussion in section [ sec-2 ] , we can immediately read off the values of all @xmath108 s , @xmath50 s and @xmath51 s from the expressions for @xmath109 and @xmath110 , indeed @xmath111 whereas @xmath112 . then ,
if we focus on the coefficients such that @xmath113 we get @xmath114 = \\ & = - \bigg [ \left ( ( \nu + 1 ) \ , a_{k - 1 , \ , \ell } + \frac{(\nu+1)^3}{2 } \ , a_{k - 1 , \ , \ell-1 }
\right ) + \\ & + \left ( -\frac{1}{2 } \ , a_{k - 2 , \ , \ell+1 } - \frac{(\nu+1)^2}{2 } \ , a_{k - 2 , \ , \ell } - \frac{(\nu+1)^4}{8 } \ ,
a_{k - 2 , \ , \ell-1 }
\right ) \bigg ] \ , .
\end{split}\ ] ] thus , @xmath115 \ , \end{split}\ ] ] now , if we set @xmath116 these coefficients reduce to @xmath117 \ , , \
] ] that corresponds exactly to the one in @xcite for the _ fractional maxwell model of order 1/2_. in order to conclude this section , we still need to explicitly all the parts of eq .
( [ eq - fin ] ) .
first of all , let us rewrite eq .
( [ eq - fin ] ) in terms of the polynomials @xmath54 , @xmath118 then , to compute @xmath119 one just need to perform the inversion of its laplace transform in eq .
( [ eq - phi - laplace ] ) .
specifically , if we consider a step input @xmath120 we have that ( see appendix [ app - a ] ) @xmath121 \ , ( t - x)^{k/2 } \ , f_{1/2 } \left(z(\nu , t , x ) , \ , \frac{k}{2 } \right ) \ , , \ ] ] where @xmath122 here we have also taken profit of the fact that for a bessel body @xmath103 , as shown above , and that @xmath123 = e^{- x \ , s } \ , { \widetilde{f } } ( s ) \ , , \ ] ] and @xmath124 given the previous results , we are now capable to write down the explicit wave - front expansion , given a step input , for a generic viscoelastic bessel medium , @xmath125 \ , \sum _ { k=0 } ^\infty \ , \sum _ { \ell = 0 } ^k a_{k , \ell } \ , \frac{x^\ell}{\ell ! } \ , ( t - x)^{k/2 } \ , f_{1/2 } \left(z(\nu , t , x ) , \ , \frac{k}{2 } \right ) \ , .\ ] ] where @xmath126 is defined as above and where the coefficients @xmath56 are determined as in eq .
( [ eq - akl ] ) . in the following
we show some plots of the wave - front expansion provided in eq .
( [ eq - short - time ] ) .
we consider the specific case of @xmath127 because it will tourn out to be relevant in biophysics
. depicted versus @xmath128 for some fixed values of @xmath129.,width=377 ] analogously , if we consider impulsive input @xmath130 , @xmath131 , one can easily show that @xmath121 \ , ( t - x)^{(k-2)/2 } \ , f_{1/2 } \left(z(\nu , t , x ) , \ , \frac{k - 2}{2 } \right ) \ , , \ ] ] where @xmath122 therefore , the wave - front expansion for a bessel model of order @xmath7 , given an impulsive input , is provided by @xmath132 \ , \sum _ { k=0 } ^\infty \ , \sum _ { \ell = 0 } ^k a_{k , \ell } \ , \frac{x^\ell}{\ell ! } \ , ( t - x)^{(k-2)/2 } \ , f_{1/2 } \left(z(\nu , t , x ) , \ , \frac{k - 2}{2 } \right ) \ , .\ ] ] where @xmath126 is defined as above and where the coefficients @xmath56 are determined as in eq .
( [ eq - akl ] ) . in the following
we show some plots of the wave - front expansion provided in eq .
( [ eq - short - time-2 ] ) . depicted versus @xmath128 for some fixed values of @xmath129.,width=377 ]
the case @xmath127 corresponds to a peculiar model for large arteries @xcite , in which they can be thought of as fluid - filled elastic tubes .
therefore , the discussion concerning the impulse response appears to be particularly relevant .
indeed , an impulsive boundary condition represents a first rough approximation for the muscular pulse sent by the heart through arteries .
clearly , this is not an exact model for such kind of impulses given that the aortic valve , for example , would not resist an `` infinite stress '' without breaking .
nonetheless , the time evolution of the system after the pulse is basically the same .
now , in fig .
[ fig-2 ] we show the impulsive response function as seen by a static `` observer '' placed in a given position @xmath129 of an artery .
this plot clearly tells us that , at @xmath129 , the signal will reach its ( finite ! ) peak at @xmath133 where @xmath134 is a time delay due to the viscosity of the fluid .
so , we have that the blood viscosity slows down the motion of the pulse and dissipates part of its energy , leading to a damping of its peak .
moreover , the further away we put our observer from the edge of the artery , the stronger is the dumping of the peak of the pulse .
it is quite known in the theory of laplace transform that , under fairly general conditions , the asymptotic behavior in the time domain is strictly related to the asymptotic behavior in the laplace domain . in particular , let @xmath135 be a suitable function in the time domain .
then , denoting with @xmath136 the laplace transform of @xmath135 , the asymptotic behavior of @xmath135 as @xmath26 is uniquely determined by the asymptotic expansion of @xmath136 as @xmath137 .
this is basically the content of the first tauberian theorem , in order to perform the wave - front expansion , we took profit the second tauberian theorem , the one that relates the asymptotic behavior of @xmath135 as @xmath26 to the asymptotic expansion of @xmath136 as @xmath137 . ] . for further informations about physical applications of this technique ,
see @xcite .
therefore , the long time asymptotic behavior of our response function @xmath138 is formally equivalent to the asymptotic behavior of @xmath139 as @xmath137 .
hence , recalling the formal solution of the equation of motion ( [ eq - motion - laplace ] ) , in the laplace domain , @xmath140 \ , , \ ] ] in order to infer the long time behavior of @xmath138 we just need to compute the expansion for @xmath31 , as @xmath137 , and then invert back to the time domain .
first , let us consider the case of a step input @xmath120 . for a body
whose viscoelastic description is modeled in terms of bessel models @xcite we have @xmath141 thus , @xmath142^{1/2 } \overset{s\to0}{= } \sqrt{4 ( \nu + 1)(\nu+2 ) } \ , s^{1/2 } \left [ 1 + \frac{1}{4 ( \nu + 1)(\nu+3 ) } \ , s \right ] + o ( s^{5/2 } ) \ , .\ ] ] then , we can compute the behavior of the solution of the wave equation for our viscoelastic model as @xmath137 , i.e. @xmath143 , indeed @xmath144 = \frac{1}{s } \ , \exp \left [ - x \ , \mu ( s ) \right ] \overset{s\to0}{\sim } \\ & \sim \frac{1}{s } \ , \exp \left [ - \sqrt{4 ( \nu + 1)(\nu+2 ) \ , x^{2 } } \ , s^{1/2 } \right ] \ , . \end{split}\ ] ] now , taking profit of the fact that ( see @xcite ) @xmath145 = \frac{1}{s } \ , \exp \left ( - \sqrt{a s } \right ) \
, , \ ] ] where @xmath146 \equiv { \widetilde{f } } ( s)$ ] , then from eq .
( [ eq - r ] ) we immediately get @xmath147 \quad \mbox{as } \,\,\ , t \to + \infty \ , .\ ] ] depicted versus @xmath128 , for @xmath148.,width=340 ] on the other hand , if we consider an impulsive input @xmath130 , @xmath131 , then eq .
( [ eq - r ] ) reads @xmath149 = \exp \left [ - x \ , \mu ( s ) \right ] \overset{s\to0}{\sim } \\ & \sim \exp \left [ - \sqrt{4 ( \nu + 1)(\nu+2 ) \ , x^{2 } } \ , s^{1/2 } \right ] \ , .
\end{split}\ ] ] then , recalling that ( see @xcite ) @xmath150 = \exp \left ( - k \ ,
\sqrt{s } \right ) \ , , \
] ] we can invert back to the time domain the expression in eq .
( [ eq - r-2 ] ) , getting @xmath151 \quad \mbox{as } \,\,\ , t \to + \infty \ , .\ ] ] depicted versus @xmath128 , for @xmath152.,width=340 ]
in this paper we discussed the propagation of transient waves in a semi - infinite linear viscoelastic medium described by a bessel model of order @xmath7 .
in particular , in section [ sec-2 ] we revisited the buchen - mainardi algorithm for the wave - front expansion for singular viscoelastic models .
then , in section [ sec-3 ] we applied the previous method in order to compute the wave - front expansion for a bessel body of order @xmath7 , given two typical input conditions .
first , we considered the `` canonical '' step input , in analogy with the seminal paper @xcite by buchen and mainardi .
this input is particularly relevant because the singular behavior of the memory function at @xmath153 induces a wave - front smoothing of any initial discontinuity . for further details on this smoothing effect found in transient waves propagating in media with singular memory ,
we refer to hanyga @xcite .
then , we analyzed an impulsive boundary condition in order to provide some general remarks concerning a peculiar hemodynamical model , first proposed by giusti and mainardi in @xcite .
the result is that this model appears to be consistent with the expected behavior of a pressure wave as it propagates within a large artery .
finally , in section [ sec-4 ] we provided an analytic asymptotic description of the long time behavior of the response function for both the analyzed scenarios .
the work of a. g. and f. m. has been carried out in the framework of the activities of the national group of mathematical physics ( gnfm , indam ) .
the authors would also like to thank v. a. diaz and c. de tommasi for valuable comments and discussions .
as we discussed in section [ sec-2 ] , in general we have that ( setting @xmath154 ) @xmath155 \ , .\ ] ] if we consider , in particular , the case in which @xmath120 , the previous expression can be rewritten as @xmath156 \ , .\ ] ] now , for a bessel solid we have @xmath157 thus , @xmath158 \ , s^{- ( \lambda _ k + 1 ) } \ , \exp \left [ - ( \nu + 1 ) \ , x \ , s^{1/2 } \right ] \ , .\ ] ] if we then consider a function @xmath136 , in the laplace domain , defined as @xmath159 and we denote with @xmath160 \ , , \ ] ] one can immediately infer that @xmath161 moreover , expanding the exponential in eq .
( [ eq-1a ] ) and then inverting term by term , one can easily prove that @xmath162 now , as suggested in @xcite , the easiest manner to compute the inverse of the wave - front expansion is thought recurrence relations .
therefore , for this purpose we can introduce the following function : @xmath163 \ , , \quad z ( \beta , \gamma , t ) = \frac{\beta}{t^{\gamma } } \ , , \ ] ] that satisfies the following recurrence relation @xmath164 i. colombaro , a. giusti , f. mainardi , a class of linear viscoelastic models based on bessel functions , _ meccanica _ , published online , 23 may 2016 . *
doi * : 10.1007/s11012 - 016 - 0456 - 5 .
[ e - print http://arxiv.org/abs/1602.04664[arxiv:1602.04664 ] ( 2016 ) ] a. giusti , f. mainardi , a dynamic viscoelastic analogy for fluid - filled elastic tubes , _ mecanica _ ( 2016 ) , 51 : 2321 . *
doi * : 10.1007/s11012 - 016 - 0376 - 4 [ e - print http://arxiv.org/abs/1505.06694[arxiv:1505.06695 ] ( 2015 ) ] .
a. giusti , f. mainardi , on infinite series concerning zeros of bessel functions of the first kind , _ eur .
j. plus _ ( 2016 ) , vol . 131 : 206 . * doi * : 10.1140/epjp / i2016 - 16206 - 4 [ e - print http://arxiv.org/abs/1601.00563[arxiv:1601.00563 ] ( 2016 ) ] .
|
in this paper we discuss the uniaxial propagation of transient waves within a semi - infinite viscoelastic bessel medium .
first , we provide the analytic expression for the response function of the material as we approach the wave - front .
to do so , we take profit of a revisited version of the so called buchen - mainardi algorithm .
secondly , we provide an analytic expression for the long time behavior of the response function of the material .
this result is obtained by means of the tauberian theorems for the laplace transform .
finally , we relate the obtained results to a peculiar model for fluid - filled elastic tubes .
|
in the standard model ( sm ) of particle physics , the higgs field @xcite provides the mechanism of spontaneous electroweak symmetry breaking and the origin of masses of the fundamental particles , but until recently the higgs boson itself left no signal in high - energy collider experiments .
the situation has clearly changed , however , as both the atlas @xcite and cms @xcite collaborations have announced observation of a bosonic particle at about the @xmath2 significance level .
the excess is driven by the two channels with the highest mass resolution @xmath3 and @xmath4 , and the equally sensitive but low resolution @xmath5 channel . assuming the boson spin is shown to be zero
, these results will provide conclusive evidence for the discovery of a higgs - like scalar particle with mass @xmath6 for atlas and @xmath7 for cms .
the discovery points to a favored mass range : @xmath8 , which is in agreement with the indirect detections from the electroweak precision constraints , @xmath9 @xcite .
it fixes the one remaining free parameter in the sm : the higgs self - coupling @xmath10 .. ] however , this low higgs mass immediately leads to the problem of the sm higgs vacuum stability which requires @xmath10 remain positive at all scales @xmath11 .
if @xmath10 becomes negative at some scale , the potential is either unbounded from below and has no state of minimum energy or has a vacuum with lower energy for the case where @xmath10 may run positive again at an even higher scale . given @xmath12 , the higgs self - coupling @xmath10 may run negative at a scale below the planck scale @xcite , necessitating new physics beyond the sm ( bsm ) .
it was shown in @xcite that absolute vacuum stability requires a higgs mass @xmath13 , by using a partial two - loop matching and three - loop renormalization group running procedure and taking into account the existing @xmath14 experimental uncertainties in the mass of the top quark and @xmath15 .
a very similar conclusion was given in ref .
additionally , ref .
@xcite studied the two - loop qcd and yukawa corrections to the relation between the higgs quartic coupling and the higgs mass so as to reduce the uncertainty in the determination of the higgs mass from @xmath16 .
the authors claimed that while @xmath10 at the planck scale is zero , the absolute stability of the sm higgs potential is excluded at @xmath17 c.l . for @xmath18 .
thus a @xmath19124 - 126 gev higgs strongly points to new physics in the desert between the fermi and planck scales .
supersymmetry and extra dimension models may provide solutions to this problem , but up to now no clear signals of such new physics have emerged from collider searches . therefore it is instructive to consider alternative simple extensions of the sm that can alleviate the vacuum stability problem . despite the absence of collider bsm signals , definitive evidence of new physics beyond the sm comes from several additional sources , including neutrino masses and dark matter .
the solar , atmospheric , reactor , and accelerator neutrino experiments have provided convincing evidence that neutrinos are massive and lepton flavors are mixed @xcite .
precise cosmological observations have confirmed the existence of non - baryonic cold dark matter with an abundance of @xmath20 @xcite .
these two important discoveries can not be accommodated in the sm without introducing extra ingredients .
doing so can affect the stability of the electroweak vacuum _ via _ one or more of the following interactions : ( 1 ) additional quartic scalar interactions associated with new scalar degrees of freedom ; ( 2 ) yukawa interactions associated with neutrino mass generation ; ( 3 ) modified or extended gauge interactions appearing in neutrino mass and/or dark matter models . in this work ,
we study examples of all three . in particular , to understand the origin of the neutrino masses , one may extend the sm with heavy majorana neutrinos so that light neutrino masses can be generated through the so - called seesaw mechanism .
there are three types of tree - level seesaw mechanisms , categorized according to the particle content of their extension to the sm : heavy majorana neutrinos ( type - i @xcite ) plus either a @xmath21 higgs triplet ( type - ii @xcite ) or a @xmath22 fermion triplet ( type - iii @xcite ) ( @xmath23 is the sm hypercharge quantum number ) . in the context of type - i models ,
the impact of heavy right - handed neutrinos , @xmath24 , on the sm higgs vacuum stability and metastability scales was studied in ref .
@xcite , with the result that the @xmath25 decrease the higgs vacuum stability scale . in this paper
we will study the effect of a tev scale type - ii seesaw model on the higgs vacuum stability .
our result shows that the the sm higgs vacuum can remain stable up to the planck scale for certain chosen parameters of the type - ii seesaw model .
we also study how to keep the sm higgs vacuum stable by making changes only to the gauge interactions . in the context of the sm gauge groups , stability up to the planck scale
can be restored by introducing a series of new electroweak multiplets , the neutral component of which may serve as the cold dark matter candidate . the presence of these multiplets may significantly alter the renormalization group ( rg ) evolution of the gauge coupling coefficients and thus indirectly keep the sm higgs vacuum stable up to the planck scale .
we also study the implications of the sm higgs vacuum stability for a new gauge symmetry . in ref .
@xcite it was shown that there are only flavor - dependent anomaly - free gauged @xmath1 symmetries in the minimal sm . in the sm with three right - handed neutrinos and a scalar singlet , there is another well - known anomaly - free gauged @xmath1 symmetry : @xmath26 @xcite .
however , the sm higgs boson carries no new @xmath1 charge in both cases .
the presence of these new @xmath1 symmetries only affects the running behavior of the yukawa couplings ( most notably that of the top quark ) . in this paper , we study the effect of a new type of @xmath1 gauge symmetry in which the sm higgs boson carries non - zero charge .
the anomalies are spontaneously cancelled by extending the sm with three right - handed neutrinos .
we derive the one - loop @xmath27-functions of the model and investigate their effects on the sm higgs vacuum stability .
we find that stability of the vacuum can be achieved for some regions of parameter space .
the paper is organized as follows : in section ii we show preliminary formulas relevant for the numerical analysis . we investigate the type - ii seesaw effect in section iii .
section iv is devoted to studying the implications of the sm higgs vacuum stability for the gauge interactions .
we summarize in section v. expressions for the relevant @xmath27-functions appear in the appendix .
we first review the stability analysis within the sm .
a constraint on the higgs mass can be obtained by the requirement that spontaneous symmetry breaking actually occurs , that is , @xmath28 be the minimum of the higgs potential @xmath29 where @xmath30 gev is the higgs vacuum expectation value ( vev ) .
this bound is essentially equivalent to the requirement that the quartic higgs coupling @xmath16 never becomes negative at any scale @xmath31 , where @xmath32 is the scale of new physics . in this paper
we will study numerically the impact of representative simple bsm scenarios on the vacuum stability of a @xmath19124 - 126 gev sm higgs . as a sm baseline
, we will use the two - loop beta functions of the higgs self coupling @xmath10 , the gauge couplings @xmath33 @xmath34 and the top quark yukawa coupling @xmath35 , as well as the one - loop matching condition for the sm higgs mass .
contributions of bsm physics are considered at the one - loop level .
the resulting stability requirements for a given bsm scenario are likely to be overly conservative , since the three - loop analyses tend to alleviate the tension of a 124 - 126 gev higgs with stability . however , since we do not presently have in hand the two - loop running for the bsm scenarios , it may not make sense to consider the sm at the three - loop and the bsm at the one - loop .
thus , for illustrative purpose we will use the one - loop matching and two - loop renormalization group running " procedure .
the @xmath27-function of @xmath36 is given to two loop order in eq .
( [ lambdabeta ] ) , in which @xmath37 represent the @xmath27-functions of @xmath10 at the one - loop and two - loop level , @xmath38 with @xmath39 being a reference energy scale , and @xmath35 is the yukawa coupling of the top quark . for illustrative purposes , we neglect the scale - dependence of the yukawa and gauge couplings , arriving at a simplified condition for vacuum stability at the scale @xmath40 : @xmath41 a meaningful and complete analysis should take into account the running behavior of all parameters .
one should also take implement the one - loop matching condition between the running higgs quartic coupling and the higgs boson pole mass @xmath42 @xcite : @xmath43 \ ; .
\end{aligned}\ ] ] where the expression of @xmath44 can be found in ref .
@xcite .
the two loop @xmath27-functions for the gauge couplings are given in eq .
( [ betag ] ) , in which @xmath45 represent the yukawa coupling matrices of up - quarks , down - quarks , and charged leptons , respectively . here
the @xmath46 gauge couplings @xmath47 are normalized based on @xmath48 ( though we do not impose any gut relations on the couplings ) , so the electroweak couplings @xmath49 and @xmath50 are related to these by @xmath51 and @xmath52 . the determination of the couplings proceeds from the relations @xmath53 , with @xmath54 at the z - pole @xcite .
the two - loop @xmath27-function of the top quark yukawa coupling is given in eq .
( [ betatop ] ) .
the initial input of @xmath35 is given by @xmath55 , where @xmath56 and @xmath57 is the top quark running mass determined from @xcite : @xmath58 in which the second and the third terms correspond to the one- and two - loop qcd corrections while the fourth term comes from the electroweak corrections at the one - loop level .
we use the running mass of the top quark value @xmath59 gev @xcite in our following numerical analysis . utilizing the foregoing rg analysis and the present range for @xmath42
, one finds that @xmath10 runs negative for @xmath60 gev .
a simple solution to the 125 gev sm higgs vacuum stability problem is obtained by introducing a new beyond - the - sm scalar that may interact with the sm higgs through a four scalar coupling vertex .
typical examples are the higgs portal dark matter models , e.g. the scalar singlet or darkon " @xcite and inert dark matter models @xcite . for
a detailed analyses of the implications of these models on the higgs vacuum stability , see refs .
@xcite .
the fact that neutrinos have tiny but non - zero masses is the first ( terrestrial ) experimental evidence of new physics beyond the sm .
the most convincing idea to understand the origin of neutrino masses is the seesaw mechanism .
the effect of a type - i seesaw model on the higgs vacuum stability was studied in ref .
@xcite , and the type - i seesaw model was found to aggravate the instability of the vacuum .
here we investigate the effect of a tev scale type - ii seesaw model , which extends the sm with a triplet scalar @xmath61 , @xmath62 transforming as ( 3,1 ) under the electroweak gauge group @xmath63 .
the additional scalar potential can be written as @xmath64 + { \lambda_1
\over 2 } \left ( { \rm tr } [ \delta^\dagger \delta ] \right)^2 + { \lambda_2 \over 2 } \left [ ( { \rm tr } [ \delta^\dagger \delta])^2 -{\rm tr } ( \delta^\dagger \delta \delta^\dagger \delta)\right ] \nonumber \\ & + & \lambda_4 h^\dagger h { \rm tr } [ \delta^\dagger \delta ] + \lambda_5 h^\dagger [ \delta^\dagger,~ \delta ] h + \left [ \sqrt{2}\lambda_6 h^t i \sigma_2 \delta^\dagger h + { \rm h.c . } \right ] \ ; .\end{aligned}\ ] ] the scalar triplet couples to the left - handed lepton doublet through the following yukawa interaction : @xmath65 where @xmath66 is the charge conjugation operator . the active neutrino mass can be derived from eq .
( [ neumass ] ) after the spontaneous breaking of electroweak symmetry : @xmath67 where @xmath68 is the vev of the scalar triplet and is constrained to be less than @xmath69 by the @xmath70 parameter . in this model
the one - loop @xmath27-function of @xmath10 can be written as @xcite @xmath71 and the one loop @xmath27-functions for the gauge couplings in this case are given by @xmath72 , with @xmath73 defined in eq .
( [ betag ] ) . notice that both @xmath74 and @xmath75 contribute to the @xmath27-function of @xmath10 . here
we mainly consider the effect of @xmath74 by working in the limit where the other triplet couplings ( @xmath76 , and @xmath75 ) are equal to zero at the input scale .
this is a reasonable simplification since @xmath77 and @xmath78 do not contribute to @xmath79 at one - loop level , and as eq .
( [ eq : beta_lambda_seesaw ] ) shows @xmath74 has a larger impact on the running of @xmath10 than @xmath75 ( assuming @xmath74 and @xmath75 are of the same order ) . as well . ]
nonetheless we do include the rg evolution of all the scalar couplings in our analysis and we ensure that all of the vacuum stability conditions in ref .
@xcite , which in our notation are given by @xmath80 are satisfied for all values of @xmath40 between @xmath42 and @xmath81 ( the parentheses in eqs .
( [ eq : seesaw_stability_4 ] ) and ( [ eq : seesaw_stability_5 ] ) indicate that there are actually two stability conditions in each equation : one with @xmath75 in the parentheses taken into account and the other one without )
. a vacuum stability analysis in which the other type - ii seesaw scalar couplings are allowed to have non - zero values at the input scale is much more complicated ( see @xcite ) and would distract from our purpose of studying the rg evolution of @xmath10 .
furthermore , we note that , as @xmath82 appears in neither the other scalar coupling @xmath27-functions nor the vacuum stability conditions in eqs .
( [ eq : seesaw_stability_1])-([eq : seesaw_stability_5 ] ) , we do not include it in our analysis .
the coupling @xmath82 effects the seesaw mechanism by giving the triplet a vev : @xmath83 and as mentioned above @xmath84 .
bounds on @xmath82 that arise from avoiding tachyonic directions in the potential at the ew minimum @xcite can always be satisfied while still obtaining the desired neutrino masses through appropriate choices of the triplet mass scale @xmath85 and the yukawa couplings @xmath86 . as the function of energy scale @xmath40 .
the solid and dashed correspond to @xmath87 and @xmath88 respectively .
the short - dashed and dotted lines correspond to @xmath89 and @xmath90 respectively .
right panel : @xmath42 as the function of cutoff scale @xmath40 .
the plus and cross signs are the perturbativity constraint with @xmath74 being @xmath91 and @xmath92 . the dashed and dotted lines are the vacuum stability constraint with @xmath93 and @xmath91 , respectively .
the horizontal band is the current experimental value of higgs boson mass.,title="fig:",width=302 ] as the function of energy scale @xmath40 .
the solid and dashed correspond to @xmath87 and @xmath88 respectively .
the short - dashed and dotted lines correspond to @xmath89 and @xmath90 respectively .
right panel : @xmath42 as the function of cutoff scale @xmath40 .
the plus and cross signs are the perturbativity constraint with @xmath74 being @xmath91 and @xmath92 .
the dashed and dotted lines are the vacuum stability constraint with @xmath93 and @xmath91 , respectively .
the horizontal band is the current experimental value of higgs boson mass.,title="fig:",width=302 ] to study the higgs vacuum stability , we first calculate the quartic coupling @xmath94 using the one - loop matching condition in eq .
( [ match ] ) , then run @xmath16 to the planck scale by solving the rg equations .
the @xmath27-function of @xmath74 is @xmath95 + 6 y_t^2 ) \lambda_4 \ ; . \label{phi}\end{aligned}\ ] ] the @xmath27-function for the top yukawa coupling is the same as that in the sm .
the @xmath27-functions for the @xmath96 @xmath97 can be found in the appendix .
we assume the scalar triplet is at the electroweak scale and therefore @xmath98 , as implied by the scale of the light neutrino masses .
consequently we can safely neglect the @xmath99 $ ] term in eq .
( [ phi ] ) and need not consider the matching condition at the seesaw threshold . in the left panel of fig .
[ triplet ] we plot @xmath10 as a function of energy scale @xmath40 .
the solid and dashed lines correspond to @xmath87 and @xmath88 , respectively .
the short - dashed and dotted lines correspond to @xmath89 and @xmath90 .
we find that the vacuum of a 124 - 126 gev higgs can be stable up to the planck scale for the case of @xmath100 , while the vacuum of the @xmath101 higgs will be unstable at the scale of @xmath102 for the case @xmath103 .
it is interesting and instructive to also study the perturbativity constraints in this model .
the perturbativity bound is defined as the highest higgs boson running mass given by the the higgs quartic coupling which satisfies the condition @xmath104
@xcite for any @xmath40 between the electroweak and planck scale , @xmath81 ( this criterion is less stringent than that used in @xcite which was based on the work of @xcite ) . with this perturbativity requirement , in conjunction with the vacuum stability conditions , we plot in the right panel of fig .
[ triplet ] the bounds on the higgs mass @xmath42 as a function of the energy scale @xmath40 .
the plus and cross signs are the perturbativity constraints with @xmath74 equal to @xmath91 and @xmath92 , respectively .
the dashed and dotted lines are the vacuum stability constraints with @xmath93 and @xmath91 , respectively .
the horizontal band is the current experimental value of higgs boson mass .
we can read from the figure that the 125 gev sm higgs mass satisfies both the perturbativity and the vacuum stability constraints up to the planck scale for appropriately chosen initial input value of @xmath74 .
note that the range of this coupling that is consistent with both vacuum stability and perturbativity is rather restricted .
we comment that our analysis is similar in spirit to that of ref .
@xcite , though with some differences .
in particular , we use two - loop @xmath27-functions for the gauge and yukawa couplings , analyze in more detail the perturbativity bounds , and a study of a range of input values for @xmath74 . on the other hand
, @xcite considers the behavior of the full set of scalar couplings , constraints for electroweak precision data , and implications for the @xmath105 rate .
in this section , we consider alternative solutions to the sm higgs vacuum stability problem by ( a ) modifying the @xmath27-functions of the gauge couplings through the introduction of new electroweak ( ew ) multiplets , which might provide a cold dark matter candidate , or ( b ) introducing a new @xmath1 gauge symmetry .
higher representation ew multiplets exist in various models .
a typical example is the gauge portal dark matter model in which dark matter annihilates into the sm particles through ew gauge interactions .
the gauge portal scenario is one genre of a more general set of dark matter models that also include higgs portal and axion portal models among others .
minimal dark matter @xcite is a typical gauge - portal dark matter model in which a high - dimension electroweak multiplet ( _ e.g. _ , a @xmath106 ) with hypercharge @xmath22 is introduced .
there are also models where an electroweak triplet @xcite or 7-plet @xcite can be dark matter candidates .
taking into account loop contributions from these new ew multiplets , the one - loop @xmath27-functions of @xmath107 and @xmath108 will be @xmath109 where @xmath110 for bosonic or fermionic dark matter respectively .
@xmath23 is the weak hypercharge of the dark matter and @xmath111 is the dimension of the @xmath112 multiplet representation .
we can conclude that the running behavior of the @xmath33 may be significantly changed for these cases and thus may have an effect on the rg evolution of the higgs quartic coupling .
to illustrate , we plot in the left - panel of fig .
[ multiplet ] the coupling @xmath10 as the function of energy scale @xmath40 , where the solid , dashed , dotted , short - dotted and dashed lines correspond to the addition of @xmath113 , @xmath114 , @xmath115 and @xmath116 ew multiplets , separately ( @xmath117 and @xmath118 represent bosonic i.e. , scalar and fermionic fields ) .
we set @xmath89 and @xmath119 for the initial input .
the mass of the ew multiplet is set to be several tev so as to be consistent with the constraint on the dark matter relic abundance @xcite .
we conclude from the figure that the @xmath120 sm higgs vacuum is stable up to the planck scale with the addition of a single 5-plet or 7-plet ; however , the higgs vacuum remains unstable with the addition of a single electroweak triplet or quadruplet .
at least two extra ew multiplets of the triplet or quadruplet variety are needed to keep the higgs vacuum stable up to the planck scale as can be seen from the right panel of fig .
[ multiplet ] . as the function of energy scale @xmath40 , where the solid , dashed , dotted and shot - dashed lines correspond to @xmath113 , @xmath114 , @xmath115 and @xmath116 ew multiplet cases respectively , where @xmath117 and @xmath118 represent bosonic or fermionic fields .
right panel : @xmath10 as the function of @xmath40 , the solid , dashed and dotted lines correspond to two quadruplet , two triplet and one triplet plus one quadruplet cases , respectively .
, title="fig:",width=302 ] as the function of energy scale @xmath40 , where the solid , dashed , dotted and shot - dashed lines correspond to @xmath113 , @xmath114 , @xmath115 and @xmath116 ew multiplet cases respectively , where @xmath117 and @xmath118 represent bosonic or fermionic fields .
right panel : @xmath10 as the function of @xmath40 , the solid , dashed and dotted lines correspond to two quadruplet , two triplet and one triplet plus one quadruplet cases , respectively . , title="fig:",width=302 ] it has been shown that there are only flavor dependent anomaly - free gauged @xmath1 symmetries , i.e. , @xmath121 @xcite , in the sm .
a @xmath26 gauge symmetry emerges in the sm plus three right - handed neutrinos .
may also be global , but we focus on the gauged case .
] however , the sm higgs boson carries no @xmath1 charge for both cases , so the only impact of the new @xmath1 symmetry relevant to stability is on the running behavior of the top quark yukawa coupling . in this paper
we investigate a new anomaly - free @xmath122 for which the sm higgs may carry charge .
such a new symmetry may originate from gut models @xcite or string inspired models @xcite . only right - handed fermions and the sm higgs boson carry a @xmath122 charge which we normalize to be multiples of ` @xmath123 ' .
the even number of fermion doublets required by the global @xmath112 anomaly @xcite is provided by the sm .
the absence of axial - vector anomalies @xcite in the presence of the @xmath122 and the gravitational - gauge anomaly @xcite requires that certain sums of the @xmath122 charges vanish .
the right - handed fermions are assigned charges of @xmath124 so these anomaly - free conditions are @xmath125 = 0 \ ; , \\ & & { u(1)^\prime}^2 u(1)_y : \hspace{1 cm } -a^2 \left [ 3 \times { 2 \over 3 } - 3 \times { 1 \over 3 } -1 \right]=0 \ ; , \\ & & u(1)^\prime : \hspace{2 cm } -\left[a + ( - a)\right ] - 3 [ a+ ( -a ) ] = 0 \ ; , \\ & & { u(1)^\prime}^3 : \hspace{2 cm } -[a^3 + ( -a)^3 ] - 3 [ a^3 + ( -a)^3 ] = 0 \ ; .\end{aligned}\ ] ] the @xmath122 charge of the sm higgs is fixed by the yukawa interactions .
we list in table i the quantum numbers of the fields under the @xmath122 . .
quantum numbers of fields under local @xmath1 . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] we have also included an additional scalar field @xmath126 which appears in table i. the @xmath122 can be spontaneously broken by the addition of this scalar @xmath126 with non - zero vev and transforming as a singlet under the sm gauge group . the number of generations of the scalar @xmath126 , @xmath127 , and its @xmath122 charge written as a multiple @xmath128 of @xmath123 are not fixed by the requirement of anomaly cancellation or any interactions with the sm fields .
the impact of the resulting singlet - higgs interactions on stability constitutes a special case of earlier work @xcite , so we do not consider it in detail here .
we only focus on the impact of the @xmath122 gauge interaction on the higgs vacuum stability .
note that the charge normalization @xmath123 can be absorbed by a redefinition of the new gauge coupling @xmath129 .
taking the coefficient of the @xmath130 operator to be negligible , the one - loop @xmath27-function of @xmath10 can be written as @xmath131 where the second and third terms are the contribution of the @xmath122 .
the @xmath27-function of @xmath132 can be written as @xmath133 where @xmath134 , @xmath135 and @xmath127 are the number of generations of fermions ( 3 ) , the sm higgs doublet ( 1 ) , and additional @xmath122-breaking singlet , respectively .
the new gauge interaction also affects the evolution of the yukawa coupling of the top quark .
its @xmath27-function is given by @xmath136 the number of generations of scalars @xmath127 , their @xmath122 charge @xmath128 , and the value of the coupling @xmath132 at the input scale are not totally arbitrary .
at one - loop order , it is straightforward to determine the scale at which @xmath132 has a landau pole ( we will leave a study of the two - loop effects of the new @xmath122 symmetry for future work ) . solving eq .
( [ eq : g4_beta ] ) for @xmath137 and equating the resulting denominator with zero , the landau pole scale is found to be @xmath138\ ] ] where @xmath139 is the value of the gauge coupling at the input scale @xmath39 . by increasing @xmath139 , @xmath127 , or @xmath128 ( the latter two increase @xmath140 ) ,
@xmath141 decreases . for certain choices it will be true that @xmath142 and as @xmath143 then surely @xmath10 becomes non - perturbative
nonetheless , our numerical analysis shows that the choice of @xmath139 has a more direct impact on the running of @xmath10 than varying @xmath127 and @xmath128 , so we fix @xmath144 to allow the greatest freedom in choosing the value of the @xmath122 gauge coupling at the input scale while avoiding the landau pole . as the function of cutoff scale @xmath40
the plus and cross signs are the perturbativity constraint with @xmath145 respectively .
the dashed and dotted lines are the vacuum stability constraint with @xmath146 respectively .
the horizontal band is the current measured value of higgs boson mass .
, width=415 ] we show in fig .
[ u1 ] the vacuum stability and perturbativity bounds on the sm higgs mass @xmath42 as a function of @xmath40 for a set of values of @xmath132 .
we set the scale of @xmath122-breaking at the @xmath147 scale and the mass of the corresponding @xmath148 boson to be @xmath149 in order to be roughly consistent with present lhc bounds@xcite .
the plus and cross signs represent perturbativity constraints ( by requiring @xmath150 ) with @xmath132 being @xmath151 and @xmath91 , respectively .
the dashed and dotted lines represent the higgs vacuum stability constraint with @xmath132 being @xmath91 and @xmath151 , respectively .
the horizontal band is the currently measured value of the higgs mass .
it is evident from the figure that the sm higgs vacuum stability predicts that @xmath132 should be roughly larger than @xmath91 in this specific model , though a smaller value could also be viable if the impact of the singlet scalar - higgs coupling is included .
alternatively , the presence of a sufficiently large gauge coupling would allow for stability in the absence of a significant singlet - higgs interaction . in either case
, such a parameter space might be accessible by the lhc , since we have assumed that @xmath152 .
a detailed analysis of the collider signatures of the model will be presented in future work .
if the observed @xmath153 gev boson is , indeed , the higgs boson , then stability of the sm electroweak vacuum up to the planck scale may require that some bsm degrees of freedom become active at scales @xmath154 1 tev .
the presence of these degrees of freedom can modify the rg evolution of the higgs quartic self - interaction that may otherwise run negative below the planck scale .
in general , three ingredients may change the running behavior of the higgs self - coupling : four - scalar interactions , yukawa interactions , and gauge interactions . in this paper , we have studied the sm higgs vacuum stability problem in representative minimal extensions of the sm that also address other phenomenological problems , accounting for the neutrino masses , dark matter from ew multiplets , and a new gauge symmetry .
we find that vacuum stability of a @xmath19124 - 126 gev higgs up to the planck scale could point to the existence of new scalars , which might be a tev scale higgs triplet in the type - ii seesaw model , whose coupling @xmath74 with the sm higgs should be in the rather restricted range @xmath155 .
alternatively , stability could be achieved through existence of a series of tev - scale ew multiplets , which can be gauge - portal dark matter candidates , or to a new @xmath122 gauge symmetry , in which the sm higgs carries non - zero charge . in the former case
, the number of new particles should be relatively large , _
i.e. _ , one generation quintuplet fermion ( or higher representational ew multiplet ) or at least two quadruplet scalar and ( or ) triplet fermion multiplets .
if the new @xmath1 symmetry is broken at the @xmath156 scale , then the sm higgs vacuum stability implies that @xmath132 can not be arbitrarily small unless the couplings of the associated sm singlets to the higgs doublet , the number of singlets , or the singlet @xmath122 charge is sufficiently large .
this work was supported in part by doe contracts de - fg02 - 08er41531 ( mjrm and wchao ) and de - sc0007983 ( mg ) , the wisconsin alumni research foundation ( mjrm and wchao ) , and by the wayne state university division of research ( mg ) .
the two loop @xmath27-function of the higgs quartic coupling @xmath10 is given by @xcite @xmath157 with @xmath158 one loop @xmath27-functions of @xmath77 , @xmath78 and @xmath75 @xcite @xmath163 \lambda_1 -8 { \rm tr } [ ( y_\delta^\dagger y_\delta^{})^2 ] \ ; , \\
16\pi^2 \beta_{\lambda_2 } = & & - \left ( { 36\over 5 } g_1 ^ 2 + 24 g_2 ^ 2 \right ) \lambda_1 + 12 g_2 ^ 4 - { 144 \over 5 } g_1 ^ 2 g_2 ^ 2 + 3 \lambda_2 ^ 2 + 12 \lambda_1 \lambda_2 \nonumber \\&&-8\lambda_5 ^ 2 + 4 { \rm tr}[y_\delta^\dagger y_\delta ] \lambda_1 + 8 { \rm tr } [ ( y_\delta^\dagger y_\delta^{})^2 ] \ ; , \\ 16\pi^2 \beta_{\lambda_5 } = & & -\left({9\over 2 } g_1 ^ 2 + { 33\over 2 } g_2 ^ 2 \right)\lambda_5 - { 18 \over 5 } g_1 ^ 2 g_2 ^ 2 + ( 2\lambda_1 -2\lambda_2 + 2 \lambda+8\lambda_4 \nonumber \\ & & + 6y_t^2 + 2 { \rm tr } [ y_\delta^\dagger y_\delta^ { } ] ) \lambda_5^ { } \ ; .\end{aligned}\ ] ] f. englert and r. brout , phys
* 13 * , 321 ( 1964 ) ; p. w. higgs , phys .
* 12 * , 132 ( 1964 ) ; phys .
* 13 * , 508(1964 ) ; phys .
rev . * 145 * , 1156 ( 1966 ) ; g. s. guralnik , phys .
* 13 * , 585(1964 ) ; t. w. b. kibble , phys . rev . * 155 * , 1554 ( 1967 ) .
j. elias - miro , j. r. espinosa , g. f. giudice , g. isidori , a. riotto and a. strumia , phys .
b * 709 * , 222 ( 2012 ) [ arxiv:1112.3022 [ hep - ph ] ] .
z. -z .
xing , h. zhang and s. zhou , phys .
d * 86 * , 013013 ( 2012 ) [ arxiv:1112.3112 [ hep - ph ] ] .
p. minkowski , phys .
b * 67 * , 421 ( 1977 ) ; t. yanagida , in _ workshop on unified theories _
, kek report 79 - 18 p.95 ( 1979 ) ; m. gell - mann , p. ramond , r. slansky , in _ supergravity _ ( north holland , amsterdam , 1979 ) eds .
p. van nieuwenhuizen , d. freedman , p.315 ; s. l. glashow , in _ 1979 cargese summer institute on quarks and leptons _
( plenum press , new york , 1980 ) eds .
m. levy , j .- l .
basdevant , d. speiser , j. weyers , r. gastmans and m. jacobs , p.687 ; r. barbieri , d. v. nanopoulos , g. morchio and f. strocchi , phys .
b * 90 * , 91 ( 1980 ) ; r. n. mohapatra and g. senjanovic , phys .
lett . * 44 * , 912 ( 1980 ) ; g. lazarides , q. shafi and c. wetterich , nucl .
b * 181 * , 287 ( 1981 ) .
w. konetschny and w. kummer , phys . lett .
b * 70 * , 433 ( 1977 ) ; t. p. cheng and l. f. li , phys .
d * 22 * , 2860 ( 1980 ) ; g. lazarides , q. shafi and c. wetterich , nucl .
b * 181 * , 287 ( 1981 ) ; j. schechter and j. w. f. valle , phys .
d * 22 * , 2227 ( 1980 ) ; r. n. mohapatra and g. senjanovic , phys .
d * 23 * , 165 ( 1981 ) .
machacek and m. t. vaughn , nucl .
b * 222 * , 83 ( 1983 ) ; nucl .
b * 236 * , 221 ( 1984 ) ; nucl .
b * 249 * , 70 ( 1985 ) ; c. ford , i. jack and d. jones , nucl .
b * 387 * , 373 ( 1992 ) ; h. arason _
et al_. , phys .
d * 46 * , 3945 ( 1992 ) ; v. barger , m. s. berger and p. ohmann , phys .
d * 47 * , 1093 ( 1993 ) ; m. x. luo and y. xiao , phys
lett * 90 * , 011601 ( 2003 ) .
v. barger , p. langacker , m. mccaskey , m. j. ramsey - musolf and g. shaughnessy , phys .
d * 77 * , 035005 ( 2008 ) [ arxiv:0706.4311 [ hep - ph ] ] .
v. barger , p. langacker , m. mccaskey , m. ramsey - musolf and g. shaughnessy , phys .
d * 79 * , 015018 ( 2009 ) [ arxiv:0811.0393 [ hep - ph ] ] .
y. cai , x. -g .
he and b. ren , phys .
d * 83 * , 083524 ( 2011 ) [ arxiv:1102.1522 [ hep - ph ] ] .
e. j. chun , h. m. lee and p. sharma , arxiv:1209.1303 [ hep - ph ] . c. cheung , m. papucci and k. m. zurek , jhep * 1207 * , 105 ( 2012 ) [ arxiv:1203.5106 [ hep - ph ] ] .
o. lebedev , eur .
j. c * 72 * , 2058 ( 2012 ) [ arxiv:1203.0156 [ hep - ph ] ] .
j. elias - miro , j. r. espinosa , g. f. giudice , h. m. lee and a. strumia , jhep * 1206 * , 031 ( 2012 ) [ arxiv:1203.0237 [ hep - ph ] ] .
k. kumericki , i. picek and b. radovcic , jhep * 1207 * , 039 ( 2012 ) [ arxiv:1204.6597 [ hep - ph ] ] .
p. fileviez perez , h. h. patel , m. .j .
ramsey - musolf and k. wang , phys .
d * 79 * , 055024 ( 2009 ) [ arxiv:0811.3957 [ hep - ph ] ] .
l. a. anchordoqui , i. antoniadis , h. goldberg , x. huang , d. lust , t. r. taylor and b. vlcek , arxiv:1208.2821 [ hep - ph ] ; l. a. anchordoqui , i. antoniadis , h. goldberg , x. huang , d. lust , t. r. taylor and b. vlcek , arxiv:1206.2537 [ hep - ph ] .
|
recent results from atlas and cms point to a narrow range for the higgs mass : @xmath0~{\rm gev}$ ] . given this range , a case may be made for new physics beyond the standard model ( sm ) because of the resultant vacuum stability problem , i.e. , the sm higgs quartic coupling may run to negative values at a scale below the planck scale .
we study representative minimal extensions of the sm that can keep the sm higgs vacuum stable to the planck scale by introducing new scalar or fermion interactions at the tev scale while solving other phenomenological problems .
in particular , we consider the type - ii seesaw model , which is introduced to explain the non - zero majorana masses of the active neutrinos .
similarly , we observe that if the stability of the sm higgs vacuum is ensured by the running of the gauge sector couplings , then one may require a series of new electroweak multiplets , the neutral component of which can be cold dark matter candidate . stability may also point to a new @xmath1 gauge symmetry , in which the sm higgs carries non - zero charge .
|
the problem of determining essential self - adjointness of smooth coefficient first - order differential operators , as well as their powers , is an important and well studied topic .
this paper considers similar problems but in the context of non - smooth coefficients .
more precisely , we consider first - order symmetric differential operators @xmath0 , as well as their powers , on smooth vector bundles @xmath1 , over smooth , noncompact manifolds @xmath2 .
we allow for the coefficients of the operator as well as the metrics on the bundle and the manifold to be non - smooth .
our primary focus is to understand the relationship between the regularity of the coefficients of @xmath0 and the essential self - adjointness of powers of @xmath0 .
there is a plethora of historical literature surrounding this subject , and therefore , we confine ourselves to presenting only the relevant references to our work .
from our point of view , it was gaffney who in @xcite made a first significant contribution by establishing the essential self - adjointness of the hodge - laplacian @xmath3 under a so - called _ negligible boundary _ condition
. moreover , gaffney allows his manifold to be @xmath4 and incomplete .
the next relevant reference to us is the work of wolf in @xcite , where he studies the essential self - adjointness of general dirac operators and their squares .
moreover , cordes in @xcite obtains the essential self - adjointness of all powers of the laplace - beltrami operator @xmath5 on functions .
later , in @xcite , chernoff studies conditions under which essential self - adjointness of _ all _ powers of first - order operators are obtained .
these last three references assume both completeness and smoothness of their metrics and the coefficients of the operators .
the study we present here is in part motivated by an attempt to recover results similar to that of @xcite and @xcite but without assuming smoothness of the operator and the metric . a priori , our differential operator @xmath6 , and such operators can arise , for instance , as operators built from levi - civita connections associated to lipschitz metrics . under a set of mild assumptions [ d : first]-[d : last ] , we proceed by demonstrating that the essential self - adjointness of this operator is equivalent to the _ negligible boundary _ property formulated with respect to the operator @xmath0 ( see definition [ def : negbdy ] ) .
this property is trivially satisfied on compact manifolds and hence , our analysis is exclusively carried out in the non - compact setting .
if the differential operator has @xmath7-coefficients , then for @xmath8 , we demonstrate that the essential self - adjointness of @xmath9 is equivalent to this aforementioned property for the operator @xmath9 .
indeed , we do not expect essential self - adjointness , in general , to survive for orders @xmath10 for a @xmath7-coefficient operator .
this is partly due to the fact that , even for an arbitrary smooth compactly supported @xmath11 , @xmath12 exists only distributionally . in [ s : essneg ] , we demonstrate this equivalence under the mild set of assumptions [ d : first]-[d : last ] , primarily motivated by features we would expect from elliptic operators .
we emphasise that in demonstrating this equivalence , we do not make any assumptions about the completeness of the underlying space .
we proceed abstractly in order to emphasise this point , and in fact , carry out this analysis on spaces with so - called _ rough metrics_. these are riemannian - like metrics that have merely measurable coefficients , which are comparable against induced euclidean metrics in small enough charts . a priori , such
a metric only induces a well defined measure , and it is unknown whether there is a naturally associated notion of length .
the significance of the negligible boundary criterion is that it can still be formulated on this measure space without alluding to a length structure . in [
s : htol ] , at this same level of generality , we are also able to show that if @xmath9 is essentially self - adjoint , then so are its lower powers .
this is done via functional calculus and operator theoretic methods to emphasise the fact that this assertion makes no geometric demands .
geometry begins to play a significant role when attempting to boost essential self - adjointness from lower to higher powers .
we carry out this study in [ s : ess ] , and the way in which we do this is to establish the negligible boundary condition for higher powers via a bootstrapping procedure . here , we are forced to assume completeness , but we are able to allow for a restricted class of rough metrics that induce a length space .
this is still a large and significant class of non - trivial metrics which , for instance , include bi - lipschitz pullbacks of smooth , complete metrics .
we emphasise that it is the lack of regularity that forces us away from the wave technique in @xcite or the pde technique in @xcite , as these techniques seem to require smoothness in a crucial way . by generalising the ideas of @xcite
, we demonstrate a certain elliptic estimate for lower powers of our operator from knowing that the higher powers have finite @xmath13-energy .
this in turn allows us to obtain negligible boundary for higher powers via the first power .
in [ s : app ] , we consider applications to dirac and other elliptic differential operators .
as a first consequence , we highlight the following result for spin manifolds with non - smooth metrics .
[ thm : main2 ] let @xmath14 be a @xmath15 complete metric on a smooth spin manifold @xmath2 , with a spin structure @xmath16 .
let @xmath17 , where @xmath18 is the minimal complex irreducible representation . then , the associated _ atiyah - singer dirac operator _
@xmath19 is essentially self - adjoint on @xmath20 . if the metric is @xmath7 for @xmath21 , then @xmath22 is essentially self - adjoint on @xmath20 for @xmath23 .
the first part of this theorem , the essential self - adjointness of @xmath19 for lipschitz metrics , was obtained by bandara , mcintosh and rosn in @xcite by other methods , and was a guiding motivation for our work here . moreover , theorem [ thm : main2 ] is a particular instance of more general results we obtain for dirac operators on general dirac bundles . as a second highlight theorem , we present the following consequence of our work to general symmetric elliptic operators .
[ thm : main1 ] let @xmath1 be a smooth bundle over a smooth manifold @xmath2 with continuous metrics @xmath24 and @xmath14 .
suppose that @xmath0 is a first - order elliptic differential operator with @xmath7-coefficients , @xmath25 , that is symmetric on @xmath26 in @xmath27 .
then , @xmath28 on @xmath20 for @xmath29 is essentially self - adjoint .
our theorems generalise the consequences of the work of wolf in @xcite and chernoff in @xcite to dirac operators to settings where smoothness assumptions on the metrics are relaxed . in theorem
[ thm : main1 ] , we recover the results of chernoff from @xcite in the elliptic setting , and we also dispense with the _ local velocity of propagation _ condition assumed in that paper .
hemanth saratchandran wishes to acknowledge support from the beijing international centre for mathematical research , and the jin guang mathematical foundation .
lashi bandara was supported by the knut and alice wallenberg foundation , kaw 2013.0322 postdoctoral program in mathematics for researchers from outside sweden .
throughout this article , we will use the analysts inequality @xmath30 to mean that @xmath31 with @xmath32 , as well as the analysts equivalence @xmath33 . the support of a section ( or function ) @xmath34 will be denoted by @xmath35 .
unless otherwise stated , we will assume einstein summation convention throughout this paper .
that is , whenever a raised index appears multiplicatively against a lowered index , we assume summation over that index . by _ precompact set _ , we mean a set whose closure is compact .
let @xmath2 be a connected , non - compact , smooth manifold and @xmath36 be a smooth vector bundle of finite dimension @xmath37 . by @xmath38 , we denote the fibre over the point @xmath39 , given by @xmath40 .
the locally euclidean structure allows us to define spaces of regularity in a local sense .
the spaces @xmath41 will be used to denote @xmath42 times differentiable sections for which the @xmath42-th derivative is @xmath43-hlder continuous _
locally_. the notation @xmath44 will be such sections that are compactly supported .
moreover , this locally euclidean structure allows us to import measure - theoretic notions in the absence of a metric : we say that a set @xmath45 is measurable if @xmath46 is lebesgue measurable for all coordinate charts @xmath47 with @xmath48 .
this allows us to define measurable functions @xmath49 , and by @xmath50 , we denote measurable sections over @xmath1 to be sections @xmath51 in continuous local frames @xmath52 with @xmath53 a measurable function .
see @xcite for a detailed construction . using our notion of measurability , we define @xmath54 to be measurable sections @xmath55 such that @xmath56 over local trivialisations @xmath57 .
the definition of local sobolev spaces follow similarly : we say that @xmath58 if @xmath59 . since our goal is to study _ global _ differential operators over @xmath1 , on letting @xmath60 be the dual bundle of @xmath1 , we define the following notion of a metric tensor .
let @xmath61 be real and symmetric , and suppose that for each @xmath39 , there exists a trivialisation @xmath62 containing @xmath63 and a constant @xmath64 satisfying : @xmath65 for almost - every @xmath66 and @xmath67 , where @xmath68 denotes the euclidean metric in @xmath69 .
then , we say that such a metric is a _ bundle rough metric _ , and such a trivialisation is said to satisfy the _
local comparability condition_. in the situation where @xmath70 , we say that @xmath14 is a _ rough metric _ on @xmath2 if in addition , it is real - valued and the trivialisations are induced by coordinate charts .
we distinguish the latter notion of a rough metric from a bundle rough metric since the chart induced trivialisations are necessary to setup a measure .
more precisely , given a rough metric @xmath14 on @xmath2 , we obtain a global measure @xmath71 by pasting together @xmath72 via a smooth partition of unity subordinate to a covering of @xmath2 by locally comparable charts .
this is readily checked to define a borel - measure that is finite on compact sets .
we emphasise that a priori , we do not know whether @xmath14 induces a length structure .
moreover , a set @xmath45 is measurable if and only if it is @xmath71-measurable .
the notion of rough metrics arose from the study of the geometric invariances of the kato square root problem .
there is a large non - trivial class of such metrics that arise as bi - lipschitz pullbacks of smooth metrics .
a detailed exposition of the properties of rough metrics can be found in @xcite . throughout the remainder of this paper ,
unless otherwise specified , we fix a rough metric @xmath14 on @xmath2 and a bundle rough metric @xmath24 on @xmath1 .
if @xmath1 and @xmath24 are real , we consider @xmath24 as a complex - valued inner product by complexifying @xmath1 .
for an open set @xmath73 , define @xmath74 spaces as the set of measurable distributions @xmath75 such that @xmath76 when @xmath77 , we simply write @xmath78 and @xmath79 . in the special case of @xmath80 ,
the space @xmath81 is a hilbert space with the inner product @xmath82 the @xmath27 inner product will be denoted by @xmath83 and the induced norm by @xmath84
. it can be readily verified that @xmath54 is the space of measurable @xmath75 such that @xmath85 for all pre - compact @xmath86 .
furthermore , the local sobolev spaces @xmath87 can be characterised by @xmath88 such that @xmath89 for every smooth connection @xmath90 over pre - compact @xmath91 .
note that any two smooth connections @xmath92 and @xmath93 over a pre - compact @xmath94 are comparable in the sense that @xmath95 for @xmath96 . in what is to follow
, we will require some notions from operator theory .
fixing a hilbert space @xmath97 , we consider operators @xmath98 , where @xmath99 is a subspace of @xmath97 called the _
domain _ of the operator .
the _ range _ of @xmath100 is denoted by @xmath101 and its _ null space _ , or _
kernel _ , by @xmath102 .
an operator is _ densely - defined _ if @xmath99 is dense in @xmath97 , and it is _ bounded _ if there exists a @xmath32 such that @xmath103 .
we say that an operator is _ closed _ if @xmath104 and @xmath105 implies that @xmath106 and @xmath107 .
this is equivalent to saying that the graph @xmath108 is a closed subset of @xmath109 .
a subspace @xmath110 is called a _ core _ for @xmath100 if it is dense in @xmath99 with respect to the graph norm @xmath111 .
an operator is _ closable _ if @xmath112 and @xmath105 implies that @xmath113 . in that case ,
@xmath114 where @xmath115 is a closed operator .
we write the closure of the operator as @xmath116 .
a densely - defined operator @xmath100 admits a closed operator @xmath117 called _ the adjoint _ of @xmath100 with domain @xmath118 the operator @xmath117 is defined as follows : for @xmath119 , there exists @xmath120 such that @xmath121 by the riesz - representation theorem , and @xmath122 . an operator @xmath123 is said to be _ an adjoint _ of @xmath100 if @xmath124 for all @xmath106 and @xmath125 .
in particular , whenever @xmath100 and @xmath123 are densely - defined , they are both closable and admit densely - defined adjoints .
an operator @xmath100 is symmetric if @xmath100 is _ an adjoint _ to itself .
however , typically @xmath126 , by which we mean that @xmath127 and @xmath128 on @xmath99 .
an operator @xmath100 is _ self - adjoint _ if @xmath128 .
a symmetric operator is _ essentially self - adjoint _ if it admits a unique self - adjoint extension @xmath129 .
in this situation , it is readily checked that @xmath130 . throughout this paper , whenever we say _ first - order differential operator _ , we will assume it is an operator that takes the form @xmath131 locally , with coefficients @xmath132 , and @xmath133 for some @xmath134 .
this is a linear map @xmath135 , that is local and for which @xmath136}}$ ] is an almost - everywhere nonzero , fibrewise , bounded multiplication operator for @xmath137 .
note that for almost - every @xmath39 and each fibre norm @xmath138 on @xmath38 , there exist @xmath139 such that @xmath140 if for @xmath25 , @xmath141 , and @xmath142 for @xmath143 , then we say that @xmath0 has @xmath7 coefficients . throughout ,
let @xmath144 for a @xmath7 coefficient operator ( or @xmath145 if the operator has @xmath146 coefficients ) and define : @xmath147 denote the @xmath148-graph norm of the operator by : @xmath149 for a function @xmath150 , define the commutators @xmath151}}u,$ ] for @xmath152 .
the commutator @xmath153 will be denoted by @xmath154 .
we present the following two axioms under which we prove the most general results of this paper . in [ s : app ] , we illustrate that these axioms are valid for a wide class of elliptic operators . 1 .
[ d : first ] [ d : close ] whenever @xmath155 and @xmath156 , @xmath157 2 .
[ d : reg ] [ d : last ] whenever @xmath158 and @xmath159 , then @xmath160 .
the condition [ d : reg ] is an @xmath13-ellipticity condition on the operator @xmath0 .
it is automatically satisfied for @xmath7 coefficient elliptic operators .
we formulate this as an axiom since we only require this weaker formulation , and as we shall see in
[ s : app ] , it can be proved in instances where the coefficients are merely @xmath146 .
for integers @xmath161 , let @xmath162 with domain @xmath163 and @xmath164 with domain @xmath165 for the case @xmath166 , [ d : close ] implies that @xmath167 and @xmath168 are closable , and on letting @xmath169 and @xmath170 , we obtain that @xmath171 for @xmath172 and @xmath173 .
moreover , using [ d : reg ] and a mollification argument , we obtain that @xmath174 .
fix @xmath156 , let @xmath175 be a pre - compact set satisfying @xmath176 , and @xmath177 be smooth partition of unity on @xmath175 .
for @xmath178 , write @xmath179 on @xmath175 . set @xmath180 , extended to zero outside of @xmath181 . since @xmath0 is a @xmath7 coefficient operator , we have that @xmath182 for @xmath183 .
then , @xmath184 that is , @xmath185 observe that if we simply asked that @xmath0 be symmetric on @xmath20 , then a similar calculation would yield [ d : close ] as a consequence . as a consequence of this symmetry condition for @xmath28
, we obtain that the operators @xmath186 and @xmath187 are closable and as for the @xmath188 case , we write @xmath189 and @xmath190 respectively to denote the closures of these operators with domains @xmath191 and @xmath192 . by viewing @xmath28 distributionally ,
i.e. , for @xmath158 defining @xmath193 for @xmath156 , define the maximal domain as @xmath194 in addition to the conditions [ d : first]-[d : last ] , a fundamental criteria we use and exploit throughout this paper is the following . [ def : negbdy ] we say that the operator @xmath9 exhibits _ negligible boundary _ if @xmath195 for all @xmath196 .
the first theorem we present is the following .
it is proved in
[ s : essneg2 ] .
[ thm : firstmain ] let @xmath9 satisfy [ d : first]-[d : last ] on a bundle @xmath1 with a bundle rough metric @xmath24 over @xmath2 with a rough metric @xmath14 .
then , the following are equivalent : a. @xmath197 , b. @xmath198 is self - adjoint , c. @xmath199 is self - adjoint , d. @xmath9 satisfies , e. @xmath200 is essentially self - adjoint .
the next theorem allows us to deduce essential self - adjointness of lower powers of the operator when we know this for higher powers .
it is proved in
[ s : core ] .
[ thm : firstmain2 ] let @xmath0 satisfy [ d : first]-[d : last ] on a bundle @xmath1 with a bundle rough metric @xmath24 over @xmath2 with a rough metric @xmath14 . if @xmath9 on @xmath20 is essentially self - adjoint , then @xmath20 is a core for @xmath201 for @xmath202 $ ] and moreover , it is a core for @xmath203 for @xmath204 .
this result requires very little geometric properties : the theorem is phrased for rough metrics which may not , a priori , even induce a length structure . in particular , it does not require completeness .
moreover , note that we only require the first power of @xmath0 to satisfy [ d : first]-[d : last ] in this theorem .
the second theorem we present is in the case that the rough metric induces a length structure . by this , we mean that @xmath205 where @xmath206 , is a distance metric which yields a topology that agrees with the topology on @xmath2 .
the quintessential nontrivial example of such a situation is the pullback metric @xmath207 , where @xmath208 is a smooth metric and @xmath209 is a lipeomorphism
. the metric @xmath14 , in general , only has measurable coefficients , and @xmath209 can be seen as a bi - lipschitz transformation of @xmath208 .
[ thm : secondmain ] let @xmath1 be a vector bundle with a bundle rough metric @xmath24 , over a manifold @xmath2 with a rough metric @xmath14 that induces a complete length space . if @xmath0 is a first - order operator satisfying [ d : first]-[d : last ] , then @xmath0 on @xmath20 is essentially self - adjoint . if @xmath0 has @xmath7 coefficients and @xmath28 satisfies [ d : first]-[d : last ] for @xmath161 , then @xmath28 on @xmath20 is essentially self - adjoint for @xmath161 .
the proof of this theorem can be found in [ s : ess - k ] . in @xcite , wolf obtains essential self - adjointness for @xmath0 and @xmath210 when @xmath0 is the dirac operator for _ smooth _ complete metrics .
following from this , chernoff in @xcite uses wave techniques to obtain such a result for all powers @xmath28 of first - order differential operators satisfying a velocity of propagation condition .
the wave techniques crucially rely on smoothness . in our theorem , we only assume completeness and allow for bad singularities at the level of the metric . we keep track of the way in which the coefficients of our differential operator affects the density of @xmath20 for higher powers .
fix a @xmath15 metric @xmath14 on @xmath2 . by the usual formula for the christoffel symbols , we define the levi - civita connection @xmath211 .
let @xmath212 denote the clifford algebra , which is the exterior algebra @xmath213 equipped with the clifford product @xmath214 given by @xmath215 by the symbol @xmath90 , we denote the canonical lift of the levi - civita connection to the bundle @xmath212 .
following the notation of @xcite , we say that a complex bundle @xmath216 is a spin bundle over @xmath217 if : a. it is a bundle of left - modules over @xmath212 with @xmath218 , b. it is equipped with a hermitian @xmath15-metric and a compatible connection @xmath219 satisfying : @xmath220 for all @xmath221 in @xmath222 , @xmath223 and for almost - every @xmath39 .
the first property says that @xmath224 is unitary and the second property says that @xmath90 is a module derivation . fixing a frame @xmath225 for @xmath226 , recall that the dirac operator @xmath0 is defined by @xmath227 it is easy to check that this is a first - order differential operator , @xmath228 as a consequence of the module derivation property in .
we now prove that @xmath0 satisfies [ d : close ] . but first , we present the following lemma that will be the primary tool used in its proof . [
lem : divlip ] let @xmath229 , a compactly supported lipschitz co - vectorfield .
then , @xmath230 since the metric @xmath14 is at least continuous , we note as in proposition 13 of @xcite that , given any @xmath231 , there is a smooth metric @xmath208 which is @xmath232-close to @xmath14 , by which we mean that : @xmath233 for every @xmath39 and @xmath234 .
note that there exists an @xmath235 symmetric such that @xmath236 , and on letting @xmath237 , we have that @xmath238 .
also , @xmath239 is bounded with respect to both metrics @xmath14 and @xmath208 .
computing in @xmath13 allows us to assert that @xmath240 ( c.f .
proposition 12 in @xcite ) .
therefore , @xmath241 but since the metric @xmath208 is smooth , @xmath242 , @xmath229 , and therefore @xmath243 .
thus , by lemma 7.113 in @xcite , we obtain that @xmath244 . with this , we prove the following . whenever @xmath245 , we have that @xmath246 .
moreover , @xmath0 satisfies [ d : close ] .
fix a point @xmath39 where @xmath14 and @xmath24 are differentiable and fix a smooth frame @xmath247 near @xmath63 .
let @xmath248 and note that by the lifting of @xmath90 to @xmath249 , we have that @xmath250 using metric compatibility , we obtain @xmath251 now , note that for @xmath252 , @xmath253 define the vectorfield @xmath254 , extended to the whole of @xmath2 by zero outside of @xmath255 .
it is easy to see that this is invariantly defined and that it is @xmath15 .
thus , from combining these calculations , we see that @xmath256 at points of differentiability .
thus , by invoking lemma [ lem : divlip ] , we obtain that @xmath257 and therefore , @xmath0 is symmetric on @xmath258 .
the proof of [ d : close ] follows from this as noted in
[ s : mainres ] . observe that unlike the smooth case , we can not assume that we can solve for a frame in which @xmath259 , since the metrics @xmath14 and @xmath24 are merely @xmath15 .
next , fix @xmath260 and @xmath261 .
let @xmath262 be a compactly supported smooth function such that @xmath263 . since @xmath0 is a first - order differential operator ,
the symbol of @xmath0 at @xmath264 is : @xmath265 the critical regularity case in the following theorem is modelled on the proof of proposition 3.4 in @xcite . [
prop : diraca2 ] @xmath0 satisfies [ d : reg ] . if the metrics @xmath14 and @xmath24 are @xmath7 for @xmath25 and @xmath0 is a @xmath7 coefficient operator , [ d : reg ] holds for @xmath9 for @xmath266 .
first , observe that any two _ smooth _
connections @xmath92 and @xmath93 are comparable on precompact subsets in @xmath13-norm . hence , to show [ d : reg ] , it suffices to show that @xmath267 implies that @xmath268 inside precompact trivialisations @xmath47 corresponding to charts .
also note that @xmath269 means exactly that @xmath270 for all @xmath271 .
suppose that @xmath272 with @xmath273 compact .
fix @xmath271 with @xmath274 .
since we are inside a precompact chart and @xmath14 is @xmath15 , we obtain a constant @xmath275 such that @xmath276 for all @xmath277 , where by @xmath68 we denote the euclidean metric
. moreover , there exists a transformation @xmath278 such that @xmath279 and so that @xmath280 , where @xmath281 and @xmath282 is the pullback of the lebesgue measure in @xmath283 by @xmath209 .
then , we obtain by the fact that @xmath267 that @xmath284 . expanding this norm , taking an orthonormal frame @xmath225 for @xmath226 (
note that this frame is lipschitz ) and @xmath285 for @xmath216 , we have that @xmath286 since @xmath287 .
furthermore , @xmath288 , and so again , we have that @xmath289 since we fixed an orthonormal frame @xmath290 , the action @xmath291 for a matrix @xmath292 , where the coefficients of @xmath292 are constant inside @xmath73 . viewing this integral on @xmath293 and extending @xmath294 to the whole of @xmath293 by setting it to @xmath295 outside of @xmath283 , we obtain for any @xmath296 ( where @xmath297 inside @xmath73 ) , @xmath298 .
since the matrix @xmath292 is constant coefficient , we have that @xmath299 for any @xmath300 , where @xmath301 denotes the fourier transform . then , @xmath302 this is valid for any @xmath296 , therefore , @xmath303 .
hence , @xmath304 which implies that @xmath305 . for a general @xmath272 with @xmath267 ,
fix @xmath306 a precompact chart and @xmath307 .
there is a smooth partition of unity @xmath308 inside @xmath306 such that @xmath309 on @xmath73 , and we extend @xmath310 outside @xmath306 .
then , for any @xmath271 , @xmath311 since @xmath267 we have @xmath312 and it is easy to see , from the cauchy - schwarz inequality , that @xmath313 ( where the constant contains @xmath314 ) .
this proves that @xmath315 , and since @xmath316 , by running our previous argument with @xmath317 in place of @xmath294 , we obtain that @xmath318 whenever @xmath267 . the estimate in [ d : reg ]
is immediate .
this proves the critical regularity case . for the case that the coefficients of the metric and operator are @xmath7 , and since @xmath319 implies that @xmath0 is an elliptic operator , [ d : reg ] follows from the classical schauder interior regularity estimate .
see theorem 10.3 in @xcite . on combining these propositions , we use theorem [ thm : firstmain ] and theorem [ thm : secondmain ] to obtain the following theorem . the metric @xmath14 appearing in this theorem automatically induces a length space as a consequence of proposition 4.1 in @xcite by burtscher .
the operator @xmath0 on @xmath258 is essentially self - adjoint .
if the metric @xmath14 and @xmath24 are of class @xmath7 for @xmath25 , and @xmath0 is a @xmath7 coefficient operator , then @xmath9 on @xmath258 is essentially self - adjoint for @xmath320 .
moreover , @xmath258 is a core for @xmath321 for @xmath322 $ ] .
the following are then two noteworthy consequences .
we remind the reader that in these corollaries , a @xmath7 coefficient metric induces a @xmath323 coefficient connection , from which the respective operators are built .
let @xmath14 be a @xmath15 complete metric on a smooth manifold @xmath2 .
then , the hodge - dirac operator @xmath324 on @xmath325 , where @xmath213 is the exterior - algebra , is essentially self - adjoint . if @xmath14 is @xmath7 , @xmath21 , then we get that @xmath9 on @xmath325 is essentially self - adjoint for @xmath326 .
gaffney in @xcite obtains this result for @xmath210 on @xmath7 manifolds .
roelcke in @xcite obtains this theorem for @xmath210 restricted to functions , and cordes @xcite obtains this for @xmath327 restricted to functions when the metric is smooth .
gaffney in @xcite obtains it for all powers under the smoothness assumption .
let @xmath14 be a @xmath15 complete metric on a smooth spin manifold @xmath2 , with a spin structure @xmath16 .
let @xmath17 , where @xmath18 is the minimal complex irreducible representation . then , the associated _ atiyah - singer dirac operator _
@xmath19 is essentially self - adjoint on @xmath20 .
if the metric is @xmath7 for @xmath21 , then @xmath22 is essentially self - adjoint on @xmath20 for @xmath23 . when the metric @xmath14 is smooth , wolf in @xcite obtains this for @xmath19 and @xmath328 , and chernoff in @xcite obtains this for all powers @xmath329 .
the @xmath15 metric case is proved in @xcite by bandara , mcintosh and rosn . here , we consider elliptic first - order differential operators @xmath0 with @xmath7 coefficients . as a consequence of theorem [ thm : secondmain ] , we obtain the following .
let @xmath1 be a smooth bundle over a smooth manifold @xmath2 with rough metrics @xmath24 and @xmath14 .
suppose that @xmath14 induces a complete length space .
let @xmath0 be a first - order elliptic differential operator with @xmath7 coefficients , @xmath25 , that is symmetric on @xmath20 in @xmath27 .
then , @xmath28 on @xmath20 for @xmath330 is essentially self - adjoint .
fix @xmath156 and @xmath331 . by a partition of unity argument
, we can find @xmath332 such that @xmath333 on @xmath334 .
thus , on @xmath334 , @xmath335 , and hence , @xmath336 where the second equality follows from the symmetry of @xmath0 on @xmath20 .
thus , [ d : close ] is satisfied .
the condition [ d : reg ] is a consequence of elliptic regularity , see theorem 10.3 in @xcite .
this theorem is a generalisation of @xcite by chernoff for elliptic operators and with minimal regularity assumptions on the underlying metrics .
moreover , chernoff assumes that @xmath337 , where @xmath338 is the `` local velocity of propagation '' for @xmath0 inside a ball of radius @xmath339 . in the elliptic context , we are able to show that such an assumption is not necessary .
[ s : essneg2 ] in this section , we establish the equivalence of negligible boundary and essential self - adjointness for powers of operators .
this is the crucial property we exploit throughout the later parts of this paper .
we emphasise that the background geometric assumptions here are minimal . in particular
, we do not assume completeness .
the central tool is for us to be able to equate @xmath340 with @xmath341 .
we establish some preliminary lemmas that will aid us in obtaining this equality .
[ lem : local ] let @xmath160 with @xmath342 compact , @xmath73 a precompact open set with @xmath343 and @xmath344 any smooth connection in @xmath73 .
then , @xmath345 and there exists a constant @xmath32 ( dependent on @xmath344 and @xmath73 ) such that @xmath346 observe that , via a partition of unity argument , we can assume that @xmath343 with @xmath73 corresponding to a chart . also , since @xmath73 is precompact , any two smooth connections , as well as all of their powers , are comparable .
thus , we assume that @xmath344 is the flat connection inside @xmath73 with respect to a fixed trivialisation .
next , note by the hypothesis on @xmath11 , we have that @xmath347 , where by what we have said we can fix the norm @xmath348 thus , there is a sequence @xmath349 such that @xmath104 in @xmath350 .
using the fact that @xmath351 inside @xmath73 , we obtain @xmath352 as @xmath353 , the right hand side of this expression tends to zero , and hence , we obtain that @xmath354 .
since @xmath9 is closable and the sequence @xmath349 , we conclude that @xmath345 and @xmath355 .
the estimate follows easily . with the aid of this lemma ,
we compute the maximal domain . our proof here is inspired by the work of masamune in @xcite ( see theorem 2 , p.114 ) .
[ prop : dommax ] under the assumptions [ d : first]-[d : last ] for @xmath9 , we have the equality @xmath356 .
we first prove the second equality .
for that , note that : @xmath357 to prove the first equality , it suffices to show that @xmath358 .
fix @xmath359 , let @xmath360 be a pre - compact open cover of @xmath2 by locally comparable trivialisations / charts for both @xmath2 and @xmath1 .
let @xmath361 be a smooth partition of unity subordinate to @xmath362 . on writing @xmath180 ,
it is clear that @xmath363 pointwise almost - everywhere and also in @xmath13 .
we show that @xmath364 . for that , observe that by [ d : reg ] , we obtain @xmath160 .
fix a choice of smooth frame @xmath365 in @xmath366 , and the flat connection @xmath90 with respect to @xmath367 inside @xmath366 .
also , we have @xmath368 where @xmath369 . since @xmath370 is smooth , @xmath371 is bounded inside @xmath366 for all @xmath372 , by considering a sequence @xmath373 in @xmath350 with @xmath374 , we deduce that @xmath375 .
therefore , by lemma [ lem : local ] , we obtain that @xmath376 .
define @xmath377 , for @xmath378 , where @xmath379 is the standard symmetric mollifier .
we can choose @xmath380 so that @xmath381 .
by lemma [ lem : local ] , there is a constant @xmath382 such that @xmath383 since we assume that @xmath366 satisfy the local comparability condition , setting @xmath384 , we have @xmath385 where @xmath386 is the pullback of the lebesgue measure inside @xmath366 . by standard mollification theory
, we have that the right hand side of this expression tends to zero as @xmath387 . also , observe that @xmath388 .
fix @xmath389 and for each @xmath134 , choose @xmath390 such that @xmath391 for each @xmath39 , define @xmath392 , this is well - defined because @xmath361 is locally finite .
it is easy to see that @xmath393 , and to show that @xmath394 , it suffices to show that @xmath395 .
fix @xmath156 and note that @xmath396 by assumption , we have that @xmath397 , and so we are reduced to showing that @xmath398 since @xmath0 is local , @xmath9 is local , and we have that @xmath399 .
choose @xmath400 such that @xmath401 on @xmath334 , we then have @xmath402 thus , @xmath394 . by a similar argument
, we obtain that @xmath403 in the graph norm @xmath404 . with the aid of this proposition , we prove theorem [ thm : firstmain ] . if @xmath197 , then @xmath405 , where the last equality follows from proposition [ prop : dommax ] .
that is , @xmath406 , which shows that @xmath198 is self - adjoint .
now , if @xmath198 is self - adjoint , we have that @xmath407 and so @xmath199 is self - adjoint .
if @xmath199 is self - adjoint , since @xmath408 , it follows that is satisfied .
recall the notation @xmath409 with domain @xmath410 . assuming the negligible boundary condition
, we obtain that @xmath411 .
but since @xmath412 , we have that @xmath413 by proposition [ prop : dommax ] .
so , @xmath414 and so by considering the bi - adjoint , we obtain that @xmath415 .
the essential self - adjointness of @xmath200 is equivalent to the self - adjointness of @xmath198 .
[ s : core ] in this section , we prove that , if @xmath416 is self - adjoint , then so is every power @xmath417 for @xmath418 . the reasons for this involve no geometry , but only properties of the operators , and to highlight that , we will keep the presentation sufficiently abstract .
the way we will proceed is to move from the operator @xmath419 to the operator @xmath420 , then use results from interpolation theory to assert that @xmath421 has @xmath20 as a core .
we then employ an alternative argument to show that this is a core for @xmath422 , then use functional calculus and operator theory to assert that @xmath423 the functional calculus we use here is the holomorphic functional calculus . for self - adjoint operators
@xmath100 , this functional calculus is given by : @xmath424 where @xmath425 is a curve cutting the spectrum at zero and infinity inside a bisector .
the functions @xmath426 are holomorphic on a bisector and decay sufficiently rapidly at @xmath295 and at @xmath427 . the functional calculus can be extended to bounded holomorphic functions @xmath34 on a bisector .
a detailed exposition of these ideas can be found in @xcite by albrecht , duong , mcintosh and in the book @xcite by haase .
[ lem : core1 ] let @xmath100 be a non - negative self - adjoint operator on a hilbert space @xmath97 , and let @xmath428 be a core for @xmath100 .
then , @xmath429 is a core for @xmath430 for all @xmath431 $ ] .
the case @xmath432 are easy , so we fix @xmath433 .
first we show that @xmath99 is dense in @xmath434 .
theorem 6.6.1 in @xcite yields that the real - interpolation space , @xmath435 has @xmath99 as a dense subspace for @xmath433 and @xmath436 . by theorem 4.3.12 in @xcite , on choosing @xmath80 , we obtain that @xmath437 .
next , note that @xmath438 by functional calculus since @xmath439 , where @xmath440 .
moreover , since @xmath441 , we get that @xmath442 for all @xmath106 .
fix @xmath443 , since we have already proved that @xmath99 is dense in @xmath434 , we can find @xmath444 such that @xmath445 moreover , since @xmath429 is a core for @xmath100 , we can find @xmath446 such that @xmath447 thus , on combining this with the estimate @xmath448 for @xmath106 , and since @xmath444 , @xmath449 this shows that @xmath429 is a core for @xmath430 . in the following lemma , we note that for a self - adjoint operator @xmath100 , @xmath450 .
[ lem : powermod ] let @xmath100 be self - adjoint on @xmath97 . then , for any integer @xmath451 , @xmath452 with @xmath453
consider the functions @xmath454 each such function is holomorphic , bounded , and hence , @xmath455 and @xmath456 .
this shows that @xmath457 in graph norm . to obtain the equivalence of norms
, we note that by the self - adjointness of @xmath100 , @xmath458 . for higher powers , @xmath459 where the second equality follows from functional calculus . for the next lemma ,
we specialise to the operator in question .
[ lem : htof ] let @xmath0 be an operator satisfying [ d : first]-[d : last ] . if @xmath198 is self - adjoint , then @xmath460 is also self - adjoint .
first , we consider the case that @xmath461 .
the symmetry condition [ d : close ] implies that @xmath460 exists and @xmath462 so that @xmath463 .
since @xmath464 is closed , @xmath465 and hence , by lemma 3 in @xcite ( with @xmath466 and @xmath467 ) , @xmath460 is self - adjoint .
thus @xmath468 .
now , for @xmath469 , we replace @xmath0 by @xmath470 and by this argument , we obtain that @xmath471 is self - adjoint whenever @xmath472 is self - adjoint . repeating this procedure
@xmath473-times , we obtain that @xmath460 is self - adjoint .
next , suppose that @xmath148 is an odd number .
assume for contradiction that @xmath460 is not self - adjoint .
then @xmath474 or @xmath475 .
without loss of generality , we assume @xmath475 and we can find a non - zero vector @xmath476 such that @xmath477 . by proposition [ prop : dommax ]
, we have that @xmath478 , so @xmath479 which implies that @xmath480 for all @xmath481 . invoking [ d : close ] , since @xmath480 for all @xmath482 , @xmath483 for all @xmath484 .
that is , @xmath485 . but @xmath198 is self - adjoint , @xmath486 , and so we have @xmath487 , which contradicts that @xmath198 is self - adjoint . for a general @xmath148 ,
write @xmath488 for @xmath489 odd , we obtain that @xmath460 is self - adjoint whenever @xmath198 is self - adjoint . with the aid of this ,
we obtain a proof of theorem [ thm : firstmain2 ] .
first , @xmath462 yields that @xmath490 and so @xmath491 .
the operator @xmath198 is self - adjoint by hypothesis and @xmath492 is self - adjoint since @xmath460 is self - adjoint .
then , by lemma [ lem : htof ] , we have that @xmath493 . by lemma [ lem : core1 ]
, @xmath20 is a core for @xmath494 . on considering the functions @xmath495 and observing that @xmath496 is bounded holomorphic on @xmath497 , we obtain @xmath498 and hence , @xmath499 thus , we have that @xmath20 is a core for @xmath500 for @xmath431 $ ] . on setting @xmath501 for integers @xmath502 , by lemma [ lem : powermod ] , we obtain that @xmath20 is a core for @xmath203 .
throughout this section , we assume that @xmath14 is a rough metric that induces a _ complete length _ space . we will denote the distance metric of this length space by @xmath503 .
furthermore , we assume the bundle @xmath1 is equipped with a rough metric @xmath24 .
we note by [ d : close ] that @xmath507 for all @xmath508 and @xmath173 .
thus , it suffices to show that @xmath174 . for that , fix @xmath509 . on fixing a smooth partition of unity subordinate to locally comparable precompact charts , we can further assume that @xmath510 , where @xmath73 is a precompact open chart .
inside there , fix the flat connection @xmath344 , and hence , by [ d : reg ] , we can deduce that that whenever @xmath511 with @xmath343 , implies that @xmath512 with the estimate @xmath513 now , let @xmath514 , where @xmath225 is the frame in @xmath73 for which @xmath515 is flat and @xmath516 is the standard symmetric mollifier @xmath517 with @xmath518 .
hence , we have that @xmath519 for @xmath520 large enough , and moreover , @xmath521 . in particular , @xmath522 and moreover , @xmath523 in @xmath27 .
thus , @xmath524 and the right hand side tends to zero as @xmath353 . by the closedness of @xmath460
, we have that @xmath173 .
it is easy to see that this now extends for @xmath172 .
when we apply this proposition in later parts , we will be taking @xmath525 and considering the section @xmath526 , where @xmath527 . in this case
we have that @xmath528 is @xmath529 and hence this lies in @xmath530 .
thus , we can apply this proposition to this case . in @xcite
, chernoff defines the _ local velocity of propagation _ by : @xmath531 and the _ velocity of propagation inside a ball _ by @xmath532 it is easy to see that @xmath338 is non - negative and monotonically increasing . since @xmath351 locally with @xmath533 for some @xmath134
, we have that @xmath534 for all @xmath339 .
also , since the operator locally takes the form @xmath131 , it is easy to see that @xmath535 . by covering the ball @xmath536 by a finite number of locally comparable precompact charts , and using that @xmath537 , we obtain : there exists @xmath538 such that @xmath539 for almost - every @xmath540 .
therefore , for each @xmath339 , the velocity of propagation satisfies @xmath541 .
if we take a smooth function @xmath34 , we then find that for any smooth section @xmath11 @xmath542 for almost - every @xmath39 .
letting @xmath543 denote the @xmath13 norm on the ball @xmath536 , we obtain @xmath544 fix a point @xmath545 in @xmath2 , and for any point @xmath39 , let @xmath546 , where @xmath503 denotes the distance function associated to the length structure induced on @xmath2 by @xmath14 .
the triangle inequality shows us that @xmath547 for any @xmath548 .
thus @xmath549 is lipschitz and hence differentiable almost - everywhere . inside a locally comparable chart @xmath550
, we have @xmath551 such that @xmath552 for almost - every @xmath277 . for @xmath339
we let @xmath536 denote the open ball centred about the fixed point @xmath545 . by the assumption that our rough metric induces a complete length space , we are able to apply the metric space version of the hopf - rinow theorem ( see proposition 3.7 in @xcite ) and obtain that metric balls @xmath536 are precompact . as a consequence , on a ball @xmath536
, there exists @xmath553 such that @xmath554 for almost - every @xmath540 .
define @xmath555 and note that @xmath556 is increasing in @xmath557 . also , define @xmath558 which is again an increasing function in @xmath557 .
choose a smooth function @xmath559 $ ] such that @xmath560 = 1 $ ] , @xmath561 , and such that @xmath489 is non - zero on the interval @xmath562 .
denote @xmath563 .
for each @xmath339 , define @xmath564 $ ] by @xmath565 .
then @xmath566 on @xmath567 and @xmath568 .
furthermore @xmath569 is non - zero on @xmath570 , and since @xmath570 is precompact , it follows that @xmath569 is compactly supported .
the lipschitz property of @xmath549 implies that @xmath569 is lipschitz and hence almost - everywhere differentiable . at points of differentiability , and points where @xmath571 is defined ,
we have @xmath572 we remind the reader that @xmath571 is defined almost - everywhere so the above gradient bound holds almost - everywhere on @xmath2 . in particular , @xmath573 is bounded above by @xmath574 .
[ d_essen_self ] let @xmath1 be a vector bundle with a rough metric @xmath24 over a manifold @xmath2 with a rough metric @xmath14 inducing a complete length space .
if the operator @xmath0 satisfies [ d : first]-[d : last ] , then it satisfies the negligible boundary condition , and @xmath0 on @xmath20 is essentially self - adjoint .
we fix two arbitrary sections @xmath580 , what we need to prove is that @xmath581 in order to do this , we consider the truncated section @xmath582 for some @xmath339 .
this section is lipschitz with compact support .
we then have that @xmath583 where the first equality comes from proposition [ lip_sym ] , and the second from the fact that @xmath584 $ ] . from
and we see that @xmath585 as @xmath586 .
furthermore , @xmath587 in @xmath13 as @xmath586 by the dominated convergence theorem , and similarly for @xmath588 .
this implies that if we let @xmath586 in the above equality , we obtain @xmath581 since @xmath589 were arbitrary sections it follows that such an equality holds for all @xmath580 , which is the negligible boundary condition .
the conclusion follows from theorem [ thm : firstmain ] . in this subsection
we will show that the powers @xmath28 on @xmath20 , for @xmath161 , are essentially self - adjoint when the operator @xmath0 has @xmath7 coefficients .
the case @xmath166 holds even for operators with @xmath146 coefficients and was obtained in theorem [ d_essen_self ] .
we do not expect the essential self - adjointness to hold in general for @xmath28 with @xmath590 since even for smooth , compactly supported sections @xmath11 , we can only make sense of @xmath591 distributionally .
our approach will be based on certain local estimates over a ball of radius @xmath557 .
we will then show that these local estimates prolong to global estimates on the whole manifold by taking the radius @xmath557 to infinity .
these global estimates allow us to prove the negligible boundary condition , and from our previous work , establish essential self - adjointness .
this idea of using local estimates to establish essential self - adjointness of an operator is inspired by wolf s work on the dirac operator in @xcite .
more specifically , in proposition 6.2 in @xcite , wolf establishes a certain global estimate of @xmath592 in terms of @xmath593 and @xmath594 and this , along with some other facts , allows him to conclude the essential self - adjointness of @xmath210 , where @xmath0 is the dirac operator .
we will show that when @xmath595 one can bound @xmath596 in terms of @xmath597 and @xmath594 .
this implies that @xmath598 , which we will then exploit to prove the negligible boundary condition for @xmath599 . we should also mention that proposition 6.2 in @xcite
is based on the method of andreotti - vesentini ( see 6 in @xcite ) .
in addition to the assumptions of theorem [ d_essen_self ] , for the remainder of this section , we assume that @xmath0 is a @xmath7 coefficient operator satisfying [ d : first]-[d : last ] for @xmath28 with @xmath600 . fix a smooth section @xmath11 of @xmath1 .
consider the compactly supported section @xmath609 . since @xmath569 is differentiable almost - everywhere , at points of differentiability
, we have the following formula : @xmath610 we then fix a smooth compactly supported function @xmath611 such that @xmath612 on @xmath613 and @xmath310 outside of @xmath614 .
we have @xmath615 where we have used proposition [ lip_sym ] to get the last equality , noting that @xmath616 . as @xmath617 has support contained in @xmath618
, we have that the above inner products are zero on the complement of @xmath618 .
this implies , using the fact that @xmath612 on @xmath613 , that @xmath619 for every @xmath378 sufficiently small
. therefore it must be true on the open ball @xmath570 , so we find @xmath620 using cauchy - schwarz and the estimate , for any @xmath621 we obtain the bound @xmath622 for any @xmath623 , the cauchy - schwarz inequality also implies @xmath624 using and in we obtain @xmath625 which we can simplify to @xmath626 by choosing @xmath557 large enough we can find a @xmath605 such that @xmath608 for any @xmath627 , and hence we have proved .
the above proposition [ dir_est ] implies that @xmath631 .
furthermore , theorem [ d_essen_self ] implies that @xmath0 satisfies the negligible boundary condition on @xmath2 . therefore ,
if we take @xmath632 we have @xmath633 which is the negligible boundary condition for @xmath210 . from theorem [ thm : firstmain ]
it follows that @xmath210 is essentially self - adjoint on @xmath2 .
[ rem_dir_est ] we observe that part of the hypothesis of proposition [ dir_est ] involved assuming that our section @xmath601 .
this was done so that in the _ global _ estimate , we knew that the right hand side was finite , hence this gives us that the left hand side is finite .
in the case of the _ local _ estimate , the assumption that @xmath601 is not needed .
it suffices to assume that @xmath11 is smooth ( in fact @xmath634 is enough ) .
this is because we can fix a smooth compactly supported function @xmath611 taking the value @xmath612 on @xmath613 and which vanishes outside of @xmath614 .
then , @xmath635 and we can use the fact that @xmath636 on @xmath570 where the local estimate holds . we will see that this is a crucial observation in the argument to obtain local estimates for higher @xmath637 . in proposition [ dir_est ] , in order to obtain our required global estimate , we first proved an estimate on the ball @xmath570 before taking a limit @xmath586 .
in order to prove the essential self - adjointness of higher powers @xmath599 for @xmath638 , we will proceed along the same lines .
our goal will therefore be to obtain a similar bound on @xmath639 in terms of @xmath640 and @xmath641 , over balls of sufficiently large radius
. we will give the details for the case @xmath642 , which involves some slight modifications of the above @xmath166 case , and then we will show how to do the general case via induction .
we start by estimating @xmath649 . as in the proof of proposition [ dir_est ] we fix a smooth compactly supported function @xmath611 such that @xmath612 on @xmath613 and @xmath310 outside of @xmath614 .
we then have @xmath650 where we have used proposition [ lip_sym ] to obtain the first equality ( noting that @xmath651 , hence is in @xmath652 ) , and the fact that at points of differentiability of @xmath653 , @xmath654 . since @xmath569 has support contained @xmath618 and @xmath612 on @xmath613 , the above gives @xmath655 for any @xmath378 sufficiently small . by taking @xmath656
we obtain @xmath657 we now want to plug in our local estimate for the @xmath661 term occurring on the right .
we remind the reader that at first sight this seems impossible . in obtaining a local estimate for @xmath662 we assumed @xmath601 , and
our hypothesis at this point is that @xmath643 .
however , we do not know , a priori , that @xmath601 and in fact , this is what we will prove when we obtain the associated _ global _ estimate .
however , as we observed in remark [ rem_dir_est ] , we actually do not require @xmath11 to be in @xmath663 for the _ local _ estimate .
it suffices to ask that @xmath11 be smooth .
thus , on substituting the local estimate into the above inequality , we obtain : @xmath664 we can re - write this to give @xmath665 where we remind the reader that @xmath666 .
we are now left with showing that we can choose @xmath557 and @xmath645 sufficiently large so that the coefficients in the estimate are positive .
it is easy to see that for large @xmath557 we have @xmath667 .
we take @xmath668 , where @xmath669 is a parameter , and observe that by taking @xmath557 and @xmath669 sufficiently large , we can make the coefficient @xmath670 on the left hand side positive .
this finishes the proof .
observe that if we take @xmath586 , we have that @xmath671 , and @xmath672 .
so by choosing @xmath645 appropriately , @xmath673 can always be made greater than zero .
therefore by taking @xmath586 in the above local estimate , we obtain the following global estimate .
the above theorem shows us that @xmath677 , and we know from proposition [ dir^2_essen_self ] that @xmath631 .
therefore , if we take @xmath680 then @xmath681 which is the negligible boundary condition for @xmath679 .
it follows from theorem [ thm : firstmain ] that @xmath679 is essentially self - adjoint . the general case of proving essential self - adjointness of @xmath599 proceeds along similar lines . in obtaining a local estimate for @xmath682
, we will meet constants of the form @xmath683 that will depend on a given set of positive numbers @xmath684 , and constants @xmath685 that arise from the local estimate for @xmath686 .
the goal will be to show that we can pick these constants to be positive , by making a suitable choice for @xmath687 .
for example , in the case of obtaining a local estimate for @xmath678 in terms of @xmath688 and @xmath11 , the constant @xmath689 takes the form @xmath690 where we remind the reader that @xmath691 and @xmath666 .
the type of question we will be faced with is : given @xmath692 can we choose @xmath693 so that @xmath694 for @xmath695 ?
the idea would be to choose @xmath557 and @xmath696 sufficiently large so that @xmath697 .
the problem is that as we vary @xmath696 the constant @xmath670 also changes , so there is a non triviality that we need to prove .
the way in which we proceed is to generalise the argument demonstrating how @xmath670 can be made positive .
we define @xmath698 , where @xmath699 is a parameter to be chosen later .
we then define @xmath700 , and substitute these into the formulas for @xmath670 and @xmath701 to obtain @xmath702 in the above formula for @xmath701 , we can see that the @xmath703 term does not depend on @xmath699 .
therefore , we start by choosing @xmath557 sufficiently large so that @xmath667
. then we choose @xmath669 and @xmath557 sufficiently large so that @xmath704 , and finally choose @xmath699 and @xmath557 sufficiently large so that @xmath697 .
we will prove this by induction .
we have already seen that it is true for the @xmath717 case ( and in fact the @xmath718 case ) .
so assume it is true for all @xmath719 , we will prove it is true for @xmath720 . the formula for @xmath721
is given by @xmath722 we observe that if we insert @xmath709 into the @xmath723-th position for @xmath724 , we only affect the @xmath725 in the above formula for @xmath721 .
our induction hypothesis tells us that this term can be made independent of @xmath726 .
if we then substitute @xmath727 into the @xmath728-th position , we see that we do not affect the @xmath725 term , because this has been made independent of @xmath729 , and then we see that we get rid of the @xmath730 dependence arising in the other terms . in particular , it follows that @xmath731 is independent of @xmath732 , and this establishes that @xmath721 satisfies the first condition . for the second condition , we observe that @xmath731 is given by @xmath733 our induction hypothesis allows us to choose @xmath734 , and @xmath557 sufficiently large so that @xmath694 for @xmath719 .
furthermore , the first condition of the induction hypothesis implies that the constant @xmath735 does not depend on @xmath729 , which in turn allows us to conclude that @xmath736 does not depend on @xmath737 .
therefore , by taking @xmath737 and @xmath557 sufficiently large we can make it so that @xmath738 this establishes the second condition for @xmath721 .
it follows by induction that it is true for all @xmath706 . as for the last statement
, simply observe that if we are given positive numbers @xmath714 we can substitute @xmath739 into the @xmath723-th position for @xmath710 . by condition ( ii ) we can then choose @xmath740 , and @xmath557 sufficiently large so that @xmath694 .
we will prove this by induction , the @xmath642 case being done in proposition [ dir^2_loc_prop ] .
so assume the proposition is true for @xmath745 .
this means that for @xmath525 , given any @xmath746 we can choose positive @xmath747 sufficiently large so that @xmath694 for @xmath748 , and so that @xmath749 it should be noted , as we observed in remark [ rem_dir_est ] , that such an estimate holds merely for @xmath152 .
when we substitute this estimate into the derived estimate for @xmath682 we will use this fact without explicit mention . by lemma [ constants ] , we can choose @xmath750 and @xmath557 sufficiently large so that @xmath694 for @xmath743 .
furthermore , in the proof of that lemma , we showed how we could choose @xmath750 so that the @xmath751 would not depend on @xmath729 .
in particular , this means that for these @xmath694 for @xmath752 the above estimate for @xmath686 holds with all the @xmath694 .
therefore , all that we need to do is show that the above estimate holds . to obtain an estimate for @xmath682 we start by estimating @xmath753 .
we begin by fixing a smooth compactly supported function @xmath611 such that @xmath612 on @xmath613 and @xmath611 vanishing outside of @xmath614 .
we then have that @xmath754 where the second equality follows from proposition [ lip_sym ] . using cauchy - schwarz , the bound , and the fact that @xmath757 does not have any zeros in @xmath570
, we have @xmath758 similarly we have @xmath759 these two bounds imply @xmath760 we now substitute the estimate for @xmath761 , as given in the induction hypothesis , to obtain @xmath762 this can be re - written as @xmath763 by definition , the coefficient on the left hand side is precisely @xmath721 , and so we have established the estimate for @xmath682 .
the local estimates we obtain , from the above proposition , can be promoted to a global estimate by taking @xmath764 .
all we have to note is that we have explicit formulas for the constants .
we know that @xmath666 , and in general @xmath765 therefore , it follows that as @xmath586 we have that @xmath671 , and in general @xmath766 one can then prove an analogous result to lemma [ constants ] for the @xmath767 .
we will not do this as the proof is exactly the same as the proof of lemma [ constants ] .
using this fact , and taking @xmath586 in the local estimate in proposition [ dir^k_loc ] , we obtain the following theorem . the above theorem shows us that @xmath774 for @xmath161 .
fixing @xmath775 , we have using theorem [ d_essen_self ] that @xmath776 this implies that @xmath28 satisfies the negligible boundary condition for @xmath777 and the conclusion follows from theorem [ thm : firstmain ] .
david albrecht , xuan duong , and alan mcintosh , _ operator theory and harmonic analysis _ , instructional workshop on analysis and geometry , part iii ( canberra , 1995 ) , proc .
centre math .
34 , austral .
, canberra , 1996 , pp . 77136 .
mr 1394696 martin r. bridson and andr haefliger , _ metric spaces of non - positive curvature _ , grundlehren der mathematischen wissenschaften [ fundamental principles of mathematical sciences ] , vol .
319 , springer - verlag , berlin , 1999 .
mr 1744486 alexander cardona , ivn contreras , and andrs f. reyes - lega , _ geometric and topological methods for quantum field theory : proceedings of the 2009 villa de leyva summer school _ , cambridge university press , 2013 .
bennett chow , sun - chin chu , david glickenstein , christine guenther , james isenberg , tom ivey , dan knopf , peng lu , feng luo , and lei ni , _ the ricci flow : techniques and applications .
part i _ , mathematical surveys and monographs , vol .
135 , american mathematical society , providence , ri , 2007 , geometric aspects .
mr 2302600 ( 2008f:53088 ) alessandra lunardi , _ interpolation theory _
, second ed .
, appunti .
scuola normale superiore di pisa ( nuova serie ) .
[ lecture notes .
scuola normale superiore di pisa ( new series ) ] , edizioni della normale , pisa , 2009 .
|
we consider first - order differential operators with locally bounded measurable coefficients on vector bundles with measurable coefficient metrics . under a mild set of assumptions ,
we demonstrate the equivalence between the essential self - adjointness of such operators to a negligible boundary property .
when the operator possesses higher regularity coefficients , we show that higher powers are essentially self - adjoint if and only if this condition is satisfied . in the case
that the low - regularity riemannian metric induces a complete length space , we demonstrate essential self - adjointness of the operator and its higher powers up to the regularity of its coefficients .
we also present applications to dirac operators on dirac bundles when the metric is non - smooth .
|
one of the important problems in the control of quantum systems is the decoherence which emerges from undesirable interactions @xcite .
specially in quantum computation processes , it is very important to find a quantum memory which is robust against emergent errors . in recent years , finding a solution for the decoherence problem has been one of the biggest challenges @xcite .
one of the options to circumvent this problem is topological quantum computation where topological nature of a physical system is used to perform a fault tolerant quantum computation @xcite .
+ the key property in the topological quantum computational model is topological order .
topological order is a new phase of matter which for the first time was presented by wen in 1980 @xcite .
it is completely different from ordinary order such as ferromagnetic order so that the topological orders can not be described by landau paradigm and symmetry breaking theory @xcite .
also there is no local order parameter which recognizes topological phase .
in fact topological order is described by a non - local order parameter and it is independent from local details of physical systems @xcite .
there are many physical systems in condensed matter physics with topological order @xcite , for example in quantum hall effects @xcite , high - temperature superconductors @xcite and frustrated magnetism @xcite . + among physical systems with topological order ,
lattice models have been taken into more consideration , because of their simplicity . in 1997 , kitaev presented a model hamiltonian on a torus which could encode the subspace of two qubits in it s ground states @xcite .
thereafter , topological color codes were introduced by bombin et al .
@xcite where the unitary clifford gates could be topologically applied while this was not possible in the toric code .
+ degeneracy of the ground state in the topological codes depend on topological nature of the model and local small perturbations can not lift such a degeneracy . the robustness of the degeneracy of the ground state is only guaranteed in small perturbations so that when perturbation increases , at a critical value , a phase transition occurs .
in fact , there is a topological phase transition ( tpt ) where the point of phase transition is a measure of the robustness of the code .
the robustness of the topological codes are studied against thermal perturbations and local perturbations @xcite . as an example ,
robustness of the toric codes against parallel and transverse magnetic fields have been studied where magnetic field can be treated as an extrinsic factor which perturbs topological order @xcite .
if the perturbation is as ising interaction , there is a phase transition between topological phase and the ferromagnetic phase .
such a problem has been studied for the toric code in an ising perturbation ( tci ) on a square lattice @xcite and for the color code in ising perturbation ( cci ) on the honeycomb lattice in @xcite .
+ in this paper , we focus on the effect of the ising perturbation on topological codes on various lattices .
we show that such models are mapped to ising model in transverse field ( it ) by re - writing the hamiltonian in a new basis .
since the approximated phase transition points for the it on different lattices are well - known @xcite , we provide a useful table to compare the phase transition points of topological codes on various lattices .
+ we consider dependence of the phase transition point and also the robustness of topological codes on triangulation of their underlying lattices .
in fact , for a pure topological model on a specific lattice , there are some topological properties which are independent of triangulation of the lattice @xcite . when the perturbation is added , the system is not topological .
however , before the phase transition occurs , topological properties are still independent of triangulation .
since the point of phase transition is a property of the perturbed topological code , it is expected that the phase transition point depends on triangulation .
we derive points of phase transition for perturbed topological codes on various triangulations and explicitly show dependence of phase transition points on triangulation of the lattice .
+ in the first step , we compare the robustness of cci s on two specific lattices .
we show that they are mapped to ising model in transverse field on two different lattices with different phase transition points .
the same result is derived for the tci s on three different lattices .
we emphasize that although it is well known that the phase transition point in an ordinary quantum phase transition depends on the underlying lattice , such a problem has not been explicitly studied for a tpt .
+ in the second step , we compare the robustness of the tci and cci .
such a work has been performed for the topological codes in a parallel field . in @xcite ,
the authors have compared the robustness of the color code and the toric code in parallel field .
they have concluded that the color code is more robust than the toric code against parallel magnetic field .
we show that underlying lattices of these codes are important in such a problem so that the robustness of tci on square lattice is less than the robustness of the cci on honeycomb lattice while the tci on honeycomb lattice is as robust as the cci on honeycomb lattice .
so the robustness of topological codes is very sensitive to triangulation of the lattices and we can not compare different topological codes without considering the underlying lattices .
the structure of this paper is as follows : + in section ( [ sec1 ] ) we review topological properties of two topological codes , the color codes and the toric codes . in section ( [ sec2 ] )
we apply ising perturbation to the topological color codes on various lattices .
then we apply the same perturbation to the toric code . in this section
we present a detailed mapping to re - write hamiltonian of the models in a new basis so that the initial models are converted to the ising model in the transverse field .
finally , in section ( [ sec3 ] ) , we provide a table of the phase transition points and compare the robustness of the topological codes .
in this section , we review some important properties of the topological quantum codes . specifically , we consider two well - known topological codes , the color codes and the toric codes .
topological order of these models is a non - local property which does not depend on the underlying lattices .
for example the color codes can be defined on the honeycomb and square - octagonal lattices so that , in both cases , topological order is the same . in this subsection ,
we review the topological color codes ( tcc ) .
it has been shown that they can be used for topological implementation of clifford unitary gates and it is an advantage of these codes compared with the toric codes @xcite .
+ the tcc models are defined on three - colorable lattices where one can color all faces ( plaquettes ) of the lattice by three different colors so that any two neighbor plaquettes are not in the same color @xcite . here
we consider the tcc model on a honeycomb lattice , but we emphasize that topological properties are the same for all tcc models . to this end , consider a honeycomb lattice on a torus where spins live on vertices of the lattice , see figure ( [ hexagonal ] ) .
there are two kinds of operators corresponding to each plaquette of the lattice as follows : @xmath0 where @xmath1 and @xmath2 denote the pauli operators @xmath3 and @xmath4 , respectively .
if the number of the plaquettes is @xmath5 , the number of the plaquette operators will be equal to @xmath6 .
the most important problem is to find the stabilized subspace of the code which is defined as follows : @xmath7 this subspace can also be considered as the subspace of ground states of a hamiltonian as : @xmath8 by the fact that @xmath9 , it is simple to check that the following state is a ground state of the hamiltonian ( [ top ] ) : latexmath:[\[\label{00 } is eigenstate of the pauli operator @xmath2 .
also there are many constraints between the plaquette operators .
as it has been shown in figure ( [ hexagonal ] ) , the honeycomb lattice is colored by three colors such as blue , green , red . according to the periodic boundry conditions on the torus ,
it is clear that the following constraints will be hold on the plaquette operators : @xmath11 where @xmath12 refers to the plaquttes which are colored by red and so on .
hence , from @xmath6 plaquette operators , only @xmath13 number of them are independent . because the number of qubits is also equal to @xmath6 , degeneracy of the ground state is sixteen - fold .
+ in the following , we show that all ground states are constructed by applying some non - local operators to state ( [ 00 ] ) . for defining these non - local operators , we consider that all edges of the honeycomb lattice can also be colored by three colors so that each edge with a specific color connects two plaquettes with the same color , see figure([hexagonal ] )
. then we define three kinds of color strings on the lattice where each string moves on the edges and plaquettes with the same color and two endpoints of the strings live in the plaquettes , see figure ( [ hexagonal ] ) .
we denote these strings by @xmath14 .
we define two string operators @xmath15 corresponding to each string in the form of : @xmath16 where the pauli operators are applied on all qubits which live on the string @xmath17 and @xmath18 can be each one of three colors , red , green or blue . if two end points of a string are connected to each other , it generates a closed loop and we will have a corresponding loop operator . since each plaquette shares in two spins with the loop operator , it is simple to show each loop operator commutes with all plaquette operators in the hamiltonian ( [ top ] ) . by attention to topology of the torus ,
the loop operators are divided to two general classes .
one class contains trivial loops which are contractible and another contains non - trivial loops which are non - contractible . +
important point is about non - trivial operators .
there are six fundamental non - trivial loops which are denoted by three colors and two directions on the torus , see figure ( [ loops ] ) . related to each kind of the non - trivial loops ,
there are two operators which are constructed by @xmath2 or @xmath1 operators .
we denote these operators as : @xmath19 where @xmath20 refers to a non - trivial loop and @xmath18 denotes color of the loop and @xmath21 denotes direction of the loop around the torus .
an important property of the non - trivial loop operators is that they can not be written as product of the plaquette operators in the hamiltonian ( [ top ] ) .
but it is simple to show from three loop operators @xmath22 and @xmath23 , only two number of them are independent and one can be written as product of two other ones and some plaquette operators .
so we select only two colors of these loops , for example red and green .
+ because of the topology of the torus , two non - trivial loops where one is in direction @xmath24 and another is in direction @xmath25 cross each other , see figure ( [ loops ] ) , so there is an anti - commutation relation between the non - trivial loop operators as follows : @xmath26 other relations between these operators are commutation . according to the above relations the non - trivial loop operators generate a @xmath27 dimensional space which can encode the space of four qubits .
also we can explicitly construct all sixteen bases of this space by the non - local operators in relation ( [ q ] ) .
in fact , it is clear that the state @xmath28 as ( [ 00 ] ) is stabilized by @xmath29 , @xmath30 , @xmath31 , @xmath32 , because they commute with operators @xmath33 then they are applied to state @xmath34 which is eigenstate of the pauli operator @xmath2 .
so all bases of the degenerate subspace are constructed as follows : @xmath35 where @xmath36 and @xmath37 are 0 or 1 .
we should emphasize that degeneracy of the ground state in this model is related to the topology of torus and the colorable structure of the lattice .
the most simple of the topological codes are the toric codes which can be defined on each oriented graph @xcite . here
we consider a honeycomb lattice on a torus with @xmath5 spins which live on the edges of the lattice .
the toric code on such a lattice is defined by the following plaquette and vertex operators : @xmath38 where @xmath39 refers to spins which belong to plaquette @xmath40 and @xmath41 refers to spins which are neighbor of vertex @xmath42 , see figure ( [ kitaev ] ) .
the stabilized subspace of this code can treated as the ground state of a hamiltonian as : @xmath43 since @xmath44 , it is simple to show that the following un - normalized state is a ground state of this hamiltonian : @xmath45 and @xmath46 .
so the number of independent operators is equal to @xmath47 and the ground state has four - fold degeneracy . to find other ground states we should consider two non - trivial loops @xmath48 around two directions of the torus where loops can be defined on the edges of lattice as well as the edges of dual lattice where the vertices , edges and faces of dual lattice are in one to one correspondence with the faces , edges and vertices of the lattice , respectively .
there are two operators corresponding to a non - trivial loop @xmath49 on the lattice and a non - trivial loop @xmath50 on the dual lattice , see figure ( [ kitaev ] ) , in the form of : @xmath51 on one hand these operators commute with the hamiltonian so they have joint eigenstates , and furthermore there exist the following anti - commutation relations between these operators : @xmath52 other relations between these operators are commutation .
according to the above relations all four ground states of the model are derived as : @xmath53 where @xmath54 are 0 or 1 , and @xmath55 is a state as in eq.([as ] ) .
thus in the toric codes , topological order of the ground states is associated to the periodic conditions of the torus and it is independent from the underlying lattice .
+ the degenerate ground state of the topological codes can be used as a robust memory for quantum computation goals .
for example , consider a perturbation as magnetic field on spins in the toric code .
such a perturbation leads to splitting between the degenerate states .
the important point in this case is that two ground states of the model are only mapped to each other by a non - local operator as product of pauli operators on a non - trivial loop which is related to the topology of the model .
therefore , if the perturbation is small enough , such a splitting can not happen .
+ it is clear that the robustness of the degeneracy is lifted when perturbation increases .
in fact a perturbation even in the first order leads to a transition term between the ground state and the exited states in the perturbed energy levels .
this problem leads to a phase transition from topological order to an ordinary order .
finally , the topological order is lifted at a phase transition point .
therefore , the point of phase transition is a measure of the robustness of the topological codes . +
in this section , we consider the topological codes on different two - dimensional lattices with ising perturbation .
we show that these models are mapped to the ising model in transverse field by re - writing the initial hamiltonian in a non - local basis . in a general form ,
the hamiltonian of the topological code with ising perturbation is in the form of : @xmath56 where @xmath57 is the hamiltonian of the topological code and @xmath58 is the hamiltonian of the ising perturbation . at the first ,
let us consider two special regimes of the couplings . in case that @xmath59 or @xmath60 , we have the ising model which have a two - fold degenerate ground state where @xmath2 components of all spins are up or down . in this case system
is in the ferromagnetic phase and a local order parameter can describe phase of the model @xcite .
now we consider a case that @xmath61 or @xmath62 , so we have the topological model where the system is in a topological phase and any local order parameter can not describe such a phase @xcite .
since in these two different regimes we have two different phases , it is a reasonable expectation that at a critical ratio of @xmath63 a phase transition occurs . + we re - write the hamiltonian ( [ to ] ) in a new basis and we show that in this new basis such a hamiltonian is mapped to the ising model in the transverse field on a 2d lattice . + our method in studying the topological codes in presence of the ising perturbation is to re - write hamiltonian of the model in a new basis .
it leads to convert the initial model to an ising model in the transverse field .
similar method has been already used in other papers about the toric code on square lattice @xcite and the color code on the honeycomb lattice @xcite . in this subsection
, we explain the main idea for the tcc model on the honeycomb lattice then we apply it to other models on various lattices . to this end
, we consider a hamiltonian as : @xmath64 where @xmath65 denotes neighbor spins on the honeycomb lattice . in this model , we have considered a positiv coupling constant @xmath66 for the tcc hamiltonian and a positiv coupling constant @xmath67 for the ising perturbation .
+ at the first , by the fact that the ising perturbation commutes with the operators @xmath68 in the tcc model , we find that ground state of the new model is in a subspace which is stabilized by all operators @xmath68 .
we re - write the hamiltonian ( [ topi ] ) in such a subspace by choosing an un - normalized basis as follows : latexmath:[\[\label{qqq } where @xmath5 is the number of plaquettes of the lattice and @xmath70 s are binary numbers which are attributed to each plaquette .
in fact we can attribute a virtual spin to each plaquette where @xmath70 denotes the value of the virtual spin in a plaquette @xmath40 .
the states in eq([qqq ] ) are non - local so that they correspond to anyonic excitations of the tcc model . + the number of states as ( [ qqq ] ) is @xmath71 and all of them are stabilized by the operators @xmath68 .
about the above basis , we should emphasize that all these states are not independent , in fact there are constraints between the plaquette operators in different colors according to relation ( [ dd ] ) .
for example product of all operators @xmath72 and @xmath73 is equal to the identity and we find that @xmath74 .
this basis would be complete if we add some new states which are constructed by applying the non - trivial loop operators on the state @xmath55 .
+ in the following we should re - write the hamiltonian ( [ topi ] ) in the basis ( [ qqq ] ) .
as we explained , all states ( [ qqq ] ) are stabilized by @xmath68 s , so @xmath75 and since the number of @xmath68 s is equal to n , we find that @xmath76 .
+ then we apply the operator @xmath33 on the states ( [ qqq ] ) . by the fact that @xmath77 , we find that @xmath78 .
we can describe this final result in an interesting view so that effect of @xmath33 in the basis ( [ qqq ] ) is equivalent to effect of the pauli operator @xmath79 in a basis as @xmath80 on the virtual spins .
+ interesting situation occurs when we apply the ising term on the basis ( [ qqq ] ) . to this end , as we explained in the previous section , we can color all links of the honeycomb lattice in three colors .
thus we divide the ising interactions to three parts corresponding to the color of their links as : @xmath81 where @xmath82 denotes sites @xmath54 which are in two endpoints of a blue link and so on .
for example , we apply the ising interaction on the blue links to the state @xmath83 .
each link @xmath84 has four neighbor plaquettes and it is simple to show that corresponding operator to @xmath65 as @xmath85 does not commute with two plaquette operators @xmath33 corresponding to two plaquettes in two endpoints of that link , see figure([triangular ] , left ) .
therefore , we find that @xmath86 , where @xmath33 is corresponding operator to one of the end points of the edge @xmath65 .
finally , we conclude when the ising interaction corresponding to a blue link of the lattice is applied to basis @xmath83 , it rises up the virtual spins @xmath87 corresponding to two blue plaquettes in two end points of that link .
therefore , we can describe the ising interaction @xmath85 as an ising interaction @xmath88 on the basis @xmath80 , where @xmath89 are two blue plaquettes which are connected by a blue link . as you can see in figure ( [ triangular ] , right ) , if we denote this ising interaction between the neighbor virtual spins by a link , we will have a triangular lattice where all virtual spins in blue plaquettes live on the vertices of that lattice .
+ the above argument can also be repeated for the ising interactions on the green and red links where , corresponding to each color , we will have an ising model on a triangular lattice .
+ by considering all terms in the hamiltonian ( [ topi ] ) in the new basis , we will have a new hamiltonian in the form of : @xmath90 and @xmath91 in two endpoints of that link .
right : the virtual spins which live in blue plaquettes are denoted by blue circles .
an ising interaction @xmath92 between two real spins on blue links does not commute with two plaquette operators @xmath33 on two it s end points and it is equivalent to an ising interaction @xmath93 between virtual spins which live on corresponding plaquettes .
finally we have an ising model on a triangular lattice . , width=302,height=132 ] where @xmath94 refers to all virtual spins on the blue plaquettes and @xmath95 refer to the neighbor plaquettes in the same color on a triangular lattice . for finding the ground state of the hamiltonian
, we use this fact that the final hamiltonian in ( [ tran ] ) on the virtual spins has been written as summation of three independent hamiltonian on three triangular lattices with different colors .
therefore , it is enough to find the ground state of each part , independently .
interesting result is that each part of the hamiltonian is an ising model in a transverse field on a triangular lattice which a quantum phase transition has been known for it at @xmath96 @xcite . at @xmath97 limit all virtual spins are in positive direction of @xmath2 and correspondingly the real spins are in ground state of the color code . at @xmath98 limit
all virtual spins are in the ground state of the ising model on the triangular lattice and correspondingly the real spins are in the ground state of the ising model on the honeycomb lattice . + the color codes can be defined on any three - colorable lattices .
definition of the stabilizers are the same for all of them so that there are two kinds of stabilizers corresponding to each plaquette @xmath99 which are as product of @xmath1 or @xmath2 operators on the spins of each plaquette .
+ in this subsection , we consider the color code on a square - octagonal lattice in a ising perturbation , see figure ( [ sqhec ] ) . in order to find the phase transition point , similar to the previous section , we re - write the hamiltonian of the model in the basis ( [ qqq ] ) .
all of arguments are similar to the previous section so that here we divide the hamiltonian to three parts corresponding to three colors of the plaquettes of the lattice .
each part of the hamiltonian in the basis ( [ qqq ] ) is as an ising model in the transverse field on the virtual spins in plaquettes with the same color .
only distinction is kind of the ising interaction pattern on the virtual spins in each color . +
as it has been shown in figure ( [ sq ] ) , links of the lattice with green color generate a square lattice where the virtual spins live on it s vertices .
moreover , blue links generate a square lattice where the virtual spins live on it s vertices , see figure([sq2 ] ) .
also we should be care that there are two blue links between each two blue neighbor plaquettes .
this final point leads to an important distinction so that two ising interactions in original model on the real spins correspond to an ising interaction in the square lattice on the virtual spins . by considering a similar argument for the red plaquettes ,
the final hamiltonian on the virtual spins will be as follows : @xmath100 where @xmath101 refers to all virtual spins on the green plaquettes and @xmath95 are neighbor plaquettes in the same color which generate a square lattice .
each part of the above hamiltonian is an ising model in the transverse field on a square lattice .
for the first part , phase transition occurs at @xmath102 @xcite and for two other parts , phase transition occurs at @xmath103 . when the ising perturbation is increased by @xmath67 , at @xmath104 for the red and blue sublattices a phase transition occurs from topological order to ferromagnetic order but the green sublattice stays in topological phase . at @xmath105 topological order
is completely removed and we have a ferromagnetic order .
we can compare this result with that for the color code on the honeycomb lattice where phase transition occurs at @xmath106 . between two real spins on green links
does not commute with two plaquette operators @xmath33 on two it s end points and it is equivalent to an ising interaction @xmath93 between virtual spins which live on corresponding plaquettes .
finally we have an ising model on a green square lattice .
, width=151,height=132 ] between real spins on two blue links which connect two blue plaquettes . by the fact that these two ising interactions does not commute with two plaquette operators @xmath33 on two it s end points ,
we find that they are equivalent to an ising interaction @xmath93 between virtual spins which live on corresponding plaquettes .
finally we have an ising model on a square lattice . , width=151,height=132 ] in this subsection , we study the toric code on a honeycomb lattice with the ising perturbation . to this end , we consider the nearest neighbors of each spin and we assume an ising interaction between them .
as it has been shown in figure ( [ tri ] , left ) , if we illustrate each ising interaction by a link , ising interaction pattern is described by a triangle - hexagonal lattice which the spins live on it s vertices . in this new lattice the vertex operators of the toric code correspond to the triangular plaquettes and the plaquette operators of the toric code correspond to the hexagonal plaquettes .
finally the toric code with the ising perturbation can be described by a hamiltonian in the form of : @xmath107 by the fact that the operator @xmath92 commute with the operators @xmath108 and the non - trivial loop operators @xmath109 , we find that the ground state of the above hamiltonian is in a subspace which is stabilized by @xmath108 s and @xmath109 .
therefore , we select the following basis to re - write the hamiltonian ( [ asd ] ) : @xmath110 ) in this basis . first , by the fact that @xmath108 s commute with all operators @xmath111 , we find that @xmath112 .
second , we apply the operators @xmath111 , by the fact that @xmath113 , we find that @xmath114 .
it is useful to describe the final relation in view of virtual spins . to this end , if we insert virtual spins on the vertices of the initial honeycomb lattice , the state @xmath115 can treated as product state @xmath116 on the virtual spins .
therefore , effect of the operator @xmath111 on the new basis is equivalent to effect of operator @xmath117 on the virtual spin @xmath118 .
+ finally , we apply the ising operator @xmath85 to the basis ( [ q1 ] ) . to this end
, we consider a special link on the triangle - hexagonal lattice .
since the corresponding ising operator to such a link does not commute with two vertex operators in two it s endpoints , see figure ( [ tri ] , right ) , we find the following relation between each ising operator and a vertex operator in one of the it s endpoints : @xmath119 between two real spins does not commute with two vertex operators on two it s endpoints and it is equivalent to an ising interaction @xmath120 between corresponding virtual spins which are denoted by red circles . , width=302,height=132 ] therefore , effect of the operator @xmath85 on the basis ( [ q1 ] ) leads to flip the virtual spins which live on two endpoints of the corresponding link .
it means that the ising interaction @xmath85 on the real spins in the initial model is equivalent to an ising interaction @xmath120 on the virtual spins .
+ in the following , we should find effect of all ising interactions in the initial model . to this end , according to figure ( [ triangle ] ) , we color vertices of the initial honeycomb lattice by two different colors , blue and green . by attention to the above argument , each ising interaction between the real spins in the initial lattice , similar to what has been shown in figure ( [ tri ] , right ) , is equivalent to an ising interaction @xmath120 between two virtual spins in the same color .
finally , the ising interactions in the initial honeycomb lattice convert to the ising interactions on two triangular lattices in two different colors and the hamiltonian ( [ asd ] ) in term of the virtual spins are wrote as follows : @xmath121 where @xmath122 ( @xmath123 ) refers to the blue ( green ) triangular lattice .
therefore , the initial hamiltonian is converted to two independent ising models in the transverse field on triangular lattices , see figure ( [ triangle ] ) . as we pointed in the previous subsection ,
the phase transition point of this model occurs at ratio @xmath124 .
we can compare this result with the toric code on square lattice where the phase transition point is at @xmath125 @xcite .
+ as a final case , we consider the toric code on the triangular lattice . in such a model , spins live on the edges of a triangular lattice and two kinds stabilizer operators are defined corresponding to plaquettes and vertices of the lattice , see figure([tr ] ) .
ising perturbation is applied between neighboring spins where pattern of the ising interactions has been shown in figure([tr2 ] , left ) .
similar to the previous subsection , we consider a non - local basis as : @xmath126 in such a basis , the plaqutte operators @xmath108 are equal to identity and the vertex operators @xmath111 are equal to pauli operator @xmath117 on virtual spins which live on vertices of the lattices .
+ for ising interactions in the initial hamiltonian , as it has been shown in figure([tr2 ] , right ) , by the fact that each ising operator @xmath92 does not commute with two neighboring vertex operators @xmath111 and @xmath127 , we find that the ising operator in the initial model is equal to an ising operator @xmath120 in the new basis .
if we apply such a transformation for all ising operators , see figure([tr3 ] ) , we will have a triangular lattice on virtual spins with ising interaction @xmath120 between neighbor spins .
also there is an important point that two ising interactions in the initial model are mapped to an new ising interaction on virtual spins .
finally , the initial model in the new basis would be an ising model in a transverse field as : @xmath128 the phase transition point for this model is at @xmath129 .
.the corresponding ising model to each topological codes and the phase transition points have been shown . [ cols="<,<,<,<",options="header " , ]
a summary of results of the previous section has been shown in table ( [ tab ] ) . according to our derivations , two well known sets of topological codes on the most important lattices has been mapped to ising model in transverse field on different lattices .
+ first , we compare the robustness of the tci s on different lattices .
the tci on square lattice is mapped to it on square lattice while the tci on the honeycomb and the triangular lattices are mapped to it on triangular lattice with different phase transition points .
therefore , the robustness of the tci s depends on triangulation of the underlying lattice .
+ second , we compare the cci s on two different lattices .
the cci on honeycomb lattice is mapped to the it on triangular lattice while the cci on square - octagonal lattice is mapped to the it on square lattice .
therefore , the robustness of the cci s also depends on triangulation of the underlying lattice .
+ third , we compare the robustness of the tci s with the cci s . the robustness of the tci on square lattice is less than the robustness of the cci on honeycomb lattice while the robustness is the same for the tci and cci on honeycomb lattice .
therefore , triangulation of the underlying lattice is important in comparison of two different topological codes so that there is not any way to compare the robustness of two different topological codes independent of their underlying lattices . +
in this paper , we derived the points of phase transition for two important topological codes on various lattices with the ising perturbation , the toric code and the color code . to this end , we presented a detailed mapping between such models and the ising model in transverse field .
our results explicitly showed that points of the phase transition and also robustness of perturbed topological codes depend on triangulation of lattices .
also , by comparing the robustness of the toric code and the color code , we showed that there is no way to compare the robustness of two different topological codes independent of triangulation of the underlying lattices .
i would like to thank a. poostforush and a. montakhab for their help in finalization of the paper .
also i thank v. karimipour for his good comments on the paper .
99 maximilian schlosshauer , rev .
76:1267 - 1305 ( 2004 ) .
julia kempe , dave bacon , daniel a. lidar , k. birgitta whaley , phys .
rev . a 63 , 042307 ( 2001 ) .
a. r. calderbank , peter w. shor , phys .
rev . a , vol .
2 , pp . 1098 - 1106 ( 1996 ) .
a. y. kitaev , ann .
( n. y. ) 303 , 2 ( 2003 ) .
e. dennis , a. kitaev , a. landahl , j. preskill , j. math .
43 , 4452 - 4505 ( 2002 ) .
g . wen and q. niu , phys . rev .
b 41 , 9377 ( 1990 ) .
l. d. landau , phys . z. sowjetunion . 11 , 26 ( 1937 ) , goldenfeld n. , `` lectures on phase transitions and critical phenomena '' , ( perseus books publishing , l.l.c . , massachusetts , 1992 ) .
wen , adv .
44 , 405 ( 1995 ) .
wen , quantum field theory of many - body systems ( oxford university press , 2004 ) .
tsui , h.l .
stormer , and a.c .
lett , 48 1559 ( 1982 ) .
x .- g . wen , int . j. mod . phys .
b 5 , 1641 ( 1991 ) . t. h. hansson , v. oganesyan , and s. l. sondhi , ann .
phys . , 313 , 497 ( 2004 ) .
d. s. rokhsar and s. a. kivelson , phys .
61 , 2376 ( 1988 ) .
n. read and b. chakraborty , phys .
b 40 , 7133 ( 1989 ) .
r. moessner and s. l. sondhi , phys .
86 , 1881 ( 2001 ) .
e. ardonne , p. fendley , and e. fradkin , ann .
310 , 493 ( 2004 ) .
h. bombin and m. a. martin - delgado , phys .
97 , 180501 ( 2006 ) ; phys .
b , 75:075103 ( 2007 ) .
h. bombin , and m. a. martin - delgado , new j. phys .
13 , 083006 ( 2011 ) ; h. bombin , ruben s. andrist , masayuki ohzeki , helmut g. katzgraber , and m. a. martin - delgado , phys .
x 2 , 021004 ( 2012 ) .
h. bombin , m.a .
martin - delgado , phys.rev.lett.98:160502 ( 2007 ) .
masayuki ohzeki , phys .
e 80 , 011141 ( 2009 ) .
s. trebst , p. werner , m. troyer , k. shtengel , and c. nayak , phys .
98 , 070602 ( 2007 ) ; a. hamma and d. a. lidar , phys . rev .
100 , 030502 ( 2008 ) .
alioscia hamma , lukasz cincio , siddhartha santra , paolo zanardi , luigi amico , local response of topological order to an external perturbation , arxiv:1211.4538 gbor b. halsz , alioscia hamma , phys . rev . a 86 , 062330 ( 2012 ) .
jahromi , m. kargarian , s farhad masoudi , and k.p .
schmidt , phys .
b 87 , 094413 ( 2013 ) .
m. d. schulz , s. dusuel , r. orus , j. vidal , and k. p. schmidt , new journal of physics 14 , 025005 ( 2012 ) .
s. dusuel , m. kamfor , r. orus , k. p. schmidt , and j. vidal , robustness of a perturbed topological phase , physical review letters , 106 , 107203 , ( 2011 ) .
a. hamma , d. a. lidar , phys .
100 , 030502 ( 2008).j .
vidal , s. dusuel and k.p .
schmidt , phys .
b 79 , 033109 ( 2009 ) .
j. vidal , k.p .
schmidt , r. thomale , and s. dusuel , phys .
b 80 , 081104 ( 2009 ) .
tupitsyn , a. kitaev , n.v .
prokofev , and p.c.e .
stamp , phys .
b 82 , 085114 ( 2010 )
. f. wu , y. deng , and n. prokofev , phys .
b 85 , 195104 ( 2012 ) .
vahid karimipour , laleh memarzadeh , parisa zarkeshian , phys .
a 87 , 032322 ( 2013 ) .
saeed s. jahromi , mehdi kargarian , s farhad masoudi , kai phillip schmidt , phys .
b 87 , 094413 ( 2013 ) . h .- x .
he , c. j. hamer , and j. oitmaa , j. phys . a 23 , 1775 ( 1990 ) .
m. s. l. du croo de jongh and j. m. j. van leeuwen , phys . rev .
b 57 , 84948500 ( 1998 ) .
r. santos , l. sneddon , r. b. stinchcome , j. phys .
a : math . gen . 14
( 1981 ) v. karimipour , phys . rev .
b 79 , 214435 ( 2009 ) .
|
we study the phase transition from two different topological phases to the ferromagnetic phase by focusing on points of the phase transition . to this end , we present a detailed mapping from such models to the ising model in a transverse field .
such a mapping is derived by re - writing the initial hamiltonian in a new basis so that the final model in such a basis has a well - known approximated phase transition point
. specifically , we consider the toric codes and the color codes on some various lattices with ising perturbation .
our results provide a useful table to compare the robustness of the topological codes and to explicitly show that the robustness of the topological codes depends on triangulation of their underlying lattices . * on the robustness of topological quantum codes : + ising perturbation + * mohammad hossein zarei + physics department , college of sciences , shiraz university , shiraz 71454 , iran +
|
the stellar group known as chromospherically active binary stars ( cabs ) is characterized by the presence of cool components which show emissions in caii h&k , h@xmath4 , mgii h&k , cii , civ and other spectral lines .
they have also other special characteristics , such as short orbital periods and high rotational velocities ( strassmeier et al .
1993 ) , and include rs cvn binaries , which contain at least an evolved component , and by dra stars , which are still in the main sequence ( ms ) .
this last group comprises young single stars with rapid rotation and close binary stars .
this paper belongs to an ongoing project designed to study the evolution and properties of cabs . in preliminary works ( fernndez figueroa et al .
1993 ; barrado et al .
1993 ) we analyzed the li abundances of a small sample , and in paper i of this series ( barrado et al .
1994 ) we described the evolutionary status of this kind of binaries , and studied a subsample of cabs having accurate mass determinations ( e.g. the eclipsing binaries ) . using the position of each component on a radii t@xmath5 plane , we classified a group of 50 cabs in 3 different types : ms stars having masses @xmath61.7 m@xmath3 , evolved stars with masses around 1.4 m@xmath3 , and giants with masses in the range 2.55 m@xmath3 .
it is also possible to discriminate 3 different evolutionary stages in the second group : subgiants evolving off the ms , subgiants at the bottom of the red giant branch ( rgb ) , and giants ascending the rgb above the previous subgroup . using evolutionary tracks to compute the ages of these stars
, we found that the ages depend essentially on the stellar mass , and that the actual values are quite close to the corresponding ages for the terminal age main sequence ( tams ) .
we interpreted this relation in the context of the evolution of the internal structure and rotation , as an effect of the increase of the stellar radius as the component evolve off the ms , and the decrease of the rotation period due to tidal effects . on the other hand ,
since the components of cabs , ( especially close binaries ) do not spin down by magnetic breaking due to the transfer of angular momentum from the orbit to the stellar rotation , these stars have the needed ingredients to show an enhanced dynamo effect .
we study here a subsample of cabs with dwarf components .
from now on , we will use the word _ dwarf _ to denote stars of luminosity classes v and iv which are evolving off the ms , but they have not reached the rgb .
we take now into account that the abundances of light elements can change in the stellar interior as a result of different nuclear reactions . in particular ,
lithium is destroyed inside main sequence stars by the reaction @xmath1li(p,@xmath4)@xmath7he when the temperature is higher than 2.6@xmath010@xmath8 k. late - type stars have a convective envelope which deepens with decreasing stellar mass , facilitating so the transport of material from the bottom of this well mixed zone to the li burning layer . because of this , li abundances depend on mass and age for ms stars , dependence very well established by numerous studies made in open clusters and field stars ( see , for a comprehensive review , balachandran 1994 ) . for the masses considered in this study , the actual internal stellar structure models do not predict temperatures high enough to burn li at the bottom of the convection zone , and there should be additional mechanisms , beside pure convection , responsible of the transport of material below the convective envelope . the aim of this work is to help to discriminate between proposed theoretical mixing mechanisms .
we deal here with the li abundances of those components of cabs classified as ms stars and stars evolving off the ms , and we will present the study of li abundances in evolved components of cabs in a forthcoming paper ( barrado y navascus et al .
1997b , paper iii ) . following paper
i , we start by studying the evolutionary status of cabs . in this case
we have used all of the data contained in the strassmeier et al.s ( 1993 ) catalog .
we have based our discussion in the photometry of each component to classify the sample .
we describe these details and present the selected sample of dwarfs in section 2 .
the observations and data reduction process are presented in section 3 .
section 4 is devoted to the measurement of li equivalent widths ( ew ) and the derivation of li abundances .
section 5 contains a comparison between cabs and single and binary stars belonging to different open clusters .
the relations to other stellar parameters are shown in section 6 , and , finally , the summary and conclusions are presented in section 7 .
in paper i we studied the evolutionary status of cabs using a sample of binaries which have accurate values of stellar masses and radii .
we concluded that cabs are in a very specific evolutionary stage , since we found a relation between their ages and the masses of the primary stars .
our goal here is to study in detail the status of the whole sample of cabs and , in particular , to define the subsample of dwarfs components . since an important part of the systems included in the catalog by strassmeier et al .
( 1993 ) has no eclipses ( then the radii and masses are not well known or are not known at all ) , the study about the evolutionary status must be conducted in a different way than in paper i. we have used here the absolute visual magnitudes ( these systems are quite near to the sun , and most of them have measured trigonometric parallaxes ) versus the ( b v ) colors to produce figure 1 .
we have overplotted the evolutionary tracks by schaller et al .
( 1992 ) and schaerer et al .
( 1993a ) for metallicities
z=0.020 ( solid lines ) and z=0.008 ( dashed lines ) , respectively
. dwarf , subgiant and giant primaries ( namely , the brightest star in the binary system ) are shown as solid pentagons , squares and triangles , whereas dwarf , subgiant and giant secondaries are shown as open pentagons , squares and triangles , respectively . in order to be able to locate most of the components of cabs in the color magnitude diagram of figure 1
, we proceeded in different ways : for an important part of the cabs there are available values of ( b - v ) for each component , which were measured during eclipses .
in other cases , the components have equal mass or spectral type .
then , the color should be the same and the magnitudes follow the relation @xmath9=@xmath10=@xmath11 + 0.753@xmath12 . under certain conditions ( when the binary is close to the zams ) it is possible to perform a deconvolution process of the photometry , as described by barrado y navascus & stauffer ( 1996 ) , to compute individual colors and magnitudes .
in a few cases , absolute magnitudes were assigned using the spectral type .
details of this process can be found in the footnotes to table 1 . as it can be seen in figure 1
, the spectral classification does not provide clear boundaries to separate evolutionary stages of cabs .
moreover , the distinction between hot and cool components , although useful to identify the stars inside the system , is quite confusing when the evolutionary status is studied . in fact , depending on the masses of both components , the cool star can be the more massive one ( and then the most evolved one in the case of a k iii@xmath13f v binary ) or the vice versa ( the case of a k v@xmath13 m v binary ) .
therefore , we will use the terms primary and secondary to refer the more and less massive star , respectively .
when the masses are not known , the primary is defined as the most evolved star . according to figure 1 , it is possible to classify cabs in 5 different groups : * stars inside the ms .
* stars evolving off the ms or crossing the giant gap ( ms - off ) .
* stars at the bottom of the red giant branch . * those ones in the top part of the rgb .
* components in the horizontal branch ( hb ) .
this classification attends to differences in the physical structure of the stars and the mechanisms they use to produce energy .
then , the cabs in the second group ( ms - off ) have finished the hydrogen combustion in the center of the core and are expanding or developing their convective envelopes .
the difference between the third and fourth groups ( b - rgb and t - rgb ) arises from the fact that b - rgb stars are burning helium , whereas t - rgb stars could burn also other elements .
note that this classification is in excellent agreement with that one described in paper i based on the positions on the radii
t@xmath5 plane .
the age of each component of the cabs was computed using again the theoretical evolutionary tracks by schaller et al .
( 1993 ) and schaerer et al .
( 1993a , b ) .
the method for estimating ages consists of an interpolation with radii and then , with masses , because both stellar parameters are well known for an important part of stars in our subsample of cabs containing dwarf components .
as shown in paper i , there is a clear correlation between mass and age for cabs .
we have used this characteristic to assign the ages of those cabs of our sample which have known values of their masses but their radii have not been measured .
the actual expression is : @xmath14 an illustration of this relation can be seen in figure 3 of paper i. the basic data namely , photometry , stellar masses , radii , distances , orbital and rotational periods , etc were selected from the catalog of cabs ( strassmeier et al .
1993 and references therein ) .
table 1 shows the systems which have been studied in this paper : column 1 lists the name of the system , column 2 the hd number , columns 3 , 4 , 5 and 6 _ our _ assignation of the ( u b ) , ( b v ) , ( v r)@xmath15 and ( v i)@xmath15 colors , respectively , for each component .
column 7 lists the distance whereas the 8th , and 9th columns contain the apparent visual magnitude and the absolute magnitude computed with the previous 2 columns . in this last case
, we also took into account other information in order to obtain the most accurate estimation of the magnitude of each component in the binary system ( see notes in table 1 ) .
finally , column 10 shows our evolutionary classification ( ms or ms - off star ) , and column 11 provides the spectral type .
the last column contains notes about the method used to compute the absolute visual magnitude and the photometric indices for each component . in total ,
our sample has 41 systems , which contain 62 dwarf components .
as far as we know , our subsample does not have any bias with respect to the whole sample of cabs .
the spectra of the stars analyzed here were collected during three different observational runs .
the two first ones were carried out at the coud focus of the 2.2 m telescope of the calar alto spanish
german observatory ( caha ) in june 1993 and may 1994 , respectively . the last run was carried out at the nordic optical telescope ( not , 2.56 m ) of the spanish roque de los muchachos observatory ( la palma ) in september 1994 .
for the first observational campaign we used the boller and chivens spectrograph equipped with a frd ccd ( 1024@xmath01024 pixels , 13@xmath16m / pixel ) at the f/3 camera and a gec ccd ( 1155@xmath0768 pixels , 22.5 @xmath16m / pixel ) at the f/12 camera .
a slit of 1 arcsec was used , and the spectral resolution achieved ( measured from the fwhm of th ar emission lines spectra ) was r@xmath1730@xmath18000 in both cases . for the second campaign we selected the rca ccd at the f/3 focus , ( 1024@xmath01024 , 15 @xmath16m / pixel , r@xmath1730@xmath19 ) and the tek ccd at the f/12 focus ( 1024@xmath01024 , 24 @xmath16m / pixel , r@xmath1750@xmath19 ) . finally , the observations at the not were performed using the iacub echelle spectrograph ( mckeith et al .
1993 ) at the cassegrain focus of the telescope , with a thomson @xmath20 ccd detector .
a final resolution of about 40@xmath18000 was obtained .
exposure times were estimated in order obtain signal to noise ( s / n ) ratios in the range 100160 .
since the visual magnitudes of our objects are very different ( 1@xmath12@xmath21v@xmath2111@xmath12 ) , we took exposures from 5 s up to 3600 s. in order to avoid the presence of cosmic rays , several exposures were taken for the longest exposures until the required s / n ratio was achieved , adding them at the end of the reduction process .
although radial velocities are important in some cases due to the fact that the orbital periods are short , the exposure times were not long enough to produce an important doppler broadening of the lines , since in those cases the rotation is itself quite important and is by far the main source of the broadening .
we reduced the data using standard procedures ( bias subtraction , flat field correction , extraction of one - dimensional spectra and continuum normalization ) and the midas package . for the caha data we transformed the pixel scale into wavelength scale by comparison with th ar lamp spectra .
for the not data , we calibrated the spectra by comparison with photospheric lines of the solar spectrum taken by observing the moon with the same instrumentation .
the wavelength calibration is extremely important , since a considerable part of our binaries are double lined systems and the theoretical positions of the li features of both stars were computed previously to compare with the observed wavelengths .
the @xmath1lii 6707.8 feature is in fact a spectroscopic doublet separated by 0.15 ( 6707.76 and 6707.91 , andersen et al .
1984 ) . with the resolutions of our spectra
, the doublet appears as a single line .
moreover , since most of our stars are rapid rotators , the li features are blended with the fei 6707.4 line . on the other hand ,
28 out of 41 systems of our sample are sb2 systems ( double lined systems ) , and the lii@xmath13fe complex can also be blended with other spectral features arising from the other component . in order to measure the li equivalent width ( ew ) , we have fitted gaussian profiles to the spectra .
phases were accounted for during this process . using the published ephemerides ( strassmeier et al .
1993 and references therein ) , we computed the positions of the lines arising from each component and used this information when fitting and measuring the equivalent widths .
different situations appeared when we were performing the fitting procedure : * sb1 systems with vsini@xmath2223 kms@xmath23 .
the blends between lii and fei lines were weak and we were able to separate the fe contribution with the fitting procedure .
we used the comparison between the measured ew(fei 6707.4 ) with empirical values obtained from ( b v ) color indices ( soderblom et al .
1990 ) as a quality control of the procedure .
we also obtained new empirical relations between ew(fei 6707.4 ) and the equivalent widths of other spectral lines such as fei 6703.6 & fei 6705.1 and different colors .
* sb1 systems with vsini@xmath2423 kms@xmath23 .
we measured ew(li+fe ) and eliminated the fe contribution taking into account the ew of stars with the same ( b v ) color , and the empirical fits described above between ew(fei 6707.4 ) and the ews of other iron lines . *
sb2 systems where we measured in the spectrum @xmath25v@xmath26@xmath2423 kms@xmath23 ( where the difference of radial velocities between components is defined as @xmath25v@xmath26 = @xmath27v@xmath28 v@xmath29 ) and no blend with other close lines from the other component . the li features from both components appeared separated .
we proceeded like in the first case .
* sb2 systems with multiple blends and narrow lines .
we fitted as many gaussian curves as necessary to obtain the ew(li ) .
* sb2 systems with multiple blends and wide lines .
we measured the whole ew and tried to take out the different contributions with empirical relations between lines and colors .
some examples of these fitting procedures can be found in figure 2 . on the other hand ,
the continuum location is very important .
there are uncertainties due to blends with other lines and the noise of the spectra . as a rule
, we located the continuum in the middle of the noise distribution for each wavelength .
since most of our spectra have a secondary component bright enough to contribute to the continuum of the primary star , a correction of the measured ew is necessary .
the continuum correction factors ccf were calculated using the equation : @xmath30 where @xmath4 is defined ( boesgaard & tripicco 1986b ) as : @xmath31 or , in the cases the radii ( @xmath32 , @xmath33 ) were not available , we used the relation between the fluxes derived from the apparent magnitude of the primary and secondary ( @xmath34 , @xmath35 ) , defined as : @xmath36 the differences between both methods are irrelevant .
then , the final ews were obtained using : @xmath37 the results are shown in table 2 . in columns 1 and 2 we list the name of the star and its hd number , while columns 3 and 4 contain the parameters effective temperature and radius needed to compute the continuum correction factors ( ccf ) , listed in column 8 . the method to compute ccf
is indicated in column 9 .
columns 5 , 6 and 7 provide the measured equivalent widths of li i 6707.8 plus fei 6707.4 , and only the iron line or the lithium doublet , respectively . in these two last cases ,
the values are not corrected by the effect of the continuum due to the presence of a companion star .
we have used different photometric temperature scales to compute the final effective temperatures .
since most of our stars have available ( b v ) colors , this index has been our basic temperature indicator .
we estimated these temperatures with the scale provided by thorburn et al .
( 1993 ) . however , for the coolest stars in our sample , we selected scales based on ( v i)@xmath15 ( cayrel at al . 1985 ) and ( r i)@xmath15 ( carney 1983 ) , and the final t@xmath5 was computed as the average of these values . for some cases , photometry is not available and we have used only the spectral type to compute the effective temperature ( schmidt kaler 1982 ) .
the standar deviation of the differences between temperatures obtained from color indices and from the spectral types , for those stars having both , is 320 k. we verified that this procedure does not introduce any bias in the final temperatures and abundances .
in fact , we computed , when possible , the different temperatures for the stars in our sample and compared them . except for stars having temperatures less than 4500 k , there are neither significant differences ( the average value is 80 k ) not any clear trend .
the differences in the lower range of temperatures are easy to understand since the thorburn et al .
( 1993 ) temperature scale was obtained using stars with higher temperatures . in order to estimate the li abundances
, we used the pallavicini et al .
( 1987 ) curves of growth ( cog ) , valid for 4500 @xmath21 t@xmath5 @xmath21 6500 k. in those cases where the effective temperature was lower , we used the cog computed by soderblom et al .
all these cog were computed under lte conditions and a gravity value of log g=4.5 .
systematic differences between the results obtained with both sets of cog were avoided by modifying the last group following the technique described by duncan & jones ( 1983 ) .
this consists in computing the differences between the curves obtained for a given temperature ( in our case , 4500 k ) and shifting a whole set with these values .
then , the positions of both curves at 4500 k would overlap .
the maximum value of the differences , in the linear part of the curves at any temperature , is 0.1 dex .
in addition to our data , we have included the available measurements present in the literature .
specifically , we have re analyzed the data by pallavicini et al .
( 1992 ) , fernndez
figueroa et al . ( 1993 ) , and randich et al . (
1993 ) using similar procedures .
the final corrected ew(li ) and abundances are listed in colums 4 and 5 of table 3 .
this table also contains information about the stellar masses , orbital and photometric periods , and ages ( computed by interpolating radii and masses with isochrones or from the relation between ages and masses for cabs , see eq . 1 in section 2 ) .
in general , spectral synthesis provides more accurate abundances because it allows to remove different effects , such as the blends with other lines at similar wavelength .
however , in the specific case of cabs , randich et al . (
1993 ) have shown that there are no important differences between abundances derived using cog and those obtained using synthetic spectra techniques .
other studies show similar results ( eg .
barrado y navascus et al .
moreover , under special circumstances it is not possible to compute a reliable synthetic spectrum for some binaries because there is not accurate information .
since the goal of this paper is to study the evolution of the li abundance in a large sample of dwarf components of cabs , it is reasonable to use cog to derive li abundances . on the other hand , as mentioned above , the use of cog allows us to include in our analysis the ew(li ) provided in the literature .
abundances were derived assuming lte conditions .
there are two ways in which the abundances could be affected by stellar activity _ per se_. on one hand , there could be significant nlte effects due to the presence of the chromosphere .
as shown by houdebine & doyle ( 1995 ) , the inward uv and optical flux can reheat the upper atmosphere and modify the continuum level .
therefore , the observed ew(li ) would be smaller than that of a non active star for the same abundance , metallicity and mass
. this effect would be relevant when the h@xmath4 line appears in emission and when the abundances are high , but none of our stars presents both characteristics simultaneously .
note that this effect is opposed to the overabundances we have found ( see next subsection ) .
then , we consider that the excesses in the li abundances are real , and the nlte is a second order effect in the li abundance determination for our stars . moreover , pavlenko et al .
( 1995 ) have shown that the differences between li abundances calculated under lte and nlte conditions for stars with t@xmath38 k are significant only for very strong lines .
on the other hand , stellar spots could modify the li abundance determinations .
there are two associated effects : first , the spots can change the photometric indices and , therefore , the inferred t@xmath5 which will be used to compute the abundances .
second , the ew(li ) is larger in spots than in the quiet photosphere , as shown by solar observations ( giampapa 1984 ) .
however , these effects will not modify the abundances more than @xmath170.30 dex altogether in the worst case ( barrado y navascus 1996 ) .
in addition to the previously described phenomenon associated with the presence of stellar spots , there are other different sources of errors in the li abundance determinations , such as the ew measurement process , the effective temperature scale adopted , the computed continuum correction factors , and the results provided by different curves of growth .
we have estimated the errors in our sample as follows : for two typical stars with 6000 and 5000 k , we have corrected ew(li)s of @xmath1750 and @xmath1720 m , respectively .
the error in these measurements is @xmath1720% .
these numbers are translated into errors in the abundances equivalent to 0.16 and 0.13 dex in each case . on the other hand
, the most important source of error appears in the temperature determination .
the average error in the adopted effective temperatures is 200 k , and it produces uncertainties of 0.22 and 0.24 dex , respectively , in the abundances for both temperatures .
the uncertainty in the continuum correction factors can be estimated in 10% , introducing additional errors of 0.06 and 0.05 dex . finally , the errors of lte curves of growth add another @xmath170.13 dex ( pallavicini et al .
the final errors can be computed using the expression : @xmath39 where @xmath25log n(li)@xmath40 are the individual errors described above . for both temperatures ,
the final value is 0.31 dex .
if the possible presence of spots is taken into account , it would lead to a maximum error of 0.43 dex . as noted by randich et al .
( 1993 ) , although these errors can be very important for individual stars , they cancel each other and do not affect to the conclusions reached when studying the whole sample of stars . as shown by cayrel et al .
( 1984 ) , balachandran et al . (
1988 ) , garca lpez et al .
( 1994 ) , and several other works , li abundances of stars belonging to open clusters are strongly related to the stellar mass or effective temperature .
figure 3a shows the li abundances of our sample of dwarf components of cabs against t@xmath5 solid circles. for comparison purposes , we have also included members of the pleiades ( 70120 myr , meynet 1993 ) and hyades ( 600800 myr ; mermilliod 1981 , gilroy 1989 ) open clusters .
abundances from pleiades were selected from butler et al .
( 1987 ) , pilachowski & hobbs ( 1987 ) , boesgaard et al .
( 1988 ) , soderblom et al .
( 1993 ) and garca lpez et al .
( 1994 ) , whereas those from hyades stars were obtained from boesgaard & tripicco ( 1986a ) , rebolo & beckman ( 1988 ) , thorburn et al .
( 1993 ) and barrado y navascus & stauffer ( 1996 ) .
pleiades stars are plotted as open triangles , whereas hyades stars are included as open circles .
the average behavior of single members of this cluster is shown as a solid line .
a first inspection of this diagram shows that the li abundance of cabs depends strongly on t@xmath5 , as happens in open cluster stars .
the estimated abundances for cabs are in some cases similar to those present in pleiades stars .
it could be concluded then that these cabs have pleiades age .
other cabs , having abundances smaller than the hyades , would be older than 600800 myr .
moreover , it would be possible to constrain the age of the cooler stars in our sample ( 4500 @xmath41 t@xmath5 @xmath41 3000 ) between 70 and 800 myr . however , this interpretation does not seem to be correct , for two reasons : first , it is a well known fact that low mass cabs belong to the old disk population ( eker 1992 ) , and , therefore , their ages are between 10@xmath42 and 5@xmath010@xmath42 yr , depending on the velocity dispersion .
second , the analysis presented in paper i shows that the ages of these systems are much older than the hyades , at least for those stars having masses larger than 0.8 m@xmath3 .
on the other hand , the scatter in the abundances of cabs is as large as that present in pleiades , and much larger than that corresponding to hyades stars .
the scatter observed in late - type stars of the pleiades has been associated with the effect that rotation can have in the li depletion phenomenon ( garca lpez et al .
young rapid rotators would inhibit the li depletion during their premain sequence phase ( pms ) due to the effect of rotation on the internal structure ( martn & claret 1996 ) . during the ms phase
, the scatter could be due to the lack of transfer of angular momentum and material between the envelope and the core ( pinsonneault et al .
it could also be very important the role of magnetic fields in the inhibition of transport of material to the stellar interior ( spruit 1987 ; schatzman 1993 ) .
part of the scatter in the abundances observed in the log n(li)t@xmath5 plot for cabs is due to a dependence on the stellar mass .
since an important part of our sample are quite evolved objects , and in fact in several cases they are leaving the main sequence , it is possible to have several cabs of different masses at the same temperature , as it is shown in figure 3b .
we have also included in the figure the sample of nearby dwarfs studied by favata et al .
( 1996 ) , and the li depletion isochrones by chaboyer ( 1993 ; from top to bottom , for 50 , 70 , 300 , 800 and 4500 myr . ) . as noticed by favata et al .
( 1996 ) , the observed dependence of li on the temperature is not real , it is introduced by the conversion on the ew upper limits into abundances .
the comparison between these two samples of old dwarfs allows us to determine that cabs have undergone a process of inhibition of their li depletion .
this phenomenon was previously established by randich and pallavicini ( 1991 ) and pallavicini et al .
( 1992 ) , using a sample of cabs which contained essentially giant stars .
our study , which makes use of dwarfs and therefore allows direct comparisons with open clusters , clearly confirms and supports this result .
a more extensive discussion will be carried out in sections 5 and 6 .
as noticed above , figures 3a , b are affected by the evolution of some stellar parameters such as temperature when a star approaches or crosses over the tams . to avoid this and the possible consequences on the interpretation of the abundances ,
we have compared them against stellar masses ( figure 4 ) . in the particular cases of cabs ,
these values have been computed directly in most of cases because there are available accurate orbits for both components , with uncertainties below @xmath170.05 m@xmath3 . for a few binaries , we estimated the masses using the m@xmath0sin@xmath43i values ( obtained from the radial velocities curves ) and the inclination of the orbit ( from the rotational periods and vsini ) , the position in color magnitude diagrams or the radii t@xmath5 plane ( see footnotes in table 3 ) .
the masses of cluster stars were estimated from isochrone fitting , following balachandran ( 1995 ) .
figure 4a compares the abundances against stellar masses for our sample of cabs and hyades binaries .
note that we have differentiated tidally locked binary systems ( tlbs ) from well separated systems ( barrado y navascus & stauffer 1996 ) .
in fact , some of the hyades tlbs have been catalogued as cabs by strassmeier et al .
barrado y navascus & stauffer ( 1996 ) have studied the li abundances of tlbs in the hyades and concluded that they show clear overabundances when the orbital period is less than 8 days , in excellent agreement with the theoretical predictions by zahn ( 1994 ) . in our case
, we have a larger sample of stars , which are older than those and have p@xmath44 ranging from less than 1 day to more than one hundred . since cabs are , in general , older than hyades stars , they have had more time to synchronize their orbital and rotational periods .
therefore , binaries with larger p@xmath44 are tlbs . as in the hyades , there is a relation between li abundances and p@xmath44 , in the sense that , for the same mass , ( e.g. @xmath170.8 m@xmath3 ) , li tends to be higher in the case of tlbs than in the case of non - tlbs . due to the fact that tlbs with p@xmath458 d have achieved synchronization during their ms life time ( zahn 1994 ) , the li inhibition should have taken place , at least in part , during this stage of the evolution . in figure 4a
, cabs show some differences with hyades binaries : first , an important part of cabs have larger abundances than hyades stars of similar mass , despite the fact that the studied cabs except 42 cap have the same age than hyades or they are older .
second , the li abundance scatter is much larger , and in the particular case of @xmath170.8 m@xmath3 li abundances change by a factor 16000 .
the abundance spread for a given mass in an old cluster is never larger that @xmath171 dex .
since cabs of equal masses have similar ages , the observed scatter should be related to other parameters .
the presence of excesses in the li abundances of at least some of our binaries at @xmath170.8 m@xmath3 can be related to the results found by garca lpez et al .
( 1994 ) for stars with similar masses in the pleiades .
they showed that the abundances of stars having this mass are quite sensitive to rotation , and that the fast rotators deplete less li than the slow ones . on the other hand
, it is worth to notice the detection of li in very low mass stars , in particular yy gem and v1396 cyg .
however , these abundances should be taken with some caveats since at these low temperatures the spectrum of late type stars presents multiple lines , together tio and other molecular bands which were not taken into account to measure ew(li+fe ) .
it is possible to conclude from figure 4a that the li depletion is also less pronounced in cabs than the hyades for those stars located in the mid f dip ( boesgaard & tripicco 1986a ) , despite that the cabs are more evolved and should have started to mix the external material with material poor in li from inside , in the so called dilution process ( iben 1965 ) .
figures 4b , c , d present a comparison between the abundances of cabs with members of the ngc752 , m67 and ngc188 open clusters , which have ages corresponding to 1.72.0@xmath010@xmath42 ( hobbs & pilachowski 1986a ; meynet et al .
1993 ) , 3.0 - 6.0@xmath010@xmath42 ( vandenberg 1985 ; hobbs & pilachowski 1986b ; spite et al .
1987 ; garca lpez et al .
1988 ; montgomery et al . 1993 ; meynet et al , 1993 ; balachandran 1995 ; barrado y navascus et al .
1997c ) , and 6.510.0@xmath010@xmath42 yr ( vandenberg 1985 ; meynet et al .
1993 ; hobbs & pilachowski 1986b ) , respectively . in these figures ,
open cluster stars are whown as open symbols , cabs having similar ages to the clusters appear as solid squares , and other cabs are overplotted as solid circles .
these figures show , as in figure 3 , that the li abundance scatter could be attributed to an age spread .
however , we have shown in paper i that dwarf cabs are near the tams , and therefore they are much older , in most of cases , than these clusters .
it is very interesting to perform a comparison using different mass intervals . except the cabs belonging to the hyades , which are outside of the validity of the mass - age relationship for cabs due to their low mass (
the relation was computed for stars having masses larger than 0.8 m@xmath3 ) , all cabs classified as ms are older than ngc752 ( see figure 4b ) . only some cabs located in figure 1 above the ms ( ms - off ) have ages comparable to this cluster .
these stars have masses in the range of the mid f dip , and should have severely depleted their li due to one of the mechanisms that work in that region ( see balachandran 1995 and references therein ) and also should have started to dilute its abundance due to the mixing of material as the star develops a deep convective envelope .
on the other hand , the comparison of masses @xmath171.4 - 1.2 m@xmath3 , corresponding to the cool side of the li dip , in figure 4c , between the cabs and m67 shows some clues about the li depletion phenomenon when binaries start to evolve off the ms . in this figure ,
binaries and single stars of m67 are represented as open triangles and squares , respectively .
as it can be seen , there is not a clear correlation between binarity and li abundances in m67 , despite the fact that in average they have higher abundances than single stars .
a more detailed study about the li abundances in binaries of m67 can be found in barrado y navascus et al .
( 1997c ) . in this mass range , which contains few cabs , only one shows a clear overabundance .
therefore , it seems that li depletion is not affected in a large extent for this mass range .
however , the li excesses are more conspicuous in the case of cabs having smaller masses , since they are older than m67 .
they present the li doublet in their spectra , so it is possible to derive abundances , when most of the m67 stars have only upper limits due to the lack of this feature in their spectrum .
finally , figure 4d contains data of the oldest open cluster considered : ngc188 .
the conclusion is the same one as above , dwarf cabs have abundances larger than the values corresponding to their age .
these results agree with the study by spite et al .
they found that close binaries belonging to the old disk population or the galactic halo have systematicaly larger abundances than binaries with larger orbital periods or single stars .
other excesses for the abundances of dwarf binaries have been found by ryan & deliyannis ( 1995 ) . another way to demonstrate the excesses in the li abundances of cabs
is shown in figure 5 . here
, the ages are plotted against the average li abundances for some mass ranges .
symbol sizes increase with stellar mass ( see key ) .
error bars ( 0.17 dex for the age ) are also shown .
we also have included the average abundances for different ages , computed from cluster stars in three mass ranges ( 0.85 m@xmath3 @xmath21 mass @xmath6 0.95 m@xmath3 , 0.95 m@xmath3 @xmath21 mass @xmath6 1.05 m@xmath3 , 1.05 m@xmath3 @xmath21 mass @xmath6 1.15 m@xmath3 ) .
these average values were obtained using li abundances derived in single stars which belong to open clusters , such as pleiades , uma group , coma , hyades , ngc 752 , m67 and ngc 188 .
we have carried out several fits of the type log n(li ) = a + b log age , one for each mass interval . despite their age
, cabs have high li abundances . in some cases
much higher than those ones characteristic of single stars at the same age .
the difference becomes extreme when the comparison is made considering stars with m @xmath6 0.95 m@xmath3 , corresponding to the two smaller circle sizes and the two bottom fits .
then , we conclude that the li depletion rate is completely different , and slower , in cabs that in single stars .
one possible explanation for the high li abundances present on the surfaces of cabs is the creation of li via spallation reactions during very active episodes , in particular flares .
in fact , pallavicini et al .
( 1992 ) and randich et al .
( 1993 ) tried to use this hypothesis to explain the overabundances that they found in their samples of cabs , essentially composed by giant stars .
several checks of this hypothesis have been carried out , since the creation of li would imply a specific ratio between the isotopes @xmath8li and @xmath1li .
also , energetic criteria seem to reject the possibility of creation of li at large scale .
however , @xmath8li has been detected in hd84937 ( smith et al .
1993 , hobbs & thorburn 1994 ) , an f - type metal - poor dwarf , and the calculations performed for population ii stars by deliyannis & malaney ( 1995 ) indicate that it is possible to create li in enough amount during the flares to substitute the depleted material , but , in principle , this mechanism is restricted to dwarfs and subgiants which have very shallow convective envelopes .
the creation of li during flares should produce a relation between the li abundances and some activity indicators such as h@xmath4 , caii h&k , the caii infrared triplet , mgii h&k , etc .
we have tried to find out if there are such relations for our sample of cabs . from the available data in the literature ,
we have found no obvious relation between the x - ray luminosity ( dempsey et al .
1993 ) , or h@xmath4 ( montes et al .
the lack of relations would indicate that there is no relevant li creation during flares .
however , there is a relation between li abundances and the surface fluxes in caii h&k , as provided by fernndez
figueroa et al . ( 1994 ) and montes et al . ( 1995 ) .
this relation only shows up when specific ranges of masses are selected . in figure 6
we have plotted the li abundances against caii h&k fluxes for stars having masses lower than 1.25 m@xmath3 and masses in the range 3.0 @xmath21 m @xmath21 5.0 m@xmath3 .
( we have included data corresponding to giant components of cabs in this figure , see paper iii . )
an obvious interpretation of this phenomenon would be that the presence of stellar spots , due to the high activity rates , could modify the measured li equivalent widths , since the lii 6707.8 feature is very sensitive to variations in the temperature ( almost all li is ionized in the considered range of temperature ) .
therefore , a stellar surface covered in a large proportion by spots would have a li equivalent width larger than that one corresponding to the quiet photosphere , for a given value of the abundance .
if this effect would not be corrected , the computed abundance would be larger than the real value .
several works have tried to estimate the importance of this effect .
soderblom et al . (
1993 ) performed different calculations to see if the observed spread in the log n(li)-t@xmath5 plane for the pleiades cluster could be attributed to this situation .
they obtained that only an extreme situation ( when the fraction of the surface covered by spots filling factor were very high ) could account for the spread .
moreover , barrado y navascus ( 1996 ) , using solar observations of the lii 6707.8 , ki 7699 and nai 5896 lines in different solar regions , computed a grid of simulations under different conditions ( spots of different temperatures , presence of plages , different filling factors for the spots and the plages ) .
he found that normal situations for cabs ( filling factor up to 0.30 ) could only introduce maximum errors smaller than 0.30 dex in the li abundances .
a different strategy was adopted by pallavicini et al .
they studied carefully four active stars , trying to find correlations between the photometric variability and changes in the shape and the equivalent widths of the li feature .
they found no proves of changes in the li related to surface inhomogeneities .
all these evidences , together with the fact that we have not measured abundances larger than the so called `` cosmic abundance '' , which corresponds to the abundance observed in young objects which have not undergone any depletion process ( log n(li)=3.2 , martn et al .
1994 ) , allow us to conclude that , with some caveats , there is no significant creation of li , if any at all , on the atmosphere of cabs , and that the surface inhomogeneities do not affect essentially the computed abundances .
our interpretation of the relation between li abundances and caii h&k fluxes goes in other direction : the synchronization between the rotational and the orbital periods .
this peculiarity would affect the transport mechanism in the stellar interior , inhibiting the depletion of li .
figure 4a shows that tlbs have , in average , larger abundances than other cabs without coupling between the orbital and rotational periods .
the stars undergoing a magnetic breaking by stellar winds would transfer angular momentum from the convective envelope to the radiative core to avoid intense radial differential rotations , and this angular momentum transfer is accompanied by mixing of material in the stellar interior ( pinsonneault et al .
1989 , 1990 , 1992 ; chaboyer et al .
if magnetic breaking is inhibitted in tlbs , similar stars in different systems deplete li in a different amount , because they have different orbital periods and the orbit works as a source of angular momentum for the spin . on the other hand , the rossby number ( the relation between the rotational period and the turnover time of the convective cells , n@xmath46 ) is strongly related to different activity indicators ( noyes et al .
then , the relation between caii h&k fluxes and li abundances in cabs is a result of the inhibition of li depletion caused by high rotation ( due to tidal forces present in cabs ) , while increasing the stellar activity .
there is an alternative explanation for the relation between activity and li abundance and for the high abundances found in cabs .
this explanation relies in the mixing due to internal gravity waves generated by the pressure fluctuation of the convective cells at the bottom of convective envelope on the stellar interior ( garca lpez & spruit 1991 ; montalbn 1994 ) .
montalbn ( 1994 ) shows how this mechanism produces a macroscopic diffusion , which transports li ( and other elements ) toward the internal layers , where it is destroyed . on the other hand , strong magnetic fields in the bottom of the convective envelope , which are responsible of the stellar activity ,
can inhibit the generation of gravity waves ( schatzman 1993 ) . in a more general way
, the presence of a strong magnetic field could also inhibit turbulent mixing mechanisms between the base of the convection zone and the stellar interior ( spruit 1987 ) .
this would explain the link between activity and li abundance and also the high abundances obtained in cabs .
the actual data can not discriminate which mechanism is actually working .
only a detailed modeling of the evolution of li abundances and other stellar properties will help to find the answer .
we have presented a study of li abundances in dwarf components of chromospherically active binary stars .
the accuracy in the measurements of masses and radii for an important part of the binaries studied here , suggests the use of these data to check different theories about the li depletion phenomenon .
since cabs as a group contains binaries in very different evolutionary stages , we first classified the totality of binaries listed in the catalog compiled by strassmeier et al .
this classification was based on the position of each component in a color magnitude diagram .
we selected stars located in the ms and those beyond the terminal age main sequence , but only if they have not reached the bottom of the red giant branch .
the study of the li abundances in our sample shows that these stars present abundances higher than the normal values characteristic of stars of the same mass and age .
this conclusion was reached by comparing our cabs , whose ages were derived by isochrone - fitting with masses and radii , with open cluster stars .
the comparison between stars in the range 1.21.4 m@xmath3 the cool side of the li gap belonging to the m67 and the cabs , of similar ages , shows that li depletion is not very sensitive to binarity in this mass range .
however , notorious overabundances appear for cabs with 0.75 @xmath21 mass @xmath21 0.95 m@xmath3 .
we have also confirmed the presence of a trend between the surface flux in caii h&k and the li abundance for cabs in specific mass ranges .
we have interpreted this relation and the li excesses in the context of the transport of angular momentum from the orbit to the spin , due to a tidal effect .
this transport produces high rotational rates in these old stars ( they do not spin down as single dwarfs do ) .
because they are , essentially , near the tams , they have deep convective envelopes and , hence , low rossby numbers ( a very enhanced activity ) . on the other hand , the interchange of angular momentum also avoids the radial differential rotation between the convective envelope and the radiative core . therefore ,
the turbulent mixing which may appear due to this effect in single stars , leading to li depletion , would be inhibited and li would be preserved till certain extent . in short , active stars would have less differential rotation and more li photospheric abundance than less active stars , explaining the relation between li and caii h&k emission .
however , the data do not rule out other explanations such as the transport induced by internal gravity waves and its inhibition due to strong magnetic fields present at the bottom of the convective envelope . @xmath18
+ cm.- mass estimated from the color magnitud diagram .
porb.- photometric preiod similar to the orbital value .
m.- age estimated from the mass age realtionship .
p.- age of the secondary equal to the computed age for the primary .
[email protected] mass computed from m@xmath0sin@xmath43i .
rt.- mass estimated from radii t@xmath5 plane .
figure 1.- color magnitude diagram for the totality of cabs listed in strassmeier et al .
this figure allows the classification of these binaries in five different groups .
symbols for the primary and secondary components are labeled with p and s , respectively .
figure 3.- lithium abundance against effective temperature . * a * in this panel , we show our dwarf components of cabs as solid circles , members of hyades as open circles and stars belonging to pleiades as open triangles .
the average abundance for single hyades stars is shown as a solid line .
* b * comparison between the abundances of li of cabs circles , the sample of nearby dwarfs by favata et al .
1995 crosses for actual data and triangles for upper limits and the lithium depletion isochrones by chaboyer ( 1993 ) for 50 , 70 , 300 , 600 and 1700 myr solid lines .
figure 4.- lithium abundance against stellar mass . *
a * tlbs cabs are shown as solid triangles , whereas those cabs without the coupling are represented as solid circles .
those hyades binaries with synchronizarion are shown as open squares .
the average li abundances of hyades stars are shown as a solid line .
this figure eliminates the evolutionary effect on the effective temperature . *
b * comparison with ngc752 .
* c * comparison with m67 . *
d * comparison with ngc188 .
figure 5.- li abundance against stellar age .
sizes increase with increasing stellar mass .
fits for single stars belonging to open clusters are shown as solid lines ( @xmath171.1 , 1.0 and 0.9 m@xmath48 ) .
error bars are also shown .
|
we present an extensive study of lithium abundances in dwarf components of chromospherically active binary stars ( cabs ) .
since most of these binaries have known radii , masses and ages , this kind of data is especially useful for comparisons with theoretical models which try to explain the li depletion phenomenon .
we show that a significant part of these stars have clear li overabundances with respect to the typical values for stars of the same mass and evolutionary stage .
these excesses are evident when comparing our sample of cabs with binary and single stars belonging to open clusters of different ages , namely pleiades , hyades , ngc752 , m67 and ngc188 , which have ages ranging from 7@xmath010@xmath1 to 10@xmath2 yr .
the li excesses are more conspicuous for masses in the range 0.750.95 m@xmath3 , indicating that the rate of li depletion has been less pronounced in cabs than in single stars .
this phenomenon is interpreted in the context of transport of angular momentum from the orbit to the stellar rotation due to tidal effects .
this angular momentum transfer would avoid the radial differential rotation and the associated turbulent mixing of material in the stellar interior .
other explanations , however , can not be ruled out .
this is the case of transport of material induced by internal gravity waves , which could be inhibited due to the presence of strong magnetic fields associated with the effective dynamo in cabs .
the confirmed existence of a relation between li abundances and the fluxes in caii h&k lines can also be accommodated within both scenarios .
|
the ising model was one of the favorite models of a. n. vassiliev .
his studies were concentrated on the scaling properties of this model @xcite .
it is remarkable that in a special limit of the anisotropy parameter tending to infinity , the hamiltonian of the spin_-1/2 _ heisenberg @xmath0 chain is transformed into the hamiltonian of one - dimensional ising model @xcite .
this limit is usually called as the _ ising limit _ of the @xmath0 chain @xcite . in our paper
we shall investigate the ising limit of the @xmath0 magnet using an effective hamiltonian which is commutative with the ising hamiltonian and has a complete system of the eigen - functions which is common with that of the ising hamiltonian .
this effective hamiltonian is an appropriate object for studying of the ising limit since it is related to the transfer - matrix of the four - vertex model , which arises in the limit of infinite anisotropy from the transfer - matrix of the six - vertex model . in their turn ,
the trace identities relate the @xmath0 hamiltonian and the transfer - matrix of the six - vertex model . the problem of calculation of the correlation functions of @xmath0 heisenberg model in the framework of the quantum inverse scattering method @xcite has required serious efforts @xcite .
we shall show that calculation of the correlation functions in the ising limit is closely related to the theory of the symmetric functions @xcite , and thus is closely connected with the problems of enumerative combinatorics , such as the random walks of vicious walkers @xcite and enumeration of boxed plane partitions @xcite . in this paper
we shall calculate the thermal correlation functions of the ferromagnetic string over the ground state using the approach developed in @xcite .
the @xmath0 spin-_1/2 _ heisenberg magnet is defined on the one - dimensional lattice consisting of @xmath1 sites labeled by elements of the set @xmath2 , @xmath3 .
the corresponding spin hamiltonian is defined as follows : @xmath4 where the parameter @xmath5 is the anisotropy of the model , and @xmath6 is the external magnetic field .
the local spin operators @xmath7 and @xmath8 , dependent on the lattice argument @xmath9 , are defined standardly ( see @xcite ) , and their commutation rules are given by the relations : @xmath10\,=\,\dl_{k , l}\,\si^z_l$ ] , @xmath11\,=\,\pm 2\,\dl_{k , l}\,\si^{\pm}_l $ ] .
the state - space is spanned over of the state - vectors @xmath12 , where @xmath13 implies either @xmath14 or @xmath15 .
the spin `` up '' and `` down '' states ( @xmath16 and @xmath17 , respectively ) provide a natural basis in the linear space @xmath18 so that @xmath19 the periodic boundary conditions @xmath20 are imposed .
the hamiltonian ( [ xxzham ] ) commutes with the operator of third component of the total spin @xmath21 : @xmath22\,=\,0\,,\qquad { \widehat s}^z\equiv \frac 12\sum_{k=0}^m\sigma_k^z\ , .
\label{spi}\ ] ] to represent @xmath23-particle state - vectors , @xmath24 , let the sites with spin `` down '' states be labeled by the coordinates @xmath25 forming a strict partition @xmath26 , where @xmath27 .
there is a correspondence between each partition and an appropriate sequence of zeros and unities of the form : @xmath28 , where @xmath29 , @xmath30 .
the hamiltonian ( [ xxzham ] ) is diagonalised _ via _ the following ansatz : @xmath31 where the sum is taken over @xmath32 strict partitions @xmath33 , and @xmath34 is the fully polarized state with all spins `` up '' .
it is proposed to use bold - faced letters as short - hand notations for appropriate @xmath23-tuples of numbers : for instance , @xmath35 instead of @xmath36 , etc .
therefore , the wave function @xmath37 in ( [ bwf ] ) is of the form : @xmath38 \displaystyle{\ca_s ( { \textbf u})\,\equiv\,\prod_{1\le j <
i\le n}\frac{1 - 2\dl u_{p_i}^2 + u_{p_i}^2 u_{p_j}^2}{u_{p_i}^2 - u_{p_j}^2}}\ , , \label{ampxxz}\end{aligned}\ ] ] where summation goes over all elements of the symmetric group @xmath39 the state - vectors ( [ bwf ] ) are the eigen - states of the hamiltonian ( [ xxzham ] ) , if and only if the variables @xmath40 satisfy the _ bethe equations _ @xmath41 the corresponding eigen - energies are given by @xmath42 the limit of zero anisotropy , @xmath43 , transforms ( [ xxzham ] ) to the hamiltonian of the @xmath44 model on the periodic chain @xmath45 this limit is known as the free fermion limit of the @xmath0 chain , and it is the most studied one @xcite . in the next section
ii we shall consider the ising limit corresponding to the limit @xmath46 .
the thermal correlation functions of the ferromagnetic string operator are obtained in section iii .
section iv presents several results for the four - vertex model which is the @xmath47 limit of the six - vertex model . in section
iv it is demonstrated that the effective hamiltonian commutes with the transfer - matrix of the four - vertex model , the orthogonality of the eigen - functions is proved , and the relation between the four - vertex model and the boxed plane partitions @xcite is discussed .
the low temperature asymptotics of the correlation function is studied in the section v. discussion in section vi concludes the paper .
less studied limit of the @xmath0 model is the _ strong anisotropy _ ( sa ) limit @xmath46 @xcite .
this limit is called the _ ising limit _ since the simplest hamiltonian in this limit turns out to be the hamiltonian of the one - dimensional ising model @xcite : @xmath49 however we shall study the strong anisotropy limit using the effective hamiltonian @xmath50 , which is formally equivalent to the @xmath44 hamiltonian ( [ anis1 ] ) , provided the latter is supplied with the requirement forbidding two spin `` down '' states to occupy any pair of nearest - neighboring sites : @xmath51 the projectors @xmath52 `` cut out '' the spin `` down '' states for any pair of nearest - neighboring sites . the local projectors onto the spin `` up '' and `` down '' states are of the form : @xmath53 where @xmath54 is the unit operator in @xmath55-fold tensor product of the linear spaces @xmath56 .
the point is that the hamiltonians ( [ ising ] ) and ( [ anis2 ] ) are commutative .
indeed , the hamiltonian ( [ ising ] ) can be represented by means of ( [ proj ] ) as follows : @xmath57 where @xmath58 is the operator of number of particles commuting with the hamiltonians @xmath59 and @xmath50 . on another hand , the definition of the projector @xmath52 yields : @xmath60
then , @xmath61=0 $ ] and so the hamiltonians ( [ ising ] ) and ( [ anis2 ] ) possess a common system of the eigen - functions .
the wave function ( [ xxzwf ] ) takes the form in the limit @xmath46 : @xmath62 where the coordinates of the spin `` down '' states form a strict decreasing partition @xmath63 , as in ( [ bwf ] ) .
it follows from ( [ xxwf1 ] ) that the wave function is not equal to zero if and only if the elements @xmath25 , @xmath64 , satisfy the _ exclusion condition _ : @xmath65 . in other words , in the considered limit occupation of nearest sites is forbidden , and the _ hard - core diameter _
equal to duplicated inter - site separation arises .
the exponential parametrization @xmath66 brings the bethe equations ( [ xxzbethe ] ) to the form : @xmath67 which is respected by the solutions @xmath68 where @xmath69 , and the @xmath70 are integers or half - integers , depending on whether @xmath23 is odd or even .
it suffices to consider a set of @xmath23 different numbers @xmath70 satisfying the condition @xmath71 .
the ground state of the model is defined by the solutions @xmath72 with the condition @xmath73 fulfilled .
the notation @xmath74 for @xmath23-tuple @xmath75 of the solutions ( [ sol ] ) will be especially convenient further in order to stress that one is concerned with the solution of the bethe equations . otherwise , it is appropriate to use @xmath35 as an indication that an arbitrary parametrization of the wave - function ( [ xxwf1 ] ) is meant .
the eigen - energy of the ground state of the model calculated by means of ( [ een ] ) and ( [ grstsa ] ) is : @xmath76 the eigen - energy of the ising hamiltonian ( [ ising ] ) has the form : @xmath77 . the two limits , ( [ anis1 ] ) and ( [ anis2 ] ) , are similar in the sense that their wave functions are expressible through the _ schur functions _ @xcite : @xmath78 where @xmath79 denotes the partition @xmath80 , which is an @xmath23-tuple of nonincreasing non - negative integers : @xmath81 .
indeed , any strict partition @xmath82 and non - strict partition @xmath83 ( denoted as @xmath63 and @xmath84 , respectively ) can be related by the formula @xmath85 , where @xmath86 . in other terms , @xmath87 , where @xmath88 is the strict partition @xmath89 .
any strict partition @xmath63 with the elements respecting the exclusion condition @xmath65 is connected with the non - strict partition @xmath90 by the relation @xmath91 , where @xmath92 .
so , the wave function ( [ xxwf1 ] ) may be represented as @xmath93 , and the corresponding state - vector ( [ bwf ] ) is specialized as follows : @xmath94 summation in ( [ vstv ] ) goes over all non - strict partitions @xmath90 into at most @xmath23 parts so that each is less than @xmath95 , i.e. , @xmath96 .
summation over the non - strict partitions @xmath90 is equivalent to that over the strict partitions @xmath63 with the elements respecting the conditions @xmath65 and @xmath97 .
we shall study the thermal correlation function of the states with no spins `` down '' on the last @xmath98 sites of the lattice ( would be called as _ survival probability of the ferromagnetic string _ ) defined by the ratio : @xmath99 where @xmath50 and @xmath100 are defined , respectively , by ( [ anis2 ] ) and ( [ proj ] ) , and @xmath101 is inverse temperature .
besides , @xmath102 indicates that the eigen - states are calculated on the solution ( [ grstsa ] ) .
our calculations will be based on the following relation for the schur functions ( [ sch ] ) which is due to the binet
cauchy formula @xcite : @xmath103 summation in ( [ schr ] ) goes over all non - strict partitions @xmath84 into at most @xmath23 parts so that each is less than @xmath104 : @xmath105 .
the notation for the range of summation in ( [ schr ] ) will be extensively used in the rest of the paper .
firstly , it is straightforward to calculate , using ( [ schr ] ) , the following average of the projector @xmath100 ( [ anis2 ] ) over the state - vectors ( [ vstv ] ) : @xmath106 where the notation for the vandermonde determinant is used : @xmath107 right - hand side of ( [ spdfpxx ] ) , taken at @xmath108 , gives the answer for the scalar product of the state - vectors ( [ vstv ] ) , i.e. , for @xmath109 .
equation ( [ spdfpxx ] ) , taken on the solutions ( [ grstsa ] ) , allows one to obtain the so - called _ emptiness formation probability _ @xmath110 , which expresses the probability of formation of a string of the empty ( i.e. , spin `` up '' ) states on last @xmath98 sites of the lattice : @xmath111 + \,\frac{1-e^{i n(k_l - k_j ) } } { ( m - n+1)(1-e^{i(k_l - k_j ) } ) } ( 1-\delta_{l j})\bigr)_{1\leq l , j\leq n}\,,\qquad k_l\equiv \frac{2\pi l}{m+1-n}. & & \nonumber\end{aligned}\ ] ] the solutions of the bethe equations ( [ bthv ] ) constitute a complete set of the eigen - states , and therefore one can obtain the resolution of the identity operator : @xmath112 where the summation over @xmath113 implies summation over all independent solutions ( [ sol ] ) . here the exponential parametrization for the solutions of the bethe equations is used in the compact form , @xmath114
. we shall calculate the nominator of ( [ ratbe0 ] ) using insertions of the resolution ( [ ratbe171 ] ) . taking into account
that @xmath115 we obtain : @xmath116 \times\,\displaystyle{\bigl| { \cv}(e^{i{\bth } } ) \bigr|^2\,\cp({\textbf v}^{-2 } , e^{i{\bth}})\,\cp(e^{-i{\bth } } , { \textbf u}^{2})}\ , , \end{array } \label{ratbe200}\ ] ] where @xmath117 the range of summation in ( [ ratbe201 ] ) is taken as follows : @xmath118 .
then , using the relation ( [ schr ] ) we obtain from ( [ ratbe200 ] ) : @xmath119 & = & \displaystyle{\frac{1}{(m+1)(m+1-n)^{n-1}\,{\cv}({\textbf u}^2){\cv}({\textbf v}^{-2 } ) } \sum\limits_{m - n\ge i_1 >
i_2 \dots > i_n\ge 0 } \!\!e^{-\be e_n(\bth)}}\\[0.6 cm ] & \times&\displaystyle{\det \bigl ( \frac{1-(e^{i\ta_{i}}v_j^{-2})^{m - n - n+1}}{1- e^{i\ta_{i}}v_j^{-2 } } \bigr)_{1\le i , j \le n}\,\det \bigl(\frac{1-(u_l^2 e^{-i\ta_{p}})^{m - n - n+1}}{1-u_l^2 e^{-i\ta_{p}}}\bigr)_{1\le l , p \le n}}\ , , \end{array } \label{ratbe20}\ ] ] where summation goes over the ordered sets @xmath120 that parametrize @xmath74 ( [ sol ] ) , and @xmath121 is given by ( [ eigenv ] ) . expression for @xmath122 ( [ ratbe0 ] ) appears from ( [ ratbe20 ] ) as follows : @xmath123 \times\,\displaystyle\biggl|{\det \bigl(\frac{1-e^{i(m - n - n+1)(\ta_l-{\ta}^{\rm v}_{p})}}{1- e^{i(\ta_l-{\ta}^{\rm v}_{p})}}\bigr)_{1\le p , l \le n}\biggr|^{2 } } , \end{array } \label{ratbe212}\ ] ] where @xmath102 implies the ground - state solution ( [ grstsa ] ) , and @xmath124 is the ground - state energy in the limit of strong anisotropy .
the _ six - vertex _ model on a square lattice is defined by six different configurations of arrows pointed both in and out of each lattice site . a statistical weight @xmath126 @xmath127 is ascribed to each admissible type of the vertices ( fig .
1 ) . representing the arrows pointing up or to the right by the solid lines one can get the alternative description of the vertices in terms of lines floating through the lattice sites . so far the bonds of a lattice may be only in two states with a line either without it , there is a one to one correspondence between the admissible configuration of arrows on a lattice and the network of lines the nest of lattice paths .
the _ four - vertex _ model is obtained in the limit when the weights @xmath128 .
the @xmath104-operator of the six - vertex model is equal to @xcite : @xmath129 where @xmath130 and @xmath131 , and it satisfies the intertwining relation : @xmath132 here , @xmath133 is the @xmath134-matrix : @xmath135 where @xmath136 and @xmath137 .
let us consider the following transformation of the @xmath104-operator ( [ ratbe50 ] ) : @xmath138 & = & \left ( \begin{array}{cc } -ue^{(h+\gamma ) \si_n^z}-u^{-1}e^{(h-\ga ) \si_n^z } & 2\sinh(2\ga)\ , e^{\omega+h\si_n^z}\,\si_n^{- } \\ 2\sinh(2\ga)\ , e^{-\omega + h\si_n^z}\,\si_n^{+ } & u e^{(h-\ga ) \si_n^z}+u^{-1}e^{(h+\ga ) \si_n^z } \end{array } \right ) .
\nonumber\end{aligned}\ ] ] the operator ( [ tlsv ] ) is intertwined by the transformed @xmath139-matrix of the form : @xmath140 if one puts @xmath141 , then the limit @xmath142 where @xmath143 ( see ( [ proj ] ) ) , gives us the @xmath104-operator of , so - called , _ four - vertex _ model .
analogously , the limit @xmath144 transforms the matrix @xmath145 into the @xmath139-matrix of the four - vertex model @xcite : @xmath146 where @xmath147 the monodromy matrix of the models is defined as the matrix product of @xmath104-operators : @xmath148 the transfer matrix is the matrix trace of the monodromy matrix : @xmath149 . in @xcite
it was proved that the transfer - matrix of the six - vertex model commutes with the hamiltonian of the @xmath0 model ( [ xxzham ] ) : @xmath150=0 $ ] .
the transfer - matrix of the four - vertex model @xmath151 satisfies the property : @xmath152 where @xmath153 is the projector defined in ( [ anis2 ] ) . indeed , using the explicit expression for the entries of the product of two @xmath104-operators ,
@xmath154 one can show that @xmath155 where @xmath156 .
it follows from ( [ cl ] ) that @xmath157 at @xmath158 .
invariance of the matrix trace with respect of cyclic permutation implies that @xmath159 , and therefore the relation ( [ lt ] ) is also fulfilled .
further , one can advance the following * proposition : * _ the transfer matrix of the four - vertex model commutes with the @xmath0 hamiltonian taken in the ising limit _ : @xmath160=0\,.\label{ratbe53}\ ] ] * proof : * in order to prove ( [ ratbe53 ] ) , it is suffices to consider @xmath161 ( [ anis2 ] ) at @xmath162 .
we shall use the method proposed in @xcite and then developed in @xcite . generalizing this
approach with respect of the problem under consideration , we shall take into account the relation ( subscripts are omitted ) : @xmath163\,\cc_{n+1 , n } \\[0.3 cm ] = \,\cc_{n+1 , n}\bigl ( q(n+1|u)l(n|u)-l(n+1|u)q(n|u)\bigr ) \cc_{n+1 , n}\ , , \end{array } \label{ratbe54}\ ] ] where the operators @xmath164 and @xmath165 are introduced in ( [ anis1 ] ) and ( [ cl ] ) , respectively .
if so , we obtain the vanishing of the commutator in question ,
@xmath166 \,=\,-\frac 12\sum_{n=0}^m \cp\,\bigl[\widehat h_{n+1,n},\tau ( u)\bigr]\,\cp\,=\,-\frac 12 \sum_{n=0}^m \tr\bigl(\cp\,\bigl [ \widehat h_{n+1,n},t(u)\bigr]\,\cp\bigr)\,=\,0\ , , \label{ratbe541}\ ] ] provided that @xmath167\,=\ , l(m|u ) \cdots l(n+2|u)\,\bigl [ \widehat h_{n+1,n},l(n+1|u)l(n|u ) \bigr]\,l(n-1|u)\cdots l(0|u)\,.\label{ratbe542}\ ] ] furthermore , the commutators of the entries from the right - hand side of ( [ ratbe543 ] ) look as follows : @xmath168\,=\,-\,\bigl [ \widehat h_{n+1,n},d_{n+1,n } \bigr ] \,=\,\check q_{n+1}-\check q_n \\[0.3 cm ] \bigl [ \widehat h_{n+1,n},b_{n+1,n}\bigr ] \,=\,-u\si_{n+1}^{-}+u^{-1 } \si_n^{-}+u\si_{n+1}^{- } \hat q_n - u^{-1 } \hat q_{n+1 } \si_n^{- } , \\[0.3 cm ] \bigl [ \widehat h_{n+1,n},c_{n+1,n } \bigr ] \,=\,-u^{-1}\si_{n+1}^{+}+u \si_n^{+}+u^{-1}\si_{n+1}^{+ } \hat q_n - u \hat q_{n+1 } \si_n^{+}\ , .
\end{array } \label{ratbe544}\ ] ] in turn , one obtains from ( [ ratbe544 ] ) : @xmath169\,\cc_{n+1 , n } \,=\,\cc_{n+1 , n}\bigl ( \check q_{n+1}-\check q_n\bigr ) \cc_{n+1 , n } , \\[0.3 cm ] \cc_{n+1 , n}\,\bigl[\widehat h_{n+1,n } , b_{n+1,n}\bigr]\,\cc_{n+1 , n } \,=\,\cc_{n+1 , n } \bigl (
-u\si_{n+1}^{-}+u^{-1}\si_n^{-}\bigr ) \cc_{n+1 , n } , \\[0.3 cm ] \cc_{n+1 , n}\,\bigl[\widehat h_{n+1,n } , c_{n+1,n}\bigr]\,\cc_{n+1 , n } \,=\,\cc_{n+1 , n}\bigl ( -u^{-1}\si_{n+1}^{+}+u\si_n^{+}\bigr ) \cc_{n+1 , n } , \\[0.3 cm ] \cc_{n+1 , n}\,\bigl[\widehat h_{n+1,n } , d_{n+1,n}\bigr]\,\cc_{n+1 , n } \,=\,\cc_{n+1 , n}\bigl ( -\check q_{n+1 } + \check q_n \bigr ) \cc_{n+1 , n}. \end{array } \label{ratbe545}\ ] ] on the other hand , using the diagonal matrix @xmath170 we obtain the following relation for the @xmath104-operators ( [ ratbe51 ] ) : @xmath171 comparison of ( [ ratbe545 ] ) and ( [ ratbe546 ] ) demonstrates that the relation ( [ ratbe54 ] ) , as well as the statement of the relation ( [ ratbe53 ] ) are valid indeed .
@xmath172 the state vector of the four - vertex model is constructed in the framework of the algebraic bethe ansatz as follows : @xmath173 this vector is the eigen - vector both of the transfer - matrix of the four - vertex model and of the hamiltonian ( [ anis2 ] ) , provided that the parameters @xmath40 fulfill the bethe equations ( [ bthv ] ) .
it was shown in @xcite that this vector can be represented in the form ( [ vstv ] ) . according to the representation ( [ spdfpxx ] ) taken at @xmath108 ,
the scalar product @xmath109 is proportional to the determinant of the matrix , say , @xmath174 .
the bethe equations ( [ bthv ] ) enable one to represent this determinant as follows : @xmath175 where the entries @xmath176 take the form : @xmath177 it is crucial that the determinant of the matrix @xmath178 vanishes provided the parameters @xmath35 and @xmath179 are independent solutions of the bethe equations ( [ bthv ] ) .
really , this matrix has a non - trivial eigen - vector @xmath180 with zero eigen - value : @xmath181 validity of ( [ normv3 ] ) is justified by two identities : @xmath182 \displaystyle{\sum\limits_{k=1}^n\biggl(\frac{y_k - x_k}{y_k } \prod\limits_{l=1,\,l\ne k}^n \frac{y_k - x_l}{y_k - y_l}\biggr)\,=\,1\,-\,\prod\limits_{l=1}^n \frac{x_l}{y_l}}\ , .
\end{array}\ ] ] therefore , the scalar product @xmath183 vanishes ( i.e. , the state - vectors are orthogonal ) provided the sets of the parameters @xmath35 and @xmath179 represent independent bethe solutions .
completeness of the system of the state - vectors was proved in @xcite .
consider the four - vertex model @xcite .
there is one to one correspondence between the strict plane partitions ( i.e. , the plane partitions that decay along each column and each raw @xcite ) and the nests of the admissible paths on a square lattice of the size @xmath184 with the following boundary conditions : all arrows on the left and right boundaries are pointing to the left , while the arrows on the top and bottom of the first @xmath23 columns ( counting from the left ) are pointing inwards and the arrows on the top and bottom of the last @xmath23 ones are pointing outwards ( see fig .
2 ) .
it was proved in @xcite that if one puts @xmath185 and @xmath186 in the scalar product of the state - vectors ( [ stva ] ) , then @xmath187 where @xmath188 is the generating function of strict plane partitions placed into a box of size @xmath189 : @xmath190 being taken at @xmath191 , this formula gives the number of strict plane partitions in a box of size @xmath192 : @xmath193 it is straightforward to obtain @xcite that the entry @xmath194 is equal to the number strict partitions in a @xmath195 box .
let us obtain the low temperature estimate of the correlation function ( [ ratbe0 ] ) .
the survival probability of the ferromagnetic string is written as follows : @xmath196 where @xmath197 . here
@xmath198 is given according to ( [ ratbe201 ] ) , provided the summation domain is given by @xmath199 .
if the chain is long enough while the number of quasi - particles is moderate , i.e. , @xmath200 , we replace the sums in ( [ ratbe61 ] ) by the integrals as follows : @xmath201 \displaystyle{\times\,\prod\limits_{i=1}^n\bigl ( \int_{0}^{2\pi}\!\frac{d\ta_i}{2\pi}\bigr ) \,e^{{\be}\sum\limits_{l=1}^n(\cos{\ta}_l-\cos{\ta}^{\rm v}_{l})}\,\bigr|{\cp } ( e^{-i{\bth } } , e^{i { \bth}^{\rm v}})\bigr|^2}\prod_{1\leq k
< l\leq n } \bigl|e^{i{\ta}_k}-e^{i{\ta}_l}\bigr|^2\ , .
\end{array } \label{ratbe62}\ ] ] in the considered limit , we may asymptotically put @xmath202 and @xmath203 , @xmath204 . at the large enough @xmath101 ,
we approximate ( [ ratbe62 ] ) in the following way : @xmath205 \displaystyle{\ci\,\equiv\ , \int\limits_{-\infty}^{\infty } \!\int\limits_{-\infty}^{\infty } \cdots\int\limits_{-\infty}^{\infty}\,e^{-\sum\limits_{l=1}^n x^2_l}\!\!\prod_{1\leq k < l\leq n } \bigl|x_k- x_l\bigr|^2 d x_1 d x_2 \dots d x_n } \ , . \label{ratbe621}\end{aligned}\ ] ] the integral @xmath206 ( [ ratbe621 ] ) is the _ mehta integral _ , and its value is known @xcite : @xmath207 where @xmath208 eventually , taking into account ( [ spdfpxx3 ] ) and ( [ ratbe620 ] ) , we express the answer in the following form : @xmath209 the low temperature decay of the survival probability of the ferromagnetic string is governed by the critical exponent @xmath210 , while its amplitude is proportional to the squared number of strict plane partitions in a box of size @xmath211 .
in this paper we studied the @xmath0 heisenberg magnetic chain in a specific limit of the anisotropy parameter : @xmath46 .
the , so - called , survival probability of the ferromagnetic string has been calculated over the @xmath23-particle ground state of the model .
it was proved that the hamiltonian of the ising limit , previously introduced in the papers @xcite , is correct as the limit @xmath46 of the @xmath0 hamiltonian .
finally , we have found the leading term of the low temperature asymptotics of the survival probability of the ferromagnetic string for the model in the finite volume and at fixed number of quasi - particles .
we have demonstrated that the amplitude of the asymptotics is proportional to the squared number of strict plane partitions in a box , while its critical exponent is proportional to the squared number of particles . comparing this result with
that obtained in @xcite , one can conclude that this type of the behavior is universal for a special class of the integrable models .
this paper was supported in part by the russian foundation for basic research ( no .
100100600 ) and by the russian academy of sciences program `` mathematical methods in non - linear dynamics '' .
a. n. vassiliev , _ the functional methods in the quantum field theory and statistics _ , leningrad university press , leningrad , 1976 .
[ in russian ] + a. n. vasiliev , _ functional methods in quantum field theory and statistical physics _ , gordon and breach , amsterdam , 1998 .
n. m. bogolyubov , v. f. bratsev , a. n. vasilev , a. l. korzhenevskii , r. a. radzhabov , _ high temperature expansions at an arbitrary magnetization in the ising model _ ,
* 26 * ( 1976 ) , no . 3 , 230237 . l. d. faddeev , _ quantum completely integrable models of field theory _ , sov .
c , * 1 * ( 1980 ) , 107160 ; in : 40 years in mathematical physics , world sci .
20th century math .
2 , world sci . , singapore , 1995 , pp .
187235 .
n. kitanine , j. m. maillet , n. slavnov , v. terras , _ correlation functions of the xxz spin-@xmath212 heisenberg chain at the free fermion point from their multiple integral representations _
b * 642 * ( 2002 ) , no .
3 , 433455 .
m. katori , h. tanemura , t. nagao , n. komatsuda , _ vicious walks with a wall , noncolliding meanders , and chiral and bogoliubov de gennes random matrices _ ,
e , * 68 * ( 2003 ) , no . 2 , 021112 [ 16 pages ] . n. m. bogoliubov , c. malyshev , _ a path integration approach to the correlators of xy heisenberg magnet and random walks _ , in : proceedings of the 9th intern .
`` path integrals : new trends and perspectives '' ( dresden , germany , september 2328 , 2007 ) .
eds . , w. janke , a. pelster ( world sci . ,
singapore , 2008 ) , pp .
arxiv:0810.4816 n. m. bogoliubov , c. malyshev , _ the correlation functions of the @xmath0 heisenberg chain in the case of zero or infinite anisotropy , and random walks of vicious walkers _ , st . petersburg math .
j. , * 22 * ( 2011 ) , no . 3 , 359 - 377 .
arxiv:0912.1138 c. malyshev , _ functional integration with an `` automorphic '' boundary condition and correlators of third components of spins in the @xmath44 heisenberg model _ , theor .
phys . , * 136 * ( 2003 ) , no . 2 , 11431154
. arxiv : hep - th/0204007 f. c. alcaraz , r. z. bariev , _ an exactly solvable constrained @xmath0 chain _ , in : _ statistical physics on the eve of the 21@xmath214 century . in honour of j.
b. mcguire on the occasion of his 65@xmath215 birthday ( series on advances in statistical mechanics ) _ , eds . , m. t. batchelor , l. t. wille ( world scientific , singapore , 1999 ) .
arxiv : cond - mat/9904042
|
the spin-_1/2 _ @xmath0 heisenberg magnet is considered for the case of the anisotropy parameter tending to infinity ( so - called , _ ising limit _ ) .
a thermal correlation function of the ferromagnetic string is calculated over the ground state .
the approach to the calculation of the correlation functions is based on the observation that the wave function in the limit considered is expressed in terms of the symmetric schur functions .
it is demonstrated that at low temperatures the amplitude of the asymptotical expression of the present correlation function is proportional to the squared numbers of strict boxed plane partitions
. 1.0 cm 1.0 cm
|
the oslo cyclotron group has established a method to deduce experimental level densities and @xmath0-ray strength functions @xcite .
these data contain essential information on nuclear structure and thermal and electromagnetic properties . in the last couple of years several fruitful applications of the method have been reported [ 2 - 10 ] .
the most efficient way to create entropy in atomic nuclei is to break @xmath1 nucleon cooper pairs in the core .
a beautiful manifestation of pairbreaking is the backbending phenomena in rapidly rotating nuclei . in this work
, the pairbreaking process is studied as function of intrinsic excitation energy ( or temperature ) . of great interest
is the quenching of the pair correlations , a topic which is directly connected to the nuclear level density . important applications of nuclear level densities and @xmath0-ray strength functions are the determination of nuclear reaction cross sections from hauser - feshbach type of calculations .
these cross sections are used as input parameters in large network calculations of stellar evolution , and in the simulation of accelerator - driven transmutation of nuclear waste .
the experiments were carried out using the ( @xmath2he,@xmath3 ) and ( @xmath2he,@xmath2he@xmath0 ) reactions at the oslo cyclotron laboratory with a 45 mev @xmath2he beam .
self - supporting metallic targets of rare earth nuclei were isotopically enriched to @xmath4% and had thicknesses of @xmath5 mg/@xmath6 .
the charged particles and @xmath0 rays were recorded with the detector array cactus , which contains eight particle telescopes and 27 nai @xmath0-ray detectors .
each telescope is placed at an angle of 45@xmath7 relative to the beam axis , and comprises one si front and one si(li ) back detector with thickness @xmath8 and @xmath9 @xmath10 m , respectively .
the nai @xmath0-detector array , having a resolution of @xmath11 % at @xmath12 mev and a total efficiency of @xmath13 % , surrounds the target and particle detectors .
in addition , two ge detectors were used to monitor the spin distribution and the selectivity of the reactions . in order to determine the true @xmath0-energy distribution , the @xmath0 spectra
are corrected for the response of the nai detectors with the unfolding procedure of ref .
in addition random coincidences are subtracted from the @xmath0 spectra .
the set of unfolded @xmath0 spectra are organized in a @xmath14 matrix , where the initial excitation energies @xmath15 are determined by means of reaction kinematics utilizing the energy of the ejectile .
this matrix comprises the @xmath0-energy distribution of the total @xmath0 cascade .
the primary @xmath0 matrix can now be found according to the subtraction technique of ref .
@xcite , see data points of fig . [
fig : fgboth ] .
the procedure is based on the assumption that the decay properties of the particular reaction - selected distribution of states within each energy bin are independent on whether the respective ensembles of states are directly populated through the nuclear reaction or by @xmath0-decay from higher lying states .
the idea is now to find two functions , the level density @xmath16 and the @xmath0-energy dependent function @xmath17 , that according to the brink - axel hypothesis @xcite , should describe the primary @xmath0-ray spectrum at excitation energy @xmath15 : @xmath18 here , @xmath19 is fitted to the observed primary @xmath0-ray matrix @xcite . in fig .
[ fig : fgboth ] the best fit to @xmath20 is shown for the @xmath21dy nucleus .
all primary @xmath0-ray spectra are seen to be very well described by the same @xmath22 and @xmath23 functions .
however , from @xmath22 and @xmath23 one can construct other functions which give identical fits to the data by @xcite @xmath24 \rho(e - e_{\gamma } ) , \label{eq : array1 } \\
\tilde{f}(e_{\gamma } ) & = & b \exp(\alpha e_{\gamma } ) f(e_{\gamma } ) .
\label{eq : array2}\end{aligned}\ ] ] here , the parameters @xmath25 and @xmath26 can be determined by fitting the level density to the number of known discrete levels at low excitation energy and to the level density estimated from neutron - resonance spacing data at high excitation energy .
furthermore , for dipole radiation , @xmath23 is proportional to @xmath27 , where @xmath28 is the @xmath0-ray strength function .
thus , according to ref .
@xcite , the parameter @xmath29 can be determined from the known total @xmath0 widths of neutron resonances . finally , the normalized level densities and @xmath0-ray strength functions are shown in fig .
[ fig : strfrho8 ] .
high nuclear level density is due to the many ways the potential and kinetic energies of individual nucleons can add up to a certain excitation energy .
each valence nucleon , i.e. nucleons not coupled in cooper pairs , carries a single - particle entropy of @xmath30 in units of the boltzmann constant @xmath31 .
thus , a broken cooper pair contributes significantly to the total entropy and heat capacity . observed structures in the caloric curve and heat capacity
may therefore indicate the breaking of cooper pairs .
the microcanonical partition function is simply the multiplicity of nuclear states , which experimentally corresponds to the level density of accessible states .
thus , the experimental level density @xmath16 is our starting point for the extraction of thermodynamic properties of nuclei .
the entropy in the micro - canonical ensemble is determined by @xmath32 where the boltzmann constant is set to unity ( @xmath33 ) , and @xmath34 is adjusted to obtain @xmath35 in the ground - state band of even - even nuclei . the nuclear temperature and heat capacity is defined by @xmath36 @xmath37 figure [ fig : termo ] shows the entropy , temperature and heat capacity for @xmath38er and @xmath39er deduced in the microcanonical ensemble .
the small bumps in the entropy curves are seen to be enhanced through the differentiations performed in eq .
( [ eq : t ] ) .
the spectacular feature of negative branches in the heat capacity is a direct consequence of negative slopes in the caloric curve . until today
, their interpretation is a controversial topic .
the canonical ensemble theory is based on the partition function @xmath40 which is a laplace transform of the multiplicity of states @xmath41 , where @xmath42 is the energy bin .
the thermal average of the excitation energy in the canonical ensemble is @xmath43 by the laplace transform in eq .
( [ eq : z ] ) much of the information contained in the micro - canonical level density is smeared out .
thus , fine structure in the thermodynamic observables in the microcanonical ensemble will not be visible in the canonical ensemble .
the lines in fig .
[ fig : termo ] display the smooth dependence of the canonical temperature on @xmath44 .
the corresponding heat capacity is the derivative by @xmath45 and is shown in fig .
[ fig : cv ] for @xmath46yb as a function of @xmath47 and @xmath44 .
the heat capacities show a pronounced peak as function of temperature at @xmath48 mev , which we interpret as the critical temperature for the pair breaking transition .
in fig . [ fig : elin ] the experimental @xmath0-strength function is fitted by a theoretical strength function taking into account both the giant electric dipole resonance and the spin - flip resonance .
in addition , a weaker resonance at lower energies is needed in order to fit the experimental data . because of the much lower strength of this resonance compared to the giant electric dipole resonance ( gedr )
, it is denoted pygmy resonance .
the e1 radiation is described by @xcite @xmath49 where we apply @xmath47 as a constant fit parameter .
the m1 radiation is described by the lorentzian @xmath50 where @xmath51 , @xmath52 and @xmath53 are giant magnetic dipole resonance ( gmdr ) parameters @xcite .
furthermore , the pygmy resonance is described with a lorentzian function @xmath54 ( similar as eq .
( [ eq : m1 ] ) ) , where the pygmy - resonance strength @xmath55 , width @xmath56 and centroid @xmath57 have been fitted in order to adjust the total theoretical strength function to the experimental data . the pygmy resonance parameters obtained for several rare earth nuclei
are shown in fig .
[ fig : pygsys ] .
the systematics indicate that the strengths @xmath58 of these soft dipole resonances are quenched when approaching the @xmath59 shell gap .
the pygmy resonance has been explained by the enhancement of the e1 @xmath0-strength function @xcite . still , m1 character can not be excluded . at an excitation energy around @xmath60 mev
, there is a concentration of orbital m1 strength in the weakly collective scissors mode @xcite .
this mode was first observed in electron - scattering experiments @xcite , and is confirmed by the @xmath61 reaction @xcite .
clarification of the electromagnetic character of the soft dipole resonance observed in our work awaits new experimental results .
thermodynamic observables have been deduced from the level density and display signatures of phase - like transitions within the microcanonical and the canonical ensemble , interpreted as the transition from a strongly pair - correlated phase to an uncorrelated phase . in the microcanonical ensemble
one may observe details of the successive breaking of nucleon pairs , information which is hidden in the canonical approach .
the canonical ensemble on the other hand , reveals the average properties of the pairing transition . using the canonical @xmath62 quantity as thermometer , a local maximum is found at @xmath63 mev , indicating the breaking of cooper pairs and quenching of pair correlations .
the experimental @xmath0-strength function is fitted by a theoretical strength function , assuming that the @xmath0 decay in the continuum is governed by dipole transitions .
the contribution of electric and magnetic dipole radiation to the @xmath0-strength function is recognized , and a bump is observed in the @xmath0-strength function at @xmath64 mev . a measurement of the electromagnetic character of this soft dipole resonance will be crucial in order to understand the physics behind this exciting phenomenon .
financial support from the norwegian research council ( nfr ) is gratefully acknowledged .
part of this work was performed under the auspices of the u.s .
department of energy by the university of california , lawrence livermore national laboratory under contract no . w-7405-eng-48 .
e. melby _ et al .
_ , phys .
lett . * 83 , 3150 ( 1999 ) .
a. schiller _ et al .
_ , phys .
c * 61 , 044324 ( 2000 ) .
m. guttormsen _ et al .
_ , phys .
c * 61 * 067302 ( 2000 ) .
m. guttormsen _ et al .
_ , phys .
c * 62 * 024306 ( 2000 ) .
a. schiller _ et al .
c * 63 * , 021306 ( 2001 ) .
m. guttormsen _ et al .
_ , phys . rev .
c * 63 * , 044301 ( 2001 ) .
e. melby _ et al .
_ , phys .
c * 63 * , 044309 ( 2001 ) .
a. voinov _ et al .
_ , phys .
c * 63 * , 044313 ( 2001 ) .
s. siem _ et al .
_ , phys .
c * 65 * , 044318 ( 2002 ) . * * m. guttormsen,_et al .
_ , nucl .
instrum .
methods phys .
a * 374 , 371 ( 1996 ) .
m. guttormsen , t. ramsy , and j. rekstad , + nucl .
instrum .
methods phys .
a * 255 * , 518 ( 1987 ) .
brink , ph.d .
thesis , oxford university ( 1955 ) .
p. axel , phys .
rev . * 126 * , 671 ( 1962 ) .
* s.g .
kadmenski , v.p .
markushev , and v.i .
furman , + yad .
37 , 277 ( 1983 ) [ sov .
j. nucl .
phys . * 37 , 165 ( 1983 ) ] .
m. igashira _ et al .
_ , nucl .
phys . * a457 , 301 ( 1986 ) .
a. richter , nucl .
* a507 , 99c ( 1990 ) .
d. bohle _ et al .
_ , phys .
b * 137 , 27 ( 1984 ) .
n. pietralla _
et al . _ ,
c * 58 , 184 ( 1998 ) , and references therein . * * * * * *
|
a method to extract primary @xmath0-ray spectra from particle-@xmath0 coincidences at excitation energies up to the neutron binding energy is described . from these spectra , the level density and @xmath0-ray strength function can be determined . from the level density ,
several thermodynamical quantities are obtained within the microcanonical and canonical ensemble . also models for the @xmath0-ray strength function
are discussed .
+ + , m. hjorth - jensen , j. rekstad and s. siem + department of physics , university of oslo , box 1048 blindern , n-0316 oslo , norway + + a. schiller + lawrence livermore national laboratory , livermore , ca-94551 , usa + + a. voinov + frank laboratory of neutron physics , joint institute of nuclear research , 141980 dubna , moscow reg . ,
russia +
|
the magnetic properties of free - standing atomic clusters of 3d tm elements have been intensively scrutinized during the last two decades .
two different but related phenomena have specifically been discussed and essentially unravelled .
the first is the modification of local magnetic moments as compared with the values found in bulk materials .
the second is the competition between the possible ferromagnetic , antiferromagnetic and non - collinear arrangements of the local spins , as well as its interplay with the geometry of the nanostructure . in the case of ferromagnetic elements like fe , co and ni ,
the increase of the average cluster magnetic moment can be easily explained in terms of the reduced atomic coordination in the low - dimensional regime , with oscillations associated to structural ( symmetry ) changes .
@xcite the case of antiferromagnets like cr and mn is much more complex .
atoms of these elements may display large magnetic moments , since they have a large number of d - holes susceptible to be polarized . on the other hand , clusters of these atoms may display tiny average magnetizations due to the tendency of their atomic moments to align in antiparallel directions .
the structure plays also a fundamental role in the magnetic behavior of these clusters , since it may originate magnetic frustration .
a conventional example of magnetic frustration in a classical spin system appears when atoms positions form triangular motifs .
the studies of these classical systems show that magnetic frustration frequently leads to non - collinear configurations of the local spin moments .
the latest theoretical studies reported in the literature show that non - collinear arrangements of quantum spins also appear as the ground or as some of the first isomers of clusters of @xmath0 atoms , including not only cr and mn , but also fe , co and ni.@xcite all materials made of @xmath1 tm elements are paramagnets , in contrast to those of the @xmath0 row .
a natural question thus arises of whether small clusters of @xmath1 elements may show low - lying magnetic states of collinear or even non - colinear nature .
bulk palladium , being a paramagnet in the brink of becoming a ferromagnet , presents one of the most intriguing and controversial magnetic behaviors in nature.@xcite it is therefore not surprising that the very few experimental and theoretical studies published so far try to clarify whether pd clusters of given sizes are magnetic or not , and what is the order of magnitude of their average magnetic moment . from the experimental side ,
most of the reports agree that only very small clusters have a net magnetic moment @xcite , with the exception of shinohara and coworkers , @xcite who found noticeable magnetic moments at the surface of pd particles as big as 79 . from the theoretical side ,
there is also consensus that very small pd clusters are indeed magnetic.@xcite futschek et al.@xcite have studied recently small pd clusters using density functional theory ( dft ) in the collinear framework , within a fixed - moment mode .
they have found that multiple spin isomers exist for each cluster size with very small energy differences .
interestingly , some of these competing isomers present ferromagnetic order , while others display antiferromagnetic alignments , with possible frustration .
although pd has tendency to ferromagentic order , this fact strongly points out to the possible existence of non - collinear magnetic structures , as a mechanism to release the frustration and competition between the different magnetic solutions .
we report in this article a thorough _ ab initio _ study of the magnetic behavior of small palladium clusters pd@xmath2 , with @xmath3 ranging from 3 to 7 .
we have performed a simultaneous optimization of the geometric and magnetic degrees of freedom fully allowing for non - collinear spin arrangements .
a debate exists currently on the accuracy of the local density approximation ( lda ) @xcite versus the generalized gradient approximation ( gga ) @xcite for the determination of the magnetic behavior of low - dimensional pd systems @xcite .
the present article shows that both approximations provide the same results for the ground state of free - standing pd atomic clusters .
moreover , even the existence and energy ordering of the first excited states depend much more weakly on the choice of correlation functional and pseudopotential in than the bulk fcc material .
we have performed our calculations using the code siesta.@xcite siesta is a dft method that employs linear combination of pseudoatomic orbitals as basis set .
the electronic core is replaced by a nonlocal norm - conserving troullier - martins@xcite pseudopotential that may include nonlinear core correction terms .
the code allows to perform , together with the electronic calculation , structural optimization using a variety of algorithms .
it also allows to simulate non - collinear spin arrangements both in the lda and in the gga approximations .
@xcite in the present calculation , we have also used a variety of pseudopotentials to test their effect on free - standing clusters and their corresponding transferability .
we have generated three different pseudopotentials using lda .
the first ( lda1 ) was built with the electronic configurations @xmath4 , @xmath5 and @xmath6 , and core - corrections matching radius @xmath7 a.u . ; the second ( lda2 ) was identical to lda1 , but with @xmath8 a.u .
; the third had a closed - shell atomic configuration ( @xmath9 , @xmath5 and @xmath10 ) and @xmath8 a.u .
we have also generated two gga pseudopotentials with electronic configuration @xmath4 , @xmath5 and @xmath6 , and @xmath11 or 1.2 a.u .
( gga1 and gga2 , respectively ) . in all five cases ,
the cutoff radii of the s , p and d orbitals were taken at 2.30 , 2.46 and 1.67 a.u .
, respectively .
we have described valence states by a double-@xmath12 polarized basis set ( e.g. : two different radial functions for s and d orbitals and a single one for p orbitals ) .
we have taken an energy cutoff of 150 ry to define the real space grid for numerical integrations , but we checked that higher cutoffs did not alter the results .
we have carried out the structural optimization using a conjugate gradient algorithm , where we have set the tolerance for the forces at 0.003 ev / , with eventual double - checks using 0.001 ev / .
we have found that the five pseudopotentials provide similar results when applied to an isolated palladium atom , being the eigenvalues of the ground state and different excited states slightly better reproduced with lda1 and gga1 ( both had @xmath7 a.u . ) .
however , we have observed that they give rise to different magnetic behaviors when applied to the bulk fcc material .
we first remind that palladium has a paramagnetic ground state with a lattice constant of 3.89 .
moreover , the latest simulations show that lda predicts a paramagnetic ground state with a lattice constant of about 3.85 - 3.91 in agreement with experiments , while gga predicts the ground state to be ferromagnetic , with a much larger lattice constant of about 4.0 and a magnetic moment of about 0.4 @xmath13 .
the fact that lda provides a slightly better description of the ground state of bulk hcp has been stressed recently by alexandre and co - workers @xcite .
our lda1 pseudopotential gives a ferromagnetic ground state with @xmath14 , while lda2 and lda3 predict the ground state to be paramagnetic , as it should .
finally , both gga pseudopotentials lead to a ferromagnetic ground state with @xmath15 . on the other hand ,
all lda approximations give a lattice constant equal to 3.90 .
in contrast , both gga pseudopotentials predict it to be equal to 4.01 .
we note here that the fact that all lda pseudopotentials predict the same lattice constant for the ground state regardless of its magnetic nature also happens with fcc iron , where lda predicts the same ground state energy and lattice constant for the paramagnetic , antiferromagnetic and ferromagnetic low spin states @xcite .
these results highlight the lack of transferability of the different pseudopotential and the sensitivity with the exchange and correlation functional used to describe the ground state of bulk palladium .
it is therefore of great importance to assess whether the same difficulties apply to the case of the atomic clusters considered here . as we shall discuss below
, atomic palladium clusters are much more insensitive to the choice of pseudopotential and exchange and correlation functional .
we shall show that the ground state is the same for the different approximations used here , in stark contrast to the case of the bulk material .
further , a large portion of the low - lying excited states are reproduced by all approximations .
notice that we have not kept fixed the magnetic moment in our simulations of the pd@xmath2 clusters , but rather have allowed it to vary freely during the non - collinear iterative selfconsistency process , in contrast to previous authors .
moreover , while we can not rule out that we may have missed low lying solutions , we have endeavored to minimize this risk by feeding a large variety of non - collinear seeds for each cluster . this effort has allowed us to find a rich and complex family of metastable solutions , that was absent in previous works .
we finally note that we have repeated all calculations with the pseudopotentials lda1 , lda3 and gga1 .
we have found that all clusters , except pd@xmath16 , share the same collinear magnetic ground state , with a total spin of 2 @xmath13 , in agreement with previous authors @xcite .
the geometry of the ground state and the average interatomic distance of the pd@xmath2 clusters is displayed in fig .
1 , where we show that these range from 2.55 in @xmath17 to 2.71 for @xmath18 . we have written the binding energies of the different clusters in table i. the table shows that gga1 gives slightly smaller values than lda1 and lda3 , as otherwise expected . moreover , the binding energies predicted by gga1 are very similar to those obtained by kumar , who also used the gga ( within an ultrasoft pseudopotentials , plane waves code ) and by futschek et al . , who used the all - electron vasp code , but did not state the approximation employed .
we should stress that all the tested pseudopotentials provide the same ground state , in stark contrast to the situation that arose for the bulk material .
moreover , we have found very similar inter - atomic distances for all pd@xmath2 clusters , as shown in table ii .
the table shows that gga1 predicts slightly longer distances than lda1 or lda3 , as otherwise
. these distances also agree with those obtained by kumar and futschek within a range of 1 per cent .
the pd@xmath16 cluster displays a behavior different from the rest , and therefore we discuss it separately .
futschek and coworkers @xcite found that pd@xmath16 was also ferromagnetic in contrast to kumar et al.@xcite , who predicted it to be paramagnetic .
we have found that both states are nearly degenerate , with the paramagnetic solution being slightly more stable .
further , we have found that the relative energy of the two solutions show a strong dependence with inter - atomic distance .
( 2 ) shows that there is a level crossing when the bonds are elongated by just 1 per cent from the equilibrium distance . the close proximity between the equilibrium distance and the level crossing distance explains the sensitivity of these results with respect to slight differences in the simulations and the discrepancies between kumar and futscheck @xcite .
aditionally , we have been unable to find non - collinear or antiferromagnetic solutions for this cluster .
nc2 ( lda1 ) , nc2 ( lda3 ) and nc1 ( gga1 ) .
the arrows are proportional to the size of the atomic moments . ] in contrast , and independently of the pseudopotential or approximation used , the rest of the clusters show a rich variety of antiferromagnetic and non - collinear solutions .
most of these solutions , though not all , exist for all lda1 , lda3 and gga1 .
we have also found that , whenever they exist , the relative order of the different solutions is maintained , and the size of the atomic moments is very similar .
these facts strengthen our belief that pd atomic clusters are much more insensitive to the pseudopotential and approximation employed than bulk pd .
it is also reassuring that most of the collinear solutions have been identified in previous calculations@xcite ( e.g. : af1 for pd@xmath19 and pd@xmath20 and af2 for pd@xmath21 ) .
the non - collinear solutions found can be classified into those that release antiferromagnetic frustration and therefore have lower excitation energy than the af solution ( nc1 and nc2 in pd@xmath19 , shown in table iii and fig .
3 ) , and radial or quasi - radial solutions , that resemble the hedgehogs found in low dimensional theories of classical or quantum antiferromagnets@xcite . hedgehogs in these theories do not release frustration but rather are excitations over the antiferromagnetic ground state .
we also find that these radial states have a higher energy that the antiferromagnetic solution , and therefore do not release frustration .
notice that the antiferromagnetic and non - collinear solutions can be reached at temperatures of the order of room temperatures ( 25 mev ) .
therefore , any measurement of the magnetization performed at room temperature should find a thermal average of all those states , many of which have a tiny magnetic moment .
it should not be surprising that such a measurement give a small net moment .
we finally discuss the relationship between magnetism and equilibrium interatomic distances .
we have found that these are essentially the same regardless of the magnetic state for the largest clusters ( @xmath3= 5 - 7 ) , the smallest ones showing slight variations of less than 0.04 , but only within the lda solutions .
we have additionally analyzed the relative stability of the different solutions as a function of the interatomic distance . to this aim ,
we plot in fig .
4 the energy per atom of the low - lying excited states of the pd@xmath20 cluster , relative to the ground state energy , as a function of an uniform volume expansion , obtained using gga1 .
the figure shows that no crossover takes place , apart from the nearly - degenerate af1 and af2 solutions , that cross at an expansion of about 4% .
moreover , the relative energy differences are essentially preserved and the local magnetic moments kept constant , except for the af2 and radial solutions , where they slightly change ( by about 10% ) .
to summarize , we have studied the geometry and magnetic properties of the ground state and lowest lying isomers of small palladium clusters pd@xmath2 , with @xmath3 ranging from 3 to 7 .
our results confirm that the ground state is indeed collinear or paramagnetic .
we have found a rich variety of non - collinear low - lying isomers , some of which efficiently release frustration , while other ( hedgehog - like solutions ) do not .
all these solutions should contribute to the room temperature magnetic behavior of the clusters , probably rendering small measured magnetic moments .
we have finally found that simulations of atomic clusters are rather insensitive to the choice of the pseudopotential and to the approximation used for the exchange and correlation potential . to provide a rough explanation of this contrasting behavior
, we first note that gga and lda provide results that are quantitatively very similar in most cases , including bulk palladium .
gga provides results that are qualitatively different from lda for bulk palladium because this material sits right at the edge of the paramagnetic - ferromagnetic transition , actually tilted slightly towards the paramagnetic side .
therefore , bulk palladium falls slightly short of fulfilling stoner s criterion .
gga and lda are mean field approximations .
hence , they tend to overstimate magnetic properties .
this error is partly compensated by lda , that predicts usually lower magnetic moments than gga .
consequently , lda predicts the correct ground state for bulk palladium , while gga provides a ground state that is slightly tilted towards the magnetic side of the phase diagram .
hence , lda and gga predict quantitatively similar , but qualitatively different results for the magnetic properties of this material .
atomic pd clusters are strongly magnetic , with the exception of pd@xmath16 .
therefore , all these approximations predict quantitative and qualitative similar results for those clusters .
this work was supported by the spanish ministerio de educacion y ciencia ( grants mat2005 - 03415 , bfm2003 - 03156 and sab2004 - 0129 ) in conjuction with the european regional development found , intas ( project 03 - 51 - 4778 ) and the mexican ( promec - sep - ca-230 and conacyt - sni ) .
we wish to acknowledge useful conversations with j. m. montejano - carrizales , l. fernndez - seivane , v. m. garca - surez and f. yndurain .
|
we present a thorough theoretical assessment of the stability of non - collinear spin arrangements in small palladium clusters .
we generally find that ferromagnetic order is always preferred , but that antiferromagnetic and non - collinear configurations of different sorts exist and compete for the first excited isomers .
we also show that the ground state is insensitive to the choice of atomic configuration for the pseudopotential used and to the approximation taken for the exchange and correlation potential .
moreover , the existence and relative stability of the different excited configurations also depends weakly on the approximations employed .
these results provide strong evidence on the transferability of pseudopotential and exchange and correlation functionals for palladium clusters as opposed to the situation found for the bulk phases of palladium .
= 10000
|
at extreme high temperature and/or chemical potential , hadronic matter undergoes a phase transition where color degrees of freedom plays important role , known as quark gluon plasma ( qgp ) .
the qcd based model calculations claim that at vanishing baryon chemical potential ( @xmath4 ) the transition is simply a crossover @xcite . at large @xmath4 ,
the phase transition is of first order @xcite .
the presence of these two types of transition , at two extreme limit of @xmath4 , urges to have an end point in the first order phase transition in qcd phase diagram , known as the qcd critical point ( cp ) . to probe the cp and
map the phase boundary in phase diagram , the beam energy scan program @xcite has been undertaken with successful data taking in year 2010 and 2011 . the star experiment has collected data for au+au collisions at @xmath0= 7.7 to 200 gev covering large range of @xmath4= 20 to 410 mev @xcite in the phase diagram . + the various qcd
based models reveal , at the cp , the susceptibilities and correlation length ( @xmath5 ) of the system diverge @xcite . the higher order susceptibilities and higher power of the correlation length
amplify the signal of divergence at the cp .
the first four moments , such as mean ( @xmath6 ) , standard deviation ( @xmath7 ) , skewness ( @xmath8 ) and kurtosis ( @xmath9 ) of the conserved charged distributions like , net - charge , net - proton and net - strangeness are very closely related to these respective higher order susceptibilities @xcite and higher power of the correlation length @xcite . for example , the variance , skewness and kurtosis have been shown to be related to the @xmath10 , @xmath11 and @xmath12 , respectively @xcite . in order to cancel the volume term in the susceptibilities , different combinations of the various moments
are constructed , some of these are : @xmath13 , @xmath2 and @xmath3 . these products , as a function of beam energy , are expected to show non - monotonic behavior near the cp . recently in the star experiment , the analysis has been made on the higher moments of the net - charge , net - proton and net - kaon @xcite for the beam energy scan program to probe the cp in phase diagram . beside the cp search , recent lattice qcd model calculations @xcite motivate to extract the freeze - out parameter like , temperature and chemical potential , from the experimentally measured higher moments of the net - charge multiplicity distribution .
the star experiment has taken data for au+au collisions at @xmath0= 7.7 to 200 gev in last two years .
the star detector system provides the excellent particle identification and large acceptance for the event - by - event fluctuation analysis .
the time projection chamber ( tpc ) is the main tracking device .
extensive quality assurance is performed for each energy in order to minimize the fluctuation of detector efficiency .
the event - by - event net - charge distribution is measured for au + au collisions occurring within @xmath1430 cm along the @xmath15 position of the interaction point from the tpc center and 2.0 cm radius in the transverse plane .
the charged particles are measured between the transverse momentum ( @xmath16 ) range 0.2@xmath17@xmath16 @xmath182.0 gev/_c _ and pseudo - rapidity ( @xmath19 ) range at @xmath20 region .
the standard star track quality cuts are used for this analysis .
the contamination of spallation protons , produced due to beam pipe interaction , are suppressed by removing protons within 200 @xmath21@xmath16 @xmath21 400 mev/_c_. to avoid the auto - correlation effect in the higher moments analysis , the centrality selection has been done by uncorrected charged particles measured within @xmath22 from the tpc detector .
the heavy ion collision geometry of each event is characterized by the number of participating nucleons ( @xmath23 ) .
the average number of participant ( @xmath24 ) for each centrality bin is calculated by the monte carlo glauber simulation .
the finite centrality bin width effect has been avoided by using centrality bin width correction @xcite . for the statistical error estimation
, delta theorem is used @xcite .
= 7.7 to 200 gev .
, scaledwidth=80.0% ]
the raw event - by - event net - charge multiplicity distributions are plotted in fig .
[ fig1 ] for three centralities , 0 - 5% , 30 - 40% and 60 - 70% for above seven colliding energies .
it is observed that the width of net - charge multiplicity distributions increases with increase in centrality and mean of the net - charge multiplicity distributions shift towards positive side with decreasing colliding energies . in fig .
[ fig2 ] , the moments such as , m , @xmath7 , @xmath8 and @xmath9 for the net - charge distributions are plotted as function of @xmath24for the above seven colliding energies .
the m and @xmath7 values increase in going from peripheral to central collisions for the above seven colliding energies , whereas @xmath8 and @xmath9 values decrease with increase in collision centrality .
the evolution of these higher moments can be understood by the central limit theorem ( clt ) @xcite , which explain the @xmath24(proxy of volume ) dependence of these moments . for au+au collisions from @xmath0=7.7 to 200 gev .
the dotted lines represent the expectation form clt.,width=340 ] are plotted as a function of @xmath24 for au+au collisions from @xmath0=7.7 to 200 gev .
the dotted lines represent the poisson expectation from the data.,width=359 ] are plotted as a function of @xmath24 for au+au collisions from @xmath0=7.7 to 200 gev .
the dotted lines represent the poisson expectation from the data.,width=359 ] for the non - cp baseline at the above energies , the poisson expectation is studied . in this approach ,
the positive and negative charged particles distributions are assumed to be independently poisson distribution where no dynamical correlation among the positive and negative charge particles are taken into account .
hence , the net - charge distribution is taken as a skellam distribution ( resultant of the difference of two poisson distributions ) which serve as one of the baseline for this analysis .
for the skellam distribution , @xmath25 , @xmath26 and @xmath27 .
here @xmath28 and @xmath29 are the means of the positive and negative charged particle distribution .
the results from the poisson expectations are discussed and compared below .
( a ) , @xmath2 ( b ) and @xmath30 ( c ) are plotted as a function of the colliding energy for three centrality bins 0 - 5% , 30 - 40% and 60 - 70% .
the blue shaded region represent the poisson expectation , reflects the range covering 0 - 5% to 60 - 70% centrality .
the red dotted lines are the hrg predictions.,width=241 ] in fig .
[ fig3 ] and [ fig4 ] , the values of @xmath2 and @xmath31 are plotted as as function of @xmath24for the above seven energies and corresponding poisson expectations are also plotted for each energy .
the deviation between @xmath2 and that of poisson expectation are observed to be increasing as decreasing colliding energy . whereas in @xmath3 , deviation from poisson expectations are observed almost similar in all energies . in fig .
[ fig5 ] , the energy dependence of the product of moments such as , @xmath32 , @xmath2 and @xmath3 are plotted along with poisson expectation and hadron resonance gas ( hrg ) model predictions .
the values of @xmath32 increases with increase in colliding energy for three centrality bins whereas in case of the @xmath2 , it increases with decreasing colliding energies .
the @xmath30 shows no energy dependence and all the values are above unity , except for most central events at 7.7 gev with large statistical uncertainty .
the energy dependence of the @xmath2 shows systematically large deviation from its poisson expectation below @xmath0=27 gev as compared at higher energies .
the hrg predictions of @xmath30 are very close to data whereas that of @xmath2 over - predicts the data .
the higher moments of the net - charge multiplicity distributions have been measured for au+au collisions at @xmath0= 7.7 to 200 gev .
the centrality dependence of the moments follows the expectation from the clt .
the values of @xmath32 increase with increase in colliding energy .
the values of @xmath2 increase with decreasing colliding energies and deviates from poisson expectation below @xmath0= 27 gev . within statistical uncertainty
, @xmath3 is seen to be independent of collision energy .
|
we report the higher order moments of the net - charge multiplicity distributions for the au+au collisions at @xmath0= 7.7 , 11.5 , 19.6 , 27 , 39 , 62.4 and 200 gev in the star experiment at the relativistic heavy - ion collider ( rhic ) .
the energy and centrality dependence of higher moments and their products ( such as @xmath1 , @xmath2 and @xmath3 ) are presented .
the data are also compared to poisson expectations and hadron resonance gas model calculations .
|
some even - even nuclei are energetically forbidden to decay via single @xmath12 emission , while the decay via emission of two electrons and two neutrinos is energetically allowed .
the experimentally observed neutrino accompanied double beta ( @xmath13 ) decay is a second order weak process with half lives of the order of 10@xmath14 yr @xcite .
the decay process without neutrino emission , neutrinoless double beta ( @xmath0 ) decay , is of fundamental relevance as its observation would imply lepton number violation indicating physics beyond the standard model of particle physics .
the gerda experiment @xcite is designed to search for @xmath0 decay in the isotope @xmath1ge .
this process is identified by a monoenergetic line in the energy sum spectrum of the two electrons at 2039 kev @xcite , the @xmath2-value of the decay .
the two precursor experiments , the heidelberg moscow ( hdm ) and the international germanium experiment ( igex ) , have set limits on the half live @xmath15 of @xmath0 decay @[email protected]@xmath17 yr @xcite and @[email protected]@xmath17 yr @xcite ( 90% c.l . ) , respectively .
a subgroup of the hdm experiment claims to have observed @xmath0 decay with a central value of the half life of @xmath15=1.19@xmath17 yr @xcite .
this result was later refined using pulse shape discrimination ( psd ) @xcite yielding a half life of @xmath15=2.23@xmath17 yr .
several inconsistencies in the latter analysis have been pointed out in ref .
@xcite .
the design of the gerda apparatus for the search of @xmath0 decay follows the suggestion to operate high purity germanium ( hpge ) detectors directly in a cryogenic liquid that serves as cooling medium and simultaneously as ultra - pure shielding against external radiation @xcite .
gerda aims in its phase i to test the hdm claim of a signal and , in case of no confirmation , improve this limit by an order of magnitude in phase ii of the experiment .
prerequisites for rare - event studies are @xmath18 extremely low backgrounds , usually expressed in terms of a background index ( bi ) measured in cts/(kev@xmath3kg@xmath3yr ) , and @xmath19 large masses and long measuring times , expressed as exposure . reducing
the background and establishing a radio - pure environment is an experimental challenge .
proper analysis methods must be applied to guarantee an unbiased analysis .
the gerda collaboration has blinded a region of @xmath2@xmath2020 kev during the data taking period @xcite . during this time , analysis methods and background models have been developed and tested .
the latter is described in this paper together with other parameters demonstrating the data quality .
the raw data are converted into energy spectra . if the energies of individual events fall within a range @xmath2@xmath2120 kev , these events are stored during the blinding mode in the backup files only .
they are not converted to the data file that is available for analysis .
this blinding window is schematically represented in fig .
[ fig : windows ] by the yellow area , including the red range .
after fixing the calibration parameters and the background model , the blinding window was partially opened except the peak range at @xmath2 , indicated in red in fig .
[ fig : windows ] .
the blue range covers the energies from 100 kev to 7.5 mev .
the data from this energy range were available for analysis all the time .
the observable @xmath11 lines can be used to identify background sources . a range between 1930 kev and 2190 kev
was then used to determine the bi .
the energy regions around significant @xmath11 lines are excluded in the latter , as shown schematically in fig [ fig : windows ] .
representation of energy spectra for definition of the energy windows used in the blind analysis . ]
data were taken until until may 2013 .
these data provide the exposure for phase i. the data used in this analysis of the background are a subset containing data taken until march 2013 . the extraction of the background model is described in detail in this paper . in the process , the necessary parameters are defined for the upcoming @xmath0 analysis .
an important feature is the stable performance of the germanium detectors enriched in @xmath1ge ; this is demonstrated for the complete data taking period ( sec .
[ sec : exp ] ) . the data with exposure is used to interpolate the background within @xmath22 .
the expectation for the bi is given in this paper before unblinding the data in the energy range @xmath22 , the region of highest physics interest .
the paper is organized as follows : after presenting the experimental details , particularly on the detectors used in phase i of the gerda experiment , coaxial and bege type ( sec . [
sec : exp ] ) , the spectra and the identified background sources will be discussed ( sec .
[ sec : spectradata ] and [ sec : back ] ) .
these are the basic ingredients for the background decomposition for the coaxial detectors ( sec .
[ sec : modeling ] ) and for the bege detectors ( sec .
[ sec : begemodel ] ) .
the models work well for both types of detectors . after cross checks of the background model ( sec .
[ sec : cross - checks ] ) the paper concludes with the prediction for the background at @xmath2 and the prospective sensitivity of gerda phase i ( sec . [ sec : extrapol ] ) .
this section briefly recalls the main features of the gerda experiment .
the main expected background components are briefly summarized . due to the screening of the components before installation
, the known inventory of radioactive contaminations can be estimated .
finally , the stable performance of the experiment is demonstrated and the data selection cuts are discussed .
l|l||c||c|r||r detector & & @xmath23 & @xmath24 & @xmath25 & + & & & g & & + + ang 1 @xmath26 ) & 0.859(29 ) & 958 & 795(50 ) & 0.830(52)&1.8(5 ) + ang 2 & 0.866(25 ) & 2833 & 2468(145 ) & 0.871(51)&2.3(7 ) + ang 3 & 0.883(26 ) & 2391 & 2070(136 ) & 0.866(57)&1.9(7 ) + ang 4 & 0.863(13 ) & 2372 & 2136(135 ) & 0.901(57)&1.4(7 ) + ang 5 & 0.856(13 ) & 2746 & 2281(132 ) & 0.831(48)&2.6(6 ) + rg 1 & 0.855(15 ) & 2110 & 1908(125 ) & 0.904(59)&1.5(7 ) + rg 2 & 0.855(15 ) & 2166 & 1800(115 ) & 0.831(53)&2.3(7 ) + rg 3 @xmath26 ) & 0.855(15 ) & 2087 & 1868(113 ) & 0.895(54)&1.4(7 ) + + ' '' '' gd32b & 0.877(13 ) & 717 & 638(19 ) & 0.890(27)&1.0(2 ) + gd32c & 0.877(13 ) & 743 & 677(22 ) & 0.911(30)&0.8(3 ) + gd32d & 0.877(13 ) & 723 & 667(19 ) & 0.923(26)&0.7(2 ) + gd35b & 0.877(13 ) & 812 & 742(24 ) & 0.914(29)&0.8(3 ) + gd35c @xmath26 ) & 0.877(13 ) & 635 & 575(20 ) & 0.906(32)&0.8(3 ) + + gtf 32 @xmath26 ) & 0.078(1 ) & 2321 & 2251(116 ) & 0.97(5 ) & 0.4(8 ) + gtf 45 @xmath26 ) & 0.078(1 ) & 2312 & & & + gtf 112&0.078 ( 1 ) & 2965 & & & + @xmath26 ) not used in this analysis the setup of the gerda experiment is described in detail in ref .
gerda operates high purity germanium ( hpge ) detectors made from material enriched to about 86% in @xmath1ge in liquid argon ( lar ) which serves both as coolant and as shielding . a schematic view is given in fig .
[ fig : exp ] . a stainless steel cryostat filled with
64 m@xmath27 of lar is located inside a water tank of 10 m in diameter
. only very small amounts of lar are lost as it is cooled via a heat exchanger by liquid nitrogen .
the 590 m@xmath27 of high purity ( @xmath160.17 m@xmath28 m ) water moderate ambient neutrons and @xmath11 radiation .
it is instrumented with 66 photo multiplier tubes ( pmt ) and operates as a cherenkov muon veto to further reduce cosmic induced backgrounds to insignificant levels for the gerda experiment .
muons traversing through the opening of the cryostat without reaching water are detected by plastic scintillator panels on top of the clean room .
schematic drawing of the main components of the gerda experiment .
for details see ref .
, height=321 ] three coaxial or five bege detectors are mounted into each of the four strings which are lowered through a lock separating the clean room from the cryostat .
the detector strings with coaxial detectors are housed in 60 @xmath29 m thin - walled copper containers permeable to lar - called mini shroud in the following - with a distance of a few mm from the detector outer surfaces .
a 30 @xmath29 m thin copper cylinder - called radon shroud in the following - with a diameter of 75 cm encloses the detector array . a picture of a detector string can be found in @xcite .
the custom made preamplifiers are operated in lar at a distance of about 30 cm from the top of the detector array .
the analog signals are digitized by 100 mhz fadcs .
all eight of the reprocessed coaxial germanium detectors from the hdm and the igex experiments @xcite were deployed on november 9 2011 , together with three detectors with natural isotopic abundance .
a schematic drawing of the coaxial detector type is shown in fig .
[ fig : detector_schematic ] , top .
two enriched detectors ( ang 1 and rg 3 ) developed high leakage currents soon after the start of data taking and were not considered in the analysis . rg 2 was taking data for about one year before it also had to be switched off due to an increase of its leakage current . in july 2012 ,
two of the low background coaxial hpge detectors with natural isotopic abundance , gtf 32 and gtf 45 , were replaced by five enriched broad energy germanium ( bege ) detectors , which follow the phase ii design of gerda ( see fig .
[ fig : detector_schematic ] , bottom ) .
the geometries and thus the pulse shape properties of the two types of detectors differ as discussed in ref .
one of these bege detectors ( gd35c ) showed instabilities during data taking and was not used for further analysis .
the most relevant properties of all the germanium detectors are compiled in table [ tab : recval ] .
note , that the numbers for dead layers @xmath30 are to be interpreted as effective values , because their determination by comparison of count rates and monte carlo ( mc ) predictions depends on the precision of the model and the geometries @xcite .
schematic sketch of a coaxial hpge detector ( top ) and a bege detector ( bottom ) with their different surfaces and dead layers ( drawings not to scale ) . ] an important source of background is induced by cosmic radiation .
muon induced background events are efficiently vetoed by identification of cherenkov light emitted by muons when they pass the water tank . the number of long lived cosmogenically produced isotopes , especially @xmath31ge and @xmath8co are minimized by minimization of the time above ground during processing of the detectors and the structural materials .
further contributions stem from radioactivity included in the detector and structural materials or the surrounding environment , i.e. the rocks of the laboratory .
the selection of materials has been described in @xcite .
the most important activities are discussed in the next section .
background from @xmath5ar present in lar was found during gerda commissioning to be more significant than anticipated .
the @xmath12 decay of its progeny @xmath5k can contribute to the background at @xmath2 if the decay happens near detector surfaces .
for gerda phase i coaxial detectors this background was significantly reduced by implementation of the mini shrouds .
however , for the bege detectors this remains an important background due to their thinner surface n@xmath32 dead layer .
another source of background stems from the calibration sources that have a typical initial activity of about 10 - 20 kbq .
when in parking position they are well shielded .
due to an accident during commissioning the experiment , one 20 kbq @xmath7th calibration source fell to the bottom of the cryostat .
the bi expected from this source is around @xmath4 cts/(kev@xmath3kg@xmath3yr ) , thus , significantly less then the phase i bi goal .
hence , the calibration source was left inside the lar cryostat .
it will be removed during the upgrade of the experiment to its second phase .
a significant fraction of the background is induced by contaminations of bulk materials and surfaces with nuclei from the @xmath33u and @xmath34th decay chains .
the @xmath33u decay chain can be subdivided into three sub decay chains : @xmath33u to @xmath10ra , @xmath10ra to @xmath35pb and @xmath35pb to @xmath36pb , due to isotopes with half lives significantly longer than the live time of the experiment .
only the two latter sub decay chains are relevant in the following .
the noble gas @xmath37rn ( t@xmath38=3.8 days ) plays a special role , as it can further break the @xmath10ra to @xmath35pb chain due to its volatility .
whenever activities of @xmath6bi are quoted it is assumed that the chain is in secular equilibrium between @xmath10ra and @xmath35pb inside metallic materials , while for non metallic materials the equilibrium can be broken at @xmath37rn .
llcccccc component & units & @xmath39k & @xmath6bi&@xmath10ra & @xmath7th & @xmath8co & @xmath37rn & bi + & & & & & & & @xmath4 cts/(kev@xmath3kg@xmath3yr ) + & + copper det .
support&@xmath29bq / det .
& @xmath407 & @xmath401.3 & @xmath401.5 & & & @xmath400.2 + ptfe det .
support & @xmath29bq / det . & 6.0 ( 11 ) & 0.25 ( 9)&0.31 ( 14 ) & & & 0.1 + ptfe in array & @xmath29bq / det & 6.5 ( 16 ) & 0.9 ( 2 ) & & & & 0.1 + mini shroud & @xmath29bq / det . & & 22 ( 7 ) & & & & 2.8 + li salt & mbq / kg & & 17(5 ) & & & & @xmath41 0.003@xmath42 + + cc2 preamps & @xmath29bq / det . & 600 ( 100 ) & 95 ( 9 ) & 50 ( 8) & & & 0.8 + cables and + suspension & mbq / m & 1.40 ( 25)&0.4 ( 2 ) & 0.9 ( 2 ) & 76 ( 16 ) & & 0.2 + + cryostat & mbq & & & & & 54.7 ( 35 ) & @xmath400.7 + copper of cryostat & mbq & @xmath40784 & 264 ( 80 ) & 216 ( 80 ) & 288 ( 72 ) & & + steel of cryostat & kbq & @xmath4072 & @xmath4030 & @xmath4030 & 475 & & + lock system & mbq & & & & & 2.4 ( 3 ) & @xmath400.03 + @xmath7th calib . source & kbq & & & 20 & & & @xmath401.0 + @xmath26 ) value derived for 1 mg of li salt absorbed into the surface of each detector + the hardware components close to the detectors and the components of the suspension system have been tested for their radio - purity prior to installation @xcite .
the hardware parts at close ( up to 2 cm ) and medium ( up to 30 cm ) distance from the detectors have been screened using hpge screening facilities or icp - ms measurements , while the parts in the lock system have been tested for @xmath37rn emanation @xcite .
some materials proved to have low , but measurable , radioactive contaminations .
table [ tab : thorium ] quotes the total measured activities and limits of the most significant screened components and their expected contribution to the bi close to @xmath2 . as the @xmath37rn emanation rate in the cryostat with its copper lining and
the lock system is on the order of 60 mbq , some @xmath6bi may be expected in the lar surrounding of the detectors . assuming a homogeneous distribution of @xmath37rn in the lar
, this would result in a contribution to the bi at @xmath2 of 7@xmath3@xmath43 cts/(kev@xmath3kg@xmath3yr ) . to reduce this latter contribution to the background
, the radon shroud was installed around the array with the intention to keep @xmath37rn transported by lar convection at sufficient distance from the detectors . additionally , li salt that is used to dope n@xmath32 surfaces of the detectors was screened .
it is not precisely known how much li diffuses into the crystal . a rough estimation assuming an n@xmath32 li doping of 10@xmath44 li nuclei per @xmath45 germanium results in an overall li weight per detector of @xmath415@xmath46 g which leads to negligible background contributions even if it is assumed that the @xmath10ra contamination diffuses into the germanium with the same efficiency as li .
the measured activities in the hardware components within 2 cm from the detectors lead to a total contribution to the bi of @xmath413@xmath3@xmath4 cts/(kev@xmath3kg@xmath3yr ) using efficiencies obtained by mc simulations @xcite . from the medium distance contributions @xmath41@xmath4
cts/(kev@xmath3kg@xmath3yr ) are expected , while the far sources contribute with @xmath40 @xmath4 cts/(kev@xmath3kg@xmath3yr ) . as detailed in ref .
@xcite the extrapolated background rates for all contaminations were predicted to be tolerable for phase i and to yield a bi of @xmath40@xmath47 cts/(kev@xmath3kg@xmath3yr ) .
live time fraction of the data acquisition for the muon veto ( top ) and for the hpge detectors ( bottom ) the spikes in the live time fraction arise from the regular calibration measurements .
the development of the exposure of the enriched detectors ( bottom ) and the total live time of the muon veto system ( top ) is also shown .
the red vertical line indicates the end of the data range for the evaluation of the background model .
, title="fig : " ] + live time fraction of the data acquisition for the muon veto ( top ) and for the hpge detectors ( bottom ) the spikes in the live time fraction arise from the regular calibration measurements .
the development of the exposure of the enriched detectors ( bottom ) and the total live time of the muon veto system ( top ) is also shown .
the red vertical line indicates the end of the data range for the evaluation of the background model .
, title="fig : " ] the muon veto system started operation in december 2010 and ran up to may 21 2013 , when the data taking for the @xmath0 analysis was stopped .
its stable performance is shown in the top graph of fig .
[ fig : live ] .
the interruptions were due to the test and installation of the plastic panel in april / may 2011 and due to short calibrations .
the probability that a muon induced event in the detectors is accompanied by a signal in the veto ( overall muon rejection efficiency ) is @xmath48 , reducing the contribution of the muons to the bi to @xmath4010@xmath49 cts/(kev@xmath3kg@xmath3yr ) @xcite .
no evidence for delayed coincidences between @xmath46 veto events and germanium events was found . the bottom graph in fig .
[ fig : live ] demonstrates the live time fraction of data taking .
the interruption in may 2012 was due to temperature instabilities in the gerda clean room , while the interruption in july 2012 was due to the insertion of five phase ii type bege detectors .
the analysis presented here considers data taken until march 3 , 2013 , corresponding to a live time of 417.19 days and an exposure of = 16.70 kg@xmath3yr for the coaxial detectors ; the four bege detectors acquired between 205 and 230 days of live time each , yielding a total exposure of = 1.80 kg@xmath3yr .
the end of run 43 in march is marked by the red vertical line in fig [ fig : live ] , bottom .
the data have been processed using algorithms and data selection procedures @xcite implemented in the gerda software framework @xcite .
a set of quality cuts , described in detail in ref .
@xcite , is applied to identify and reject unphysical events , e.g. generated by discharges or by electromagnetic noise .
the cuts take into account several parameters of the charge pulse , such as rise time , baseline fluctuations and reconstructed position of the leading edge .
the cuts also identify events having a non - flat baseline , e.g. due to a previous pulse happening within a few hundreds of @xmath29s .
moreover , events in which two distinct pulses are observed during the digitization time window ( 80 @xmath29s ) are marked as pile - up and are discarded from the analysis . from the total number of triggers roughly 91% are kept as physical events . due to the very low counting rate ,
the gerda data set has a negligible contamination of accidental pile - up events and the selection efficiency for genuine @xmath0 events is hence practically unaffected by the anti pile - up cuts .
similarly , the loss of physical events above 1 mev due to mis - classification by the quality cuts is less than 0.1% .
the linearity and the long term stability of the energy scale as well as the energy resolution given as full width at half maximum ( fwhm ) were checked regularly with @xmath7th sources . between calibrations
the stability of the gain of the preamplifiers was monitored by test pulses induced on the test inputs of the preamplifiers .
whenever unusual fluctuations on the preamplifier response were observed , calibrations were performed .
the linearity of the preamplifier has been checked using test pulses up to an energy range of 6 mev .
it was found that between 3 and 6 mev the calibration has a precision of better than 10 kev ; above 6 mev some channels exhibit larger non - linearity .
physical events passing the quality cuts are excluded from the analysis if they come in coincidence within 8 @xmath29s with a valid muon veto signal ( muon veto cut ) or if they have energy deposited in more than one hpge detector ( anti - coincidence cut ) .
the anti - coincidence cut does not further affect the selection efficiency for @xmath0 decays , since only events with full energy deposit of 2039 kev are considered .
the dead time induced by the muon veto cut is practically zero as the rate of ( [email protected])@xmath310@xmath50 /s of events coincident between germanium detectors and the gerda muon veto system is very low .
the stability of the energy scale was checked by the time dependence of the peak position for the full energy peak at 2614.5 kev from the @xmath7th calibration source .
the maximal shifts are about 2 kev with the exception of 5 kev for the gd32b detector .
the distributions of the shifts are fitted by a gaussian with fwhm amounting to 1.49 kev for the coaxial and to 1.01 kev for the bege detectors .
the respective uncertainties are smaller than 10% .
the shifts are tolerable compared to the energy resolution . to obtain the energy resolution at @xmath2 the results from the calibration measurements are interpolated to the energy @xmath2 using the standard expression fwhm=
@xmath51 @xcite .
the energy resolution during normal data taking is slightly inferior to the resolution during calibration measurements .
the resulting offset was determined by taking the difference between the resolution of the @xmath5k line and the interpolated resolution determined from calibrations .
the scaled offset is added to the resolution at @xmath2 expected from calibration measurements .
the fwhm of all enriched detectors at 2614.5 kev is determined to be between 4.2 and 5.8 kev for the coaxial detectors and between 2.6 and 4.0 kev for the beges .
the resolutions are stable in time to within 0.3 kev for the beges and to within 0.2 kev for the coaxial detectors .
the resolutions of all relevant enriched detectors are shown in table [ tab : fwhm ] . .
[ tab : fwhm ] energy resolutions ( fwhm ) in kev of the enriched detectors at @xmath2 . for definition of the data sets
see sec .
[ sec : datasets ] . [ cols="<,<,<,<",options="header " , ] source distance : @xmath52 close ( @xmath402 cm ) ; @xmath53 medium ( 2 - 30 cm ) ; @xmath54 far ( @xmath1630 cm ) + @xmath26 ) prior : discussed in the text + @xmath55 ) single : obtained from coincident spectrum with histogram entries for each detector event separately .
a comparison of the resulting activities in table [ tab : fits_results_golden ] with the known inventory of radio contaminations shown in table [ tab : thorium ] shows that all contaminations expected from screening are seen in the background spectra .
however , the activities identified by screening measurements are not sufficient to explain the total background seen .
the minimum model describes the background spectrum well without any medium distance and distant contaminations .
also if medium distance and distant sources are added , the largest fraction to the background comes from close sources , especially on the p@xmath32 and n@xmath32 surfaces .
note , that the activity obtained for @xmath5k and hence for the @xmath5ar contamination of lar is higher than the previously most stringent limit reported in ref .
@xcite . in the maximum model
, strong correlations are found between several background sources .
contaminations of @xmath5k on the n@xmath32 surface and in lar can not be distinguished .
similarly , the model has no distinction power between contaminations of the radon shroud , the heat exchanger and the detector assembly with @xmath6bi , @xmath7th and @xmath7ac .
this explains the differences of the derived activities in the two models .
the main difference between minimum and maximum models is the number of events on the p@xmath32 surface of the detectors .
a fit of the background model has also been made to the _ silver - coax _ data set .
its overall spectral shape can only be described sufficiently well , if either an additional @xmath5k contamination of the p@xmath32 surface and/or an additional @xmath6bi contamination of the lar is assumed . both additional contaminations seem plausible after a modification of the experimental surrounding like the insertion of bege detectors to the cryostat .
an equivalent procedure as for the coaxial detectors was used to model the energy spectrum observed for the _ sum bege _ data set . since the exposure collected with the bege detectors is much smaller than for the coaxial data set , only a qualitative analysis is possible for this data set .
the lower mass of the bege detectors with respect to the coaxial detectors reduces the detection efficiency for full energy peaks .
hence , fewer @xmath11 lines are positively identified in the bege spectrum .
this makes it even more difficult to establish and to constrain possible background components .
the contributions to the bege background model were simulated using an implementation of the gerda phase i detector array containing the three coaxial detector strings and an additional string with the five bege detectors .
the n@xmath56 dead layer thicknesses used in the mc are listed in table [ tab : recval ] .
the effective p@xmath56 dead layer thickness was set to 600 nm .
the minimum model contributions were considered .
additionally , two contributions were added to the bege model : @xmath31ge decays in germanium and @xmath5k decays on the n@xmath32 surface . a contribution from @xmath31ge
is expected due to cosmogenic activation above ground , analogously to @xmath8co in germanium . due to the rather short half life of 271
d the @xmath31ge contribution can be neglected for the coaxial detectors , which have been stored underground for several years . for the newly produced bege detectors , however , these decays and the subsequent decay of @xmath31ga have to be taken into account .
the contribution from @xmath5k decays on the n@xmath32 surface , on the other hand , is enhanced with respect to the coaxial detectors due to the thinner dead layer and has to be taken into account for the model .
the n@xmath32 surface dead layer is partially active @xcite , which in particular affects the detection efficiency for surface @xmath12 interactions .
thus , the mc simulation used for @xmath5k on n@xmath32 surface included an approximation of this effect .
40% of the dead layer thickness ( as stated in tab .
1 ) has been modeled with zero charge collection efficiency , the other part with a linearly increasing charge collection efficiency from 0% to 100% .
the contributions of @xmath8co and @xmath31ge to the model are limited to 0.05 cts / day and 0.32 cts / day , respectively .
the upper values for these cosmogenically produced isotopes are derived from an assumed activation rate for these isotopes according to ref .
@xcite and the known histories of exposure to cosmic rays of the individual detectors .
the procedure to obtain the best fit was equivalent to the model definition of the coaxial detectors .
the best fit model for bege detectors is shown in fig .
[ fig : bege_best_fit ] . around @xmath2 the largest contribution arise from @xmath5k on the n@xmath32 surfaces ( see last column of table [ tab : background_components ] ) .
the presented bege background model is consistent with a background decomposition obtained by pulse shape discrimination of the data @xcite .
the background model developed has some predictive power that can be checked with the available data .
this section describes cross checks performed on the background model . from the best fit models the resulting half life @xmath57 for @xmath13 decay can be extracted .
@xmath57 is calculated using the relation @xmath58 where @xmath59 is the best fit number of @xmath13 decays derived from the individual model .
the efficiency @xmath60 is given by the weighted detection efficiency of @xmath13 decays in the fit range , @xmath61 , @xmath62 table [ tab:2vbb_half_lives ] gives the half lives extracted from the different background models .
all results are consistent with the earlier @xmath13 analysis @xcite within the uncertainties .
note , that a three times larger exposure was available for this analysis as compared to the analysis in ref .
@xcite while systematic uncertainties are not considered .
lrl model & [ kg@xmath3yr ] & @xmath57@xmath63yr + ' '' '' _ gold - coax _ minimum & 15.40 & 1.92@xmath64 + ' '' '' _ gold - coax _ maximum & 15.40 & 1.92@xmath65 + ' '' '' _ gold - nat _ minimum & 3.13 & 1.74@xmath66 + ' '' '' _ sum - bege _ & 1.80 & 1.96@xmath67 + ' '' '' analysis in ref .
@xcite & 5.04 & 1.84@xmath68 + at energies below 600 kev the energy spectrum is dominated by @xmath69ar with an activity of @xmath70=[[email protected](stat ) @xmath210.08(syst ) ] bq / kg @xcite homogeneously distributed in lar .
this part of the spectrum has not been included into the background fit to avoid uncertainties due to the n@xmath32 dead layer thickness and the theoretical shape of the beta decay spectrum . a strong @xmath11 line at 352 kev is , however , expected from decays of @xmath6bi in the vicinity of the detectors .
the intensity of this line depends strongly on the distance of the @xmath6bi contamination from the detectors .
hence , this cross check can give a hint on how realistic the assumed distribution of the @xmath6bi contamination is .
the minimum ( maximum ) model predicts 20.1@xmath71 ( 17.5@xmath72 ) counts/(kg@xmath3yr ) in the peak while a fit of a gaussian plus a linear background to the data gives 20.4@xmath73 counts/(kg@xmath3yr ) for the _ gold - coax _ data set .
[ fig:352kev ] shows the energy spectrum of the _ gold - coax _ data set in the energy region between 310 and 440 kev .
the gaussian plus linear background fit to the data as well as the minimum model prediction without the @xmath69ar contribution dominating the spectrum in this energy region is shown .
for the _ gold - nat _ data set the minimum model prediction of ( 22.1@xmath74 counts/(kg@xmath3yr ) is also consistent with the observed ( 25.6 @xmath75 6.2 ) counts/(kg@xmath3yr ) .
this cross check makes it possible to distinguish between the locations of @xmath6bi contaminations if it is assumed that the decays of @xmath6bi and @xmath6pb happen at the same location .
it excludes the results for @xmath6pb contamination in or on the radon shroud as the best fit maximum model for the _ gold - nat _ data set predicts .
this model predicts only ( 4.6@xmath76 counts/ ( kg@xmath3yr ) .
this cross check confirms the indication from the background model that close sources are responsible for most of the @xmath6bi background contribution .
energy spectrum of the _ gold - coax _ data set ( filled histogram ) and the minimum model prediction ( red histogram ) .
the data and the model spectrum are fitted with a gaussian plus linear background ( dashed lines ) . ]
r|r|ccc|ccciso- & energy & & + tope & ' '' '' [ kev ] & + & ' '' '' & global & global & fit to & global & global & fit to + & & analysis & analysis & data & analysis & analysis & data + & & ( min .
fit ) & ( max .
fit ) & & ( min .
fit ) & ( max .
fit ) & + @xmath39k & 1460.8 & 11.9 [ 10.6,13.6 ] & 11.8 [ 10.6,13.6 ] & 13.6[12.5,15.0 ] & 19.4 [ 16.7,23.0 ] & 19.8 [ 16.7,23.1 ] & 18.3 [ 15.7,21.4 ] + @xmath5k & 1524.7 & 61.2 [ 59.3,63.8 ] & 61.4 [ 48.5 , 66.2 ] & 60.3[58.1 , 62.5 ] & 75.0 [ 70.2 , 82.4 ] & 75.6 [ 49.1 , 78.3 ] & 73.8 [ 69.1 , 80.1 ] + @xmath8co & 1173.2 & 2.5 [ 1.6 , 3.7 ] & 1.6 [ 0.6 , 2.9 ] & 4.2 [ 2.8 , 5.6 ] & 2.0 [ 0.1 , 4.6 ] & 1.9 [ 0.1 , 3.2 ] & @xmath403.8 + & 1332.3 & 2.3 [ 1.4 , 3.4 ] & 1.5 [ 0.5 , 2.6 ] & @xmath401.6 & 1.9 [ 0.1 , 4.4 ] & 1.7 [ 0.1 , 2.8 ] & 3.3 [ 1.6 , 5.2 ] + @xmath7ac & 911.2 & 3.8 [ 2.2 , 5.8 ] & 3.4 [ 1.0 , 3.6 ] & 3.9 [ 2.2 , 5.6 ] & 6.1 [ 3.9 , 8.6 ] & 5.4 [ 0.5 , 9.8 ] & 4.9 [ 2.7 , 7.3 ] + & 969.0 & 2.7 [ 1.5 , 4.1 ] & 2.8 [ 0.8 , 2.8 ] & 3.5 [ 1.8 , 5.0 ] & 4.4 [ 2.8 , 6.2 ] & 3.9 [ 0.4 , 7.1 ] & 4.8 [ 2.6 , 7.4 ] + @xmath77tl & 583.2 & 2.9 [ 2.5 , 3.5 ] & 1.3 [ 0.3 , 2.0 ] & 3.9 [ 1.8 , 5.7 ] & 3.0 [ 1.8 , 4.7 ] & @xmath400.9 & 2.5 [ 0.4 , 4.6 ] + & 2614.5 & 1.4 [ 1.2 , 1.7 ] & 1.0 [ 0.2 , 1.5 ] & 1.2 [ 0.9 , 1.5 ] & 1.5 [ 0.9 , 2.4 ] & 1.5 [ 0.0 , 6.7 ] & 1.4 [ 0.6 , 2.3 ] + @xmath6pb & 351.9 & 20.1 [ 17.9,22.7 ] & 17.5 [ 4.2,24.5 ] & 20.4 [ 16.2,24.8 ] & 22.1 [ 17.7,27.3 ] & 4.6 [ 3.1,6.0 ] & 25.6 [ 18.1,34.1 ] + @xmath6bi & 609.3 & 11.2 [ 10.0 , 12.6 ] & 8.0 [ 2.0 , 11.2 ] & 10.0 [ 8.0 , 12.3 ] & 11.2 [ 9.0 , 13.8 ] & 6.5 [ 4.6 , 8.6 ] & 7.6 [ 4.8 , 11.0 ]
+ & 1120.3 & 2.6 [ 2.3 , 2.9 ] & 1.8 [ 0.4 , 2.5 ] & @xmath403.1 & 2.6 [ 2.1 , 3.2 ] & 3.0 [ 2.1 , 3.9 ] & 4.0 [ 1.8 , 6.3 ] + & 1729.6 & 1.0 [ 0.9 , 1.1 ] & 1.0 [ 0.2 , 1.4 ] & 0.5 [ 0.2 , 0.9 ] & 1.3 [ 1.0 , 1.6 ] & 0.8 [ 0.5 , 1.1 ] & 0.9 [ 0.3 , 1 . ]
+ & 1764.5 & 3.6 [ 3.2 , 4.1 ] & 2.7 [ 0.7 , 3.8 ] & 3.1 [ 2.6 , 3.7 ] & 4.1 [ 3.3 , 5.1 ] & 3.7 [ 2.5 , 4.9 ] & 3.5 [ 2.4 , 5.0 ] + & 1847.4 & 0.6 [ 0.5 , 0.7 ] & 0.6 [ 0.1 , 0.8 ] & 0.6 [ 0.3 , 1.0 ] & 0.7 [ 0.6 , 0.9 ] & 0.5 [ 0.3 , 0.6 ] & 1.2 [ 0.5 , 2.1 ] + & 2204.2 & 1.0 [ 0.9 , 1.1 ] & 0.8 [ 0.2 , 1.1 ] & 0.8 [ 0.5 , 1.2 ] & 1.2 [ 1.0 , 1.5 ] & 1.3 [ 0.9 , 1.7 ] & 0.8 [ 0.2,1.6 ] + & 2447.9 & 0.3 [ 0.27 , 0.34 ] & 0.2 [ 0.05 , 0.3 ] & 0.2 [ 0.1 , 0.4 ] & 0.3 [ 0.2 , 0.4 ] & 0.4 [ 0.3 , 0.5 ] & @xmath401.8 + as the fit has been performed with a binning larger than the energy resolution of the detectors , the information from the line intensities is not maximized in the fitting procedure .
hence , it is instructive to cross check the line intensities obtained from fitting the peaks with a gaussian plus linear background in the different data sets with the expectation from the models .
table [ tab : backgroundlines ] compares the @xmath11-line intensities from the minimum and maximum models to those obtained from a fine binned analysis , i.e. a fit to data .
note , that for some of the @xmath11 peaks no fit could be performed due to limited number of events in the peak region or the low intensity of the @xmath11 line compared to the other background contributions . in those cases
the number of counts in the @xmath213@xmath78 energy range around the peak positions were used .
the background has been estimated according to the continuum seen in the @xmath215@xmath78 side bands at lower and higher energies around the peak .
a narrower side band is used when there is a second line close to the peak .
the intensities of the @xmath11 lines are obtained by marginalizing the posterior probability distribution of the signal rate .
the uncertainties on the predicted rate by the global models are due to the fit uncertainty on the parameters of the model components that give contribution to the @xmath11-ray line .
the statistical uncertainties due to the simulated number of events is on the order of 0.1% , i.e. negligible compared to the fit uncertainty .
there is excellent agreement between the numbers from the global analysis and those from the fine - binned analysis . to check for stability
, the fits were performed using different binnings . as the energy resolution of the detectors is around 4.5 kev at @xmath2 and the calibration at higher energies e@xmath165 mev , relevant for the @xmath9 model
is precise to about 10 kev , the lowest binning chosen was 10 kev .
also a 50 kev binning was performed .
the activities of different components derived from the fits with different binnings do not vary outside the uncertainties given for the 30 kev binning fits .
additionally it was checked whether the overall goodness of fit and the predicted bi and individual contributions in the region of interest changes if biases are introduced to the fits by single strong assumptions on individual background components .
minimum model fits for the _ gold - coax _ and _ gold - nat _ data sets were performed with the following individual modifications : @xmath7th and @xmath7ac are only in or on the radon shroud ; no @xmath6bi is present on the p@xmath56 surface ; @xmath5k is only on the p@xmath56 surface ; @xmath5k is present on the p@xmath56 surface ; @xmath8co is only inside the crystal , @xmath8co is only inside the detector assembly . except for the unrealistic assumption that all @xmath5k comes from p@xmath56 surface contaminations
all fits describe the measured spectrum reasonably well .
the prediction for the bi at the region of interest varies by 10% between the different models for the _ gold - coax _ data set and by 15% for the _ gold - nat _ data set .
the predictions for the activities of the individual components of the different models are consistent within the 68% uncertainty range quoted in table [ tab : fits_results_golden ] .
an important contribution to the background model are surface events from @xmath10ra daughters .
these include the decays of @xmath6bi and @xmath6po .
@xmath6po has a half life of only 164.4 @xmath29s .
hence , a number of events is expected where a low energy event from the @xmath6bi decay is followed by a high energy @xmath9 event from the @xmath6po decay in the same detector , a bipo tag .
the data reduction includes cuts that remove pile up events .
as the half life of @xmath6po is of the order of the decay time of the preamplifier signal , the analysis is not tailored to reconstruct typical bipo events .
hence it is difficult to quantify precisely the efficiency for this tag .
nevertheless , for the purpose of a qualitative statement an order of magnitude guess is made : an efficiency of 50% is assumed for the bipo recognition efficiency . using the number for the _ gold - coax _ data set obtained by the marginalized probability density function of the fit the number of detected @xmath79po surface events is 13.5 being reduced to approximately seven events from @xmath6bi decays on the surface that can lead to energy deposition in the detector active volume ( the decrease being due to the decay nucleus recoiling away from the surface ) . with an efficiency of the order of 50% to detect the bipo tag only roughly 3 to 4 bipo events
are expected . in the _ gold - coax _
data set in total 5 events have been found ( 2 in ang 2 , 2 in ang 3 and 1 in rg 1 ) that satisfy the criteria for a bipo tag . a mono - parametric pulse shape analysis technique for the identification of surface interactions on the p@xmath32 electrode of coaxial detectors has been recently developed and applied on phase i data @xcite .
the method is based on a cut on the rise time of the charge pulses computed between 5% and 50% of the maximum amplitude .
the cut level is calibrated on experimental data using the pure sample of high - energy @xmath9-induced events .
the analysis has been extended to the entire _ gold - coax _ data set . fixing
the cut to accept 95% of the events occurring in the proximity of the @xmath80 electrode , in the energy region of interest 43% of the events survive the cut .
part of the events surviving the cut is expected to be due to @xmath11 interactions in the proximity of the @xmath80 surface . applying the corrections described in ref .
@xcite , the total amount of @xmath9 and @xmath12 induced events on the p@xmath32 electrode is estimated to be between 15% and 35% of the number of events in the energy region of interest .
this result is consistent with the number of decays on the p@xmath32 surface predicted by the minimal model ( [email protected])% , given by the @xmath9 emitting isotopes plus @xmath6bi .
it is slightly lower compared to the maximal background model that requires 50% considering @xmath9 , @xmath6bi and @xmath5k on the p@xmath32 surface . as a large fraction of the contaminations are , according to the background model(s ) , located inside the detector array ( i.e. in the detector assembly , or in lar close to the surfaces of the detectors ) , a significant number of events are expected to have coincident hits in two detectors .
the efficiency to detect coincident events is expected to be increased with respect to single @xmath11 emitters for decays of isotopes with emission of multiple @xmath11 rays such as @xmath5k , @xmath8co , @xmath6bi and @xmath77tl .
coincident spectra are , thus , sensitive to differences in source locations .
a sum coincidence spectrum was produced for the _ gold - coax _ data set by summing the energies of all detectors in an event and filling the corresponding bin of the histogram .
also a single coincidence spectrum was produced by filling the corresponding bin of the histogram for each individual detector separately . the same procedure as for the minimum fit model ( see sec .
[ sec : modeling ] ) was applied to get best fit coincidence models for the single and sum spectra .
the results for the activities obtained from the minimum model best fit to the sum spectrum ( except for @xmath8co , where the single spectrum was used ) are summarized in table [ tab : fits_results_golden ] .
the obtained activities from coincident and single detector spectra are consistent with each other .
note , that the simulations were not tuned for the coincidence analysis .
the background source distribution was simplified in the simulation , while small changes in source location , especially within the detector array , can have significant effects on the coincidence efficiencies .
the fact that the @xmath5k activity derived from the coincidence fit is slightly higher than for the minimum and maximum models of the _ gold - coax _ and _ gold - nat _ data sets may be a hint that the distribution of @xmath5k in lar is not homogeneous . the consistency between the derived activities from coincident and single detector spectra support the result of the background model that the spectrum around @xmath2 is dominated by contaminants close to the detectors .
ll|rcrc|rc|rc|c & & & & & exp . from + component & location & & & & & screening + & & + ' '' '' total & & 18.5 & [ 17.6,19.3]&21.9 & [ 20.7,23.8 ] & 29.6 & [ 27.1,32.7 ] & 38.1 & [ 32.2,43.3 ] & + ' '' ''
@xmath5k & lar homogeneous & 3.0 & [ 2.9,3.1 ] & 2.6 & [ 2.0,2.8 ] & 2.9 & [ 2.7,3.2 ] & 2.0 & [ 1.8,2.3 ] & + @xmath5k & p@xmath32 surface & & & 4.6 & [ 1.2,7.4 ] & & & & &
+ @xmath5k & n@xmath32 surface & & & 0.2 & [ 0.1,0.4 ] & & & 20.8 & [ 6.8,23.7 ] & + @xmath8co & det .
assembly & 1.4 & [ 0.9,2.1 ] & 0.9 & [ 0.3,1.4 ] & 1.1 & [ 0.0,2.5 ] & & @xmath404.7 &
+ @xmath8co & germanium & 0.6 & @xmath160.1 @xmath26 ) & 0.6 & @xmath160.1 @xmath26 ) & 9.2 & [ 4.5,12.9 ] & 1.0 & [ 0.3,1.0 ] & + @xmath31ge & germanium & & & & & & & & 1.5 ( @xmath406.7 ) &
+ @xmath6bi & det .
assembly & 5.2 & [ 4.7,5.9 ] & 2.2 & [ 0.5,3.1 ] & 4.9 & [ 3.9,6.1 ] & 5.1 & [ 3.1,6.9 ] & @xmath412.8 + @xmath6bi & lar close to p@xmath32 & & & 3.1 & @xmath404.7 & & & & & @xmath400.7 + @xmath6bi & p@xmath32 surface & 1.4 & [ 1.0,1.8 ] @xmath26 ) & 1.3 & [ 0.9,1.8 ] @xmath26 ) & 3.7 & [ 2.7,4.8 ] @xmath26 ) & 0.7 & [ 0.1,1.3 ] @xmath26 ) & + @xmath6bi & radon shroud & & & 0.7 & @xmath403.5 & & & & & + @xmath7th & det .
assembly & 4.5 & [ 3.9,5.4 ] & 1.6 & [ 0.4,2.5 ] & 4.0 & [ 2.5,6.3 ] & 4.2 & [ 1.8,8.4 ] & @xmath400.3 + @xmath7th & radon shroud * ) & & & 1.7 & @xmath402.9 & & & & & @xmath411.0 + @xmath9 model & p@xmath32 surface and & 2.4&[2.4,2.5 ] & 2.4 & [ 2.3,2.5 ] & 3.8 & [ 3.5,4.2 ] & 1.5 & [ 1.2,1.8 ] & + & lar close to p@xmath32 & & & & & & & & & + @xmath26 ) prior : discussed in sec .
[ sec : modeling ] + @xmath81 ) representing all distant sources including the heat exchanger , the wall of the steel cryostat and the calibration source at the bottom of the tank .
the background models obtained by global fits in the 570 - 7500 kev region allow to predict the individual background contributions and the total background at @xmath2 .
table [ tab : background_components ] lists the predictions for the bi from different contributions in a 10 kev window for coaxial detectors and in a 8 kev window for bege detectors around @xmath2 for different data sets .
the results obtained from the best fit parameters are quoted together with the smallest 68% interval of the marginalized distributions of the parameters .
if the maximum of the marginalized distribution is at zero a 90% upper limit is given .
for the case of an internal @xmath8co contamination a 90% lower limit is given , because a higher contamination gives a better fit , however , is constrained by prior knowledge of above ground exposure to cosmic rays .
according to the models the main contributions to the background at @xmath2 are due to the @xmath9-emitting isotopes in the @xmath10ra decay chain , @xmath5k , @xmath8co , @xmath6bi and @xmath7th .
the fraction with which each component contributes depends on the assumed source location . tab .
[ tab : background_components ] also lists the bis expected from the screening measurements as reported in tab . [
tab : thorium ] .
the bis due to the individual identified components do not match well with the bis derived from the background model , indicating that the unidentified close - by @xmath6bi and @xmath7th contributions have to be present .
[ fig:0nbb_region ] shows the best fit minimum and maximum models and the individual contributions together with the observed spectrum around @xmath2 for the _ gold - coax _ data set .
the spectral shapes of the best fit models are constant around @xmath2 .
no peaks are predicted in the blinded regions .
this indicates that the bi at @xmath2 can be estimated by interpolating the results of a fit to the observed number of events outside the signal search window .
the window to be used for this estimation is chosen as sum of the 19302019 kev , 20592099 kev , 21092114 kev and 21242190 kev intervals for a total width of 200 kev .
the bi evaluation window excludes the central 40 kev window around @xmath2 and the regions within @xmath215 kev from the @xmath11 lines expected from the background model
namely , single escape peak from @xmath77tl at 2104 kev and the @xmath6bi @xmath11 line at 2119 kev .
the resulting background indices from the interpolation are listed in table [ tab : extrapolated_bi ] for different data sets together with the prediction of the background models for comparison .
the lower panels of the plots in fig . [ fig:0nbb_region ] demonstrate that the background model can be described by a constant in the bi evaluation windows .
the difference in the resulting bi is less than 1% if a linear instead of a constant distribution is assumed .
the statistical uncertainty for the approximation of the bi by an interpolation is of the same size as the systematic uncertainty expected by the model .
lccc & _ gold - coax _ & _ gold - nat _ & _ sum - bege _ + ' '' '' & + & + interpolation & 17.5 [ 15.1,20.1]&30.4 [ 23.7,38.4 ] & 36.1 [ 26.4,49.3 ] + minimum & 18.5 [ 17.6,19.3]&29.6 [ 27.1,32.7 ] & 38.1 [ 32.2,43.3 ] + maximum & 21.9 [ 20.7,23.8]&37.1 [ 32.2,39.2 ] & + ' '' '' & + & 30 kev & 40 kev & 32 kev + data & 13 & 5 & 2 + minimum & 8.6 [ 8.2,9.1 ] & 3.5 [ 3.2,3.8 ] & 2.2 [ 1.9,2.5 ] + maximum & 10.3 [ 9.7,11.1 ] & 4.2 [ 3.8,4.6 ] & + the global fits were performed by excluding the central 40 kev region around @xmath2 which was completely blinded until may 2013 .
thereafter , a 30 kev ( 32 kev ) window was opened for analysis by keeping the central 10 kev ( 8 kev ) window still blinded for the enriched coaxial ( bege ) detectors .
the natural detector gtf 112 was completely unblinded .
the first step of unblinding gives the possibility to compare the model predictions to the observed number of events in those regions as a consistency check for the model .
table [ tab : extrapolated_bi ] also lists the predicted and observed number of events in these energy regions for different data sets . in total 13 events
were observed in the unblinded 30 kev window of the _ gold - coax _ data set .
the predictions in this window were 8.6 events from minimum and 10.3 events from maximum model .
the probability to observe 13 events or more given the predictions are 10% and 24% , respectively .
in the _ gold - nat _ data set 5 events were found in the 40 kev unblinded window , resulting in a 27% probability for the minimum model prediction of 3.5 events and a 41% probability for the maximum model prediction of 4.2 events .
for the _ sum - bege _ data set there is a perfect agreement between the observed two events and the expectation of 2.2 events from the model . if the additional events seen in the 30 kev unblinded window are included to the interpolation the expected bi at @xmath2 increases to 19@xmath3@xmath4 cts/(kev@xmath3kg@xmath3yr ) for the _ gold - coax _ data set
. given , for the _ gold - coax _ data set , the background prediction of the minimum model of 18.5@xmath310@xmath49 cts/(kev@xmath3kg@xmath3yr ) and the known 17.90 kg@xmath3yr exposure at the end of phase i , the sensitivity for the @xmath0 decay half life @xmath15 was calculated .
the value of the exposure - averaged total efficiency ( see eq . [ equ : efficiency ] ) for the _ gold - coax _ data set is @xmath820.688 . from the energy spectrum
an upper limit on the @xmath0 signal strength @xmath83 at specified probability or confidence level can be derived and converted to a half life limit @xmath84 using @xmath85 in order to estimate the limit setting sensitivity without pulse shape analysis , 10@xmath86 mc realizations of gerda were generated assuming no @xmath0 signal . for each realization ,
the number of events was allowed to fluctuate according to a poisson distribution with expectation given by the number of predicted background events .
the expected lower limit for @xmath15 was estimated by using both , bayesian and frequentist analyses .
in both analyses the signal and background strengths were free parameters .
for the frequentist analysis , the @xmath0 decay rates were estimated from a profile likelihood fit to the unbinned energy spectrum of each realization .
the 90% c.l .
lower limit @xmath15 @xmath161.9@xmath310@xmath87 yr ( 90% c.l . ) corresponds to the median of the 90% quantile of the profile likelihood . in the bayesian analysis ,
the 90% probability lower limit for @xmath15 was calculated as the median of the 90% quantiles of the posterior marginalized probabilities p@xmath88 , where @xmath89 is the hypothesis that both background and @xmath0 events contribute to the spectrum .
the result is @xmath15 @xmath161.7@xmath310@xmath87 yr ( 90% c.l . ) .
the difference in the numerical values from the bayesian and frequentist analysis ( which have conceptually a different meaning ) is mainly due to the behavior of the two approaches in the cases when the number of observed counts is smaller than the background expectation .
the sensitivity is expected to be about 10% better than calculated from the _ gold - coax _ data set only , because of the extra exposure available in the _ silver - coax _ and _ sum - bege _ data sets .
also the sensitivity might further increase by applying pulse shape discrimination techniques to the _ gold - coax _ , _ silver - coax _ and _ sum - bege _ data sets @xcite .
the background measured by the experiment has been presented in an energy range between 100 and 7500 kev .
it has been demonstrated that stable low background data taking with the innovative technique of operating bare hpge detectors in a cryogenic liquid is possible over a time period of about 1.5 yr .
more than 20 kg@xmath3yr of data have been acquired by the experiment with six enriched coaxial detectors of a total mass of 14.6 kg and with four enriched bege detectors of a total mass of 3.0 kg .
a background model has been developed with the @xmath22=40 kev blinded that allows to predict the bi in this energy range .
the predictions of the models have been tested for consistency on a 30 kev ( 32 kev ) range for the coaxial ( bege ) detectors , while the central @xmath22=10 kev region of interest and @xmath22=8 kev region for the bege detectors around @xmath2 was still blinded .
the model describes the background in an energy range from 570 to 7500 kev well .
the only significant background contributions in originate from decays of @xmath5k in the lar bath , from @xmath6bi in the detector assembly , from residual @xmath37rn dissolved in lar , from @xmath7th and @xmath8co in the detector assembly , and from surface @xmath9 particles .
the largest contributions come from contaminants located close to the detectors .
several cross checks confirm the validity of the background model .
the 68% credibility intervals of the bi expected due to the minimum and maximum models at @xmath2 of @xmath1ge span the range between 17.6 and 23.8@xmath3@xmath4 cts/(kev@xmath3kg@xmath3yr ) .
this range includes the systematic uncertainty due to different source location assumptions .
predictions for the number of events in the blinded region around @xmath2 have been made .
it could be shown that the expected background is flat in a region of @xmath41 200 kev around @xmath2 and that no significant peak like structures are expected in the blinded energy region . the background model and an interpolation of a fit to data from a 200 kev energy window into the blinded energy window
give compatible results .
the bi interpolated into the region of interest are + ( 1.75@xmath90)@xmath3@xmath47 cts/(kev@xmath3kg@xmath3yr ) for the coaxial detectors and ( 3.6@xmath91)@xmath3@xmath47 cts/(kev@xmath3kg@xmath3yr ) for the bege detectors .
the statistical uncertainty on the bi prediction from interpolation is of the same size as the systematic uncertainty from the choice of the background model .
the bi obtained from interpolation of the spectrum in a 200 kev window around @xmath2 will be used in the @xmath0 analysis of the phase i data .
given the expected background rate without pulse shape discrimination and assuming no signal , the sensitivity for the _ gold - coax _ data set is @xmath92 yr ( 90% c.l . ) using a profile likelihood fit and @xmath93 yr ( 90% c.i . ) using a bayesian analysis .
the gerda experiment is supported financially by the german federal ministry for education and research ( bmbf ) , the german research foundation ( dfg ) via the excellence cluster universe , the italian istituto nazionale di fisica nucleare ( infn ) , the max planck society ( mpg ) , the polish national science centre ( ncn ) , the foundation for polish science ( mpd programme ) , the russian foundation for basic research , and the swiss national science foundation ( snf ) .
the institutions acknowledge also internal financial support .
particle data group , review of particle physics , k. nakamura _ et al .
_ , j. phys .
g * 37 * ( 2010 ) k.h .
et al . _ ,
c * 73 * ( 2013 ) 2330 g. douysset , t. fritioff and c. carlberg , phys .
86*(2001)4259 hdm - collaboration : h. v. klapdor - kleingrothaus _ et al .
j. * a12 * ( 2001 ) 147 igex collaboration : c. e. aalseth _
et al . _ , phys .
d * 65 * ( 2002 ) 092007 h. v. klapdor - kleingrothaus _ et al .
b * 586 * ( 2004 ) 198 h. v. klapdor - kleingrothaus and i.v .
krivosheina , mod .
a21 * ( 2006 ) 1547 b. schwingenheuer , annalen der physik * 525 * ( 2013 ) 269 g. heusser , annu .
* 45 * ( 1995 ) 543 the gerda collaboration , m. agostini _
_ , submitted to epj c , arxiv:1307.2610 m. bernab heider , phd thesis , university of heidelberg ( 2009 ) g. zuzel and h. simgen , appl .
* 67 * ( 2009 ) 889 d. lenz , phd thesis , technical university munich ( 2008 ) i. barabanov _
et al . _ ,
methods a * 606 * ( 2009 ) 790 l. pandola _
et al . _ ,
methods a * 570 * ( 2007 ) 149 m agostini , l pandola , p zavarise , j. phys .
( conf . ser . )
* 368 * ( 2012 ) 012047 p. zavarise _ et al . _ ,
ser * 375 * ( 2012 ) 042028 m. agostini _ et al .
_ , jinst * 6 * ( 2011 ) p08013 k. debertin and r. g. helmer , _ gamma and x - ray spectroscopy with semiconductor detectors _ , north holland , ( 1988 ) collaboration , m. agostini _
et al . _ , j. phys .
g : nucl . part .
* 40 * ( 2013 ) 035110 n. becerici - schmidt for the gerda collaboration , proceedings of the workshop in low radioactivity techniques , lngs , assergi , italy , april 10 - 12 , 2013 , aip conf .
1549 , 156 - 160 ( 2013 ) doi : 10.1063/1.4818098 , preprint arxiv:1306.2302 m. boswell _ et al .
_ , ieee trans .
* 58 * ( 2011 ) 1212 geant collaboration , s. agostinelli _
inst . meth .
a * 506 * ( 2003 ) 250 ; j. allison _ et al . _ ,
ieee trans .
* 53 * ( 2006 ) 270 g. meierhofer _
et al . _ ,
j. a * 48 * ( 2012 ) 20 and references therein .
et al . _ ,
c * 77 * ( 2008 ) 054614 i. barabanov _ et al .
b * 251 * ( 2006 ) 115 o.a .
ponkratenko , v.i .
tretyak and yu.g .
zdesenko , phys . atom .
nuclei * 63 * ( 2000 ) 1282 a. caldwell , d. kollar , and k. krninger , comput .
* 180 * ( 2009 ) 2197 r. aggarwal and a. caldwell , eur .
j. plus * 127 * ( 2012 ) 24 v.d .
et al . _ ,
instr . meth .
a * 416 * ( 1998 ) 179 e. aguayo _
et al . _ ,
instr . meth .
a * 701 * ( 2013 ) 176 p. benetti _ et al .
a * 574 * ( 2007 ) 83 m. agostini , phd thesis , technical university munich ( 2013 )
|
the germanium detector array ( ) experiment at the gran sasso underground laboratory ( lngs ) of infn is searching for neutrinoless double beta ( @xmath0 ) decay of @xmath1ge .
the signature of the signal is a monoenergetic peak at 2039 kev , the @xmath2 value of the decay . to avoid bias in the signal search
, the present analysis does not consider all those events , that fall in a 40 kev wide region centered around @xmath2 .
the main parameters needed for the @xmath0 analysis are described .
a background model was developed to describe the observed energy spectrum .
the model contains several contributions , that are expected on the basis of material screening or that are established by the observation of characteristic structures in the energy spectrum .
the model predicts a flat energy spectrum for the blinding window around @xmath2 with a background index ranging from 17.6 to 23.8@xmath3@xmath4 cts/(kev@xmath3kg@xmath3yr ) . a part of the data not considered before
has been used to test if the predictions of the background model are consistent .
the observed number of events in this energy region is consistent with the background model .
the background at @xmath2 is dominated by close sources , mainly due to @xmath5k , @xmath6bi , @xmath7th , @xmath8co and @xmath9 emitting isotopes from the @xmath10ra decay chain .
the individual fractions depend on the assumed locations of the contaminants .
it is shown , that after removal of the known @xmath11 peaks , the energy spectrum can be fitted in an energy range of 200 kev around @xmath2 with a constant background .
this gives a background index consistent with the full model and uncertainties of the same size .
example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore
|
matter - antimatter interaction produces more energy per unit mass than any other means of energy production . for example , proton - antiproton annihilation releases 275 times more energy in the form of kinetic energy of charged particles than nuclear fission or @xmath0 fusion .
this energy is released by simple contact of antimatter with matter so that , in principle , no ignition energy is required to start the reaction .
it is therefore not surprising that the concept of using antimatter as an energy source has been in scientific literature for decades @xcite .
other practical applications of antimatter are under consideration .
for example , antimatter propulsion systems @xcite , space based power generators @xcite , directed energy weapons @xcite , cancer therapy @xcite . finally , both edward teller @xcite and andrei sakharov @xcite
, the key scientists in charge of the development of the h - bomb in their respective countries , show in their published scientific works a big interest in the annihilation properties of antimatter , the nuclear process that after fission and fusion could lead to a third generation of nuclear bombs .
this paper is a summary of a comprehensive assessment of the feasibility of producing large quantities of antiprotons and using them for igniting inertial - confinement fusion pellets or triggering large scale thermonuclear explosions @xcite .
when a particle meets it s antiparticle they annihilate and the energy equivalent to their total mass ( @xmath4 ) is converted into various new particles and kinetic energy @xcite . in the case of proton - antiproton annihilation , many different reaction channels are possible , each resulting in the production of a different number of charged and neutral particles .
a good approximation is that three charged and two neutral pions are produced on the average . since neutral pions quickly decay into photons , the typical @xmath5 annihilation process is as follows : @xmath6 where @xmath7 mev and @xmath8 mev .
an antiproton can also annihilate with a neutron , in which case mostly pions are produced again , in numbers , on the average , similar to @xmath5 annihilation .
antiprotons , antineutrons and positrons can combine to form antinuclei , antiatoms , antimolecules .
annihilation occurs when the two kinds of matter come sufficiently close to one other . even at some distance
, a neutral atom and a neutral antiatom will attract each other by van der waals forces @xcite . as a consequence , storage of _ antiatoms _ in a container made of matter
is impossible in general .
however , there may exist metastable states of _ antiprotons _ in normal matter @xcite .
when a @xmath9 annihilates in a hydrogen plasma , essentially all the annihilation energy is radiated in the form of very energetic pions and photons . at solid hydrogen densities ,
the mean free path of the 187 mev photons is 25 m , so that they will not loose energy in the plasma . however , the three 236 mev charged pions will loose energy by multiple coulomb interactions with the electrons at a rate approximately given by : @xmath10 mev / cm in solid @xmath11 or @xmath0 and 2.06 mev / cm in @xmath1 .
if we now assume that annihilation takes place at the center of a sphere , the energy @xmath12 deposited within a radius @xmath13 cm is only 1.5 mev out of the total 1876 mev annihilation energy .
there are however several ways to improve energy deposition , and thus plasma heating .
firstly , the fuel to be heated may be compressed by a factor @xmath3 , @xmath14 will then be multiplied by @xmath3 , and thus @xmath12 by @xmath15 .
but compression requires energy .
secondly , fuels such as @xmath1 , which contain more electrons , have a proportionally larger @xmath14 .
however , their thermonuclear ignition temperature is also higher .
finally , annihilation may take place with a nucleus .
when a @xmath9 annihilates with a nucleon from a nucleus , because of the fermi motion of the annihilated nucleon , the nucleus will recoil with an energy of about 20 mev .
furthermore , each of the 5 annihilation pions has a probability of colliding with the rest of the nucleus .
hence , the average total energy deposition in a sphere is @xmath16 where @xmath17 is the number of charged pions and @xmath18 the local energy deposition by the recoiling nucleus and the various pion - nucleus interaction debris . in the case of @xmath9 annihilation with deuterium or tritium
@xmath18 is approximately 12 mev on the average , about half of the fermi energy . with heavy nuclei
there have been many theoretical speculations in the absence of measurements .
the first of these was introduced by duerr and teller @xcite , who speculated that an antiproton would find a very strong ( 900 mev ) attractive potential when getting close to a nucleus .
more recently @xcite , los alamos scientists have calculated that annihilation in carbon would result in the local energy deposition of about 100 mev .
recent measurements at cern show that it is in fact only 33 mev in carbon @xcite .
low energy @xmath9 s annihilate mostly at the surface of nuclei , and thus local energy deposition follows a @xmath19 dependence on atomic weight . in effect , the cern data is compatible with the expression : @xmath20}.\ ] ] hence , for @xmath9 annihilation in @xmath11 , @xmath0 or @xmath1 , @xmath21 is always about 3 and @xmath18 is approximately equal to 0 , 12 or 22 mev respectively .
a matter - antimatter plasma is obtained if some initially stable particle - antiparticle mixture is suddenly ignited .
the annihilation rate of two interacting species , with number densities @xmath22 and @xmath23 , is @xmath24 where @xmath25 is the annihilation reaction rate averaged over the maxwell distribution . in a @xmath26 plasma
, equation holds for both protons and electrons with @xmath27 initially .
hence , for a given temperature @xmath28 if we assume @xmath29 kev , @xmath25 is approximately the same for both @xmath30 and @xmath5 annihilation .
thus the electron and the proton populations deplete at the same rate , with a time constant of 5 ns for @xmath31 g/@xmath32 .
when matter and antimatter come into contact , annihilation primarily takes place in a boundary layer in which particles and antiparticles are mixing .
the thickness of this matter - antimatter plasma is of the order of the antimatter mean - free - path in matter , i.e. , @xmath33 . a first approximation , assuming that whenever an antiparticle penetrates into the boundary layer it instantly annihilates , is an annihilation rate per element area given by the total number of antiparticles impinging on that surface . from the maxwell velocity distribution one
gets @xmath34 the @xmath35 annihilation rate is thus @xmath36 times the @xmath9 annihilation rate .
however , since the @xmath37 plasma debye length is much smaller than the boundary layer thickness , plasma charge neutrality insures that the antimatter flow rate is determined by the slowest annihilation rate .
therefore , if @xmath37 s interact with the walls of a closed cavity , annihilation results in an overall decrease of the antimatter density within the cavity .
let us now take the case of a sphere of solid antihydrogen that is suddenly put in contact with a collapsing spherical shell of compressed @xmath0 . to solve eq .
one has to calculate the increase in the @xmath37 plasma internal energy by the pions and other particles from @xmath9 annihilation in the surrounding @xmath0 : @xmath38 where @xmath39 cm is the approximate range of the 20 mev recoil protons from @xmath9 annihilation in @xmath0 , and @xmath40 ( initially equal to @xmath41 ) the number of @xmath37 atoms . for hydrogen @xmath42
, we get a system of equations for the @xmath37 plasma density and temperature . if annihilation is much faster than the collapse of the cavity ( @xmath43 constant ) the solution of eqs . and
is @xmath44 for @xmath45 , which corresponds to @xmath46 cm , we find @xmath47 kev and @xmath48 ns .
thus , in about @xmath49 ns , over 90% of the antihydrogen in the sphere is annihilated .
this time constant is compatible with the requirements of instantaneous thermalization and inertial confinement of the plasma .
the most efficient way to trigger a thermonuclear explosion is probably to start a thermonuclear detonation wave in @xmath1 by collapsing a hollow sphere of that material on a tiny spherical pellet of solid antihydrogen . in the spark model of thermonuclear ignition ,
an outgoing spherical detonation wave starts if : ( a ) a critical amount of energy @xmath50 is deposited in the center of the sphere ( the `` spark '' region ) and ( b ) if the temperature within this volume is higher than a critical temperature @xmath51 . without compression , one has @xmath52 kev and @xmath53 kev for solid @xmath0 , and @xmath54 kev and @xmath55 kev for @xmath1 .
however , for a compressed thermonuclear fuel at temperature @xmath51 , the critical energy decreases with the square of the compression factor @xmath3 .
the number @xmath40 of @xmath9 annihilations necessary to induce a thermonuclear burn wave can be estimated by supposing that annihilation takes place at the center of the sphere to be ignited .
thus , from equation , condition ( a ) is satisfied if @xmath56 since the pions originate from the center , the temperature in the fuel goes as @xmath57 .
therefore , for simplicity , we require that condition ( b ) is satisfied for the average temperature within the critical volume .
thus @xmath58 where @xmath59 and @xmath60 are respectively equal to 2 and 2.5 for @xmath0 , and 6 and 9.5 for @xmath1 .
taking @xmath61 , a modest compression factor , and solving eqs . and for @xmath40 and the spark radius @xmath62 , one finds @xmath63 and @xmath64 cm for @xmath0 , and @xmath65 and @xmath66 cm for @xmath1 .
however , because of some of the simplifying assumptions made , these results may be somewhat pessimistic .
hence , we will assume that @xmath67 @xmath9 s are sufficient to trigger the thermonuclear explosion of compressed @xmath0 or @xmath1 pellets . for thermonuclear explosions in the kiloton range
, chemical explosives may be used to implode the @xmath1 shells . for low yield explosions such as in x - ray laser pumping or icf ,
compression factors higher than 30 can be achieved using magnetic compression , beams or other techniques .
however , antiproton induced fusion will remain an attractive alternative to normal icf only if the compression factor is kept relatively small , i.e. , less than 300 , giving a number of @xmath9 s of the order of @xmath68 .
the production of @xmath68 @xmath9 s for each antimatter triggered icf pellet would require an energy investment of at least @xmath69 mj @xcite .
it will therefore be very difficult to achieve energy break - even in power generating reactors using annihilation techniques .
moreover , the technologies for producing @xmath9 s with high energy accelerator systems , and the means for manipulating and storing sizable amounts of @xmath37 are extremely complicated .
for instance , a plant of the size required to produce the antimatter needed for one thermonuclear bomb trigger a day ( @xmath70 g of @xmath37 or @xmath67 @xmath37 atoms per day ) could consist of several 10 s of accelerators and storage rings , and could require as many as several large nuclear power plants to supply the electricity @xcite .
a study by the rand corporation gives a cost estimate of $ 500 to 1000 million for a prototype factory providing 10 to 100 micrograms , and $ 5 to 15 billion for a full production factory with an output of about 10 mg per year @xcite . as a consequence ,
civilian applications of antimatter for power production are very unlikely .
directed energy weapons applications may include the triggering of thermonuclear plasma jets , and x - ray or gamma - ray laser pumping . in the event of a comprehensive test ban treaty
, antimatter would provide a means for inducing laboratory and small scale thermonuclear explosions in a yield range which can not easily be covered by underground explosions or classical icf systems @xcite .
of course , many technical problems will have to be solved @xcite .
in particular , the levitation of a frozen @xmath37 pellet within a 1 mm diameter cryostat at the heart of a complex thermonuclear device is a tremendous challenge for materials microtechnology .
however , if metastable states of @xmath9 s in @xmath71 , @xmath72 or possibly @xmath73 compounds are discovered , much simpler designs could be considered .
special issue of the j. of the brit .
interplanetary soc .
on antimatter propulsion .
jbis * 35 * ( 1982 ) 387 - 424 .
r.l . forward : _ making and storing antihydrogen for propulsion _ , workshop on the design of a low - energy antimatter facility in the usa , university of wisconsin , october 3 - 5 ( 1985 ) .
sullivan : _ a measurement of the local energy deposition by antiprotons coming to rest in tissue - like material_. cern preprint tis
rp/149/cf ( 22 april 1985 ) .
biol . * 30 * ( 1985 ) 12971303 .
morgan and v.w .
hugues : _ atomic processes involved in matter - antimatter annihilation_. phys .
d2 * ( 1970 ) 13891399 . for e.
teller see page 1399 .
morgan , v.w .
hughes : _ atom - antiatom interactions_. phys . rev . * a7 * ( 1973 ) 18111825 .
a. gsponer and j .- p .
hurni : _ antimatter induced fusion and thermonuclear explosions_. atomkernenergie @xmath74 kerntechnik ( independent journal on energy systems and radiation ) * 49 * ( 1987 ) 198203 .
e - print arxiv : physics/0507125 available in pdf format at + a. gsponer , b. jasani , and s. sahin : _ emerging nuclear energy systems and nuclear weapon proliferation_. atomkernenergie @xmath74 kerntechnik ( independent journal on energy systems and radiation ) * 43 * ( 1983 ) 169174 .
a. gsponer et j .- p .
hurni : _ les armes antimatire_. la recherche * 17 * ( novembre 1986 ) 14401443 .
english translations : _
antimatter weapons_. the world scientist ( new delhi , india , 1987 ) 7477 , and bulletin of peace proposals * 19 * ( 1988 ) 444450 .
e - print arxiv : physics/0507132 available in pdf format at +
|
the possibility of using antihydrogen for igniting inertial confinement fusion pellets or triggering large - scale thermonuclear explosions is investigated .
the number of antiproton annihilations required to start a thermonuclear burn wave in either @xmath0 or @xmath1 is found to be about @xmath2 , where @xmath3 is the compression factor of the fuel to be ignited .
we conclude that the financial and energy investments needed to produce such amounts of antiprotons would confine applications of antimatter triggered thermonuclear devices to the military domain .
|
the relaxation of non - equilibrium spin polarization is the central phenomenon in spin - dependent transport in semiconductor nanostructures .
@xcite one of the most efficient mechanisms of electron spin relaxation in iii - v semiconductors is the well - known dyakonov - perel mechanism @xcite based on the classical picture of angular diffusion of the spin vector in a random magnetic field .
the field originates from the momentum - dependent spin - orbit splitting of the conduction band in the crystals with zinc - blend structure .
@xcite while passing through the crystal , the electron is scattered by impurities and its momentum changes randomly with time . as a consequence , the effective magnetic field also changes randomly with a correlation time of the order of the momentum relaxation time @xmath1 .
the spin relaxation time @xmath2 is a characteristic time of the spin angular diffusion @xcite @xmath3 , where @xmath4 is the typical angle of the spin precession for the momentum relaxation time and @xmath5 is the frequency of precession in the effective magnetic field proportional to the conduction band splitting .
for a two - dimensional ( 2d ) case , when the electron motion in one direction is confined by the quantum well , the spin splitting and , hence , the precession frequency are proportional to the in - plane electron velocity @xcite @xmath6 so , at low temperatures , when inelastic processes can be disregarded , the spin relaxation rate for a 2d electron with a given energy @xmath7 is proportional to the particle diffusion coefficient @xmath8 the effects of localization on the particle diffusion have been discussed in a great number of publications . the first - order term in a series expansion of @xmath9 in @xmath10 ( @xmath11 is the mean free path
, @xmath12 is the electron wave vector ) is known as the weak localization correction @xcite ( for a review , see ref . ) , coming from the coherent enhancement of the backscattering amplitude .
a remarkable feature of this correction is the logarithmic divergence at low temperatures in the 2d case .
equation implies a similar divergence of the spin relaxation rate .
such a divergence was first predicted by singh @xcite in spin correlation functions for a system with spin - dependent impurity scattering .
it was shown , however , that the quantum correction to the spin relaxation rate is not proportional to the quantum correction to the diffusion coefficient , as one might expect from eq . .
a similar result was obtained in ref . for a system with a spin - split spectrum .
it was found in ref . that the weak localization slows down the spin relaxation of excitons in quantum wells , which leads to a @xmath0 power tail in the spin orientation .
a similar effect was also predicted for electrons in 2d semiconductors with a zinc - blend crystal structure . in this paper , we consider the effects of localization on the spin relaxation for a 2d semiconductor with a spin - split spectrum .
we show that the spin dynamics is described by a boltzmann - type equation . in the first order in @xmath13 the localization effects
can be taken into account by a nonlocal - in - time correction to the boltzmann collision integral .
this correction is expressed in terms of the spin - dependent return probability .
we discuss the role of coherent returns at different scattering angles and show that the `` nonbackscattering '' contribution to the collision integral plays a key role .
we solve the generalized kinetic equation and demonstrate that , at large times , the spin polarization decays as @xmath14 the magnetic field , both transversal and longitudinal , is found to suppress the long - living tail in the spin - relaxation .
the hamiltonian of a 2d with a spin - split spectrum is given by @xmath15 here @xmath16 is the in - plane electron momentum , @xmath17 is the electron effective mass and @xmath18 is a vector consisting of pauli matrices .
the spin - orbit interaction is described by the second term , in which @xmath19 depends on the direction of the electron momentum @xmath20 the matrix @xmath21 is the sum of two terms : the so - called bychkov - rashba term @xcite @xmath22 ( with nonzero components @xmath23 ) and @xmath24 which is the dresselhaus term @xcite averaged over the electron motion along the @xmath25-direction perpendicular to the quantum well plane .
the bychkov - rashba coupling depends on the asymmetry of the quantum well confining potential .
its strength can be tuned by varying the gate voltage.@xcite the dresselhaus term is present in semiconductors with no bulk inversion symmetry .
the components of the matrix @xmath26 are also linear in the in - plane electron momentum @xmath27 and vary with well plane orientation with respect to the crystallographic axes @xcite ( we neglect cubic dresselhaus terms , assuming that the electron concentration is relatively small ) .
we consider the scattering by the short - range impurity potential with the correlation function @xmath28 , where the coefficient @xmath29 is related to the transport scattering time by @xmath30 the classical spin dynamics is described by the kinetic equation .
@xcite for a homogeneous case , this equation is @xmath31 here @xmath32 is the boltzmann collision integral and @xmath33 is the spin density in the momentum space , related to the averaged spin by @xmath34 we assume that the spin splitting is relatively small : @xmath35 this inequality provides @xmath36 the relationship between @xmath2 and the inelastic scattering time @xmath37 varies with temperature . here
we focus on the case of low temperatures , assuming that @xmath38 then spins with different energies do not correlate with each other , and the solution of eq .
( [ boltz ] ) at @xmath39 yields @xmath40 , where @xmath41 , @xmath42 is the initial spin density averaged over the momentum direction , and @xmath43 is the spin relaxation tensor ( tensor of inverse relaxation times ) given by @xcite @xmath44~\tau/2 .
\label{gamma}\ ] ] the conventional approach to the calculation of the correlations functions in weakly localized systems is based on the kubo formula .
@xcite an alternative approach @xcite is to generalize the boltzmann equation to include weak localization effects in the kinetic description .
this approach may turn out to be more convenient when studying nonlinear and strongly nonequilibrium phenomena . to describe quantitatively the weak localization phenomenon in the kinetic picture
, one has to modify the boltzmann equation by introducing a nonlocal - in - time correction to the collision integral .
@xcite these corrections can be derived @xcite from the diagrammatic structure of linear - response functions .
diagrammatically , the inclusion of a weak localization correction to the effective collision integral requires the consideration of the irreducible diagrams @xcite in fig .
[ fig1]b , c , in addition to the diagram for the born scattering , shown in fig . [ fig1]a . the crossed - ladder diagrams ( 1b )
are usually considered to describe the coherent backscattering of the electron wave .
a physical interpretation of the diagrams ( 1b , c ) in terms of a small change in the effective differential cross - section for a single impurity was suggested in ref . .
it was based on an analysis of the interference contribution of trajectories propagating in the opposite directions along closed paths in terms of the phase stationarity requirement . and the respective scattering processes @xmath45 .
the born collision process @xmath46 is independent of the electron spin ( its contribution is proportional to @xmath47 ) .
coherent backscattering @xmath48 ( @xmath49 ) , as well as the processes @xmath50 describing coherent scattering at an arbitrary angle ( @xmath51 ) , are spin - dependent due to rotation of the spin of an electron passing along the closed path.,scaledwidth=60.0% ] is due to the coherent backscattering shown in fig .
[ fig1]b@xmath52 .
the enhancement of the coherent backscattering is accompanied by a reduction of scattering at other angles ( the process in fig .
[ fig1]c@xmath52 ) , the total cross - section being unchanged.,scaledwidth=50.0% ] this analysis shows that the diagrams ( 1b ) correspond to a process shown in fig .
[ fig1]b@xmath52 , which is indeed a coherent backscattering . diagrams ( 1c ) were found to describe coherent scattering processes with arbitrary scattering angles , shown in fig .
[ fig1]c@xmath53 next , we discuss the key points of a quantitative description of the weakly localized regime within the kinetic approach .
we start with a brief discussion of the zero spin - orbit coupling .
as can be seen from figs .
[ fig1]b@xmath52,c@xmath52 , the relevant processes contain the same closed paths , so the effective change in the differential cross - section of impurity @xmath54 coming from both ( 1b@xmath52 ) and ( 1c@xmath52 ) , is expressed in terms of the return probability @xcite @xmath55 . \label{dsdiff}\ ] ] here @xmath56 is the isotropic cross - section in the drude approximation , @xmath57 is the impurity concentration , @xmath58 is the electron wavelength , and @xmath59 is given by @xmath60 where @xmath61 is the probability density for a diffusing particle to return after the time @xmath62 to the origin @xmath63 the coefficient @xmath64 in the cross - section correction was found in ref . by integration over small deviations of the electron trajectories from the trajectories ( 1b@xmath52 ) , ( 1c@xmath52 ) meeting the phase stationarity requirement .
physically , @xmath65 is the characteristic area of the region around the origin into which the diffusing electron should return for the effective interference to occur .
the calculations show that the contributions of ( 1b@xmath52 ) and ( 1c@xmath52 ) have different signs .
the positive contribution represented in eq . by @xmath66
comes from the process ( 1b@xmath52 ) , while the negative one ( the term @xmath67 ) , from the process ( 1c@xmath52 ) . in other words ,
the enhancement of the differential cross - section at the angle @xmath68 due to the coherent backscattering is accompanied by a reduction of the scattering in other directions , the total cross - section remaining unchanged ( see fig .
[ fig2 ] ) .
we see that the correction to the effective impurity cross - section is @xmath69-dependent .
therefore , the correction to the collision integral in the time representation turns out to be nonlocal in time : @xmath70 here @xmath71 is the electron distribution function ( we consider a homogeneous case , assuming that @xmath72 is independent of @xmath73 ) and @xmath74 are the angles of @xmath75 and @xmath76 here and further we omit the `` outflux '' term @xmath77 in the collision integral .
it makes a zero contribution , because the total cross - section is not changed by the weak localization .
the probability density @xmath78 is found as a sum over the paths involving different numbers of scattering events ( see , for example , ref . )
@xmath79 where @xmath80 and @xmath81 at @xmath82 the function @xmath83 is the classical probability density for an electron starting to move from @xmath84 to experience the first collision around the point @xmath73 . in the framework of a diagrammatic approach ,
the function @xmath83 arises as the product of two spherical waves ( retarded and advanced green s functions ) @xmath85 here @xmath86 , @xmath87 and we took into account that @xmath88 a path involving @xmath89 scatterings contains @xmath90 functions @xmath91 for the clockwise propagation along the path and @xmath90 functions @xmath92 for the counterclockwise propagation . as a result , @xmath93 contains @xmath90 functions @xmath94 using eqs . , , , and , one can show that in the diffusion approximation ( @xmath95 ) , when the typical number of scattering events along a path is large ( @xmath96 ) , the function @xmath97 obeys the diffusion equation @xmath98 solving eq . and
taking @xmath99 we find that the return probability is given by @xmath100 , and @xmath101 a generalization of the above results to a system with a spin - split spectrum is straightforward .
since the electron spin rotates while passing along a closed loop , electron green s functions become operators with respect to the spin variables : @xmath102 they are the products of spherical waves and spin - rotation matrices describing the electron spin precession . the precession frequency
@xmath103 depends on the propagation direction @xmath104 due to the spin precession , the quantum correction to the effective cross section becomes spin - dependent : @xmath105 where the spin indices @xmath106 correspond to the electron trajectories , as shown in figs .
[ fig1]b@xmath52,c@xmath52 : @xmath107 , \label{sabcd}\ ] ] the correction to the boltzmann collision integral can be written as @xmath108_{\beta\gamma}=n_iv\int \delta \mathcal{s}^{\beta\gamma\alpha\theta}_{\omega}(\phi-\phi ' ) f_{\alpha\theta}(\mathbf p',\omega)d\phi ' , \label{dj}\ ] ] where @xmath109 is the momentum - dependent spin - density matrix . to derive the expression for @xmath110
, we first introduce the probability density for a diffusing electron to arrive after @xmath89 collisions at the point @xmath111 with the spin rotated by an angle @xmath112 @xmath113 the angle @xmath114 changes with the coordinates of the scattering points @xmath115 .
one can find it from the matrix equation @xmath116 where @xmath117 .
the spin - dependent return probability is then expressed via the total probability density @xmath118 taken at @xmath84 @xmath119 here @xmath120 and function @xmath121 is defined in appendix [ apa ] .
( note that , in the absence of the spin - orbit coupling , @xmath122 and @xmath110 is expressed as @xmath123 ) . what remains to be done
is to find an equation for @xmath124 to this end , the probabilities @xmath93 are related to each other by the recurrent equations @xmath125 the vector @xmath126 describes the change of the spin rotation angle in a ballistic path between two scattering . one can find it from the equation @xmath127 next , we sum eq . over @xmath89 and
take into account that @xmath128 after combersome but straightforward calculations , we find that @xmath129 ( the probability density to arrive after time @xmath62 to the point @xmath73 with the spin rotated by the angle @xmath112 ) is described by an equation similar to eq . : @xmath130 here @xmath131 is the angular momentum operator acting on the functions of vector @xmath112 ( see ref . ) and @xmath132 the explicit expressions for @xmath131 and for the common eigen - functions of @xmath133 and @xmath134 are presented in appendix [ apa ] .
expanding @xmath135 in a series over these functions , we keep the term with @xmath136 the corresponding eigenfunction @xmath137 is independent of the angle @xmath112 and @xmath138 .
this is the only term which survives at @xmath139 the other terms decay exponentially with characteristic times of the order of @xmath140 ( this statement is not true for a degenerate case , when @xmath141 depends on the component of @xmath131 along a single axis .
this case is discussed below ) . as a result ,
we obtain the following expression for the asymptotical behavior of @xmath142 @xmath143 now , we write the distribution function as @xmath144 where @xmath72 is the particle density in the momentum space , related to the electron concentration by @xmath145 @xmath146 is the unit matrix , and @xmath147 is the spin density . substituting eq . and eq . into eq .
and eq . , respectively ,
making a fourier transform , and taking into account eq . , we obtain the weak - localization - induced correction to the collision integral @xmath148 where @xmath149 and @xmath150 are given by @xmath151 here @xmath152 . using eq . , we find the asymptotical behavior of these functions @xmath153 note also that the spin - orbit coupling can be neglected in the time interval @xmath154 .
hence , @xmath155 and the expressions for @xmath156 and @xmath157 become @xmath158 we see that the difference between eqs . and
is in the numerical coefficients only .
for the case with a spin polarization uniform in space , the generalized kinetic equation is @xmath159 where @xmath160 is given by eq . .
this equation can be solved in the usual way . @xcite since @xmath161 the spin density can be represented as a sum of the isotropic part @xmath162 which depends on the electron energy only , and a small anisotropic correction @xmath163 which is linear in the electron momentum @xmath164 @xmath165 substituting eq . into eq . , using of eq . , and taking into account the equalities @xmath166 ( here the angular brackets stand for averaging over the momentum direction ) we obtain a closed relation for @xmath167 @xmath168 assuming that the spin polarization was created at @xmath169 with a density @xmath170 and neglecting the quantum correction , we get the exponential relaxation @xmath171 ( here @xmath172 is the theta - function ) .
this solution is valid until @xmath173 . for larger times
, the spin polarization should be found from the condition that the right - hand side of eq
. equals zero : @xmath174 so we find that the spin polarization has a long - living power tail at large times @xmath175 to conclude this section , we note that we neglected in our calculations the electron dephasing due to inelastic scattering .
such dephasing can be accounted for phenomenologically by introducing the factor @xmath176 into the right - hand side of eq . .
here @xmath177 is the phase - breaking time .
equation is invalid for the degenerate case , when the spin precession frequency @xmath103 is parallel to a certain vector @xmath178 for any electron momentum : @xmath179 for any @xmath180 in the classical limit , the component of the electron spin parallel to @xmath178 does not relax @xcite @xmath181 and the two perpendicular components relax with equal rates @xmath182 the off - diagonal components of @xmath183 being equal to zero ( @xmath184 ) .
this happens in symmetric quantum wells grown in the [ 110 ] direction , @xcite as well as in asymmetric quantum wells grown in the [ 001 ] direction , due to the interplay between dresselhaus and rashba couplings .
@xcite to find the long - time asymptotic of the return probability , we write the formal solution of eq .
as @xmath185 in the degenerate case , each of the three operators @xmath186 and @xmath187 is proportional to @xmath188 which is the component of @xmath131 along the precession axis
. therefore , these three operators commute with each other . changing the integration variables @xmath189 in eq .
we obtain @xmath190 thus , after travelling around a closed loop , the electron spin does not rotate at all .
this can be interpreted as follows .
@xcite for the degenerate case , the spin rotation angle is simply given by @xmath191 for a closed loop , we have @xmath192 and , as a consequence , @xmath193 after substituting eq . into eq . and using eqs .
, , one can see that the weak localization does not affect the longitudinal component of the spin density @xmath194 while the relaxation of the perpendicular component is described by the following equation @xmath195 the integral in the right - hand side of eq .
contains an additional factor @xmath196 as compared with that in eq . .
it can be shown that this factor arises from the contribution of the eigenfunctions with @xmath197 to the long - time asymptotic of the return probability .
@xcite the long - time asymptotic of the spin polarization is given by @xmath198
next , we consider the influence of the external magnetic field on the long - living power tail in the spin polarization . when an external field @xmath199 is present , the spin rotation matrices appearing in @xmath200 and @xmath201 become different : @xmath202 for @xmath200 and @xmath203 for @xmath204 here
, @xmath205 is the frequency of the spin precession in the field @xmath199 .
so , eq . is modified as @xmath206 where @xmath207 here @xcite @xmath208 \right),\ ] ] and @xmath209 is the component of the magnetic field normal to the quantum well plane .
the vectors @xmath114 and @xmath210 should be found from eq . with @xmath211r / v$ ] for @xmath114 and @xmath212r / v$ ] for @xmath210 .
writing out the recurrent equations for the consecutive terms in eq . , we can derive the following expression @xmath213 for the probability density @xmath214 here the operators @xmath131 and @xmath215 are given by eq .
( with the replacement @xmath216 for @xmath215 ) .
the term @xmath217 in eq . is responsible for the magnetic field effect on the electron orbital motion . @xcite here @xmath218/2 $ ] is the vector potential .
we assume that the magnetic field is small @xmath219 for @xmath220 the solution of eq .
is @xmath221 where @xmath222 obeys eq . .
using eq . , we rewrite the long - living solution ( which corresponds to zero total angular momentum @xmath223 ) in terms of two angles @xmath224 : @xmath225 here @xmath226 is the eigenvalue of the operator @xmath227 at @xmath228 and @xmath229 is the respective eigenfunction . for @xmath230
the long time dynamics of the spin relaxation is determined by the eigenvalues of the operator @xmath231 first , we neglect the term @xmath232 in this approximation , the eigenvalues of @xmath233 corresponding to @xmath234 are given by @xmath235 where @xmath236 and @xmath237 since the operator @xmath238 has a discrete spectrum , we have to make the replacement @xmath239 in eq . .
the second step is to take into account the term @xmath240 taking it to be a small perturbation . in the first order of the perturbation theory , this term leads to the mixing of the eigenfunctions with @xmath241 to the eigenfunction with @xmath228 .
this mixing can be disregarded at @xmath242 corrections to the eigenvalues @xmath243 arise in the second order only .
they are calculated in appendix [ apb ] .
using eq . , we can show that the spin polarization dynamics at large times is described as @xmath244 thus , we see that the magnetic field does suppress the long - living tail in the spin polarization .
next we discuss briefly the physical meaning of the results obtained .
our calculations were based on the interpretation of the weak localization phenomena in terms of two scattering processes : coherent backscattering ( see fig . [ fig1]b@xmath52 ) and coherent scattering at an arbitrary angle ( see fig . [ fig1 ] @xmath50 ) . the existence of the long - living spin polarization can be explained as follows .
both coherent scattering processes were shown to be nonlocal in time .
in other words , the transition of a spin @xmath245 to a spin @xmath246 caused by coherent scattering , takes a certain time @xmath247 which may be relatively long : @xmath139 the coherent scattering events do not change the direction of the electron spin . indeed , as seen from eq . , @xmath248 and , as a consequence , @xmath249 therefore
, the electrons involved in such a scattering can keep memory about the initial spin polarization during the time much longer than @xmath140 the power law @xmath0 is due to the proportionality of the probability of coherent scattering to the probability of diffusive return .
the collision integral accounting for both coherent processes does not change the total scattering cross - section .
therefore , @xmath250 using eqs .
, , , and taking into account that @xmath251 and @xmath252 , we obtain @xmath253 eqs . , imply that the effect of localization on the angular spin diffusion , as well as that on the particle diffusion , can be accounted for by the @xmath69-dependent renormalization of the transport scattering time
. however , the quantum corrections to this time have differnt signs for the particle and angular spin diffusion : @xmath254 for the particle diffusion and @xmath255 for the spin diffusion .
this implies that the eq . is invalid in the quantum case , or , more precisely , it relates @xmath2 and @xmath256 rather than @xmath2 and @xmath257 .
an important comment should be made concerning the role of the coherent nonbackscattering contribution ( processes @xmath50 in fig .
[ fig1 ] ) . neglecting this contribution
, we have @xmath258 it can be easily shown that this leads to a physically meaningless result for the spin relaxation rate .
therefore , the correct treatment of the effect of weak localization on the spin relaxation is only possible when the nonbackscattering coherent processes are taken into account ( the role of such effects for particle diffusion was discussed in ref . ) .
it is worth noting that the weak localization effects are usually considered to be due to the coherent backscattering only .
the point is that the quantum correction to the conductivity is usually calculated by means of the kubo formula , which expresses conductivity in terms of the current - current correlation function .
this approach focuses on the calculation of the velocity correlation function , which depends on the anisotropic part of the distribution function @xmath259 ; so , there is no need to know corrections to @xmath260 the situation is quite different for spin relaxation which is due to the relaxation of the isotropic part of the distribution function .
@xcite to conclude , we have discussed the long - time dynamics of the spin polarization in a 2d disordered semiconductor .
it is shown that , at large times , the spin relaxation slows down due to weak localization effects .
an analytical expression for the long - living tail of the spin - polarization has been derived . the magnetic field
, both transversal and longitudinal , suppresses this tail , restoring the exponential relaxation .
the authors are grateful to a.p .
dmitriev and k. kavokin for useful discussions .
this work has been supported by rfbr , a grant of the ras and a grant of the russian scientific school 2192.2003.2 .
99 semiconductor spintronics and quantum computation , eds .
d. d. awschalom , d. loss , and n. samarth , in the serie `` nanoscience and technology '' , eds . k. von klitzing , h. sakaki , and r. wiesendanger ( springer , berlin , 2002 ) .
m. i. dyakonov and v. i. perel , fiz .
tela , * 13 * , 3581 ( 1971 ) [ sov .
phys . solid state , * 13 * , 3023 ( 1972 ) ] .
g. dresselhaus , phys .
rev . , * 100 * , 580 ( 1955 ) .
a. bychkov and e. i. rashba , pisma zh .
teor . fiz . ,
* 39 * , 66 ( 1984 ) [ jetp lett .
v. 39 , 78 ( 1984 ) ] .
m. i. dyakonov and v. yu .
kachorovskii , fiz .
pluprovodn.,*20 * , 178 ( 1986 ) [ sov .
semiconductors , * 20 * , 110 ( 1986 ) ] .
l. p. gorkov , a. i. larkin , and d. e. khmelnitskii , pisma zh .
, * 30 * , 248(1979)[jetp lett.,*30 * , 248 ( 1979 ) ] p. a. lee and t. v. ramakrishnan , rev .
phys . , v. 57 , 287 ( 1985 ) .
a. singh , phys .
b , * 39 * , 505 ( 1989 ) ; * 40 * , 783 ( 1989 ) a. g. malshukov , k. a. chao and m. willander , phys . rev .
lett . , v. 76 , 3794 ( 1996 ) a. g. malshukov , k. a. chao and m. willander , phys .
b , v. 52 , 5233 ( 1995 ) j. nitta , t. akazaki , h. takayanagi , and t. enoki , phys .
lett . , * 78 * , 1335 ( 1997 ) . s. hershfield and v. ambegaokar , physical review b , * 34 * , 2147 ( 1986 ) g. strinati , c. castelani , c. di castro , phys .
b , * 40 * , 12237 ( 1989 ) a. p. dmitriev , i. v. gornyi , and v. yu . kachorovskii , phys
b , * 56 * , 9910 ( 1997 ) eq . is valid in the diffusion approximation ( @xmath261 ) when the typical number of scattering events along the closed path is large . in the ballistic case ( @xmath262 )
the change of the effective cross - section can also be expressed via return probability .
@xcite however , one should take into account the dependence of the return probability on @xmath263 s. chakravarty and a. schmid , physics report , * 140 * , 195 ( 1986 ) n. ja .
vilenkin , _ representation of lie groups and special functions _ , kluwer academic publishers group ( 1992 ) more precisely , a slow relaxation of this component takes place due to the turms cubic in @xmath75 , neglected in the hamiltonian .
f. g. pikus and g. e. pikus , phys .
b , * 51 * , 16928 ( 1995 ) n. s. averkiev and l. e. golub , phys
b , * 60 * , 15582 ( 1999 ) in the conventional terminology of the weak localization , this means that in the degenerate case the triplet channel makes the same contribution as the singlet one , @xcite while in the non - degenarate case the non - zero contribution comes from the singlet channel only .
b. l. altshuler , a. g. aronov , and d. e. khmelnitskii , j. phys .
c : solid state phys .
, * 15 * , 7367 ( 1982 ) note that the account of diagrams ( 1c ) is also necessary for calculation of weak localization correction to the density - density correlation function ( see ref . ) .
m. hamermesh , _ group theory and its application to physical problems _ , dover publications inc .
, new york ( 1964 ) s. v. iordanskii , yu . b. lyanda - geller , and g. e. pikus , jetp letters,*60 * , 206 ( 1994 ) ( pisma v jetp,*60 * , 199 ( 1994 ) ) w. knap _ et al _ , phys .
b * 53 * 3912 ( 1996 ) a. g. malshukov and k. a. chao , phys .
b , * 61 * , r2413 ( 2000 )
the analytical expression for the angular momentum operator @xmath131 is @xmath264\right]- \left[\frac{\boldsymbol{\phi}}{2}\times\frac{\partial}{\partial \boldsymbol{\phi}}\right]\right ) , \label{l}\ ] ] with @xmath265 the components of this operator obey the usual commutation rules , @xmath266=i\epsilon_{ijl}\hat l_l.$ ] the common eigenfunctions of @xmath133 and @xmath134 are wigner s rotation matrices @xcite @xmath267 : @xmath268 the values of @xmath269 may be integer and half - integer @xmath270 ( as usual , @xmath271 ) .
there is also @xmath272 degeneracy with respect to @xmath273 the eigenfunction corresponding to @xmath274 is equal to unity @xmath275 the orthogonality conditions are @xcite @xmath276 here the integration is made over all possible transformations of the @xmath277 group .
the integration measure is taken in the invariant form : @xmath278 note that @xmath277 transformations are usually parameterized by euler angles @xmath279 .
for such a parametrization , the invariant integration measure is given by @xcite @xmath280 ( the expression for @xmath131 via the euler angles is also given in ref . ) .
it is convenient for us to use the components of vector @xmath112 ( @xmath281 ) instead of euler angles .
this gives@xcite @xmath282 the expansion of the delta function @xmath283 in wigner s functions is @xmath284 the solution to eq .
can be represented as a sum over all angular momenta @xmath285 however , as follows from eq .
only two terms ( @xmath274 and @xmath197 ) contribute to the spin - dependent return probability .
indeed , the action of the operator @xmath286 on the spin - rotation matrices is given by @xmath287 as follows from eq .
, the projection of the operator @xmath286 onto the subspace formed by products of two rotation matrices is given by @xmath288 where @xmath289 and @xmath290 are the pauli matrices acting on the first and second rotation matrices , respectively .
therefore , the angular momentum @xmath291 can only be 0 or 1 ( singlet and triplet contributions ) .
using eq . and the property @xmath292 valid for arbitrary functions @xmath293 and @xmath294 we can see that the function @xmath295 obeys the equation similar to eq .
@xmath296_{\beta \beta ' \theta \theta ' } w^{\beta'\gamma \alpha \theta ' } = \delta(t ) \delta(\mathbf r)\delta_{\alpha \beta } \delta_{\gamma \theta } , \label{drdif1}\ ] ] where matrix elements @xmath297 are given by eq . .
note that eq . was used in refs . for calculation of weak localization corrections to conductivity .
the alternative derivation of the operator @xmath298 was given in ref .
using eq . , we expand eq . as @xmath299 where @xmath300 is the full set of the eigenfunctions ( @xmath301 ; @xmath302 ; @xmath303 ) for zero total angular momentum @xmath304 in the second order of the perturbation theory
, the term @xmath305 leads to the corrections to the eigenvalues of the operator @xmath306 .
these corrections are different for the functions @xmath307 with different values of @xmath308 they are calculated as @xmath309 next , we diagonalize the operator @xmath310 in the subspace formed by the three functions @xmath311 , and @xmath312 direct calculation gives @xmath313 the expression for the long - living solution becomes @xmath314 where @xmath315 . substituting eq . into eq . and using eqs . , , and
we derive eq . .
|
the effect of weak localization on spin relaxation in a two - dimensional system with a spin - split spectrum is considered .
it is shown that the spin relaxation slows down due to the interference of electron waves moving along closed paths in opposite directions . as a result
, the averaged electron spin decays at large times as @xmath0 .
it is found that the spin dynamics can be described by a boltzmann - type equation , in which the weak localization effects are taken into account as nonlocal - in - time corrections to the collision integral .
the corrections are expressed via a spin - dependent return probability .
the physical nature of the phenomenon is discussed and it is shown that the `` nonbackscattering '' contribution to the weak localization plays an essential role .
it is also demonstrated that the magnetic field , both transversal and longitudinal , suppresses the power tail in the spin polarization .
|
recently , the investigation of spin - orbit ( so ) coupling in neutral atoms has attracted much attentions @xcite . in particular , a one - dimensional ( 1d ) equal rashba and dresselhaus so coupling has been first realized in the ultracold @xmath0rb atoms by a couple of raman lasers @xcite . by applying the same laser technique , this 1d so
coupling has been also achieved experimentally in the degenerate fermi gas with @xmath1k @xcite and @xmath2li @xcite .
theoretical investigations have been revealed that in the presence of so coupling , the degenerate fermi gas can exhibit the interesting physics in both three jpv , gm , yzq , hh , jpv1,iskin , lj , wy1,ld , li , ks , kz , rl , zhangp1,peng1,hly1,ld1 and lower dimensions @xcite .
for example , by increasing the strength of so coupling , the density of state at the fermi surface is increased , and the cooper paring gap can be thus enhanced significantly @xcite .
more importantly , this system may be changed from the bardeen - cooper - schrieffer ( bcs ) superfluid to the bose - einstein condensate ( bec ) with a new two - body bound state called rashbon @xcite . when an effective zeeman field is applied , the 2d degenerate fermi gas with the rashba so coupling exhibits an exotic topological superfluid supporting the majorana fermions @xcite , which is the heart for realizing the topological quantum computing @xcite .
recently , a universal midgap bound state in the topological superfluid has been predicted @xcite . to illustrate the so - driven fundamental physics ,
a generalized one - channel model has been introduced in all previous considerations jpv , gm , yzq , hh , jpv1,iskin , lj , wy1,ld , li , ks , kz , rl , zhangp1,peng1,hly1,ld1,chen , mg , hly2,jz , wy2,xyang , zjn , fw . in this one - channel model ,
only the atoms tuned via the feshbash - resonant technique are taken into account .
however , it is valid for the _ broad _ feshbash - resonant regime with @xmath3 @xcite , where the dimensionless parameter is defined as @xmath4 , @xmath5 is the bohr magneton , @xmath6 is the atom mass , @xmath7 is the background @xmath8-wave scattering strength , @xmath9 is the resonant width and @xmath10 is the fermi energy .
in fact , to get a more realistic and complete description of the degenerate fermi gas , especially in the _ narrow _ feshbash - resonant limit with @xmath11 , we must introduce a two - channel model @xcite , which includes both the atoms in the open channel and the molecules in the closed channel .
moreover , in the narrow feshbash - resonant regime , some fundamental properties can be observed experimentally by detecting the striking molecular signature @xcite , additional to measuring the superfluid pairing gap applied usually in the one - channel model @xcite .
more importantly , new quantum phase transitions have been predicted yn , zs , attributed to the existence of extra u(1 ) symmetry for the molecular field . on the experimental side ,
the degenerate fermi gas in the narrow feshbach - resonant regime has been also reported successfully in @xmath2li @xcite and the fermi - fermi mixture of @xmath2li and @xmath1k @xcite .
thus , it is crucially important to explore the so - induced exotic physics in this regime @xcite . in the present paper
we investigate the so - induced bound state and molecular signature of the degenerate fermi gas in the narrow feshbach resonance .
the main results are given as follows .
( i ) without the atom - atom interactions , only one bound state can be found even if so coupling exists .
moreover , the corresponding bound - state energy depends strongly on the strength of so coupling , but is influenced slightly on its type .
( ii ) with the increasing of the strength of so coupling , the critical point at which the molecular fraction vanishes shifts from zero to the negative detuning . in the weak so coupling ,
this shifting is proportional to the square of its strength .
( iii ) finally , we also show that the molecular fraction can be well controlled by so coupling .
we believe that in experiments it is a good signature to detect the so - induced physics .
for the so - driven two - channel model , the total hamiltonian can be written formally as @xmath12 in hamiltonian ( [ total h ] ) , @xmath13is the atom hamiltonian , where @xmath14 is the creation operator for a atom with the momentum @xmath15 and @xmath16 , and @xmath17 * * * * is the kinetic energy of the atom . @xmath18is
the molecular hamiltonian , where @xmath19 is the creation operator of a molecule with the momentum @xmath20 , @xmath21 with @xmath22 is the kinetic energy of the molecule , and @xmath23 is the bare detuning determined by the feshbash - resonant position @xmath24 via a relation @xmath25without so coupling , the system has the bcs superfluid in the positive detuning ( @xmath26 ) , and enters into the bec regime in the negative * * * * detuning ( @xmath27 ) * * * * @xcite*. * at lower energy , the position is given approximately by @xmath28 , where @xmath29 is the magnetic field at which the resonance is at zero energy , and @xmath30 is the tunable magnetic field @xcite .
the atom - molecule interconversion term is governed by the following hamiltonian @xmath31where @xmath32 is the coupling constant that measures the amplitude of the decay of the molecule in the closed channel into a pair of the open - channel atoms . * * * * finally , the so coupling is chosen as a generalized rashba and dresselhaus type .
the corresponding hamiltonian is given by @xmath33 \label{hs}\]]with @xmath34 and @xmath35 , where @xmath36 and @xmath37 are the so coupling strengths for the rashba and dresselhaus types , respectively .
clearly , @xmath38 is the generalized strength of so coupling and the dimensionless parameter @xmath39reflects the competition between these different types of so coupling .
for example , for @xmath40 ( @xmath41 ) , the 2d rashba so coupling can be found . whereas , for @xmath42 ( @xmath43 ) , the 1d equal rashba and dresselhaus so coupling can be generated .
fortunately , this 1d so coupling has been realized experimentally in the ultracold neutral atoms @xcite . in the absence of so coupling ,
the limit @xmath44 can be applied usually to discuss the standard two - channel model including the effective zeeman field @xcite . however , in the presence of so coupling , the result is quite complicated .
if both so coupling and the effective zeeman field are taken into account , the parity and time - reversal symmetries are broken . as a result , the @xmath20-dependent order parameter should be introduced dfa . however , in this paper we do not consider the effect of the effective zeeman field and thus may focus on the case of @xmath44 .
we begin to discuss the two - body bound state of the generalized two - channel model ( [ total h ] ) by introducing the ansatz wavefunction . in the absence of so coupling ( @xmath45 ) ,
hamiltonian ( [ total h ] ) reduces to the standard two - channel model @xmath46 @xcite , in which only the singlet cooper paring can be formed .
however , in the presence of so coupling , both the singlet and triplet cooper parings can coexist @xcite . as a result
, the ansatz wavefunction should be written formally as @xmath47where @xmath48 and @xmath49 ( @xmath50 and @xmath51 ) represent the probability amplitude of the singlet ( triplet ) cooper paring , @xmath52 stands for the probability amplitude of the molecule , and @xmath53 is the direct multiple of the fermion vacuum with spin flipping and the molecule vacuum . substituting the wavefunction in eq .
( [ ws ] ) into the stationary schrdinger equation @xmath54we find that the six coefficients determining the ansatz wavefunction @xmath55 and the energy @xmath56 are governed by the following equations : @xmath57where @xmath58 and @xmath59 .
( [ eqc ] ) can not be solved directly because of lack of a coefficient equation . if we define the spin symmetry and anti - symmetry vectors as @xmath60the stationary schrdinger equation is rewritten as @xmath61 in the representation of @xmath62^{t}$ ] .
this leads to another equations for the coefficients @xmath63 and @xmath64 , that is , @xmath65substituting eq .
( [ ace ] ) into eq .
( [ eqc ] ) yields @xmath66 ) as a function of the detuning @xmath67 for the different strengths of so coupling ( a ) @xmath68 , ( b ) @xmath69 , and ( c ) @xmath70 when @xmath71 . in fig .
1(a ) , the red open symbol corresponds to the analytical result ( ar ) and the black solid line represents the direct numerical simulation ( ns).,width=283 ] equation ( [ meq ] ) , which is the main result of this paper , determines the so - induced bound - state energy of the generalized two - channel model ( total h ) .
the procedure is given as follows .
( i ) we first obtain the energy @xmath56 from eq .
( [ meq ] ) .
( ii ) then , we introduce the threshold energy @xmath72 , which is the lowest energy of the free particle ( i.e. , the lowest band ) , to judge whether this energy is the bound - state energy or not . if @xmath73 , the bound state exists and the corresponding energy @xmath56 is called the bound - state energy , and vice versa @xcite . according to its definition ,
the threshold energy @xmath72 of the generalized two - channel model ( [ total h ] ) is evaluated as @xmath74 in the absence of so coupling ( @xmath45 ) , this threshold energy becomes @xmath75 , as expected . in the following discussions , we mainly consider two interesting cases , including the 2d rashba so coupling ( @xmath40 ) and 1d equal rashba and dresselhaus so coupling ( @xmath42 ) , to reveal the fundamental properties of the bound state .
we first address the case of the 2d rashba so coupling ( @xmath40 ) . in the absence of so coupling ( @xmath45 ) ,
the analytical bound - state energy is derived from eq .
( [ meq ] ) by @xmath76it implies that in such a case only one bound state can be found @xcite . in the presence of so coupling ( @xmath77 ) , the explicit expression for the bound - state energy
can not be obtained . however , for the weak so coupling , eq .
( [ meq ] ) is simplified as @xmath78}{8\pi \hbar ^{5}\sqrt{-e}}=\frac{e-\delta } { g^{2 } } \label{eeq}\]]with the help of a taylor expansion with respect to the strength of so coupling . in this case , hamiltonian ( [ total h ] ) also exhibits one bound state , as shown in fig .
1(a ) . by further solving eq .
( [ eeq ] ) approximately , we find that the bound - state energy is proportional to @xmath79
. this behavior agrees well with the numerical simulation , as also shown in fig .
it implies that the bound - state energy can decrease by increasing the strength of so coupling . for the strong so coupling
, the perturbation method is invalid .
here we numerically solve eq .
( [ meq ] ) to evaluate the bound - state energy @xmath56 .
even if the strong so coupling exists , only one bound state can be found , as shown in figs .
1(b ) and 1(c ) for the different types of so coupling including @xmath80 ( the 2d rashba so coupling ) and @xmath81 ( the 1d equal rashba and dresselhaus so coupling ) when @xmath71 and @xmath82.,width=279 ] in fig .
[ fig2 ] , we plot the bound - state energy with respect to the detuning @xmath24 for the different types of so coupling including @xmath83 ( the 2d rashba so coupling ) and @xmath42 ( the 1d equal rashba and dresselhaus so coupling ) . it can be seen clearly that the bound - state energy is affected slightly by the type of so coupling .
when the feshbach - resonant width parameter @xmath84 increases , the system changes from the narrow limit ( @xmath11 ) to the broad limit ( @xmath85 ) .
especially , for the broad limit ( @xmath3 ) , eq .
( [ meq ] ) becomes @xmath86which is similar to the result of ref .
@xcite . in fig .
[ fig3 ] , we plot the bound - state energy for the 1d equal rashba and dresselhaus so coupling as a function of the detuning @xmath24 for the different feshbach - resonant width parameters ( a ) @xmath71 , ( b ) @xmath87 , and ( c ) @xmath88 .
this figure shows again that for the broad limit our considered two - channel model reduces to the single - channel model .
however , it should be remarked that the energy @xmath56 for the broad feshbach resonance is not a bound - state energy , but is a two - body interaction energy approaching infinitely the bound - state energy @xcite as a function of the detuning @xmath67 for the different feshbach - resonant widths ( a ) @xmath89 , ( b ) @xmath90 , and ( c ) @xmath88 when @xmath91 . in ( b ) and ( c ) , the red dash line corresponds to the numerical solution of eq .
( [ eeqn ] ) . for the 2d rashba so coupling
, the similar conclusion can be also found.,width=298 ]
having obtained the bound - state energy in the generalized two - channel model , it is conveniently to discuss the experimentally - measurable molecular signature . in terms of the hellmann - feymann theorem ,
the molecular fraction is obtained by @xmath92where the bound - state energy @xmath56 can be derived from eq .
( [ meq ] ) .
figure [ fig4 ] is plotted the scaled molecular fraction @xmath93 of both the 2d rashba so coupling ( @xmath40 ) and the 1d equal rashba and dresselhaus so coupling ( @xmath42 ) with respect to the detuning @xmath24 for the different strengths of so coupling . in the absence of so coupling ( @xmath94 ) , the molecule exists in the negative detuning ( @xmath27 ) .
however , for the positive detuning ( @xmath95 ) , the physical bound state vanishes @xcite , i.e. , there is no real molecular fraction . with the increasing of the strength of so coupling , the critical point
at which the molecular fraction vanishes shifts from zero to the negative detuning .
the physics can be understood as follows . in the generalized two - channel model ,
the molecules play two roles .
one is that they interact directly with the atoms via hamiltonian @xmath96 .
the other ( the most important ) is that they induce the indirect atom - atom interactions , which generate the cooper pairing .
when the cooper pairing is enhanced by so coupling gm , yzq , hh , the molecules are thus suppressed because the system need guarantee a conserved number @xmath97 . in order to see clearly
this behavior induced by so coupling , we introduce a key parameter @xmath98 . this parameter describes the maximum detuning at which the molecular fraction exists . in terms of the definition , the parameter @xmath98 is given by@xmath99_{e = e_{t}}. \label{dm}\]]in the case of the weak so coupling , the parameter @xmath98 is obtained explicitly by@xmath100eq .
( [ dmeq ] ) shows clearly that @xmath98 decreases with the increasing of the strength of so coupling . and ( b ) the 1d equal rashba and dresselhaus so coupling with @xmath81 as a function of the detuning @xmath67 for the different strengths of so coupling when @xmath89.,width=302 ] in fig .
[ fig5 ] , we plot the molecular fraction of both the 2d rashba so coupling ( @xmath40 ) and the 1d equal rashba and dresselhaus so coupling ( @xmath42 ) with respect to the strength of so coupling for the different detunings . in the negative detuning ( @xmath101 ) we shows again that the so coupling suppresses the molecular fraction . with the increasing of the strength of so coupling , the molecular fraction also vanishes .
it means that in experiment the molecular fraction can be well controlled by tuning the so strength .
in addition , in the positive detuning , no molecular fraction can be found even if so coupling exists . and ( b ) the 1d equal rashba and dresselhaus so coupling with @xmath81 as a function of the strength @xmath102 of so coupling for the different detunings when @xmath71.,width=302 ]
in summary , motivated by the recent experimental developments , we have investigated the so - driven degenerate fermi gas in the narrow feshbash resonance based on the generalized two - channel model .
we have found that in the absence of the atom - atom interactions , only one bound state can be found even if so coupling exists .
in addition , we have shown that the molecular fraction can be well controlled by so coupling .
we believe that in experiments it is a good signature to explore the so - induced physics .
we thank professors peng zhang , wei yi , wei zhang , shizhong zhang and doctor zengqiang yu for their helpful discussions .
this work was supported partly by the 973 program under grant no .
2012cb921603 ; the nnsfc under grants no .
10934004 , no . 11074154 , and no .
61275211 ; nnsfc project for excellent research team under grant no . 61121064 ; and
international science and technology cooperation program of china under grant no.2001dfa12490 .
_ note added_during preparing this paper , we noticed that two bound states for the so - driven two - channel model with the atom - atom interactions was predicted by v. b. shenoy in terms of a renormalizable quantum field theory @xcite .
however , that paper does not consider the molecular fraction .
|
in this paper we explore the spin - orbit - induced bound state and molecular signature of the degenerate fermi gas in a narrow feshbach resonance based on a generalized two - channel model . without the atom - atom interactions ,
only one bound state can be found even if spin - orbit coupling exists .
moreover , the corresponding bound - state energy depends strongly on the strength of spin - orbit coupling , but is influenced slightly by its type .
in addition , we find that when increasing the strength of spin - orbit coupling , the critical point at which the molecular fraction vanishes shifts from zero to the negative detuning . in the weak spin - orbit coupling ,
this shifting is proportional to the square of its strength .
finally , we also show that the molecular fraction can be well controlled by spin - orbit coupling .
|
the existence of vortex glass phase in the mixed state of quasi two - dimensional ( 2d ) high-@xmath0 superconductors such as yba@xmath9cu@xmath10o@xmath11 in the presence of magnetic field ( @xmath1 ) attracts considerable attention @xcite .
mller et al .
@xcite have pointed out for the first time the existence of the glassy phase in la@xmath9cuo@xmath12:ba .
they have shown that the boundary between the vortex liquid phase and the vortex glass phase corresponds to the de almeida - thouless ( at ) @xcite line for the spin glass behavior . since
then there have been a number of experimental works on the vortex glass phase in quasi 2d superconductors .
experimentally it has been confirmed that the @xmath1-@xmath2 diagram consists of the vortex glass , vortex lattice ( abrikosov lattice ) , and vortex liquid phases .
the boundary between the vortex lattice phase and the vortex liquid phase ( the line @xmath13 ) is of the first order , while the boundary between the vortex lattice phase and the vortex glass phase ( the line @xmath14 ) and the boundary between the vortex glass phase and the vortex liquid phase ( the line @xmath5 ) are of the second order .
these boundaries merge into a multicritical point at @xmath2 = @xmath3 and @xmath1 = @xmath4 .
sn - metal graphite ( mg ) constitutes a novel class of materials having unique layered structures .
this system can be prepared by reduction of an acceptor - type sncl@xmath9 graphite intercalation compound ( gic ) as a precursor material .
ideally , the staging structure of sn - mg would be the same as that of sncl@xmath9 gic .
sn layer is sandwiched between adjacent graphene sheets .
these sandwiched structures are periodically stacked along the @xmath15 axis perpendicular to the basal plane of graphene sheets .
sn - mg undergoes a superconducting transition at a critical temperature @xmath0 (= 3.75 k ) , which is close to that of bulk sn @xcite .
the superconductivity occurs mainly in sn metal layers , forming a quasi 2d superconductor ( type - ii ) .
in this paper we report results on the magnetic properties of sn - mg from measurements of dc and ac magnetic susceptibility .
three methods are used to determine the @xmath1-@xmath2 diagram : ( i ) the @xmath2 dependence of the magnetization ( @xmath16 ) in various states such as the fc ( field - cooled ) state , zfc ( zero - field cooled ) state , ir ( isothermal remnant ) state , and tr ( thermoremnant ) state , ( ii ) the @xmath2 , @xmath1 , @xmath17 , and @xmath18 dependence of the dispersion ( @xmath19 or @xmath20 ) and absorption ( @xmath21 or @xmath22 ) of the ac magnetic susceptibility , where @xmath17 and @xmath18 are the amplitude and frequency of the ac magnetic field , respectively , and ( iii ) the @xmath1 dependence of @xmath16 .
we find that there are four lines ( @xmath14 , @xmath5 , @xmath6 and @xmath13 ) in the @xmath1-@xmath2 phase diagram , separating vortex liquid , vortex glass , and vortex lattice phase .
these lines merge at the multicritical point located at @xmath3 = 3.4 k and @xmath4 = 40 oe .
the lines @xmath5 and @xmath6 are irreversibility lines , intersecting at a critical point at @xmath23 = 2.5 k and @xmath8 = 160 oe .
the line @xmath5 has an at - power law form,@xcite while the line @xmath6 has a gabay - thouless ( gt)-power law form @xcite .
the nature of possible slow dynamics and nonlinearity in the vortex glass phase is examined from the measurement of @xmath19 and @xmath21 as a function of @xmath24 , @xmath17 , @xmath2 , and @xmath1 .
the method of sample preparation in sn - mg is similar to that in other mg s previously reported @xcite .
sncl@xmath9 graphite intercalation compound ( gic ) samples as a precursor material , were prepared by heating a mixture of natural graphite flakes ( grade : rfl 99.9 s ) from kropfmhl , germany and an excess amount of sncl@xmath9 .
the reaction was continued for three days at 400 @xmath25c in an ampoule filled with chlorine gas .
the synthesis of sn - mg was made by the reduction of sncl@xmath9 gic .
sncl@xmath9 gic samples were kept for two days in a solution of lithium diphenylide in tetrahydrofuran ( thf ) at room temperature .
then the samples were filtered , rinsed by thf , and dried in air .
the structure of sn - mg thus obtained was examined by ( 00@xmath26 ) x - ray diffraction , and bright field images and selected - area electron diffraction ( saed ) ( hitachi h-800 transmission electron microscope ) operated at 200 kv .
no pristine sncl@xmath9 and its hydrolisation products were left in the sn layers .
the details of sample preparation and characterization in sn - mg will be reported elsewhere .
the sn - mg sample consists of many small flakes .
each flake has a well - defined @xmath15 axis .
the sample in the present work is regarded as a powdered sample with the @xmath15 axis randomly distributed in all directions , since these flakes are randomly piled inside the sample capsule .
the measurements of dc and ac magnetic susceptibility were carried out using a squid magnetometer ( quantum design mpms xl-5 ) . before setting up a sample at 298 k , a remnant magnetic field
was reduced to less than 3 moe using an ultra - low field capability option . for convenience , hereafter this remnant field is noted as the state @xmath1 = 0 .
( i)_zfc and fc magnetic susceptibility _ : the sample was cooled from 298 to 1.9 k at @xmath1 = 0 .
after @xmath1 was applied at 1.9 k , the zero - field cooled magnetization ( @xmath27 ) was measured with increasing @xmath2 from 1.9 to 6 k. the sample was kept at 20 k for 20 minutes in the presence of @xmath1 .
then the field cooled magnetization ( @xmath28 ) was measured with decreasing @xmath2 from 6 to 1.9 k. ( ii ) @xmath16-@xmath1 loop : the sample was cooled from 298 k to @xmath2 at @xmath29 .
the measurement was carried out with varying @xmath1 from 0 to 3 koe at @xmath2 , from @xmath1 = 3 to -3 koe , and from @xmath30 to 3 koe .
( iii ) @xmath27 vs @xmath1 : the sample was cooled from 298 k to @xmath2 at @xmath29 .
the zfc magnetization @xmath27 at @xmath2 was measured with increasing @xmath1 from 0 to 250 oe .
( iv ) _ ac magnetic susceptibility_. the dispersion ( @xmath19 ) and the absorption ( @xmath21 ) was determined from the in - phase and out - of phase first harmonic ( @xmath24 ) component in the ac magnetization measurement @xcite .
the dispersion ( @xmath19 ) and the absorption ( @xmath21 ) were described by @xmath31 and @xmath32 , where @xmath33 and @xmath34 ( @xmath35 ) are the real and imaginary parts of high - order nonlinear magnetic susceptibility , respectively . in the limit of @xmath36 , @xmath19 and @xmath21
coincide with the linear susceptibility @xmath20 and @xmath22 , respectively . before the measurement ,
the sample was cooled from 298 to 1.9 k at @xmath29 .
( i ) the @xmath2 dependence of @xmath19 and @xmath21 was measured with increasing @xmath2 from 1.9 to 6 k in the presence of @xmath1 .
after the measurement , the sample was cooled from 6 to 1.9 k in the presence of the same @xmath1
. then the measurement was repeated at different @xmath1 .
( ii ) the @xmath17 dependence was carried out at fixed @xmath2 by changing @xmath17 ( @xmath37 oe ) at @xmath38 hz .
( iii ) the @xmath18 dependence was carried out at fixed @xmath2 in the presence of @xmath1 by changing @xmath18 ( @xmath39 hz ) .
dependence of @xmath40 and @xmath41 at low @xmath1 ( @xmath42 oe ) for sn - mg.,width=302 ] dependence of @xmath40 and @xmath41 at high @xmath1 ( @xmath43 koe ) for sn - mg .
the inset shows the plot of the peak temperature of @xmath40 vs @xmath2 as a function of @xmath1 .
the dotted line is a least - squares fitting curve .
see the text for detail .
( b ) @xmath2 dependence of @xmath40 and @xmath41 at @xmath1 = 3 koe , which is a blow - up of fig .
[ fig : two](a).,width=302 ] figure [ fig : one ] shows the @xmath2 dependence of @xmath40 and @xmath41 at low @xmath1 .
the susceptibility @xmath41 deviates from @xmath40 below a characteristic temperature for @xmath42 oe , suggesting the irreversibility line in the @xmath1-@xmath2 plane which separates the reversible region from the irreversible region .
figure 2 shows the @xmath2 dependence of @xmath40 and @xmath41 for @xmath44 koe . for @xmath45
oe there is no noticeable difference between @xmath40 and @xmath41 at any @xmath2 .
the susceptibility ( @xmath46 ) exhibits a very broad peak , which shifts to the low-@xmath2 side as @xmath1 increases . in the inset of fig .
[ fig : two](a ) we plot these peak temperatures as a function of @xmath1 .
the value of @xmath1 (= @xmath47 ) for the peak is much larger than that of a upper critical field @xmath48 ( whose definition will be given in sec .
[ disa ] ) at the same @xmath2 , and it tends to zero around @xmath0 .
this result suggests that an antiferromagnetic ( af ) short - range order survives at high @xmath1 where the superconductivity disappears
. the origin may be due to the possible antiferromagnetism of nanographites in sn - mg ( see sec . [ disc ] ) .
dependence of the difference @xmath49
(= @xmath50 ) at low @xmath1 for sn - mg .
the solid lines are guides to the eyes.,width=302 ] figure [ fig : three ] shows the @xmath2 dependence of the difference @xmath49 (= @xmath51 ) at various @xmath1 .
the difference @xmath49 for @xmath52 ( @xmath53 oe ) decreases with increasing @xmath2 and reduces to zero at a characteristic temperature .
in contrast , the difference @xmath49 for @xmath54 oe decreases in two steps with increasing @xmath2 , showing a local minimum and a local maximum .
the difference @xmath49 shows a small peak at 2.25 k for @xmath55 oe and is almost equal to zero for any @xmath2 for @xmath56 oe .
dependence of @xmath27 , @xmath57 , @xmath28 , and @xmath58 for sn - mg .
the definition of these magnetizations and the method of the measurement are described in the text .
@xmath1 = 100 oe .
( b ) @xmath2 dependence of @xmath49 (= @xmath59 ) , @xmath60 (= @xmath61 ) , @xmath57 , and @xmath58 .
( c ) @xmath2 dependence of ( @xmath62 ) .
the solid line is guide to the eyes.,width=302 ] in a mixed state of conventional type - ii superconductors that have well - defined lower critical field @xmath63 and upper critical field @xmath48 ( @xmath64 ) , the thermodynamically most stable state involves filling the system with fluxoids , exhibiting only a partial meissner effect with the magnetic induction @xmath65 .
even a small degree of inhomogenity is sufficient to pin some fluxoids at local free energy minima , so that some fluxoids remain trapped in place where @xmath1 is removed . in sn - mg
the values of @xmath16 and @xmath66 are strongly dependent on the magnetic states such as the zfc , fc , ir ( isothermal remnant ) , and tr ( thermoremnant ) states .
experimentally , the values of @xmath16 and @xmath66 for these states are obtained as follows .
first the sample was cooled from 298 to 1.9 k at @xmath1 = 0 .
then @xmath1 was applied .
the measurements of @xmath27 and @xmath57 were done with increasing @xmath2 from 1.9 to 4 k. at each @xmath2 , @xmath27 was measured at @xmath1 and then @xmath57 was measured 100 sec later after the field was changed from @xmath1 to 0 oe .
second , the sample was annealed at 20 k for 1200 sec at @xmath1 .
the measurements of @xmath28 and @xmath58 were done with decreasing @xmath2 from 4 to 1.9 k. at each @xmath2 , @xmath28 was measured at @xmath1 = 100 oe and then @xmath58 was measured 100 sec later after the field was changed from @xmath1 to 0 oe .
the values of @xmath16 and @xmath66 for each state thus obtained are described by @xmath67 , @xmath68 , @xmath69 , and @xmath70 . here
@xmath71 and @xmath72 are the magnetic inductions in the fc and zfc states with @xmath1 , while @xmath73 and @xmath74 are the magnetic inductions in the ir and tr states with @xmath29 , respectively .
these magnetic inductions result from the flux trapping at pinning centers .
figure 4(a ) shows the @xmath2 dependence of @xmath27 and @xmath28 at @xmath75 oe ( @xmath76 ) , and @xmath57 and @xmath58 at @xmath29 .
the @xmath2 dependence of @xmath57 and @xmath58 is different from @xmath73 and @xmath74 by only a factor of 4@xmath77 .
the @xmath2 dependence of @xmath57 and @xmath58 is similar to that of typical spin glass systems such as 3d ising spin glass fe@xmath78mn@xmath79tio@xmath10 @xcite , where the competition between the nearest neighbor antiferromagnetic and ferromagnetic interactions leads to the spin frustration effect .
the magnetization @xmath57 exhibits a broad peak around 3.0 k , while @xmath58 decreases with increasing @xmath2 . both @xmath57 and @xmath58
reduce to zero at @xmath0 , implying the appearance of both @xmath73 and @xmath74 below @xmath0 .
the deviation of @xmath57 from @xmath58 occurs below 3.1 k , showing the irreversible effect of magnetization .
note that the point at @xmath2 = 2.94 k and @xmath1 = 100 oe is located on the line @xmath5 in the @xmath1-@xmath2 diagram ( see sec .
[ disa ] ) .
figure [ fig : four](b ) shows the @xmath2 dependence of @xmath60 [ = @xmath80 and @xmath49 [ = @xmath81 , as well as @xmath57 and @xmath58 . figure [ fig : four](c ) shows the @xmath2 dependence of ( @xmath82 ) .
the difference ( @xmath62 ) is equal to zero at @xmath2 above 3.25 k and exhibits a peak at 3.0 k where @xmath57 has a peak .
it becomes negative below 2.7 k. the temperature region between 3.0 and 2.4 - 2.7 k corresponds to the vortex glass phase for @xmath75 oe .
dependence of @xmath27 , @xmath57 , @xmath28 , and @xmath58 for sn - mg .
@xmath1 = 30 oe .
( b ) @xmath2 dependence of @xmath49 (= @xmath83 ) , @xmath60 (= @xmath61 ) , @xmath57 , and @xmath58 .
( c ) @xmath2 dependence of ( @xmath62 ) .
the solid line is guide to the eyes.,width=302 ] figures [ fig : five](a ) - ( c ) show the @xmath2 dependence of @xmath28 , @xmath27 , @xmath57 , @xmath58 , @xmath49 , @xmath60 , and @xmath62 for @xmath1 = 30 oe .
the magnetization @xmath57 appears below @xmath0 , exhibiting a peak at 3.5 k. note that a point at @xmath2 = 3.5 k and @xmath1 = 30 oe is located on the line @xmath13 in the @xmath1-@xmath2 diagram ( see sec .
[ disa ] ) .
the deviation of @xmath57 from @xmath58 occurs at @xmath2 below 3.6 k , while the deviation of @xmath27 from @xmath28 occurs below 3.55 k. the difference ( @xmath62 ) exhibits a very sharp peak at 3.5 k , being equal to zero above 3.6 k and below 3.4 k. vs @xmath1 ) for sn - mg . ( a ) @xmath2 = 1.9 k and ( b ) 5 k. the method of the measurement is described in the text .
the solid lines are guides to the eyes.,width=302 ] figure [ fig : six ] shows the hysteresis loop of the magnetization @xmath16 at @xmath84 and 5 k. the @xmath16-@xmath1 curve at 1.9 k shows a typical superconducting behavior with a negative local minimum around @xmath85 oe , corresponding to the phase boundary ( the line @xmath14 ) between the vortex lattice and vortex glass phase ( see sec .
[ disa ] ) .
the @xmath16-@xmath1 curve is not reversible , partly because some fluxoids remain trapped at pinning centers when @xmath1 is removed .
the @xmath16-@xmath1 curve at 5 k is not perfectly proportional to @xmath1 .
the susceptibility @xmath86 is estimated to be @xmath87 emu / g at @xmath88 koe , which is in contrast to the negative diamagnetic susceptibility of pristine graphite : @xmath89 emu / g for @xmath1 along the @xmath15 axis and @xmath90 emu / g along the basal plane ( so - called @xmath15 plane ) perpendicular to the @xmath15 axis below 100 k @xcite .
vs @xmath1 at various @xmath2 .
the solid lines are guides to the eyes .
the inset of fig .
[ fig : seven](b ) shows the jump in magnetization @xmath91 ( which is defined in the text ) as a function of @xmath2.,width=302 ] figures [ fig : seven](a ) and ( b ) show typical data of @xmath27 vs @xmath1 at various @xmath2 .
the sample was cooled from 298 k to @xmath2 ( @xmath92 k ) at @xmath93 .
the magnetization @xmath27 at @xmath2 was measured with increasing @xmath1 ( @xmath94 oe ) .
the magnetization @xmath27 exhibits a single local minimum at a characteristic field for @xmath95 .
similar behavior has been reported in yba@xmath9cu@xmath10o@xmath96 by shibata et al .
it seems that a discontinuous change in @xmath27 is observed just above the characteristic field for @xmath97 , where @xmath98 k and @xmath99 k. the field of the local - minimum for @xmath27 vs @xmath1 , which is plotted as a function of @xmath2 , forms the line @xmath13 of the first order transition in the @xmath1-@xmath2 plane ( see sec . [ disa ] ) .
the jump height @xmath91 , which is defined as the difference between the value of @xmath27 at which @xmath27 starts to be proportional to @xmath1 and the local - minimum value of @xmath27 vs @xmath1 , rapidly decreases with increasing @xmath2 and reduces to zero at @xmath0 .
the jump height @xmath91 may be described by @xmath100 @xcite , where @xmath101 and @xmath102 denote the magnetization in the vortex liquid and the vortex lattice phase , respectively .
the positive value of @xmath91 implies that the vortex density in the vortex liquid phase is higher than that in the vortex lattice phase .
the magnetization @xmath27 at @xmath103 and 3.8 k is almost proportional to @xmath1 with a positive slope .
the derivative d@xmath27/d@xmath1 at 3.7 and 3.8 k shows a broad peak around @xmath75 oe , suggesting the existence of af short - range order .
the origin of the af short - range order is due to nanographites in sn - mg ( see sec . [ disc ] ) .
note that a discontinuous jump of @xmath27 in the vicinity of @xmath104 is an artifact due to the squid measurement .
such a jump in the magnetization is frequently observed when the magnetization changes its sign .
the lower critical field @xmath105 is defined as a field at which the curve of @xmath16 vs @xmath1 starts to deviate from the linear portion due to the penetration of the magnetic flux into the system : @xmath106 k@xmath107 oe .
dependence of @xmath20 at various @xmath1 for sn - mg .
@xmath108 oe . @xmath38 hz . the solid lines are guides to the eyes.,width=302 ] figure [ fig : eight ] shows the @xmath2 dependence of the dispersion @xmath20 at various @xmath1 , respectively , where @xmath38 hz and @xmath108 oe . the ac response to such a small @xmath17 is found to be linear in the whole range of temperatures and fields .
the derivative d@xmath20/d@xmath2 shows a sharp peak at @xmath0 at @xmath1 = 0 , shifting to the low-@xmath2 side with increasing @xmath1 . the peak temperature for d@xmath20/d@xmath2 vs @xmath2 plotted as a function of @xmath1 in the @xmath1-@xmath2 diagram forms the line @xmath5 for @xmath109 and the line @xmath13 for @xmath52 ( see sec .
[ disa ] ) .
dependence of @xmath22 at various @xmath1 for sn - mg .
@xmath17 = 0.1 oe . @xmath38 hz .
the solid lines are guides to the eyes.,width=302 ] figure [ fig : nine ] shows the @xmath2 dependence of the absorption @xmath22 at various @xmath1 , where @xmath38 hz and @xmath110 oe . the absorption @xmath22 at @xmath29 seems to have two peaks at @xmath0 and 3.5 k. the peak at @xmath0 at @xmath1 = 0 shifts to the low-@xmath2 side with increasing @xmath1 . in contrast , the peak at 3.5 k at @xmath29 remains unshifted , becoming a shoulder even at 10 oe , and tends to disappear above @xmath111 oe .
note that the peak temperature of the main peak for @xmath22 vs @xmath2 is slightly higher than that for d@xmath20/d@xmath2 vs @xmath2 at the same @xmath1 .
the peak temperature for @xmath22 vs @xmath2 is plotted as a function of @xmath1 in the @xmath1-@xmath2 diagram , forming the line @xmath13 for @xmath52 and a new line @xmath6 for @xmath109 which deviates from the line @xmath5 for @xmath54 oe ( see sec . [ disa ] ) .
the main peak of @xmath22 vs @xmath2 is sharp for @xmath52 but becomes much broad for @xmath109 .
dependence of ( a ) the dispersion @xmath19 and ( b ) the absorption @xmath21 at various @xmath18 ( @xmath39 hz ) for sn - mg .
@xmath75 oe . because of the nonlinearlity , @xmath21 at @xmath112 oe is not equal to @xmath22 at @xmath110 oe ( see fig .
[ fig : twelve](b ) for the nonlinearity at @xmath29 ) .
the solid lines are guides to the eyes.,width=302 ] the ac response ( @xmath19 and @xmath21 ) for sufficiently large amplitude of the ac magnetic field becomes nonlinear below @xmath0 .
in fact , as it will be shown later , the @xmath2 dependence of @xmath21 is strongly dependent on @xmath17 for @xmath113 oe .
this is in contrast to the case of @xmath110 oe , where the linearity is valid : @xmath114 and @xmath115 . here
a large amplitude of the ac magnetic field ( @xmath112 oe ) was used to examine the effect of nonlinearity on the vortex glass phase .
figure [ fig : ten ] shows the @xmath2 dependence of @xmath19 and @xmath21 for sn - mg at various @xmath18 ( @xmath116 hz ) in the presence of @xmath1 (= 100 oe ) , where @xmath112 oe .
the @xmath2 dependence of @xmath19 is almost independent of @xmath18 , while the @xmath2 dependence of @xmath21 is strongly dependent on @xmath18 .
the shape of @xmath21 vs @xmath2 does not change with @xmath18 but the basement of @xmath21 proportionally increases with increasing @xmath18 for @xmath117 hz at any @xmath2 below @xmath0 . as shown in fig .
[ fig : ten](b ) there are two broad peaks at 2.90 and 3.15 k for @xmath21 with @xmath112 oe .
these peaks do not shift with @xmath18 for @xmath117 hz within the experimental error in @xmath2 ( typically @xmath118 k ) .
no relaxation time effect is observed , suggesting @xmath119 , where @xmath120 is the relaxation time of the vortex system and @xmath24 ( = @xmath121 ) is the measured angular frequency .
this is very different from the case of typical spin - glass systems where @xmath122 and the peak of @xmath21 shifts to the high @xmath2 side with increasing @xmath18 @xcite .
we show that the @xmath2 and @xmath18 dependence of @xmath19 and @xmath21 are explained in terms of the screening behavior arising from skin size effects .
the absorption peak occurs when the electromagnetic penetration depth @xmath123 becomes comparable to the sample size @xmath26 ( see sec .
[ disb ] ) .
the peak at 3.15 k coincides with that of @xmath22 at @xmath110 oe and @xmath38 hz ( see fig . [
fig : nine ] ) .
the point at @xmath75 oe and @xmath124 k is located on the line @xmath6 . on the other hand , the peak at 2.9 k is newly observed for @xmath112 oe .
the point at @xmath1 = 100 oe and @xmath125 k is located on the line @xmath5 .
note that the data of d@xmath20/d@xmath2 vs @xmath2 shows a peak at 2.94 k and d@xmath49/d@xmath2 has a local minimum at 3.0 k. in this sense , the line @xmath5 is the irreversibility line .
dependence of @xmath21 at various @xmath18 ( @xmath117 hz ) for sn - mg .
@xmath126 oe .
@xmath112 oe . the solid lines are guides to the eyes.,width=302 ] dependence of ( a ) @xmath19 and ( b ) @xmath21 at various @xmath2 for sn - mg . @xmath38
@xmath93.,width=302 ] figure [ fig : eleven ] shows the @xmath2 dependence of @xmath21 for sn - mg at various @xmath18 ( @xmath116 hz ) in the presence of @xmath1 (= 70 oe ) , where @xmath112 oe . there are two peaks at 3.1 - 3.15 k and 3.3 - 3.4 k.
it seems that these peaks do not shift with @xmath18 .
this result suggests no occurrence of the relaxation time effect for @xmath117 hz .
the peak at 3.3 - 3.4 k coincides with that at 3.34 k of @xmath22 at @xmath110 oe and @xmath38 hz ( the data at @xmath127 oe are not shown in fig . 9 ) .
the point at @xmath127 oe and @xmath128 - 3.4 k is located on the line @xmath6 . on the other hand , the peak at 3.1 - 3.15 k is newly observed for @xmath112 oe .
the point at @xmath127 oe and @xmath129 - 3.15 k is located on the line @xmath5 . figure [ fig : twelve ] shows the @xmath17 dependence of ( a ) @xmath19 and ( b ) @xmath21 for sn - mg at various @xmath2 in the absence of @xmath1 , where @xmath37 oe .
both @xmath19 and @xmath21 are not independent of @xmath17 for @xmath130 oe , indicating the nonlinearity of the ac response . the dispersion @xmath19 linearly changes with @xmath17 below @xmath0 , but is almost independent of @xmath17 at 3.8 k just above @xmath0 .
-@xmath2 diagram for sn - mg .
the peak temperatures of d@xmath20/d@xmath2 vs @xmath2 ( @xmath131 ) , @xmath22 vs @xmath2 ( @xmath132 ) , and the local - minimum temperatures of d@xmath49/d@xmath2 vs @xmath2 ( @xmath133 ) are plotted as a function of @xmath1 .
the local - minimum fields of @xmath27 vs @xmath1 ( @xmath134 ) are plotted as a function of @xmath2 . the vortex lattice ( abrikosov lattice ) ,
vortex liquid , and vortex glass phases are separated by the lines @xmath13 , @xmath14 , @xmath5 , and @xmath6 .
these lines merge at the multicritical point located at ( @xmath135 k , @xmath136 oe ) . the line @xmath48 is a crossover line denoted by closed squares ( @xmath137 ) . the lower critical line @xmath63 is not shown in this figure .
@xmath138k@xmath139 koe .
the solid lines for the lines @xmath5 , @xmath6 , @xmath14 , and @xmath13 denote the least - squares fitting curves of the data for @xmath140 vs @xmath2 to eq.([eq : one ] ) with @xmath141 and @xmath142 , respectively .
the corresponding parameters are given in the text.,width=302 ] our results obtained above are summarized in the @xmath1-@xmath2 diagram . in figs .
[ fig : thirteen](a ) and ( b ) , we make a plot of the peak temperatures of d@xmath20/d@xmath2 vs @xmath2 and @xmath22 vs @xmath2 , and the local - minimum temperatures of d@xmath49/d@xmath2 vs @xmath2 as a function of @xmath1 , and the local - minimum fields of @xmath27 vs @xmath1 as a function of @xmath2 .
we find that these data lie on the phase boundaries denoted by the lines @xmath14 , @xmath5 , @xmath6 , and @xmath13 , which separate the vortex lattice , vortex glass , and vortex liquid phase .
these lines merge at the multicritical point at @xmath98 k and @xmath143 oe .
the lines @xmath14 , @xmath5 , and @xmath6 are of the second order , while the line @xmath13 is of the first order .
the line @xmath48 is a crossover line separating the normal phase at higher temperatures from the vortex liquid phase .
( i ) on the line @xmath5 , the data of d@xmath20/d@xmath2 vs @xmath2 have a peak at fixed @xmath1 , the data of d@xmath49/d@xmath2 vs @xmath2 have a local minimum , and the data of @xmath22 vs @xmath2 have a peak for @xmath144 oe .
( ii ) on the line @xmath6 , the data of @xmath22 vs @xmath2 have a peak at fixed @xmath1 .
( iii ) on the line @xmath14 , the data of d@xmath49/d@xmath2 vs @xmath2 have a local minimum at fixed @xmath1 and the data of @xmath27 vs @xmath1 have a local minimum at fixed @xmath2 .
( iv ) on the line @xmath13 , the data of d@xmath20/d@xmath2 vs @xmath2 and @xmath22 vs @xmath2 have a peak at fixed @xmath1 , the data of d@xmath49/d@xmath2 vs @xmath2 have a local minimum at fixed @xmath1 , and the data of @xmath27 vs @xmath1 have a local minimum at fixed @xmath2 .
( v ) on the line @xmath48 , the data of @xmath41 vs @xmath2 come to take a saturated value @xmath145 at fixed @xmath1 .
the susceptibility @xmath145 is positive and weakly dependent on @xmath1 .
it takes @xmath146 emu / g at @xmath147 oe , increasing with increasing @xmath1 , and reaches a peak ( @xmath148 emu / g ) at @xmath55 oe . in turn
it decreases with further increasing @xmath1 , and reaches @xmath149 emu / g at @xmath88 koe ( see fig .
[ fig : three](a ) ) . using an empirical relation given by werhammer et al .
@xcite , @xmath150 k ) @xmath151d@xmath48/d@xmath152 , the value of @xmath48 at @xmath153 k can be estimated as @xmath154 oe , where we use @xmath99 k and a slope ( d@xmath48/d@xmath2)@xmath155 ( oe / k ) obtained from the linear relation in @xmath48 vs @xmath2 in the vicinity of @xmath156 and @xmath157 .
the coherence length @xmath158 can be estimated as @xmath159 using a relation @xmath160 , where @xmath161 gauss @xmath162 ) is a fluxoid .
note that our value of @xmath48(0 ) for sn - mg is on the same order as that of sn films grown on an insb ( 110 ) surface @xcite .
the value of @xmath163 strongly depends on the thickness @xmath164 of the sn films : @xmath165 oe for @xmath166 and 2540 oe for @xmath167 for @xmath1 parallel to the sn film . here
we assume that the @xmath2 dependence of the line @xmath140 ( @xmath141 and @xmath142 ) may be described by a power law form given by @xmath168 where @xmath169 is an exponent , and @xmath170 and @xmath171 are characteristic field and temperature , respectively .
the least squares fit of the data of @xmath1 vs @xmath2 ( @xmath172 oe ) obtained from the peak temperature of d@xmath20/d@xmath2 vs @xmath2 at @xmath1 , to eq.([eq : one ] ) with @xmath173 yields the parameters @xmath174 , @xmath175 k , and @xmath176 oe . the exponent @xmath177 is close to that (= 1.50 ) predicted from the at theory @xcite , suggesting that the line @xmath5 is the so - called irreversibility line separating a higher - temperature vortex liquid phase from a lower - temperature vortex glass phase .
the least squares fit of the data of @xmath1 vs @xmath2 ( @xmath178 oe ) obtained from the peak temperature of @xmath22 vs @xmath2 at @xmath1 , to eq.([eq : one ] ) with @xmath179 yields the parameters @xmath180 , @xmath181 k , and @xmath182 oe . the exponent @xmath183 is close to that (= 0.50 ) predicted from the gabay - thoules ( gt ) theory @xcite .
such an occurrence of the at - gt crossover has been reported in granular high @xmath0 superconductors @xcite .
no gt - power law form has been observed in a single crystal of yba@xmath9cu@xmath10o@xmath11 without granularity @xcite .
these results suggest that the at - gt crossover behavior is indicative of the frustrated nature of the system due to disorder and granularity .
this crossover is an analogy to the ising - heisenberg crossover behavior observed in heisenberg - like spin glasses @xcite .
such a crossover occurs when the magnetic field collapses the random local anisotropy field .
the at and gt lines in sg systems represent the longitudinal and transverse freezing , respectively .
similarly , the least - squares fit of the data of @xmath1 vs @xmath2 ( @xmath184 oe ) obtained from the peak temperature of d@xmath20/d@xmath2 vs @xmath2 at @xmath1 , to eq.([eq : one ] ) with @xmath185 yields the parameters @xmath186 , @xmath187 k , and @xmath188 oe .
note that the lower critical field @xmath189 is much lower than @xmath190 .
the least - squares fit of the data of @xmath1 vs @xmath2 ( @xmath191 oe ) obtained from the peak temperature of d@xmath20/d@xmath2 vs @xmath2 at @xmath1 , to eq.([eq : one ] ) with @xmath192 yields the parameters @xmath193 , @xmath194 k , and @xmath195 oe .
the value of @xmath196 is in good agreement with the mean - field value (= 1 ) derived from the ginzburg - landau theory @xcite , where the correlation length varies with @xmath197 .
the @xmath1-@xmath2 diagram of sn - mg is similar to that of a quasi-2d superconductors ( type - ii ) .
the field for the multicritical point ( @xmath198 ) is described by the dimensional crossover field that separates the region of 2d thermal fluctuations ( for @xmath109 ) and the 3d thermal fluctuations ( @xmath52 ) . in this sense
, the @xmath1-@xmath2 diagram may be universal for any quasi-2d superconductors when the @xmath1 axis and @xmath2 axis are scaled by @xmath199 and @xmath200 , respectively .
for yba@xmath9cu@xmath10o@xmath11 with @xmath201 k , the multicritical point is located at @xmath202 k and @xmath203 koe @xcite .
the ratio @xmath204 for yba@xmath9cu@xmath10o@xmath11 is 0.84 , which is almost equal to that (= 0.91 ) for sn - mg with @xmath98 k and @xmath99
k. here we discuss the origin of the irreversibility lines @xmath6 and @xmath5 for @xmath205 oe .
we find the following results on the @xmath2 dependence of ac magnetic susceptibility at small ac field ( @xmath108 oe ) and large ac field ( @xmath112 oe ) .
( i ) the linear absorption @xmath22 at @xmath110 oe exhibits a peak only on the line @xmath6 .
( ii ) the nonlinear absorption @xmath21 exhibits a peak on the line @xmath6 and a peak on the line @xmath5 .
it is assumed that there is an irreversibility line @xmath206 ( or the line @xmath207 ) in the @xmath1-@xmath2 plane which separates an ohmic region [ @xmath208 from an non - ohmic region [ @xmath209 . above @xmath206 ,
a voltage @xmath210 is proportional to current @xmath211 , whereas below @xmath206 , @xmath210 shows an extremely nonlinear dependence where the linear resistivity is zero and the electric field @xmath212 is described by @xmath213 @xcite .
here @xmath214 and @xmath215 are positive constants and @xmath216 is the current density . why does the peak of the absorption for small ac field amplitude appear on only the line @xmath6 for @xmath109 ? why does the peak of the absorption for large ac field amplitude appear on both the lines @xmath5 and @xmath6 for @xmath109 ?
similar problem was addressed first by geshkenbein et al .
@xcite and later by steel and graybeal @xcite .
according to their skin size effect hypothesis , the peak of @xmath21 appears when the electromagnetic penetration depth @xmath217 is on the order of the system size @xmath26 , or when @xmath24 coincides with the characteristic angular frequency @xmath218 given by @xmath219 where @xmath220 is the electrical resistivity .
then @xmath21 is a function of only @xmath221 , and has a peak at @xmath222 .
when @xmath223 becomes infinity , the field penetrates the system completely and @xmath21 tends to zero . in the opposite limiting case ( @xmath224 ) ,
the screening is complete and @xmath225 .
first we consider the peak temperature @xmath226 of @xmath21 at any @xmath17 , where @xmath227 .
we assume that the response of the system is linear ( ohmic ) .
the current density is of the order of @xmath228
. the resistivity @xmath229 , which is independent of @xmath230 , decreases with decreasing @xmath2 and tends to zero at @xmath206 .
the peak temperature @xmath226 can be determined from the matching condition ( @xmath222 ) given by eq.([eq : two ] ) : @xmath231 for @xmath232 hz and at @xmath233 for @xmath234 khz .
then we have a relation @xmath235 .
since there is no shift of @xmath226 in @xmath21 with @xmath18 ( @xmath236 hz ) in the present work , it follows that the difference of peak temperatures defined by @xmath237 should be less than 0.01 k. because of the sharp drop in resistivity with decreasing @xmath2 , @xmath226 thus obtained is very close to but above @xmath206 .
the value of @xmath226 is independent of @xmath17 because eq.([eq : two ] ) is independent of @xmath17 .
the resistivity is finite at @xmath238 : @xmath239 . in summary ,
@xmath21 vs @xmath2 has a peak at @xmath226 just above @xmath206 , which is independent of @xmath17 .
next we consider the peak temperature @xmath226 of @xmath21 at any @xmath17 , where @xmath240 .
we assume that the response of the system is highly nonlinear ( non - ohmic )
. the effective resistivity is defined by @xmath241 , where @xmath242 . as a function of @xmath2
, @xmath243 decreases with decreasing @xmath2 and tends to zero at a temperature below @xmath206 .
as a function of @xmath230 , @xmath243 has a local minimum at @xmath244 , and is equal to zero at both @xmath245 and @xmath246 .
when @xmath230 is smaller than @xmath247 [ or @xmath248 , @xmath249 decreases with decreasing @xmath230 ( or with decreasing @xmath17 ) .
the peak temperature @xmath226 can be estimated from eq.([eq : two ] ) with @xmath250 : @xmath251 at @xmath238 .
the value of @xmath230 ( or @xmath17 ) is uniquely determined because of the relation between @xmath230 and @xmath17 .
the peak temperature @xmath226 , which is lower than @xmath206 , increases with decreasing @xmath17 and approaches the line @xmath206 from the low-@xmath2 side . in summary , @xmath21 vs @xmath2 has a peak at @xmath226 just below @xmath206 for sufficiently large
@xmath17 satisfying the above condition . the lines @xmath5 and @xmath6 intersect at a critical point located at @xmath252 k and @xmath253 oe .
the line @xmath5 line for @xmath254 no longer separates the irreversible region of the vortex glass phase from the reversible region of the vortex liquid phase . in this sense
, the vortex glass phase may connect smoothly to the vortex liquid phase at the critical point at @xmath7 and @xmath8 @xcite .
as shown in fig .
[ fig : three ] , in fact the @xmath2 dependence of @xmath49 at @xmath255 oe is rather different from that at 150 oe .
the difference @xmath49 at @xmath256 oe is almost equal to zero at any @xmath2 , indicating the reversibility of magnetization on the line @xmath207 .
the line @xmath5(@xmath2 ) may correspond to the at line of spin glasses @xcite .
then it is expected that the relaxation time of sn - mg , @xmath120 , diverges as a power law form @xmath257^{-x}$ ] @xcite , as @xmath2 approaches the irreversibility line @xmath206 from the high-@xmath2 side . here
@xmath258 is a characteristic relaxation time and @xmath259 is a dynamic critical exponent .
experimentally , no relaxation time effect is observed in the @xmath2 and @xmath18 dependence of @xmath19 and @xmath21 on the line @xmath5(@xmath2 ) ( see sec .
[ resulte ] ) .
this result indicates that the condition @xmath119 is satisfied , in spite of crossing the line @xmath207 .
thus the power law form is not the case of the present system .
what is the origin of @xmath120 satisfying the condition @xmath260 ?
the absorption of the ac field is mainly governed by the hopping of the vortices over the barriers between different metastable states .
the depinning time @xmath120 for such intervalley transition is obtained from a characteristic relaxation time @xmath258 by an arrhenius form @xmath261 , where @xmath262 is the activation barrier energy between the different metastable states @xcite .
then @xmath120 smoothly changes as @xmath2 changes and @xmath119 on the line @xmath207 . in summary ,
it is concluded from the experimental results described above that the properties of the vortex glass phase in sn - mg are characterized as follows , although most of them are not sufficiently understood at present .
( i ) the magnetization @xmath27 deviates from @xmath28 below @xmath206 for @xmath109 .
( ii ) the magnetization @xmath57 deviates from @xmath58 below @xmath206 .
the @xmath2 dependence of @xmath57 and @xmath58 is very similar to that of @xmath27 and @xmath28 in typical spin glass systems , respectively @xcite .
( iii ) the difference ( @xmath62 ) takes a positive large value for the vortex glass phase , while it is equal to zero for the vortex liquid phase and it takes a negative value in the vortex lattice phase .
( iv ) the relaxation time @xmath120 shows no behavior of divergence on the line @xmath206 . ( v ) the non - ohmic behavior is observed in the vortex glass phase , while the ohmic behavior is observed in the vortex liquid phase .
there are several results supporting the existence of the af short - range order .
the first evidence is the @xmath2 dependence of the peak field @xmath47 as shown in the inset of fig .
the field @xmath47 is much higher than @xmath48 at the same @xmath2 : @xmath263 koe and @xmath264 oe at 3.0 k. the field @xmath47 increases with decreasing @xmath2 and may be described by a power law form given by @xmath265 , where @xmath266 is an exponent , and @xmath267 and @xmath268 are characteristic field and temperature , respectively .
the least squares fit of the data of @xmath47 vs @xmath2 ( @xmath269 koe ) to the power law form yields the parameters @xmath270 , @xmath271 k , and @xmath272 koe .
the exponent @xmath266 is close to @xmath273 for the molecular field theory .
the temperature @xmath268 is a little higher than @xmath0 .
the second evidence is the magnitude of the dc magnetic susceptibility : @xmath274 emu / g at @xmath88 koe and @xmath275 k. this value of @xmath276 is positive and is very different from that of pristine graphite which is diamagnetic ( see sec . [ resultc ] ) @xcite . what is the origin of the af short - range order ?
the existence of the af short - range order has been observed in other mg s such as bi - mg @xcite . in these mg s
the metal layers would generate internal stress inside the graphite lattice , leading to the break up of adjacent graphene sheets into nanographites .
fujita et al .
@xcite and wakabayashi et al .
@xcite have theoretically suggested that the electronic structures of finite - size graphene sheets depend crucially on the shape of their edges ( zigzag type and armchair type edges ) .
finite graphite systems having zigzag edges exhibit a special edge state .
the corresponding energy bands are almost flat at the fermi energy , thereby giving a sharp peak in the density of states ( dos ) at the fermi energy . as a result
, conduction electrons are localized around the zig - zag edges of nanographites , forming localized magnetic moments .
recently harigaya @xcite has theoretically predicted that the magnetism in nanographites with zigzag edge sites depends on the stacking sequence of nanographites .
the antiferromagntic spin alignment becomes energetically favorable when the adjacent graphene sheets have a a - b type stacking sequence along the @xmath15 axis , like pristine graphite @xcite .
the af short - range order in nanographites may coexist with the superconductivity occurring in sn sheets even at low @xmath2 and low @xmath1 .
sn - mg is a quasi-2d superconductor , showing a superconducting transition at @xmath99 k for @xmath29 .
the @xmath1-@xmath2 diagram of sn - mg consists of four lines ( @xmath14 , @xmath5 , @xmath6 , and @xmath13 ) separating vortex lattice phase , vortex glass phase , and vortex liquid phase .
these lines merge into the multicritical point located at @xmath98 k and @xmath277 oe .
the irreversibility lines @xmath5 and @xmath6 intersect at a critical point at @xmath7 = 2.5 k and @xmath253 oe .
the line @xmath5 has an at - power law form , while the line @xmath6 has a gt - power law form .
the slow dynamics and nonlinearity in the vortex glass phase is an analogy to that in the spin glass phase observed in 3d ising spin glasses .
further studies will be required to understanding the nature of the vortex glass phase from measurements of resistivity ( the nonlinear relation of current - voltage ) and the absorption and dispersion at frequencies ( @xmath278 khz ) . the work at binghamton ( m.s . and i.s.s . )
was supported by the research foundation of suny - binghamton ( contract number 240 - 9522a ) .
the work at osaka ( j.w . ) was supported by the ministry of education , science , sports and culture , japan [ the grant for young scientists ( no . 70314375 ) ] and by kansai invention center , kyoto , japan .
00 g. blatter , m.v .
feigelman , v.b .
geshkenbein , a.i .
larkin , v.m .
vinokur , rev .
66 ( 1994 ) 1125 .
see also references therein .
gammel , d.a .
huse , d.j .
bishop , in : a.p .
young(ed ) , spin glasses and random fields , world scientific publishing , singapore , 1998 , p.299 . see also references therein .
g. menon , phys .
b 65 ( 2002 ) 104527 . k.a .
mller , m. takashige , j.g .
bednorz , phys .
58 ( 1987 ) 1143 .
de almeida , d.j .
thouless , j. phys . a 11 ( 1978 ) 983 .
e. maxwell , phys .
86 ( 1952 ) 235 .
m. gabay , g. toulouse , phys .
47 ( 1981 ) 201 .
j. walter , h. shioyama , phys .
lett . a 254 ( 1999 ) 65 .
j. walter , adv .
12 , ( 2000 ) 31 .
j. walter , phil . mag . lett .
80 ( 2000 ) 257 .
j. walter , j. heiermann , g. dyker , s. hara , h. shioyama , j. catalysis 189 ( 2000 ) 449 .
m. suzuki , i.s .
suzuki , t.y .
huang , j. phys .
: condensed matter 14 ( 2002 ) 5583 .
k. gunnarsson , p. svedlindh , p. nordblad , l. lundgren , h. aruga , a. ito , phys .
61 ( 1988 ) 754 .
j. heremans , c.h .
olk , d.t .
morelli , phys .
b 49 ( 1994 ) 15122
. k. shibata , t. nishizuka , t. sasaki , n. kobayashi , phys .
b 66 ( 2002 ) 214518 .
see also references therein .
u. welp , j.a .
fendrich , w.k .
kwok , g.w .
crabtree , b.w .
veal , phys .
76 ( 1996 ) 4809 .
mydosh , spin glasses : an experimental introduction , taylor & francis , london , 1993 .
werthamer , e. helfand , p.c .
hohenberg , phys .
147 ( 1966 ) 2953 .
i. didschuns , k. fleischer , p. schilbe , n. esser , w. richter , k. luders , physica c 377 ( 2002 ) 89 .
vieira , j. schaf , physica c 384 ( 2003 ) 514 .
vieira , j.p .
da silva , j. schaf , phys .
rev.b 64 ( 2001 ) 094516 .
kenning , d. chu , r. orbach , phys .
66 ( 1991 ) 2923 .
see , for example , j.b .
ketterson , s.n .
song , superconductivity , cabmridge university press , cabmridge , 1998 .
geshkenbein , v.m .
vinokur , r. fehrenbacher , phys .
b 43 ( 1991 ) 3748 . d.g .
steel , j.m .
graybeal , phys .
b 45 ( 1992 ) 12643 .
m. suzuki , i.s .
suzuki , r. lee , j. walter , phys .
b 66 ( 2002 ) 014533 .
m. fujita , k. wakabayashi , k. nakada , k. kusakabe , j. phys .
( 1996 ) 1920 .
k. wakabayashi , m. fujita , h. ajiki , m. sigrist , phys .
b 59 ( 1999 ) 8271 .
k. harigaya , j. phys .
condensed matter 13 ( 2001 ) 1295 .
t. enoki , m. suzuki , m. endo , graphite intercalation compounds and applications , oxford university press , new york , 2003 .
|
the magnetic properties of a quasi two - dimensional superconductor , sn - metal graphite ( mg ) , are studied using dc and ac magnetic susceptibility .
sn - mg has a unique layered structure where sn metal layer is sandwiched between adjacent graphene sheets .
this compound undergoes a superconducting transition at @xmath0 = 3.75 k at @xmath1 = 0 .
the @xmath1-@xmath2 diagram of sn - mg is similar to that of a quasi two - dimensional superconductors .
the phase boundaries of vortex liquid , vortex glass , and vortex lattice phase merge into a multicritical point located at @xmath3 = 3.4 k and @xmath4 = 40 oe . there are two irreversibility lines denoted by @xmath5 ( de almeida - thouless type ) and @xmath6 ( gabay - toulouse type ) , intersecting at @xmath7= 2.5 k and @xmath8 = 160 oe .
the nature of slow dynamic and nonlinearity of the vortex glass phase is studied .
vortex glass phase , phase diagram , magnetic irreversibility 74.25.ha , 74.25.dw , 74.60.ec , 74.80.dm
|
single particle interference is one of the most genuine quantum mechanical effects . since the original double - slit experiment @xcite
, it has been observed with electrons in vacuum @xcite and even with the more massive @xmath0 molecules @xcite .
mesoscopic rings threaded by a magnetic flux provided the solid - state analogous @xcite .
intra - molecular interference has been recently discussed in molecular junctions for the case of strong @xcite and weak @xcite molecule - lead coupling . what unifies these realizations of quantum interference is that the travelling particle has two ( or more ) _ spatially equivalent _ paths at disposal to go from one point to another of the interferometer .
interference , though is hindered by decoherence .
generally , for junctions in the strong coupling regime decoherence can be neglected due to the short time of flight of the particle within the interferometer . in the weak coupling case , instead
, the dwelling time is long .
usually , the decoherence introduced by the leads dominates , in this regime , the picture and the dynamics essentially consists of sequential tunnelling events connecting the many - body eigenstates of the isolated system . yet
, interference is achieved whenever two _ energetically equivalent _ paths involving degenerate states contribute to the dynamics ( see fig .
[ fig : paths ] ) .
interference survives as far as the splitting between the many body levels is smaller that the tunnelling rate to the leads since in this limit the system can not distinguish between the two paths .
thus , in such devices , that we called interference single electron transistors@xcite ( iset ) , interference effects show up even in the coulomb blockade regime .
they can _ e.g. _ yield a selective spin blockade in an iset coupled to ferromagnetic leads@xcite .
similar blocking effects have been found also in multiple quantum dot systems in dc @xcite and ac @xcite magnetic fields . in the present paper
we develop a general theory of interference blockade .
we give in fact an _ a priori _ algorithm for the detection of the interference blocking states of a generic iset . as a concrete example
, we analyze the triple dot iset ( see section [ sec : triple_dot ] ) as it represents the simplest structure exhibiting interference blockade .
in particular we concentrate on the blockade that involves an _ excited _ triplet state and we show how to prepare the system in each of the three spin states without application of any external magnetic field . thus we obtain an interference mediated control of the electron spin in quantum dots , a highly desirable property for spintronics @xcite and spin - qubit applications @xcite .
the method of choice for the study of the dynamics in those systems is the generalized master equation approach for the reduced density matrix ( rdm ) , where coherences between degenerate states are retained @xcite .
such coherences give rise to precession effects and ultimately cause interference blockade .
the paper is organized as follows : in section [ sec : model ] we introduce a generic model of iset . in section [ sec : blocking_states ] we set the necessary and sufficient conditions which define the interference blocking states and a generic algorithm to detect them . in section [ sec : triple_dot ] we apply the theory to the triple dot molecule as archetypal example of iset .
section [ sec : conclusions ] closes the paper with a summary of the results and conclusive remarks .
let us consider the interference single electron transistor ( iset ) described by the hamiltonian : @xmath1 where @xmath2 represents the central system and also contains the energy shift operated by a capacitively coupled gate electrode at the potential @xmath3 .
the hamiltonian @xmath4 is invariant with respect to a set of point symmetry operations that defines the symmetry group of the device .
this fact ensures the existence of _ degenerate _ states . in particular , for essentially planar structures belonging to the @xmath5 group ,
the ( non - accidental ) orbital degeneracy is at maximum twofold and can be resolved using the eigenvalues @xmath6 of the projection of the angular momentum along the principal axis of rotation .
a generic eigenstate is then represented by the ket @xmath7 where @xmath8 is the number of electrons on the system , @xmath9 is the spin and @xmath10 the energy of the state .
the size of the fock space can make the exact diagonalization of @xmath2 a numerical challenge in its own .
we will not treat here this problem and concentrate instead on the transport characteristics .
@xmath11 describes two reservoirs of non - interacting electrons with a difference @xmath12 between their electrochemical potentials .
finally , @xmath13 accounts for the weak tunnelling coupling between the system and the leads , characteristic of sets , and we consider the tunnelling events restricted to the atoms or to the dots closest to the corresponding lead : @xmath14 where @xmath15 creates an electron with spin @xmath9 and momentum @xmath16 in lead @xmath17 , @xmath18 creates an electron in the atom or dot closest to the lead @xmath19 and @xmath20 is the bare tunnelling amplitude that we assume for simplicity independent of @xmath19 , @xmath16 and @xmath9 . in the weak coupling regime the dynamics essentially consists of sequential tunnelling events at the source and drain lead that induce a flow of probability between the many - body eigenstates of the system .
it is natural to define , in this picture , a blocking state as a state which the system can enter but from which it can not escape .
when the system occupies a blocking state the particle number can not change in time and the current vanishes .
if degenerate states participate to transport , they can lead to interference since , like the two arms of an electronic interferometer , they are populated simultaneously .
in particular , depending on the external parameters they can form linear superpositions which behave as blocking states . if a blocking state is the linear combination of degenerate states we call it _ interference blocking state_. the coupling between the system and the leads not only generates the tunneling dynamics described so far , but also contributes to an internal dynamics of the system that leaves unchanged its particle number .
in fact the equation of motion for the reduced density matrix @xmath21 of the system can be cast , to lowest non vanishing order in the coupling to the leads , in the form @xcite : @xmath22 - \frac{i}{\hbar}[h_{\rm eff},\rho ] + \mathcal{l}_{\rm tun}\rho .
\label{eq : gme - small}\ ] ] the commutator with @xmath2 in eq
. represents the coherent evolution of the system in absence of the leads .
the operator @xmath23 describes instead the sequential tunnelling processes and is defined in terms of the transition amplitudes between the different many - body states .
finally , @xmath24 renormalizes the coherent dynamics associated to the system hamiltonian and is also proportional to the system - lead tunnelling coupling .
the specific form of @xmath25 depends on the details of the system , yet in all cases it is bias and gate voltage dependent and it vanishes for non degenerate states .
for the description of the tunnelling dynamics contained in the superoperator @xmath23 it is convenient to classify all possible tunnelling events according to four categories : i)creation ( annihilation ) tunnelling events that increase ( decrease ) by one the number of electrons in the system , ii ) source ( drain ) tunnelling that involves the lead with the higher ( lower ) chemical potential , iii ) @xmath26 ( @xmath27 ) tunnelling that involves an electron with spin up ( down ) with respect of the corresponding lead quantization axis , iv ) gain ( loss ) tunnelling that increases ( decreases ) the energy in the system .
using categories i)-iii ) we can efficiently organize the matrix elements of the system component of @xmath13 in the matrices : @xmath28 where @xmath29 means source and drain respectively and @xmath30 is a matrix in itself , defined for every creation transition from a state with particle number @xmath8 and energy @xmath10 to one with @xmath31 particles and energy @xmath32 .
we indicate correspondingly in the following transitions involving @xmath33 and @xmath34 as source - creation and drain - creation transitions .
the compact notation @xmath35 indicates all possible combination of the quantum numbers @xmath6 and @xmath36 .
it follows that the size of @xmath37 is @xmath38 where the function @xmath39 gives the degeneracy of the many - body energy level with @xmath8 particles and energy @xmath10 .
analogously @xmath40 accounts for the annihilation transitions .
the fourth category concerns energy and it is intimately related to the first and the second .
not all transitions are in fact allowed : due to the energy conservation and the pauli exclusion principle holding in the fermionic leads , the energy gain ( loss ) of the system associated to a gain ( loss ) transition is governed by the bias voltage .
these energy conditions are summarized in the table [ tab : en_condition ] and illustrated in fig .
[ fig : energy_cond ] . .energy conditions for tunnelling transitions between the many - body eigenstates of the system .
the quantity @xmath41 is the difference between the energies of the final and initial many - body states of the system involved in the transition . [
cols="^,^,^",options="header " , ] chosen to favor a configuration with two electrons .
the other parameters in the system are @xmath42 and @xmath43 , where @xmath44 is the hopping integral between the different dots . ] in fig .
[ fig : current ] we show the stationary current through the triple dot iset as a function of bias and gate voltage . at low bias
the current vanishes almost everywhere due to coulomb blockade .
the particle number is fixed within each coulomb diamond by the gate voltage and the zero particle diamond is the first to the right .
the zero current lines running parallel to the borders of the 6 , 4 and 2 particle diamonds are instead signatures of ground state interference that involves an orbitally non - degenerate ground state ( with 2 , 4 , and 6 particle ) and an orbitally double - degenerate one ( with 3 and 5 particles ) . in appendix
[ app : equivalence ] we illustrate how to obtain an expression for the blocking states in these cases .
the striking feature in fig . [ fig : current ] is the black area of current blocking sticking out of the right side of the two particles coulomb diamond .
it is the fingerprint of the occupation of an excited interference blocking state .
[ fig : stab_diagr ] is a zoom of the current plot in the vicinity of this excited state blocking .
the dashed lines indicate at which bias and gate voltage a specific transition is energetically allowed , with the notation @xmath45 labelling the @xmath46th excited many - body level with @xmath8 particles .
these lines are physically recognizable as abrupt changes in the current and run all parallel to two fundamental directions determined by the ground state transitions . for positive bias ,
positive ( negative ) slope lines indicates the bias threshold for the opening of source - creation ( drain - annihilation ) transitions .
the higher the bias the more transitions are open , the higher , in general , the current . .
the other parameters are the same as the ones in fig .
[ fig : spectrum ] . ]
the anomalous blockade region is delimited on three sides by transitions lines associated to the first excited two particle level @xmath47 .
our group theoretical analysis shows that the two particle first excited state is a twofold orbitally degenerate spin triplet ( see table [ tab : degeneracies ] ) .
in other terms we can classify its six states with the notation @xmath48 with @xmath49 being the projection of the angular momentum along the main rotation axis , perpendicular to the plane of the triple dot , and @xmath50 the component of the spin along a generic quantization axis .
the @xmath51 energy level is instead twice spin degenerate and invariant under the symmetry operations of the point group @xmath52 . in order to identify the 2 particle blocking states we perform
the analysis presented in the previous section for the @xmath47 energy level with the gate and bias in the blocking region .
firstly , we find that the @xmath47 energy level can be reached from @xmath53 via the drain - annihilation transition @xmath54 followed by the source - creation transition @xmath55 .
secondly , the space of the decoupled states @xmath56 is empty and the only energetically allowed outgoing transition is the drain - annihilation @xmath57 transition .
thus the blocking space is given by the expression : @xmath58\ ] ] and has dimension three .
for clearness we give in the appendix [ app : t_matrices ] the explicit expression of @xmath59 and the corresponding vectors that span @xmath60 . essentially , there is a blocking state for each of the three projection of the spin @xmath61 .
this result is natural since , for unpolarized or parallel polarized leads , coherences between states of different spin projection along the common lead quantization axis do not survive in the stationary limit . outside the blocking region
either the first or the second blocking state conditions are violated . in particular , below the lower right border the state @xmath47 can not be reached from the global minimum since the @xmath55 source - creation transition is forbidden while above the upper left ( right ) borders the state @xmath47 can be depopulated towards the @xmath62 ( @xmath51 ) states via a source - creation ( drain - annihilation ) transition .
the orbital interference blocking presented in the previous section acquires a spin dependence in presence of polarized leads .
the lead polarization @xmath63 with @xmath17 is defined by means of the density of states @xmath64 at the fermi energy for the different spin states : @xmath65 and is taken equal for the two leads @xmath66 . finally , the spin polarization influences the dynamics of the system via the spin dependent bare tunnelling rates @xmath67 that enter the definition of the tunnelling component of the liouvillian @xmath23 and the renormalization frequencies @xmath68 .
we assume the leads to be parallel polarized so that no spin torque is active in the device and we can exclude the spin accumulation associated to that@xcite . in fig .
[ fig : i_vs_pvb ] we show the current in the excited state blocking region as a function of the bias and of the ( parallel ) lead polarization @xmath69 . for
non - polarized leads the current is blocked at a single bias , while for finite values of @xmath69 the blocking is threefold .
for the same bias and polarization ranges we present in fig .
[ fig : sz_vs_pvb ] the @xmath70 component of the spin for the triple dot
. the spin projection @xmath61 assumes , exactly in correspondence of the current blocking , the values @xmath71 , respectively , as the bias is increased . for @xmath50 from left to right , respectively . ] as a function of the polarization and of the bias in proximity of the excited state interference blocking .
notice by comparison with fig .
[ fig : i_vs_pvb ] the correspondence between current blocking and spin preparation . ]
the explanation of this effect relies on the third blocking condition , eq . , and concerns the form of the effective hamiltonian introduced in eq . .
the latter can be written , due to the rotational symmetry of the system , in the form : @xmath72 where @xmath73 is the projection of the angular momentum in the direction of the lead @xmath19 and it does not depend on the spin degree of freedom @xmath61 .
moreover , @xmath68 is the frequency renormalization given to the states of spin projection @xmath61 by their coupling to the @xmath19 lead . in the appendix [ app : effective hamiltonian ] we give an explicit expression for @xmath68 and @xmath73 . in fig .
[ fig : wls ] we plot instead @xmath74 as a function of the bias for different polarizations .
the gate is fixed at @xmath75 . in the leads .
the gate is @xmath76 . at @xmath77
all the renormalization frequencies coincide ( full line ) .
the @xmath78 frequency increases ( decreases ) monotonously with the polarization .
the @xmath79 frequency is instead independent of it . ]
since the two particle ground state is totally symmetric ( @xmath80 symmetry ) , a three particle blocking state must be antisymmetric with respect to the vertical plane that intersects the center of the system and the drain dot .
for this reason a blocking state is also an eigenstate of the projection @xmath81 of the angular momentum in the direction of the drain lead .
consequently , the last blocking condition is satisfied only if : @xmath82 and the effective hamiltonian is proportional to @xmath81 . for zero polarization in the leads the condition holds at the same bias for the three spin projections @xmath61 and
the blocking state is a statistical mixture of the three spin projections . for polarized leads , instead
, each spin projection is blocked at a specific bias and the spin on the system is controlled simply by changing the bias across the device .
the dashed lines in figs . [ fig : i_vs_pvb ] and [ fig : sz_vs_pvb ] represent the solutions of equations for @xmath83 from left to right , respectively .
clearly they also indicate in fig .
[ fig : i_vs_pvb ] the zeros of the current and in fig .
[ fig : sz_vs_pvb ] the fully populated spin states .
in this paper we addressed the interference effects that characterize the transport through a symmetric single electron transistor .
in particular we gave the generic conditions for interference blockade and an algorithm for the identification of the interference blocking states as linear combination of degenerate many - body eigenstates of the system .
as an application of the theory we studied the triple dot iset . despite its relative simplicity ,
this system exhibits different types of interference blocking and it represents an interesting playground of the general theory . specifically , we concentrated on the interference blockade that involves an excited triplet state . in presence of polarized leads
we exploited the interference blocking in order to access each of the triplet states by all electrical means .
the theory is sufficiently general to be applied to any device consisting of a system with degenerate many - body spectrum weakly coupled to metallic leads _
e.g. _ molecular junctions , graphene or carbon nanotube quantum dots , artificial molecules .
in particular , the algebraic formulation of the blocking condition in terms of kernels of the tunnelling matrices @xmath84 , eq . , allows a straightforward numerical implementation and makes the algorithm directly applicable to complex junctions with highly degenerate spectrum .
we acknowledge financial support by the dfg under the programs sfb689 , spp1243 . we derive here the equation ( 1 ) in [ ] as a specific example of the general theory presented in the paper .
that equation represents the interference blocking condition for the simplest possible configuration involving only a non degenerate and a doubly degenerate state .
let us consider for simplicity a spinless @xcite system and a gate and bias condition that restricts the set of relevant many - body states to three : one with @xmath8 particles and two ( degenerate ) with @xmath31 or @xmath85 particles .
the interference blocking state , if it exists , belongs to the @xmath86 level .
there is only one interesting tunnelling matrix to be analyzed , namely @xmath87 .
let us take for it the generic form : @xmath88 where @xmath89 and @xmath90 indicate source and drain respectively and 1 and 2 label the two degenerate states with @xmath86 particles .
@xmath91 are the elements of the @xmath92 matrices introduced in eqs . and .
this condition is identical to eq .
( 1 ) in @xcite .
the blocking state can finally be calculated as : @xmath96 where the @xmath97 is , in the relevant case , the entire space and the projector @xmath98 simply removes the last component of the vector that defines the one dimensional kernel .
we give here explicitly the @xmath59 matrix necessary for the calculation of the triplet blocking states and the associated blocking states .
the states in the @xmath51 doublet and in the two times orbitally degenerate triplet @xmath47 are labelled and ordered as follows : @xmath99 the elements of the @xmath100 matrices that compose @xmath59 have thus the general form : @xmath101 by orbital and spin symmetry arguments it is possible to show that @xmath102 where @xmath103 the subscript @xmath104 labels a reference dot and @xmath105 is the angle of the rotation that brings the dot @xmath19 on the dot @xmath104 .
the explicit form of @xmath106 reads : the rank of this matrix is 6 since all columns are independent .
thus @xmath108 coincides with the full hilbert space of the first excited 2 electron energy level .
the blocking space @xmath109 reads : @xmath110 where @xmath111 reads in accordance to its general definition given in eq . , and the projector @xmath113 removes the last four components from the vectors that span @xmath114 .
it is then straightforward to calculate the vectors that span the blocking space @xmath109 : the vectors @xmath116 and @xmath117 are the components of the blocking states written in the @xmath47 basis set presented in .
thus , the three blocking states correspond to the three different projectors of the total spin @xmath118 , respectively .
we present here explicitly the renormalization frequency @xmath68 and the projection of the angular momentum @xmath73 which appear in the expression of the effective hamiltonian .
the frequency @xmath68 is defined for the degenerate two particle excited level @xmath47 in terms of transition amplitudes to all the states of neighbor particle numbers : where @xmath120 is the projector on the @xmath8-particle level with energy @xmath10 and @xmath121 destroys an electron of spin @xmath9 in the middle dot @xmath104 .
we defined the function @xmath122 $ ] ,
where @xmath123 , @xmath124 is the temperature and @xmath125 is the digamma function .
moreover @xmath126 is the bare tunnelling rate to the lead @xmath19 of an electron of spin @xmath127 , where @xmath20 is the tunnelling amplitude and @xmath128 is the density of states for electrons of spin @xmath127 in the lead @xmath19 at the corresponding chemical potential @xmath129 . due to the particular choice of the arbitrary phase of the @xmath130 particle excited states ,
@xmath68 does not depend on the orbital quantum number @xmath6 .
it depends instead on the bias and gate voltage through the energy of the @xmath131 , @xmath47 and @xmath132 particle states . where @xmath135 is the angle of which we have to rotate the triple dot system to bring the middle dot @xmath104 into the position of the contact dot @xmath19 . for a derivation of ( [ eq : angular_momentum ] ) see the supplementary material of [ ]
for all degenerate subspaces , if no accidental degeneracy is present ( like for our parameter choice ) , the effective hamiltonian has the form given in , , , with the renormalization frequencies calculated using the appropriate energies and matrix elements .
if the equilibrium chemical potential is not set to zero the many - body energy spectrum should be substituted with the spectrum of the many - body free energy ( @xmath137 ) where @xmath138 is the chemical potential of the leads at zero bias .
the rest of the argumentation remains unchanged .
|
we consider nanojunctions in the single - electron tunnelling regime which , due to a high degree of spatial symmetry , have a degenerate many body spectrum . as a consequence ,
interference phenomena which cause a current blocking can occur at specific values of the bias and gate voltage .
we present here a general formalism to give necessary and sufficient conditions for interference blockade also in the presence of spin polarized leads . as an example we analyze a triple quantum dot single electron transistor ( set ) . for a set - up with parallel polarized leads ,
we show how to selectively prepare the system in each of the three states of an excited spin triplet without application of any external magnetic field .
|
molecule - based magnets present an intriguing field of materials due to their combination of magnetism with various other physical phenomena .
specifically , those materials which exhibit photomagnetic properties provide a vast potential for technological applications . in recent years , prussian blue analogues ( pbas )
have piqued the interest of the molecular magnetism community due to their range of magnetic properties .
although examples exist that are known to order magnetically at or above room temperature @xcite , the feature that distinguishes a particularly appealing subset of pbas is the phenomenon of a charge - transfer - induced spin transition ( ctist ) @xcite , which is linked with that of persistent photoinduced magnetism ( ppim ) @xcite . when these properties are combined with long - range magnetic order , the potential for information storage applications begins to emerge .
a quintessential example of such a material is the cofe - pba , a@xmath9co[fe(cn)@xmath1@xmath10h@xmath3o , where a is a monovalent alkali cation .
cofe - pbas are well - known to exhibit a photoinduced ctist up to a temperature of @xmath11 k @xcite , as well as a similar photoinduced transition within the hysteresis of the thermal ctist near , and even at , room temperature for certain stoichiometries @xcite .
however , these photoinduced effects are weaker above @xmath12 k , which is the critical temperature , @xmath13 , for the long - range magnetic ordering .
the fact that the photoeffect in the cofe - pba is also structural in nature has been utilized to effect an increase in the temperature at which photocontrol is possible , by incorporating this material into heterostructures @xcite with pbas having higher ordering temperatures and known to be pressure - sensitive @xcite . to guide work pursuing room temperature photocontrolled magnetism
, we seek a deeper fundametntal understanding of the mechanism leading to these novel effects in the heterostructures , focusing here on quantifying the amount of material , in the non - photoactive layers , affected by this photoinduced strain from the cofe - pba . in the present study ,
core ( * b * ) and core - shell ( * ba * ) particles of pbas ( fig .
[ fig1 ] ) , composed of shells of k@xmath0ni[cr(cn)@xmath1@xmath2h@xmath3o ( * a * ) , which orders ferromagnetically with @xmath4 k , surrounding bulk cores of rb@xmath5co[fe(cn)@xmath1@xmath6h@xmath3o ( * b * ) , which is photoactive and orders ferrimagnetically with @xmath7 k , have been investigated .
below @xmath4 k of the shell , these samples exhibit a persistent photoinduced decrease in low - field susceptibility , similar to those results from other core - shell particles @xcite and analogous * aba * heterostructured films @xcite .
this net decrease is attributed to the photoinduced lattice expansion in the * b * layer , which generates a strain - induced decrease in the magnetization of the * a * layer , similar to a pressure - induced decrease observed by others in a pure * a*-like material , attributed to spin - canting @xcite . upon further examination
, the core - shell susceptibility data also reveal a significant portion of the * a * material whose superexchange , @xmath8 , is perturbed with the application of this photoinduced strain from the * b * material .
k@xmath14cr(cn)@xmath15 was synthesized by treating aqueous solutions of potassium cyanide with crcl@xmath16h@xmath3o and was used after recrystallization from methanol @xcite .
deionized water used in synthetic procedures was obtained from a barnstead nanopure .
all of the other reagents were purchased from sigma - aldrich or fisher - acros and were used without further purification .
the filters used during the synthesis were fast pes bottle top filters with @xmath17 m pore size ( nalgene ) .
the synthesis of the core particles was performed at room temperature .
an aqueous solution ( @xmath18 ml ) of cocl@xmath19h@xmath3o ( @xmath20 mmol , @xmath21 mm ) , and an equal volume of an aqueous solution containing k@xmath14fe(cn)@xmath15 ( @xmath22 mmol , @xmath23 mm ) were simultaneously added dropwise to @xmath24 ml of nanopure water . to substitute the rb cation , rbcl ( @xmath25 mmol , @xmath26 mm ) was included in the cocl@xmath3 solution .
the final solution was kept under vigorous stirring for one hour after all additions were complete .
the particles were subsequently filtered under vacuum using a @xmath17 m filter before being washed with ultrapure water and redispersed in @xmath18 ml of water for use in the next step . to isolate the particles , instead of water ,
the particles were concentrated by centrifugation and dried under a flow of air . to prepare the core - shell particles , the previously synthesized core particles solution was diluted with water to 400 ml . an aqueous solution ( @xmath27 ml ) of nicl@xmath19h@xmath3o ( @xmath28 mmol , @xmath29 mm ) and an aqueous solution ( @xmath27 ml ) of k@xmath14cr(cn)@xmath15 ( @xmath30 mmol , @xmath31 mm ) were added dropwise ( @xmath32 ml / hour , peristaltic pump ) under stirring at room temperature .
the final solution was kept under vigorous stirring for one hour after all additions were complete .
once the addition was complete , the particles were filtered using a @xmath17 m filter and rinsed with ultrapure water . to isolate the particles , instead of water ,
the particles were concentrated by centrifugation and dried under a flow of air .
transmission electron microscopy ( tem ) was performed on a jeol-2010f high - resolution transmission electron microscope at @xmath24 kv .
tem grids used were carbon film on a holey carbon support film , 400 mesh , copper from ted - pella , inc .
grids were prepared by dropping @xmath33l of a solution containing @xmath34 mg of sample dispersed in @xmath35 ml of etoh by sonication for @xmath36 min .
energy dispersive x - ray spectroscopy ( eds ) was performed with an oxford instruments eds x - ray microanalysis system coupled to the tem .
room temperature powder x - ray diffraction ( xrd ) was performed on a bruker duo diffractometer .
powder samples were packed into @xmath37 mm diameter boron - rich thin walled capillary tubes purchased from the charles supper company .
diffraction patterns were collected in the range @xmath38 with a @xmath39 sec / image collection time .
the raw data from the ccd detector were converted to powder patterns using powdll converter .
peaks of the xrd spectra for ( a ) the core ( * b * ) sample and ( b ) the core - shell ( * ba * ) sample , normalized to the amplitude of the * b * peak in each , as a function of @xmath40.,scaledwidth=50.0% ] fourier - transform infrared ( ftir ) spectroscopy was performed on a nicolet 6700 thermo scientific spectrophotometer .
powder samples were mixed with kbr and pressed into pellets under @xmath12 mpa .
scans ( typically 64 ) were performed between @xmath41 and @xmath42 @xmath43 , in the region of the cn stretching bands , with a precision of @xmath44 @xmath43 .
corresponding scans of a pure kbr pellet were taken as the background measurement .
magnetization measurements were performed on a quantum design mpms xl-7 superconducting interference device ( squid ) magnetometer using a homemade fiber optic sample rod with a bundle of 10 optical fibers to bring light into the sample space @xcite .
powder samples were spread between two pieces of transparent tape and held in transparent drinking straws directly below the optical fibers .
visible white light was supplied by a @xmath11 w tungsten - halogen lamp located at room temperature .
the temperature dependence of the low - field susceptibility was measured in an applied field of @xmath18 g while warming the sample from base temperature , @xmath35 k. isothermal magnetization was measured at @xmath35 k while sweeping the field between @xmath45 kg and @xmath46 kg . all light state measurements were performed subsequent to 8 hours of isothermal irradiation at @xmath47 k , after irradiation had been ceased .
the sample was then cooled again to base temperature , where the same protocol that was used for the dark state measurements was repeated to acquire the light state data set . , of the * ba * sample near the transition temperature of the * a * constituent , where the shoulder between @xmath48 k and @xmath49
k is indicative of the shift in @xmath13 due to the photoinduced strain .
the bottom left inset shows the untreated susceptibility , @xmath50 , in the dark and light states .
the top right inset shows @xmath51 over the entire range of temperatures.,scaledwidth=50.0% ]
the tem images reveal contrast between core and shell in the * ba * particles not present for the pure * b * particles ( fig . [ fig1 ] ) .
additionally , eds line scans of individual particles show co and fe content concentrated in the cores , with ni and cr content predominately occupying the edges @xcite . the ftir spectra display a new cn stretching frequency , near @xmath52 @xmath43 , arising after addition of the * a * shell , consistent with observations for the pure * a * material .
powder xrd patterns also present two distinct lattice constants consistent with the single - phase * b * and * a * materials , respectively , while only the * b * phase is exhibited in the core particles ( fig .
[ fig2 ] ) .
the low - field susceptibility of the * ba * sample exhibits a persistent photoinduced decrease ( fig .
[ fig3 ] ) .
although the * b * material is a well - established photomagnet @xcite , these heterostructures invert the sign of the effect in addition to increasing the temperature of photocontrol @xcite .
this photoeffect possesses a significant magnitude up to @xmath53 k of the * a * material , attributed to a strain - induced canting of the * a * spins , similar to pressure - induced effects in a similar material @xcite .
additionally , a subtle shoulder near the @xmath13 of the * a * material emerges in the photoinduced decrease , @xmath51 , indicating a shift in the superexchange of some portion of the shell . in order to quantify the amount of affected material , the derivative of the susceptibility was taken and normalized to the same area , in the range @xmath47 k to @xmath54 k , for each data set ( fig .
[ fig4 ] ) .
the validity of normalizing the light and dark data sets to equal areas is substantiated by the corresponding saturation magnetizations , which indicate no loss of spins upon irradiation .
thus , taking the area under these curves to represent the total number of spins in the shell , this area must remain constant for a given sample between the dark and light states . upon inspection of these derivatives , an aspect of the effect not immediately obvious in the raw data ( fig .
[ fig3 ] inset ) becomes apparent .
namely , the magnetic transition of the * a * material sharpens significantly when the * b * material is transitioned to the light state . again
, this observation indicates that a significant portion of the * a * material experiences a change in its superexchange .
, in the dark and light states near the transition temperature of the * a * constituent .
the inset shows @xmath55 over the entire range of temperatures.,scaledwidth=50.0% ] guided by these observations , we apply a simple model of the shell comprising two distinct regions : an inner , strain - sensitive interface region and an outer , rigid bulk - like region , which remains unaffected by switching between the dark and light states . taking these regions to represent distinct superexchange values , the areas under the curves in fig .
[ fig4 ] , representing the total number of spins in the shell constituent , must account for these two regions .
fitting these curves with two peaks , corresponding to the two assumed superexchange values , provides a measure of the percentage of the shell in each region .
the bulk - like value is taken to be the higher @xmath13 , which does not shift , and this value remains constant between the fits , while the total area is constrained to unity .
these fits reveal 50% of the material with each @xmath13 , where the bulk - like material has @xmath56 k and the strained material has @xmath57 k shifting to @xmath58 k , which is closer to that attributed to the bulk .
the direction of this shift may be explained phenomenologically in the following manner . upon cooling ,
the * b * material undergoes a lattice contraction from a @xmath59 lattice constant to @xmath60 , thus stretching the * a * lattice near the interface . with irradiation ,
the * b * lattice again expands to its room temperature size of @xmath59 , allowing the interfacial * a * lattice to relax closer to its bulk state .
by analyzing the first derivative of the susceptibilities in the dark and light states of the * ba * particles , we have quantified the amount of the * a * material affected by the strain arising from the photinduced expansion of the * b * lattice . although the shift in superexchange is found to be small ( @xmath61 k ) , nearly _
half _ of the volume of the * a * shell experiences this superexchange perturbation upon photoswitching the * b * core .
a forthcoming report will detail the investigation of the same cores with increasing shell thicknesses .
this work was supported by nsf dmr-1005581 ( drt ) , dmr-1202033 ( mwm ) , the national high magnetic field laboratory via cooperative agreement nsf dmr-0654118 , and the state of florida .
the authors would like to thank dr .
khalil a. abboud , for the use of the bruker duo instrument , and kerry siebein , at the university of florida major analytical instrumentation center , for tem and eds work .
s. ferlay , t. mallah , r. ouahs , p. veillet , m. verdaguer , inorg . chem . * 38 * , 229 - 234 ( 1999 ) .
m. verdaguer , a. bleuzen , c. train , r. garde , f. f. biani , c. desplanches , phil .
a * 357 * , 2959 - 2976 ( 1999 ) .
n. shimamoto , s. ohkoshi , o. sato , k. hashimoto , chem . *
41 * , 678 - 684 ( 2002 ) .
d. a. pejacovi , j. l. manson , j. s. miller , a. j. epstein , phys .
lett . * 85 * , 1994 - 1997 ( 2000 ) .
o. sato , t. iyoda , a. fujishima , k. hashimoto , science * 272 * , 704 - 705 ( 1996 ) .
o. sato , y. einaga , a. fujishima , k. hashimoto , inorg .
chem . * 38 * , 4405 - 4412 ( 1999 ) .
a. bleuzen , c. lomenech , v .
escax , f. villain , f. varret , c. cartier dit moulin , m. verdaguer , j. am .
soc . * 122 * , 6648 - 6652 ( 2000 ) .
n. shimamoto , s. ohkoshi , o. sato , k. hashimoto , chem . lett . * 4 * , 486 - 487 ( 2002 ) .
d. m. pajerowski , m. j. andrus , j. e. gardner , e. s. knowles , m. w. meisel , d. r. talham , j. am .
soc . * 132 * , 4058 - 4059 ( 2010 ) .
m. f. dumont , e. s. knowles , a. guiet , d. m. pajerowski , a. gomez , s. w. kycia , m. w. meisel , d. r. talham , inorg .
50 * , 4295 - 4300 ( 2011 ) .
d. m. pajerowski , j. e. gardner , f. a. frye , m. j. andrus , m. f. dumont , e. s. knowles , m. w. meisel , d. r. talham , chem . mater . * 23 * , 3045 - 3053 ( 2011 ) .
m. zentkov , z. arnold , j. kamard , v. kaveansk , m. lukov , s. mat@xmath62a , m. mihalik , z.mitrov , a. zentko , j. phys . : condens .
matter * 19 * , 266217 ( 2007 ) .
j. h. bigelow , inorg . synth . * 2 * , 203 - 205 ( 1946 ) . j .-
. park , ph.d .
dissertation , university of florida ( 2006 ) , http://etd.fcla.edu/uf/ufe0013792/park_j.pdf .
m. f. dumont , ph.d . dissertation , university of florida ( 2011 ) , http://etd.fcla.edu/uf/ufe0042791/dumont_m.pdf .
|
cubic heterostructured ( * ba * ) particles of prussian blue analogues , composed of a shell of ferromagnetic k@xmath0ni[cr(cn)@xmath1@xmath2h@xmath3o ( * a * ) , @xmath4 k , surrounding a bulk core of photoactive ferrimagnetic rb@xmath5co[fe(cn)@xmath1@xmath6h@xmath3o ( * b * ) , @xmath7 k , have been studied . below @xmath4 k , these samples exhibit a persistent photoinduced decrease in low - field magnetization , and these results resemble data from other core - shell particles and analogous * aba * heterostructured films .
this net decrease suggests that the photoinduced lattice expansion in the * b * layer generates a strain - induced decrease in the magnetization of the * a * layer , similar to a pressure - induced decrease observed by others in a pure * a*-like material and by us in our * ba * cubes . upon further examination
, the data also reveal a significant portion of the * a * material whose superexchange , @xmath8 , is perturbed by the photoinduced strain from the * b * constituent .
|
one of the first wavelets ever considered was the mexican hat ( see figure [ mexican ] ) , which is @xmath3 the mexican hat function @xmath4 , and its fourier transform @xmath5 . ]
the mexican hat function @xmath4 , and its fourier transform @xmath5 . ]
morlet @xcite generated the mexican hat system @xmath6 and applied it to geophysical problems as if it were orthonormal ( which it is not ) .
daubechies @xcite explained why morlet s approach succeeded : the system is an almost - tight _ frame _ , with its frame operator @xmath7 being a bijection on @xmath8 , and indeed being rather close to the identity . in other words ,
analysis followed by synthesis with the mexican hat system provides almost - perfect reconstruction in @xmath8 .
hence the frame operator is surjective , and so every square integrable function can be decomposed into a norm convergent series in terms of the wavelet system @xmath9 .
that is , each @xmath10 has an unconditionally convergent series expansion @xmath11 for some coefficients @xmath12 .
do such series expansions hold also when @xmath13 for @xmath14 ?
meyer highlighted the extent of our ignorance on this issue when he remarked that _
`` we do not know whether the functions @xmath15 , form a complete set in @xmath16 for @xmath17 '' _ @xcite .
in other words , are finite linear combinations of the @xmath9 dense in @xmath0 ?
note that proving existence of series expansions in @xmath0 would be even stronger than completeness .
two counterexamples reveal the subtlety of the problem .
tao @xcite showed for each @xmath18 that there exists a smooth , compactly supported synthesizer @xmath19 that is `` good '' on @xmath8 ( generating a wavelet frame there ) but is `` bad '' on @xmath0 ( having non - bijective frame operator ) . for @xmath20 , tchamitchian and lemari
@xcite found examples with even worse behavior .
their system @xmath21 fails even to be complete in @xmath0 . despite these obstacles , we recently answered meyer s completeness question in the affirmative , by proving @xmath0-completeness criteria for general wavelet systems @xcite , and verifying those criteria for the specific mexican hat example . yet
the stronger question of series expansions in @xmath0 remained open .
this paper answers the stronger question , by proving criteria for wavelet series expansions of arbitrary functions in @xmath22 , as well as in the hardy space @xmath1 and @xmath2 ( see theorem [ invertible ] ) .
these bijectivity criteria on the frame operator are explicitly computable .
we verify them numerically for the mexican hat system , in section [ mexicanexample ] . to prove the main results ( theorem [ invertible ] and corollary [ cor - expansions ] ) we develop explicit norm estimates on general caldern
zygmund operators ( proposition [ czatomicbound ] ) and computable bounds on the caldern
zygmund constants for the wavelet frame kernel ( theorem [ frameczofreq ] ) .
an innovative feature of our work is that we estimate the wavelet frame kernel in the frequency domain .
prior approaches proceeded in the time domain : see ( * ? ? ?
* ch.5 ) , @xcite , and related approaches @xcite .
estimates in the frequency domain are better for our purposes , for three reasons , first , we can estimate the caldern
zygmund constants while keeping the analyzer and synthesizer together instead of splitting them apart ( as discussed in section [ freqestimates ] ) .
second , in the frequency domain we need not preserve cancelations , because @xmath23 is nonnegative for many examples ( such as the mexican hat in figure [ mexican ] ) .
third , the construction of a `` good '' analyzer is easier in the frequency domain , where in order to obtain almost - perfect reconstruction of @xmath8 we need only satisfy a discrete caldern condition and ensure that the overlaps " are sufficiently small ( see section [ mexicanexample ] ) .
the techniques in this paper are largely unrelated to our earlier work @xcite , which relied on @xmath8-type methods in the frequency domain and did not use the caldern zygmund theory . to state the main result , we need some definitions .
fix @xmath24 with @xmath25 call @xmath26 the _ dilation factor _ , and @xmath27 the _ translation step_. we rescale functions @xmath28 by @xmath29 the wavelet _ synthesis operator _ associated with @xmath19 is the map @xmath30 where the coefficients @xmath12 are complex numbers .
the _ analysis operator _ associated with @xmath31 is the map @xmath32 where @xmath33 .
( the `` @xmath27 '' in the definition of analysis appears only for later convenience .
some authors use @xmath34 instead , to emphasize their riemann sum interpretation with respect to the continuous parameter case ; see @xcite . )
sometimes we write @xmath35 and @xmath36 , to emphasize the dependence of these operators on the underlying synthesizer and analyzer . the _ wavelet frame operator _ consists of analysis followed by synthesis : @xmath37 write @xmath38 for the identity function on @xmath39 , and @xmath40 for the usual sobolev space .
also , write @xmath41 to mean @xmath42
. two quantities @xmath43 and @xmath44 will be defined later , in section [ mainproofs ] .
our main result says that if these quantities are less than @xmath45 , then the wavelet frame operator is bijective on @xmath1 and @xmath2 and hence on @xmath0 for @xmath17 .
[ invertible ] assume @xmath46 with @xmath47 , and that the derivatives decay near the origin and infinity according to @xmath48 for some @xmath49 .
if @xmath50 and @xmath51 , then @xmath52 extends from @xmath8 to a bounded , linear and bijective operator on @xmath1 , on @xmath0 for @xmath53 , and on @xmath2 .
the proof , in section [ mainproofs ] , involves a quantitative perturbation estimate for wavelet frames : we show @xmath52 lies within distance @xmath45 of the identity operator on @xmath1 and hence is invertible there .
clearly this approach depends sensitively on the choice of norm for @xmath1 .
we employ the @xmath8-atomic norm , as defined in the following section .
of course , after the frame operator has been proved bijective , it remains bijective under any equivalent norm .
the constants @xmath54 and @xmath55 in theorem [ invertible ] depend on two types of information : ( i ) calderon zygmund bounds on the frame operator , which we control with frequency - domain estimates on the analyzer and synthesizer , and ( ii ) @xmath8-norm bounds on the frame operator , which we obtain in practice from daubechies - type frame conditions . ( the original daubechies condition can be found in @xcite . a later variant is due to casazza and christensen @xcite . ) the precise dependence of @xmath54 and @xmath55 on these quantities is explained in the definitions in section [ mainproofs ] , and then revisited in a specific situation in section [ mexicanexample ] , when we treat the mexican hat .
next we give a strong answer to meyer s completeness question about non - orthogonal wavelets , for the class of wavelets in our main theorem .
[ cor - expansions ] assume all the hypotheses of theorem [ invertible ] hold .
then every function @xmath56 in @xmath1 , or @xmath2 , or @xmath0 for @xmath17 , can be expressed as a norm convergent wavelet series of the form @xmath57 .
we apply the corollary in section [ mexicanexample ] to the mexican hat example .
there we construct an analyzer @xmath31 which , when combined with the mexican hat synthesizer , generates a frame operator lying less than @xmath58 from the identity on @xmath8 and less than about @xmath59 from the identity in @xmath1 .
thus the frame operator is a bijection , and so the mexican hat wavelet can reproduce each function in @xmath60 and @xmath2 .
note that wavelet systems can not reproduce @xmath61-functions with nonzero integral , because the wavelet @xmath19 and all its dilates have integral zero .
interestingly , reproduction in @xmath0 with @xmath62 can be achieved by using only the small scales ( terms with @xmath63 ) , as shown in @xcite .
open problems that a motivated reader might pursue include higher dimensional analogues of theorem [ invertible ] , and the case of @xmath64 spaces for @xmath65 . motivating literature for our approach includes : the @xmath8 frame work of chui and shi @xcite , the phi - transform theory of band limited dual frames for triebel
lizorkin spaces by frazier and jawerth @xcite with recent extension to irregular translations by cabrelli , molter and romero @xcite , the co - orbit theory of feichtinger and grchenig @xcite , and the non - band limited but oversampled `` approximate duals '' of gilbert , han , hogan , lakey , weiland and weiss @xcite .
note that gilbert _
et al . _ treated translations in a lattice for the full scale of triebel lizorkin spaces .
their work was extended to irregular translation grids ( still oversampled ) by li and sun @xcite , in the restricted case of @xmath66 . underlying the approaches of feichtinger and grchenig and gilbert _ et al . _
( and frazier and jawerth @xcite , bui and paluszyski @xcite , and li and sun @xcite ) is the fact that a sufficiently dense oversampling of translations and dilations must yield approximate reconstruction in any reasonable function space , for any reasonable analyzer and synthesizer .
the reason is that by the caldern reproducing formula , perfect reconstruction holds in the limit of infinite oversampling ( the case of continuous parameter translations and dilations ) .
unfortunately , those papers do not specify the degree of oversampling required to guarantee their conclusions .
in other words , the dilation factor @xmath26 must lie sufficiently close to @xmath45 and the translation step @xmath27 sufficiently close to @xmath67 , but _ how _ close would constitute `` sufficiently close '' is simply not specified .
theorem [ invertible ] avoids oversampling completely , by constructing a criterion that is usable for fixed values of @xmath26 and @xmath27 .
further , the theorem relaxes the band limitation of frazier and jawerth , so that one may treat non - band limited examples such as the mexican hat .
the cost of that relaxation is merely that the frame operator yields approximate rather than perfect reconstruction : our frame operator is bijective rather than equalling the identity .
let us now contrast our work with the established theory of _ orthonormal _ wavelets .
in that setting , the frame operator gives perfect reconstruction on @xmath8 and is bounded on other function spaces such as @xmath1 and @xmath0 , assuming reasonable smoothness and decay of the wavelet .
the frame operator therefore equals the identity on those spaces , simply by density of @xmath8 , and thus one obtains wavelet expansions in whole families of function spaces @xcite . in this current paper
, we do not assume orthonormality or perfect reconstruction on @xmath8 .
we assume instead that the difference between the frame operator and the identity has small norm in @xmath8 .
the challenge in theorem [ invertible ] is to show ( under suitable hypotheses ) that the norm of this difference remains small when considered in the hardy space and @xmath2 , so that the frame operator is bijective on those spaces .
our work is unrelated also to multiresolution analyses that generate pairs of perfectly reconstructing analyzers and synthesizers , such as @xcite .
such methods do not apply to the mexican hat wavelet , for example , because it satisfies no scaling or refinement equation .
theorem [ invertible ] takes a different approach : it finds sufficient conditions for a given analyzer and synthesizer to generate a frame - like structure .
we employ the fourier transform with @xmath68 in the exponent , @xmath69 so that parseval s identity says @xmath70 .
next we specify the norms to be used on hardy space and @xmath2 , and define the caldern zygmund operators that will act upon those spaces . as always , we write @xmath71 .
we work with the hardy space @xmath72 defined by @xmath8 atoms , as follows .
call @xmath73 an _ atom _ if @xmath74 is supported in some bounded interval @xmath75 with @xmath76 and @xmath77 .
notice @xmath78 .
define @xmath79 ( note the series @xmath80 converges in @xmath61 . )
then @xmath1 is a banach space under the atomic norm @xmath81 and it imbeds continuously into @xmath61 with @xmath82 the dual of the hardy space is @xmath83 , the space of functions with bounded mean oscillation .
we employ the dual space norm on @xmath2 : @xmath84 call @xmath85 a _ generalized caldern zygmund operator _ if assumptions 1a , 1b , 1c , 2 and 3 below all hold , for some constants @xmath86 .
suppose @xmath87 is a measurable , complex valued function on @xmath88 .
_ assumption 1a .
_ @xmath89 is a bounded linear operator such that if @xmath90 has compact support then @xmath91 _ assumption 3 . _ if @xmath90 has compact support and integral zero , then @xmath92 has @xmath93 , and @xmath94 has @xmath95 .
note that the integral representation need not hold when @xmath96 , for example if @xmath85 is the identity operator and @xmath97 .
_ this definition of generalized czo follows the treatment by hernndez and weiss @xcite .
a more general distributional definition was developed earlier by coifman and meyer @xcite ; see also grafakos @xcite .
define @xmath98 so that @xmath99 as @xmath100 .
our wavelet results depend on boundedness of generalized caldern zygmund operators
. boundedness of czos on @xmath101 and @xmath2 is well known @xcite , but boundedness is not enough for our purposes .
we need a certain operator norm less than @xmath45 , in order to prove invertibility of the wavelet frame operator by using a neumann series .
thus we must have _ explicitly computable _ norm bounds on czos .
the point of the next proposition is to prove such explicit bounds .
the constant used below is @xmath102 notice @xmath103 .
write @xmath104 for the conjugate exponent : @xmath105 .
[ czatomicbound ] assume @xmath106 is a generalized calderon
zygmund operator with constants @xmath107 .
fix @xmath108 .
then @xmath85 has a bounded linear extension to @xmath1 , to @xmath0 for each @xmath53 , and to @xmath2 , with norms : @xmath109 equality holds for @xmath110 . the purpose of including the free parameter @xmath111 in the proposition is to improve the bounds by allowing trade offs between the size of @xmath112 and of @xmath113 and @xmath114 .
for instance , @xmath115 is a nearly - optimal choice for the mexican hat example in section [ mexicanexample ] , whereas choosing @xmath116 would give results an order of magnitude worse . in order to prove proposition [ czatomicbound ] at the end of this section
, we will find explicit bounds saying ( roughly ) that generalized czos map atoms to molecules , and that molecules decompose into atoms , so that molecules belong to the hardy space .
the next two lemmas phrase these known results in a way that avoids the molecular norm , so as to prove the proposition more directly and obtain computable constants .
note that here , and in later proofs , we must not switch between different types of atoms , because doing so could introduce large additional constants into the norm estimates .
we must consistently use @xmath8 atoms .
the first lemma says that @xmath85 maps each atom to a function that decays like the square of the distance from the support of the atom .
[ atomsmolecules ] let @xmath108 .
if assumptions 1a , 1c and 2 hold and @xmath117 is an atom supported in a bounded interval @xmath75 centered at @xmath118 , then @xmath119 with @xmath120 where the constant is @xmath121 here @xmath122 denotes the interval having the same center as @xmath75 and @xmath111 times the length .
we adapt parts of the standard proof ; see for example ( * ? ? ?
* theorem 5.6.8 ) .
first , @xmath123 by assumption 2 , since the atom @xmath74 belongs to @xmath8 .
next , if @xmath124 lies outside the support @xmath75 of the atom @xmath74 then @xmath125a(y ) \ , dy \label{eq - kerneldiff}\end{aligned}\ ] ] since @xmath126 .
now suppose @xmath127 and @xmath128 .
then @xmath129 where the middle inequality uses that @xmath108 .
assumption 1c then implies with that @xmath130 by a change of variable , where @xmath131 evaluating the last integral shows that @xmath132 since the atom @xmath74 has @xmath8-norm at most @xmath133 .
note @xmath134 ( since @xmath135 ) , and that the function @xmath136 is decreasing for @xmath137 .
hence we may replace @xmath138 with @xmath111 , inside the square root in the last formula .
the lemma now follows by substituting the definition of @xmath138 into the remaining factor of @xmath139 .
notice @xmath140 is locally integrable because it belongs to @xmath8 , and is globally integrable because it decays like @xmath141 at infinity .
thus @xmath142 .
[ moleculesatoms ] suppose @xmath143 and @xmath75 is a bounded interval centered at @xmath118 .
let @xmath144 and @xmath108 . if @xmath145 and @xmath146 , then @xmath147 with norm @xmath148 the proof adapts the essential ideas from coifman and weiss s decomposition of a molecule into a sum of atoms ( * ? ? ? * theorem c ) .
they used dyadic scaling , @xmath149 , whereas we obtain a better constant by optimizing over @xmath150 in the proof below . also , they chose the initial radius @xmath151 to depend on the @xmath8 norm of @xmath152 , whereas our choice of @xmath151 will depend on @xmath111 and @xmath75 .
but the central idea of decomposing @xmath152 into atoms supported on annuli comes directly from their work .
we may suppose @xmath153 , and that the interval is centered at @xmath154 , by a translation .
let @xmath155 .
( we will later choose @xmath156 . )
define @xmath157 , \qquad j=0,1,2,3,\dots \\ \omega_0 & = i_0 = [ -r , r ] = \zeta i , \\
\omega_j & = i_j \setminus i_{j-1 } = [ -\gamma^j r,\gamma^j r ] \setminus [ -\gamma^{j-1 } r,\gamma^{j-1}r ] , \qquad j=1,2,3,\dots \\ { { 1}}_j & = { { 1}}_{\omega_j } , \\ m_j & = \big ( m - \frac{1}{|\omega_j| } \int_{\omega_j } m(y ) \ , dy \big ) { { 1}}_j .\end{aligned}\ ] ] clearly @xmath158 is supported in @xmath159 , and @xmath158 is square integrable and has mean value zero .
we break the proof into several steps .
we will prove @xmath160 is an atom , for numbers @xmath161 that satisfy @xmath162 to start proving these claims , let @xmath163 so that @xmath160 satisfies the definition of an atom supported in the interval @xmath164 .
( this choice of @xmath161 is positive except when @xmath165 , and in that exceptional case we may instead choose @xmath161 to be any positive number satisfying or . )
thus the task is to estimate the norm of @xmath158 and hence to establish and .
first notice that for @xmath166 , @xmath167 hence in particular , @xmath168 by and the triangle inequality .
thus @xmath169 , which is the desired estimate .
now consider @xmath170 .
note that @xmath171 if @xmath172 then latexmath:[\[\label{mest } @xmath154 , and so by , @xmath174 squaring and integrating shows that @xmath175 substituting this estimate into the definition of @xmath161 implies the bound .
step 2 . from step 1
we conclude that @xmath176 belongs to @xmath1 , with norm @xmath177 step 3 . to construct the remaining parts of the decomposition of @xmath152 , we let @xmath178 so that @xmath179 we will show @xmath180 belongs to @xmath1 . let @xmath181
. then @xmath182 by summation by parts and using that @xmath183 by hypothesis .
moreover , @xmath184 is an atom supported in the interval @xmath185 , because it has integral zero and because the disjointness of @xmath186 and @xmath187 implies that @xmath188 hence @xmath180 belongs to @xmath1 , with atomic norm at most @xmath189
lastly , we observe that @xmath190 hence by and a geometric sum , the atomic norm of @xmath180 is at most @xmath191 step 4 .
clearly we may decompose @xmath152 as @xmath192 and so the bounds from step 2 and step 3 show that @xmath193 the right side is minimal when @xmath194 .
we choose @xmath156 , which lies close to the minimum point and gives a simple formula .
the lemma follows immediately .
now we can deduce boundedness of @xmath85 on atoms .
[ atombound ] suppose assumptions 1a , 1c , 2 , 3 hold , and that @xmath117 is an atom .
let @xmath108 .
then @xmath195 with @xmath196 since @xmath74 is an atom , we have @xmath119 with @xmath197 by assumption 3 , and so @xmath198 for all @xmath128 , by lemma [ atomsmolecules ] ( using assumptions 1a , 1c and 2 ) .
hence by lemma [ moleculesatoms ] , we conclude @xmath195 with @xmath199 now we finish the proof by substituting for @xmath200 in terms of @xmath114 ( using the formula in lemma [ atomsmolecules ] ) and recalling that @xmath76 . next we develop a bound from @xmath61 to @xmath201 .
the result is standard ( * ? ? ?
* theorem 8.2.1 ) , and we include the lemma only to get a formula for the constant @xmath202 . [ weakl1 ]
let @xmath108 . if assumptions 1a , 1c , 2 hold and @xmath203 , then @xmath204 where @xmath205 .
hence @xmath85 extends to a linear operator from @xmath61 to weak-@xmath61 that satisfies the same bound .
first we show that if @xmath206 is supported in a bounded interval @xmath75 centered at @xmath118 and @xmath207 has integral @xmath67 , then @xmath208 indeed , by we have @xmath209 when @xmath135 .
integrating gives @xmath210 we evaluate the inner integral by changing variable like in the proof of lemma [ atomsmolecules ] : @xmath211 since @xmath212 when @xmath127 .
hence we have proved . to prove the lemma
, we now apply the proof of ( * ? ? ?
* theorem 8.2.1 ) to @xmath56 , except changing the interval @xmath213 in that proof to @xmath75 , and changing the interval @xmath214 to @xmath122 , and using as part of our estimate .
one hence obtains the bound @xmath215 where @xmath216 is a free parameter .
( if we directly followed @xcite then we would obtain @xmath217 instead of @xmath114 , but we gain a factor of @xmath218 by noting that the `` bad '' function in the caldern
zygmund decomposition has @xmath61-norm at most @xmath219 . ) minimizing over the choice of @xmath220 yields the constant @xmath202 stated in the lemma .
boundedness of @xmath85 on @xmath0 for @xmath18 , as stated in estimate , holds by combining the weak @xmath221 bound in lemma [ weakl1 ] and the strong @xmath222 bound in assumption 2 with the explicit interpolation result in proposition [ interpr ] ( taking @xmath223 there ) .
boundedness for @xmath224 as in follows by duality , because @xmath225 satisfies caldern zygmund assumptions 1 , 2 and 3 for the kernel @xmath226 , except with the constants @xmath113 and @xmath114 interchanged .
now we prove @xmath85 is bounded on @xmath1 .
let @xmath227 .
then @xmath56 belongs to @xmath61 , and so @xmath228 belongs to weak-@xmath61 by lemma [ weakl1 ] . to obtain a formula for @xmath228
, we consider an atomic decomposition @xmath229 of @xmath56 .
recall @xmath230 with its norm bounded by an absolute constant , by lemma [ atombound ] .
hence the series @xmath231 converges absolutely in @xmath1 .
in fact , this series equals @xmath228 a.e . by an argument that uses the weak-@xmath61 result ( see @xcite ) .
therefore the formula @xmath232 holds , and holds for each choice of atomic decomposition .
the @xmath1 norm estimate for @xmath85 in now follows from the bound on atoms in lemma [ atombound ] .
boundedness of @xmath85 on @xmath2 as in is immediate from duality and the boundedness of @xmath233 on @xmath1 .
the wavelet frame operator @xmath52 is known to be a caldern zygmund operator , under suitable hypotheses on the synthesizer and analyzer .
the existing proofs proceed in the spatial ( or time ) domain , using @xmath61 majorants : see @xcite , ( * ? ? ?
* chapter 7 ) or the explicit version in ( * ? ? ?
* lemma 3.1 ) .
unfortunately , in practice such time domain estimates are too big ( by an order of magnitude ) to prove invertibility for examples such as the mexican hat .
we need better estimates that can capture cancelations in the analyzer and synthesizer .
theorem [ frameczofreq ] in this section develops improved estimates on those caldern zygmund constants of the wavelet frame operator , by working in the frequency domain and keeping @xmath23 and @xmath234 together there ( rather than `` splitting apart '' @xmath19 and @xmath31 in the time domain , as one does in the standard approach ) .
then later we use the theorem to prove our main result , in section [ mainproofs ] .
recall the _ frame operator _
associated with synthesizer @xmath19 and analyzer @xmath31 : @xmath235 formally , we may rewrite the frame operator as an integral operator @xmath236 where the _ wavelet frame kernel _ is defined formally by @xmath237 denote the portion of the kernel at level @xmath238 by @xmath239 we will prove absolute convergence of the series defining the kernel @xmath240 , in the time domain , and of the series for its partial derivatives . even though we ultimately want estimates in the frequency domain
, the convergence of the kernel in the time domain ensures that the calderon
zygmund assumptions 1 and 2 make sense .
[ k0properties ] assume @xmath19 is bounded and @xmath241 .
then the series @xmath242 is absolutely convergent , for each @xmath243 .
further : \(i ) if @xmath19 and @xmath31 are continuous , then so is @xmath244 .
\(ii ) if @xmath19 is continuously differentiable , @xmath245 is bounded , and @xmath31 is continuous , then the partial derivative @xmath246 exists and is continuous . the assumption @xmath241 has been chosen for its simplicity , and could certainly be weakened .
it suffices to assume that @xmath31 has a bounded , symmetric decreasing , integrable majorant , such as @xmath247 .
similar comments apply to other lemmas below . _ note . _
a formula for @xmath248 follows by interchanging the roles of @xmath19 and @xmath31 . for the absolute convergence , observe that @xmath249 \(i ) now suppose in addition that @xmath19 and @xmath31 are continuous .
then @xmath250 as @xmath251 by an application of dominated convergence .
\(ii ) by definition of the partial derivative , @xmath252 \overline{\phi(y - bk ) } \\ & = b \sum_{k \in { { \mathbb z } } } \psi^\prime(x - bk ) \overline{\phi(y - bk)}\end{aligned}\ ] ] using dominated convergence , since the difference quotient for @xmath19 is bounded by @xmath253 .
continuity follows by part ( i ) , since @xmath245 is continuous .
now we show that the series for the full kernel @xmath240 converges .
[ kproperties ] assume @xmath254 and @xmath255 are @xmath256 .
then the series for @xmath87 converges absolutely whenever @xmath257 , with @xmath258 further : \(i ) if @xmath19 and @xmath31 are continuous , then so is @xmath87 wherever @xmath257 .
more precisely , the series @xmath259 converges uniformly on the set @xmath260 , for each @xmath49 , and hence @xmath240 is continuous there .
\(ii ) if @xmath19 is continuously differentiable , @xmath31 is continuous , and @xmath261 and @xmath255 are @xmath262 , then the partial derivative @xmath263 exists by term - by - term differentiation wherever @xmath257 , and @xmath263 is continuous .
this term - by - term derivative converges absolutely , and satisfies @xmath264 interchanging the roles of @xmath19 and @xmath31 gives an analogous formula for @xmath265 .
we adapt ideas from ( * ? ? ?
* lemma 5.3.12 and 5.6 ) ; or see @xcite .
write @xmath266 . for each @xmath267
, the triangle inequality implies @xmath268 , and so either @xmath269 or else @xmath270 .
hence we may `` split apart '' @xmath19 and @xmath31 , or rather the @xmath271-terms that bound them ; that is , @xmath272 where @xmath273 is some constant ( which changes from line to line , and depends on @xmath27 ) . summing the last estimate over @xmath274
gives that @xmath275 by lemma [ jsum ] in the appendix .
\(i ) let @xmath49 and @xmath276 .
since @xmath271 is symmetric decreasing , the proof above gives @xmath277 hence the series @xmath278 converges uniformly on the set where @xmath279 . since @xmath244 is continuous by lemma [ k0properties](i ) , it follows that @xmath240 is continuous when @xmath276 .
\(ii ) write @xmath280 for the majorant of @xmath245 and @xmath31 . by arguing as in the first part of this proof
, one can show that the term - by - term derivative series @xmath281 converges absolutely on @xmath282 , and is bounded by @xmath283 . by lemma [ k0properties](ii )
we know that this term - by - term derivative equals @xmath284 and that @xmath244 is continuously differentiable with respect to @xmath124 .
this last series converges uniformly when @xmath276 , by arguments similar to part ( i ) above , and hence it converges to a continuous function .
this continuous function is the derivative @xmath263
. a function will possess majorants of the above type as soon as its fourier transform possesses enough derivatives in @xmath61 , as the next lemma summarizes .
[ sobolevmajorant ] if @xmath285 then @xmath19 is continuous and @xmath286 . if @xmath287 then @xmath19 is continuous and @xmath288 .
from @xmath289 we have that @xmath290 is continuous ( possibly after redefining @xmath19 on a set of measure zero ) .
integrating by parts twice gives that @xmath291 hence @xmath292 if @xmath287 , then integrating by parts a third time yields a majorant that decays like @xmath293 at infinity . the hypotheses in the last lemma can be weakened to assume just that @xmath23 belongs to an inhomogeneous besov space , @xmath294 or @xmath295 respectively .
those weaker assumptions do not appear useful for our work , though , because the besov space norm seems too complicated for explicit estimations to be practical .
[ frameczokernelrep ] assume @xmath296 decay near the origin and infinity according to @xmath297 for some @xmath49 .
assume @xmath298 .
then the frame operator @xmath52 satisfies caldern zygmund assumption 2 : it is linear and bounded on @xmath8 , and if @xmath90 has compact support and @xmath299 , then @xmath300 the frame operator @xmath52 is linear and bounded on @xmath8 , for example by ( * ? ? ?
* propositions 6 and 7 ) . we know
@xmath19 and @xmath31 are @xmath256 , by lemma [ sobolevmajorant ] and the hypothesis that @xmath298 .
thus by lemma [ kproperties ] , the kernel satisfies @xmath301 .
if @xmath90 has compact support and @xmath302 , then the function @xmath303 is bounded on the support of @xmath56 , and so by dominated convergence , @xmath304 [ frameczointzero ] assume @xmath306 with @xmath47 , and that @xmath307 and @xmath308 decay near the origin and infinity according to .
then the frame operator @xmath52 satisfies caldern zygmund assumption 3 : if @xmath90 has compact support and integral zero , then @xmath309 with @xmath310 , and @xmath311 with @xmath312 .
if @xmath90 has compact support and integral zero , then its fourier transform vanishes at the origin and belongs to the sobolev space @xmath313 .
hence by boundedness of analysis and synthesis on the space of such `` littlewood paley functions '' ( as in ( * ? ? ? * propositions 5 and 6 ) ) , the wavelet series defining @xmath314 converges in the littlewood paley space .
then the wavelet series converges also in @xmath61 , by an easy imbedding ( * ? ? ?
* formula ( 1 ) ) .
integrating the series term - by - term gives that @xmath315 because @xmath316 . by interchanging the roles of @xmath31 and @xmath19 we see the same conclusion holds for the @xmath8-adjoint , since @xmath317 .
as explained at the beginning of the section , we need good , explicit estimates on the caldern zygmund constants of the wavelet frame kernel .
our estimates involve the following frequency - domain norm quantities : @xmath318 and @xmath319^{\prime \prime } \big\rvert_{l^1 } \\
\tau_2(\psi,\phi ) & = \frac{1}{4\pi^2 } \sum_{l \in { { \mathbb z } } } \big\lvert \big [ x(\cdot ) \widehat{\psi}(\cdot ) \overline{\widehat{\phi}(\cdot+lb^{-1})}\big]^{\prime \prime \prime } \big\rvert_{l^1 }
\\ \tau_3(\psi,\phi ) & = \frac{1}{4\pi^2 } \sum_{l \in { { \mathbb z } } } \big\lvert \big [ x(\cdot ) \widehat{\psi } ( \cdot + lb^{-1 } ) \overline{\widehat{\phi}(\cdot ) } \big]^{\prime \prime \prime } \big\rvert_{l^1}\end{aligned}\ ] ] where @xmath38 denotes the identity function . now we can estimate the caldern
zygmund constants @xmath107 . [ frameczofreq ]
assume @xmath320 with @xmath321 .
then the wavelet frame kernel @xmath87 in satisfies caldern zygmund assumptions 1a , 1b , 1c with the constants @xmath322{\sigma_2(\psi,\phi ) \tau_2(\psi,\phi)^2 } , \\
c_3(\psi,\phi ) & = \frac{4|a| ( 2|a|+1)}{|a|^2 - 1 } \sqrt[3]{\sigma_3(\psi,\phi ) \tau_3(\psi,\phi)^2 } , \end{aligned}\ ] ] provided
these constants are finite .
( _ aside . _
these constants blow up as @xmath323 .
this apparent singularity would disappear if we had multiplied the frame operator by @xmath324 , as discussed in the introduction .
anyhow , the issue is irrelevant because we fix the dilation factor @xmath26 in this paper . )
the proof will build on off - diagonal decay lemmas for @xmath244 and its derivatives .
the first lemma explains the central idea : we invoke poisson summation to transfer to the frequency domain , and then integrate by parts .
[ kernel1 ] assume @xmath320 with @xmath325 , and that @xmath326 and @xmath327 .
then for all @xmath243 , @xmath328 the hypotheses guarantee that @xmath19 and @xmath31 are continuous and are @xmath256 , by lemma [ sobolevmajorant ] .
hence the series for @xmath329 converges absolutely and yields a continuous function , by lemma [ k0properties ] .
we have @xmath330 by the poisson summation formula ( justified below ) , where @xmath331 . by writing out the convolution , we find @xmath332 this last series for @xmath244 converges absolutely , with @xmath333 by definition of @xmath334 .
now we justify the use of poisson summation .
let @xmath335 , so that @xmath336 is continuous and @xmath27-periodic .
the fourier coefficients of @xmath336 have @xmath337-norm at most @xmath326 , by the calculations above .
thus the fourier series of @xmath336 is a continuous function , which therefore equals @xmath338 for every @xmath339 .
choosing @xmath340 gives the poisson summation formula as used in , since @xmath341 .
returning now to the formula for @xmath329 in , integrating by parts twice gives that @xmath342^{\prime \prime } \
, d\xi , \ ] ] which implies that @xmath343 , and finishes the proof of the lemma .
[ kernel2 ] assume @xmath320 with @xmath344 and @xmath345 , and that @xmath346 and @xmath347 .
then for all @xmath243 , @xmath348 the hypotheses @xmath344 imply by lemma [ sobolevmajorant ] that @xmath349 and @xmath31 are continuous and are @xmath256 .
thus the series for @xmath244 and @xmath350 converge absolutely and yield continuous functions , by lemma [ k0properties ] .
since @xmath351 , the lemma follows by applying the proof of lemma [ kernel1 ] to @xmath245 instead of @xmath19 , and integrating by parts three times instead of twice .
the hypotheses of the theorem ensure ( by lemma [ sobolevmajorant ] ) that @xmath352 are all @xmath262 .
recall that the kernel @xmath353 has partial derivative @xmath354 by lemma [ kproperties](ii ) .
the first caldern zygmund estimate follows by combining lemmas [ kernel1 ] and [ jsum ] ( in the appendix ) , with @xmath355 and with @xmath356 as defined in theorem [ frameczofreq ] .
next , @xmath357 by combining lemmas [ kernel2 ] and [ jsum ] , with @xmath113 as defined in theorem [ frameczofreq ] .
this derivative bound implies the second caldern
zygmund condition ; for example , see @xcite .
the third estimate follows by swapping the roles of @xmath19 and @xmath31 .
the theorem is now proved .
recall the function @xmath358 defined for @xmath359 in section [ definitions ] , and the constants @xmath360 and so on introduced in section [ freqestimates ] . define @xmath361 discarding the terms with @xmath358 gives an inequality on the @xmath8 norm : @xmath362 obviously , if @xmath363 is finite for any value of @xmath111 then it is finite for every value of @xmath111 , and similarly for @xmath364 .
[ bounded ] assume @xmath19 and @xmath31 satisfy the regularity and decay hypotheses in theorem [ invertible ] , that is , @xmath46 and @xmath47 , along with estimate .
let @xmath108 and assume @xmath365 and @xmath366 .
then the wavelet frame operator @xmath52 on @xmath8 extends to a bounded linear operator on the hardy space @xmath1 and on @xmath2 , with norm bounds : @xmath367 the frame operator @xmath52 satisfies caldern zygmund assumptions 1 , 2 and 3 , by theorem [ frameczofreq ] , proposition [ frameczokernelrep ] and proposition [ frameczointzero ] , respectively .
now proposition [ czatomicbound ] completes the proof .
next we define the constants in the statement of theorem [ invertible ] .
assume @xmath19 and @xmath31 satisfy the regularity and decay hypotheses of theorem [ invertible ] ( as in proposition [ bounded ] ) .
choose some other functions @xmath368 and @xmath369 that satisfy the same assumptions , and which in addition give perfect reconstruction on @xmath8 , meaning @xmath370 ( such functions certainly exist , for example , smooth orthonormal wavelets with sufficiently many vanishing moments . ) define the constant @xmath371 when @xmath372 .
( this notation suppresses the dependence of @xmath373 on our choices of @xmath368 and @xmath369 , but no harm will come from that omission . ) roughly speaking , if @xmath373 is `` small '' then the perturbations @xmath374 and @xmath375 are `` small '' .
note that theorem [ invertible ] assumes @xmath54 and @xmath55 are both rather small , in fact , less than @xmath45 in magnitude .
that means the perturbations are `` small '' , so that @xmath19 and @xmath31 are `` close '' to giving perfect reconstruction . in view of the linearity of @xmath376 with respect to the synthesizer @xmath19 , and linearity of @xmath377 with respect to @xmath31 ,
the perfect reconstruction in implies a decomposition @xmath378 valid on @xmath8 .
thus by proposition [ bounded ] , @xmath379 extends to a bounded operator on @xmath1 with norm bound @xmath380 for each @xmath381 .
taking the infimum of the right side with respect to @xmath111 and @xmath382 gives @xmath43 , and so we have @xmath383 by hypothesis and .
we conclude the frame operator @xmath52 is bijective on @xmath1 and on @xmath8 , with inverse given by the norm convergent neumann series @xmath384 notice this inverse operator is well defined on @xmath385 , because if @xmath386 then the inverse series @xmath387 converges to the same limit function in @xmath1 as in @xmath8 , by considering pointwise convergence of a subsequence of partial sums .
since @xmath52 and @xmath388 are bounded on @xmath1 and on @xmath8 , they extend to bounded operators on @xmath389 , by interpolation ( either the complex method ( * ? ? ?
* corollary 1 ) , or the real method ( * ? ? ?
* theorem iii.6.1 ) ) .
these extended operators remain inverses of each other on @xmath0 , by density of @xmath390 .
thus @xmath52 is bijective on @xmath0 .
argue similarly for @xmath391 , using @xmath364 and @xmath55 and then interpolating between @xmath8 and @xmath2 ( * ? ? ?
* corollary 2 ) . _ note . _ an alternative to decomposition is @xmath392
if this formula is used then one obtains an alternative version of theorem [ invertible ] , as one verifies by adapting the proof in the obvious manner .
the quantities @xmath54 and @xmath55 determine the rate of convergence of the neumann series for @xmath388 , in the proof above , for @xmath1 and @xmath2 .
to get convergence rates on @xmath0 one may interpolate @xmath393 as follows .
( this next material is not needed elsewhere in the paper . )
define @xmath394 when @xmath395 and let @xmath396 be as in . here
@xmath397 is the constant from section [ boundedczo ]
. then @xmath398 by proposition [ czatomicbound ] , and so the decomposition implies @xmath399 .
hence if @xmath400 then the neumann series for @xmath388 converges with geometric rate @xmath396 .
this approach breaks down for @xmath401 near @xmath45 because the interpolation constant @xmath397 blows up there , and so one gets @xmath402 . to obtain a convergence rate on the neumann series for such @xmath401-values
, one should instead interpolate _
powers _ of @xmath393 , as explained in @xcite .
formally , one merely notes that bijectivity of @xmath52 on each function space in question ( as given by theorem [ invertible ] ) implies surjectivity of @xmath376 , which means that each function @xmath56 can be written as @xmath403 for some coefficients @xmath12 . to make this deduction rigorous
, one needs @xmath376 and @xmath377 to be extended separately , whereas theorem [ invertible ] says only that the composition @xmath52 extends from @xmath8 to be bounded on @xmath404 and @xmath0 .
fortunately , the extension of analysis and synthesis separately to those function spaces ( and indeed to a whole scale of triebel
lizorkin spaces ) has been proved by frazier and jawerth ( * ? ? ?
* theorems 3.5 , 3.7 , and p.81 ) .
one can verify their hypotheses straightforwardly , given the assumptions in theorem [ invertible ] .
in particular they proved that @xmath405 and @xmath406 for a certain homogeneous triebel lizorkin sequence space @xmath407 , and similarly for @xmath1 when @xmath408 and @xmath2 when @xmath409 .
thus for all @xmath56 we have @xmath410 and so @xmath411 with unconditional convergence , where the coefficient sequence @xmath412 belongs to the space @xmath407 .
the mexican hat synthesizer @xmath413 is shown in figure [ mexican ] along with its fourier transform @xmath414 we assume dyadic dilations and unit translations , so that @xmath415 we wish to apply our results to this example , in particular to get expansions of arbitrary @xmath416 in terms of the mexican hat system @xmath21 . in order to apply our results , though
, we must construct a suitable analyzer @xmath31 .
first we will cut off the mexican hat in the frequency domain , to obtain a band - limited synthesizer @xmath368 . towards that end , define a `` ramp '' function @xmath417 so that @xmath418 and @xmath419 for @xmath420 .
this formula for @xmath421 implies that @xmath422 increases from @xmath67 to @xmath45 as @xmath423 increases from @xmath67 to @xmath45 , and that @xmath424 next define a cut - off function ( see figure [ kappabeta ] ) by @xmath425 so that @xmath426 increases from @xmath67 to @xmath45 as @xmath423 increases from @xmath427 to @xmath428 and similarly as @xmath423 decreases from @xmath429 to @xmath430 .
further define @xmath431 obviously @xmath368 is band - limited , with @xmath432 supported in the interval @xmath433 $ ] .
notice @xmath434 on the interval @xmath435 $ ] , because @xmath436 there . the cut - off function @xmath437 and the double bump @xmath438 , in the frequency domain . ]
the cut - off function @xmath437 and the double bump @xmath438 , in the frequency domain . ] \2 .
next we construct a @xmath31 such that analysis with @xmath31 followed by synthesis with @xmath368 gives perfect reconstruction on @xmath8 .
the method depends on a dyadic partition of unity , constructed as follows .
start by defining a @xmath439-smooth `` double bump '' function supported in the interval @xmath435 $ ] : @xmath440 as plotted in figure [ kappabeta ] .
we claim that @xmath438 generates a dyadic partition of unity , with @xmath441 indeed , when @xmath442 we have @xmath443 for all @xmath444 and the terms with @xmath445 give @xmath446 by definition of @xmath438 and .
( _ note . _
the double bump function @xmath438 is a particular case of the `` bell function '' construction used for meyer lemari wavelets @xcite . )
we define the analyzer @xmath31 by choosing its fourier transform to be @xmath447 as shown in figure [ muphi ] , which is a kind of `` band - limited reciprocal '' of @xmath23 .
hence @xmath448 on @xmath39 , because @xmath449 on the interval @xmath435 $ ] , which contains the support of @xmath438 .
hence by the partition of unity property , we see @xmath368 and @xmath31 satisfy the `` discrete caldern condition '' @xmath450 since also @xmath432 and @xmath234 are supported in the interval @xmath433 $ ] of length @xmath45 , the discrete caldern condition implies that they give perfect reconstruction , meaning @xmath451 on @xmath8 .
( this fact , that band limitation together with discrete caldern implies perfect reconstruction , can be found in ( * ? ? ? * chapter 6 ) , or else use ( * ? ? ?
* formula ( 14 ) ) , for example . )
choose @xmath452 , so that @xmath368 and @xmath369 give perfect reconstruction on @xmath8 .
then let @xmath453 note that @xmath454 .
we plot @xmath455 in figure [ muphi ] .
the fourier transforms of the cut - off synthesizer , and of the analyzer . ]
the fourier transforms of the cut - off synthesizer , and of the analyzer . ] \3 . next , for the synthesizer @xmath456 and analyzer @xmath31 we estimate the quantities @xmath457 } < 0.000045 \\ \sigma_2(\mu,\phi ) & = 2\pi \sum_{l \neq 0 } \lvert \beta(\cdot ) x(\cdot+l ) \widehat{\psi}(\cdot+l ) / \widehat{\psi}(\cdot ) \rvert_{l^1[-1/3,1/3 ] } < 0.00022 \\ \sigma_3(\mu,\phi ) & = 2\pi \sum_{l \neq 0 } \lvert \beta(\cdot ) x(\cdot ) \widehat{\psi}(\cdot + l ) / \widehat{\psi}(\cdot ) \rvert_{l^1[-1/3,1/3 ] } < 0.000067\end{aligned}\ ] ] @xmath458^{\prime \prime } \big\rvert_{l^1[-1/3,1/3 ] } < 0.00086 \\ \tau_2(\mu,\phi ) & = \frac{1}{4\pi^2 } \sum_{l \neq 0 } \big\lvert \big [ \beta(\cdot ) x(\cdot+l ) \widehat{\psi}(\cdot+l ) / \widehat{\psi } ( \cdot)\big]^{\prime \prime \prime } \big\rvert_{l^1[-1/3,1/3 ] } < 0.036 \\ \tau_3(\mu,\phi ) & = \frac{1}{4\pi^2 } \sum_{l \neq 0 } \big\lvert \big [ \beta(\cdot ) x(\cdot ) \widehat{\psi } ( \cdot + l ) / \widehat{\psi}(\cdot ) \big]^{\prime \prime \prime } \big\rvert_{l^1[-1/3,1/3 ] } < 0.014\end{aligned}\ ] ] as we now justify . the first quantity @xmath334 was defined in section [ freqestimates ] , giving @xmath459 we may exclude @xmath460 from the sum because the supports of @xmath461 and @xmath234 do not overlap ( see figure [ muphi ] ) .
further , when @xmath462 we may replace @xmath461 ( which equals @xmath463 ) with @xmath23 , because these functions agree outside the interval @xmath433 $ ]
. then we may substitute @xmath464 .
these steps lead to the formula given above for @xmath465 , which we have estimated numerically using mathematica to obtain the upper bound @xmath466 .
we proceed similarly for each of the other quantities @xmath467 . from this numerical work
we can estimate the caldern
zygmund constants in theorem [ frameczofreq ] , taking @xmath468 there : @xmath469{\sigma_2(\mu,\phi ) \tau_2(\mu,\phi)^2 } < 0.088 , \\
c_3(\mu,\phi ) & = \frac{40}{3 } \sqrt[3]{\sigma_3(\mu,\phi ) \tau_3(\mu,\phi)^2 } < 0.032 .\end{aligned}\ ] ] 4 .
we have @xmath470 on @xmath8 , since @xmath471 and @xmath451 on @xmath8 .
hence @xmath472 where @xmath473 denotes a daubechies type sufficient frame estimate ; see ( * ? ? ?
* theorem 1 ) with @xmath110 .
the quantity @xmath473 can be evaluated numerically , giving @xmath474 .
using this fact and our estimates from step 3 , we calculate from the definitions of @xmath363 and @xmath364 that @xmath475 where we chose the values of @xmath476 and @xmath477 to approximately minimize @xmath363 and @xmath364 .
( recall also here that @xmath478 , which eliminates the second term in the definitions of @xmath54 and @xmath55 . )
thus @xmath52 provides almost - perfect reconstruction on @xmath1 and @xmath2 , with error of at most @xmath479% . hence the operator
@xmath52 is bijective on @xmath1 and @xmath2 , and on @xmath0 for each @xmath53 , by theorem [ invertible ] .
we conclude from corollary [ cor - expansions ] that every function in @xmath0 can be expressed as a norm convergent series in terms of the mexican hat wavelet system @xmath21 .
thus our work provides a strong positive answer to meyer s question in the introduction .
let us state some direct estimates for the frame operator on @xmath0 , in the mexican hat example , even though we do not need them for our work : @xmath480 when @xmath481 we can not argue directly that the neumann series for @xmath388 converges , because we can not show @xmath482 in that range . that explains why in the example above we use theorem [ invertible ] , whose proof interpolates the invertibility property between @xmath1 and @xmath8 , rather than trying to prove directly that the neumann series converges .
laugesen thanks the department of mathematics and statistics at the university of canterbury , new zealand , for hosting him during much of this research .
this work was partially supported by a grant from the simons foundation ( # 204296 to richard laugesen ) .
in order to extend our decay estimates from @xmath244 to the full kernel @xmath240 , earlier in the paper , we summed over dilation scales @xmath483 . here
we collect elementary estimates used in that task .
[ jsum ] assume @xmath484 and let @xmath485 .
\(i ) if @xmath486 for all @xmath487 , then @xmath488 \(ii ) if @xmath489 for all @xmath487 , then @xmath490{\sigma \tau^2}}{|z|^2 } , \qquad z \neq 0 .\ ] ] \(i ) fix @xmath487 and let @xmath491 be any integer .
we split the sum into two pieces and estimate each piece using the assumption on @xmath492 : @xmath493 by the geometric series .
choose @xmath491 to be the smallest integer satisfying @xmath494 , so that @xmath495 .
( to motivate this choice , notice that @xmath496 if and only if @xmath497 . )
then @xmath498 from which part ( i ) of the lemma follows .
\(ii ) adapt part ( i ) , except choose @xmath491 to be the smallest integer satisfying @xmath499{\tau/\sigma}$ ] .
the marcinkiewicz interpolation theorem implies in particular that if a sublinear operator satisfies weak-@xmath500 and weak-@xmath8 bounds , then strong-@xmath0 bounds hold for @xmath401 between @xmath501 and @xmath218 . by strengthening the second hypothesis to strong-@xmath8 and calling also on riesz thorin interpolation ( which requires the operator to be linear )
, we will obtain an explicit estimate on the @xmath0 norm such that equality holds when @xmath110 .
the bound will involve the following constants , for @xmath502 : @xmath503 notice @xmath110 gives @xmath103 , and indeed @xmath504 for each @xmath501 . at the other extreme of @xmath401-values ,
we see @xmath505 blows up like @xmath506 as @xmath507 , and @xmath397 blows up like @xmath508 as @xmath509
. see figure [ cpfig ] .
if @xmath85 is weak type @xmath515 and strong type @xmath222 , then @xmath85 is strong type @xmath516 for @xmath517 , with @xmath518/\left[(2/r)-1 \right ] } \ , \lvert z \rvert_{l^2(x ) \to l^2(y)}^{\left [ ( 2/r ) - ( 2/p ) \right]/\left [ ( 2/r)-1 \right ] } .\ ] ] ( equality holds for @xmath110 . ) in particular , when @xmath223 the conclusion says for @xmath395 that @xmath519 the case @xmath223 of the proposition improves by a factor of about @xmath521 ( when @xmath401 is close to @xmath218 ) on the best bound we found in the literature , which is an exercise in the monograph by grafakos ( * ? ? ? *
exercise 1.3.2 ) .
we follow the same approach as grafakos , except in the proof below we employ the harmonic mean of @xmath45 and @xmath401 instead of the arithmetic mean ; the harmonic mean yields simpler formulas .
suppose @xmath522 .
then marcinkiewicz interpolation applied with @xmath523 ( see ( * ? ? ?
* theorem 1.3.2 ) ) implies boundness of @xmath85 on @xmath524 , with @xmath525/\left [ ( 2/r)-1 \right ] } \ , \lvert z \rvert_{l^2 \to
l^2}^{\left [ ( 2/r ) - ( 2/q ) \right]/\left [ ( 2/r)-1 \right ] } .\ ] ] next , riesz thorin interpolation applied with @xmath526 ( see ( * ? ? ?
* theorem 1.3.4 ) ) says that @xmath527/\left [ ( 2/q)-1 \right ] } \lvert z \rvert_{l^2 \to l^2}^{\left [ ( 2/q)-(2/p ) \right]/\left [ ( 2/q)-1 \right ] } .\ ] ] we substitute the marcinkiewicz bound into this last riesz thorin bound , giving that @xmath528 / ( 2-q ) } \label{interim } \\ & \qquad \qquad \qquad \times \lvert z \rvert_{l^r \to \text{weak-}l^r}^{\left[(2/p ) - 1\right]/\left [ ( 2/r)-1 \right ] } \ , \lvert z \rvert_{l^2 \to l^2}^{\left [ ( 2/r ) - ( 2/p ) \right]/\left [ ( 2/r)-1 \right ] } .
\notag\end{aligned}\ ] ] we would like to choose @xmath529 $ ] to minimize the right side of bound . that coefficient seems too complicated for analytical minimization to be feasible , but
numerical work suggests that a good ( _ i.e. _ , within a factor of about @xmath218 of being minimal ) choice for @xmath530 is the harmonic mean of @xmath501 and @xmath401 : @xmath531 substituting this choice of @xmath530 into yields the constant @xmath505 claimed in the proposition .
m. frazier , b. jawerth and g. weiss .
littlewood paley theory and the study of function spaces .
cbms regional conference series in mathematics , 79 .
published for the conference board of the mathematical sciences , washington , dc ; by the american mathematical society , providence , ri , 1991 .
j. e. gilbert , y. s. han , j. a. hogan , j. d. lakey , d. weiland and g. weiss .
_ smooth molecular decompositions of functions and singular integral operators . _
742 , 74 pp .
, 2002 .
j. garca cuerva and j. l. rubio de francia .
weighted norm inequalities and related topics .
north - holland mathematics studies , 116 .
notas de matemtica , 104 .
holland publishing co. , amsterdam , 1985 .
y. meyer and r. coifman . wavelets .
caldern zygmund and multilinear operators .
translated from the 1990 and 1991 french originals by david salinger .
cambridge studies in advanced mathematics , 48 .
cambridge university press , cambridge , 1997 .
|
we prove a sufficient condition for frame - type wavelet series in @xmath0 , the hardy space @xmath1 , and @xmath2 .
for example , functions in these spaces are shown to have expansions in terms of the mexican hat wavelet , thus giving a strong answer to an old question of meyer .
bijectivity of the wavelet frame operator acting on hardy space is established with the help of new frequency - domain estimates on the caldern zygmund constants of the frame kernel .
|
the concept of proton driven plasma wakefield acceleration emerged five years ago @xcite .
the main motivation for using protons is that the energy content of state - of - the - art proton beams is sufficiently high to bring a substantial amount of electrons to tev energy scale in a single plasma section .
since plasmas can support extremely strong accelerating fields @xcite , this concept may open a path to the future of high - energy particle physics @xcite . in the initial proposal @xcite ,
the proton beam was assumed to be longitudinally compressed to the sub - millimeter length . that was necessary for driving the plasma wave with the wavelength of 1 mm or shorter .
however , as available proton beams have lengths of about 10 cm , such strong compression seems unrealistic to obtain with conventional techniques @xcite .
soon it was realized that the plasma wave can slice the initially long proton beam into the train of equally spaced micro - bunches which in turn resonantly drives the plasma wave @xcite .
the slicing is caused by the saturated self - modulation instability @xcite . .
parameters of the baseline awake scenario and the notation . [ cols="<,<",options="header " , ] to quantify the sensitivity of wakefield phase to parameter variations , we introduce the phase sensitivity @xmath0 as @xmath1 where @xmath2 is the shift of some constant phase point .
the phase curves may change in a complicated way due to variations of beam or plasma parameters , as exemplified by fig.[f8-phases ] . for most curves ,
the strongest phase deviation is located near the amplitude maximum ( at @xmath3 m ) , while at the end of the plasma section the phase shift is small .
we therefore calculate @xmath0 in two plasma cross - sections : @xmath4 m and @xmath5 m . to this end
we take the local maximum of @xmath6 located at @xmath7 cm .
phase sensitivities listed in table [ t3 ] display several noticeable features .
the most flaring one is the high sensitivity to plasma density variations . as shown in ref.@xcite , a small density variation @xmath8 results in the relative change @xmath9 of the plasma wavelength and the forward shift of the wakefield pattern by @xmath10 , where @xmath11 is the distance between the observation point and the area of wakefield formation .
this distance was not defined strictly in ref.@xcite , and the phase sensitivity can be used to refine the definition . here
@xmath12 cm , so the full distance to the seed laser pulse is the `` effective '' wakefield length if the beam self - modulates in the plasma of a detuned density .
this contrasts to the case considered in ref.@xcite , where an already modulated beam enters a perturbed density plasma , and @xmath11 is roughly twice shorter .
the extreme phase sensitivity to plasma density variations makes it challenging to deterministically inject externally generated short electron bunches into the wakefield of self - modulating proton beam .
assume that the bunch must be placed into a certain accelerating bucket with a longitudinal precision of about @xmath13 .
then it straightforwardly follows that the plasma density must be held to a designated value within @xmath14% .
another noticeable feature is that all other sensitivities are small at @xmath5 m .
values less than unity means that even a twofold change of a parameter moves a fixed phase point by a small fraction of the wakefield period .
thus the wakefield structure established after partial destruction of micro - bunches is well phase locked to the seed perturbation . at @xmath4 m sensitivities to beam population , beam length at fixed current , and beam radius are high since these parameters determine the beam density @xmath15 which in turn governs the theoretically predicted phase velocity @xmath16 of the growing self - modulation mode @xcite : @xmath17 from ( [ e7 ] ) we may expect @xmath18 to relate as @xmath19 .
this reasonably agrees with the simulations except for the sensitivity to the beam energy which is roughly twice lower .
note also the relatively high sensitivity to the atomic weight of plasma ions at @xmath4 m . in comparison to the amplitude sensitivity
this shows that ion motion manifests itself much stronger in phase shifts than in the amplitude change .
the performed study has not only clarified the sensitivity of the excited wakefield to beam and plasma parameters ( summarized in table [ t3 ] ) , but also helped to identify physical effects limiting the wakefield amplitude in different parts of the parameter space .
the most important effect is the nonlinear elongation of the wakefield period @xcite . as long as this effect has not come into play ,
the wakefield amplitude is directly proportional to the number of protons driving the wakefield .
once the limit is achieved by improving some of the parameters , further growth of the accelerating field drastically slows down .
two other important effects are emittance driven divergence and motion of plasma ions .
both effects are characterized by some threshold values or , to be exact , surfaces in the multi - dimensional parameter space .
if the threshold is crossed , the wakefield amplitude rapidly drops .
if not , there is no influence on the wakefield amplitude .
the baseline parameter set of the awake experiment ( table [ t1 ] ) falls exactly at the onset of nonlinear period elongation and close to thresholds of emittance driven divergence and ion motion . in other words , this is the point beyond which efforts in increasing beam charge or peak current will not result in the proportional increase of the wakefield amplitude .
also there is no much safety margin in beam emittance and ion weight .
a way to stronger wakefields may probably be opened by operating at lower beam emittances and higher plasma densities , but this is a subject of a separate study .
very high sensitivity of the wakefield phase to plasma density variations stimulates interest in injection methods based on selective trapping of electrons from an initially long ( several wavelengths ) beam @xcite .
the authors are grateful to participants of awake collaboration for stimulating discussions .
18 a.caldwell , k.lotov , a.pukhov , and f.simon , nature phys . * 5 * , 363 ( 2009 ) .
k.v.lotov , phys . rev .
beams * 13 * , 041301 ( 2010 ) .
e. esarey , c. b. schroeder , and w. p. leemans , rev .
phys . * 81 * , 1229 ( 2009 ) .
awake collaboration , proton - driven plasma wakefield acceleration : a path to the future of high - energy particle physics ( to appear in plasma physics and controlled fusion , 2014 ) .
g.xia , a. caldwell , producing short proton bunch for driving plasma wakefield acceleration .
proceedings of ipac2010 ( kyoto , japan ) , p.43954397 .
a. caldwell , k. lotov , a. pukhov and g. xia , plasma phys .
controlled fusion * 53 * , 014003 ( 2011 ) .
n.kumar , a.pukhov , and k.lotov , phys .
lett . * 104 * , 255003 ( 2010 ) .
k.v.lotov , instability of long driving beams in plasma wakefield accelerators .
6th european particle accelerator conference ( stockholm , 1998 ) , p.806 - 808 .
a. pukhov , n. kumar , t. tuckmantel , a. upadhyay , k. lotov , p. muggli , v. khudik , c. siemon , and g. shvets , phys .
. lett . * 107 * , 145003 ( 2011 ) .
c.b.schroeder , c.benedetti , e.esarey , f.j.gruener , and w. p. leemans , phys .
lett . * 107 * , 145002 ( 2011 ) .
p. muggli , a. caldwell , o. reimann , e. oz , r. tarkeshian , c. bracco , e. gschwendtner , a. pardons , k. lotov , a. pukhov , m. wing , s. mandry , j. vieira , physics of the awake project .
proceedings of ipac2013 ( shanghai , china ) , p.1179 - 1181 .
awake collaboration , awake design report : a proton - driven plasma wakefield acceleration experiment at cern .
cern - spsc-2013 - 013 ; spsc - tdr-003 .
c.bracco , e.gschwendtner , a.petrenko , h.timko , t.argyropoulos , h.bartosik , t.bohl , j.e.mueller , b.goddard , m.meddahi , a.pardons , e.shaposhnikova , f.m.velotti , h.vincke , nucl . instr . meth .
a * 740 * , 48 ( 2014 ) .
e.oz , p.muggli , nucl .
instr . meth .
a * 740 * , 197 ( 2014 ) .
k.v.lotov , g.z.lotova , v.i.lotov , a.upadhyay , t.tuckmantel , a.pukhov , a.caldwell , phys . rev .
beams 16 , 041301 ( 2013 ) .
a. caldwell and k. v. lotov , phys .
plasmas * 18 * , 103101 ( 2011 ) .
j. beringer et al .
( particle data group ) , phys .
d * 86 * , 010001 ( 2012 ) .
k.v.lotov , phys .
plasmas * 5 * , 785 ( 1998 ) .
k.v.lotov , phys . rev .
beams * 6 * , 061301 ( 2003 ) .
lotov , a. sosedkin , e.mesyats , simulation of self - modulating particle beams in plasma wakefield accelerators .
proceedings of ipac2013 ( shanghai , china ) , p.1238 - 1240 . `
www.inp.nsk.su/~lotov/lcode ` .
r.keinigs and m.e.jones phys .
fluids * 30 * , 252 ( 1987 ) .
k.v.lotov , phys .
plasmas 20 , 083119 ( 2013 ) .
k.v.lotov , phys .
plasmas 18(2 ) ( 2011 ) 024501 .
j.vieira , r.a.fonseca , w.b.mori , and l.o.silva , phys .
lett . * 109 * , 145005 ( 2012 ) .
j. vieira , y. fang , w. b. mori , l. o. silva , and p. muggli , phys .
plasmas * 19 * , 063105 ( 2012 ) .
k.v.lotov , a.pukhov , and a.caldwell , phys .
plasmas 20(1 ) , 013102 ( 2013 ) .
k.v.lotov , j. plasma phys .
78(4 ) , 455 - 459 ( 2012 ) .
|
the dependence of wakefield amplitude and phase on beam and plasma parameters is studied in the parameter area of interest for self - modulating proton beam - driven plasma wakefield acceleration .
the wakefield sensitivity to small parameter variations reveals the expected level of shot - to - shot jitter of experimental results .
of all the parameters , the plasma density stands out , as the wakefield phase is extremely sensitive to this parameter .
the study of large variations determines the effects that limit the achievable accelerating field in different parts of the parameter space : nonlinear elongation of the wakefield period , insufficient charge of the drive beam , emittance - driven beam divergence , and motion of plasma ions .
|
symbiotic stars are characterized by peculiar spectra where molecular absorption bands - the characteristic features of a cool giant - are present together with emission lines corresponding to a high excitation level .
the red and infrared spectra of symbiotic stars are typical of cool giants while in the uv range they are characterized by a very hot continuum .
it is widely assumed that most symbiotic stars are detached binaries consisting of a cool giant and a white dwarf surrounded by a nebulosity [ [ boyar67 ] ] .
mass exchange in these systems is driven by stellar wind .
the goal of this work is to carry out a study of classical symbiotic systems using as example one of the most well - studied representatives of this class - z and . according to the analyses of its energy distribution in a wide spectral range [ [ fc88 ] ] , the components of z and have the following characteristics : the cool m3.5iii giant has a mass @xmath3 and a radius @xmath4 100@xmath5 , while the white dwarf has a mass @xmath6 , a radius @xmath4 0.07@xmath5 and a temperature @xmath7 k. the giant loses mass at the rate of @xmath8/yr .
gas in the circumbinary envelope has an electron density of @xmath9 and a temperature @xmath10 k. the separation of the system is 482@xmath5 , the orbital period is 758 days .
more than 100 years of observations have shown that this star presents outbursts .
the last of them took place in 2000 and has been actively observed at different wavelengths .
following the characteristics of their outbursts , symbiotic stars can be divided in two types [ [ pr80 ] , [ mk92 ] ] .
the first type includes so - called classical symbiotics ( e.g. z and , ag peg , ag dra , ci cyg , ax per ) .
their outbursts typically last a few months and have amplitudes of 2 - 4@xmath11 .
the typical feature of such outbursts is the weakening of high excitation lines while the brightness increases .
the detailed description of the behaviour of these stars is given in [ [ boyar69 ] , [ kenyon86 ] ] .
symbiotic novae ( e.g. v1016 cyg , v1329 cyg , rs oph , hm sge ) show the second type of outbursts .
these outbursts are longer and brighter .
their energetics can even exceed that of novae outbursts .
the principal difference of these outbursts from those of the first type is the increase of the ionization degree while the brightness increases .
the behaviour of these stars during outbursts is described in [ [ baratta ] , [ mammano ] , [ ciatti],[puetter ] ] .
the second type of outbursts is similar to slow classical novae outbursts [ [ allen ] , [ kenyon83 ] ] ( for differences between symbiotic novae and slow classical novae outbursts see [ [ ibentu96 ] ] ) .
the most probable mechanism describing the observational manifestations of classical symbiotics as well as of symbiotic novae is thermonuclear burning on the accretor s surface [ see , e.g. [ kenyon86 ] , [ ibentu96 ] ] .
it has been shown in various works [ [ tu ] , [ zytkow ] , [ pr80 ] , [ iben ] , [ ibentu89 ] ] that the process of thermonuclear burning of hydrogen on the white dwarf surface strongly depends on the accretion rate @xmath12 [ [ pr80 ] ] . only , in a narrow range of @xmath12 is the steady hydrogen burning possible [ [ zytkow ] , [ sion ] , [ fuji ] ]
. the lower limit of this range is given by the expression [ [ sienk],[iben ] ] @xmath13 while the higher limit is given by [ [ pacz ] , [ uus ] ] @xmath14 for z and where @xmath15 , steady burning is possible in the range @xmath16 .
if the accretion rate is below this interval , thermonuclear burning exhausts the matter faster than accretion can replenish it bringing the burning to a halt . following this ,
hydrogen begins to accumulate on the white dwarf surface until the pressure near the hydrogen envelope base reaches a critical value and a hydrogen shell flash occurs .
the bolometric luminosity increases by a factor 10100 times in the course of @xmath4 1 year and stays at the active state for @xmath4 10 years .
the total duration of the outburst is dozens of years that corresponds to the nuclear time scale .
this is observed in symbiotic novae .
if for some reason @xmath12 becomes larger than @xmath17 , the accreted matter accumulates above the burning shell and expands to giant dimensions .
such behaviour is typical for classical symbiotic stars .
though the bolometric luminosity stays constant , the visual brightness increases by 1 - 2 magnitudes and the effective temperature decreases .
optical outburst develops on the thermal timescale . in accordance with existing models of thermonuclear burning on a white dwarf surface [ [ sienk],[zytkow ] ] , small variations of the accretion rate
can lead to significant changes of the temperature at practically constant bolometric luminosity . for a white dwarf of @xmath18 mass , an envelope mass change of @xmath19 can result in a 3@xmath11 luminosity increase [ [ pr80 ] ] .
this value corresponds to the typical amplitude of z and outbursts .
in order to study the gas flow structure in the symbiotic system z and 3d numerical simulations have been carried out .
the zero point of the coordinate system was placed at the centre of the accretor , the @xmath21- axis was directed along the line connecting centres of the components and oppositely to the mass - losing star , the @xmath22-axis along the accretor orbital motion , and the @xmath23-axis along the axis of rotation of the binary system .
the flow was described by the system of euler equations in corotating coordinate frame : @xmath24 @xmath25 @xmath26 @xmath27 @xmath28 here @xmath29 is the velocity vector , @xmath30 the pressure , @xmath31 the density , @xmath32 the specific total enthalpy , @xmath33 the specific total energy , @xmath34 the specific internal energy , @xmath35 the angular velocity of binary system s rotation , and @xmath36 the force potential . in the standard definition
when only the gravitational forces from the point mass components and the centrifugal force are taken into account , the force potential is given by : @xmath37 here @xmath38 is the mass of the accretor , @xmath39 the mass - losing star s mass , @xmath40 , @xmath41 the radius - vectors of the centres of components , @xmath42 the radius - vector of the centre of mass of the system .
this is the so - called roche potential .
but in our case the additional force responsible for the donor s wind acceleration should also be taken into account .
therefore , the form of the potential changes .
previous studies ( e.g. , [ [ gs][dima96 ] ] ) have shown that the general flow structure in the system where components do not fill their roche lobes is defined first and foremost by the stellar wind parameters . since the mechanism of gas acceleration is poorly known for cool giants , we mimic the radiation pressure by reducing the gravitational attraction force to the donor star by a factor ( @xmath43 ) .
so , the modified force potential looks the following : @xmath44 to close the system the perfect gas equation of state was used .
@xmath45 the value of adiabatic index has been accepted @xmath46 , that corresponds to the case close to the isothermal one [ [ gamma1][gamma3 ] ] .
we adopted for the parameters of the binary system those of z and . in our numerical simulations
all variables were put in dimensionless form as follows : the length was normalized by the separation of two stars , @xmath47 , and the time by @xmath48 , and so the velocity was normalized by @xmath49 .
the density was normalized by that at the surface of the mass - losing star .
the method of computational fluid dynamics is described in [ [ nagae ] ] and details are explained in [ [ makita ] , [ fujiwara ] ] .
we use the simplified flux vector splitting ( sfs ) finite volume scheme to discretize the euler equation ( the description of the sfs scheme is given in the appendix of [ [ makita ] ] ) .
we assume a symmetry about the orbital plane , and therefore only compute the upper half - space of the computational domain , which is @xmath50 , @xmath51 , @xmath52 .
the region is divided into @xmath53 cells .
the inner numerical boundary surrounding the mass - losing star is represented by an equipotential surface with a mean radius @xmath54 ( we set the numerical boundary at a little larger radius , @xmath55 , rather than @xmath56 ) .
the radius of the mass - accreting star is too small to fit in our numerical grid so it is represented by one cell . the cells just inside of the inner numerical boundary about the mass - losing star are assumed to be filled with gas with dimensionless density @xmath57 , dimensionless sound speed @xmath58 ( corresponding to @xmath59 ) , and a dimensionless normal speed to the surface , @xmath60 .
the real normal velocity of the gas on the numerical boundary is determined by solving riemann problems between the cells adjacent to the inner boundary surface .
the cell representing the mass - accreting star and the cells just outside of the outer numerical boundary are filled by a gas with density @xmath61 , pressure @xmath62 , and three velocity components @xmath63 .
this assumption does not mean that accretion onto the mass accreting star nor that escape from the computational domain do not occur .
the velocity of the gas on the boundaries are computed by solving riemann problems . at the initial stage @xmath64 , the computational domain except the region inside of the inner boundaries
is assumed to be filled by a gas with @xmath61 , @xmath62 , and @xmath65 this gas is gradually replaced by the gas supplied from the mass - losing star .
we compute up to a few rotation periods that is generally long enough for the system to reach a quasi - steady state .
the velocity of the mass - losing star s wind in z and is @xmath425 km / s [ [ fc88 ] ] .
it is presumed in the framework of the considered mechanism that for slightly smaller velocities an accretion disc will form while for slightly greater values the disc will be destroyed and a wind accretion type of flow will occur . in order to check the applicability of the proposed mechanism ,
numerical simulations for velocities of 25@xmath66km / s have been carried out.5 km / s can be easily explained by an activity of the giant [ [ mk92 ] ] . ]
calculations for the wind velocity @xmath67 km / s were necessary for studying the possibility of disc formation in the system , accretion rate estimation and to check if the accretion rate is indeed within the limits of the steady burning range .
calculations for @xmath68 km / s were conducted in order to obtain parameters of the flow structure in the system without disc . to answer the questions concerning processes taking place after the wind velocity change if the disc will be destroyed , how the accretion rate will grow , if the accretion rate change will go above the upper limit of the steady burning range calculations with a wind velocity jump from @xmath69 km / s have been carried out the results of our numerical modelling have shown that for wind velocity @xmath67 km / s and values of the parameter @xmath70 $ ] a steady accretion disc forms in the system . in figures 1,2 the flow structure for the case @xmath67 km / s and @xmath71=0.94 is shown .
density contours and velocity vectors in the equatorial plane of the system are presented for the whole computational domain ( fig . 1 ) and for the near - accretor area @xmath72\times[-a\ldots 0.5a]$ ] ( fig .
the results of the calculations are presented at the time @xmath73 when a steady regime was already established in the system .
these results show that for this wind velocity value a bow shock and an accretion disc of approximately @xmath74 form in the system .
note , that two spiral density waves are seen in the accretion disc . in this solution
the accretion rate is @xmath0 of matter lost by the donor .
similar values of the accretion efficiency for solution with accretion disc were obtained by mastrodemos and morris [ [ mm98 ] ] for detached binaries . for z and where the mass loss rate is estimated to be @xmath75 @xmath2/yr , the accretion rate will be @xmath1 @xmath2/yr , i.e. corresponding to the steady burning range ( near the upper limit ) .
as one can see , due to the influence of the pressure gradient and the presence of spiral shocks , the disc structure is far from a conventional keplerian disc . since we used the full set of euler equations that incorporate advection term in energy equation ,
the obtained flow structure is more similar to advection - dominated discs as discussed by walder and folini [ [ wf99 ] ] .
of course , the system of euler equations with adiabatic energy equation does not incorporate viscous heating nor radiative cooling , but the numerical viscosity , on the one hand , and the choice of the adiabatic index @xmath76 close to 1 , on the other hand , provide an implicit account of these processes .
thus , we believe that our mathematical model is adequate to the physics of accretion discs .
the solution with wind velocity @xmath68 km / s is presented in figures 3 and 4 . just as in figures 1,2 , here density contours and velocity vectors are presented for all the computational domain ( fig .
3 ) and for the area near the accretor @xmath77\times[-a\ldots 0.5a]$ ] ( fig . 4 ) .
the situation presented corresponds to the moment when the steady regime has already become established .
the analysis of these results show that in the case when wind velocity equals 30 km / s the cone shock is close to the accretor and does not leave any room for the formation of a disc . in this solution
a wind accretion type of flow instead of disc accretion takes place with the accretion rate being @xmath78 of the matter leaving donor s surface .
similar flow structures were obtained in 3d simulations performed by dumm et al . [ [ dumm ] ] . according to their results , the accretion efficiency equals 6% for the case of wind accretion . the analysis of results presented above allows to conclude that even small change in the donor s wind velocity ( within @xmath66 km / s around the observed value 25 km / s ) leads to a flow structure and accretion regime changeover , namely , to the transition from a disc to a wind accretion flow . in order to consider the process of transition between these two regimes after the wind velocity increase
, we took the stationary solution for the case 20 km / s after @xmath79 from the beginning of calculations ( situation presented in figures 1a,1b ) and then raised the velocity up to 30 km / s .
after the increase of the wind velocity on the inner boundary the flow structure changes , evolving from a state with an accretion disc to the one with a cone shock presented in figures 2a , 2b .
let us consider the process of the flow restructuring .
after @xmath80 ( @xmath4133 days ) since the wind velocity change , the matter moving with increased velocity reaches the vicinity of the accretor , namely , the bow shock located in front of it . then the wind continues to move further crushing the accretion disc and making the matter of the disc fall on the accretor s surface .
a snapshot of this flow rearrangement is presented in figure 5 . the area size and
all the designations are the same as in figures 2,4 .
this situation corresponds to @xmath4180 days after the wind velocity increase .
it should be mentioned that the study of the transition period is limited in the framework of this model .
after the accretion rate jump the accepted boundary conditions on the accretor change and the model used does nt describe the real physical situation any more .
correspondingly , the presented results of calculations of the flow rearrangement period are correct only at first stages .
the behaviour of the accretion rate with time is shown in figures 4a , b . here
the time passed from the beginning of the calculation in units of @xmath81 is shown on the x axis .
the accretion rate is given in dimensionless units . from the moment when the matter with increased velocity reaches the vicinity of the accreting component , the accretion rate begins to grow and reaches its maximum after approximately @xmath82 ( @xmath447 days ) .
the maximum value of the accretion rate is 2 - 2.2 times as compared to the state with disc accretion . as shown above , for accretor of 0.6@xmath2 mass , the ratio of the upper limit of the steady burning range to its lower limit @xmath83 .
for a white dwarf of 0.55@xmath2 mass ( the value given in [ [ mk92 ] ] ) an increase of 10% only should be enough to exceed the upper limit .
so , the approximately twofold accretion rate growth obtained in our calculations is large enough to transfer the system from the quiescent to active state .
the analysis of the data presented in figures 4a , b shows that the time of full disc destruction is @xmath84 days .
the time during which the accretion rate exceeds the upper limit of the steady burning range should correspond to the time of outburst development .
if we suppose that the exceedance of the limit of the steady burning interval occurs when the accretion rate increases by 1.5 times , the time during which accretion rate is above this value will be approximately 100 days and @xmath85 will be accreted during this time .
outbursts in classical symbiotic stars occur when the accretion rate becomes greater than the upper limit of the steady burning range . for transition from quiescent to active state
it is necessary to provide an increase in the accretion rate on a rather long time interval corresponding to the characteristic time of the outburst development ( @xmath4100 days ) .
earlier , in [ [ bisikalo2002 ] ] , a mechanism providing the required accretion rate increase in the system has been proposed on the basis of 2d calculations . according to this mechanism even minor change of
the donor s wind velocity is enough for the accretion regime to change . during the transition from disc to wind accretion
the accretion disc is destroyed and the wind with increased velocity makes the matter of the disc fall onto the accretor s surface .
the analysis of 2d calculations has shown that during this process the accretion rate growth is enough for system to overstep the limits of steady burning , therefore , leading to the development of an outburst . in order to answer
if this mechanism can work in observed astrophysical objects , the gasdynamic modelling of the flow structure in the classical symbiotic system z and using more realistic 3d model has been carried out .
the results of our calculations allowed us to draw following conclusions : 1 .
it was found that for the donor wind velocity 20 km / s and @xmath86 the accretion disc forms in the system .
the accretion rate for the solution with disc is @xmath0 of the matter lost by the donor .
if the wind velocity equals 30 km / s the accretion disc disappears and a cone shock forms .
the accretion rate for this case is @xmath78 of the matter that left the donor .
these two solutions show that even minor change in the donor s wind velocity ( within the limits of @xmath66km / s from the observed value 25 km / s [ [ fc88 ] ] ) leads to the change in flow structure and to the accretion regime change , namely , to the transition from the disc accretion to the wind accretion .
4 . in accordance with the proposed mechanism , the solution with the accretion disc should correspond to a quiescent state of the system . according to our results ,
the accretion rate in this case is @xmath0 .
so , for z and , where the mass loss rate is estimated to be @xmath87/yr , the accretion rate will be @xmath88/yr , in good agreement with the steady burning interval for the most probable white dwarf masses . 5 .
in the solution where the wind velocity increases from 20 to 30 km / s the accretion disc is destroyed and the matter it contained falls onto the accretor s surface . 6 .
the detailed study of the accretion regime change after the wind velocity increase has shown that the process of disc destruction is followed by an accretion rate jump .
the value of maximum accretion rate is @xmath42 - 2.2 times as much as the initial value .
as it has been mentioned above , the steady burning range is rather narrow and if the accretor s mass equals 0.6@xmath2 the increase of the accretion rate by 2.2 times is unambiguously enough for overstepping the upper limit of this range .
results of our calculations give an accretion rate increase large enough to put the system outside the steady burning range and to move it up into the active state .
the time during which the accretion rate is above the steady burning range should correspond to the outburst development time , that is approximately 100 days for z and . according to our results
the full time of disc destruction is approximately @xmath84 days .
if we assume that the system leaves the steady burning range after the accretion rate increases in 1.5 times , the corresponding time will be @xmath20 days , in good agreement with observations .
the typical amplitude of z and outburst is @xmath89 . for a white dwarf of 1.2@xmath2 mass such brightness increase
is provided by an envelope mass @xmath90@xmath2 [ [ pr80 ] ] . in our calculations
the envelope mass is smaller , @xmath91@xmath2 .
it is obvious , that for correct comparison of envelope masses the calculations of thermonuclear burning on the surface of a white dwarf with the mass equal to that of the accretor in z and are required .
moreover it is necessary to take into account the asymmetry of accretion .
unfortunately , we do not know any works where such estimations were made , so , the question on the envelope mass remains open .
given the accretion rate corresponding to the upper limit of the steady burning range and given the typical time of the outburst of @xmath92 year , the envelope mass should not exceed significantly @xmath91@xmath2 , in good agreement with our results .
the analysis of results presented here allows us to draw the main conclusion that the transition from quiescent to active state in symbiotic stars can be concerned with the accretion regime change ( transition from the disc accretion to wind accretion ) as the result of insignificant variations of donor s wind velocity .
this work was supported by the `` 21-st century coe program of origin and evolution of planetary systems '' of the ministry of education , culture , sports , science and technology ( mext ) of japan .
in particular , t.m . , d.v.b . and
h.b . acknowledge its financial support .
t.m . was supported by the grant in aid for scientific research of the japan society of promotion of science ( 13640241 ) .
calculations were performed on the nec sx-6 super - computer at the isas / jaxa japan .
the work was also supported by russian foundation for basic research ( projects 03 - 02 - 16622 , 03 - 01 - 00311 , 05 - 02 - 16123 , 05 - 02 - 17070 , 05 - 02 - 17874 ) , by science schools support program ( project n 162.2003.2 ) , by presidium ras programs `` mathematical modelling and intellectual systems '' , `` nonstationary phenomena in astronomy '' .
1 . [ boyar67 ]
a. a. boyarchuk , izv .
crao * 38 * , 155 ( 1967 ) .
2 . [ fc88 ] t. fernandez - castro , a. cassatella , a. gimenez , _ et al .
_ , astrophys
. j. * 324 * , 1016 ( 1988 ) .
3 . [ pr80 ] b. paczyski and b. rudak , astron . & astrophys . *
82 * , 349 ( 1980 ) .
4 . [ mk92 ] j. mikoajewska and s. j. kenyon , monthly notices roy .
. soc . * 256 * , 177 ( 1992 ) .
[ boyar69 ] a. a. boyarchuk , izv .
crao * 39 * , 124 ( 1969 ) . 6 .
[ kenyon86 ] s. j. kenyon , the symbiotic stars , cambridge university press , cambridge ( 1986 ) .
[ baratta ] g. b. baratta , a. cassatella , and r. viotti , astrophys .
j. * 187 * , 651 ( 1974 ) .
[ mammano ] a. mammano and f. ciatti , astron . & astrophys . * 39 * , 405 ( 1975 )
[ ciatti ] f. ciatti , a. mammano , and a. vittone , astron . & astrophys . *
68 * , 251 ( 1978 ) . 10 .
[ puetter ] r. c. puetter , r. w. russell , b. t. soifer , and s. p. willner , astrophys .
j. * 223 * , l93 ( 1978 ) . 11 .
[ allen ] d. a. allen , monthly notices royal astron .
soc . * 192 * , 521 ( 1980 ) . 12 .
[ kenyon83 ] s. j. kenyon and j. w. truran , astrophys .
j. * 273 * , 280 ( 1983 ) . 13 .
[ ibentu96 ] i. iben , jr . and a. v. tutukov , astrophys
* 105 * , 145 ( 1996 )
[ tu ] a. v. tutukov and l. r. yungelson , astrofiz .
* 12 * , 521 ( 1976 )
[ zytkow ] b. paczyski and a. ytkow , astrophys .
j. * 222 * , 604 ( 1978 ) . 16 .
[ iben ] i. iben , jr .
, astrophys .
j. , * 259 * , 244 ( 1982 ) . 17 .
[ ibentu89 ] i. iben , jr . and a. v. tutukov , astrophys .
j. * 342*,430 ( 1989 ) .
[ sion ] e. m. sion , m. j. acierno , and s. tomczyk , astrophys .
j. * 230 * , 832 ( 1979 ) . 19 .
[ fuji ] m. y. fujimoto , astrophys .
j. * 257 * , 767 ( 1982 ) . 20 .
[ sienk ] r. sienkiewicz , w. dziembowski , 1977 , iau coll . * 42 * , 327 ( 1977 ) . 21 .
[ bisikalo2002 ] d. v. bisikalo , a. a. boyarchuk , e. yu . kilpio , and o. a. kuznetsov , astron .
reports * 46 * , 1022 ( 2002 ) .
[ pacz ] b. paczyski , acta astron . * 20 * , 47 ( 1970 ) . 23 .
[ uus ] u. s , nauchn . inf .
* 17 * , 32 ( 1970 ) . 24 .
[ gs ] g. s. bisnovatyi - kogan , ya .
m. kazhdan , a. a. klypin , a. e. lutskii , sov . astron . * 233 * , 201 , ( 1979 ) . 25 .
[ dima94 ] d. v. bisikalo , a. a. boyarchuk , o. a. kuznetsov _ et al .
_ , astron .
reports , * 38 * , 494 ( 1994 ) 26 .
[ dima96 ] d. v. bisikalo , a. a. boyarchuk , o. a. kuznetsov , and v.m.chechetkin , astron .
reports , * 40 * , 662 ( 1996 ) 27 . [ gamma1 ]
k. sawada , t. matsuda , and hachisu , monthly notices roy .
* 219 * , 75 ( 1986 ) . 28 .
[ gamma2 ] d. molteni , g. belvedere , and g. lanzafame , monthly notices roy .
. soc . * 249 * , 748 ( 1991 ) . 29 .
[ gamma3 ] d. v. bisikalo , a. a. boyarchuk , o. a. kuznetsov _ et al .
_ , astron .
reports , * 39 * , 325 ( 1995 ) 30 .
[ nagae ] t. nagae , k. oka , t. matsuda _ et al .
_ , astron . & astrophys .
* 419 * , 335 ( 2004 ) . 31 .
[ makita ] m. makita , k. miyawaki , t. matsuda , monthly notices roy .
. soc . * 316 * , 906 ( 2000 ) . 32 .
[ fujiwara ] h. fujiwara , m. makita , t. matsuda , progress of theoretical physics * 106 * , 729 ( 2001 ) .
[ mm98 ] n. mastrodemos , m. morris , astrophys .
j. * 497 * , 303 ( 1998 ) .
[ wf99 ] r. walder , d. folini , thermal and ionization aspects of flows from hot stars , edited by henny lamers and arved sapar , asp conference series , * 204 * , 331 ( 2000 ) 35 .
[ dumm ] t. dumm , d. folini , h. nussbaumer _ et al .
_ , astron . & astrophys .
* 354 * , 1014 ( 2000 ) . *
fig . 1 . * density contours and velocity vectors in the equatorial plane of the system for the case @xmath67 km / s and @xmath71=0.94 .
the empty circle centered at ( -1,0 ) corresponds to the donor ( radius of the circle equals donor s radius ) , the diamond at ( 0,0 ) point marks the accretor .
|
the results of 3d modelling of the flow structure in the classical symbiotic system z andromedae are presented .
outbursts in systems of this type occur when the accretion rate exceeds the upper limit of the steady burning range .
therefore , in order to realize the transition from a quiescent to an active state it is necessary to find a mechanism able to sufficiently increase the accretion rate on a time scale typical to the duration of outburst development .
our calculations have confirmed the transition mechanism from quiescence to outburst in classic symbiotic systems suggested earlier on the basis of 2d calculations ( bisikalo et al , 2002 ) .
the analysis of our results have shown that for wind velocity of 20 km / s an accretion disc forms in the system .
the accretion rate for the solution with the disc is @xmath0 of the mass loss rate of the donor , that is , @xmath1 @xmath2/yr for z and .
this value is in agreement with the steady burning range for white dwarf masses typically accepted for this system .
when the wind velocity increases from 20 to 30 km / s the accretion disc is destroyed and the matter of the disc falls onto the accretor s surface .
this process is followed by an approximately twofold accretion rate jump .
the resulting accretion rate growth is sufficient for passing the upper limit of the steady burning range , thereby bringing the system into an active state .
the time during which the accretion rate is above the steady burning value is in a very good agreement with observations .
the analysis of the results presented here allows us to conclude that small variations in the donor s wind velocity can lead to the transition from the disc accretion to the wind accretion and , as a consequence , to the transition from quiescent to active state in classic symbiotic stars .
|
throughout this paper , @xmath3 is a field or the ring of integers @xmath4 .
@xmath5={\mathbf{k}}[v_1,\dots , v_m]$ ] is the graded polynomial algebra on @xmath6 variables , deg@xmath7 .
the _ face ring _
( also known as the _ stanley - reisner ring _ ) of a simplicial complex @xmath0 with @xmath6 vertices is the quotient ring @xmath8/\mathcal { i}_k\ ] ] where @xmath9 is the ideal generated by those square free monomials @xmath10 for which @xmath11 is not a simplex in @xmath0 . for any simple polytope @xmath12 , davis and januszkiewicz @xcite introduced a @xmath13-manifold @xmath14 with orbit space @xmath12 .
after that buchstaber and panov @xcite generalized this definition to any simplicial complex @xmath0 with vertex set @xmath15=\{1,2,\dots , m\}$ ] , and named it the _ moment - angle complex _ associated to @xmath0 : @xmath16 where @xmath17 , @xmath18 if @xmath19 and @xmath20 if @xmath21 .
the following theorem is proved by buchstaber and panov @xcite for the case over a field by using eilenberg - moore spectral sequence , @xcite for the general case .
let k be a simplicial complex with @xmath6 vertices .
then the following isomorphism of algebras holds : @xmath22}({\mathbf{k}}(k),{\mathbf{k}}).\ ] ] since @xmath23}({\mathbf{k}}(k),{\mathbf{k}})$ ] has a natural @xmath24-bigrade .
so the bigraded cohomology ring can be decomposed as follows : @xmath25}({\mathbf{k}}(k),{\mathbf{k}})=\bigoplus\limits_{i\geq0}\bigoplus\limits_{j\subseteq[m ] } \mathrm{tor}_{i , j}^{{\mathbf{k}}[m]}({\mathbf{k}}(k),{\mathbf{k}}).\ ] ] hochster @xcite gave a combinatorial description of the tor - groups @xmath1}({\mathbf{k}}(k),{\mathbf{k}})$ ] .
@xmath26}({\mathbf{k}}(k),{\mathbf{k}})\cong\w{h}^ { |j|-i-1}(k_j;{\mathbf{k}}),\ ] ] where @xmath27 , @xmath28 is the full subcomplex of @xmath0 corresponding to @xmath29 .
hochster proved this for the case @xmath3 is a field by analyzing the betti number of both sides .
buchstaber and panov ( * ? ? ?
* theorem 3.2.9 ) generalize this result to @xmath30 by means of the koszul resolution @xmath31\otimes{\mathbf{k}}[m]$ ] of @xmath3 .
recently , wang and zheng @xcite gave another way to calculate @xmath23}({\mathbf{k}}(k),{\mathbf{k}})$ ] by using taylor resolution on the stanley - reisner ring @xmath2 .
this method was presented firstly by yuzvinsky in @xcite . in this paper
we give a combinatorial description of this method , and then we prove hochster s theorem in a different way .
let @xmath0 be a simplicial complex on the set @xmath15 $ ] .
a _ missing face _ of @xmath0 is the subset @xmath32 $ ] where @xmath33 and every proper subset of @xmath34 is a simplex of @xmath0 .
denote by @xmath35 the missing face set of @xmath0 .
clearly @xmath35 is uniquely determined by @xmath0 . a _ simplicial complement _
@xmath36 of @xmath0 is a subset of @xmath37\mid\tau\not\in k\}$ ] so that @xmath38 . usually @xmath36 is not unique .
denote by @xmath39 the set of simplicial complements of @xmath0 .
given a simplicial complement @xmath36 of @xmath0 , one can define an exterior algebra @xmath40 $ ] generated by all elements of @xmath36 . for a monomial @xmath41 $ ] , let @xmath42 be the _ total set _ of @xmath43 .
so @xmath40 $ ] has a natural @xmath24-bigrade , which means @xmath44=\bigoplus\limits_{i\geq 0}\bigoplus\limits_{j\subseteq[m]}\lambda^{i , j}[\mathbb{p}]\ ] ] where @xmath45 $ ] is generated by monomials @xmath43 satisfying @xmath46 and deg@xmath47 .
one can make @xmath40 $ ] into a chain complex by defining a differential @xmath48 on it .
the differential @xmath49\to \lambda^{n-1,*}[\mathbb{p}]$ ] is generated by @xmath50 where @xmath51 , @xmath52 ( particularly , if @xmath53 , @xmath54 ) , @xmath55 if @xmath56 and zero otherwise .
[ thm:1 ] let @xmath0 be a simplicial complex on the set @xmath15 $ ] .
given a simplicial complement @xmath57 .
then @xmath58}_{i,\,j}({\mathbf{k}}(k),{\mathbf{k}})\cong h_i\big ( \lambda^ { * , j}[\mathbb{p}],d\big).\ ] ] from theorem [ thm:1 ] , we know that the homology groups @xmath59,d\big)$ ] of a simplicial complement @xmath36 is not depend on the choice of @xmath36 .
it just depend on the simplicial complex @xmath0 .
if we can prove that @xmath60,d\big)\cong\w{h}^{|j|-i-1}(k_j;{\mathbf{k}}),\ ] ] then the hochster theorem is proved .
the following theorem ensure these isomorphisms hold .
[ thm:3 ] let @xmath0 be a simplicial complex on the set @xmath15 $ ] , and let @xmath36 be one of the simplicial compliments of @xmath0 .
then we have the following group isomorphisms : @xmath61}[\mathbb{p}],d\big)\cong\w{h}^{m - i-1}(k;{\mathbf{k}}).\ ] ] let everything be as before
. then @xmath62,d\big)\cong\w{h}^{|j|-i-1}(k_j;{\mathbf{k}}).\ ] ]
if @xmath0 is a simplex , the theorem is trivial . so in this paper
, we assume that @xmath0 is a simplicial complex on the set @xmath15 $ ] but not a simplex .
given a simplicial complement @xmath36 of @xmath0 , suppose @xmath63 for any @xmath64 , we have a simplicial complex @xmath65 clearly @xmath66 is a triangulation of @xmath67 .
we denote by @xmath68 , the interior of the geometric realization of @xmath66 .
if @xmath73 , there is a simplex @xmath74}\setminus k$ ] satisfying @xmath75 in the relative interior of @xmath34 . from the definition of a simplicial complement
, there exists a simplex @xmath76 , such that @xmath77 .
thus , @xmath78 . for any topological space @xmath79 , let @xmath80 be an open cover of the space @xmath79 indexed by a set @xmath81 .
we define the _ nerve _ @xmath82 to be the abstract simplicial complex on the set @xmath81 @xmath83 for a simplex @xmath84 , denote by @xmath85 .
[ thm:2 ] let @xmath0 be a simplicial complex on @xmath15 $ ] , @xmath91 be a simplicial complement of @xmath0 . by proposition 2.1
, @xmath71 forms an open cover of the topological space @xmath72
. then we have the following isomorphisms : @xmath92}[\mathbb{p}],d\big)\cong\widetilde{\check{h}}_{n-2}(u(k),{\mathbf{u}};{\mathbf{k}}):=\widetilde{h}_{n-2}({\mathcal{n}}({\mathbf{u}});{\mathbf{k}}).\ ] ] from definition 2.2 , we know that @xmath101}\mid u_{i_1}\cap u_{i_2}\cap\ldots\cap u_{i_n}\neq{\varnothing}\big\}.\ ] ] lemma [ lem:1 ] shows that @xmath102 if and onely if @xmath103 $ ] .
so the nerve can be written as @xmath101}\mid\tau_{i_1}\cup\tau_{i_2}\ldots\cup\tau_{i_n}\neq [ m]\big\}.\ ] ] let @xmath104 $ ] be the exterior algebra generated by @xmath91 .
we define another differential @xmath105\rightarrow \lambda^{n-1}[p_0]$ ] generated by @xmath106 where @xmath52 for any monomial @xmath51 .
define a homomorphism @xmath107,\ ] ] generated by @xmath108 $ ] , where @xmath109 is the reduced simplicial chain complex of @xmath82 .
obviously , @xmath110 is a monomorphism .
let @xmath111 .
then we get a short exact sequence of chain complexes , @xmath112,\,\partial\big ) \rightarrow\big(\lambda^{*+1}[\mathbb{p}]/\gamma,\,\partial)\rightarrow 0.\ ] ] apparently , @xmath113/\gamma$ ] is generated by all monomials @xmath114}[\mathbb{p}]$ ] ( i.e. , @xmath115 $ ] ) .
it is easy to see that there is a chain isomorphism @xmath116/\gamma,\,\partial \big)\cong\big(\lambda^{*+1,[m]}[\mathbb{p}],\,d\big),\ ] ] where @xmath117}[\mathbb{p}],\,d\big)$ ] is as in theorem [ thm:1 ] .
it is easy to see that @xmath118,\,\partial \big)=0 $ ] .
thus from the long exact sequence induced by the short exact sequence above , we get that @xmath92}[\mathbb{p}],\,d\big)\cong\w{h}_{n-2}({\mathcal{n}}({\mathbf{u}});{\mathbf{k}}).\ ] ] note first that @xmath119 , and @xmath120 for each @xmath121 , @xmath122 .
we combine the results of theorem [ thm:4 ] , theorem [ thm:2 ] and lemma [ lem:1 ] to get that @xmath92}[\mathbb{p}],\,d\big)\cong\widetilde h_{n-2}(u(k);{\mathbf{k}}).\ ] ] from alexander duality theorem , we have @xmath123 m. hochster .
cohen macaulay rings , combinatorics , and simplicial complexes . in _ in : ring theory ,
ii , proc . second conf .
oklahoma , norman , okla .
, 1975 _ , volume 26 of _ in : lect .
notes pure appl
_ , pages 171233 .
dekker , new york , 1977 .
|
in this paper , we consider homology groups induced by the exterior algebra generated by a simplicial compliment of a simplicial complex @xmath0 .
these homology groups are isomorphic to the tor - groups @xmath1}({\mathbf{k}}(k),{\mathbf{k}})$ ] of the face ring @xmath2 , which is very useful and much studied in toric topology . by using ech homology theory and alexander duality theorem , we prove that these homology groups have dualities with the simplicial cohomology groups of the full subcomplexes of @xmath0 .
then we give a new proof of hochster s theorem .
|
the @xmath0 superconformal algebra ( sca ) is a basic tool in the world - sheet analysis of string compactifications on calabi yau ( cy ) manifold with complex dimension-@xmath1 .
it is well - known that in @xmath0 sca there exist two types of representations : bps ( massless ) and non - bps ( massive ) representations .
bps representations appear when the conformal weight @xmath7 of their highest - weight state hits the unitarity bound @xmath8 ( in the ramond sector ) where @xmath9 denotes the central charge of sca . on the other hand
non - bps representations appear at @xmath10 .
we study the decomposition of the elliptic genus for calabi
yau manifold @xmath11 in terms of characters of these representations ; @xmath12 \\ + \sum_{n=1}^\infty \sum_{q } p_{d , q}(n ) \
, \left [ \text{non - bps representations at $ h = n + \frac{d}{8}$ with $ u(1)$ charge-$q$ } \right ] .\end{gathered}\ ] ] since there exists only a finite number ( of order @xmath1 ) of bps representations , the set of their multiplicities @xmath13 is finite .
on the other hand , we have an infinite series of multiplicities @xmath14 for non - bps representations and we define the intrinsic cy entropy @xmath15 by the rate of its exponential growth @xmath16 such an analysis has been done for the case of hyperkhler manifolds @xcite based on earlier works on the representation theory of the @xmath17 superconformal algebra @xcite .
it was pointed out that the predicted entropy of hyperkhler manifolds coincides with that of the standard @xmath18-@xmath19 black hole @xcite when we consider the symmetric product of @xmath20 surfaces .
see ref . for a similar idea .
we also note that the expansion of elliptic genera in terms of theta functions has been discussed in ref . .
a key in our analysis is the fact that the characters of the bps representations are the mock theta functions which were first introduced by ramanujan ( see , _ e.g. _ , refs . ) .
we rely on recent developments @xcite on the understanding of the mock theta function ( see refs . for review ) : namely , the mock theta function is a holomorphic part of the harmonic maass form , and that it has a vector - valued modular form with weight-@xmath21 as its `` shadow '' .
this paper is organized as follows . in section [ sec : sca ] we review the characters of the @xmath0 superconformal algebra and the elliptic genera of the cy manifolds .
we make an extensive use of jacobi forms , whose properties are collected in appendix .
decompositions of the elliptic genera for the cy manifold are studied in sections [ sec : cy_odd ] and [ sec : cy_even ] .
we treat the odd - dimensional cy manifolds in section [ sec : cy_odd ] .
we notice the close relationship between the characters of @xmath0 and @xmath17 sca in this case and gritsenko s result on the space of jacobi forms .
it turns out that the entropy of cy @xmath1-folds coincides with the entropy of the corresponding hyperkhler manifolds in @xmath2-dimensions .
in particular the entropy of cy 3-folds vanishes identically . in section
[ sec : cy_even ] we discuss the even - dimensional cy manifolds .
adopting the same strategy as in our previous paper @xcite , we compute the entropy by use of the poincar maass series and obtain the result @xmath22 the last section is devoted to concluding remarks .
the @xmath0 superconformal algebra with the central charge @xmath23 is a fundamental tool studying the compactification of string theory on calabi - yau manifold @xmath11 with complex dimension @xmath1 .
it is well - known that there exists an isomorphism @xcite of the algebra , @xmath24 j_n & \to j_n + \frac{c}{3 } \ ,
\alpha \ , \delta_{n,0 } , \\[2 mm ] g_r^{\pm } & \to g_{r\pm \alpha}^\pm , \end{aligned}\ ] ] where @xmath25 .
ramond and ns sectors are exchanged when @xmath26 . for our purpose of studying string compactification , we need the extended @xmath0 sca which is invariant under the integral spectral flow @xmath27 .
odake discussed such series of extended sca s by adding spectral flow generators to the standard @xmath0 sca @xcite .
the highest weight state in the extended @xmath0 sca are labeled by conformal weight @xmath7 and @xmath28 charge @xmath29 ; @xmath30 j_0 \left| \omega \right\rangle = q \left| \omega \right\rangle .
\end{gathered}\ ] ] in the ramond sector , due to a relation between zero - modes of the supercurrents , @xmath31 , we have the unitarity condition @xmath32 characters are defined by @xmath33 where @xmath34 denotes the spin structure , and @xmath35 is the hilbert space of the representation . under the spectral flow , the characters in the ramond / ns sectors
are transformed to each other as ; @xmath36 \operatorname{ch}_\bullet^{r}(z;\tau ) = q^{\frac{d}{8 } } \ , { \mathrm{e}}^{d \pi { \mathrm{i}}z } \ , \operatorname{ch}_\bullet^{ns } \left ( z + \frac{\tau}{2 } ; \tau \right ) , \\[2 mm ] \operatorname{ch}_\bullet^{\widetilde{r}}(z;\tau ) = { \mathrm{e}}^ { \frac{d}{2 } \pi { \mathrm{i } } } \ , \operatorname{ch}_\bullet^{r } \left ( z + \frac{1}{2 } ; \tau \right ) . \end{gathered}\ ] ] here the phase factor in the @xmath37-sector is our convention . by construction ,
the superconformal character is given by the irreducible characters of @xmath0 sca @xcite summed over spectral flow .
this fact can be directly checked in the work of refs . .
in the ramond sector , characters are given explicitly as follows . here
the @xmath28 charge takes values @xmath38 * massive ( non - bps ) representations : + @xmath39 ; @xmath40 and
@xmath41 , @xmath42 ^ 3 } \ , { \mathrm{e}}^{2 \pi { \mathrm{i}}\left ( q - \frac{1}{2 } \right ) z }
\\ \times \sum_{n \in \mathbb{z } } q^{\frac{d-1}{2 } n^2 + \left ( q - \frac{1}{2 } \right ) n } \ , \left ( - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \right)^{(d-1 ) n } , \end{gathered}\ ] ] * massless ( bps ) representations : + @xmath43 ; @xmath44 , @xmath45 ^ 3 } \ , { \mathrm{e}}^{2 \pi { \mathrm{i}}\left ( q + \frac{1}{2 } \right ) z } \\ \times
\sum_{n \in \mathbb{z } } q^{\frac{d-1}{2 } n^2 + \left ( q+\frac{1}{2 } \right ) n } \ , \frac { \left ( - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \right)^{(d-1 ) n } } { 1 - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \ , q^n } , \end{gathered}\ ] ] and for @xmath43 ; @xmath46 @xmath47 ^ 3 } \ , { \mathrm{e}}^{2 \pi { \mathrm{i}}\frac{d+1}{2 } z } \\ \times \sum_{n \in \mathbb{z } } q^{\frac{d-1}{2 } n^2 + \frac{d+1}{2 } n } \ , \frac { \left ( 1-q \right ) \ , \left ( - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \right)^{(d-1 ) n } } { \left ( 1 - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \ , q^n \right ) \ , \left ( 1 - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \ ,
q^{n+1 } \right ) } .
\end{gathered}\ ] ] see appendix [ sec : theta ] for the notation of theta functions .
the characters for @xmath48 are given by @xmath49 the witten index of massless representations are given by @xmath50 1 + ( -1)^d , & \text{for $ q=\frac{d}{2}$ , } \end{cases}\ ] ] while all massive representations have a vanishing index .
at the unitarity bound @xmath43 , a massive character decomposes into a sum of massless characters as @xmath51 where @xmath52 , and @xmath53 the elliptic genus of the cy manifold with complex dimension @xmath1 is identified with @xcite @xmath54 , \ ] ] where @xmath55 . due to the supersymmetry
, only the ground state contributes in the right - moving sector , and the elliptic genus is independent of @xmath56 .
it is known @xcite that the elliptic genus is the jacobi form with weight-@xmath57 and index-@xmath58 .
see appendix [ sec : jacobi ] for the definition of the jacobi form .
the elliptic genus is related to the topological invariants of manifolds .
we have @xmath59 q^{\frac{d}{4 } } \ , z_{cy_d } \left ( z=\frac{1+\tau}{2 } ; \tau \right ) = \widehat{a}_{cy_d}+\cdots , \end{gathered}\ ] ] where @xmath60 and @xmath61 are respectively the euler characteristic and the @xmath62-genus .
we study odd - dimensional calabi - yau manifolds @xmath11 ( @xmath63 odd ) throughout this section .
the @xmath28 charge is @xmath64 .
we see from that the combination of @xmath0 massless characters which is even in @xmath65 can be written as @xmath66 & \operatorname{ch}_{d , h=\frac{d}{8},q=\frac{1}{2}}^{\widetilde{r}}(z;\tau ) + \operatorname{ch}_{d , h=\frac{d}{8},q=-\frac{1}{2}}^{\widetilde{r}}(z;\tau ) \\ & = ( -1)^{\frac{d+1}{2 } } \ , \frac{{\mathrm{i}}\ , \theta_{11}(z;\tau ) } { \left [ \eta(\tau ) \right]^3 } \ , \sum_{n \in \mathbb{z } } q^{\frac{d-1}{2 } n^2 } \ , { \mathrm{e}}^{2 \pi { \mathrm{i}}(d-1 ) n z } \ , \frac { 1 + { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \ , q^n } { 1 - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \ , q^n } \\ & = ( -1)^{\frac{d+1}{2 } } \ , \phi_{0,\frac{3}{2}}(z;\tau ) \ , c^{\mathcal{n}=4}_{\frac{d-3}{2}}(z;\tau ) , \end{aligned}\ ] ] where @xmath67 , defined in @xcite , is the isospin-@xmath57 massless character in @xmath68 @xmath17 sca @xcite .
@xmath69 is a weight-@xmath57 index-@xmath70 jacobi form listed in appendix [ sec : jacobi ] .
similarly the even-@xmath65 combination of massive characters can be written as @xmath71 where @xmath72 , and @xmath73 , defined in , is the basis of massive characters in @xmath17 sca @xcite .
so both the ( even-@xmath65 part of ) @xmath0 massless and massive characters coincide with those of @xmath17 sca up to the jacobi form @xmath74 .
this also happens to the elliptic genus @xmath75 .
the elliptic genus for @xmath11 is a jacobi form with weight-@xmath57 and index-@xmath58 and is an even function of @xmath65 because of .
now there exists a structural theorem on the space of jacobi forms with half - integral index @xmath76 @xcite .
this space is isomorphic to the space @xmath77 @xmath78 in view of the relationship , and the above theorem , we can conclude that the character decomposition of the elliptic genus of the calabi
yau @xmath1-fold is essentially the same as that of the hyperkhler manifolds with complex dimension-@xmath2 , which was studied in our previous paper @xcite .
let us consider the case when the elliptic genus @xmath75 includes a piece @xmath79 .
\label{dominant}\ ] ] the above combination contains the identity representation in the ns sector and gives the dominant contribution to the entropy . using the results of ref .
we can derive the entropy of cy manifolds @xmath15 from the increase in the multiplicity of massive representations with @xmath28 charge @xmath72 @xmath80 the elliptic genus of the calabi
yau @xmath3-fold @xmath81 is given by @xcite @xmath82 where @xmath83 is the euler number of @xmath81 . with a help of we
have @xmath84 which proves a simple decomposition formula for the elliptic genus , @xmath85 .\ ] ] there is no contribution from the massive representations in this decomposition , and we conclude that the entropy of the cy 3-fold vanishes identically , @xmath86 the elliptic genus of the calabi yau @xmath87-fold @xmath88 is @xmath89 where @xmath90 is the euler number of @xmath88 and @xmath91 is one half of the elliptic genus of @xmath20 surface . as was shown in ref .
( see also ref . ) , character decomposition of the @xmath20 elliptic genus is given by @xmath92 \
, b^{\mathcal{n}=4}_{1,1}(z;\tau ) .\ ] ] here the positive integers @xmath93 are given by the rademacher expansion as @xmath94 where @xmath95 is the legendre symbol , and @xmath96 denotes the modified bessel function , @xmath97 since @xmath0 characters coincide with those of @xmath17 up to a factor @xmath98 , the character decomposition of becomes exactly the same as in .
we conclude that @xmath99 s_{cy_5 } & = s_{k3}= \log a_n \\ & \sim 2 \ , \pi \sqrt { \frac{1}{2 } \ , \left ( n - \frac{1}{8 } \right ) } . \end{aligned}\ ] ]
let us next study the even - dimensional calabi
yau manifolds @xmath11 .
@xmath1 is even throughout this section .
the @xmath28 charge is integral @xmath100 .
we follow the method of ref . to construct the character decomposition of the elliptic genus . for notational simplicity we set @xmath101 as the massless character with @xmath28 charge-@xmath57 , @xmath99 { c}_d(z;\tau ) & = ( -1)^{\frac{d}{2 } } \ , \operatorname{ch}_{d , h=\frac{d}{8 } , q=0}^{\widetilde{r}}(z;\tau ) \\ & = \frac{{\mathrm{i}}\ , \theta_{11}(z;\tau ) } { \left [ \eta(\tau ) \right]^3 } \ ,
{ \mathrm{e}}^{\pi { \mathrm{i}}z } \sum_{n \in \mathbb{z } } ( -1)^n \ , q^{\frac{d-1}{2 } n^2 + \frac{1}{2 } n } \ , \frac { { \mathrm{e}}^{2 \pi { \mathrm{i}}(d-1)n z } } { 1 - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \ , q^n } .
\end{aligned}\ ] ] it is identified as @xmath102 ^ 3 } \ , { \mathrm{e}}^{\pi { \mathrm{i}}z } \ , f_{d-1 } \left ( \frac{-1+\tau}{2 \ , ( d-1 ) } , z+ \frac{-1+\tau}{2 \ , ( d-1 ) } ; \tau \right ) , \ ] ] where @xmath103 is the appell function defined in .
appell function undergoes a modular transformation with a mordell s integral ( [ mordell - int ] ) and is thus a typical mock theta function .
one can cure its modular property by the process of completion @xcite . by using ,
we define the completion @xmath104 \ , \phi_{-1,\frac{1}{2}}(z;\tau ) \ , \widetilde{\vartheta}_{\frac{d-1}{2},a}(z;\tau ) , \end{gathered}\ ] ] where @xmath105 is a ( modified ) theta series defined as , and @xmath106 denotes the non - holomorphic function .
( note that the combination @xmath107 is invariant under @xmath108 ) .
one finds that @xmath109 is a real analytic jacobi form satisfying @xmath110 \widehat{{c}}_d(z+\tau ; \tau ) = q^{-\frac{d}{2 } } \ , { \mathrm{e}}^{- 2 \pi { \mathrm{i}}d z } \ , \widehat{{c}}_d(z;\tau ) , \\[2 mm ] \widehat{{c}}_d(z;\tau+1 ) = \widehat{{c}}_d(z+1;\tau ) = \widehat{{c}}_d(z;\tau ) .
\end{gathered}\ ] ] we note that the shadow @xcite , a weight-@xmath21 vector modular form , of our system of mock theta functions is given by @xmath111 & { \mathrm{i}}\ , \sqrt{2 \ , ( d-1 ) } \ , \overline { \sqrt{\im \tau } \ , \frac{\partial}{\partial \overline{\tau } } \left [ { \mathrm{e}}^{- \frac{a-1}{d-1 } \pi { \mathrm{i } } } \ , q^{-\frac{1}{8(d-1 ) } } \ , r_{\frac{d-1}{2},a-1 } \left ( \frac{-1+\tau}{2(d-1 ) } ; \tau \right ) \right ] } \\ & = \sum_{n \in \mathbb{z } } \left ( ( d-1 ) \ ,
n + a-\frac{1}{2 } \right ) \,(-1)^n \ , q^{\frac{1}{2 ( d-1 ) } \left ( ( d-1 ) n + a - \frac{1}{2 } \right)^2 } \\ & = \frac{1}{2 } \left [ \eta(\tau ) \right]^3 \ , \frac { \widetilde{\vartheta}_{\frac{d-1}{2},a } + \widetilde{\vartheta}_{\frac{d-1}{2},d - a } } { \widetilde{\vartheta}_{\frac{1}{2},1 } } ( 0 ; \tau ) , \end{aligned}\ ] ] where @xmath112 .
the theta series @xmath113 , which is a basis for jacobi form with index @xmath114 , may be regarded as the massive characters up to a prefactor ; @xmath115 combining the theta series , we define the basis functions as @xmath116 , & \text{for $ 1 \leq a < \frac{d}{2}$ } , \\[6 mm ] \displaystyle \phi_{-1,\frac{1}{2}}(z;\tau ) \ , \widetilde{\vartheta}_{\frac{d-1}{2},\frac{d}{2}}(z;\tau ) , & \text{for $ a = \frac{d}{2 } $ . }
\end{cases}\ ] ] note that @xmath117 and @xmath118 is an even - function of @xmath65 with a 2nd order zero at @xmath119 .
one has @xmath120 thus they describe the even-@xmath65 part of the massive characters .
they form a set of vector - valued jacobi forms satisfying @xmath121 { b}_{d , a}(z+1;\tau ) = { b}_{d , a}(z;\tau ) , \\[2 mm ] { b}_{d , a}(z+\tau ; \tau ) = q^{- \frac{d}{2 } } \ , { \mathrm{e}}^{- 2 \pi { \mathrm{i}}d z } \ , { b}_{d , a}(z;\tau ) , \\[2 mm ] { b}_{d , a}(z;\tau ) = \sqrt{\frac{\tau}{{\mathrm{i } } } } \ , { \mathrm{e}}^{- \pi { \mathrm{i}}d \frac{z^2}{\tau } } \ , \sum_{b=1}^{\frac{d}{2 } } \frac { \delta_{a , \frac{d}{2 } } -2}{\sqrt{d-1 } } \ , \sin \left ( \frac{(2 \ , a - 1 ) \ , ( 2 \ , b - 1)}{2 \ , ( d-1 ) } \ , \pi \right ) \ , { b}_{d , b } \left ( \frac{z}{\tau } ; - \frac{1}{\tau } \right ) , \\[2 mm ] { b}_{d , a}(z;\tau+1 ) = { \mathrm{e}}^ { \frac{\left ( a - \frac{1}{2 } \right)^2}{d-1 } \pi { \mathrm{i } } } \ , { b}_{d , a}(z;\tau )
. \end{gathered}\ ] ] by using , the latter two identities are summarized as @xmath122_{a_1,a_2 } + \left [ \rho(\gamma ) \right]_{d - a_1,a_2 } \right ) \ , { b}_{d , a_2}(z;\tau ) , & \text{for $ a_1<\frac{d}{2 } $ , } \\[6 mm ]
\displaystyle \frac{1}{\sqrt{c \ , \tau+d } } \ , { \mathrm{e}}^{d \frac{c z^2}{c \tau+d } \pi { \mathrm{i } } } \sum_{a_2=1}^{\frac{d}{2 } } \left [ \rho(\gamma ) \right]_{\frac{d}{2},a_2 } \ , { b}_{d , a_2}(z;\tau ) , & \text{for $ a_1=\frac{d}{2}$. } \end{cases}\end{gathered}\ ] ] by choosing a set of coordinates on the torus , @xmath123 for @xmath124 , we define functions @xmath125 as @xmath126 j_d(z;w_1 , \dots , w_{\frac{d}{2}};\tau ) & = \widehat{{c}}_{d}(z;\tau ) - \sum_{a=1}^{\frac{d}{2 } } \widehat{h}_{d , a}(w_1,\dots , w_{\frac{d}{2 } } ; \tau ) \ , { b}_{d , a}(z;\tau ) \\ & = { { c}}_{d}(z;\tau ) - \sum_{a=1}^{\frac{d}{2 } } { h}_{d , a}(w_1,\dots , w_{\frac{d}{2 } } ; \tau ) \ , { b}_{d , a}(z;\tau ) .
\end{aligned}\ ] ] here @xmath127_{ab } \ ,
\widehat{c}_d(w_b;\tau ) , \\[2 mm ] h_{d , a}(w_1,\dots , w_{\frac{d}{2}};\tau ) = \sum_{b=1}^{\frac{d}{2 } } \left [ \mathbf{{b}}_d(\boldsymbol{w};\tau)^{-1 } \right]_{ab } \ , { c}_d(w_b;\tau ) , \end{gathered}\ ] ] where the matrix @xmath128 is defined by @xmath129_{ab } = { b}_{d , b}(w_a ; \tau ) .\ ] ] we note that @xmath130_{ab } \ , \left ( \widehat{c}_d(w_b;\tau)- { c}_d(w_b;\tau ) \right ) \nonumber \\ & = -{1\over 2 } \sum_b \left [ \mathbf{{b}}_d(\boldsymbol{w};\tau)^{-1 } \right]_{ab } \sum_c { \mathrm{e}}^{-{c-1\over d-1}\pi { \mathrm{i } } } \ , q^{-{1\over 8(d-1 ) } } \ , r_{{d-1\over 2},c-1}\left({-1+\tau \over 2(d-1)};\tau\right)b_{d , c}(w_b;\tau ) \nonumber \\ & = -{1\over 2 } \ , { \mathrm{e}}^{-{a-1\over d-1}\pi { \mathrm{i } } } \ , q^{-{1\over 8(d-1 ) } } \ , r_{{d-1\over 2},a-1}\left({-1+\tau \over 2(d-1)};\tau\right).\end{aligned}\ ] ] thus @xmath131 is a completion of @xmath132 . by use of and
the fact that @xmath132 is @xmath133-independent , we can check that it is annihilated by the hyperbolic laplacian @xmath134 we note that @xmath135 \begin{aligned } j_d(z+1;w_1,\dots , w_{\frac{d}{2 } } ; \tau ) & = j_d(z;w_1,\dots , w_a+1 , \dots , w_{\frac{d}{2 } } ; \tau ) \\ & = j_d(z;w_1,\dots , w_a+\tau , \dots , w_{\frac{d}{2 } } ; \tau ) \\ & = j_d(z;w_1,\dots , w_{\frac{d}{2 } } ; \tau+1 ) \\ & = j_d(z;w_1,\dots , w_{\frac{d}{2 } } ; \tau ) , \end{aligned } \\[2 mm ] j_d(z+\tau;w_1,\dots , w_{\frac{d}{2 } } ; \tau ) = q^{-\frac{d}{2 } } \ , { \mathrm{e}}^{-2 \pi { \mathrm{i}}d z } \ , j_d(z;w_1,\dots , w_{\frac{d}{2 } } ; \tau ) .
\end{gathered}\label{j - prop1}\ ] ] by construction the function @xmath136 vanishes at @xmath137 for @xmath138 @xmath139 and we also have @xmath140 if we choose @xmath141 to be half - periods @xmath142 and use the notation @xmath143 , @xmath144 it is possible from the above conditions , , to show that @xmath145 where @xmath146 .
in this manner the functions @xmath147 generate the basis vectors of the space @xmath148 and their linear combination gives the elliptic genera of @xmath11 .
we set the euler number of @xmath11 to be @xmath60 and consider a function @xmath149 which vanishes at @xmath119 .
completion of this function @xmath150 is a real analytic jacobi form with weight-@xmath57 and index-@xmath58 and can be expanded in terms of @xmath151 .
this is because a function @xmath152 , which is a real analytic jacobi form with weight-@xmath153 and index-@xmath154 , can be expanded in terms of @xmath155 as discussed in .
we have @xmath156 which reduces to @xmath157 by taking the holomorphic part , we obtain a character decomposition of the elliptic genus @xmath75 as @xmath158 where the function @xmath159 is computed by the fourier integral @xmath160 with an arbitrary @xmath161 . since @xmath118 is ( the even part of ) massive characters , the integral fourier coefficients of @xmath159 denotes the multiplicities of the massive representations .
we note that the lowest power @xmath162 of @xmath159 corresponds to the unitarity bound , and that it decomposes into a sum of massless characters as and , @xmath163 , & \text{for $ 1 \leq a < \frac{d}{2}$ } , \\[6 mm ] \displaystyle - \left [ \operatorname{ch}^{\widetilde{r}}_{d , h=\frac{d}{8},q=\frac{d}{2}}(z;\tau ) + \sum_{\varepsilon = \pm 1 } \operatorname{ch}^{\widetilde{r}}_{d , h=\frac{d}{8},q=\varepsilon \left(\frac{d}{2}-1\right)}(z;\tau ) \right ] , & \text{for $ a=\frac{d}{2}$}. \end{cases}\end{gathered}\ ] ] both sides of are real analytic jacobi forms with weight-@xmath57 and index-@xmath58 . because of the transformation formula of @xmath118 , transformation law of the functions @xmath164 is determined as @xmath165_{a_1 , a_2 } \ , \widehat{\sigma}_{d , a_2}(\tau),\ ] ] where the multiplier system @xmath166 is given by @xmath167_{a_1,a_2 } \\
= \begin{cases } \displaystyle \begin{aligned } & \frac { { \mathrm{e}}^{\frac { \left(a_1-\frac{1}{2}\right)^2 } { d-1 } \frac{d}{c } \pi { \mathrm{i } } } } { 2 \ , \sqrt{{\mathrm{i } } } \ , \sqrt{c \ , ( d-1 ) } } \ , \sum_{j=0}^{2c-1 } \sum_{\varepsilon=\pm1 } ( -1)^j \ , \varepsilon \ , { \mathrm{e}}^ { ( d-1 ) \left ( j+ \varepsilon \frac{a_2-\frac{1}{2}}{d-1 } \right)^2 \frac{a}{c } \pi { \mathrm{i}}-2 \pi { \mathrm{i}}\frac{a_1-\frac{1}{2}}{c } \left ( j + \varepsilon \frac{a_2-\frac{1}{2}}{d-1 } \right ) } , \\ & \hspace{100 mm } \text{for $ a_2 < \frac{d}{2 } $ , } \end{aligned } \\[6 mm ] \displaystyle \begin{aligned } & \frac { { \mathrm{e}}^ { \frac{\left(a_1-\frac{1}{2}\right)^2}{d-1 } \frac{d}{c}\pi { \mathrm{i } } } } { 2 \ , \sqrt{{\mathrm{i } } } \ , \sqrt{c \ , ( d-1 ) } } \ , \sum_{j=0}^{2c-1 } ( -1)^j { \mathrm{e}}^{(d-1 ) \left(j+\frac{1}{2}\right)^2 \frac{a}{c } \pi { \mathrm{i}}-2 \pi { \mathrm{i}}\frac{a_1-\frac{1}{2}}{c } \left(j+\frac{1}{2 } \right ) } , \\ & \hspace{100 mm } \text{for $ a_2 = \frac{d}{2 } $ . }
\end{aligned } \end{cases}\end{gathered}\ ] ] in view that the fact that the basis of jacobi forms @xmath168 are decomposed as and the elliptic genus is given by the linear combination of these jacobi forms , we deduce that the functions @xmath164 are also a vector - valued harmonic maass form as in the case of @xmath169 @xmath170 we construct a solution of above differential equations using the poincar
maass series @xmath171 @xcite .
we note that the polar part of @xmath171 has the form of @xmath172 the fourier coefficients @xmath173 of polar parts are fixed from once the elliptic genus @xmath75 is given .
for instance , if the elliptic genus is given by a jacobi form @xmath174 , \end{gathered}\ ] ] where @xmath175 and without loss of generality @xmath176 , we find that @xmath177 where @xmath178 denotes the binomial coefficient . by use of the function @xmath179 as an eigenfunction of the hyperbolic laplacian , a solution of @xmath180 which transforms like with the boundary condition can be constructed in the form of the poincar
maass series as @xmath181_{a_1,a_2 } \
, \frac{1 } { \sqrt{c \ , \tau+ d } } \ , \varphi _ { m - \frac{\left(a_2-\frac{1}{2}\right)^2}{2(d-1 ) } , \frac{3}{4 } } ^{\frac{1}{2 } } \left ( \gamma(\tau ) \right ) , \end{gathered}\ ] ] where @xmath182 is the stabilizer of @xmath183 , @xmath184 .
we have @xmath185 here a weight-@xmath186 theta series @xmath187 having the same transformation law as , and @xmath188 being a modular forms of @xmath189 may appear in the right - hand - side of the equation . to have a non - vanishing theta series @xmath187 , @xmath190 is divisible either by @xmath191 with odd prime @xmath192 , or by @xmath193 with distinct odd primes @xmath192 and @xmath194 , due to the serre stark theorem @xcite . the holomorphic part is thus given by @xmath99 \sigma_{d , a}(\tau ) - \theta_{d , a}(\tau ) & = \left .
p_{d , a}(\tau ) \right|_{\text{holomorphic } } \\ & = q^{-\frac{\left(a -\frac{1}{2}\right)^2}{2 ( d-1 ) } } \sum_{n=0}^\infty p_{d , a}(n ) \ , q^n .
\end{aligned}\ ] ] following the same analysis with our previous paper @xcite ( see ref . for a general treatment ) , we can compute the coefficients @xmath173 as @xmath195 where @xmath196_{a_1,a_2 } \
, \frac{2 \ , \pi}{\sqrt{{\mathrm{i } } } } \ , \left ( \frac { \left(a_2-\frac{1}{2}\right)^2 -2 \ , ( d-1 ) \,m } { 2 ( d-1 ) \ , n - \left(a_1-\frac{1}{2 } \right)^2 } \right)^{\frac{1}{4 } } \\ \times \frac{1}{c } \ , i_{\frac{1}{2 } } \left ( \frac{4 \ , \pi}{c } \ , \sqrt { \left ( n - \frac{\left(a_1-\frac{1}{2}\right)^2}{2 \ , ( d-1 ) } \right ) \ , \left ( \frac{\left(a_2-\frac{1}{2}\right)^2}{2 \ , ( d-1 ) } -m \right ) } \right ) \\ \times { \mathrm{e}}^{- 2 \pi { \mathrm{i}}\left ( \frac{\left(a_2-\frac{1}{2}\right)^2}{2 \ , ( d-1 ) } -m \right ) \frac{a}{c } + 2 \pi { \mathrm{i}}\left ( n - \frac{\left(a_1-\frac{1}{2}\right)^2}{2 \ , ( d-1 ) } \right ) \frac{d}{c } } .\end{gathered}\ ] ] we mean @xmath197 in the summand . since the fourier coefficients of theta series stay constant , they are negligible as compared with @xmath173 which increases exponentially at large @xmath198
. the dominant terms at large @xmath198 read as @xmath199 one finds that the exponential growth of @xmath173 is determined by the maximum value @xmath200 such that @xmath201 .
as in the case of hyperkhler manifolds , @xmath202 is non - zero when the elliptic genus @xmath75 includes a jacobi form @xmath203 as we see from . in this case
we conclude from that the entropy @xmath15 from the @xmath28 charge-@xmath29 is given by @xmath99 s_{cy_d } & = \log \left| p_{d , q}(n ) \right| \\ & \sim 2 \ , \pi \sqrt { \frac{d-1}{2 } \ , n - \left ( \frac{q-\frac{1}{2}}{2 } \right)^2 } .
\end{aligned}\ ] ] it is known that in this case the @xmath28 current algebra is enhanced to a level-1 affine @xmath206 current algebra , and that we have @xmath17 sca @xcite .
@xmath205 is either a complex @xmath204-tori or @xmath20 surface .
the elliptic genus for the former vanishes . in the latter case
, we have @xmath207 , and its character decomposition was studied in detail in ref . .
our result for @xmath208 indeed reproduces multiplicities @xmath93 .
the dimension of the space of jacobi forms @xmath211 is two , and we set the bases of the jacobi forms as @xcite @xmath212 , \\[2 mm ] z_{x_4^{(2)}}(z;\tau ) & = 2 \ , \biggl [ \left ( \frac{\theta_{10}(z;\tau)}{\theta_{10}(0;\tau ) } \cdot \frac{\theta_{00}(z;\tau)}{\theta_{00}(0;\tau ) } \right)^2 + \left ( \frac{\theta_{00}(z;\tau)}{\theta_{00}(0;\tau ) } \cdot \frac{\theta_{01}(z;\tau)}{\theta_{01}(0;\tau ) } \right)^2 \nonumber \\ & \hspace{60 mm } + \left ( \frac{\theta_{01}(z;\tau)}{\theta_{01}(0;\tau ) } \cdot \frac{\theta_{10}(z;\tau)}{\theta_{10}(0;\tau ) } \right)^2 \biggr ] .\end{aligned}\ ] ] the eichler zagier bases are given by @xmath213 ^ 2 \\[1 mm ] \left [ \phi_{-2,1 } \right]^2 \ , e_4 \end{pmatrix } = \begin{pmatrix } 1 & 16 \\ 1 & - 8 \end{pmatrix } \ , \begin{pmatrix }
z_{x_4^{(1)}}(z;\tau ) \\[1 mm ] z_{x_4^{(2)}}(z;\tau ) \end{pmatrix } , \ ] ] in order to derive the character decomposition of these jacobi forms , we use and find explicit results ( we use @xmath214 for brevity ) .
@xmath215 = \left ( - \zeta^{\frac{3}{2 } } - 5 \ , \zeta^{\frac{1}{2 } } + 5 \ , \zeta^{-\frac{1}{2 } } + \zeta^{-\frac{3}{2 } } \right )
\\ + \left ( -5 \ , \zeta^{\frac{5}{2 } } - 207 \ , \zeta^{\frac{3}{2 } } + 790 \ , \zeta^{\frac{1}{2 } } - 790 \ , \zeta^{-\frac{1}{2 } } + 207 \ , \zeta^{-\frac{3}{2 } } + 5 \ , \zeta^{-\frac{5}{2 } } \right ) \,q \\ + \left ( 5 \ , \zeta^{\frac{7}{2 } } + 790 \ , \zeta^{\frac{5}{2 } } - 5724 \ , \zeta^{\frac{3}{2 } } + 13955 \ , \zeta^{\frac{1}{2 } } - 13955 \ , \zeta^{-\frac{1}{2 } } + 5724 \ , \zeta^{-\frac{3}{2 } } - 790 \ , \zeta^{-\frac{5}{2 } } - 5 \ , \zeta^{- \frac{7}{2 } } \right ) \,q^2 \\ + \left ( \zeta^{\frac{9}{2 } } - 790 \ , \zeta^{\frac{7}{2 } } + 13955 \ , \zeta^{\frac{5}{2 } } - 65385 \ , \zeta^{\frac{3}{2 } } + 132909 \ , \zeta^{\frac{1}{2 } } \right .
\\ \left . - 132909 \ , \zeta^{-\frac{1}{2 }
} + 65385 \ , \zeta^{-\frac{3}{2 } } - 13955 \ , \zeta^{-\frac{5}{2 } } + 790 \ , \zeta^{- \frac{7}{2 } } - \zeta^{-\frac{9}{2 } } \right ) \,q^3 + \cdots,\end{gathered}\ ] ] @xmath216 \\
= \left ( - \zeta^{\frac{1}{2 } } + \zeta^{-\frac{1}{2 } } \right ) + \left ( - \zeta^{\frac{5}{2 } } + 6 \ , \zeta^{\frac{3}{2 } } + 5 \ , \zeta^{\frac{1}{2 } } - 5 \ , \zeta^{-\frac{1}{2 } } -6 \ , \zeta^{-\frac{3}{2 } } + \zeta^{-\frac{5}{2 } } \right ) \ , q \\ + \left ( \zeta^{\frac{7}{2 } } + 5 \ , \zeta^{\frac{5}{2 } } + 18 \ , \zeta^{\frac{3}{2 } } + 10 \ , \zeta^{\frac{1}{2 } } - 10 \ , \zeta^{-\frac{1}{2 } } -18 \ , \zeta^{-\frac{3}{2 } } -5 \ , \zeta^{-\frac{5}{2 } } - \zeta^{- \frac{7}{2 } } \right ) \,q^2 \\ + \left ( -5 \ , \zeta^{\frac{7}{2 } } + 10 \ , \zeta^{\frac{5}{2 } } + 30 \ , \zeta^{\frac{3}{2 } } + 21 \ , \zeta^{\frac{1}{2 } } - 21 \ , \zeta^{-\frac{1}{2 } } -30 \ , \zeta^{-\frac{3}{2 } } -10 \ , \zeta^{-\frac{5}{2 } } + 5 \ , \zeta^{- \frac{7}{2 } } \right ) \,q^3 + \cdots .\end{gathered}\ ] ] fourier coefficients in the above expressions give the multiplicities of massive representations , and we obtain the character expansion @xmath217 @xmath218 in tables [ tab : cy4_case1 ] and [ tab : cy4_case2 ] , we confirm numerically that the poincar
maass series reproduce the above fourier coefficients in and accurately , when we set @xmath219 to be @xmath220 and @xmath221 , respectively .
we plot in fig .
[ fig : logcy4 ] absolute values of the fourier coefficients , and , together with the prediction of the dominant term of the poincar
maass series .
@xmath222 @xmath223 -folds .
blue ( resp .
red ) dots show the exact multiplicities of massive representations associated with @xmath224 ( resp .
@xmath225 ) in ( upper part ) and ( lower part ) .
blue and red curves show the approximation by the dominant term of the poincar
maass series with the coefficients of the polar part @xmath226 ( upper part ) and @xmath221 ( lower part ) . ]
following the convention of ref . , we set the bases of the elliptic genera as @xmath229 , \\
z_{x_6^{(2)}}(z;\tau ) = 8 \ , \left [ \left ( \frac{\theta_{10}(z;\tau)}{\theta_{10}(0;\tau ) } \right)^4 \ , \left ( \frac{\theta_{00}(z;\tau)}{\theta_{00}(0;\tau ) } \right)^2 + \text{other 5 terms } \right ] , \\
z_{x_6^{(3)}}(z;\tau ) = 4 \ , \left ( \frac{\theta_{10}(z;\tau)}{\theta_{10}(0;\tau ) } \right)^2 \ , \left ( \frac{\theta_{00}(z;\tau)}{\theta_{00}(0;\tau ) } \right)^2 \ , \left ( \frac{\theta_{01}(z;\tau)}{\theta_{01}(0;\tau ) } \right)^2 .\end{gathered}\ ] ] we note that these bases are related to the eichler zagier bases @xcite by @xmath230 ^ 3 \\[1 mm ] \left [ \phi_{-2,1 } \right]^2 \ , \phi_{0,1 } \ , e_4 \\[1 mm ] \left [ \phi_{-2,1 } \right]^3 \ , e_6 \end{pmatrix } = \begin{pmatrix } 1 & 24 & 96 \\ 1 & 0 & -48 \\ 1 & -12 & 96 \end{pmatrix } \ , \begin{pmatrix } z_{x_6^{(1 ) } } \\[1 mm ] z_{x_6^{(2 ) } } \\[1 mm ] z_{x_6^{(3 ) } } \end{pmatrix } .\ ] ] for these jacobi forms , we have character decompositions as follows ; @xmath231 @xmath232 @xmath233 we have checked numerically that the poincar maass series generates the above multiplicities of the massive representations quite accurately , when we set the polar part @xmath234 to be @xmath235 , @xmath236 , and @xmath237 , respectively .
we have computed the entropy of the calabi
yau @xmath1-folds by use of the poincar
maass series .
we have found that , when @xmath1 is odd , the entropy coincides with that of the hyperkhler manifolds with a complex ( @xmath238)-dimension . especially the entropy of the calabi
yau 3-folds vanishes identically
. it will be interesting to provide physical intepretation of this result . in summary ,
the entropy of calabi
yau @xmath1-folds is given by @xmath239
this work is supported in part by grant - in - aid from the ministry of education , culture , sports , science and technology of japan .
the jacobi theta functions are defined by @xmath240 \theta_{10}(z;\tau ) & = \sum_{n \in \mathbb{z } } q^{\frac{1}{2 } \left ( n + \frac{1}{2 } \right)^2 } \ , { \mathrm{e}}^{2 \pi { \mathrm{i}}\left ( n+\frac{1}{2 } \right ) z } , \\[2 mm ] \theta_{00 } ( z;\tau ) & = \sum_{n \in \mathbb{z } } q^{\frac{1}{2 } n^2 } \ , { \mathrm{e}}^{2 \pi { \mathrm{i}}n z } , \\[2 mm ] \theta_{01 } ( z;\tau ) & = \sum_{n \in \mathbb{z } } q^{\frac{1}{2 } n^2 } \ , { \mathrm{e}}^{2 \pi { \mathrm{i}}n \left ( z+\frac{1}{2 } \right ) } . \end{aligned}\ ] ] throughout this paper , we set @xmath241 with @xmath242 in the upper half plane , @xmath243 .
we define the theta function for @xmath244 and @xmath245 by @xmath246 we have @xmath247 note that @xmath248 for our notational convention , we introduce another set of theta series @xmath105 for @xmath249 and @xmath250 @xmath251 \widetilde{\vartheta}_{\frac{d-1}{2},a}(z;\tau ) & = { \mathrm{e}}^{\frac{a-1}{d-1 } \pi { \mathrm{i } } } \ , { \mathrm{e}}^{\pi { \mathrm{i}}z } \ , q^{\frac{1}{8(d - 1 ) } } \ , \vartheta_{\frac{d-1}{2},a-1 } \left ( z + \frac{-1+\tau}{2 \ , ( d-1 ) } ; \tau \right ) \\ & = \sum_{n \in \mathbb{z } } ( -1)^n \ , q^{\frac{1}{2 ( d-1 ) } \left ( ( d-1 ) n + a - \frac{1}{2 } \right)^2 } \ , { \mathrm{e}}^{2 \pi { \mathrm{i}}\left ( ( d-1 ) n + a - \frac{1}{2 } \right ) z }
. \end{aligned}\ ] ] we see that @xmath252 and we have @xmath253 \widetilde{\vartheta}_{\frac{d-1}{2},a}(z+\tau ; \tau ) = - q^{-\frac{d-1}{2 } } \ , { \mathrm{e}}^{-2 \pi { \mathrm{i}}(d-1 ) z } \ , \widetilde{\vartheta}_{\frac{d-1}{2},a}(z;\tau ) , \\[2 mm ] \widetilde{\vartheta}_{\frac{d-1}{2},a}(z;\tau+1 ) = \begin{cases } \displaystyle { \mathrm{e}}^{\pi { \mathrm{i}}\frac{\left(a-\frac{1}{2}\right)^2}{d-1 } } \ , \widetilde{\vartheta}_{\frac{d-1}{2},a}(z;\tau ) , & \text{for $ 1\leq a \leq \frac{d}{2}$ } , \\[4 mm ] \displaystyle { \mathrm{e}}^{\pi { \mathrm{i}}\frac{\left(d - a - \frac{1}{2}\right)^2}{d-1 } } \ , \widetilde{\vartheta}_{\frac{d-1}{2},a}(z;\tau ) , & \text{for $ a > \frac{d}{2}$ } , \end{cases } \\[2 mm ] \widetilde{\vartheta}_{\frac{d-1}{2},a_1}(z;\tau ) = \sqrt{\frac{{\mathrm{i}}}{\tau } } \ , { \mathrm{e}}^{- \pi { \mathrm{i}}(d-1 ) \frac{z^2}{\tau } } \ , \frac{1}{\sqrt{d-1 } } \sum_{a_2=1}^{d-1 } { \mathrm{e}}^{\frac{(2 a_1 -1 ) ( 2 a_2 -1)}{2(d-1 ) } \pi { \mathrm{i } } } \ , \widetilde{\vartheta}_{\frac{d-1}{2},a_2 } \left ( \frac{z}{\tau } ; - \frac{1}{\tau } \right ) .
\label{theta - def } \end{gathered}\ ] ] generally for @xmath254 , we have ( see , _
e.g. _ , refs . , also ref .
in which it is used as surgery data in constructing the @xmath255 quantum invariants of 3-manifolds ) @xmath256_{a_1,a_2 } \ , \widetilde{\vartheta}_{\frac{d-1}{2},a_2}(z;\tau),\ ] ] where @xmath257 is a @xmath258 matrix defined by @xmath259_{a_1,a_2 } \\
= \begin{cases } \displaystyle \frac { { \mathrm{e}}^{\frac{d}{c } \frac { \left ( a_2 - \frac{1}{2 } \right)^2 } { d-1 } \pi { \mathrm{i } } } \ , } { 2 \ , \sqrt{{\mathrm{i } } } \ , \sqrt{c \ , ( d-1 ) } } \sum_{j=0}^{2c-1 } ( -1)^j \ , { \mathrm{e}}^{\pi { \mathrm{i}}\frac{a}{c } ( d-1 ) \left ( j+\frac{a_1-\frac{1}{2}}{d-1 } \right)^2 - 2 \pi { \mathrm{i}}\frac{a_2-\frac{1}{2}}{c } \ , \left ( j+\frac{a_1-\frac{1}{2}}{d-1 } \right ) } , & \text{for $ a_2 \leq \frac{d}{2}$ } , \\[6 mm ] \displaystyle - \frac { { \mathrm{e}}^{\frac{d}{c } \frac { \left ( a_2 -d + \frac{1}{2 } \right)^2 } { d-1 } \pi { \mathrm{i } } } \ , } { 2 \,\sqrt{{\mathrm{i } } } \ , \sqrt{c \ , ( d-1 ) } } \sum_{j=0}^{2c-1 } ( -1)^j \ , { \mathrm{e}}^{\pi { \mathrm{i}}\frac{a}{c } ( d-1 ) \left ( j+\frac{a_1-\frac{1}{2}}{d-1 } \right)^2 - 2 \pi { \mathrm{i}}\frac{a_2-d+\frac{1}{2}}{c } \ , \left ( j+\frac{a_1-\frac{1}{2}}{d-1 } \right ) } , & \text{for $ a_2>\frac{d}{2}$}. \end{cases}\end{gathered}\ ] ] ( note : the symbol @xmath260 should not be confused with the symbol @xmath261 which signifies the completion of mock theta functions ) .
the jacobi form with weight-@xmath262 and index-@xmath263 fulfills @xmath264 f(z+s \ , \tau + t ; \tau ) = ( -1)^{2 m ( s+t ) } \ , { \mathrm{e}}^{-2 \pi { \mathrm{i}}m ( s^2 \tau + 2 s z ) } \ , f(z;\tau ) , \label{jacobi - def } \end{gathered}\ ] ] where @xmath265 and @xmath266 .
the weak jacobi form has a non - negative power of @xmath267 .
see ref . as a basic reference .
we denote @xmath268 as the space of weak jacobi forms with weight-@xmath269 and index-@xmath270 .
the space @xmath268 with even weight @xmath269 and integral index @xmath271 is generated by @xcite @xmath272 namely bases of @xmath268 are @xmath273^a \ , \left [ e_6(\tau ) \right]^b \ , \left [ \phi_{-2,1}(z;\tau ) \right]^c \ , \left [ \phi_{0,1}(z;\tau ) \right]^d,\ ] ] with non - negative integers @xmath274 , @xmath275 , @xmath9 , @xmath276 satisfying @xmath277 here @xmath278 and @xmath279 are the eisenstein series defined by @xmath280 where @xmath281 and @xmath282 are respectively the bernoulli number and the divisor function @xmath283 \sigma_k(n ) = \sum_{1 \leq r | n } r^k .\end{gathered}\ ] ] the remaining two functions with index-@xmath153 are @xmath284 ^ 2 } { \left [ \eta(\tau ) \right]^6 } , \\[2 mm ] \phi_{0,1}(z;\tau ) & = 4 \ , \left [ \left ( \frac{\theta_{10}(z;\tau)}{\theta_{10}(0;\tau ) } \right)^2 + \left ( \frac{\theta_{00}(z;\tau)}{\theta_{00}(0;\tau ) } \right)^2 + \left ( \frac{\theta_{01}(z;\tau)}{\theta_{01}(0;\tau ) } \right)^2 \right ] . \label{phi01}\end{aligned}\ ] ] in the case of the index being half - odd integral , the space @xmath285 is isomorphic to the space with an integral index as follows @xcite ; @xmath286 \mathbb{j}_{2k+1,m+\frac{1}{2 } } = \phi_{-1,\frac{1}{2}}(z;\tau ) \cdot \mathbb{j}_{2k+2,m } , \end{gathered}\ ] ] where @xmath74 and @xmath287 are defined by @xmath288 \phi_{-1,\frac{1}{2}}(z;\tau ) & = \frac{{\mathrm{i}}\ , \theta_{11}(z;\tau ) } { \left [ \eta(\tau ) \right]^3 } .
\end{aligned}\ ] ]
the jacobi form @xmath289 with weight-@xmath269 and index-@xmath270 can be expanded in terms of the theta functions @xmath290 ( resp .
@xmath291 ) when the index is @xmath271 ( resp .
@xmath292 ) . 1 . when @xmath271 , we have @xmath293 here with an arbitrary @xmath294 , we have @xmath295 2 .
when @xmath292 , we have @xmath296 here with an arbitrary @xmath294 , we have @xmath297 the former is a standard result ( see , _
e.g. _ , ref . ) , and the proof is as follows . in the case of an integral index-@xmath271 ,
the jacobi form is periodic @xmath298 , and we have the fourier expansion @xmath299 where the prefactor is for our convention . then the function @xmath300 is given by . using @xmath301 in the integrand
, we obtain @xmath302 which gives . in the same manner
, we can prove the case ( b ) . when the index @xmath270 is half - odd integral @xmath303 , the jacobi form @xmath289 is anti - periodic , @xmath304 .
then we have the fourier expansion @xmath305 here again the prefactor is chosen for our convention , and the function @xmath300 is defined by the fourier integral . substituting @xmath306 for
, we find @xmath307 which proves .
for @xmath308 , we set @xmath309 to be the hyperbolic laplacian @xmath99 \delta_\ell & = - v^2 \left ( \frac{\partial^2}{\partial u^2 } + \frac{\partial^2}{\partial v^2 } \right ) + { \mathrm{i}}\ , \ell \ , v \ , \left ( \frac{\partial}{\partial u } + { \mathrm{i}}\ , \frac{\partial}{\partial v } \right ) \\ & = -4 \ , \left ( \im \tau \right)^{2 - \ell } \frac{\partial}{\partial \tau } \left ( \im \tau \right)^\ell \frac{\partial}{\partial \overline{\tau } } . \end{aligned}\ ] ] following ref . , we set for @xmath310 @xmath311 here the function @xmath312 is defined by @xmath313 where @xmath314 is the @xmath315-whittaker function @xcite .
the function @xmath316 is an eigenfunction of @xmath309 ; @xmath317 \ , \varphi_{-h , s}^{\ell}(\tau ) , \ ] ] which behaves at @xmath318 as @xmath319
we set the appell function @xcite for @xmath320 , @xmath321 by use of watson s method @xcite , we can check that it transforms under the @xmath322-transformation as @xmath323 here the integration in the right hand side is the mordell integral @xcite . following zwegers @xcite ( see refs . for a review ; also refs . for a classical review ) , we define the completion @xmath324 by @xmath325 here the non - holomorphic partner is defined by @xmath326 \ , q^{- \frac{m^2}{2 d } } \ , { \mathrm{e}}^{-2 \pi { \mathrm{i}}m u } , \label{r - func}\ ] ] where @xmath327 then the completion @xmath328 has the following modular transformation @xmath329
\widehat{f}_d\left ( \frac{u}{\tau } , \frac{z}{\tau } ; - \frac{1}{\tau } \right ) = \tau \ , { \mathrm{e}}^{\pi { \mathrm{i}}d \frac{z^2-u^2}{\tau } } \ , \widehat{f}_d(u , z;\tau ) .
\end{gathered}\ ] ]
in our previous studies on the @xmath17 superconformal algebras @xcite with central charge @xmath68 , we have used @xmath330 ^ 2 } { \left [ \eta(\tau ) \right]^3 } \ , \frac { \vartheta_{k+1,a } - \vartheta_{k+1 , -a } } { \vartheta_{2,1 } - \vartheta_{2,-1 } } ( z;\tau ) , \\[2 mm ] c^{\mathcal{n}=4}_k(z;\tau ) = \frac { \left [ \theta_{11}(z ; \tau ) \right]^2 } { \left [ \eta(\tau ) \right]^3 } \ , \frac{{\mathrm{i}}}{\theta_{11}(2 \ , z;\tau ) } \ ,
\sum_{n \in \mathbb{z } } q^{(k+1 ) n^2 } \ , { \mathrm{e}}^{4 \pi { \mathrm{i}}(k+1 ) n z } \ , \frac { 1+{\mathrm{e}}^{2 \pi { \mathrm{i}}z } \ , q^n } { 1 - { \mathrm{e}}^{2 \pi { \mathrm{i}}z } \ , q^n } .
\label{define_c_k}\end{gathered}\ ] ] here @xmath73 with @xmath331 is the basis function for the massive characters , and @xmath67 is the massless character with isospin-@xmath57 .
base functions @xmath73 are vector - valued jacobi forms satisfying @xmath332 b^{\mathcal{n}=4}_{k , a}(z;\tau+1 ) = { \mathrm{e}}^ { \frac{a^2}{2(k+1 ) } \pi { \mathrm{i } } } \ , b^{\mathcal{n}=4}_{k , a}(z ; \tau ) , \\[2 mm ] b^{\mathcal{n}=4}_{k , a}(z+1 ; \tau ) = b^{\mathcal{n}=4}_{k , a}(z ; \tau ) , \\[2 mm ] b^{\mathcal{n}=4}_{k , a}(z+\tau ; \tau ) = q^{-k } \ , { \mathrm{e}}^{-4 \pi { \mathrm{i}}k z } \ , b^{\mathcal{n}=4}_{k ,
a}(z;\tau ) , \end{gathered}\ ] ] while the massless character is a mock theta function satisfying @xmath333 completion of @xmath67 is given by @xmath334 which is a real analytic jacobi form satisfying @xmath335 \widehat{c}^{\mathcal{n}=4}_k(z;\tau+1 ) = \widehat{c}^{\mathcal{n}=4}_k(z+1 ; \tau ) = \widehat{c}^{\mathcal{n}=4}_k(z ; \tau ) , \\[2 mm ] \widehat{c}^{\mathcal{n}=4}_k(z+\tau ; \tau ) = q^{-k } \ , { \mathrm{e}}^{-4 \pi { \mathrm{i}}k z } \ , \widehat{c}^{\mathcal{n}=4}_k(z ; \tau ) . \end{gathered}\ ] ] g. e. andrews , _ mock theta functions _ , in l. ehrenpreis and r. c. gunning , eds .
, _ theta functions bowdoin 1987 _ , vol .
49 ( part 2 ) of _ proc .
pure math .
_ , pp . 283298 , amer .
soc . , providence , 1989 .
l. a. borisov and a. libgober , _ elliptic genera of toric varieties and applications to mirror symmetry _ , http://dx.doi.org/10.1007/s002220000058[invent .
] * 140 * , 453485 ( 2000 ) , http://jp.arxiv.org/abs/math/9904126 [ math/9904126 ] .
m. cveti and f. larsen , _ near horizon geometry of rotating black holes in five dimensions _ , http://dx.doi.org/10.1016/s0550-3213(98)00604-x[nucl .
b ] * 531 * , 239255 ( 1998 ) , http://jp.arxiv.org/abs/hep-th/9805097[hep-th/9805097 ] .
t. eguchi and k. hikami , _ superconformal algebras and mock theta functions _ , http://dx.doi.org/10.1088/1751-8113/42/30/304010[j .
a : math . theor . ] * 42 * , 304010 ( 2009 ) , 23 pages , http://jp.arxiv.org/abs/0812.1151[arxiv:0812.1151 [ math - ph ] ] .
, _ superconformal algebras and mock theta functions 2 .
rademacher expansion for k3 surface _
number theory phys .
* 3 * , 531554 ( 2009 ) , http://jp.arxiv.org/abs/0904.0911 [ arxiv:0904.0911 [ math - ph ] ] . , _ @xmath17 superconformal algebra and the entropy of hyperkhler manifolds _ , http://dx.doi.org/10.1007/jhep02(2010)019[j .
high energy phys . ] * 2010:02 * , 019 ( 2010 ) , 28 pages , http://jp.arxiv.org/abs/0909.0410[arxiv:0909.0410 [ hep - th ] ] .
t. eguchi , h. ooguri , a. taormina , and s .- k .
yang , _ superconformal algebras and string compactification on manifolds with su(@xmath198 ) holonomy _ , http://dx.doi.org/10.1016/0550-3213(89)90454-9[nucl .
b ] * 315 * , 193221 ( 1989 ) .
t. kawai , y. yamada , and s .- k .
yang , _ elliptic genera and @xmath338 superconformal field theory _ , http://dx.doi.org/10.1016/0550-3213(94)90428-6[nucl .
b ] * 414 * , 191212 ( 1994 ) , http://jp.arxiv.org/abs/hep-th/9306096[hep-th/9306096 ] .
e. b. kiritsis , _ character formulae and the structure of the representations of the @xmath340 , @xmath338 superconformal algebras _ , http://dx.doi.org/10.1142/s0217751x88000795[int .
a ] * 3 * , 18711906 ( 1988 ) . c. d. d. neumann , _ the elliptic genus of calabi
yau @xmath3- and @xmath209-folds , product formulae and generalized kac
moody algebras _ , http://dx.doi.org/10.1016/s0393-0440(98)00015-1[j .
* 29 * , 512 ( 1999 ) , http://jp.arxiv.org/abs/hep-th/9607029 [ hep - th/9607029 ] .
, _ unearthing the visions of a master : harmonic maass forms and number theory _ , in d. jerison , b. mazur , t. mrowka , w. schmid , r. p. stanley , and s .- t .
yau , eds .
, _ current developments in mathematics 2008 _ , pp . 347454 , intl .
press , boston , 2009 .
l. rozansky , _
witten s invariants of rational homology spheres at prime values of k and trivial connection contribution _ , http://dx.doi.org/10.1007/bf02099715[commun .
phys . ] * 180 * , 297324 ( 1996 ) , http://jp.arxiv.org/abs/q-alg/9504015[q-alg/9504015 ] .
a. m. semikhatov , i. y. tipunin , and a. taormina , _ higher - level appell functions , modular transformations , and characters _ , http://dx.doi.org/10.1007/s00220-004-1280-7[commun .
phys . ] * 255 * , 469512 ( 2005 ) , http://jp.arxiv.org/abs/math/0311314[math/0311314 ] .
serre and h. m. stark , _ modular forms of weight @xmath186 _ , in j .- p . serre and d. zagier ,
eds . , _ modular functions of one variable vi _ ,
627 of _ lecture notes in mathematics _ , pp .
2767 , springer , berlin , 1977 .
a. strominger and c. vafa , _ microscopic origin of the bekenstein hawking entropy _ , http://dx.doi.org/10.1016/0370-2693(96)00345-0[phys .
b ] * 379 * , 99104 ( 1996 ) , http://jp.arxiv.org/abs/hep-th/9601029[hep-th/9601029 ] .
n. yokoi and t. nakatsu , _ three - dimensional extremal black holes and the maldacena duality _ , http://dx.doi.org/10.1143/ptp.104.439[prog .
theor . phys . ]
* 104 * , 439458 ( 2000 ) , http://jp.arxiv.org/abs/hep-th/9912096[hep-th/9912096 ] .
|
we use the representation theory of @xmath0 superconformal algebra to study the elliptic genera of calabi yau ( cy ) @xmath1-folds .
we compute the entropy of cy manifolds from the growth rate of multiplicities of the massive ( non - bps ) representations in the decomposition of their elliptic genera .
we find that the entropy of cy manifolds of complex dimension @xmath1 behaves differently depending on whether @xmath1 is even or odd .
when @xmath1 is odd , cy entropy coincides with the entropy of the corresponding hyperkhler @xmath2-folds due to a structural theorem on jacobi forms . in particular
, we find that the calabi
yau @xmath3-fold has a vanishing entropy . at @xmath4 , using our previous results on hyperkhler manifolds , we find @xmath5 .
when @xmath1 is even , we find the behavior of cy entropy behaving as @xmath6 .
these agree with cardy s formula at large @xmath1 .
yitp report 10 - 14
|
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.