article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the laser interferometer space antenna ( lisa ) is a planned gravitational wave observatory in the frequency range of 0.1mhz100mhz that consists of three spacecraft separated by 5 million km in a nearly equilateral triangle whose center follows the earth in a heliocentric orbit with an orbital phase offset of 20 degrees .
gravitational waves will be detected as distance fluctuations between test masses moving along geodetic trajectories that are located in different spacecraft .
lisa will require low power ultra - low noise photoreceivers for precision inter - spacecraft heterodyne laser interferometry .
quadrant photoreceivers will be used to measure the test mass motion with a sensitivity of 8@xmath1 in angular orientation and 10@xmath2 in displacement over the frequency range of 0.1mhz100mhz@xcite .
the laser beam at the transmitting spacecraft will have a diameter of approximately 40 cm and an output laser power at the telescope of the order of 1w .
given the laser beam propagation over @xmath3 m and accounting for losses on the beam path , from the remote optical signal approximately 50pw will be detected on the entire quadrant photoreceiver .
lisa will use heterodyne laser interferometry ( see figure [ hetdetect ] ) for the inter - spacecraft displacement measurement .
the incoming weak signal will optically interfere with a stronger local oscillator @xmath4 .
the combined signal @xmath5 measured at the photoreceiver can be expressed as @xmath6 where @xmath7 is the frequency difference between the interfering laser beams ( heterodyne frequency ) , and @xmath8 is the interferometer phase containing the gravitational wave information .
both ports of the beamsplitter will be measured by quadrant detectors .
their combined information can also be used for common - mode rejection of laser amplitude noise .
the main task of the photoreceiver development is to maintain nearly shot - noise limited performance over a measurement bandwidth from 220mhz@xcite .
this frequency range is driven by the doppler induced frequency variations of the optical beat note signal due to the lisa constellation armlength changes , given by the orbits of each spacecraft .
descriptive diagram of laser heterodyne interferometric detection . ] the local oscillator power @xmath4 can be adjusted , according to the required photoreceiver performance . to reduce power consumption , temperature gradients at the optical bench due to hot spots at the photoreceivers , and to provide additional design margin
, we aim for a low - noise wide - bandwidth photoreceiver development operating at low @xmath4 levels . using 0.5mw local oscillator optical power on the entire quadrant photoreceiver ( of the order of 100@xmath9w per quadrant ) , and assuming a responsivity @xmath10 of 0.7a / w for ingaas photodiodes at a laser wavelength of 1064 nm , the shot - noise @xmath11 can be computed as @xmath12 allocating @xmath13 of the shot - noise level @xmath14 to the input current noise contribution of a quadrant photoreceiver , and considering this is the quadrature sum of the current noise of the individual quadrants , we set an input current noise goal of @xmath15 for the single - quadrant photoreceiver transimpedance amplifier ( tia ) .
we have chosen a conventional dc - coupled tia topology with a single ultra - low noise / wide - bandwidth operational amplifier ( op - amp ) , as shown in figure [ tia ] .
topology and noise model of photoreceiver tia . ] for the tia topology shown in figure [ tia ] , two main input current noise sources have been identified in the electronics : * johnson noise ( @xmath16 ) from feedback resistor @xmath17 : @xmath18 where @xmath19 is the boltzmann s constant , and @xmath20 is the temperature ( in kelvin ) . *
op - amp noise properties : the current @xmath21 and voltage @xmath22 noise properties of the op - amp contribute to the total tia input current noise .
* * the op - amp current noise @xmath21 sums directly to the tia input . * * the op - amp voltage noise @xmath22 translates to current noise at the tia input @xmath23 over the input and feedback impedances as @xmath24 where @xmath25 is the frequency , and @xmath26 is the total circuit capacitance @xmath27 including the photodiode capacitance @xmath28 , feedback impedance @xmath29 , op - amp common - mode input capacitance @xmath30 , and stray capacitances @xmath31 from the board , components and packaging . the bandwidth @xmath32 of the photoreceiver can be estimated as @xmath33 where @xmath34 is the gain - bandwidth product of the operational amplifier .
the total tia input current noise @xmath35 model can be expressed as @xmath36 where @xmath37 is the normalized tia transfer function , @xmath23 is a frequency dependent component of the input current noise ( see equation [ i_tia ] ) , and @xmath38 is the quadrature sum of various contributors that can be approximated by neglecting their frequency dependency for modeling purposes .
for example , the expected current noise @xmath38 in the photoreceiver shown in figure [ tia ] , can be computed as @xmath39 where @xmath21 is the op - amp current noise , @xmath16 is the johnson noise of the feedback resistor @xmath17 ( see equation [ i_j ] ) , and @xmath40 is the shot - noise from the photodiode dark current . it can be seen from equation [ ct ] that the photodiode capacitance and the op - amp common - mode input capacitance are crucial factors in the total noise budget .
the challenge for the photodiode manufacture lays in achieving a minimum capacitance per unit area while maintaining high responsivity and low leakage if reverse - biased . for the tia electronics ,
it is necessary to identify an op - amp with minimal common - mode input capacitance , current and voltage noise , and a gain - bandwidth product large enough to maintain the required sensitivity over the required measurement bandwidth of 220mhz .
under a small business innovation research ( sbir ) grant , the company discovery semiconductors has developed a large - area quadrant photodiode ( qpd ) of 1 mm diameter and a quadrant capacitance of 2.5pf when reverse - biased at 5v .
a first fully integrated quadrant photoreceiver ( figure [ ds_quad ] : qpd + tia electronics ) performs with an equivalent input current noise of less than 3.2pa/@xmath41 below 20mhz@xcite .
the characteristics of this prototype quadrant photoreceiver are : * diameter of 1 mm with a 20@xmath9 m inter - quadrant gap . * individual quadrant capacitance @xmath28=2.5pf when reverse - biased at 5v . *
dark current : 140na when reverse - biased at 5v .
* responsivity at 1064 nm : @xmath42a / w ( quantum efficiency of 0.8 ) .
* tia characteristics : * * feedback impedance : @xmath43 , @xmath44 . * * op - amp ada4817 : @xmath45 , @xmath46 , @xmath30=1.4pf , @xmath47 .
discovery semiconductors has been awarded a second stage grant to further develop quadrant photoreceivers .
given the successful development of a large - area low - capacitance qpd in the first step , the next stage will be focused on the noise reduction of the electronics , e.g. by integrating op - amps with better noise properties and studying alternative tia topologies .
we expect to receive additional devices with lower noise electronics at a later date .
photograph of a prototype quadrant photoreceiver manufactured by discovery semiconductors . ] working in parallel with discovery semiconductors to try to understand the noise and bandwidth trade - offs in more detail , we have identified the ultra - low noise / high - bandwidth op - amp el5135 from intersil@xcite with the following nominal noise properties : @xmath48 , @xmath49 , @xmath30=1pf , @xmath50 . a significant reduction in the op - amp voltage noise together with the bandwidth enhancement ( given by the higher @xmath34 ) were the main factors considered for selecting the el5135 for laboratory prototype photoreceivers . despite the higher op - amp current noise @xmath21 of the el5135 compared to the ada4817 ( mentioned above )
, the lower voltage noise @xmath22 of the el5135 enables us to achieve significantly better performance over the entire bandwidth from 220mhz .
the penalty is somewhat higher noise at low frequencies . according to equation [ i_noise ] , the term @xmath38 ( corresponding to the quadrature sum of various current noise contributions , including the op - amp current noise @xmath21 ) dominates at lower frequencies , while the frequency dependent term @xmath51 ( consisting of the op - amp voltage noise @xmath22 swing across the total equivalent tia impedance ) becomes the dominant noise contribution at higher frequencies .
+ we have designed a tia with a feedback impedance @xmath52 , @xmath44 , and an expected bandwidth of 40mhz ( according to equation [ bw ] ) .
we have built two different prototype boards to test the noise properties of a tia with the el5135 op - amp : * * gap500q photoreceiver board * : we have chosen the commercially available qpd gap500q from gpd optoelectronics@xcite with a diameter of 0.5 mm and a responsivity of approximately 0.7a / w ( quantum efficiency of 0.8 ) at 1064 nm . when reverse - biased at 5v , this device has a nominal quadrant capacitance @xmath28=2.0pf and a dark current of 2.0na , according to the manufacturer .
the purpose of this investigation is to operate the tia electronics with a photodiode that approximates the per - quadrant capacitance and package parasitic capacitance of the larger area discovery semiconductors detector . *
* mock - up tia board * : we have built a board for controlled noise investigations of the tia performance , shown in figure [ mockup_tia ] .
+ schematics of tia mock - up board for noise measurements and frequency response measurements . ]
+ this board has two inputs : 1 .
input 1 : an input capacitor ( 2.4pf ) of similar quadrant capacitance is used to replace the photodiode . for noise measurements
, this input can be grounded while maintaining input 2 open .
2 . input 2 : this is used to measure the expected photoreceiver transfer function ( tf ) by injecting a signal ( maintaining input 1 open ) , and scaling it accordingly by the feedback gain ( @xmath53 ) .
the equivalent input current noise can be obtained , by dividing the output voltage noise by the scaled transfer function .
the gap500q photodiode was reverse - biased with a battery power supply at 5v , and the op - amps in both circuits were driven with a power supply at @xmath545v .
we operated the photoreceiver using only one quadrant of the gap500q qpd with the el5135 op - amp for the tia electronics .
the photoreceiver output voltage noise @xmath55 is given by @xmath56 where @xmath11 is the photocurrent shot - noise of the incident light , @xmath57 is the photoreceiver transfer function , and @xmath58 is the equivalent input current noise of the tia electronics .
operating under dark conditions , the photoreceiver output voltage noise @xmath59 is given by @xmath60 by dividing equations [ vn ] and [ vdn]@xcite , we obtain that the input current noise of the tia electronics can be computed as @xmath61 for equivalent input current noise measurements , we used a light - emitting diode ( led ) at a center wavelength of 1050nm(@xmath5450nm)@xcite as shot - noise - limited light source .
we measured the gap500q photoreceiver output voltage noise with ( @xmath55 ) and without ( @xmath59 ) led light , operating at two different optical power levels of 90@xmath9w and 60@xmath9w that are representative for the expected nominal 100@xmath9w per quadrant .
these measurements showed equivalent input current noise levels @xmath58 that did not scale with the dc optical power level ( 90@xmath9w and 60@xmath9w ) .
this also shows that the measured current noise @xmath58 upon subtraction of the shot - noise contribution , is not dependent on the optical power , which is consistent with a shot - noise behavior of the light source .
+ we also measured the output voltage noise and the transfer function ( @xmath57 ) of our mock - up tia board ( see figure [ mockup_tia ] ) .
analogous to equation [ vdn ] , the input current noise can be computed by referring the output voltage noise ( upon subtraction in quadrature of the rf spectrum analyzer voltage noise floor ) to the input dividing by the transfer function .
figure [ noise_meas ] shows the noise measurements .
input current noise measurements of photoreceiver prototypes .
the dashed trace is the photoreceiver with one quadrant of the gap500q qpd and the el5135 op - amp tia .
the solid trace is the mock - up test board ( see figure [ mockup_tia ] ) .
the dashed - dotted trace is the photoreceiver noise model for a tia design with an el5135 op - amp and a quadrant capacitance of 2.5pf ( equations [ i_noise ] ) .
the traces with crosses and circles show the corresponding photoreceiver noise models using parameter values obtained from a fit to the data .
the dotted trace is the equivalent current noise floor of the rf spectrum analyzer used as measurement instrument , referred to the input by the tia transfer function .
the thick solid traced is our tia input current noise goal of @xmath62 in the measurement band 2 - 20 mhz . ]
the gap500q photoreceiver reaches a level of about @xmath62 up to @xmath6310mhz , increasing at higher frequencies .
it exceeds the noise goal by approximately 20% ( @xmath64 ) at 20mhz .
the mock - up tia circuit meets our noise goal over the entire bandwidth ( 2 - 20mhz ) , however , it shows noise in excess of the model . at lower frequencies the equivalent input noise
is determined by excess current noise ( @xmath38 : assumed to have no frequency dependence , equation [ it ] ) , while at higher frequencies , the increasing slope is dominated by the op - amp voltage noise swing across the total circuit capacitance ( @xmath51 : dependent on frequency , equation [ i_tia ] ) .
we run a set of measurements in order to determine some of the involved unknowns : 1 .
op - amp voltage noise ( @xmath22 ) : we measured the op - amp voltage noise on a separate sample of the el5135 , by driving the op - amp as a voltage follower with grounded input . a low - noise 10x amplifier in series with the op - amp output was necessary for a voltage noise measurement above the spectrum analyzer noise floor .
the voltage noise level measured was @xmath65 at 10mhz , which is significantly higher than the specified @xmath66 .
2 . photodiode and feedback capacitances ( @xmath28 , @xmath29 ) : using a lcr - bridge impedance measurement instrument , we measured the photodiode quadrant capacitance @xmath28 on a separate sample of the gap500q to be 3.2pf with a 5v reversed bias , which is higher than the nominal 2pf .
we also measured the capacitor @xmath28 of the mock - up board and the feedback capacitor @xmath29 to be 2.2pf ( nominally 2.4pf ) and 0.1pf , respectively .
these noise properties are higher than specified and may account for part of the excess noise . a direct measurement of the op - amp current noise @xmath21 , the op - amp input capacitance @xmath30 , and the stray capacitances @xmath67 of the circuit involve the development of dedicated electronic boards ( currently on - going ) for well - controlled measurements .
these measurements will be conducted at a later stage .
however , it is possible to obtain an estimate of these values by fitting them as parameters of the noise model ( equations [ i_noise ] and [ it ] ) to the two measured data sets . the noise level difference ( offset ) in the data of the gap500q photoreceiver and the mock - up tia is an indicator of a significant excess current noise contribution , @xmath68 , present in the photo - measurement and not in the measurement of the mock - up tia .
this can also be included into the fit by considering the following non - frequency dependent contributions @xmath38 ( equation [ it ] ) for each case : @xmath69 the capacitance values @xmath70 and @xmath71 are assumed to be known from the lcr - bridge measurements .
we also assume similar op - amp and board noise properties ( @xmath22 , @xmath21 , @xmath72 ) for the two circuits . from the fit
, we obtained an op - amp current noise of @xmath73 ( about 20% higher than nominal @xmath74 ) and a combined stray plus op - amp input capacitance @xmath75pf ( 1pf nominal @xmath30 ) .
we also fit a common op - amp voltage noise for the two data sets to be @xmath76 , which is comparable ( within @xmath77 ) to the independent measurement ( @xmath78 ) , but about 25% higher than nominal ( @xmath66 ) .
table [ fitvalues ] summarizes the current best estimates ( cbe ) of the photoreceiver noise properties .
c | c c c + parameter & nominal & cbe & method + + @xmath79 $ ] & 2.0 & 3.2 & measured + + @xmath80 $ ] & 1.5 & 1.9 & fit + + @xmath81 $ ] & 0.9 & 1.1 & fit + + @xmath82 $ ] & 1.0 & 1.3 & fit + + @xmath83 $ ] & - & 0.7 & fit + + the excess current noise contribution , @xmath68 , present in the gap500q photoreceiver data was determined to be of the order of @xmath84 .
we have reversed - biased the photodiode with a battery power supply , therefore , noise on the bias voltage translating to excess current noise is not the cause .
additional tests are required to determine the origin of this contribution .
we have presented the results of noise measurements conducted on different photoreceiver prototypes .
the measurements showed approximately 20% noise in excess of our goal between 1020mhz .
direct measurements of the op - amp voltage noise and the reverse - biased qpd quadrant capacitance evidenced noise levels higher than nominal , accounting for part of the excess noise at higher frequencies . by fitting the parameters of the noise model to the data , we obtained estimates for the combined stray plus op - amp input capacitance and the op - amp current noise @xmath21 , which was determined to be approximately 20% higher than nominal . significant excess current noise ( 50% of total )
@xmath68 was determined between photoconductive ( gap500q photoreceiver ) and electronic ( mock - up tia ) noise measurements .
additional testing is required to determine its origin .
the measured photoreceiver performance is of the order of @xmath62 below 10mhz , increasing up to @xmath64 at 20mhz with a 3db bandwidth of 34mhz .
however , the mock - up tia performs at a level of @xmath85 below 20mhz ( 10% higher than expected from the nominal model ) with a measured 3db bandwidth of 38mhz .
this suggests a significantly better performance of a real photoreceiver with the current tia design , depending upon clarification and , if viable , mitigation of the excess current noise @xmath68 .
in addition , as following steps , we plan to conduct spatial scanning of the photodiode surfaces , measurement of inter - quadrant cross - talk , and differential wavefront sensing angle measurements .
this research was supported in part by nasa contract atfp07 - 0127 .
f. guzmn cervantes is supported by an appointment to the nasa postdoctoral program at the goddard space flight center , administered by oak ridge associated universities ( orau ) through a contract with nasa .
we thank a. joshi , s. datta and j. rue for stimulating discussions .
jennrich o , lisa technology and instrumentation , class .
quantum grav . 26
shaddock d , ware b , halverson p , spero r , klipstein b , overview of the lisa phasemeter , _ aip conf proc .
_ 873 654 - 60 ( 2006 ) .
bykov i , esteban delgado j , garca marn a , heinzel g , danzmann k , lisa phasemeter development : advanced prototyping , _ j. phys .
_ 154 ( 2009 ) .
joshi a , rue j , and datta s , low - noise large - area quad photoreceivers based on low - capacitance quad ingaas photodiodes , _ ieee photonics technology letters _ , vol.21 , no.21 ( 2009 ) .
datasheet of intersil device * el5135 * : http://www.intersil.com/data/fn/fn7383.pdf datasheet of gpd optoelectonics device * gap500q * : http://www.gpd-ir.com/ diekmann c , phasenstabilisierung und -auslesung fr lisa , diploma thesis , leibniz universitt hannover , germany ( 2008 ) .
datasheet of thorlabs device * led1050e * : http://www.thorlabs.com/thorcat/16300/16388-s01.pdf | lisa will use quadrant photoreceivers as front - end devices for the phasemeter measuring the motion of drag - free test masses in both angular orientation and separation .
we have set up a laboratory testbed for the characterization of photoreceivers . some of the limiting noise sources have been identified and their contribution has been either measured or derived from the measured data .
we have built a photoreceiver with a 0.5 mm diameter quadrant photodiode with an equivalent input current noise of better than @xmath0 below 20mhz and a 3db bandwidth of 34mhz . |
leptogenesis @xcite is one of the most attractive scenarios to explain the origin of the observed matter - antimatter asymmetry of the universe .
it follows from the seesaw mechanism @xcite , which gives a natural and simple explanation to the small neutrino masses observed in neutrino experiments , and relies on the conversion of a lepton asymmetry to a baryon asymmetry thanks to the non - perturbative sphaleron processes @xcite . for a recent review on leptogenesis ,
see @xcite .
if neutrinos are massive majorana particles , then lepton number must be violated . since neutrino masses are observed experimentally to be tiny , a slightly broken lepton number symmetry , e.g. a global @xmath4 , could provide the explanation . in this case ,
small neutrino masses are not explained by a ` seesaw ' mechanism , but rather by a cancellation mechanism @xcite .
such a symmetry was introduced in the context of the @xmath5msm @xcite in @xcite to explain at the same time the kev scale of the dark matter sterile neutrino @xcite and the quasi - degeneracy of the heavier two rh neutrinos , supposed to explain the baryon asymmetry of the universe by means of leptogenesis via neutrino oscillations @xcite .
the work @xcite also made use of a slightly broken lepton symmetry to motivate large yukawa couplings with tev masses for the rh neutrinos , making them in principle accessible at the lhc .
one consequence of the slightly broken @xmath4 symmetry is the existence of two quasi - degenerate rh neutrinos .
this is interesting in the context of leptogenesis because it allows for an enhancement of the @xmath6 asymmetry parameter @xcite , and hence successful leptogenesis is possible much below the usually quoted bounds on the mass scale and on the reheat temperature of the universe after inflation assuming hierarchical rh neutrinos , @xmath7 @xcite .
the tension with the gravitino overproduction in msugra scenarios @xcite is thus relaxed .
the second consequence of the broken symmetry is the presence of large washout parameters , implying that the asymmetry will be completely independent of the initial conditions , i.e. both the initial number of rh neutrinos and any previously generated asymmetry , even taking into account flavor effects @xcite . in this paper
, we study in detail the mechanism of leptogenesis in the presence of a slightly broken @xmath4 symmetry within the supersymmetric ( susy ) and non - supersymmetric two rh neutrino ( 2rhn ) model @xcite .
note that the 2rhn model is physically equivalent to the @xmath0 decoupling limit ( @xmath8 ) @xcite .
a related study was performed in @xcite with three quasi - degenerate rh neutrinos , which have to be motivated by a larger symmetry group , e.g. @xmath9 . there , the focus was to find numerical examples where resonant leptogenesis was possible with at the same time phenomenological consequences like observable lepton - flavor - violating signals in the non - susy setup . here
, we introduce only a @xmath4 symmetry , and tackle the problem with only two rh neutrinos in both susy and non - susy cases .
proceeding analytically , we keep under control the full parameter space of the problem , and we do not focus on resonant leptogenesis , which corresponds to the maximal possible enhancement of the @xmath6 asymmetry for quasi - degenerate heavy neutrinos .
the 2rhn model implies one massless light neutrino and a reduction of the numbers of parameters compared to the model with three rh neutrinos from 18 to 11 , among which 7 ( 2 neutrino masses , 3 mixing angles and 2 @xmath6-violating phases ) are accessible in experiments . the lower number of parameters will allow us to have a perfect handle on the problem , and thus we will be able to derive an expression for the baryon asymmetry predicted by leptogenesis where the dependence on each parameter is simple .
in particular , the high - energy parameters will factorize from the low - energy ones and from flavor effects altogether .
this will make possible to study in detail the dependence of the predicted baryon asymmetry on the dirac and majorana @xmath6-violating phases as well as the unknown angle @xmath10 .
note that , since the high - energy phase will be required to be non - zero for successful leptogenesis , leptogenesis from exclusively low - energy @xmath6 violation @xcite is not viable here .
finally , we will use the maximal enhancement of the @xmath6 asymmetry in the resonant limit @xcite to obtain a constraint on one of the breaking parameters .
interestingly , this constraint implies that successful leptogenesis is incompatible with the possible observation of rh neutrinos at future colliders @xcite as well as sizable lepton - flavor - violating signals .
this conclusion holds also in the susy case , if we assume the scale of leptogenesis to be below @xmath11 gev to avoid the gravitino problem . in section 2
, we introduce the parametrization of the 2rhn model which we will use throughout the paper . in section 3 , we describe how the lepton number symmetry affects the structure of the neutrino yukawa matrix and the majorana mass matrix .
then , we discuss how to parametrize the small breaking of this symmetry . in section 4
, we turn to leptogenesis , and estimate the baryon asymmetry predicted in this model with a special emphasis on the role of the pmns phases .
we also constrain the size of the breaking parameters using the maximal enhancement of the @xmath6 asymmetry . in section 5 ,
we extend the results to the supersymmetric version of the model .
finally , we summarize our main results and conclude in section 6 .
consider the seesaw model with two rh neutrinos . in the basis
where the charged leptons and rh majorana neutrino mass matrices are both diagonal , the `` mass '' basis , the seesaw mass matrix is given by m_=-m_d m_m^-1m_d^t , where m_md_n = diag(m_1,m_2 ) .
this matrix can be diagonalized as u^m_u^d_=diag(m_1,m_2,m_3 ) , where @xmath12 is the pmns matrix
. we will adopt the parametrization @xcite @xmath13 where @xmath14 , @xmath15 and , neglecting statistical errors , we will use @xmath16 and @xmath17 , compatible with the results from neutrino oscillation experiments . moreover
, we will adopt the @xmath18 range @xmath19 @xcite .
neutrino oscillation experiments measure two neutrino mass - squared differences , @xmath20 ev@xmath21 and @xmath22 ev@xmath21 @xcite .
recall that with only two rh neutrinos the lightest active neutrino is massless .
therefore , if the neutrino mass hierarchy is normal , one has @xmath23 , @xmath24 and @xmath25 , whereas if it is inverted , @xmath26 , @xmath27 and @xmath28 .
let us now introduce the so - called casas - ibarra parametrization @xcite @xmath29 the @xmath30 matrix is a @xmath31 matrix which can be parametrized as @xcite @xmath32 in the normal and inverted hierarchy , respectively , and where @xmath33 is a complex parameter .
@xmath34 is a discrete parameter that accounts for a discrete indeterminacy in @xmath35 .
following @xcite ( see also @xcite ) in the limit of only two rh neutrinos , we introduce a global @xmath4 lepton number symmetry for the leptonic fields , lepton doublets @xmath36 and right - handed neutrinos @xmath37 .
the @xmath4 charge assignments are as follows : @xmath38 the charges of all the other standard model fields are zero .
note that we work in the basis where the mass matrix of the charged leptons is real and diagonal . from now on , we shall call the basis with @xmath39 the `` flavor '' basis . in this basis ,
the relevant terms in the lagrangian are @xmath40_{ij } \overline{\tilde{n}_i^c}\tilde{n}_j + h.c.\quad ( i=1,2,\quad \a = e,\m,\t),\ ] ] where @xmath41 and @xmath42 are the yukawa coupling matrix and the majorana mass matrix in the flavor basis , respectively .
first of all , let us consider the @xmath4 symmetry to be exact . in this case
, the allowed structure of @xmath41 is given by @xmath43 without loss of generality , by using the field redefinition of @xmath44 , the non - zero components @xmath45 can be taken real and positive . on the other hand , the allowed form of @xmath46 is @xmath47 where the mass parameter @xmath48 can be taken real and positive by field redefinitions of @xmath49 and @xmath50 .
it should be noted that , when @xmath4 is exact , the masses of the active neutrinos are exactly zero .
this is because lepton doublets carry non - zero charges , and hence a majorana mass term for them is forbidden .
let us now consider the effects of @xmath4 symmetry breaking .
such effects induce additional contributions to @xmath41 and @xmath46 : @xmath51 where @xmath52 are complex parameters with modulus of order one , @xmath53 is real and positive , @xmath54 is an arbitrary phase , and @xmath3 and @xmath55 are breaking parameters , which are supposed to be much smaller than one .
note that we will not deal here with the issue of the @xmath4 breaking mechanism , and hence just parametrize the breaking parameters as given .
we go now from the flavor basis to the mass basis for the rh neutrinos .
the majorana mass matrix @xmath42 is symmetric , and therefore can be diagonalized using a unitary matrix @xmath56 : @xmath57 thus , the flavor eigenstates @xmath39 are related to the mass eigenstates @xmath58 by @xmath59_{ij } n_j$ ] .
it is easy to show that , to first order in @xmath3 , the eigenvalues of @xmath42 are given by @xmath60 where @xmath61 .
so the degeneracy parameter @xmath62 is equal to the breaking parameter @xmath3 up to a factor @xmath63 supposed to be of order one . for simplicity
we will assume in the remainder of the paper that @xmath64 .
next , we turn to the neutrino yukawa matrix @xmath41 , which we want to express in the mass basis , i.e. @xmath65 . in a first approximation , we take @xmath66 , so that the mixing matrix of @xmath67 and @xmath68 is simply given by @xmath69 where the second matrix ensures that the two eigenvalues of the majorana mass matrix @xmath70 are positive . using the casas - ibarra parametrization eq .
( [ h ] ) , one can translate eq .
( [ tildef ] ) into [ condom ] |_2|~_h|_1| . where @xmath71 . from eq .
( [ omega ] ) for normal hierarchy , and choosing the ` ' sign and @xmath72 , one obtains in the limit @xmath73 @xmath74 note that the case ` + ' and @xmath75 would have given the same structure , though with different signs . on the other hand , for the two other choices , ` + ' and @xmath72 , and ` ' and @xmath75 , the columns 1 and 2 would have been exchanged .
the case of inverted hierarchy is obtained by exchanging the rows @xmath76 and @xmath77 .
it is easy to see now that the condition ( [ condom ] ) is indeed satisfied when @xmath73 , for ` ' and @xmath72 as well as ` + ' and @xmath75 , and the breaking parameter @xmath55 corresponds to _ h~14||^2 .
it is widely known that , in order to generate a baryon asymmetry in the early universe , one needs to satisfy the three sakharov s conditions @xcite .
baryon number violation is provided by the non - perturbative sphaleron processes @xcite .
the @xmath6 symmetry is violated in the decay of the heavy rh neutrinos @xcite @xmath78}\nonumber\\ & = & \frac{3}{16 \p ( h^{\dag}h)_{ii } } \sum_{j\neq i } \left\ { { \rm im}\left[h_{\a i}^{\star } h_{\a j}(h^{\dag}h)_{i j}\right ] \frac{\x(x_j / x_i)}{\sqrt{x_j / x_i}}\right.\nonumber \\ & & \hspace{3 cm } \left.+ \frac{2}{3(x_j / x_i-1)}{\rm i m } \left[h_{\a i}^{\star}h_{\a j}(h^{\dag}h)_{j i}\right]\right\ } , \end{aligned}\ ] ] where @xmath79 and [ xi ] ( x)= 23x . as for the third condition , namely departure from thermal equilibrium , it can be conveniently described by the so - called decay parameter @xmath80 , given by the ratio of the decay widths to the expansion rate when the rh neutrinos start to become non - relativistic at @xmath81 . in terms of yukawa couplings ,
the decay parameters can be written as k_i = v^2 m _ m_i(h^h)_ii , where @xmath82 is the equilibrium neutrino mass , given by @xcite @xmath83 with @xmath84 gev and @xmath85 .
notice that , using the casas - ibarra parametrization , eq .
( [ h ] ) , the decay parameters @xmath86 can be expressed as linear combinations of the neutrino masses @xcite [ k ] k_i = _ jm_jm_|_ji^2| .
we will assume in the following that the generation of asymmetry takes place at temperatures @xmath87 , in which case flavor matters in the leptogenesis process , and three flavors are distinguished , denoted @xmath88 @xcite .
important parameters in flavored leptogenesis are the flavored decay parameters , given by [ kflav ] k_i = v^2 m _ m_i|h_i|^2 . for quasi - degenerate rh neutrinos it will prove useful to calculate the sum of the decay parameters . using eqs .
( [ k ] ) and ( [ kflav ] ) with the matrix @xmath30 given in eq .
( [ omega ] ) fixing the sign to @xmath89 and @xmath72 compatible with the lepton symmetry , one obtains @xmath90 and @xmath91 the case of inverted hierarchy is obtained by changing the labels @xmath77 and @xmath76 . in the limit of large @xmath92 required by the symmetry , one obtains @xmath93 and @xmath94 which does not depend on the phase of @xmath95 , and where we defined the dimensionless quantity @xmath96/m_{\star}.\ ] ] it is interesting to notice that for a normal hierarchy the small entry @xmath97 makes possible a cancellation in @xmath98 such that it can be much smaller than @xmath99 ; the other two flavors can not be smaller than roughly @xmath100 .
on the other hand , when the hierarchy is inverted , a cancellation can occur in all flavors . for quasi - degenerate rh neutrinos and with a strong washout in each flavor
, the solution to the flavored boltzmann equations can be written in the form @xcite @xmath101 where @xmath102 is the so - called efficiency factor , given by @xcite @xmath103 where @xmath104 note that in the strong washout regime , for @xmath105 , the following approximation holds @xcite : @xmath106 as for the @xmath6 asymmetries , in the limit @xmath107 and for the particular @xmath30 matrix in eq .
( [ omega ] ) ] , they are given by @xmath108+{\rm im}\left[h_{\a 1}^{\star } h_{\a 2 } ( h^{\dag}h)_{2 1}\right]\right),\\ \ve_{2\alpha}&\simeq & \frac{3}{16 \p ( h^{\dag}h)_{22 } } \frac{1}{3\d_2 } \left({\rm i m } \left[h_{\a 1}^{\star } h_{\a 2}(h^{\dag}h)_{1 2}\right]+{\rm im}\left[h_{\a 1}^{\star } h_{\a 2 } ( h^{\dag}h)_{2 1}\right]\right),\end{aligned}\ ] ] where @xmath109 .
using the casas - ibarra parametrization eq .
( [ h ] ) the flavored @xmath6 asymmetries are given by @xmath110 in the limit of large @xmath92 and defining @xmath111 , one obtains @xmath112 where we defined another dimensionless quantity , [ f ] f_(m_2,m_3 , u_2,u_3 ) & & g_2m_2-m_3m_2+m_3 .
it is interesting to notice first that the @xmath6 asymmetry is independent of @xmath92 , and second that the high - energy contribution is a necessary ingredient to have a non - zero flavored @xmath6 asymmetry ; in other words , @xmath95 has to be complex .
hence , leptogenesis from exclusively low - energy phases @xcite is not possible in our model . we have pointed out earlier that @xmath113 can be much smaller than @xmath114 in certain situations , which could lead to a big enhancement of the predicted baryon asymmetry compared to an unflavored calculation .
however , it turns out that in all such cases the flavored @xmath6 asymmetry is suppressed as well , so that the final effects are never much larger than the typical enhancement of a factor three in the three - flavor regime , as we will show more precisely below .
since the lepton number symmetry implies @xmath115 , the asymmetry will be typically produced in the strong washout regime , with no dependence on the initial number of rh neutrinos and on any previously generated asymmetry . with an account of flavor effects ,
the second feature is only rigourously possible in the three - flavor regime , where the washout occurs in all directions in flavor space @xcite .
the only exception will be in some very marginal regions of the parameter space where huge flavor effects imply @xmath116 , even though @xmath117 , so that the strong washout and independence of the initial conditions are no longer guaranteed .
we will come back to this point at the end of the section .
since the final asymmetry will be almost exclusively produced in the strong washout , using eqs .
( [ nf ] ) , ( [ kappaapprox ] ) and ( [ epslargeo ] ) , one obtains the simple expression @xmath118 which depends on the fundamental quantities @xmath48 , @xmath119 , @xmath92 and @xmath120 in a very simple way . in terms of the breaking parameters @xmath3 and @xmath55 introduced in the last section , one has @xmath121 interestingly , all low - energy parameters ( neutrino masses , pmns matrix elements ) appear exclusively in the factor @xmath122 , and are thus decoupled from the high - energy parameters ( @xmath48 , @xmath119 , @xmath92 and @xmath120 ) .
actually , flavor effects altogether only appear in this factor , and we can thus easily estimate the maximal difference with an unflavored calculation , where we would have instead a factor @xmath123 on the other hand , one can find numerically the maximal factor in the flavored calculation : @xmath124 i.e. maximal flavor effects lead to a factor 56 enhancement of the final asymmetry both for normal and for inverted hierarchy . as we said earlier ,
larger effects are not possible here . finally , assuming a standard thermal history and accounting for the sphaleron conversion coefficient @xmath125 , the baryon - to - photon ratio can be calculated as [ etab ] _
b = a_sphn_b - l^fn_^rec 0.9610 ^ -2n_b - l^f , to be compared with the measured value @xcite [ etabobs ] _
b^cmb = ( 6.2 0.15)10 ^ -10
. it can be useful to make a numerical estimation of @xmath126 using eq .
( [ resultanal ] ) , with @xmath127 , and @xmath128 : @xmath129 this result tells us that a hierarchy in the breaking parameters @xmath130 is needed if one wants to relax the scale of leptogenesis . for instance , with the breaking parameters @xmath131 and @xmath132 , one obtains @xmath133 it can be easily seen from eq .
( [ resultanal ] ) that the scale of the rh neutrinos @xmath48 can be lowered if we decrease by the same factor @xmath3 . hence , with @xmath134 , we can reach the tev scale for the heavy neutrinos , making them at least in principle accessible at the lhc .
the @xmath6 asymmetry is enhanced in the degenerate limit inversely proportionally to @xmath119 , as can be seen from eq .
( [ epslargeo ] ) . however , this effect is not unlimited .
there is a maximal enhancement which leads to resonant leptogenesis @xcite .
the condition to be on the resonance is given by @xmath135 @xcite , where @xmath136 is the decay width of @xmath137 .
in our model and for the case of normal hierarchy , this condition can be conveniently translated into @xmath138 where we introduced the parameter @xmath139 as in @xcite to account for a controversy in the literature about whether it is allowed or not to reach the resonance @xcite . in the following
, we will use conservatively @xmath140 , so that the validity of eq .
( [ veia ] ) is ensured .
note that in the case of inverted hierarchy @xmath141 occurs at a value twice as large . in the following , we will assume that @xmath142 .
plugging the condition ( [ condres ] ) in eq .
( [ resultanal2 ] ) , we find for normal hierarchy @xmath143 where we used eq .
( [ maxfoverg ] ) in the second step . since we want to be consistent with the @xmath18 range of eq .
( [ etabobs ] ) , we find that @xmath144 in the case of inverted hierarchy , we find @xmath145 this implies @xmath146 remember that the lepton number symmetry implied @xmath73 , so that the allowed range in the case of inverted hierarchy is quite constrained ! note also that the case of inverted hierarchy is more constrained than normal hierarchy due to the smallness of the factor @xmath147 , never larger than @xmath2 .
this behavior is different from the 2rhn model in the hierarchical limit @xmath148 , where huge flavor effects make possible that inverted hierarchy yields almost the same bounds as normal hierarchy @xcite , contrary to the unflavored result @xcite .
the results of eqs .
( [ resnormal ] ) and ( [ resinverted ] ) show explicitly that , in our simple model with two rh neutrinos or when @xmath0 is decoupled , successful leptogenesis is only possible for relatively large values of the breaking parameter @xmath55 , especially in the case of inverted hierarchy .
this is particularly interesting in view of the possible observation of rh neutrinos at future colliders , as recently investigated in a number of papers @xcite . to have even a small chance of observing rh neutrinos
, some of the active - sterile mixing angles @xmath149 should not be much smaller than 0.01 . in our case , with the constraint from successful leptogenesis , we have at most @xmath150 for @xmath151 gev , which is much too small .
lepton - flavor - violating signals in the non - supersymmetric case under discussion are expected to be very suppressed for the same reason .
finally , we would like to present examples compatible with the constraints ( [ resnormal ] ) and ( [ resinverted ] ) which show the explicit dependence of @xmath126 on the angle @xmath10 and on the pmns phases @xmath152 and @xmath153 , which only appear in the factor @xmath154 , as already mentioned .
we present in fig .
[ fig : correlnormal ] two such examples , for a normal hierarchy of light neutrinos and two choices of @xmath155 , 0.2 and 0.02 . note that the results are compatible with our rough estimation in eq .
( [ estimation ] ) . in fig .
[ fig : correlinverted ] we display the case of inverted hierarchy .
one notices from the figures that the two @xmath6-violating phases in the pmns matrix only yield small corrections to the predicted baryon asymmetry .
when the hierarchy of light neutrinos is normal and for the maximal allowed value @xmath156 , these phases can change the final asymmetry by 40% .
when the hierarchy is inverted , the effect can be more than a factor three , but only in a very restricted region of the parameter space . before concluding the section , one comment concerning figs .
[ fig : correlnormal ] and [ fig : correlinverted ] is in order .
we found marginal regions where @xmath116 for some flavor @xmath157 , implying that the strong washout regime and thus eq .
( [ kappaapprox ] ) no longer hold . in these regions of dependence on the initial conditions
, we used the more general eq .
( [ kappa ] ) , which is also valid for @xmath158 in the case of thermal initial @xmath67 and @xmath68 abundances . in practice
, these regions of huge flavor effects on the washout are found very close to the ones where the asymmetry is maximal , and they can be seen in the left panels of figs . [ fig : correlnormal ] and [ fig : correlinverted ] . note that a very small @xmath159 as in the right panels forbids such huge suppressions of the washout .
more specifically , in the left panel of fig .
[ fig : correlnormal ] , @xmath160 can be as low as @xmath161 in the middle of the diagonal strips , implying some dependence on the initial conditions . in the left panel of fig . [
fig : correlinverted ] , @xmath160 can be as low as @xmath162 in the darker regions ( red to black ) , so the dependence on the initial conditions is even larger there .
but apart from these marginal regions , the strong washout holds , and one can safely use eq .
( [ resultanal ] ) .
we turn now to the supersymmetric version of the model . one can introduce four different @xmath6 asymmetries , which by supersymmetry are all equal : @xmath163}=-{\g(n_i\to \tilde{\ell}_{\alpha}\tilde{h}_u-\g(n_i\to \tilde{\ell}^{\dagger}_{\alpha } \bar{\tilde{h}}_u ) ) \over \sum_{\a}\left[\g(n_i\to \tilde{\ell}_{\alpha } \tilde{h}_u)+\g ( n_i\to \tilde{\ell}^{\dagger}_{\alpha } \bar{\tilde{h}}_u)\right ] } \nonumber\\ & = & -{\g(\tilde{n}_i\to \ell_{\alpha}\tilde{h}_u)-\g(\tilde{n}^{\star}_i\to \bar{\ell}_{\alpha } \bar{\tilde{h}}_u ) \over \sum_{\a}\left[\g(\tilde{n}_i\to \ell_{\alpha}\tilde{h}_u)+\g(\tilde{n}^{\star}_i\to \bar{\ell}_{\alpha } \bar{\tilde{h}}_u)\right]}=-{\g(\tilde{n}_i\to \tilde{\ell}_{\alpha}h_u)-\g(\tilde{n}^{\star}_i\to \tilde{\ell}^{\dagger}_{\alpha } h_u^{\dagger } ) \over \sum_{\a}\left[\g(n_i\to \tilde{\ell}_{\alpha } h_u)+\g ( \tilde{n}^{\star}_i\to \tilde{\ell}^{\dagger}_{\alpha } h_u^{\dagger})\right]}\nonumber \ , , \end{aligned}\ ] ] where @xmath164 , @xmath165 and @xmath166 denote sleptons , higgsinos and rh sneutrinos , respectively .
the @xmath6 asymmetries @xmath167 were calculated in @xcite to be @xmath168 g(x_j / x_i)+ \frac{2}{(x_j / x_i-1)}{\rm i m } \left[h_{\a i}^{\star}h_{\a j}(h^{\dag}h)_{j i}\right]\right\ } , \ ] ] where @xmath79 and [ xi ] g(x)= . in the limit @xmath107
we are interested in and for the particular @xmath30 matrix in eq .
( [ omega ] ) , it can be easily obtained that @xmath169+{\rm im}\left[h_{\a 1}^{\star } h_{\a 2 } ( h^{\dag}h)_{2 1}\right]\right),\\ \tilde{\ve}_{2\alpha}&\simeq & \frac{1}{8 \p ( h^{\dag}h)_{22 } } \frac{1}{\d_2 } \left({\rm i m } \left[h_{\a 1}^{\star } h_{\a 2}(h^{\dag}h)_{1 2}\right]+{\rm im}\left[h_{\a 1}^{\star } h_{\a 2 } ( h^{\dag}h)_{2 1}\right]\right),\end{aligned}\ ] ] where we recall that @xmath109 .
so the @xmath6 asymmetries are a factor of two larger than in the non - supersymmetric case , just like in the hierarchical limit @xmath170 .
let us see how the efficiency factor is affected by supersymmetry .
assuming maxwell - boltzmann distributions , rh neutrinos and rh sneutrinos follow exactly the same evolution .
the end result is that the source term in the boltzmann equation for the @xmath171 asymmetry ( see e.g. @xcite ) is a factor 4 larger than in the non - susy case , whereas the washout term is a factor 2 larger .
the reason is that , first , the @xmath6 asymmetry is a factor 2 larger , and second there are twice as many decay modes as in the non - susy case .
the latter of course affects both source and washout terms . finally , the equilibrium neutrino mass in eq .
( [ d ] ) is numerically different since @xmath172 , and @xmath173 , implying @xmath174 altogether , multiplying the @xmath6 asymmetry by a factor 2 , and replacing @xmath175 , the semi - analytical expression for the final asymmetry becomes [ cf .
( [ resultanal ] ) ] @xmath176 where @xmath177 and @xmath178 are the functions defined in eqs .
( [ f ] ) and ( [ g ] ) , but with the replacement @xmath179 .
note that the dependence on @xmath180 in this expression is extremely mild , @xmath181 .
the extension of the results presented in the last section is then straightforward . from the maximal enhancement of the @xmath6 asymmetry in the degenerate limit [ cf .
( [ condres ] ) ] , we obtain slightly modified bounds on the breaking parameters , @xmath182 in the case of normal hierarchy , and @xmath183 in the case of inverted hierarchy
. since the limits on the breaking parameters are very similar to the non - susy case , the active - sterile mixing angles are necessarily small here as well .
hence , the conclusion about the incompatibility of leptogenesis with the observation of rh neutrinos at colliders is still valid . in the case of lepton - flavor violation ,
the discussion is somewhat different since new diagrams contribute that are not suppressed by the small mixing between active and sterile neutrinos @xcite .
for example , the rate of @xmath184 will actually depend on the combination @xmath185 .
however , for the range of rh neutrino masses we are interested in to avoid the gravitino problem , i.e. @xmath186 gev , the yukawa couplings are too small ( @xmath187 ) to give an observable signal .
we have studied the mechanism of leptogenesis in the presence of a slightly broken lepton number symmetry .
two almost degenerate right - handed neutrinos result from the symmetry , with the small breaking parameter @xmath3 essentially describing the mass splitting , and @xmath55 fixing the size of the ( inverse of the ) washout .
two well - known consequences follow : first , the scale of leptogenesis can be as low as the electroweak scale .
second , the baryon asymmetry predicted through leptogenesis is independent of the initial conditions , i.e. both the initial number of rh neutrinos and any previously generated asymmetry . with only two heavy neutrinos , which is equivalent to the @xmath0 decoupling limit
, the model contains few unknown parameters in addition to the two breaking parameters : one rh neutrino mass , one ` high - energy ' phase , the angle @xmath10 and two @xmath6-violating phases in the pmns matrix .
the relatively low number of parameters has allowed us to have a perfect handle on the problem , and we have been able to obtain semi - analytical formulae for the final asymmetry , eqs .
( [ resultanal ] ) and ( [ resultanalsusy ] ) in the non - susy and susy case , respectively , which disclose a simple dependence on each one of these parameters . in particular , we have studied in detail the role of the pmns phases and @xmath10 ( see figs . [
fig : correlnormal ] and [ fig : correlinverted ] ) .
interestingly , the high - energy phase is required to be non - zero for successful leptogenesis , implying that leptogenesis from exclusively low - energy @xmath6-violation is not viable in this context . finally , we derived from successful leptogenesis and the maximal enhancement of the @xmath6 asymmetry in the resonant limit that the breaking parameter @xmath55 must be relatively large , @xmath188 for normal and @xmath2 for inverted hierarchy ( both in susy and non - susy cases ) . as a consequence ,
leptogenesis is not compatible with the observation of rh neutrinos at future colliders and with a sizable lepton - flavor violation signal .
the other breaking parameter , @xmath3 , can be much smaller , and actually needs to be so in order to have low - scale leptogenesis .
it is a pleasure to thank pasquale di bari , jrn kersten , and georg raffelt for useful discussions .
s. b. is grateful to the max - planck - institute for physics in munich where most of this work was done .
m. fukugita and t. yanagida , _ baryogenesis without grand unification _ , _ phys .
lett . _ * b174 * ( 1986 ) 45 .
p. minkowski , _
mu @xmath189 e gamma at a rate of one out of 1-billion muon decays ? _ , _ phys .
lett . _ * b67 * ( 1977 ) 421 . t. yanagida in _ workshop on unified theories , kek report 79 - 18 _ , p. 95
, 1979 .
r. barbieri , d. v. nanopoulos , g. morchio and f. strocchi , _ neutrino masses in grand unified theories _
* b90 * ( 1980 ) 91 .
r. n. mohapatra and g. senjanovic , _ neutrino masses and mixings in gauge models with spontaneous parity violation _
_ * d23 * ( 1981 ) 165 .
v. a. kuzmin , v. a. rubakov and m. e. shaposhnikov , _ on the anomalous electroweak baryon number nonconservation in the early universe _ , _ phys .
* b155 * ( 1985 ) 36 . s. davidson , e. nardi and y. nir , _ leptogenesis _ , _ phys . rept . _
* 466 * ( 2008 ) 105177 [ http://arxiv.org/abs/0802.2962 [ 0802.2962 ] ] .
d. wyler and l. wolfenstein , _ massless neutrinos in left - right symmetric models _ , _ nucl .
phys . _ * b218 * ( 1983 ) 205 . j. bernabeu , a. santamaria , j. vidal , a. mendez and j. w. f. valle , _ lepton flavor nonconservation at high - energies in a superstring inspired standard model _ , _ phys
* b187 * ( 1987 ) 303 . g. c. branco , w. grimus and l. lavoura , _ the seesaw mechanism in the presence of a conserved lepton number _ , _ nucl .
_ * b312 * ( 1989 ) 492 .
t. asaka , s. blanchet and m. shaposhnikov , _ the numsm , dark matter and neutrino masses _ , _ phys .
lett . _ * b631 * ( 2005 ) 151156 [ http://arxiv.org/abs/hep-ph/0503065[hep-ph/0503065 ] ] .
m. shaposhnikov , _ a possible symmetry of the numsm _ , _ nucl .
phys . _ * b763 * ( 2007 ) 4959 [ http://arxiv.org/abs/hep-ph/0605047 [ hep - ph/0605047 ] ] . t. asaka and m. shaposhnikov , _ the numsm , dark matter and baryon asymmetry of the universe _ , _ phys
. lett . _ * b620 * ( 2005 ) 1726 [ http://arxiv.org/abs/hep-ph/0505013[hep-ph/0505013 ] ] .
e. k. akhmedov , v. a. rubakov and a. y. smirnov , _ baryogenesis via neutrino oscillations _ , _ phys .
* 81 * ( 1998 ) 13591362 [ http://arxiv.org/abs/hep-ph/9803255[hep-ph/9803255 ] ] . j. kersten and a. y. smirnov , _ right - handed neutrinos at lhc and the mechanism of neutrino mass generation _ ,
_ * d76 * ( 2007 ) 073005 [ http://arxiv.org/abs/arxiv:0705.3221 [ hep - ph ] [ arxiv:0705.3221 [ hep - ph ] ] ] .
l. covi , e. roulet and f. vissani , _ cp violating decays in leptogenesis scenarios _ , _ phys .
_ * b384 * ( 1996 ) 169174 [ http://arxiv.org/abs/hep-ph/9605319[hep-ph/9605319 ] ] .
a. pilaftsis , _ cp violation and baryogenesis due to heavy majorana neutrinos _ , _ phys .
* d56 * ( 1997 ) 54315451 [ http://arxiv.org/abs/hep-ph/9707235[hep-ph/9707235 ] ] .
s. davidson and a. ibarra , _ a lower bound on the right - handed neutrino mass from leptogenesis _
_ * b535 * ( 2002 ) 2532 [ http://arxiv.org/abs/hep-ph/0202239[hep-ph/0202239 ] ] .
w. buchmuller , p. di bari and m. plumacher , _ cosmic microwave background , matter - antimatter asymmetry and neutrino masses _ , _ nucl .
_ * b643 * ( 2002 ) 367390 [ http://arxiv.org/abs/hep-ph/0205349 [ hep - ph/0205349 ] ] .
s. blanchet and p. di bari , _ flavor effects on leptogenesis predictions _ , _ jcap _ * 0703 * ( 2007 ) 018 [ http://arxiv.org/abs/hep-ph/0607330[hep-ph/0607330 ] ] .
s. antusch and a. m. teixeira , _ towards constraints on the susy seesaw from flavour- dependent leptogenesis _
, _ jcap _ * 0702 * ( 2007 ) 024 [ http://arxiv.org/abs/hep-ph/0611232[hep-ph/0611232 ] ] .
m. y. khlopov and a. d. linde , _ is it easy to save the gravitino ? _ , _ phys .
lett . _ * b138 * ( 1984 ) 265268 .
j. r. ellis , d. v. nanopoulos , k. a. olive and s .- j .
rey , _ on the thermal regeneration rate for light gravitinos in the early universe _ , _ astropart
. phys . _ * 4 * ( 1996 ) 371386 [ http://arxiv.org/abs/hep-ph/9505438[hep-ph/9505438 ] ] .
t. moroi , h. murayama and m. yamaguchi , _ cosmological constraints on the light stable gravitino _
* b303 * ( 1993 ) 289294 .
j. pradler and f. d. steffen , _ constraints on the reheating temperature in gravitino dark matter scenarios _ , _ phys .
_ * b648 * ( 2007 ) 224235 [ http://arxiv.org/abs/hep-ph/0612291[hep-ph/0612291 ] ] .
e. nardi , y. nir , e. roulet and j. racker , _ the importance of flavor in leptogenesis _
, _ jhep _ * 01 * ( 2006 ) 164 [ http://arxiv.org/abs/hep-ph/0601084[hep-ph/0601084 ] ] .
a. abada , s. davidson , f .- x .
josse - michaux , m. losada and a. riotto , _ flavour issues in leptogenesis _
, _ jcap _ * 0604 * ( 2006 ) 004 [ http://arxiv.org/abs/hep-ph/0601083[hep-ph/0601083 ] ] .
p. h. frampton , s. l. glashow and t. yanagida , _ cosmological sign of neutrino cp violation _ , _ phys .
_ * b548 * ( 2002 ) 119121 [ http://arxiv.org/abs/hep-ph/0208157[hep-ph/0208157 ] ] .
a. ibarra and g. g. ross , _ neutrino properties from yukawa structure _ , _ phys .
* b575 * ( 2003 ) 279289 [ http://arxiv.org/abs/hep-ph/0307051[hep-ph/0307051 ] ] .
a. ibarra and g. g. ross , _ neutrino phenomenology : the case of two right handed neutrinos _
* b591 * ( 2004 ) 285296 [ http://arxiv.org/abs/hep-ph/0312138[hep-ph/0312138 ] ] .
p. h. chankowski and k. turzynski ,
_ limits on t(reh ) for thermal leptogenesis with hierarchical neutrino masses _ , _ phys .
_ * b570 * ( 2003 ) 198204 [ http://arxiv.org/abs/hep-ph/0306059 [ hep - ph/0306059 ] ] .
a. pilaftsis and t. e. j. underwood , _ electroweak - scale resonant leptogenesis _ , _ phys .
* d72 * ( 2005 ) 113001 [ http://arxiv.org/abs/hep-ph/0506107[hep-ph/0506107 ] ] .
s. pascoli , s. t. petcov and a. riotto , _ connecting low energy leptonic cp - violation to leptogenesis _ , _ phys .
* d75 * ( 2007 ) 083511 [ http://arxiv.org/abs/hep-ph/0609125[hep-ph/0609125 ] ] .
s. pascoli , s. t. petcov and a. riotto , _ leptogenesis and low energy cp violation in neutrino physics _ , _ nucl .
phys . _ * b774 * ( 2007 ) 152 [ http://arxiv.org/abs/hep-ph/0611338[hep-ph/0611338 ] ] . g. c. branco , r. gonzalez felipe and f. r. joaquim , _ a new bridge between leptonic cp violation and leptogenesis _ , _ phys .
_ * b645 * ( 2007 ) 432436 [ http://arxiv.org/abs/hep-ph/0609297 [ hep - ph/0609297 ] ] .
e. molinaro , s. t. petcov , t. shindou and y. takanishi , _ effects of lightest neutrino mass in leptogenesis _ , _ nucl .
* b797 * ( 2008 ) 93116 [ http://arxiv.org/abs/0709.0413[0709.0413 ] ] .
a. anisimov , s. blanchet and p. di bari , _ viability of dirac phase leptogenesis _ , _ jcap _ * 0804 * ( 2008 ) 033 [ http://arxiv.org/abs/0707.3024[0707.3024 ] ] .
a. pilaftsis and t. e. j. underwood , _ resonant leptogenesis _ , _ nucl .
phys . _ * b692 * ( 2004 ) 303345 [ http://arxiv.org/abs/hep-ph/0309342[hep-ph/0309342 ] ] .
a. anisimov , a. broncano and m. plumacher , _ the cp - asymmetry in resonant leptogenesis _ , _ nucl .
* b737 * ( 2006 ) 176189 [ http://arxiv.org/abs/hep-ph/0511248[hep-ph/0511248 ] ] .
f. del aguila , j. a. aguilar - saavedra , a. martinez de la ossa and d. meloni , _ flavour and polarisation in heavy neutrino production at e+ e- colliders _ , _ phys .
lett . _ * b613 * ( 2005 ) 170180 [ http://arxiv.org/abs/hep-ph/0502189[hep-ph/0502189 ] ] .
t. han and b. zhang , _ signatures for majorana neutrinos at hadron colliders _ , _ phys .
* 97 * ( 2006 ) 171804 [ http://arxiv.org/abs/hep-ph/0604064[hep-ph/0604064 ] ] .
f. del aguila , j. a. aguilar - saavedra and r. pittau , _ heavy neutrino signals at large hadron colliders _ , _ jhep _ * 10 * ( 2007 ) 047 [ http://arxiv.org/abs/hep-ph/0703261[hep-ph/0703261 ] ] . w
_ , _ review of particle physics _ , _
journal of physics g _ * 33 * ( 2006 ) 1 + .
m. c. gonzalez - garcia and m. maltoni , _ phenomenology with massive neutrinos _ , _ phys .
_ * 460 * ( 2008 ) 1129 [ http://arxiv.org/abs/0704.1800[0704.1800 ] ] .
j. a. casas and a. ibarra , _ oscillating neutrinos and mu @xmath189 e , gamma _ , _ nucl .
_ * b618 * ( 2001 ) 171204 [ http://arxiv.org/abs/hep-ph/0103065[hep-ph/0103065 ] ] .
s. t. petcov , t. shindou and y. takanishi , _ majorana cp - violating phases , rg running of neutrino mixing parameters and charged lepton flavour violating decays _ , _ nucl .
phys . _ * b738 * ( 2006 ) 219242 [ http://arxiv.org/abs/hep-ph/0508243[hep-ph/0508243 ] ] .
a. d. sakharov , _ violation of cp invariance , c asymmetry , and baryon asymmetry of the universe _
, _ pisma zh . eksp .
fiz . _ * 5 * ( 1967 ) 3235 .
w. buchmuller , p. di bari and m. plumacher , _
leptogenesis for pedestrians _ , _ ann .
* 315 * ( 2005 ) 305351 [ http://arxiv.org/abs/hep-ph/0401240[hep-ph/0401240 ] ] .
m. fujii , k. hamaguchi and t. yanagida , _ leptogenesis with almost degenerate majorana neutrinos _
rev . _ * d65 * ( 2002 ) 115012 [ http://arxiv.org/abs/hep-ph/0202210[hep-ph/0202210 ] ] .
s. blanchet and p. di bari , _ leptogenesis beyond the limit of hierarchical heavy neutrino masses _ , _ jcap _ * 0606 * ( 2006 ) 023 [ http://arxiv.org/abs/hep-ph/0603107[hep-ph/0603107 ] ] .
di bari , _ leptogenesis , neutrino mixing data and the absolute neutrino mass scale _ , http://arxiv.org/abs/hep-ph/0406115 [ hep - ph/0406115 ] . g. f. giudice , a. notari , m. raidal , a. riotto and a. strumia , _ towards a complete theory of thermal leptogenesis in the sm and mssm _ , _ nucl .
* b685 * ( 2004 ) 89149 [ http://arxiv.org/abs/hep-ph/0310123[hep-ph/0310123 ] ] .
g. engelhard , y. grossman , e. nardi and y. nir , _ the importance of n2 leptogenesis _ , _ phys .
* 99 * ( 2007 ) 081802 [ http://arxiv.org/abs/hep-ph/0612187[hep-ph/0612187 ] ] .
collaboration , e. komatsu _ et .
_ , _ five - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological interpretation _ , http://arxiv.org/abs/0803.0547[0803.0547 ] .
s. blanchet and p. di bari , _ new aspects of leptogenesis bounds _ , http://arxiv.org/abs/0807.0743[0807.0743 ] .
s. t. petcov , w. rodejohann , t. shindou and y. takanishi , _ the see - saw mechanism , neutrino yukawa couplings , lfv decays l(i ) @xmath189 l(j ) + gamma and leptogenesis _ , _ nucl .
* b739 * ( 2006 ) 208233 [ http://arxiv.org/abs/hep-ph/0510404[hep-ph/0510404 ] ] . s. antusch , s. f. king and a. riotto , _ flavour - dependent leptogenesis with sequential dominance _ , _ jcap _ * 0611 * ( 2006 ) 011 [ http://arxiv.org/abs/hep-ph/0609038[hep-ph/0609038 ] ] .
a. abada , p. hosteins , f .- x .
josse - michaux and s. lavignac , _ successful leptogenesis in so(10 ) unification with a left- right symmetric seesaw mechanism _ , http://arxiv.org/abs/0808.2058[0808.2058 ]
. f. borzumati and a. masiero , _ large muon and electron number violations in supergravity theories _ , _ phys .
* 57 * ( 1986 ) 961 . | seesaw models with a slightly broken lepton number symmetry can explain small neutrino masses , and allow for low - scale leptogenesis .
we make a thorough analysis of leptogenesis within the simplest model with two right - handed ( rh ) neutrinos ( or with @xmath0 decoupled ) .
we obtain a semi - analytical formula for the final asymmetry in both supersymmetric and non - supersymmetric cases with a simple dependence on each parameter .
the low - energy parameters factorize from the high - energy ones , and the high - energy phase must be non - zero . the role of the pmns phases is carefully studied . moreover , we find that the breaking parameter in the yukawa coupling matrix must be relatively large , @xmath1 for normal and @xmath2 for inverted hierarchy
. therefore , leptogenesis in our simple model is incompatible with rh neutrino signals at future colliders or sizable lepton - flavor violation .
the other breaking parameter , @xmath3 , which appears in the rh neutrino mass matrix , can be much smaller , and actually needs to be so in order to have low - scale leptogenesis .
c i u
# 1 # 1#1 # 1 # 1#2#3phys .
lett .
* b # 1 * ( # 2 ) # 3 # 1#2#3nucl .
phys .
* b # 1 * ( # 2 ) # 3 # 1#2#3phys .
rev .
lett . *
# 1 * ( # 2 ) # 3 # 1#2#3phys . rev . * d # 1 * ( # 2 ) # 3 # 1#2#3z .
phys . * c # 1 * ( # 2 ) # 3 # 1#2#3class . and quantum grav . *
# 1 * ( # 2 ) # 3 # 1#2#3commun .
math .
phys . *
# 1 * ( # 2 ) # 3 # 1#2#3j . math .
phys . *
# 1 * ( # 2 ) # 3 # 1#2#3ann . of phys .
* # 1 * ( # 2 ) # 3 # 1#2#3phys .
rep . * # 1c * ( # 2 ) # 3 # 1#2#3progr .
theor .
phys . *
# 1 * ( # 2 ) # 3 # 1#2#3int .
j. mod .
phys . * a # 1 * ( # 2 ) # 3 # 1#2#3mod .
phys .
lett . * a # 1 * ( # 2 ) # 3 # 1#2#3nuovo cim .
* # 1 * ( # 2 ) # 3 # 1#2#3_ibid .
_ * # 1 * ( # 2 ) # 3 |
the standard model is not a fundamental theory and there have been many proposals to go beyond it .
it is expected that a field theory with extra dimensions arises as the low energy limit of a fundamental string theory .
for this reason extra dimensions are a common feature of any theory valid at high energies .
almost all problems of particle physics have been reformulated in this context giving new possibilities .
in particular , physics of large extra dimensions can provide solutions to the hierarchy problem @xcite , @xcite and new mechanism of symmetry breaking @xcite , that can be tested in the next high energy experiments .
in this note we developed a formalism that allows us to compute loop corrections in field theories with large extra dimensions , separating uv divergencies from finite contributions , in a direct product space .
usually , higher dimensional theories are formulated using kaluza - klein decomposition .
instead in this work we will use winding modes . in a five dimensional theory these modes are obtained propagating around the circle of the extra dimension .
two paths with different windings are topologically different .
ultraviolet divergencies in loop integrals are associated to the zero winding mode .
non zero modes are long distance and each of them will give finite terms @xcite . then this formalism is very useful to separate finite from divergent contributions .
we will use winding decomposition to compute radiative corrections on 5d and 6d theories .
when the theory is compactified on an orbifold translation symmetry is broken on the fixed points ; we will show how localized terms are generated at one loop .
we also analyze the role of localized kinetic terms , parallel and perpendicular to the brane , interpreting them in a very intuitive way , and showing their physical effects .
we show these terms can provide new mechanisms of symmetry breaking .
we will compute general finite masses and also self and gauge couplings using this method . when there are more than one extra dimensions , winding modes prove to be very useful .
in particular , we consider a 6d theory and after obtaining the propagator in mixed representation , we use it to compute radiative corrections , the finite piece of the masses and the couplings . as an interesting application we will apply this formalism to compute a two loop effective potential .
we will prove that this potential can stabilize large extra dimensions when there are brane terms . in section 2
we define winding modes working on a mixed momentum - coordinate representation .
we apply this idea to a 5d theory in section 3 . in section 4
we show how to work with more than one extra dimension . in section 5
we compute a two loop effective potential and section 6 is for conclusions . during the writing of this work
another paper appeared @xcite on similar subjects , reaching the same results about mixed momentum - coordinate representation and mass terms .
let s consider a 5d theory compactified on @xmath0 , where @xmath1 is 4d minkowsky space and @xmath2 is a compact 1-d manifold .
if we can write @xmath3 , where @xmath4 is the real line and @xmath5 is a discrete group acting freely on @xmath4 , then we can define winding modes .
the simplest example is @xmath6 ( a circle ) and @xmath7 ( the set of integer numbers , with the sum defined as the group product ) . in this case
we obtain the compact space identifying @xmath8 . due to the identification @xmath9 , the index @xmath10 labels winding modes .
this idea suggests the following procedure : we compute on the infinite space and identify @xmath8 to obtain the physical magnitudes on the compact space . for a general compact space @xmath11 of dimension @xmath12 ( @xmath13 a non compact m - dimensional manifold ) , we identify @xmath14 , @xmath15 and @xmath16
. then we can define winding modes for every @xmath17 , and associate divergencies to the zero mode @xmath18 . to show how this formalism works we will compute the massless scalar propagator on euclidean 5d spacetime .
we work on a mixed representation @xmath19 , where @xmath20 is the 4d momentum and @xmath21 the coordinate in the extra dimension .
then we can obtain the green s function from the following equation @xmath22 where we have fourier transform to momentum space only on the 4d space , and @xmath23 . solving this equation
we get @xmath24 due to translation invariance we see that the propagator only depends on @xmath25 . if we consider a massive scalar field we just have to replace @xmath26 . to get the propagator on the compact space we identify @xmath27 and sum over windings @xmath28 we can solve this sum , but it is more useful to consider the contribution of each mode . in the last equation we can see that for @xmath29 the propagator
is exponentially damped , therefore loop integrals over momenta will be finite .
on the other hand , for @xmath30 , the propagator goes as @xmath31 when evaluated in @xmath32 . for these reasons , compactifying on a circle , we will obtain divergent contributions for the winding 0-modes , and finite contributions for the other modes . and @xmath33 .
in the vertical axis we represent the number of windings.__,width=302,height=56 ] therefore , using the winding modes formalism , is very easy to separate divergent from finite contributions .
the divergent terms are associated to short distances , that is why the divergencies resulting from the 0-mode are the same as the divergencies of the uncompactified theory . for the masses ,
the radiative corrections are dominated by high energy effects .
high energy implies small distances , then for the masses , the small @xmath10 contributions dominate over large @xmath10 .
this is not the case when we do kaluza - klein decomposition .
this is one of the advantages of winding decomposition .
* orbifold compactification + orbifolds are used to obtain 4d chiral fermions from a higher dimensional theory . in general
we can obtain an orbifold with a discrete group @xmath34 acting non freely on the compact space @xmath35 .
points of @xmath35 left invariant by @xmath34 are fixed points , and there @xmath36 is singular .
the simplest example is the orbifold @xmath37 , where @xmath38 is the parity transformation on the extra dimension . due to the identification
@xmath39 $ ] and the fixed points are @xmath40 . to give a complete description we also have to specify field parities . *
we first consider a scalar field on @xmath37 , with definite parity @xmath41 . due to the identification @xmath42
, we can propagate from @xmath21 to @xmath33 and from @xmath21 to @xmath43 , then the propagator is @xcite @xmath44 where @xmath45 $ ] , and @xmath46 depends on the field parity .
this propagator depends on @xmath47 due to the breaking of translation invariance . in this equation
we can see that the propagator goes as @xmath31 for the following limits of the coordinates and windings : @xmath48 and @xmath49 .
then we expect divergencies localized on the fixed points of the orbifold @xcite .
we can expect this terms because they do nt break any symmetry of the theory .
we compute one - loop radiative corrections of the scalar interacting theory defined by @xmath51.\ ] ] we will consider our toy model an effective theory valid to some cut - off @xmath52 , that by naive dimensional analysis should be @xmath53 .
furthermore , the lagrangian in eq .
( [ teorialambdaphi4 ] ) shall be defined as the effective theory at the scale @xmath52 . in performing quantum corrections
we will cut off the 4d momentum integral at the scale @xmath52 .
the loop contributions that depend on @xmath52 will signal the divergencies of the 5d theory . in fig .
[ fig : lazos5d ] we can see the feynman diagrams that renormalize the two point function and the coupling at one loop .
then we can write the effective action with one - loop quantum corrections as @xmath54 let s consider first , the extra dimension compactified on a circle , then @xmath55 . for the circle
@xmath56 is constant and is given by ( the one - loop mass was calculated in ref .
@xcite using k - k decomposition ) @xmath57 where we have thrown away terms that cancel when @xmath58 .
as we argued in the previous section , divergencies are associated to @xmath30 and finite terms to @xmath59 .
this expresion is valid for general @xmath60 , because @xmath56 is independent of @xmath21 .
if there is a symmetry ( like supersymmetry or a gauge symmetry ) prohibiting divergent masses , then the finite term is a prediction of the theory .
now we compute @xmath61 at one loop . for simplicity , we will take @xmath62constant . according to this , expanding in powers of external momenta
we just keep the zero order terms .
then we get for @xmath61 @xmath63 there are two propagators involved , so there are two winding indexes . when the topology of the extra space is more complicated ( for example when there are more than one extra dimension ) it is useful to express this equation in terms of just one propagator as @xmath64 where @xmath65 is given by ) and eq .
( [ i ] ) are the same we can write eq .
( [ loop ] ) in k - k modes without external momenta as @xmath66 , with @xmath67 the momentum in the extra dimension .
the integrand can be written as @xmath68 , and by fourier transformation we obtain eq .
( [ i ] ) . ]
@xmath69 integrating coordinates and momenta we get the desired result , and the one - loop contribution to @xmath70 is @xmath71 the sum over @xmath10 is logarithmically divergent in the ir , so we have to introduce an ir cut - off , that means that we sum to @xmath72 , regulating the long distance behavior .
if the field is massive the mass is the natural cut - off .
the ir logarithm is the same as in 4d , and this can be easily understood in terms of k - k modes .
the zero k - k mode is massless , and this is the mode that propagates long distances .
the one - loop contribution to @xmath73 on @xmath37 are generated by the following expression ( in ref .
@xcite are given the masses for k - k modes ) @xmath74 where @xmath75 is the same as @xmath73 for the circle , except that @xmath76 .
the second term depends on @xmath21 and it s new .
as we argued , it has divergencies for @xmath77 . to obtain the divergencies we expand @xmath78 in powers of @xmath21 around the fixed points @xmath79 . expanding @xmath80 to second order ,
@xmath81 is given by @xmath82 + \partial_y^2\phi^2(y_{fp } ) \frac{\log(\lambda r)}{16}\big\}.\ ] ] the @xmath83 divergencies are associated to @xmath30 when @xmath84 and to @xmath85 when @xmath86 .
then these divergencies are localized on the fixed points of the orbifold .
this is because the orbifold compactification breaks translation invariance in these points . from eq .
( [ masaz1 ] ) we can see that @xmath56 has two contributions .
the first one is the same as in the circle , eq .
( [ 0masas1 ] ) .
the second contribution depends on @xmath21 and , from eq .
( [ s2z ] ) , we can see the divergent terms .
we split @xmath56 in divergent and finite contributions as @xmath87 .
then the divergent contribution is given by @xmath88\big\}.\ ] ] the finite term can be computed for a constant field , and is given by @xmath89,\;\;\;\;\;\;\;\;\;\;\phi=\phi_c.\ ] ] to get @xmath61 , we again expand the fields in powers of @xmath21 around the fixed points @xmath79 .
we just take the zero order term in the power series and use eqs .
( [ i ] ) and ( [ simple2 ] ) with the orbifold propagator . then @xmath61 is given by @xmath90.\ ] ]
the linear uv divergence is due to the zero winding mode .
the logarithmic divergence is localized on the fixed points , then it is 4d , and is associated to @xmath77 .
therefore we can write @xmath61 as @xmath91 @xmath92,\ ] ] where @xmath93 is a finite coupling . for a constant field @xmath94
is given by @xmath95,\;\;\;\;\;\;\;\;\;\;\phi=\phi_c.\ ] ] if the field is even there are logarithmic ir divergences , as in the circle , but this does nt happen in the odd case .
this is easier to understand in terms of k - k modes : only the even field has a massless mode . in dimensional regularization scheme , the effective scale @xmath52 in the 4d logarithm of the coupling , eq .
( [ loopsz ] ) , is replaced by an arbitrary scale @xmath96 .
further discussions in this scheme can be found in ref .
@xcite .
we have seen that new terms localized on the branes have been induced , showing that these terms should be taken from the begining . here
we want to consider the effect of these terms on the physical parameters .
we are interested in the kinetic terms : they can be parallel @xmath97 or perpendicular @xmath98 to the brane ( see @xcite for the most general case ) , where @xmath99 are the couplings .
perpendicular kinetic terms generate classical divergencies , similar to classical divergencies in electromagnetism .
they can be regularized with a fat brane , for example defining a @xmath100 when @xmath101 . then
, including the right counterterms , we can renormalize the theory ( for classical renormalization with branes of codimension bigger than one see @xcite ) .
the renormalization can be performed in the following way : we obtain the propagator @xmath102 ( with perpendicular terms ) as a perturbative series with @xmath103-vertex insertions , as is shown in fig .
[ figperturbaciones ] .
every term of order @xmath10 , has divergent contributions of order @xmath104 , where @xmath105 is a brane scale . to cancel the divergent terms when @xmath106 , we have to add counterterms proportional to @xmath107 , where @xmath10 is the number of vertex insertions .
the divergencies can be interpreted as contributions coming from processes at energy @xmath105 on the branes .
then , including these counterterms , we are neglecting high energy contributions that can feel the brane structure . but
this is exactly what we want , an approximation valid for energies @xmath108 .
after that , in the theory there are only parallel terms . to see this we can resum the perturbative series ( without the divergent terms ) @xcite and get @xmath109 where @xmath110 is the free propagator .
then @xmath102 is the same propagator as @xmath111 ( with parallel kinetic terms ) changing @xmath112 , as we will see in the following paragraphs , eq .
( [ gkinetic ] ) .
so there is nothing new considering perpendicular couplings , we can obtain all the relevant information analyzing parallel terms .
for these reasons we will concentrate on parallel kinetic terms . * parallel kinetic terms + let s consider a scalar interacting theory @xmath113.\ ] ] to obtain the propagator we apply the ideas of the previous sections .
for the moment we suppose that the extra dimension is infinite and compute the propagator with just one delta on one of the fixed points @xmath114 . in this case
the green s function equation in mixed momentum - coordinate representation is @xmath115 this equation is similar to the free one but with a new source of magnitude @xmath116 in @xmath114 .
then the propagator is @xmath117 it is immediate to read the second term as a reflection of magnitude @xmath118 on the brane on @xmath79 .
once we realize this , we can compute the propagator in the compact space in a perturbative way ( in @xcite we can see the series with mass insertions ressumed , but we want to keep our intuition working with winding modes ) .
we have to sum over all the contributions coming from reflections on @xmath119 , in a similar way we do with light travelling in a medium with different indexes of reflection @xmath120 and transmission @xmath121 .
a wave of amplitude @xmath122 arrives to a @xmath123 in @xmath124 , there @xmath125 is reflected and @xmath126 is transmited , and due to the propagation the wave amplitude is damped @xmath127 after travelling a distance @xmath128 . with these rules we can obtain the propagator to any order in @xmath129 .
* we will study the limit of @xmath130 for fixed 4d coupling @xmath131 .
the propagator in the orbifold ( with 4d kinetic term canonically normalized ) is given by @xmath132,\ ] ] where @xmath133 is given in eq .
( [ gkinetic ] ) . for simplicity
we consider first a kinetic term on @xmath134 .
the propagator of eq .
( [ gkinetic ] ) has two regimes depending on wether @xmath135 or @xmath136 .
we define a critic winding @xmath137 , then @xmath138 corresponds to the high energy regime and @xmath139 to the low energy . therefore the propagator is given by @xmath140 thus we see that at short distances the field is similar to an odd field . when @xmath141 @xmath142 is more complicated , but for a rough estimation we can consider eq .
( [ gkinetic2 ] ) valid for all @xmath10 .
it is important to note that the two pieces of equation ( [ gkinetic2 ] ) are sensible to different values of momenta .
* effect of parallel terms on the physical parameters + once we have the propagator we can compute the mass and selfcoupling at one - loop with brane kinetic terms . repeating the steps done in the previous section we calculate the one - loop mass .
we consider a constant field @xmath143 , and integrating over the extra dimension we obtain the 4d mass .
there are divergent and finite contributions . here
we show the finite terms , splitted in two contributions , depending on the winding values @xmath144 , \end{aligned}\ ] ] where the subindex @xmath46 is for the even and odd contribution .
then eq .
( [ map ] ) gives the one - loop finite contribution to the 4d mass , with brane kinetic terms .
* we can compare the finite 4d mass term with brane kinetic couplings with the one without them . to do this
we consider that the 4d couplings of both theories are the same ( for constant field @xmath143 ) .
then we get @xmath145,\;\;\;\ ; \;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;a\rightarrow \infty , \end{aligned}\ ] ] where @xmath146 and @xmath147 in eq .
( [ m4d ] ) we have put the first correction in powers of @xmath148 .
the odd field does nt coupled to the brane at @xmath149 , that s why the mass does nt change for this mode .
now we compute the vertex with constant field @xmath143 .
first we discuss the high energy regime . as we said before , in this regime the field seems an odd field .
then we get the linear and logarithmic dependence on @xmath52 in the same way we did in the previous section .
second , we consider the low energy regime that corresponds to windings @xmath150 in both propagators involved in the feynman diagram .
this case is the same as the one without brane terms , but summing over windings @xmath150 .
if the field is even under @xmath151 , there is an ir logarithm .
then , the finite and logarithmic one loop contribution to the 4d coupling are given by @xmath152,\ ] ] where @xmath153 and the first logarithm is the ir long distance .
for an odd field we get a similar result without the ir logarithm .
let s discuss now what happen with brane kinetic terms in @xmath154 . in this case
the limit of @xmath155 corresponds to an opaque brane , and the reflection has a minus sign .
then the effect of this brane is again the same as considering an odd field .
so we can get an odd field puting ( almost ) opaque branes .
this suggests a new way o symmetry breaking : let s suppose that the components of a multiplet have different brane couplings .
then the effect of these couplings will be the same as choosing different boundary conditions for the fields of a given multiplet , breaking the symmetry under which the multiplet transforms .
we consider as an application of the previous formalism , a 5d theory with gauge fields and a scalar charged field , transforming with representation s. we want to get the logarithmic divergencies of the gauge coupling , due to the scalar fields , in an orbifold . then we consider the one loop scalar contribution to the vacuum polarization @xmath156 .
this have been computed with k - k modes @xcite , here we get the same result with winding modes .
we consider the effective 4d theory , that is the theory obtained after integration over the extra dimension with constant fields ( zero k - k modes ) .
we define @xmath157 as the effective 4d coupling of an abelian theory , at the scale @xmath52 .
then , after some manipulations , we can write the divergent part of the one - loop vacuum polarization as @xmath158 the momentum integral is a loop with two scalar propagators without external momenta , then it is the same as the one loop @xmath159 contribution . on the r.h.s . of eq .
( [ pi ] ) we have written the vacuum polarization in terms of one propagator , as we did for the scalar coupling .
the gauge coupling of the effective 4d theory is at the one loop level given by @xmath160,\ ] ] where @xmath161 in eq .
( [ gl ] ) , as in eq .
( [ loopsz ] ) for the one - loop scalar coupling , the @xmath162 comes from brane effects .
if the gauge group is non - abelian , then we only have to modify the charges and multiply by @xmath163 , where @xmath164=t(s){\delta}_{ab}$ ] .
( [ gl ] ) gives the scalar contribution to the 4d effective coupling at one loop , for a theory with a cut - off scale @xmath52 .
we can consider a theory with a different cut - off @xmath165 , and one - loop coupling @xmath166 .
then the relation between the couplings @xmath17 and @xmath166 is given by @xmath167.\ ] ]
we apply winding modes formalism to a space with two extra - dimensions . given an infinite plane @xmath168 we can obtain a two - dimensional torus @xmath169 identifying with @xmath170 ( @xmath171 the integer numbers ) . the identification is @xmath172 , where @xmath173 and @xmath174 , @xmath175 .
@xmath176 and @xmath177 meassure the size of the extra dimensions compactified in a torus , @xmath178 .
there is one parameter more to obtain a complete description of @xmath169 , the angle betwen the directions of identification in the plane , as shown in fig .
[ id6d ] . to obtain a bidimensional torus @xmath169 , with parameters @xmath179.__,width=264 ] in the same way as in 5d
we can find the scalar green s function in euclidean infinite space , in mixed representation ( momenta in minkowski directions and coordinates in extra directions ) @xmath180 where @xmath23 is the 4d momentum and @xmath181 numbers the extra dimensions .
the solution to this equation is @xmath182 where @xmath183 is @xmath184 bessel function of zero order .
now we identify and get the propagator in the compact space @xmath185 with @xmath186 the modulus of the vector messured with the flat metric of a torus @xmath187 as in the 5d case , the propagator for non - zero winding is exponentially damped at high energies @xmath188 @xmath189 showing that winding contributions will be always finite . for arbitrarily small argument @xmath190 .
then the propagator diverges at short distances in the extra dimensions . therefore to compute feynman integrals with @xmath191 we have to regulate the propagator , where we see that evaluating @xmath192 we get logarithmic divergences .
we can regulate the short distance behaviour with an uv cut - off @xmath193 , then the propagator becomes @xmath194 .
] when @xmath195 . to compactify on an orbifold we have to introduce new identifications .
a simple possibility is obtained introducing two @xmath151 groups , one acting on @xmath196 and the other on @xmath197 , in this way @xmath198 .
then , due two @xmath171 and @xmath151 action on each direction , the fundamental domain becomes @xmath199 $ ] .
we can act with each @xmath151 independently , then we identify four different points on @xmath168 : @xmath200 . according to this , the orbifold 6d propagator is @xmath201 , \end{aligned}\ ] ] where @xmath202 and @xmath203 is the field parity in @xmath204 direction . here
we can make the same analysis as in 5d : the new propagator terms will give localized divergent contributions . for
zero winding the terms with just one coordinate identified under @xmath151 will give 5d divergencies ( localized in one extra dimension ) , and the term with both coordinates inverted will give 4d divergencies localized in a point of the extra space .
using the winding modes we can easily separate cut - off dependent from finite contributions in 6d . to show this
let s compute the radiative corrections of the scalar theory at one - loop with the extra space compactified on a torus @xmath169 .
the effective action is similar to eq .
( [ seff ] ) but with two extra dimensions .
for a torus @xmath56 is constant , and is given by @xmath205 again the divergent term is due to the zero winding mode and winding modes different from zero give the finite contributions . let s consider a constant field @xmath143 . integrating over the extra space we can compare the finite 4d masses obtained from a 5d theory with the ones obtained from a 6d theory . making @xmath206 and summing over windings
we get @xmath207 . a continuos variable
, then it takes values over the plane .
as we have separated the zero mode , we do nt integrate over the parallelogram in the origin.__,width=151 ] now we consider @xmath61 for a constant field @xmath143 .
using the analog to equation ( [ simple2 ] ) but with one more extra dimension we can write the 6d one - loop coupling as @xmath208 the sum over modes is logarithmically divergent , in the same way as in 5d .
we approximate the sum with an integral , then @xmath209 . we have to exclude the zero mode , then the domain of intagration is shown in fig . [ guraco ] .
therefore the second term of the r.h.s . of eq .
( [ verticecorto ] ) is @xmath210,\;\;\gamma = arctg(\frac{2r_1 r_2 \sin\theta}{r_1 ^ 2-r_2 ^ 2}),\ ] ] where @xmath211 is an ir cut - off , the inverse of @xmath212 and @xmath213 is the torus volume .
we repeat the steps done for the 6d torus , using the orbifold 6d propagator of eq .
( [ gtz2 ] ) . for simplicity
we consider a constant field @xmath143 .
then we can write @xmath73 as @xmath215\big\},\;\;\;\;\;\;\;\;\;\;\phi=\phi_c\ ] ] where @xmath216 is the orbifold volume and @xmath217 is a finite term given by @xmath218\big\ } \\ & + \frac{\lambda}{32\pi^3}\int \!\
! d^2y\sum_{\bar{w}_{fin}}[\frac{p_1}{|2(y^1,0)+\bar{w}|^4}+\frac{p_2}{|2(0,y^2)+\bar{w}|^4}+\frac{p_1p_2}{|2\bar{y}+\bar{w}|^4 } ] , \end{aligned}\ ] ] the function @xmath219 is defined by @xmath220 the sum in eq .
( [ mf ] ) is over windings that do not give divergencies .
then we have to exclude windings given by @xmath221 , when these windings give divergencies . the @xmath222 divergencies in eq .
( [ resultadomt / z2 ] ) are localized in one direction , this can be seen considering fields @xmath223 , and expanding them in power series around the fixed points .
this divergencies are similar to the bulk divergencies in 5d . using the series expansion
it can be seen that @xmath83 terms are 4d , they are localized on the four fixed points @xmath224 . to obtain divergent localized kinetic terms in the directions of the extra dimensions , we have to consider the terms of second order in the series expansion .
now we consider @xmath61 for the orbifold with constant @xmath143 . integrating ( [ simple2 ] ) in 6d with the orbifold propagator we obtain @xmath61 at one - loop .
we approximate the sum over windings with an integral .
the result depends on the @xmath151 parity @xmath203 , and is given by @xmath225 \\ & -(1+p_1+p_2+p_1p_2)[\frac{\pi-\gamma}{2}\log(\mu_{ir } r_1)+\frac{\gamma}{2}\log(\mu_{ir } r_2)]\ } , \end{aligned}\ ] ] where we have regularized the winding sum with an ir cut - off @xmath211 .
the ir logarithmic contributions are cancelled if the scalar field is odd in any of the directions .
this again is easier to understand with k - k decomposition , there is zero mode just for the even - even case .
evaluating this equation for @xmath226 is very easy to compare the linear and logarithmic divergencies with the 5d case .
the radiative corrections show that we should include bulk , 5d and 4d localized masses from the begining , and also localized vertices and kinetic terms .
as an application of the winding formalism , we compute in this section the leading two loop contributions to the effective potential for the radion , in a product space @xmath227 .
we will see that under certain symmetry assumptions we can get a coleman - weinberg potential .
therefore the size of the extra dimension can be stabilized at large values .
let s consider a scalar 5d theory , as the one in section [ orbren ] , compactified on an orbifold .
the effective potential for the radion at tree level is zero , so we compute loops to obtain a sensible effective potential ( see fig . [ figloopsescalares ] ) , we also have to include a two loop diagram , with two three - point vertices , each of them proportional to the vev @xcite . ] .
we calculate these quantum corrections using the winding formalism .
let s start with the one - loop term .
in k - k modes we can write the one loop effective potential as @xmath228 to obtain it in winding representation we can write the last equation as @xmath229 the last factor is the scalar propagator in k - k modes , then we can replace it by the one with winding modes and integrate @xmath230 \\ & = \frac{1}{8\pi^2}\big[\frac{\lambda^5}{5}\pi r-\frac{3\zeta(5)}{8\pi^4 r^4}\mp\frac{\lambda^4}{16}\big ] , \end{aligned}\ ] ] where @xmath142 is defined in eq .
( [ gr1 ] ) , @xmath231 is the riemmann zeta function ( @xmath232 ) , and we get the finite term from no zero windings . after that we compute the two loop term @xmath233 , shown in ( c ) of fig . [ figloopsescalares ] .
it is given by @xmath234\\ \times & [ { \widetilde{g}}(q,2n'\pi r)\pm{\widetilde{g}}(q,2y+2n'\pi r ) ] .
\end{aligned}\ ] ] once more zero winding modes are divergent and non zero modes give the finite contributions
. if we call @xmath235 the first term of the orbifold propagator ( identic to the circle propagator ) and @xmath236 the second term ( obtained with @xmath151 identification ) , then we can write equation ( [ v2top ] ) as @xmath237 .
let s consider first the term @xmath238 .
every loop is similar to the one loop contribution to the mass .
then @xmath238 of eq . ( [ v2top ] ) , can be interpreted as one loop with a massless propagator @xmath239 and a mass @xmath240 $ ] , as is shown in the following feynman diagram [ vssmasa ] the mass @xmath241 is of order @xmath52 , then it can not be considered a perturbation .
therefore we have to consider terms with arbitrary number of mass insertions , as is shown in fig . [ figgs2 ] .
itself is one loop , it is given by the one loop contribution to the mass .
the continuos lines are for masless propagators and the dashed line is for the massive one.__,height=98 ] if we sum the series of fig .
[ figgs2 ] , we obtain a loop with a massive propagator and a mass vertex . as we discussed in section [ windings ] , a massive propagator is obtained replacing @xmath242 . if the mass is large , @xmath243 ( as is the case because @xmath244 ) , we can approximate the propagator by @xmath245 . in this case
the propagator is exponentially damped and cancels before making windings .
then the only relevant contributions are divergent , and @xmath246 .
if there is a symmetry prohibiting divergent masses , as a local gauge symmetry , then the finite contributions are given by @xmath247 .
we can make a similar analysis for the other topologies , obtaining the same results .
summing over topologies we get @xmath248,\ ] ] where @xmath249 , and the sign of @xmath250 depends on field parity under @xmath151 . if there is no symmetry protecting @xmath251 from divergencies , the divergent terms are dominant , in the other case we only get the @xmath252 finite term .
we want to obtain a potential able to stabilize a large extra dimension .
then we consider new brane terms : we add to the previous set - up fermion fields localized on the fixed points . if these 4d fields couple to the bulk ones , there are new contributions to the two loop effective potential .
there is a new term with a fermionic - loop localized on the branes and a scalar - loop on the bulk , as is shown in fig .
[ figloopferm ] . to be more precise we consider the following interaction @xmath253 , localized on the brane at @xmath134 , and compute the two loop contribution .
the 4d fermionic loop has several terms given by @xmath254.\ ] ] we are interested in the logarithmic kinetic contributions .
the first and second terms are local , but the third is not local and will give us new things .
the term of the scalar effective action that couples to the brane loop is @xmath255 then , the two loop contribution with a loop localized on one of the branes , is given by @xmath256 , \end{aligned}\ ] ] as usual , the divergent term is due to the zero winding .
now we have to sum @xmath257 to get the effective potential to two loops .
let s suppose that for each boson there is a fermion with equal boundary conditions .
then the finite part of @xmath258 cancels out , because fermionic loops have a minus sign .
furthermore , if there is a local symmetry ( like local supersymmetry ) protecting the effective potential from divergent terms , then only the finite terms in @xmath233 and @xmath259 remain , and the two - loop effective potential is given by @xmath260 , \;\;\;\;\;\;\;\ ; cte .
1.\ ] ] this is a coleman - weinberg potential @xcite .
we know that this kind of potential can stabilize @xmath261 with a large value @xmath262 , that is consistent with our renormalization assumptions and the large extra dimensions scenario . in this way
quantum corrections in a product space with matter fields localized on the branes can stabilize the size of the extra space , with large radius .
we have used the winding formalism to compute radiative corrections on a theory with extra dimensions .
it allows us to separate , in a very clear and intuitive way , cut - off dependent from finite corrections .
it is also very easy to see how the brane terms are generated at one loop .
we extended this formalism to a higher dimensional space , and showed that it is immediate to separate finite from cut - off dependent contributions , making this method very useful .
we explored the effects of parallel and perpendicular kinetic brane terms .
our conclusions are that , whenever there are perpendicular kinetic terms on the branes , the theory can be redefined , in such a way that there only remain parallel kinetic terms .
we also argued that brane kinetic terms can provide a new way of symmetry breaking .
we applied the winding formalism to compute the finite terms of scalar masses , in theories with one and two extra dimensions .
if there is a symmetry ( like supersymmetry , or other global or gauge symmetries ) that protects masses from divergencies , this finite terms are predictions of the theory .
we also get , in theories with one and two extra dimensions , the logarithmic contributions to the 4d couplings in a simple way .
we analyzed the possibility of getting a potential stabilizing the size of the extra space , when the higher dimensional space can be approximated by a direct product .
we saw that it is very simple to compute the two loop effective potential with winding modes .
we also showed an scenario with brane terms and bulk fields where the extra volume can be stabilized with a coleman - weinberg potential .
i would like to thank lex pomarol for many ideas , discussions and constant advise .
i also acknowledge fruitful discussions with a. flachi and o. pujols .
this work was supported by the spanish education office ( mecd ) under an fpu scolarship .
99 n. arkani - hamed , s. dimopoulos , g. r. dvali , phys .
b 429(1998)263 , [ arxiv : hep - ph/9803315 ] ; i. antoniadis , n. arkani - hamed , s. dimopoulos , g. r. dvali , phys . lett .
b 436(1998)257 , [ arxiv : hep - ph/9804398 ] ; n. arkani - hamed , s. dimopoulos , g. r. dvali , phys . rev .
d 59(1999)86004 , [ arxiv : hep - ph/9807344 ] .
l. randall , r. sundrum , phys .
83(1999)3370 , [ arxiv : hep - ph/9905221 ] .
i. antoniadis , phys .
b246(1990)377 ; j. scherk , j. h. schwarz , phys .
b82(1979)60 ; m. quiros , `` new ideas in symmetry breaking '' , lectures given at tasi 2002 , [ arxiv : hep - ph/0302189 ] , and references therein . n.
arkani - hamed , l. hall , y. nomura , smith , weiner , nucl .
b605(2001)81 , [ arxiv : hep - ph/0102090 ] . m. puchwein , z. kunszt , [ arxiv : hep - th/0309069 ] .
r. barbieri , r. contino , p. creminelli , r. rattazzi , c.a .
scrucca , phys .
d66(2002)024025 , [ arxiv : hep - th/0203039 ] ; d. marti , a. pomarol , phys . rev .
d66(2002)125005 , [ arxiv : hep - ph/0205034 ] .
h. georgi , a. k. grant , g. hailu , phys .
b506(2001)207 , [ arxiv : hep - ph/0012379 ] . i. antoniadis , s. dimopoulos , a. pomarol , m. quiros , nucl .
b544(1999)503 , [ arxiv : hep - ph/9810410];a .
delgado , a. pomarol , m. quiros , phys .
d60(1999)095008 , [ arxiv : hep - ph/9812489 ] .
h. cheng , k. t. matchev , m. schmaltz , phys .
d66(2002)036005 , [ arxive : hep - ph/0204342 ] . f. del guila , m. prez victoria , j. santiago , jhep 0302(2003)051 , [ arxiv : hep - th/0302023 ] . w. d. goldberger , m. b. wise , phys . rev .
d65(2002)025011 , [ arxiv : hep - th/0104170 ] .
see for example l. j. hall , y. nomura , phys .
d64(2001)055003 , [ arxive : hep - ph/0103125 ] ; r. contino , l. pilo , r. rattazzi , e. trincherini , nucl . phys .
b622(2002)227 , [ arxiv : hep - ph/0108102 ] , j.f .
oliver , j. papavassiliou , a. santamaria , phys .
d67(2003)125004 , [ arxive : hep - ph/0302083 ] .
r. jackiw , phys .
d 9(1974)1686 .
s. coleman , e. weinberg , phys .
d 7(1973)1888 . | we compute radiative corrections in five and six dimensional field theories , using winding modes in mixed momentum - coordinate space .
this method provides a simple way of finding uv divergencies , finite corrections and localized terms when the space is compactified on orbifolds . as an application we compute the finite piece of scalar masses , the logarithmic contributions to the couplings and the effect of localized parallel and perpendicular kinetic terms .
we apply it to get a two loop effective potential that can stabilize large extra dimensions . |
entanglement is one of the most striking aspects of quantum mechanics @xcite . in classical physics ,
two interacting systems retain their individuality during the interaction process and become completely independent of each other after their coupling has been switched off . by way of contrast ,
quantum theory predicts a completely different behaviour of interacting systems .
when two quantum systems are brought to interact , their identities become in the course of time more and more entangled , so that a state - vector description of each system is in general precluded .
they build an entangled composite system , whose state - vector can not be separated into a product of the states of the subsystems .
more surprising , the entanglement is preserved even after switching off the interaction , so that a measurement on one system will affect the other one .
the individual systems can be characterised by ( reduced ) density matrices . however , even when the reduced density matrices of both subsystems are known , important information on the physics of their interaction and the related entanglement is lost .
recently , entangled quantum states of an electronic system and a harmonic oscillator have been realized with trapped atoms @xcite and in cavity qed @xcite .
the signatures of the entanglement have been partially demonstrated by quantum measurements .
the determination of the full quantum statistical information of such quantum states , however , requires new types of quantum measurements .
moreover , it is interesting in this context that with both types of systems the jaynes cummings model @xcite and its nonlinear trapped - atom counterpart @xcite could be realized @xcite . in the jaynes
cummings interaction the quantum correlation between the interacting systems has been predicted to show surprising features @xcite . in all these cases methods for determining the full information on the quantum state of entangled systems are desired for getting more insight into interaction phenomena in the quantum world .
several approaches have been proposed for determining the complete information on the quantum state of single quantum systems . in particular ,
homodyne tomography has been established for optical fields @xcite .
further simplifications of the method have been introduced , including the direct statistical sampling of the density matrix in field - strength ( quadrature ) representation @xcite , photon - number representation @xcite , and symplectic tomography @xcite .
tomographic methods have also been realized for molecular vibrations @xcite and proposed for reconstructing the motional quantum state of trapped atoms @xcite .
alternatively , phase - space distributions can be determined by measuring the number statistics of the quantum state of interest , after introducing appropriate coherent displacements @xcite .
a method of the latter type has recently been used to reconstruct the motional state of a trapped atom @xcite . for quantum systems undergoing a jaynes - cummings type dynamics @xcite , such as high-@xmath0 cavity fields @xcite or trapped atoms @xcite ,
various methods have been considered to study the quantum statistics of the bosonic subsystem ( the field or the center - of - mass motion ) , cf .
all these methods , however , give no insight into the entanglement of quantum systems . in this contribution
we propose a method for determining entangled vibronic quantum states of a trapped atom . in sec .
[ sec : charac ] of this paper we introduce a wigner - function matrix , which contains the full information on the composite system under study .
after a brief discussion of its properties we show that it can be determined via measurements of entangled , motional number statistics . in the following section we present a concrete scheme for measuring those quantities .
it relies on the nonlinearities appearing in the motional dynamics of the trapped atom interacting with laser light , which have been predicted for atomic localisations beyond the lamb dicke regime @xcite and confirmed in recent experiments @xcite . by monitoring the electronic dynamics , combined with coherent displacements of the motional subsystem
, one can perform a direct sampling of the wigner - function matrix of the vibronic state . in sec .
[ sec : numerics ] we present a numerical simulation of the reconstruction of the wigner - function matrix and briefly discuss the practical aspects of our method .
a summary and some conclusions are given in sec .
[ sec : summary ] .
the motional quantum state of a trapped atom is usually described by a density operator @xmath1 , which is an operator in the hilbert space @xmath2 of the motion of the atom in the trap . to get a phase - space description of the motional state of the atom
one is lead to quasiprobability distributions such as for example the wigner function .
this function can be obtained from the density operator as follows @xcite @xmath3 , \ ] ] where the operator - valued delta - function @xmath4 is the fourier - transform of the displacement operator @xmath5 and reads as @xcite @xmath6 equation ( [ eq : wigner1 ] ) can be inverted to get the density operator out of the wigner function @xmath7 so that the knowledge of @xmath8 is equivalent to the knowledge of the motional density operator @xmath1 .
even though the wigner - function @xmath8 displays all the obtainable knowledge about the motional state of the atom , it gives no information on its electronic degrees of freedom . for a complete description of the atomic state
i.e. including the electronic degrees of freedom one has to generalise the concepts leading to the definition of the wigner - function . for this purpose
we introduce the density operator @xmath9 of the whole system , which is now an operator in the product - space of motional and electronic hilbert spaces @xmath10 .
the reduced density operators for the motional and electronic subsystems can be obtained by taking the trace over the hilbert spaces of the electronic system ( el ) and the center - of - mass motion ( cm ) , respectively , @xmath11 , \quad \hat{\sigma } = { \rm tr}_{\rm cm } \left [ \hat{\varrho } \right ] .\ ] ] we now define a wigner - function matrix describing the complete quantum state of the trapped atom by generalising eq .
( [ eq : wigner1 ] ) , @xmath12 , \ ] ] where @xmath13 is now an operator in the product - space @xmath10 and is defined by @xmath14 here we have used the electronic flip - operators @xmath15 to the state @xmath16 . in close analogy with the case of the motional sub - system ( [ eq : inversion1 ] ) , the density operator of the composite system can be obtained from this wigner - function matrix as @xmath17 showing that the complete information on the quantum state of the atom is contained in the wigner - function matrix @xmath18 .
let us now discuss some properties of the wigner - function matrix @xmath18 .
it is easy to see that its diagonal elements @xmath19 represent the wigner functions of the conditioned density operators @xmath20 correlated to the electronic state @xmath21 .
more precisely , this means that if one would measure the electronic state of the trapped atom to be in state @xmath21 , the corresponding ( unnormalised ) conditioned wigner function of the motional sub - system would be @xmath19 .
the norm of this conditioned wigner function is simply the occupation probability @xmath22 of the electronic state @xmath23 , whereas the norm of the off - diagonal elements @xmath18 ( @xmath24 ) represents the electronic coherence @xmath25 = \sigma_{ij } .\ ] ] moreover , from eqs .
( [ eq : wigner2 ] ) and ( [ eq : top2 ] ) it is seen that the wigner - function matrix is hermitian @xmath26 so that in order to have the full information on the quantum state under consideration , it is sufficient to know the real - valued diagonal elements of the wigner - function matrix @xmath19 and real and imaginary part of the off - diagonal elements @xmath18 @xmath27 .
if one measures the electronic state to be in an arbitrary superposition @xmath28 the conditioned density matrix of the motional sub - system is @xmath29 = \sum_{ij } \psi_i^\ast \psi_j \ , \hat{\varrho}_{ij } .\ ] ] the wigner function of this motional quantum state can now be represented in terms of the wigner - function matrix ( [ eq : wigner2 ] ) @xmath30 this result shows that the off - diagonal elements of the wigner - function matrix , which are in general complex - valued , contain the information on the electronic coherences and the corresponding motional states , which are entangled with these coherences . on the other hand ,
if one has no information about the electronic state of the atom , the density operator of the motional sub - system is given by @xmath31 = \sum_i \hat{\varrho}_{ii } .\ ] ] the wigner function of this reduced density operator reads as @xmath32 and is simply the trace of the wigner - function matrix . in figure 1
we show an example for the wigner - function matrix of an entangled state corresponding to the so called `` schrdinger - cat '' states .
[ fig1 ] the quantum state considered here has been recently prepared experimentally with a trapped atom by monroe et al .
it is a quantum superposition of two motional coherent states , @xmath33 , of amplitudes @xmath34 and @xmath35 , entangled with the upper and lower electronic states , respectively .
this state can be given by the following expression @xmath36 the wigner - function matrix for this entangled quantum state can be explicitely given by @xmath37 .\end{aligned}\ ] ] it is clearly seen , that @xmath38 and @xmath39 represent the coherent states with amplitudes @xmath35 and @xmath34 , respectively , whereas @xmath40 contains the information on the electronic coherence and the quantum interference effects inherent in the entangled state under study , see fig . 1 .
if the quantum state under consideration can be written as a product state @xmath41 the corresponding wigner - function matrix reads as @xmath42 states of this type contain no entanglement between the motional and electronic degrees of freedom .
the corresponding wigner - function matrix @xmath18 is of the same shape for all indices @xmath43 , determined by the motional wigner function @xmath8 and weighted by the electronic density matrix elements @xmath44 .
for determining the wigner - function matrix @xmath18 from measured data we note that it can be represented in terms of matrix elements of the coherently displaced density operator @xmath9 , where only the motional diagonal elements are needed . to show this we make use of eqs .
( [ eq : wigner2 ] ) , ( [ eq : top2 ] ) together with eq . ( [ eq : top ] ) .
by taking the trace over the electronic sub - system , the wigner - function matrix can be written as @xmath45 , \ ] ] where @xmath46 is still an operator acting on @xmath2 . by using the cyclic property of the trace and performing the trace in terms of number - states of the harmonic vibration in the trap
, the wigner - function matrix reads as @xmath47 here we have made use of the coherently displaced density operator @xmath48 for representing the wigner - function matrix we only need the diagonal elements of @xmath49 with respect to the motional degree of freedom @xmath50 for convenience we will call these quantities displaced , entangled number statistics . to get the full information on the entangled quantum state , it is sufficient to measure the displaced , entangled number statistics ( [ eq : dispmatrix ] ) , in generalisation of the method presented in ref .
the initial coherent displacement can be realized by applying a radio - frequency field , as has been done in the experimental determination of the quantum state of the motional subsystem @xcite . in the following
we present a measurement scheme for determining the displaced , entangled number statistics @xmath51 .
the basic scheme consists in a three - level electronic system of @xmath52-type which is superimposed by the energy levels of the motion of the atom in the harmonic trap potential , see fig . 2 . [ fig2 ]
a weak electronic transition @xmath53 is the one of interest with respect to its entanglement with the motional degree of freedom .
the dynamics of this transition is monitored with very high quantum efficiency by testing a strong , auxiliary transition @xmath54 for the appearance of resonance fluorescence @xcite .
we assume that the weak transition is driven in the resolved sideband limit .
when the laser is tuned on resonance with the vibrationless line , the corresponding interaction hamiltonian ( in the interaction picture ) reads as @xcite @xmath55 here @xmath56 is the rabi frequency of the laser interacting with the weak electronic transition and @xmath57 is the phase of the laser field @xcite .
the operators @xmath58 and @xmath59 are the annihilation and creation operators of a vibrational quantum , respectively .
the function @xmath60 , describing the nonlinearities in the vibronic coupling , is given in its normally ordered form by @xmath61 with @xmath62 being the lamb dicke parameter , characterising the localisation of the atom in the trap .
the interaction hamiltonian ( [ eq1 ] ) fulfils the condition @xmath63 \!=\ ! 0
$ ] of back - action evasion for the vibrational number operator @xmath64 , so that the latter is a constant of motion .
consequently , this interaction couples only between vibronic density matrix elements @xmath65 having the same motional indices @xmath66 .
this renders it possible to determine the entangled motional number statistics @xmath67 of the initial vibronic state by monitoring the dynamics of the electronic transition @xmath53 .
one possibility to obtain the matrix elements @xmath68 is the measurement of the time dependence of the occupation of state @xmath69 .
this dynamics is completely determined by those matrix elements . however
, their reconstruction requires some effort of data analysing by fourier or other appropriate techniques , as considered for the jaynes - cummings dynamics @xcite . in the following
we will deal with an alternative method that directly yields the quantities desired in eq .
( [ eq : wigseries ] ) .
it is related to the scheme proposed in ref .
@xcite for the quantum nondemolition measurement of the motional energy of a trapped atom . after a well controlled interaction time of the laser resonant to the weak electronic transition
, a laser pulse is applied on the strong transition in order to probe for the appearance of resonance fluorescence .
the occurrence of fluorescence detects the atom in the electronic ground state @xmath70 and its absence in the excited electronic state @xmath69 .
we will focus our attention to the no - fluorescence events , since they preclude any disturbance of the motional state via light scattering by the strong transition .
alternating sequences of light pulses on the weak and the strong transition allow to reduce the motional state to a fock state , which is predetermined by time control of the pulse sequence on the weak transition .
eventually , the desired information on the entangled number statistics @xmath68 is directly given by the probability of realizing such a sequence of interaction - probe cycles , for each predetermined fock state @xmath71 .
assume that , subsequent to the coherent displacement of the motional subsystem , the atom is probed for fluorescence on the strong transition after an interaction time @xmath72 of the laser with the weak transition .
provided the atom is detected in the electronic state @xmath69 ( no fluorescence ) , the density operator of the system reduces to the ( unnormalised ) operator @xmath73 where @xmath74 is the corresponding ( unnormalised ) density operator of the motional subsystem related to the electronic state @xmath69 . in view of the interaction hamiltonian ( [ eq1 ] ) , the diagonal elements of @xmath75 in number - state representation are given by @xmath76 \sin \left ( \omega_n \tau \right),\end{aligned}\ ] ] where @xmath77 are the nonlinear vibronic rabi frequencies , @xmath78 being laguerre polynomials .
let us consider the effect of @xmath79 of these interaction - probe cycles with interaction times @xmath80 , each one accompanied by no fluorescence @xcite .
the resulting ( unnormalised ) motional number statistics is conditioned on the times @xmath80 , at which the reductions to the state @xmath69 occurred , and reads as @xmath81 the probability @xmath82 to obtain such a sequence of interaction - probe cycles is given by the trace of @xmath83 , @xmath84 and can be experimentally determined by repeating this procedure many times and counting the number of times one was successful in obtaining such a sequence .
an adequate choice of the interaction times @xmath85 allows to map @xmath86 onto the probability @xmath87 . for mapping the particular element @xmath88 , it is useful to choose the interaction times @xmath89 as @xmath90 where @xmath91 can be set to a different integer at each interaction cycle . from eq .
( [ eq : matcond ] ) it is seen that after an appropriate number @xmath92 of interaction - probe cycles @xmath93 reduces to @xmath94 the other matrix elements ( @xmath95 ) are suppressed by the product of cosines in eq .
( [ eq : matcond ] ) , provided that different motional number states are efficiently discriminated by the vibronic rabi frequencies @xmath96 @xcite .
it is easy to see [ cf .
eqs ( [ eq : prob ] ) , ( [ eq : matcondred ] ) ] that in this case the probability @xmath82 reflects the quantity @xmath97 and that further interaction - probe cycles will not change anything . since the interaction times
@xmath85 are specified from the beginning , the minimum number @xmath98 of cycles needed to complete the mapping process can be evaluated .
the number statistics @xmath97 obtained in this manner can be used to determine the entangled , motional number statistics @xmath99 of the displaced initial vibronic state via eq .
( [ eq : redmot ] ) . for this purpose ,
three different choices of the first interaction time @xmath100 are desired : * [ item1 ] @xmath101 : after detecting the atom in the excited electronic state @xmath69 ( no detection of fluorescence ) , @xmath97 is given by @xmath102 * [ item2 ] @xmath103 : for this choice the diagonal elements of the motional density matrix are given by @xmath104 * @xmath105 : in this case one arrives at @xmath106 \nonumber \\ & & + { \rm i m } \left [ \varrho^{mm}_{12}(\alpha ) e^{-i \varphi } \right ] .
\end{aligned}\ ] ] choosing two laser phases , @xmath107 and @xmath108 , one can obtain the imaginary and real part of @xmath109 , respectively , by subtracting one half of the outcomes from ( i ) and ( ii ) .
eventually , any positive event ( series of no fluorescence ) recorded in this manner can be stored in the computer for a direct sampling of the wigner - function matrix ( to be normalised by the number of trials ) according to eq .
( [ eq : wigseries ] ) .
a typical example for an entangled vibronic quantum state of a trapped atom is the state given in eq .
( [ eq : ent ] ) . in fig .
3 we show a simulation of all steps of the scheme for the determination of the wigner - function matrix of the state @xmath110 according to the method proposed above .
[ fig3 ] the maximum pulse length of the laser resonant to the weak transition is of the order of @xmath111s for a typical rabi frequency of @xmath112 khz @xcite . for interaction times of this order
the electronic relaxation of the weak transition is negligible .
the time needed for the data acquisition of a point in phase space is approximately @xmath113 seconds , so that the complete sampling of one matrix element of the wigner - function matrix @xmath18 as shown in fig .
3 would take about @xmath114 hours .
this is comparable with the data - acquisition time needed for the reconstruction of the quantum state of the motional subsystem , by measurement of the time evolution of the electronic inversion @xcite . the advantage of our scheme , however , consists in the fact that the desired number statistics is directly observed , with no need for inverting systems of linear equations with noisy data input .
the diagonal elements @xmath39 and @xmath38 simply represent the wigner functions of the coherent states @xmath115 and @xmath116 respectively . the off - diagonal elements @xmath18 contain the more interesting information on quantum interferences inherent in the entangled state ( [ eq : ent ] ) . comparing with the theoretical result of @xmath18 given in fig . 1 ,
a good agreement between theory and simulation is found .
note that the entanglement is reflected by the fact that the wigner - function matrix is different for different electronic indices .
if one would measure the wigner function of the motional sub - system , @xmath117 , the result would be nothing but an incoherent superposition of two coherent states , completely concealing the nonclassical character of the atomic state .
that is , in the chosen example the nonclassical nature of the system is manifested by the entanglement .
our measurement scheme is directly suited to demonstrate these features .
in conclusion we have proposed a measurement technique for obtaining the full information on entangled vibronic states of a trapped atom .
first of all , this requires a characterisation of the full quantum state of the composed system which can be related to an appropriate measurement scheme . for this purpose
we define a wigner - function matrix which has the character of a density matrix with respect to the electronic degrees of freedom , and of a wigner function for the motional subsystem .
this wigner - function matrix is readily related to the displaced , entangled motional number - statistics of the atom . for the determination of the wigner - function matrix
we have presented a method that allows the measurement of the displaced , entangled motional number - statistics .
it is based on the nonlinearities appearing in the vibronic coupling of a resonantly driven , weak electronic transition .
this coupling fulfils the back - action evasion criterion for the number of motional quanta .
it eventually allows , in combination with electronic - state reductions by fluorescence measurements , the direct statistical sampling of a wigner - function matrix displaying the entanglement between the motional and the electronic degrees of freedom of the atom .
the method proposed here opens new perspectives for experimental demonstrations of fundamental properties of interacting quantum systems .
s. wallentowitz and w. vogel , phys .
lett . * 75 * , 2932 ( 1995 ) ; phys .
a * 54 * , in press ; j.f .
poyatos , r. walser , j.i .
cirac , and p. zoller , phys .
a * 53 * , r1966 ( 1996 ) ; c. dhelon and g.j .
milburn , phys .
a * 54 * , r25 ( 1996 ) .
s. wallentowitz and w. vogel , phys .
a * 53 * , 4528 ( 1996 ) .
k. banaszek and k. wdkiewicz , phys .
lett . * 76 * , 4344 ( 1996 ) . for the representation of the quasidistributions
underlying the method see also k.e .
cahill and r.j .
glauber , phys .
rev . * 177 * , 1882 ( 1969 ) ; h. moya - cessa and p.l .
knight , phys .
a * 48 * , 2479 ( 1993 ) .
w. nagourney , j. sandberg , and h.g .
dehmelt , phys .
lett . * 56 * , 2797 ( 1986 ) ; th .
sauter , w. neuhauser , r. blatt , and p.e .
toschek , phys .
lett . * 57 * , 1696 ( 1986 ) ; j.c .
bergquist , r.g .
hulet , w.m .
itano , and d.j .
wineland , phys .
lett . * 57 * , 1699 ( 1986 ) . | we propose a method for measuring entangled vibronic quantum states of a trapped atom .
it is based on the nonlinear dynamics of the system that appears by resonantly driving a weak electronic transition .
the proposed technique allows the direct sampling of a wigner - function matrix , displaying all knowable information on the quantum correlations of the motional and electronic degrees of freedom of the atom .
it opens novel possibilities for testing fundamental predictions of the quantum theory concerning interaction phenomena . |
as the quality of photometric data has improved over the years ( largely due to the use of ccds ) , the applicability of a fitting - function which can account for variations in the curvature of a light profile has been demonstrated for elliptical galaxies ( capaccioli 1987 , 1989 ; davies et al .
1988 ; caon , capacciolli & donofrio 1993 ; young & currie 1994 ; graham et al .
1996 ) , and for the bulges of spiral galaxies ( andredakis , peletier & balcells 1995 ; seigar & james 1998 ; moriondo , giovanardi & hunt 1998 ; khosroshahi , wadadekar & kembhavi 2000 ; prieto et al . 2001 ; graham 2001 ; mllenhoff & heidt 2001 ) .
these systems are not universally described with either an exponential profile or an @xmath2 law ( de vaucouleurs 1948 , 1959 ) , but rather a continuous range of light profile shapes exist which are well described by the srsic ( 1968 ) @xmath0 model . in ellipticals , the shape parameter @xmath1 from the srsic model is strongly correlated with the other global properties derived independently of the @xmath0 model , such as : total luminosity and effective radius ( caon et al.1993 ; young & currie 1994 , 1995 ; jerjen & bingelli 1997 ; trujillo , graham & caon 2001 ) , central velocity dispersion ( graham , trujillo & caon 2001 ) and also central supermassive black hole mass ( graham et al.2001 ) .
additionaly , the spiral hubble type has been shown to correlate with the bulge index @xmath1 such that early type spiral galaxy bulges have larger values of @xmath1 than late type spiral galaxy bulges ( andredakis et al .
1995 ; graham 2001 ) .
this correlation arises from the fact that the index @xmath1 is well correlated with the bulge to disc luminosity ratio ( b / d ; see , e.g. simien & de vaucouleurs 1986 ) and this is one of the parameters used to establish morphological type ( sandage 1961 ) .
given the abundance of observational work and papers now using the srsic model , it seems timely that a theoretical study is performed on realistic , analytical models following the @xmath0 law .
structural and dynamical properties of isotropic , spherical galaxies following @xmath0 models have already been studied in detail in the insightful paper of ciotti ( 1991 )
. however , most elliptical galaxies and bulges of spiral galaxies are known to be non spherical objects .
typically , the mass models which have been used for the study of triaxial galaxies have followed analytical expressions which were selected to reproduce the properties of the de vaucouleurs @xmath2 profile ( e.g. jaffe 1983 ; hernquist 1990 ; dehnen 1993 ) , or more recently the modified hubble law ( chakraborty & thakur 2000 ) .
for that reason , previous studies based on these kinds of analytical models , although certainly useful , are however unable to probe the full range of properties which are now observed in real galaxies .
consequently , it is of importance to know how much room for improvement exits in the study of triaxial objects following the @xmath0 family of profiles . due to
the fact that the observed @xmath0 luminosity profiles can not be deprojected to yield analytical expression for the spatial density is an integer . ] , the @xmath0 law has been considered less useful for studies of detailed modelling .
an analytical representation ( approximation ) for the mass density profiles which accurately reproduces the observed @xmath0 luminosity profiles would be of great interest for simulations of real galaxies .
we have therefore derived just such an analytical expression for the mass density profiles of the srsic family of models .
our approximation surpasses the accuracy of both the dehnen models for representing the specific @xmath2 profile and also their extension to the double power law models of zhao ( 1997 ) .
in this paper we present a detailed study of how the physical properties of triaxial stellar systems change as a function of the index @xmath1 .
an accurate analytical expression for modelling the spatial density is presented in section 2 . in section 3
we explore the axisymmetric and the non axisymmetric components of the potential , forces and torques associated with a srsic light distribution .
finally , by using literature available k band observations of a sample of 80 spiral galaxies , the physical basis to the @xmath1@xmath3 ( or @xmath1@xmath4 ) relation is investigated in section 4 .
the projected , elliptically symmetric srsic @xmath0 intensity distribution @xmath5 can be written in terms of the projected , elliptical radial coordinate @xmath6 ( see details in trujillo et al .
2001 ) such that : @xmath7 where @xmath8 is the central intensity , and @xmath9 is the effective radius of the projected major axis . curves of constant @xmath6 on the plane of the sky are the isophotes .
the quantity @xmath10 is a function of the shape parameter @xmath1 , and is chosen so that the effective radius encloses half of the total luminosity . the exact value is derived from @xmath11@xmath12@xmath13 , where @xmath14 and @xmath15 are the gamma function and the incomplete gamma function respectively ( abramowitz & stegun 1964 , p. 260 ) .
the index @xmath1 increases monotonically with the central luminosity concentration of the surface brightness distribution ( trujillo , graham , & caon 2001 ) .
the total projected luminosity l associated with this model is given by @xmath16 where @xmath17 is the ellipticity of the isophotes .
for a homologous triaxial ellipsoid , the spatial ( deprojected ) luminosity density profile @xmath18 can be obtained by an abel integral equation ( stark 1977 ) : @xmath19(\xi^2-\zeta^2)^{-1/2}d\xi , \ ] ] where @xmath20 is a constant that depends on the three - dimensional spatial orientation of the object ( varela , muoz - tuoz & simonneau 1996 ; simonneau , varela & muoz - tuoz 1998 ) and @xmath21 parametrizes the ellipsoids of constant volume brightness .
@xmath20 equals 1 when the proper axis frame of the object has the same orientation as the observer axis frame ( i.e. when the euler angles between the two frames equal zero ) .
assume a triaxial object whose mass is stratified over ellipsoids with axis ratios a : b : c ( a@xmath22b@xmath23c ) and the x ( z ) is the long ( short ) axis ( see fig . 1 ) .
the symmetry of the problem motivates us to work with ellipsoidal coordinates where : @xmath24 and where @xmath25=b / a and @xmath26=c / a .
the mass models considered in this study are the triaxial generalizations of the spherical models discussed in detail by ciotti ( 1991 ) .
the mathematical singularities present in eq . 3 were considered and solved by simonneau & prada ( 1999 , eq .
substituting eq . 1 into eq .
3 , letting @xmath27 , and multiplying by the mass to light ratio @xmath28m / l , we obtain a similar expression to the one found by these authors : @xmath29 with @xmath30 the dimensionless mass density profiles @xmath31 , where @xmath32 is the total mass , are shown for different values of @xmath1 in fig .
it should be noted that the inner density profile decreases more slowly with increasing radius for systems having lower values of @xmath1 .
the mass density profiles of the @xmath0 family ( eq .
5 ) can be extremely well approximated by the analytical expression : @xmath33 where @xmath34 , @xmath35 and @xmath36 is the @xmath37th order modified bessel function of the third kind ( abramowitz & stegun 1964 , p. 374 ) . in the appendix
a we show the values of the parameters ( @xmath37,p , h@xmath38,h@xmath39,h@xmath40 ) as function of the index @xmath1 .
this approximation contains two exact cases : @xmath1=0.5 and @xmath1=1 , and provides relative error less than 0.1% for the rest of the cases ( fig 2b ) in the radial range 10@xmath41 .
this approximation surpasses ( by a factor of 10@xmath4210@xmath43 ) the expression presented in lima neto , gerbal & mrquez ( 1999 ) .
for three different triaxiality mass distributions : a ) spherical ( @xmath25=@xmath26=1 ) ; b ) moderately triaxial ( @xmath25=0.75 , @xmath26=0.5 ) ; c ) highly triaxial ( @xmath25=0.5 , @xmath26=0.25 ) , we have explored , in detail , the non axisymmetric gravitational field over the z=0 plane ( i.e. the disc plane when studying spiral galaxies ) .
we evaluate this quantity by calculating : @xmath44 where @xmath45 and @xmath46 are the m=2 and m=0 component of the gravitational potential , such that the nth order term @xmath47 is evaluated from the gravitational potential on the z=0 plane @xmath48 by using the fourier decomposition ( see , e.g. combes & sanders 1981 ) .
gravitational potential and gravitational force expressions are shown on appendix b. the profiles of @xmath49 for different triaxialities and values of @xmath1 are shown in fig .
3 . as it is expected , as the triaxiality increases the non spherical nature of the gravitational field increases . also
, we highlight that for a given triaxiality , smaller values of @xmath1 ( i.e less concentrated mass distribution ) give greater non - spherical gravitational fields .
the maximum non axisymmetrical behavior of the potential is obtained at radial distances less than 2 @xmath9 .
this radial distance is also a function of the index @xmath1 , decreasing a @xmath1 increases and remains quite independent of the triaxiality of the object . for a moderately triaxial object with @xmath1=1 ,
the non axisymmetrical component of the potential can vary some 6% between r=0 and r=2@xmath9 , and varies some 15% for our highly triaxial model . for an @xmath1=1 model , and starting from our moderately triaxial case ( @xmath25=0.75 , @xmath26=0.50 ) , we increased the value of @xmath26 to 0.75 .
the results are shown in fig .
3c and reveal that g(r ) varied only mildly .
this figure shows that the non axisymmetric effect ( along the radial distance ) in the z=0 plane is mainly due to how the mass of the bulge is distributed in this plane .
the non spherical component of the radial gravitational forces in the z=0 can be estimated by : @xmath50 in fig .
3 the n(r ) profiles ( eq
. 9 ) are evaluated for the same cases as was the g(r ) profiles .
a remarkable point is that n(r ) reaches its maximum value in the radial range 2 @xmath9@xmath51r@xmath514 @xmath9 . for a spiral bulge structure
, this means that the most important non axisymmetric effects take place in a zone which is dominated by the disc . as with the g(r ) parameter , stronger distortions occur as the triaxiality increases and the index @xmath1 decreases .
the mechanism which controls this distortion is basically determined by the mass distribution in the z=0 plane ( fig 3c ) .
it is noted that the relative ( i.e. % change ) non axisymmetric effects on the radial forces are larger than the relative distortion on the potential . as an example , for a moderately triaxial structure with @xmath1=1 the non axisymmetric component of the radial forces can reach 8% .
the torques provoked by the triaxial structures along the angular coordinate are evaluated around the circle of radius @xmath52 where the maximum non axisymmetrical distortion of the radial forces is produced ( i.e. at the peak of the n(r ) profile ) .
given the gravitational potential @xmath48 in the z=0 plane , we have at the radius @xmath52 : @xmath53 where @xmath54/r_{max}$ ] represents the amplitude of the tangential force along the angular coordinate at radius @xmath52 , and @xmath55 is the radial force at this radius . due to the symmetry of the ellipsoid , the values of @xmath56 need only be plotted for one quadrant in the z=0 plane ; we use 0@xmath57@xmath5890@xmath57 ( fig . 3 ) .
depending on the quadrant , @xmath56 is either negative or positive because the sign of the tangential force changes from quadrant to quadrant .
the maximum torque around a circle of radius @xmath52 depends on the triaxiality of the object .
as the triaxiality increases the maximum torque tends to be closer to the major axis as one would expect .
the position of this peak is quite independent of the value of @xmath1 .
the absolute value of the torque for any given triaxiality increases as @xmath1 decreases .
for our highly triaxial bulge , @xmath56 ranges from 0.17 ( @xmath1=10 ) to 0.24 ( @xmath1=1 ) , which would be considered a `` bar strength '' class of 2 in the classification scheme of buta & block ( 2001 ) . in the case of our moderately triaxial object
, the maximum absolute value of @xmath56 ranges between 0.06 and 0.09 .
these values correspond to a `` bar strength '' class of 1 .
thus , even a moderately triaxial bulge is capable of provoking non
negligible torques on a disk that is to say , a bar is not neccessarily required .
a detailed study separating the torque contribution from both bars and bulges would of course be of interest , and it is our intention to add a range of bar potentials to our models in the future . as with the previous parameters , for the range of triaxialities investigated and a given @xmath1 , varying the mass distribution along the z axis ( i.e. varying the triaxiality parameter @xmath26 ) only results in a slight change to @xmath56 ( see fig .
3c ) . for a spherical distribution all above parameters
in the previous section we have seen how the non axisymmetrical effects ( in the z=0 plane ) from a triaxial bulge increase as @xmath1 decreases . taken with the correlation between @xmath1 and galaxy type ( andredakis , peletier & balcells 1995 ) shown in fig .
4 , this invokes the natural question : how , if at all , are the structural properties of the bulges ( i.e. @xmath1 ) related ( that is , cause and effect ) with the non axisymmetrical components ( i.e. arms ) observed in the disc ?
the results obtained in the previous sections were evaluated without any mention to the relative mass of the bulge and disc .
it turns out that the axisymmetrical mass distribution of the disc causes a strong softening of the non axisymmetrical perturbation caused by the non sphericity of the bulge .
the degree of `` smoothing '' is an increasing function of the @xmath59 ratio .
5 shows the n(r ) profile for a moderately triaxial bulge with @xmath60 and @xmath4=0.1 and 0.01 , and for a bulge with @xmath61 and @xmath4=1.0 and 0.1 ( fig .
n(r ) was evaluated here assuming the disc follows an exponential surface brightness distribution .
the ratio between the length scale of the disc and the effective radius of the bulge is assumed to be constant and with a value of @xmath62=5 . in the @xmath63-band
is 5 ( graham & prieto 1999 ) . ] the expressions for the potential and the radial force of these structures can be found in binney & tremaine ( 1987 , p. 77 and 78 ) .
5 illustrates that for @xmath4 luminosity ratios typical of real galaxies , the non axisymmetrical effects on the disc largely disappear ( @xmath64 ) . for @xmath4=1 ,
the values of the n(r ) profile remain basically unchanged to that seen in fig . 3 , but for @xmath4=0.1 these values decrease approximately by a factor 2 , and for @xmath4=0.01 this factor is bigger than 10 .
thus , although the non radial effects on the z=0 plane increase as @xmath1 decreases , the smoothing effects of the increasingly dominant disc are stronger .
bulges with small values of @xmath1 are unable to produce significant non axisymmetrical effects on a massive disc . to explore the connection between bulges and discs in spiral galaxies ( see , e.g. fuchs 2000 ) , we have used the data from two independent samples of galaxies observed in the k band .
the k band provides a good tracer of the mass due to the near absence of dust extinction and the reduced biasing effect of a few per cent ( in mass ) of young stars .
we used the data from andredakis et al .
( 1995 ) and the structural analysis of the de jong ( 1996 ) data performed by graham ( 2001 ) .
both studies were done fitting a seeing convolved srsic law to the spiral galaxy bulges . in both samples
we have removed those objects which contained a clear bar structure , leaving a total of 28 objects from andredakis et al .
( 1995 ) and 52 objects from graham ( 2001 ) .
the relations present in fig .
4 between @xmath1 and the @xmath4 luminosity ratio , and @xmath1 versus the morphological type t have a spearman rank
order correlation coefficient of @xmath66 and @xmath67 , respectively , for the combined sample .
andredakis et al .
( 1995 ) suggest `` although other possibilities can not be excluded , the most straightforward explanation for this trend is that the presence of the disc affects the density distribution of the bulge in such a way as to make the bulge profile steeper in the outer parts .
one mechanism to produce such an effect might be that a stronger disc truncates the bulge , forcing its profile to become exponential '' .
following this line of thought , via collisionless n
body simulations , andredakis ( 1998 ) studied the adiabatic growth of the disc onto an existing @xmath2 spheroid .
he found that the disc potential modifies the bulge surface brightness profile , lowering the index @xmath1 .
this decrease was larger with more massive and more compact discs .
this mechanism , however , saturated at around @xmath1=2 and exponential bulges could not be produced .
we believe that this line of reasoning is not the most appropiate explanation for the relation between @xmath1 and @xmath4 .
firstly , we find that the index @xmath1 is not only well correlated with the luminous @xmath4 ratio , but is equally well correlated ( @xmath68 ) with bulge luminosity @xmath69 ( fig .
additionally , the correlation between @xmath1 and disc luminosity @xmath70 is relatively poor ( @xmath71 ) .
secondly , @xmath69 is more strongly correlated ( @xmath72 ) with the @xmath4 ratio than @xmath70 and the @xmath4 ratio ( @xmath73 ) ( fig .
hence , it is variations in the bulge which are predominantly responsible for variations in the @xmath4 ratio .
these above two correlations seem to indicate that @xmath1 may be related directly with the properties of the bulge rather than with the combined @xmath4 ratio .
consequently , as @xmath1 is correlated with the total bulge luminosity , the correlation between @xmath1 and @xmath4 is a result of the more fundamental correlation between @xmath69 and @xmath4 .
that is , it is not the relative increase in disc to bulge luminosity which produces bulges with smaller values of @xmath1 , but simply that bulges with larger values of @xmath1 are more luminous ( or vice versa ) and this produces the correlation between @xmath1 and the @xmath4 luminosity ratio .
favouring this argument , we note that among the elliptical galaxies ( without the need to invoke any disc ) there exists a strong correlation ( pearson s r=-0.82 ; graham , trujillo & caon , 2001 ) between @xmath1 and the total luminosity of these objects . the index @xmath1 of pressure supported stellar systems are related with the total luminosity of these structures . in agreement with this , aguerri , balcells & peletier ( 2001 ) have found ( using collisionless n body simulations ) that the bulges of late type galaxies can increase their @xmath1 values via dense satellite accretions where the new value of @xmath1 is found to be proportional to the devoured satellite mass . due to the strong correlation between the @xmath4 luminosity ratio and @xmath69 , it might be of interest to ask which one of these quantities is preferred to establish the morphological type t of a galaxy . working from b band images ( which is good for observing the young star population , and consequently the spiral arm structure , which is one of the basic criteria to the hubble galaxy classification ) , simien & de vaucouleurs ( 1986 ) fitted @xmath2 profiles and exponential discs to a sample of 64 spiral galaxies and 34 s0 type galaxies .
they presented a good correlation between the bulge to disc luminosity and @xmath3 , but not between @xmath74 and @xmath3 .
consequently , their b band observations suggested that the @xmath4 luminosity ratio was preferred to @xmath74 for establishing the morphological type t. in fig .
7 we show the relation between the @xmath4 luminosity ratio and t ( @xmath75 ) and between @xmath69 and t ( @xmath76 ) . thus , from k band observations , and fitting @xmath0 bulge profile models , the change in the luminous mass of the bulge along the hubble sequence appears equally as important as the combined change in the bulge and disc luminosity between the @xmath4 luminosity ratio and t , but only @xmath77 between @xmath74 and t. ] .
it would then follow that the luminous mass of the bulge ( i.e. @xmath69 ) is related with the spiral arm structure .
the main results of this work are the following : \a ) we have generalised the analysis of the physical properties of spherical stellar systems following the @xmath0 luminosity law to a homologous triaxial distribution . the density distribution , potential , forces and torques are evaluated and compared with the spherical case when applicable ( ciotti 1991 ) .
an extremely accurate analytical approximation ( relative error less than 0.1% ) for the mass density profile is provided .
\b ) we derive an exact expression showing how the central potential decreases as triaxiality increases .
we also show that for a fixed triaxiality , as the index @xmath1 decreases the non axisymmetrical effects in the z=0 plane increase .
even for a moderately triaxial object , the non axisymmetrical component of the potential and the radial forces are not negligible for small values of @xmath1 .
these components can range from 6 to 8% , respectively , compared to the value of the spherical component . for our highly triaxial model
, they can range over some 20% .
\c ) the non axisymmetrical effects in the disc plane due to the bulge structure are strongly reduced when an axisymmetrical disc mass is added .
for this reason , bulges with smaller values of @xmath1 appear unlikely to produce any significant non axisymmetrical effect on their disc , which is typically 10 to 100 more times more massive than the bulge . in this regard , the @xmath4 mass ratio and the triaxiality of the bulge are more important , that is , can dominate over the effects of small @xmath1 .
\d ) the correlation found between @xmath1 and the @xmath4 luminosity ratio found in spiral galaxies is explained here not as a consequence of the interplay between the bulge and the disc , but due to the strong correlation between @xmath1 and @xmath78 , and between @xmath78 and @xmath4 .
also , k band data do not support the idea that the @xmath4 luminosity ratio can be preferred over @xmath78 as an indicator to establish galaxy morphological type ( t ) .
both parameters present equally good correlations with galaxy type t. 99 abramowitz m. , stegun i. , 1964 , handbook of mathematical functions .
dover , new york aguerri , j.a.l . , balcells m. , peletier r. , 2001 , a&a , 367 , 428 andredakis y.c .
, peletier r. , balcells m. , 1995 , mnras , 275 , 874 andredakis , y.c .
, 1998 , mnras , 295 , 72 binney , j. , tremaine , s. : 1987 , galactic dynamics , princeton university press , princenton buta r. , & block d.l . , 2001 , apj , 550 , 243 block , d.l . ,
puerari , i. , frogel j.a . , eskridge , p.b .
, stockton a. , burkhard f. , 1999 , astroph . and
space sci , 269270 , 5 caon n. , capaccioli m. , donofrio m. , 1993 , mnras , 265 , 1013 capaccioli m. , 1987 , in structure and dynamics of elliptical galaxies , iau symp .
127 , reidel , dordrecht , p.47 capaccioli m. , 1989 , in the world of galaxies , ed . h. g. corwin , l. bottinelli ( berlin : springer - verlag ) , p.208 chakraborty d. k. & thakur p. , 2000 , mnras , 318 , 1273 chandrasekhar , s. 1969 , ellipsoidal figures of equilibrium ( new york : dover ) ciotti , l. 1991 , a&a , 249 , 99 combes , f. , & sanders , r.h .
1981 , a&a , 96 , 164 davies j. , philips s. , cawson m. , disney m. , kibblewhite e. , 1988 , mnras , 232 , 239 dehnen , w. , 1993 , mnras , 265 , 250 de jong , r.s . , 1996 , a&as , 118 , 557 de vaucouleurs g. , 1948 , ann .
astrophys , 11 , 247 de vaucouleurs g. , 1959 , hand .
53 , p. 311
fuchs , b. , 2000 , in : f. combes and g. mamon ( eds . ) , galaxy dynamics : from the early universe to the present , asp conference series , vol .
197 graham a.w . , lauer t. , colles m.m . , postman m. , 1996 , apj , 465 , 534 graham a.w . , prieto m. , 1999 , apj , 524 , l23 graham a.w .
, 2001 , aj , 121 , 820 graham a.w . , trujillo , i. , caon , n. , 2001 , aj , 122 , 1707 graham a.w .
, erwin , p. , caon , n. , trujillo , i. , 2001 , apj , 563 , l11 hernquist , l. , 1990 , apj , 356 , 359 jaffe , w. , 1983 , mnras , 202 , 995 jerjen , e. , bingelli , b. 1997 , in the nature of elliptical galaxies ; the second stromlo symposium , asp conf . ser . , 116 , 239 khosroshahi h.g .
, wadadekar y. , kembhavi a. , 2000 , apj , 533 , 162 lima neto g. b. , gerbal d. & mrquez i. , 1999 , mnras , 309 , 481 mazure a. & capelato h.v .
2002 , a&a , in press mllenhoff c. , heidt j. , 2001 , a&a , 368 , 16 moriondo g. , giovanardi c. , hunt l.k . , 1998 , a&as , 130 , 81 prieto , m. , aguerri , j.a.l . ,
varela , a.m. & munoz - tunon , c. , 2001 , a&a , 367 , 405 sandage , a. , 1961 .
the hubble atlas of galaxies , washington : carnegie institution seigar , m. & james , p. , 1998 ,
mnras , 299 , 672 srsic j. 1968 . atlas de galaxias australes crdoba : obs .
astronmico simien , f. , de vaucouleurs , g. , 1986 , apj , 302 , 564 simonneau e. , varela a. , muoz
tuoz c. , 1998 , il nuovo cimento , 113b , 927 simonneau e. , prada f. , 1999 , ( astro ph/9906151 ) stark a. , 1977 , apj , 213 , 368 trujillo , i. , aguerri , j.a.l .
, cepa , j. , gutierrez , c.m .
, 2001 , mnras , 321 , 269 trujillo , i. , graham , a.w . ,
caon , n. , 2001 , mnras , 326 , 869 varela a. , muoz
tuoz c. , simonneau e. , 1996 , a&a , 306 , 381 young , c.k . ,
currie , m.j .
, 1994 , mnras , 268 , l11 young , c.k . ,
currie , m.j .
, 1995 , mnras , 273 , 1141 zhao h. , 1997 , mnras , 287 , 525
the table a1 shows the values of the parameters that appear in the mass density approximation ( eq . 7 ) . [ cols="<,^,^,^,^,^,^,^",options="header " , ]
( chandrasekhar 1969 , p. 52 , theorem 12 ) , with @xmath81 it follows from eq .
( b1 ) that the potential at an internal point is a result of two contributions : that due to the ellipsoid interior to the point @xmath82 considered , and that due to the homoeoidal shell exterior to @xmath82 : @xmath83,\end{aligned}\ ] ] with f(p , q ) the elliptic integral of the first kind ( abramowitz & stegun 1964 , p. 589 ) , and with the restriction @xmath84 as reported in ciotti ( 1991 ) , the models with low @xmath1 have an inner ( @xmath86 ) potential which is much flatter than models with high @xmath1 . as the triaxiality increases there is no important change to the gradient of the gravitational potential along the semimajor axis ; the main effect is to shift the gravitational potential profile inwards from the spherical case , resulting in a lower potential at intermediate radii .
the gravitational forces for a triaxial structure are given by the expression : @xmath87 with @xmath88 a , @xmath89 b , @xmath90 c , and @xmath91 ^ 2}.\ ] ] the restrictions given in eqs .
( 6 ) and ( b4 ) also apply here . | we have investigated the structural and dynamical properties of triaxial stellar systems whose surface brightness profiles follow the @xmath0 luminosity law extending the analysis of ciotti ( 1991 ) who explored the properties of spherical @xmath0 systems . a new analytical expression that accurately reproduces the spatial ( i.e. deprojected ) luminosity density profiles ( error less than 0.1% )
is presented for detailed modelling of the srsic family of luminosity profiles .
we evaluate both the symmetric and the non axisymmetric components of the gravitational potential and force and compute the torques as a function of position . _ for a given triaxiality , stellar systems with smaller values of @xmath1 have a greater non axisymmetric gravitational field component_. we also explore the strength of the non axisymmetric forces produced by bulges with differing @xmath1 and triaxiality on systems having a range of bulge to disc ratios . the increasing disc to bulge ratio with increasing galaxy type ( decreasing @xmath1 ) is found to heavily reduce the amplitude of the non axisymmetric terms , and therefore reduce the possibility that triaxial bulges in late
type systems may be the mechanism or perturbation for non symmetric structures in the disc . using seeing
convolved @xmath0bulge plus exponential disc fits to the k band data from a sample of 80 nearby disc galaxies , we probe the relations between galaxy type , srsic index @xmath1 and the bulge to disc luminosity ratio .
these relations are shown to be primarily a consequence of the relation between @xmath1 and the total bulge luminosity . in the k
band , the trend of decreasing bulge to disc luminosity ratio along the spiral hubble sequence is predominantly , although not entirely , a consequence of the change in the total bulge luminosity ; the trend between the total disc luminosity and hubble type is much weaker .
= = = = = = = = # 1 # 1 # 1 # 1 @mathgroup@group @mathgroup@normal@groupeurmn @mathgroup@bold@groupeurbn @mathgroup@group @mathgroup@normal@groupmsamn @mathgroup@bold@groupmsamn = `` 019 = ' ' 016 = `` 040 = ' ' 336 = " 33e = = = = = = = = # 1 # 1 # 1 # 1 = = = = = = = = [ firstpage ] celestial mechanics : stellar dynamics galaxies : structure of galaxies : photometry galaxies : elliptical and lenticular , cd galaxies : spiral galaxies : kinematics and dynamics |
the long - running debate on the heating of chromospheric layers by acoustic waves is fed by the fact that even the largest measured energy fluxes ( found by bello gonzlez et al.2009,2010 ) are nearly an order of magnitude smaller than the amount needed to balance the chromospheric energy losses of 14000wm@xmath1 @xcite .
bello gonzlez et al . found an acoustic energy flux of @xmath03000 wm@xmath1 at 250 km and of @xmath02000 wm@xmath1 at 500 km from velocity fluctuations measured from narrow - band ( fwhm=1.8pm ) data at @xmath004 spatial resolution on the fei lines at 557.6 nm and 543.4 nm , formed in mid - photospheric and low - chromospheric layers , respectively .
this is a factor of 46 larger than that given by @xcite and @xcite from intensity fluctuations in continuum bands at 160 nm ( of @xmath01@xmath2 spatial resolution ) and in caiih formed at heights of 430 km and 200 km , respectively .
we refer the reader to @xcite for a detailed introduction to the acoustic - wave heating debate .
+ in this contribution , we present results on acoustic waves from data taken in the fei line at 525.02 nm with the highest spatial resolution considered so far .
the data analysed in this study were taken on june 9 , 2009 with the imax 2d - spetropolarimeter @xcite onboard sunrise @xcite .
this study is performed on a time series of @xmath023min with 33s cadence , taken from 00:36ut on in full - polarimetric mode , at quiet disc centre .
the data consist of narrow ( fwhm=6pm ) spectropolarimetric filtergrams taken at [ @xmath38 , @xmath34 , + 22.7]pm around the fei525.02 nm line ( land factor @xmath4=3 ) .
the capabilities of retrieving the physical parameters with the given spectral sampling have been studied by @xcite .
the data were calibrated according to @xcite ( sect.9.2 ) and possess a spatial resolution of 015018 , achieved by combining the onboard image stabilization system @xcite and phase - diversity techniques @xcite .
the field of view ( fov ) considered is [email protected]@xmath6 ( 815@xmath5815pix@xmath6 ) .
the line - of - sight ( los ) velocities were determined by measuring doppler shifts of the _ minimum _ position of a gaussian fitting the 4 spectral points within the fei line plus continuum @xcite .
we base our study of wave phenomena on the analysis of fluctuations in the measured los velocities . for a better comprehension of the atmospheric layer where the fei line at 525.02 nm is sensitive to velocity fluctuations , we calculated velocity response functions ( @xmath7 ) , applying the method described by @xcite .
following @xcite , we considered granule ( gr ) and intergranule ( igr ) model atmospheres based on 3d simulations from @xcite .
results are shown in fig.[rfs]a . in the case of a spectrometer with infinite spectral resolution ,
the @xmath7 calculated for the line minimum in both atmospheres are centred around 400 km height , in agreement with results by @xcite .
however , the @xmath7 are shifted to lower layers and partly broadened when ( a ) the spectral transmission of the imax instrument and ( b ) the gaussian - fit used in the determination of the line - minimum velocities , are applied to the emergent intensities in order to mimic the measurements from imax observations . for gr ,
the @xmath7 is centred at @xmath0325 km and for igr at @xmath0175 km .
+ the broad extent in height of the @xmath7 indicates the large atmospheric range where the measured velocities are referred to , i.e. indicates the difficulties in detecting the signal of small - scale velocity fluctuations along the los .
it is then advantageous to estimate the transmission of the solar atmosphere to ( wave ) velocity amplitudes , i.e. the ratio of the observed to actual velocities , taking into account the spectral characteristics of our optical system and the velocity determination method .
fig[rfs ] .
at line minimum of fei 525.02 nm : _ solid _ and _ dashed _
, for granule ( gr ) and intergranule ( igr ) atmospheres , respectively ; _ thick _ , after convolution of line profiles with spectrometer function and gaussian fit , to mimic the imax observed profiles . *
b ) _ * transmission of solar atmosphere to wave ( velocity ) amplitudes vs. wave frequency after convolution with the spectral transmission of the imax instrument ; _ solid _ : gr , _ dashed _ igr ; _ thick _ : waves/ with amplitudes constant in height ; _ thin _ : waves with exponentially increasing amplitude . _
dotted _ lines indicate the frequency ( acoustic ) range under study , i.e. from the acoustic cut - off frequency , @xmath8=5.2mhz ( u=192s ) up to @xmath9=14.3mhz ( u=70s ) .
[ rfs],width=453 ] the transmission function is modeled by computing velocities from synthesised imax profiles from a model atmosphere including velocity fluctuations ( waves ) along the los .
we consider both the gr and igr atmospheres following two approaches ( i ) waves with height - independent amplitude and ( ii ) with exponentially increasing velocity amplitudes .
the latter mimics an acoustic energy flux constant through the stratified atmosphere .
we refer to @xcite for detailed descriptions on the transmission function calculations and the determination of group velocities derived from the dispersion relation of waves .
figure[rfs]b depicts the resulting functions .
they show similar transmissions for the gr atmosphere when considering height - independent or exponentially increasing velocity amplitude .
the igr atmosphere shows a lower transmission for waves with height - independent amplitude .
lower transmission means larger correction in the determination of the energy flux ( see sect.[flux ] ) .
we performed fourier and wavelet analyses of the velocity fluctuations over the time series , as in bello gonzlez et al .
( 2009 , 2010 ) .
the temporal power spectra were calculated at each pixel and averaged over the fov . a separate study on _ granular _ and _ intergranular _
areas has been performed .
granules ( gr ) are considered to be those regions in the continuum intensity maps where @xmath10 and intergranules ( igr ) where @xmath11 .
granular dark dots and forming lanes are thus treated as intergranules .
the total flux of acoustic energy ( per element of area on the sun ) is estimated from the spatially averaged fourier power @xmath12 by @xmath13 with mass density @xmath14 , group velocity @xmath15 , and frequency intervals @xmath16 .
@xmath17 represent the _ transfer functions _ , i.e. the square of the transmission functions in fig.[rfs]b . note that the transfer functions express the smoothing by radiative transfer for small wavelengths and we correct our measurements from this effect . from the gr and igr models ,
we adopt the values @xmath18
kgm@xmath19 at 325 km and @xmath20 kgm@xmath19 at 175 km .
we limit our study to the acoustic frequency range , i.e. from 5.215.1mhz , the latter value being the nyquist frequency of our observations .
wavelets are an appropriate tool to retrieve information on the distribution of power in space and time . with this technique we calculated 2d maps of power spectra in period bins of 20s from 70s up to 190s , i.e. the acoustic cutoff period .
we employed the code by @xcite with morlet wavelets and a level of significance of 95% . an analogous equation to eq.[eq1 ] is then used in terms of periods @xmath21 ( with @xmath22=20s ) to determine the acoustic flux from the wavelet power averaged over time and fov .
fig[flux ] . ] figure [ flux ] shows averages of velocity flux spectra as functions of period ( frequency ) from the wavelet and fourier analyses .
we differentiate between gr ( _ thin _ ) and igr ( _ thick _ ) flux before ( _ dashed _ ) and after ( _ solid _ ) correction of atmospheric transmission , in the case of exponentially increasing amplitudes .
this correction is more conservative than that for waves with height - independent amplitude ( see fig.[rfs]b ) , therefore it will provide us with lower - limit corrected estimates . for the differentiation we obtain the power above grs and igrs from the wavelet analysis ( cf .
[ pow_cont ] ) .
the resulting weights are transferred to the power in the fourier analysis ( see * ? ? ?
* ) . before correction
, we find in both the fourier and wavelet analyses , an igr flux larger than the gr flux by a factor 1.7 .
the correction increases the flux in the igr more strongly , as expected from the transmission functions ( fig.[rfs]b ) . after the correction ,
the ratio between igr and gr fluxes amounts to 1.82.0 .
table[table1 ] .
lcccc wavelet : & & uncorr & corrected + gr & & 4430 & 50004970 + igr & & 6400 & 77258630 + average & & 5495 & 64706945 + 5.210mhz & & 5455 & 63606765 + 1014.3mhz & & 40 & 110180 + fourier : & & uncorr & corrected + gr & & 4930 & 70006865 + igr & & 7540 & 1307520165 + average & & 6340 & 1028014050 + 5.210mhz & & 5630 & 77208900 + 1014.3mhz & & 710 & 25605150 + the results on the total acoustic flux summed over all frequencies are collected in table[table1 ] .
the average values refer to the flux over the fov filled by gr over 46% of the area and by igr over the remaining 54% .
results from the wavelet and fourier analyses agree for the most part at low frequencies , i.e. for periods @xmath23100s .
they amount to @xmath05500wm@xmath1 before correction and to @xmath064007700wm@xmath1 after the conservative correction .
yet , a large difference appears at high frequencies , where the treatment of the fourier and wavelet analysis is different : ( 1 ) the wavelet approach is more restrictive in terms of noise when using a statistical significance level of 95% , therefore the retrieved values at these frequencies are low , ( 2 ) the wavelets were applied in period bins which reduces the spectral sampled points at high frequencies to just the bin at 80s and half of the bin at 100s , ( 3 ) the values at high frequencies are strongly corrected , i.e. the originally larger values in the fourier analysis spectra are highly increased .
we are confident in the correction of the flux spectra for frequencies @xmath2410mhz .
the large correction of flux for frequencies @xmath2510mhz above igrs from the fourier analysis , see the inset in fig .
[ flux ] , turns the flux spectrum upward .
this also was noted by @xcite who point out the difficulty in determining the noise level .
in addition , aliasing from power at frequencies above our nyquist frequency may play a role . with these caveats ,
we consider the large fluxes at high frequencies as upper limits .
the high frequency waves are most prone to radiative damping
. their amplitudes will thus be strongly reduced at the base of the chromosphere . in any case , even before correction , the directly measured energy fluxes are larger than those found in any previous observational study of acoustic waves .
the nature of these results and their consequences are discussed in sect.[discussion ] .
the wavelet analysis provides maps of wave power for each time step . with this technique applied to the high - spatial resolution imax data
, we can study in detail the distribution of power ( flux ) over the observed fov and its evolution with time .
figure[pow_cont ] shows an example of different levels of acoustic power over the period range [ 150190]s overlaid on continuum images , and an animation on the time evolution of such power patches is available at url .
fig[pow_cont ] .
the time lag corresponding approximately to the travel time from the bottom of the photosphere to 250 km height is about 42s .
the cadence of the time series is 33s , therefore no delay between intensity and power maps has been taken into account .
the wave power , far from being homogenously distributed , appears intermittently in space and time as already found by , e.g. , @xcite , @xcite and @xcite .
histograms of the acoustic power occurrence over intensity represented in fig.[pow_cont ] show that most of the power , at all frequencies , is associated with _ dark _
structures , i.e. structures with intensities @xmath26 . we have referred to these structures as _
intergranules _ , but close inspection of the power maps overlaid on the continuum intensities ( see animation available at url , attached to this manuscript ) reveal that significant power is found in forming granular ( dark ) dots and lanes . a second
, less prominent component or shoulder is seen in the histograms at @xmath27 .
it has not been reported before , we see it thanks to the high contrast of the imax images .
this power is found in small rapidly evolving granules , and in large granules ( 1 ) prior to their splitting and ( 2 ) at their bright borders . yet
not all splitting granules and bright borders show a wave power component . studying the evolution of acoustic power reveals patches of intense signal .
they appear intermittently and randomly distributed , covering areas of typically 1arcsec@xmath28 and with power at all periods lasting for about 5min .
such strong signals are found to be generated by rapid fluctuations in velocities up to @xmath04kms@xmath29 , that are of a diverse nature : ( 1 ) sudden ( within 23min ) up - down - up flows , ( 2 ) sudden down - up - down flows , ( 3 ) some associated to magnetic field concentrations , ( 4 ) some with no significant magnetism , ( 5 ) some of the latter related to imploding granular areas , etc . in a first analysis ,
we have identified 32 of these events within the 23min series . masking these events and calculating the acoustic flux from their wave power
, we find that they carry 2023kwm@xmath1 ( estimated values before and after correction , respectively ) .
yet , their small area coverage , 0.1% of the fov , implies that their global contribution to the total flux is not significant .
nevertheless , the identification of these energetic features at small scales from their strong power signature has opened a new field of study and we refer to a forthcoming contribution on the nature of these events .
in this study we have taken advantage of the high quality data of the imax / sunrise instrument . the high spatial resolution provided new exciting findings , which ( a ) confirm that acoustic waves play an important role in the energy transport through the quiet solar atmosphere , ( b ) shed new light on the spatial and temporal distribution of propagating wave phenomena and
( c ) reveal the existence of very energetic processes occurring at small scales .
table[table2 ] summarises the values of acoustic energy flux measured by different authors in the last years .
averaged values from the present study before and after correction of the transmission of the solar atmosphere to the velocity wave amplitudes are also listed for comparison .
lcc & uncorr . & corrected + wavelet : binned ( 2@xmath52 ) & 3975 & 46004900 + wavelet : binned ( 1@xmath51 ) & 5495 & 64706945 + fourier : binned ( 1@xmath51 ) & 6340 & 1028014050 + + bello gonzlez et al .
( 2010 ) + @xmath30 , 04 & & 17002000 + bello gonzlez et al .
( 2009 ) + @xmath31 , 04 & 2500 & 30003650 + straus et al .
( 2008 ) + @xmath31 , @xmath004 & & 1400 + @xmath32 , @xmath005 & & 1000 + carlsson et al ( 2007 ) + @xmath33 , @xmath0022 & & 800 + fossum & carlsson ( 2006 ) + @xmath34 , @xmath01@xmath2 & & 510 + wunnenberg et al .
( 2002 ) + @xmath35 , @xmath005 & & 900 + we note that the use of transmission functions for correcting velocity amplitudes at _ short periods _ is controversial .
the results from 3d hydrodynamic simulations by @xcite suggest that the observed high frequency doppler shifts are caused by rapid variations of the contribution functions in an atmosphere with high velocity gradients .
these authors conclude that one should re - evaluate the cause of doppler shifts with periods @xmath36 s , i.e. periods below the nyquist period of our observations .
we restrict the following discussion to the flux values from the wavelet analysis , which we consider to be _
lower limits_. the average acoustic flux at 250 km is @xmath37 , after a conservative . ]
it is mainly found at low frequencies ( periods @xmath23100s ) . yet
, the uncorrected flux already amounts to 5500wm@xmath1 .
these large values were never reported before , probably due to limited spatial resolution .
@xcite pointed out from 3d simulations that a significant amount of acoustic power is lost when degrading the spatial resolution .
this effect can be seen when applying to our data a 2@xmath52 pixel binning .
the signal is reduced by a factor of 1.4 .
we note that @xcite , analysing hinode data , found a lower power reduction upon binning . also compare the resulting acoustic flux with that found by @xcite at the same height .
there the spatial resolution is at least a factor 2 less , although the spectral sampling is better .
after transmission correction , the imax data provide fluxes larger by a factor of 2 .
we conclude that much of the power in the quiet sun occurs at small scales and high spatial resolution is mandatory to detect it .
+ simulations by @xcite give an estimate of 30@xmath38 for energy losses between 250 km up to 500 km due to radiative damping .
applying the same correction , the averaged acoustic flux we observe at 250 km would then be reduced to 4500wm@xmath1 at the bottom of the chromosphere . on the one hand ,
the flux reduction between 250 km and 500 km may be larger because of the very small spatial scales in the present data set .
on the other hand , the estimated flux at 250 km is only a lower - limit value since we have considered : ( i ) wavelet power , ( ii ) the conservative correction .
this acoustic flux is comparable to the radiative losses of 4600wm@xmath1 quantified from the standard semi - empirical hydrostatic model of a quiet average chromosphere by @xcite .
chromospheric models calculated including radiative losses from feii lines , arrive at energy needs of 14000wm@xmath1 @xcite . in our wavelet analysis , the contribution from high - frequency ( @xmath2510mhz ) waves is missing , but likely present in the solar atmosphere as suggested , e.g. the fourier analysis .
let us recall the main findings of this first study on acoustic waves from imax / sunrise observations at disc centre : + 1 ) the improvement in at least a factor of 2 in spatial resolution and the image quality of imax / sunrise data have revealed a total acoustic power larger by at least a factor of 2 than the largest power observed before . in view of the results obtained on the total acoustic power ,
we conclude that , at photospheric level , acoustic waves with periods @xmath25100s carry at least half of the flux needed to balance the observed radiative energy losses of the quiet chromosphere according to @xcite .
the possibility that the quiet solar chromosphere is supplied with acoustic wave energy to cover its radiative losses , as suggested many years ago by l. biermann and m. schwarzschild and later on pursued intensively by p. ulmschneider , needs to be revived and intensively studied , probing which fraction of the transported energy reaches the chromosphere .
+ 2 ) the main contribution to the acoustic energy flux is found in intergranules , as is extensively known .
in addition , forming dark dots and lanes above splitting granules as well as small fast - evolving granules and the brighter borders of large granules are found to be sources of propagating waves . these are regions of strong dynamics ( turbulence ) as pointed out in the studies on the same imax data series by @xcite , and @xcite .
+ 3 ) thanks to the high quality of the imax data , it has been possible to detect intermittent sources of strong energy fluxes at all frequencies in the acoustic domain .
a preliminary analysis of their nature shows that they are generated by rapid ( 23min ) fluctuations in their los velocity occurring at small scales , i.e. typically of 03@xmath503 .. they show a diverse origin , in some cases they are related to small magnetic field concentrations , in others to imploding granules , etc .
we refer to a forthcoming contribution on a detailed analysis of these events .
nbg wants to thank f. kneer for his invaluable help and interest in this work .
nbg also wants to thanks r. schlichenmaier and b. louvel for helpful suggestions on the manuscript .
the german contribution to @xmath39 is funded by the bundesministerium fr wirtschaft und technologie through deutsches zentrum fr luft- und raumfahrt e.v .
( dlr ) , grant no . 50 ou 0401 , and by the innovationsfond of the president of the max planck society ( mpg ) .
the spanish contribution was funded by the spanish micinn under projects esp2006 - 13030-c06 and aya2009 - 14105-c06 ( including european feder funds ) .
the hao contribution was partly funded through nasa grant number nnx08ah38 g .
support by the wcu grant no.r3110016 by the korean ministry of education , science and technology is acknowledged . ,
s. , steiner , o. , bruls , j. , & rammacher , w. 2007 , in astronomical society of the pacific conference series , vol .
368 , the physics of chromospheric plasmas , ed .
p. heinzel , i. dorotovi , & r. j. rutten , 93 | we study the energy flux carried by acoustic waves excited by convective motions at sub - photospheric levels .
the analysis of high - resolution spectropolarimetric data taken with imax / sunrise provides a total energy flux of @xmath064007700wm@xmath1 at a height of @xmath0250 km in the 5.210mhz range , i.e. at least twice the largest energy flux found in previous works .
our estimate lies within a factor of 2 of the energy flux needed to balance radiative losses from the chromosphere according to @xcite and revives interest in acoustic waves for transporting energy to the chromosphere .
the acoustic flux is mainly found in the intergranular lanes but also in small rapidly - evolving granules and at the bright borders , forming dark dots and lanes of splitting granules . |
ever since the discovery of the first high - redshift quasar ( schmidt 1965 ) , quasars have maintained their title as the objects with the highest known redshift ; the present record holder is a quasar at @xmath6 ( schmidt , schneider , & gunn 1991 ) . nevertheless , the highest known redshifts of galaxies have followed closely behind , with bright radio galaxies having been found up to @xmath7 ( lacy et al . 1994 ; rawlings et al .
1996 ) ; more recently , galaxies with high star formation rates have been identified from interstellar absorption lines at @xmath8 to 3 ( steidel et al . 1996 ) and from the ly@xmath0 emission line at @xmath9 ( hu & mcmahon 1996 ) .
in fact , if quasars are related to supermassive black holes that formed in the centers of high - redshift galaxies , we should expect that many galaxies already existed before the first quasars appeared . in any ` bottom - up ' theory where the observed structure in the universe forms by hierarchical gravitational collapse , and the primordial density fluctuations extend to sufficiently small scales ,
the first galaxies to form must have had much smaller masses than the present galaxies .
the first stars should have formed in systems with velocity dispersions of @xmath10 or lower , corresponding to the lowest temperatures ( @xmath11 k ) that allow cooling and dissipation of the gas by atomic processes ( systems with even lower virial temperatures can cool and dissipate through molecular hydrogen , but this cooling process should be suppressed by photodissociation of the molecules after emission of a number of uv photons that is much smaller than that needed to reionize the universe ; see haiman , rees , & loeb 1996 ) .
these systems would be very unlikely to form quasars , because even a small fraction of their baryons turning into stars should provide sufficient energy ( via ionization , stellar winds or supernovae ) to expel the remaining gas from the shallow potential well ( e.g. , couchman & rees 1986 ; dekel & silk 1986 ) .
deeper potential wells , forming at later epochs , are probably needed to form supermassive black holes in galactic centers . even if all the baryons were converted into stars very efficiently in these early dwarf galaxies , with a baryonic mass @xmath12 , their total stellar luminosity would be much smaller than in @xmath13 galaxies at present , simply due to their small mass . since only a small fraction of the baryons in these systems is likely to turn into stars before the gas is ejected
, the total stellar mass in the first galaxies to form in the universe should be much smaller than @xmath14 .
this implies that a supernova in one of these first galaxies to form will be far brighter than the galaxy itself .
thus , the brightest probes of the era when the reionization of the intergalactic medium ( igm ) started should be supernovae in very small galaxies , caused by the death of the same stars responsible for the first ionizing photons . in this paper
, we shall estimate the number of supernovae that should have taken place in these galaxies and should be observable at very high redshift , and their apparent magnitudes .
the scenario where these first small galaxies caused the reionization of the universe is strongly supported by recent evidence that the metal abundance in the ly@xmath0 forest absorption lines with @xmath15 is typically @xmath16 , from observations of civ lines ( see tytler et al .
1995 , songaila & cowie 1996 and references therein ) .
there is relatively little uncertainty in the number of uv photons that were emitted by the stars that produced a given mass of heavy elements , because the heavy elements originate from the supernovae resulting from the same stars that emit most of the ionizing photons ( although the c / o ratio is more uncertain because carbon is more abundantly produced in lower mass stars ) .
according to the most recent calculations ( madau & shull 1996 and references therein ) , the energy of lyman continuum photons emitted is 0.2% of the rest - mass energy of the heavy elements produced .
thus , the energy emitted in ionizing photons per baryon is @xmath17 mev , where @xmath18 is the average metallicity of all baryons in the universe , so only @xmath19 is needed to have emitted one ionizing photon for each baryon . if @xmath20 , then 20 ionizing photons must have been emitted per baryon when the heavy elements were made .
furthermore , if these were the photons responsible for reionizing the universe , then each baryon must have recombined 20 times on average during the reionization epoch .
this is a reasonable number , because a fraction of these photons were probably absorbed in the systems where the stars were born before the gas was expelled , and those that escaped could also have been absorbed in lyman limit systems before the universe became transparent .
thus , there is no need to invoke ejection of gas by more massive galaxies that can accrete the ionized igm to explain a metallicity @xmath21 .
we can now calculate the number of supernovae that were required for enriching the gas in the igm to the average metallicity of @xmath22 that is observed in the ly@xmath0 forest at @xmath23 .
this number should depend only on the igm density and the supernova yields , and should be independent of any other details related to the theory for galaxy formation and the type of galaxies that ejected the enriched gas .
recent simulations of cold dark matter models show that the absorption lines in the ly@xmath0 forest can be identified with the igm , with density fluctuations caused by gravitational collapse , and that most of the baryons should be in the igm in these models ( cen et al .
1994 , hernquist et al .
1996 , miralda - escud et al.1996 ) .
thus , it is reasonable to assume that the high - z supernovae enriched all the baryons in the universe to a mean at metallicity at least as high as that of the ly@xmath0 forest .
since each supernova produces an average of @xmath24 of heavy elements ( with uncertainties depending on the assumed initial mass function and supernova models ; see woosley & weaver 1995 ) , this implies that a supernova took place at high redshift for each @xmath25 of baryons in the universe .
we shall assume a high baryon density @xmath26 , in agreement with the primordial deuterium abundance measured by burles & tytler ( 1996 ) and by tytler , fan , & burles ( 1996 ) .
notice that this implies that most of the baryons at the present time are dark , so many more baryons than those we observe in galaxies had to be enriched at high redshift .
assuming the @xmath27 cosmological model , the total mass of baryons in a redshift shell of width @xmath28 around us is @xmath29 ^ 2/(1+z)^{3/2}\ , \delta z$ ] , where @xmath30 is the present hubble constant . with the above rate of supernovae per baryon mass ( assumed to take place within the epoch corresponding to the redshift shell @xmath28 ) , and taking into account that the supernovae within the shell would be seen by us over a time interval @xmath31 ,
the total supernova rate observed over all the sky is @xmath32 ^ 2 { \ , { \rm yr}}^{-1 } ~.\ ] ] or , for @xmath33 , about 1 supernova per square arc minute per year .
most of these supernovae should be type ii which , if the progenitor is a red supergiant , have a plateau of the luminosity in their lightucurves from 1 to 80 days after the explosion , with @xmath34 erg / s ( woosley & weaver 1986 ) .
a note of caution should be made here , in that the low - metallicity progenitors of these early supernovae could be very different from the high metallicity counterparts .
as illustrated by the case of sn1987a , it is probably not possible at this stage to predict the type of supernovae we should expect from this first generation of stars ; in particular , the possibility that some supernovae might reach higher intrinsic luminosities than regular type iis should be kept in mind .
a duration of 80 days would be redshifted to more than a year at redshifts @xmath35 , where the supernovae from stars responsible for the reionization of the universe should occur . at any random time there
should therefore be one or more supernovae per square arc minute visible in the sky .
of course , many more supernovae should have occurred in more massive galaxies at later epochs ( producing the metals in stars and interstellar gas at present ) , and possibly in small systems that ejected their gas and continued to enrich the igm .
the main difficulty in detecting these supernovae is obviously the extremely faint flux expected .
supernovae should be the brightest objects in the universe at very high redshift , but they are of course much fainter than quasars and , as the redshift increases , the bolometric fluxes decrease at least as rapidly as @xmath36 .
one should point out , however , that when observing at a fixed , long wavelenth such that the supernovae are observed on the rayleigh - jeans part of the spectrum , the flux actually becomes brighter as @xmath37 , in the limit of high redshift . to estimate the apparent magnitudes of the supernovae
, we assume a blackbody spectrum , which is a sufficiently close approximation for our purpose . in figure 1
we show the apparent magnitudes in several bands for a luminosity @xmath38 erg / s , and temperatures @xmath39 k and @xmath40 k. the high temperature is reached @xmath5 1 day after the explosion , when the luminosity drops to the value in the plateau part of the lightcurve ; during the next few days the temperature cools , reaching a value near @xmath41 k after about a week , and then it stays constant until the luminosity starts decreasing . immediately after the explosion , when the shock reaches the surface of the star , the luminosity and temperature can be much higher and the apparent flux can be brighter by a factor @xmath42 relative to the values in fig . 1 , but this phase only lasts for @xmath43 hour .
the apparent magnitudes have been calculated for the @xmath27 model with @xmath44 km / s / mpc , from the flux at the central wavelengths of the bands , using the central wavelengths and zero - magnitude fluxes given in allen ( 1973 ) .
the high - redshift cutoff in the curves in figure 1 indicate the redshift where the supernova light would be absorbed by the ly@xmath0 forest . at high redshift , the supernovae
must obviously be searched in the infrared .
the faintest galaxy surveys from the ground have reached magnitudes @xmath45 ( cowie et al . 1994 ) . from fig . 1 , type ii supernovae would have similar magnitudes at @xmath46 , although searching for variable objects should require more telescope time than simple object identification . in order to find supernovae at redshifts higher than known quasars , fainter fluxes by a factor of @xmath47
need to be detected .
this might be achieved with implementation of adaptive optics on large ground - based telescopes ; in the longer term , the new generation space telescope should certainly be capable to observe these supernovae ( see mather & stockman 1996 ) .
we emphasize again the large uncertainty in predicting the types and absolute magnitude of supernovae from these early generation of stars .
one possibility to detect these supernovae before more powerful telescopes and cameras in the infrared can be built is to use the magnifying power of gravitational lensing in galaxy clusters .
the deflection angles of the most massive clusters of galaxies are as large as @xmath48 , with critical lines of total length of several arc minutes .
the cross section for magnifying a source by more than a factor @xmath49 is @xmath50 , or 0.01 square arc minutes for @xmath51 .
thus , any rich lensing cluster should have a 1% chance of having a high - redshift supernova magnified by a factor larger than 10 at any time .
the highly magnified images would always appear in pairs around critical lines , and would be simple to identify only from their positions and colors given a lensing model of the cluster ( see miralda - escud & fort 1993 ) .
later , variability would have to be detected to distinguish the supernova from a faint , compact galaxy .
we have suggested in this paper that , under reasonable assumptions , supernovae should be the brightest objects in the universe beyond some redshift , in particular during the early phases of the reionization .
the supernovae might therefore be the first observational evidence we shall have of this epoch , when the very faint apparent magnitudes expected are observable .
the other observable signature of this epoch may be the 21 cm absorption or emission by the neutral intergalactic gas ( scott & rees 1994 ; madau , meiksin & rees 1996 ) .
the uv and heavy elements abundance inferred from quasar absorption lines allow us , as we have seen , to draw quantitative conclusions about the minimum number of high - mass stars formed beyond @xmath3 : to produce a metallicity @xmath22 requires one supernova for each @xmath52 of baryons .
this inference is quite robust , being insensitive to the details of structure and galaxy formation at high redshift , which of course depend on the cosmological assumptions .
however , the total mass of stars formed depends on the imf , and is therefore much more uncertain . for a standard imf ,
@xmath53 of stars need to be formed to produce one supernova , so 2% of the baryons should have been turned into stars by the time the igm reached this level of enrichment .
notice that this is equivalent to @xmath54 of the _ observed _ baryons in galaxies today , given our adopted value of @xmath55 which implies that only @xmath5 10% of the baryons are in known stars and gas in galaxies .
it is of course quite possible that the imf was different for these early stars , given the different physical environment ( a higher ambient temperature , absence of heavy elements to act as coolants and provide opacity , and no significant magnetic fields ) .
direct clues to the slope of the high - mass imf may come from ( for instance ) the relative ionization levels of h and he and heavy elements , which depend on the background radiation spectrum shortward of the lyman limit , or from relative abundances of heavy elements relative to carbon .
conceivably , all the early stars might be of high - mass , so that no coeval low - mass stars survive ; at the other extreme , the early imf could have been much steeper than the standard one , in which case there could be many pregalactic brown dwarfs . would any of these `` population iii '' stars be observable today ?
let us consider the observational consequences of the simplest assumption : that the early imf was the same as in the solar neighborhood . in that case , most of the present luminosity from the population iii would arise from red giants and stars at the tip of the main - sequence , with @xmath56 .
where should these stars be today ?
after the first galaxies ejected all their gas back to the igm , the stars that had been formed should have remained in orbit near the center of the dark matter halos .
the stars then behave as collisionless matter as the halos merge with larger objects , until the present galaxies are formed .
we would therefore expect that these stars would at present be distributed approximately like the dark matter in galactic and cluster halos , and in addition there should be some surviving galaxies from that epoch which have not merged into much larger objects ( or have survived in orbit after merging with a large halo , having escaped tidal disruption ) and still have the population iii stars in their centers .
the halos of stars formed in this way around galaxies might be somewhat more centrally concentrated than the dark matter , if many mergers take place with only a moderate increase of the halo mass at each merger ( so that dynamical friction is effective after each merger and it can bring the stars near the center of the newly formed halo before tidal disruption occurs ) .
in fact , particles that start near the centers of halos that merge tend to end up near the center of the merger product ( e.g. , spergel & hernquist 1992 ) .
the known halo stars have a very steep density profile , @xmath57 , and their total mass is @xmath58 ( e.g. , morrison 1993 ) .
this mass is comparable to the total mass we would expect in the halo in the population iii stars , if the total mass of the halo of our galaxy is @xmath59 , with a baryon fraction of 10% , and if 2% of the baryons formed population iii stars .
therefore , if the stellar mass function in the first galaxies was normal , a sizable fraction of the halo stars should have originated there ( this is not surprising , because it is derived from the assumption that the halo stars created their own metal abundance ) .
it seems difficult that the process of dynamical friction alluded to above can result in the steep slope of the halo stars .
however , the halo density profile might become shallower at large radius ( see hawkins 1983 and norris & hawkins 1991 for current observational evidence on this possibility ) , and a second halo population in the outer part of the galaxy ( @xmath60 kpc ) might be the remnant of the population iii .
these halo stars could be found in the hubble deep field ( hdf , williams et al .
if the stellar mass of this outer halo is @xmath61 , there should be @xmath62 stars near the main - sequence turnoff , i.e. , we expect a few stars in the hdf ( with area 4.4 arcmin@xmath63 ) ; these would have colors @xmath64 , @xmath65 at distances of 100 kpc . from fig . 2 in flynn , gould , & bahcall
, we see that there is at least one stellar object with these characteristics in the hdf .
several other observations may help to test the existence of the population iii stars .
an outer stellar halo would also imply a certain number of high - velocity stars near the solar neighborhood .
a stellar population may be found in the halos of external galaxies , with density profiles similar to the dark matter .
sackett et al .
( 1994 ) found a luminous halo in the galaxy ngc 5907 with @xmath66 ( i.e. , about ten times more light than what we expect for the population iii ) .
planetary nebulae could also be found in nearby halos of galaxies or galaxy groups ; several of them were reported recently by by theuns & warren 1996 ) in the fornax cluster .
there is also the possibility that the imf in the early galaxies produced a large number of brown dwarfs . in this case
, a large fraction of the baryons could have been turned into brown dwarfs , and these could be detected in ongoing microlensing experiments towards the lmc ( see paczyski 1996 ) . if the baryon fraction in the universe is 10% , the optical depth of these brown dwarfs toward the lmc could be as high as a few times @xmath67 .
finally , we notice that the metallicity distribution of the population iii stars is difficult to predict .
if only a small fraction of the neutral igm collapsed to galaxies before the reionization , then the gas in these galaxies could reached high metallicities and formed stars , and the metallicity could have diluted in the igm when the gas was ejected . at the same time , the metal abundance of the igm after reionization could be highly inhomogeneous , so some galaxies formed later could have very low metallicities .
therefore , it is difficult to predict even if the average metal abundance of the population iii stars should be higher , lower or similar to the more centrally concentrated halo stars , let alone the distribution of these metallicities .
as the observational techniques improve our ability to detect extremely faint sources , and higher redshift objects can be searched for to continue unravelling the history of galaxy formation , supernovae should become the brightest observable sources .
these supernovae created the heavy elements that were expelled to the igm , and their progenitor stars are the most likely sources of the photons that reionized the universe .
the expected rates of these supernovae , calculating under the assumption of a high baryon density ( @xmath68 ) , and an average metal production of @xmath21 , is as high as 1 supernova per square arc minute per year . to detect the supernovae
, the flux limits of the faintest sources detectable with our telescopes will probably need to be pushed by another @xmath69 magnitudes , although the first examples might be discovered at brighter fluxes behind clusters of galaxies , using the lensing magnification .
any low - mass stars that were formed in the first small galaxies where these supernovae took place should be observable today .
we have argued that , if the imf in these galaxies was similar to the present one in our galactic disk , the population iii stars are likely to account for a large fraction of the stars in our galactic halo , although most of them should be in an as yet undetected outer halo with a shallower density profile than the known , inner stellar halo .
allen , c. w. 1973 , _ astrophysical quantities _
( london : athlone press ) burles , s. , & tytler , s. 1996 , submitted to science ( astroph 9603069 ) cen , r. , miralda - escud , j. , ostriker , j. p. , & rauch , m. 1994 , apj , 437 , l9 couchman , h. m. p. , & rees , m. j. , 1986 , mnras , 221 , 53 cowie , l. l. , gardner , j. p. , hu , e. m. , songaila , a. , hodapp , k .- w . , & wainscoat , r. j. 1994 , apj , 434 , 114 dekel , a. , & silk , j. 1986 , apj , 303 , 39 flynn , c. , gould , a. , & bahcall , j. n. 1996 , apj , 466 , l55 haiman , z. , rees , m. j. , & loeb , a. 1996 , apj , submitted ( astroph-9608130 ) hawkins , m. s. 1983 , mnras , 206 , 433 hernquist , l. , katz , n. , weinberg , d. h. , & miralda - escud , j. 1996 , apj , 457 , l51 hu , e. m. , & mcmahon , r. g. 1996 , nature , 382 , 231 lacy , m. , et al .
1994 , mnras , 271 , 504 madau , p. , meiksin , a. , & rees , m. j. 1996 , apj , submitted ( astroph 9608010 ) madau , p. , & shull , j. m. 1996 , , 457 , 551 mather , j. , & stockman , h. 1996 , nasa report .
miralda - escud , j. , & fort , b. 1993 , apj , 417 , 5 miralda - escud , j. , cen , r. , ostriker , j. p. , & rauch , m. 1996 , apj , 471 , 582 morrison , h. l. 1993 , aj , 106 , 578 norris , j. , & hawkins , m. s. 1991 , apj , 380 , 104 paczyski , b. 1996 , ara&a , 34 , xxx rawlings , s. , lacy , m. , blundell , k. m. , eales , s. a. , bunker , a. j. , & garrington , s. t. 1996 , nature , 383 , 502 sackett , p. d. , morrison , h. l. , harding , p. , & boroson , t. a. 1994 , nature , 370 , 441 schmidt , m. 1965 , apj , 141 , 1295 schneider , d. p. , schmidt , m. , & gunn , j. e. 1991 , aj , 101 , 2004 scott , d. , & rees , m. j. 1990 , mnras , 247 , 510 songaila , a. , & cowie , l. l. 1996 , aj , in press ( astro - ph 9605102 ) spergel , d. n. , & hernquist , l. 1992 , apj , 397 , l75 steidel , c. c. , giavalisco , m. , pettini , m. , dickinson , m. , & adelberger , k. l. 1996 , apj , 462 , l17 theuns , t. , & warren , s. j. 1996 , submitted to mnras ( astro - ph 9609076 ) tytler , d. , fan , x .-
m . , burles , s. , cottrell , l. , davis , c. , kirkman , d. , & zuo , l. 1995 , in qso absorption lines , ed .
g. meylan , p. 289 tytler , d. , fan , x .-
m . , & burles , s. 1996 , nature , 381 , 207 williams , r. , et al .
1996 , science with the hubble space telescope ii , eds .
p. benvenuti , f. d. macchetto , & e. j. schreier ( baltimore : stsci ) , in press woosley , s. e. , & weaver , t. a. 1995 , apjs , 101 , 181 woosley , s. e. , & weaver , t. a. 1986 , ara & a , 24 , 205 | recent evidence on the metal content of the high - redshift ly@xmath0 forest seen in quasar spectra suggests that an early generation of galaxies enriched the intergalactic medium ( igm ) at @xmath1 .
we calculate the number of supernovae that need to have taken place to produce the observed metallicity .
the progenitor stars of the supernovae should have emitted @xmath2 ionizing photons for each baryon in the universe , i.e. , more than enough to ionize the igm .
we calculate that the rate of these supernovae is such that about one of them should be observable at any time per square arc minute .
their fluxes are , of course , extremely faint : at @xmath3 , the peak magnitude should be @xmath4 with a duration of @xmath5 1 year .
however , these supernovae should still be the brightest objects in the universe beyond some redshift , because the earliest galaxies should form before quasars and they should have very low mass , so their luminosities should be much lower than that of a supernova .
we also show that , under the assumption of a standard initial mass function , a significant fraction of the stars in the galactic halo should have formed in the early galaxies that reionized and enriched the igm , and which later must have merged with our galaxy .
these stars should have a more extended radial distribution than the observed halo stars . |
with the next generation of large - volume proton - decay searches and neutrino experiments currently in the r&d phase ( in particular , lbne @xcite , lena @xcite and hyper - k @xcite ) there are good prospects to push the current lower bounds on the proton lifetime to the unprecedented level of @xmath2 years . on the theory side ,
the new information may be , at least in principle , used for further testing of the grand unification paradigm ; however , this would require a very good grip on the proton lifetime predictions supplied by specific guts . unfortunately , the quality of the existing estimates is rather limited even in very simple scenarios , see figure [ estimates ] , and it is namely due to the low accuracy of the leading - order methods used in most of the relevant calculations . on the other hand , consistent next - to - leading - order ( nlo ) proton lifetime estimates
are parametrically more difficult : first , at the nlo level , the gut scale @xmath3 must be determined at two - loops ; this , however , requires a detailed understanding of the one - loop theory spectrum .
second , the flavour structure of the relevant baryon - number - violating ( bnv ) currents must be constrained by the existing data to a maximum attainable degree .
third , one has to account for several classes of almost irreducible uncertainties related to the planck - scale physics ( such as , e.g. , gravity smearing of the gauge unification pattern @xcite ) which are often comparable to the nlo effects . a simple illustration of the typical size of uncertainties in proton lifetime estimates obtained in some of the most popular su(5 ) and so(10 ) guts .
rows [ a]-[d ] depict the results obtained in the georgi - glashow model and some of its simplest extensions @xcite , @xcite , @xcite and @xcite ; in row [ e ] we quote the range given in @xcite ; in the susy case the [ f ] and [ h ] correspond to the estimates given in @xcite ; finally , [ g ] and [ i ] refer to studies @xcite and @xcite , respectively . for
more information see , e.g. , @xcite and references therein.,width=264 ] therefore , the only foreseeable way to overcome this conundrum is to focus on the simplest possible guts .
in contrast to the minimal @xmath4 georgi - glashow model @xcite which was shown to be incompatible with the electroweak data already back in mid 1980 s , the history of the minimal @xmath0 guts is rather non - linear and even after almost 40 years it is still lively and evolving .
interestingly , this can be partly attributed also to the fact that , in the @xmath0 context , the very meaning of minimality is not entirely agreed upon .
this owes namely to the relatively large number of potentially viable symmetry breaking chains in @xmath0 characterized by different effective scenarios emerging at intermediate scales .
let us recall that this is not the case in su(5 ) simply because there the need to preserve rank reduces the set of higgs representations available for the gut symmetry breaking to just few . in a certain sense
, this is not the case in supersymmetric theories either because the rigidity of the mssm gauge unification pattern calls for a single - step breaking where most of the details of the gut - scale dynamics remain obscured .
thus , besides very special features like natural r - parity conservation etc .
, the main distinctive characteristics of many models is namely their flavour structure .
hence , with the spectacular failure @xcite of the simplest potentially realistic renormalizable susy @xmath0 @xcite ( advocated by many to be even the very minimal susy gut @xcite ) , and , in particular , with no signs of susy at the lhc so far , the community s attention naturally drifts back to non - supersymmetric guts . in this review
, we shall comment in brief on the status of the simplest non - susy @xmath0 scenarios and on the latest developments including , in particular , the new upper limits on the seesaw scale obtained recently in the work @xcite and possible future prospects of accurate proton lifetime calculations in this scenario .
the simplest multiplet that can consistently support spontaneous breaking of the @xmath0 gauge symmetry in the sm direction is the 45-dimensional adjoint representation .
together with either the 16-dimensional spinor or the 126-dimensional self - dual part of the maximally antisymmetric tensor the models based on the combinations @xmath5 or @xmath6 are often regarded to as the minimal renormalizable realizations of the higgs mechanism in the @xmath0 guts . in this respect , it is important to recall that this is not the case in susy where the @xmath7-flatness conditions align the vev of @xmath8 along that of @xmath9 which , although providing the desired rank reduction , leaves a full su(5 ) as an unbroken subgroup .
remarkably enough , in the non - supersymmetric higgs model based on @xmath5 or @xmath6 the su(5 ) trap can not be entirely avoided either .
the point is that there are two states in the scalar spectrum of either of the two variants of the minimal model that can be simultaneously non - tachyonic only in a narrow region of the parameter space which , unfortunately , happens to support only su(5)-like symmetry - breaking chains .
indeed , the masses of the color - octet and the @xmath10-triplet components of @xmath8 are at the tree level given by @xcite @xmath11 where @xmath12 is a coupling in the relevant scalar potential ( see , e.g. , @xcite ) and @xmath13 and @xmath14 are the two independent sm - compatible vevs in @xmath8 @xmath15 ( with @xmath16 denoting the second pauli matrix ) which , if hierarchical enough , break the @xmath0 gauge symmetry along two different symmetry breaking chains @xmath17 given that , it is clear that the right hand sides ( rhs ) of both equations in ( [ pgbmasses ] ) are positive if and only if @xmath18 , i.e. , when there is essentially no hierarchy between @xmath14 and @xmath13 , otherwise the corresponding vacuum is unstable . obviously , in such a case , @xmath19 is almost homogeneous and the descent is @xmath4-like ; this , however , conflicts with the gauge unification constraints as in the georgi - glashow model . until recently , the argument above was taken as a no - go for any potential viability of the minimal @xmath0 model including the adjoint scalar as a higgs field responsible for the initial @xmath0 symmetry breakdown . however , as shown in @xcite this was premature because the hierarchy between @xmath13 and @xmath14 may be stabilized by quantum effects because loop corrections such as those diplayed in figure [ loops ] provide non - negligible positive contributions to the rhs of eq .
( [ pgbmasses ] ) .
sample topologies of loop diagrams providing the quantum - level stabilization of the potentially realistic vacua in the minimal @xmath0 gut .
the generic symbol @xmath20 stands for the components of the 45-dimensional adjoint higgs representation while @xmath21 denotes components of the complex scalar ( either @xmath9 or @xmath22 ) responsible for the @xmath1 symmetry breakdown.,title="fig:",width=132 ] sample topologies of loop diagrams providing the quantum - level stabilization of the potentially realistic vacua in the minimal @xmath0 gut .
the generic symbol @xmath20 stands for the components of the 45-dimensional adjoint higgs representation while @xmath21 denotes components of the complex scalar ( either @xmath9 or @xmath22 ) responsible for the @xmath1 symmetry breakdown.,title="fig:",width=132 ] sample topologies of loop diagrams providing the quantum - level stabilization of the potentially realistic vacua in the minimal @xmath0 gut . the generic symbol @xmath20 stands for the components of the 45-dimensional adjoint higgs representation while @xmath21 denotes components of the complex scalar ( either @xmath9 or @xmath22 ) responsible for the @xmath1 symmetry breakdown.,title="fig:",width=132 ] a thorough effective potential analysis @xcite in the simplest @xmath5 variant yields ( in the notation of @xcite ) @xmath23+{\rm logs}\ , , \;\;\ ; \nonumber\\ \label{810onthevac } \delta m^2(8,1,0)_{45}&= & \frac{1}{4\pi^2 } \left [ \tau^2 + \beta^2(\omega_r^2-\omega_r\omega_{bl}+3\omega_{bl}^2 ) + g^4 \left(13 \omega_r^2+\omega_{bl } \omega_r+22 \omega_{bl}^2\right)\right]+{\rm logs}\,,\nonumber \end{aligned}\ ] ] where the `` logs '' denote the typically sub - leading logarithmic corrections .
hence , for small - enough @xmath12 in ( [ pgbmasses ] ) the two problematic states may have non - tachyonic masses even for a large hierarchy between @xmath14 and @xmath13 , thus avoiding the tree - level su(5 ) trap .
let us also note that , up to the obvious differences in the @xmath24 factors , the same dynamical mechanism works in the @xmath6 higgs model .
however , the vacuum stability was not the only issue that plagued the @xmath0 guts for years .
the enormous progress in neutrino physics in the last two decades pinned the light neutrino masses into the sub - ev domain with the upper bound ( namely , from cosmology and double - beta - decay searches ) in the 1 ev ballpark . in the seesaw picture ,
this typically translates into a lower bound on the scale of the underlying dynamics somewhere in the @xmath25 gev domain .
this , however , was long ago claimed to be incompatible with the basic features of the symmetry - breaking pattern in the minimal so(10 ) models . without any detailed information about the scalar spectrum of a theory under consideration ,
the best one can do in order to study the relevant gauge coupling unification patterns is to employ the minimum survival hypothesis ( msh ) @xcite , i.e. , to assume that the components of the unified - theory multiplets cluster around the specific symmetry - breaking scales .
as rough as this approximation sounds , it often gives a qualitatively good first look at the salient features of the unification pattern . in the non - susy @xmath0 framework ,
the `` natural '' positions of the seesaw and grand unification scales have been , under this assumption , studied in @xcite and later reviewed in @xcite . in order to retain a grip on neutrinos and keep the theory well under control , in what follows
we shall focus entirely on the @xmath6 realization of the higgs mechanism in the minimal @xmath0 gut scheme in which the type - i+ii seesaw mechanism is supported at the renormalizable level . for more information about this framework an interested reader
is referred to the relevant literature @xcite .
the resulting constraints on the unification and intermediate scales obtained in this scenario ( in the minimal - survival approximation ) are displayed in figure [ fig : minimalsurvival ] .
constraints on the gut scale @xmath3 and intermediate scales ( @xmath26 , @xmath27 ) in the minimal renormalizable @xmath0 gut derived under the assumption of minimal survival @xcite .
the left panel depicts the situation in the @xmath28 symmetry breaking chains , the right panel gives the same for the @xmath29 descents .
the shaded area defines the frontier between consistent @xmath30 and inconsistent @xmath31 settings .
the three different types of horizontal curves ( dotted , dashed , solid ) correspond , consecutively , to a one - loop analysis without @xmath32 mixing effects taken into account , full - featured one - loop approach and a full two - loop calculation .
the horizontality of the latter two can be justified by a simple diagrammatic argument , see , for instance , @xcite.,title="fig:",width=226,height=188 ] constraints on the gut scale @xmath3 and intermediate scales ( @xmath26 , @xmath27 ) in the minimal renormalizable @xmath0 gut derived under the assumption of minimal survival @xcite .
the left panel depicts the situation in the @xmath28 symmetry breaking chains , the right panel gives the same for the @xmath29 descents .
the shaded area defines the frontier between consistent @xmath30 and inconsistent @xmath31 settings .
the three different types of horizontal curves ( dotted , dashed , solid ) correspond , consecutively , to a one - loop analysis without @xmath32 mixing effects taken into account , full - featured one - loop approach and a full two - loop calculation .
the horizontality of the latter two can be justified by a simple diagrammatic argument , see , for instance , @xcite.,title="fig:",width=226,height=188 ] remarkably enough , for both descends of interest there turn out to be stringent upper limits on the seesaw scale in the minimal survival picture well below @xmath33 gev and , moreover , for the chains passing through the intermediate @xmath34 stage , the upper limit for @xmath3 is in the region which tends to be problematic from the proton lifetime perspective .
this , however , implies that seesaw scale is far outside the @xmath35 gev domain favoured by the light neutrino masses unless the dirac neutrino mass terms are artificially suppressed . although there is nothing a - priori wrong about this option we shall not entertain it here . rather than that
, we shall attempt to do better than the naive estimates above by exploiting the main drawbacks of the minimum survival approach : first , the msh does not reflect many important features of realistic spectral patterns ( such as , e.g. , splitting among different components of multiplets below the relevant symmetry - breaking scales ) .
second , it is totally ignorant of special regions of the parameter space where the scalar spectrum exhibits _ unexpected _ features such as , e.g. , accidentally light states deep in the desert .
however , these are exactly the cases when the unification picture can be altered considerably . beyond the minimum - survival approximation ,
the only guiding principle left for an adventurous parameter - space explorer is the overall consistency of the theory .
this has several basic aspects : [ [ non - tachyonic - scalar - spectrum ] ] non - tachyonic scalar spectrum + + + + + + + + + + + + + + + + + + + + + + + + + + + + + first , all potentially interesting regions of the parameter space should support stable ( or at least metastable ) vacua . since the full - featured vacuum stability analysis is very difficult , we shall stick only to the necessary condition , i.e. , that there should be no tachyonic states in the scalar spectrum . let us point out that , for each such vacuum configuration at hand , one can obtain other viable settings by , e.g. , rescaling all dimensionful parameters in the scalar potential by a common factor .
similarly , it is clear that fiddling around with the mass of an accidentally light state within a range well below the typical mass - scale of all other heavy states does not destabilize specific vacua either because such variations correspond to only very small shifts in the fundamental parameters of the theory .
these two `` degrees of freedom '' can subsequently be used as an efficient tool for reducing the complexity of the numerical analysis of consistent unification patterns .
[ [ current - proton - decay - limits ] ] current proton decay limits + + + + + + + + + + + + + + + + + + + + + + + + + + + another obvious constraint on the parameter space of the minimal @xmath0 gut comes from the proton decay ; in particular , the current best limit for the @xmath36 mode from super - kamiokande @xcite should be accommodated . in what follows
we shall use this together with two assumed future limits that hyper - kamiokande ( hk ) @xcite may reach by 2025 and 2040 ( if built ) : @xmath37 furthermore , we shall for simplicity neglect all the details related to the flavour structure of the baryon - number - violating currents so that the numbers above translate directly to the bounds on the position of the gut scale . in the relevant plots ( namely , figures [ figureremnants ] and [ figure2 ] ) , the points falling between these limits will be distinguished by a simple colour - code where the light grey is used for proton lifetimes between @xmath38 and @xmath39 years , dark grey corresponds to lifetimes between @xmath39 and @xmath40 years and black points yield more than @xmath40 years .
[ [ big - bang - nucleosynthesis ] ] big - bang nucleosynthesis + + + + + + + + + + + + + + + + + + + + + + + + third , accidentally light coloured states should not be too - long - lived otherwise their late decays may interfere with the highly successful classical big - bang - nucleosynthesis ( bbn ) account of the light elements abundances .
actually , as we shall see , this is not a problem here because the accidentally light multiplets in all fully consistent cases originate in @xmath22 and , thus , couple directly to the sm matter fields through the same yukawa couplings that give rise to , e.g , right - handed neutrino masses .
thus , all the light remnants should decay well before the bbn epoch .
[ [ consistent - unification - patterns ] ] consistent unification patterns + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the simple constraints above are enough to filter out all but two qualitatively different settings with a single accidentally light scalar multiplet well below the @xmath1 symmetry - breaking scale : a scenario with a very light colour octet @xmath41 and another scheme with an intermediate - mass - scale colour sextet @xmath42 . the typical shapes of the gauge unification patterns in these two cases are shown in figure [ running ] .
the results of a detailed numerical scan over extended regions supporting these solutions are given in figure [ figureremnants ] .
interestingly , the mass range of the octet solution ( on the left panel in figure [ figureremnants ] ) can stretch as low as to the tev domain so , in principle , it can even leave its imprints in the lhc searches ; however , the sextet is not allowed below roughly @xmath43 gev .
the one - loop gauge unification patterns in the two cases of our main interest with the light @xmath41 [ left panel ] or the light @xmath42 [ right panel ] multiplet in the gut desert .
the almost vertical curves on the right correspond to the naive bounds on the `` unification point '' position derived from the current super - k and assumed future hyper - k limits quoted in eq .
( [ protondecay ] ) .
there are only three trajectories drawn up to the gut scale because the calculation has been conveniently performed in the effective sm picture .
the gray circles depict positions of various heavy scalar and gauge multiplets ( as listed in table [ tablespectra ] ) supporting these patterns . for more details an interested reader
is referred to the original work @xcite.,title="fig:",width=302,height=200 ] the one - loop gauge unification patterns in the two cases of our main interest with the light @xmath41 [ left panel ] or the light @xmath42 [ right panel ] multiplet in the gut desert .
the almost vertical curves on the right correspond to the naive bounds on the `` unification point '' position derived from the current super - k and assumed future hyper - k limits quoted in eq .
( [ protondecay ] ) .
there are only three trajectories drawn up to the gut scale because the calculation has been conveniently performed in the effective sm picture .
the gray circles depict positions of various heavy scalar and gauge multiplets ( as listed in table [ tablespectra ] ) supporting these patterns . for more details an interested reader
is referred to the original work @xcite.,title="fig:",width=302,height=200 ] mass - ranges for the accidentally light states in the light - octet [ left panel ] and light sextet [ right panel ] scenarios .
@xmath44 . in both cases ,
the variable on the vertical axis plays a role of the gut scale ; this is why the proton lifetime limits cut the points from below .
remarkably enough , the octet can be rather light and , in principle , it can be pulled close to the electroweak scale where it can , e.g. , leave its imprints in the lhc searches . , title="fig:",width=302,height=200 ] mass - ranges for the accidentally light states in the light - octet [ left panel ] and light sextet [ right panel ] scenarios .
@xmath44 . in both cases ,
the variable on the vertical axis plays a role of the gut scale ; this is why the proton lifetime limits cut the points from below .
remarkably enough , the octet can be rather light and , in principle , it can be pulled close to the electroweak scale where it can , e.g. , leave its imprints in the lhc searches . ,
title="fig:",width=302,height=200 ] finally , the allowed ranges for the @xmath1 -breaking vev ( denoted @xmath45 ) in these two scenarios are depicted in figure [ figure2 ] .
remarkably enough , _ the naive msh - based upper bounds on the seesaw scale are in both cases relaxed by as much as four orders of magnitude _ as they stretch up to about @xmath46 gev in the light - octet case and up to almost @xmath47 gev in the case of the light sextet .
this , however , makes the implementation of the standard seesaw mechanism possible even without resorting to the excessive fine - tuning in the yukawa sector implied in previous studies . the @xmath48 and @xmath49 cuts of the parameter space corresponding to solutions with the light @xmath41 [ left panel ] or @xmath42 [ right panel ] multiplets in the desert where @xmath45 denotes the @xmath1-breaking vev of @xmath22 .
various levels of gray correspond to domains accessible for different gut - scale limits , cf .
( [ protondecay]).,title="fig:",width=302,height=200 ] the @xmath48 and @xmath49 cuts of the parameter space corresponding to solutions with the light @xmath41 [ left panel ] or @xmath42 [ right panel ] multiplets in the desert where @xmath45 denotes the @xmath1-breaking vev of @xmath22 .
various levels of gray correspond to domains accessible for different gut - scale limits , cf .
( [ protondecay]).,title="fig:",width=302,height=200 ]
even though the minimal so(10 ) models have been recently revived as consistent quantum field theories free of inherent tachyonic instabilities , this beautiful and simple framework has never been rehabilitated as a potentially realistic and predictive gut scheme .
this was namely due to the old studies of the relevant gauge unification patterns which revealed a generic tendency for the @xmath1 symmetry breaking scale to be confined below @xmath33 gev , apparently too low for a reasonable implementation of the seesaw mechanism for neutrino masses .
however , all these early studies based on the minimal survival hypothesis suffer from a generic incapability to account for the effects of accidentally light multiplets with masses detached from any specific symmetry breaking scale .
as we argued , a closer inspection of the unification constraints reveals a much wider room for the @xmath1 breaking vev stretching up to the @xmath46 gev ballpark which , in turn , allows the seesaw picture to be implemented without excessive fine - tuning .
remarkably enough , the same pair of scalars that we identified as powerful `` running helpers '' in the minimal so(10 ) framework was recently singled out in the work @xcite from a totally different perspective . there
, the apparent enhancement in the @xmath50 rate indicated by the current lhc data was shown to be attributable to just this couple of states if any of them falls into the vicinity of the tev scale . however , at the current level of accuracy only the octet @xmath41 can be light enough in the minimal @xmath0 to play any role in the higgs physics because the sextet is not allowed below about @xmath43 gev , see figure [ figureremnants ] .
nevertheless , the proton - decay limits used in cutting the low - mass - sextet region therein are rather naive and it can happen that a more - sophisticated analysis including flavour effects opens a bigger room for the light sextet too .
[ [ suppression - of - the - planck - scale - induced - unification - smearing - effects - in - the - minimal - so10-guts ] ] suppression of the planck - scale induced unification - smearing effects in the minimal @xmath0 guts + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + besides simplicity , the @xmath0 models in which the gut - scale symmetry breaking is driven by a vev of the 45-dimensional adjoint representation have another very interesting feature .
this has to do with the general fragility of grand unification with respect to the planck - scale ( @xmath51 ) effects which , given the proximity of @xmath3 and @xmath51 , may not be entirely negligible .
concerning their possible impact on , e.g. , proton lifetime estimates , the most important of these is namely the planck - scale induced violation of the canonical normalization of the heavy gauge fields @xcite due to the higher - order corrections to the gauge kinetic form emerging already at the @xmath52 level : @xmath53/m_{pl}$ ] ; here @xmath54 is any scalar in the theory which can couple to a pair of adjoint representations of a specific gut symmetry group , i.e. , any field appearing in the symmetric part of the decomposition of their tensor product . for a gut - scale vev of @xmath54 ,
this induces a percent - level effect which , after a suitable redefinition of the gauge fields , leads to similar - size shifts in the gut - scale matching conditions .
such a `` unification smearing effect '' can , subsequently , play a significant role in an accurate nlo gut - scale determination which , in turn , further adds to the existing theoretical uncertainties in the absolute proton lifetime estimates .
however , if in the so(10 ) the gut - scale symmetry breaking is triggered by the vev of @xmath8 , this problem is absent because @xmath55=0 $ ] due to the fact that 45 is not in the symmetric part of the @xmath56 decomposition [ recall that @xmath57 .
thus , the minimal @xmath0 scheme with the adjoint - driven higgs mechanism is uniquely robust with respect to this class of quantum gravity effects .
this makes the symmetry - breaking analysis more reliable and , hence , admits in principle for a strong reduction of this type of theoretical uncertainties in the proton lifetime estimates .
the simplicity of the minimal @xmath0 scenario advocated in this study , together with its rather unique robustness with respect to the planck - scale - induced unification smearing effects make this class of models particularly suitable for a possible next - to - leading - order ( nlo ) proton lifetime analysis .
that , however , is far from trivial . to this end , let us just note that the main source of the large theoretical uncertainties in the existing proton lifetime estimates , cf .
figure [ estimates ] , is the inaccuracy of the gut - scale determination , partly due to the uncertainties in the low - energy inputs ( especially in @xmath58 ) and , in particular , the limited precision of the one - loop approach given the logarithmic nature of the renormalization - group evolution , both these errors are exponentially amplified in the resulting proton decay amplitude .
the only way to keep such uncertainties under control is thus a careful two - loop renormalization - group calculation including , as a prerequisite , the one - loop spectrum of the theory resulting from a dedicated analysis ( like , e.g. , that in @xcite ) together with the proper one - loop matching @xcite conditions . in this respect , the minimal renormalizable @xmath0 gut of our main concern here can be _ the _ scenario in which a decisive nlo proton lifetime analysis can be just at the verge of tractability .
the option of a gut - scale baryogenesis in so(10 ) models , recently revived in @xcite , is another interesting feature of the model under consideration .
on the technical side , however , the amount of thus generated baryon asymmetry depends on the size of the quartic coupling @xmath59 in the scalar potential , cf .
formula ( 3 ) in reference @xcite , which , unfortunately , turns out to be one of the most elusive of all the theory parameters as it was argued in @xcite , @xmath59 does not enter the tree - level scalar spectrum and , so , the one - loop unification pattern does not impose any constraints on it .
however , this coupling does take part in the decomposition of the light higgs boson in terms of its defining components in @xmath60 and , thus , it may be constrained indirectly by the flavour structure of the effective theory . hence , a dedicated proton lifetime analysis advocated above ( which , inevitably , must include a detailed account of the flavour structure of the model ) may , as one of its by - products , provide also a better grip on the gut - scale baryogenesis in the minimal @xmath0 gut .
m.m . is grateful for the invitation to cetup12 and support during his stay in lead , sd .
is partially supported by miur and the eu unilhc - grant agreement pitn - ga-2009 - 237920 .
the work of l.dl . was supported by the dfg through the sfb / tr 9 `` computational particle physics '' .
the work of m.m . is supported by the marie curie intra european fellowship within the 7th european community framework programme fp7-people-2009-ief , contract number pief - ga-2009 - 253119 , by the eu network grant unilhc pitn - ga-2009 - 237920 , by the spanish micinn grants fpa2008 - 00319/fpa and multidark cad2009 - 00064 ( consolider - ingenio 2010 programme ) and by the generalitat valenciana grant prometeo/2009/091 . .[tablespectra]the scalar spectra underpinning the gauge - coupling evolution in figure [ running ] .
the `` type '' column encodes the nature of the relevant multiplet as follows : cs = complex scalar , rs = real scalar , gb = goldstone boson and vb = vector boson , @xmath61 corresponds to the changes in the beta - functions across each of the thresholds .
the accidentally light scalars ( in the first row ) and the gauge bosons defining the gut scale are in boldface . for further details
see @xcite . [ cols="^,^,^,^",options="header " , ] | we recapitulate the latest results on the class of the simplest @xmath0 grand unified models in which the gut - scale symmetry breaking is triggered by an adjoint higgs representation .
we argue that the minimal survival approximation traditionally used in the gut- and seesaw - scale estimates tends to be blind to very interesting parts of the parameter space in which some of the intermediate - scale states necessary for non - supersymmetric unification of the sm gauge couplings can be as light as to leave their imprints in the tev domain .
the stringent minimal - survival - based estimates of the @xmath1 scale are shown to be relaxed by as much as four orders of magnitude , thus admitting for a consistent implementation of the standard seesaw mechanism even without excessive fine - tuning implied by the previous studies .
the prospects of the minimal renormalizable @xmath0 gut as a potential candidate for a well - calculable theory of proton decay are discussed in brief .
address = ahep group , instituto de fsica corpuscular
c.s.i.c./universitat de valncia , edificio de institutos de paterna , apartado 22085 , e 46071 valncia , spain address = infn , sezione di trieste , sissa , via bonomea 265 , 34136 trieste , italy address = institut fr theoretische teilchenphysik , karlsruhe institute of technology ( kit ) , d-76128 karlsruhe , germany |
the initial problem is part of the collection of the moscow institute of physics and technology @xcite .
a voltage @xmath0 is applied to the corners a and b of a semi - infinite long metallic ruler as shown in fig .
[ fig : ruler ] ) . is applied between a and b. between c and d the voltage @xmath1 is being measured . ]
the question is : when one measures the voltage difference @xmath1 between c and d , how does the voltage difference behave further down the ruler ? assuming that the thickness of the ruler can be neglected , this can be reduced to a two - dimensional problem .
a. czarnecki derived the correct solution to this problem conformal mapping @xcite . in this tutorial , we present the experimental verification of these new calculations .
conformal mapping is a mathematical technique which is widely used not only in physics but also in engineering . the main idea behind this technique is to simplify a certain problem by mapping it to a better suited geometry in order to simplify its solution .
a complex function @xmath2 is called holomorphic , if it is complex differentiable at any point in its domain . or in other words , if the following limit exists : @xmath3 a holomorphic function @xmath4 is said to be conformal if @xmath5 .
conformal functions have the properties to preserve locally angles and the shapes of infinitesimally small figures .
the mercator projection is a cylindrical map projection .
it is probably the most common way to map the spherical earth surface in two dimension .
the corresponding map is : @xmath6 @xmath7\ ] ] where @xmath8 is the earth radius , @xmath9 is the latitude , @xmath10 the longitude and @xmath11 an arbitrary central meridian ( commonly chosen to be the one of greenwich ) .
this mapping satisfies the above condition of a conformal map and visualises well its properties .
the circles of longitude and latitude are perpendicular on the map and on small scales the shapes of objects are preserved . whereas large objects can change their shape and size depending where they are located on the globe .
for example according to fig .
[ fig : mercator ] the size of greenland and africa would be of the same order .
the solution of the problem presented in sec .
[ sec : problem ] as suggested in @xcite starts with comparing the potential differences @xmath12 at two nearby pairs of points : @xmath13 it is then argued that since the ruler is semi - finite , the coefficient @xmath14 is constant and does not depend on position .
this leads to the differential equation : @xmath15 if this equation would hold everywhere one would get the final solution : @xmath16 where @xmath0 is the voltage applied at the edge and @xmath1 the measured voltage difference at a given point with a distance @xmath17 from the edge .
however , the ruler is only semi - infinite and not infinite , so the assumption that the voltage does not depend on the position is not fulfilled .
in fact , points very close to the beginning of the ruler do not have the same neighbourhood as points further down the ruler .
czarnecki @xcite derived the correct solution to this problem using conformal mapping .
complex coordinates @xmath18 were introduced such that the corner b corresponds to @xmath19 and the upper corner a to @xmath20 .
looking at the image of the ruler under the mapping @xmath21 , one sees that the corners a and b are mapped on the @xmath22-axis , @xmath23 and @xmath24 as shown in fig .
2 . ]
if an infinite ruler would be mapped with this function , this would cover the whole upper half - plane .
the actual advantage of this mapping is that every function that only depends on the distance to the corners is now symmetric to the real axis since both corners lie on that axis .
this symmetry leads to the solution : @xmath25 where @xmath26 is the length of the contacts and @xmath27 is a constant depending on their detailed geometry but not on their size if @xmath26 is small .
the expression @xmath28 describes physical rather than idealized contacts and we thus refer to it as the reality factor .
these contacts parameters are dependent on the width of the ruler .
if one assumes that the distance between the left corner and the position where we measure the voltage difference is sufficiently large ( more then a third of the width of the stripe ) , this formula can be approximated to : @xmath29 where @xmath30 and @xmath31 are the voltage differences measured at distances @xmath32 and @xmath33 from the edge .
to realize this experiment the following basic laboratory equipment is required : * standard power supply ( 30 v , 3 a ) * standard voltmeter ( @xmath34 mv ) * aluminium foil ( 10 - 50 @xmath35 m , typical thickness of aluminium kitchen foil ) the experiment consists of applying and measuring the voltage difference at different points . to simulate a semi - infinite metallic ruler ,
the metal stripe made of aluminium foil was cut much longer than actually needed .
the wires , used to apply the voltage difference , were pulled through holes in the contact stripes as shown in fig .
[ setup ] .
the experimental configuration can be described by the following parameters : the length of the ruler @xmath36 , its width @xmath37 , its thickness @xmath38 and the width of the contact stripes @xmath39 ( see fig . [ setup ] ) .
the voltage difference was measured in 5 mm steps form 0 to 50 mm .
these measurements were performed with different settings in order to investigate the influence of the following factors : @xmath36 , @xmath38 , @xmath37 and @xmath39 .
the experimental uncertainties to be taken into account are the ones of the voltmeter ( @xmath400.01 mv ) and the one of the measuring position ( @xmath400.3 mm ) .
the results for @xmath36 and @xmath38 are presented in figs .
[ vvsl]-[vvst ] . as
one can see , the measured points for ruler lengths of 300 and 500 mm are the same within the experimental errors , one can thus conclude that a ruler with more than 300 mm is sufficiently long for our experiment and it is a good approximation of a semi infinite ruler .
the results are also unaffected when using two different thicknesses of @xmath38=0.01 and 0.05 mm .
therefore aluminium kitchen foil is thin enough to approach a 2-dimensional problem as required by this experiment . on the voltage difference . ] on the voltage difference ] to study the effect of the contacts geometry and thus the @xmath28 parameter of eq.[eq : andrzejsol ] , the width @xmath39 was varied .
apart from the first measured point at @xmath41 mm , the obtained values are the same for all the others distances within the experimental errors ( see fig .
[ vvswc ] ) . on the voltage difference .
] the last parameter to be investigated is the influence of the width @xmath37 of the ruler . according to the original calculation ( eq . [ eq : originalsol ] )
one would expect that the measured values are independent on @xmath37 .
this is in contradiction with the data as shown in fig .
[ wr ] confirming the inadequacy of this solution and correctness of the newer calculations on the voltage difference and fit to the data using eq .
[ eq : andrzejsol ] ]
our measurements are in very good agreement with the predictions of the new calculations obtained applying conformal mapping by czarnecki and point out that the original solution in which the problem of calculating the distribution of the electric potential in a very long plate was proposed is not adequate .
this problem is a very nice example of the application of conformal mapping and since this experiment is very simple and uses only basic equipment , it can reproduced in any undergraduate laboratory thus providing a very good introduction to students on this subject .
99 a. czarnecki , canadian journal of physics 92 , 1297 ( 2014 ) s. b. cahn and b. e. nadgorny , `` a guide to physics problems '' , part 1 : mechanics , relativity and electrodynamics ( plenum , new york , 1994 ) http://mathworld.wolfram.com/images/eps-gif/mercatorprojection_1000.gif | we report an experimental verification of conformal mapping with kitchen aluminum foil . this experiment can be reproduced in any laboratory by undergraduate students and it is therefore an ideal experiment to introduce the concept of conformal mapping .
the original problem was the distribution of the electric potential in a very long plate .
the correct theoretical prediction was recently derived by a. czarnecki @xcite . |
the global structure of multiple hadroproduction in annihilation , has been well understood with remarkable results concerning the total cross section and angular distribution calculations .
perturbative quantum electrodynamics and perturbative quantum chromodynamics provide a good description of the initial process involving short distances @xmath2 quark - antiquark ( @xmath3 ) and @xmath2 quark - antiquark - gluon ( @xmath4 ) interactions , main responsible for the global characteristics of this kind of process .
the quarks produced off - shell initialize a cascade of several quarks and gluons with complex interactions themselves .
this stage is in general described in probabilistic terms , based on leading - log quantum chromodynamics approximation .
the outgoing colored partons are transformed into color singlet hadrons through the soft hadronization process , forming a jet of particles traveling approximately in the initial parton direction . however , hadronization require the quantum chromodynamics in the soft regime , where the coupling constant become large and there is no easy way to understand it from first principles . for
more than a decade several different models making use of monte carlo techniques were developed to describe such processes @xcite .
these methods avoid the microscopic complexity by being able to define probability distributions for each stage of the jet evolution . in these approaches ,
all known aspects involving quantum chromodynamics were included , the unknown aspects like hadronization being modeled phenomenologically .
each model has a set of free parameters which enables it to reproduce many distributions over a wide range of energy .
in particular , the transverse momenta of the hadrons with respect to the jet direction , main subject of this paper , has been accurately parameterized .
nevertheless , there is no fundamental understanding of the matter @xcite .
one alternative approach to understand the transverse momentum distribution of the charged hadrons produced in annihilation is to look at the whole process in a macroscopic way , trying to extract information through a thermodynamical treatment , not taking into account the microscopy interactions between partons , governed by the quantum chromodynamics . and
@xmath5 because we are looking at the global characteristics of the events ] fermi @xcite was the pioneer in this kind of approach , followed by hagedorn @xcite who provided the most consistent and well succeeded model . nowadays , the thermodynamical formalism is widely used in heavy ions collisions @xcite and has also been applied in @xmath6 , proton - proton ( @xmath7 ) and proton - antiproton ( @xmath8 ) interactions to determine the multiplicity of hadrons produced , with noteworthy results @xcite .
according to hagedorn , a high energy collision produces an _ unlimited _ and _ undetermined _ number of different excited hadron _ fireballs _ that reach a thermodynamical equilibrium . in this relativistic statistical _ ansatz _ , not only the number of particles but also the number of available types of particles itself grows with energy .
the immediate and important consequence of this production mechanism is that _ the temperature is independent of the primary energy _
@xcite . a similar statement
was also made , a few years later , by field and feynman @xcite .
one should expect that this temperature governs the transverse momentum ( @xmath0 ) distribution of the outgoing particles with respect to the jet axis . in a modern ansatz @xcite
, one can assume that the thermodynamical equilibrium is reached in each jet produced in @xmath9 interaction .
we consider in this paper that the thermalization begins just after the hadronization process takes place and finishes when the hadrons are no longer interacting .
since the range of this interaction is of order of a few fermi , the resonances produced , with a typical lifetime of order of @xmath10 s , should decay while the particles are still interacting .
thus , the particles produced from these decays should also participate in the thermalization process . as most charged particles observed in the detector
are produced prompt or through the decay of hadronic resonances , we could expect that the charged particles observed in the detectors have their distributions dictated by the previous ` thermodynamical ' equilibrium .
it is then important to determine the thermodynamical properties of the event , without any kinematical effect caused by the fast relative motion between the hadrons . using the boltzmann - gibbs statistics in this large grand - canonical ensemble for a relativistic gas , the transverse momentum distribution with temperature @xmath11 ,
can be written as ( see hagedorn @xcite ) : @xmath12 where @xmath13 means the longitudinal momentum , @xmath14 and @xmath15 is the mass ( @xmath16 ) @xcite .
after the integration indicated in eq .
( [ eq : pl ] ) and for @xmath17 we obtain an exponential asymptotic distribution for the transverse momentum @xcite : @xmath18 as the temperature should not depend on the center - of - mass energy ( @xmath19 ) , we can expect that eq .
( [ eq : ptb ] ) presents the same shape for all @xmath20 energies spectrum varying only the parameter @xmath21 , which depends on the average multiplicity of the events and consequently on the @xmath19 .
the measured transverse momentum distributions of charged hadrons with respect to the jet axis , defined as the sphericity axis @xcite , for center - of - mass energies in the range of 14 gev to 161 gev are displayed in figure 1 .
one notices the clear experimental deviation , for high values of @xmath0 , from the exponential behavior predicted by eq .
( [ eq : ptb ] ) and indicated by a dotted line .
notice that we could also get decent fits for low values of data point of @xmath0 ( @xmath22 gev ) using the eq .
( [ eq : ptb ] ) , but in this case the best values for the parameter @xmath23 would clearly increase with energy , a feature which disagrees with hagedorn s physically reasonable assumption .
the boltzmann - gibbs statistics , used to derive eq . ( [ eq : ptb ] ) , is suitable for an enormous number of systems , but it is known that it could lead to unphysical results for some systems that present long - range interactions ( lr ) .
in fact , lr interaction is the usual interpretation for the intermittency phenomena @xcite , large particle density fluctuations @xcite , present in hadroproduction .
long - range ( lr ) interactions can occur in many situations , particularly in processes where the interacting volume is of order of the dimension of the hadrons themselves , resulting in a high hadronic density . in this way the gradual deviation of the observed distribution from the exponential function , shown in figure 1 , could be interpreted as a gradual increase of the multiplicity and the hadronic density in the thermalization mechanism .
therefore it is useful , to look for another statistics that , unlike the boltzmann - gibbs , could treat these systems .
recently tsallis @xcite proposed a generalized statistics whose formalism @xcite has been able to describe many systems having as common features the presence of lr interactions and/or non - markovian processes @xcite .
this formalism is based on the entropic form @xmath24 where @xmath25 is the density matrix ( state ) of the system and @xmath26 is a real positive parameter .
when @xmath27 the above entropy tends to the well - known boltzmann - gibbs - shannon ( bgs ) entropy @xmath28 . in this sense the @xmath29 entropy extends the bgs entropy to the nonextensive cases corresponding to @xmath30 .
in fact , in this case the additivity property of the entropy fails because it appears a new term proportional to @xmath31 .
for example , if we look the way the composed entropy of two sub - systems @xmath32 and @xmath33 , @xmath34 , is related to the entropy of each of the two sub - systems , @xmath35 and @xmath36 , when the sub - systems @xmath32 and @xmath33 are statistically independents , one gets the expression @xmath37 .
it is easy here to see that the case @xmath38 reobtains the standard result ( additivity ) from boltzmann - gibbs entropy @xcite .
the extremization of the @xmath29 entropy with adequate constraints lead us to a probability distribution that exhibits a power - law decay ( instead of the exponential decay presented by the bgs entropy ) .
examples of application of this statistics are available in several papers @xcite , whose studied systems are , in general , nonextensive .
thus , assuming a thermal equilibrium of a relativistic gas ( the _ fireball _ ) that obeys the tsallis statistics we obtain , observing the derivation made by @xcite , the following equation for the transverse momentum distribution ( which replaces eq .
( [ eq : pl ] ) ) : @xmath39^{\frac{q}{1-q } } \hspace{0.1 cm } , \label{eq : pt}\ ] ] where @xmath26 is related to the degree of non - extensivity of the system @xcite .
@xmath11 , @xmath13 , @xmath21 and @xmath40 have the same meaning as in the previous equations .
the boltzmann - gibbs limit - eq .
( [ eq : pl ] ) - is achieved as @xmath41 .
the exact expression of the transverse momentum distribution , eq .
( [ eq : pt ] ) , is written in the appendix .
with this expression we fit the experimental data leaving @xmath26 , @xmath23 and @xmath21 as free parameters and the best fits are presented in the figure 1 .
we can see that the asymptotic behaviour of the new distribution , for larger values of @xmath0 , represents very - well the deviation from the exponential for _ all _ center - of - mass energies . .
[ cols="^,^,^,^",options="header " , ] in figure 2 we present @xmath26 and @xmath11 values as functions of the center - of - mass energy .
their numerical values are shown in table 1 .
we can see that the @xmath26 parameter increases smoothly with the energy , saturating around @xmath42 . on the other hand ,
the temperature has a pretty constant behaviour , @xmath43 gev , for the higher energy range .
we note that this constant value for the temperature is not an ad hoc hypothesis ; rather , it was obtained by the fit , indicating that all the energy given to the system is being used to create new _ fireballs _ , as predicted by hagedorn @xcite .
we note that our @xmath11 value is lower than the value obtained in reference @xcite , and this difference is probably due to the fact that , unlike us , the author considered the thermal equilibrium just after the hadronization , i.e. , before the decay of the hadronic resonances . in our process ,
the interaction volume continue to expand , lowering the temperature , until the particles no longer interact among themselves .
also , the statistics used in both papers are different . the smooth increase of @xmath26 with the center of mass energy can be understood as the expected increase of the influence of the lr interactions among the hadronic particles produced in the event . the deviation from an exponential behaviour observed in the transverse momentum distribution in the interaction , is also observed for hadrons produced in other interactions like @xmath44 or heavy ions .
it indicates a universal characteristic of the dynamics of hadron production , probably due to the relevance of the lr regime , related to the high hadronic density .
the thermodynamical treatment using the tsallis statistics discussed here presented a remarkable agreement with the experimental data in more than 4 orders of magnitude in the differential cross section - emphasized with the predicted constant behavior of the temperature @xmath11 .
therefore , the results of this approach suggest an alternative statistical treatment for the transverse momentum distribution within the accepted theory of strong interaction . acknowledgments .
we thank dr .
oliver passon for detailed delphi experimental data and carlos aguiar for useful discussion .
one of us , e.m.f.c . , thanks pronex / mct for the partial financial support .
e - mail : [email protected] , [email protected] , [email protected] 99 marchesini , g. and webber , b. r. , nucl .
b * 310 * , 461 - 526 ( 1988 ) .
sjostrand , t. , int .
phys . a * 3 * , 751 - 823 ( 1988 ) .
chun , s. and buchanan , c. , phys . rep . * 292 * , 239 - 317 ( 1998 ) .
fermi , e. , phys .
* 81 * , 683 - 687 ( 1951 ) .
hagedorn , r. , suppl .
nuovo cimento * 3 * , 147 - 186 ( 1965 ) .
schnedermann , e. , sollfrank , j. and heinz , u. , phys .
c * 48 * , 2462 - 2475 ( 1993 ) .
becattini , f. , z. phys .
c * 69 * , 485 - 492 ( 1996 ) .
becattini , f. , giovannini , a. and lupia , s. , z. phys .
c * 72 * , 491 - 496 ( 1996 ) .
becattini , f. and heinz , u. , z. phys .
c * 76 * , 269 - 286 ( 1996 ) .
field , r.d . and
feynman , r.p .
, n. phys .
b * 136 * , 1 - 76 ( 1978 ) .
althoff , m. et al .
( tasso ) , z. phys .
c * 22 * , 307 - 340 ( 1984 ) .
delphi data get from : untersuchung der energieabhangigkeit inklusiver spektren in der annihilation .
phd thesis from oliver passon ,
november 1997 , wuppertal university- germany .
also private communication from oliver passon .
de wolf , e.a .
dremin , l.m . and kittel , w. , physics reports * 270 * , 1 - 141 ( 1996 ) .
adamus , m. et al .
( na22 ) , phys .
b * 185 * , 200 - 204 ( 1987 ) .
burnett , t.h .
( jacee ) , phys .
lett . * 50 * , 2062 - 2065 ( 1983 ) .
tsallis , c. , j. stat .
phys . * 52 * , 479 - 487 ( 1988 ) .
curado , e. m. f. and tsallis , c. , j. phys .
a * 24 * , l69-l72 ( 1991 ) ; plastino , a. and plastino , a. r. , phys .
a * 226 * , 257 - 263 ( 1997 ) ; mendes , r.s .
, physica a * 242 * , 299 - 308 ( 1997 ) ; tsallis , c. mendes , r. s. and plastino , a. r. , physica a * 261 * , 534 - 554 ( 1998 ) .
plastino , a. r. and plastino , phys .
a * 174 * , 384 - 386 ( 1993 ) ; lavagno , a. kaniadakis , g. rego - monteiro , m. quarati , p. and tsallis , c. , astrophys . lett . and commun .
* 35 * , 449 - 455 ( 1998 ) ; boghosian , b. m. , phys . rev .
e * 53 * , 4754 - 4763 ( 1996 ) ; zanette , d. h. and alemany , p. a. , phys .
* 75 * , 366 - 369 ( 1995 ) ; lyra , m.l . and tsallis , c. , phys .
* 80 * , 53 - 56 ( 1998 ) .
tirnakli , u. zeren , s. bykkili , f. and demirhan , d. , z. phys .
b * 104 * , 341 - 345 ( 1997 ) .
tsallis , c. , braz .
* 29 * , 1 - 35 ( 1999 ) .
after some manipulation , the solution of the integral in eq .
( [ eq : pt ] ) was obtained by the software mathematica 3.0 and the transverse momentum distribution can be written as : | we show that the inclusion of long distance effect , expected in strong interactions , through a nonextensive thermodynamical approach is able to explain the experimental distribution of the transverse momentum of the hadrons with respect to the jet axis ( @xmath0 ) @xmath1 reaction .
the observed deviation from the exponential behavior , predicted by the boltzmann - gibbs thermodynamical treatment , is automatically recovered by the nonextensive tsallis statistics used here .
we fitted the observed @xmath0 spectrum in the range of 14 gev to 161 gev and obtained , besides a good fit , the theoretical important fact that the temperature becomes independent of the primary energy .
xxxxxxxxxxxxxxxxxx= high energy ; hadroproduction ; fireball ; statistical models .
+ 24.10.pa , 13.65.+i , 13.60.hb , |
type ia supernovae are thought to be the result of thermonuclear explosion of carbon - oxygen white dwarfs as they reach or exceed the chandrasekhar mass limit ( see , e.g. , @xcite for reviews ) .
broadly speaking , there are two main classes of models : the accreting models , in which the white dwarf more or less steadily accretes matter from a ( hydrogen - rich ) companion star @xcite , and the merging models , in which two white dwarfs coalesce under the influence of angular momentum losses due to the emission of gravitational waves @xcite .
deciding which of the scenarios contribute to the observed sn ia rates is difficult from a purely theoretical point of view .
one may , however , expect observable differences between the accreting and merging scenarios . in an effort to constrain the progenitor scenario for individual sne ia
, @xcite recently started a search for progenitor detections of newly reported , nearby sne ia in archival _ chandra x - ray observatory _
images , arguing that the x - ray luminosities of progenitors in the accreting models are expected to be much higher than those of the progenitors in the merging models .
their search was successful : the nearby type ia supernova 2007on in ngc1404 @xcite turned out to have an x - ray counterpart in a combined 75ks _ chandra_/acis exposure , taken 4.5 years before the explosion .
the x - ray counterpart was detected at the 5-@xmath3 level , and within @xmath4 of the optical supernova .
based on the low chance alignment probability of @xmath21% , it was claimed to be the likely progenitor , favouring an accreting white dwarf scenario over a merger scenario for 2007on @xcite . given the importance of the detection of the progenitor to a type - ia supernova , we obtained follow - up _ chandra _ observations ( section [ ddtsection ] ) and performed accurate astrometry of the field of 2007on using existing and new imaging ( section [ astrometrysection ] ) .
we briefly review the further evidence as to the nature of the x - ray source in section [ discussionsection ] , and conclude with a summary of our results in section [ conclusionsection ] .
following the detection of the likely progenitor to 2007on by @xcite , we were granted 40ks of _
chandra_/acis observations through the director s discretionary time ( ddt ) programme . due to pointing restrictions ,
2007on was observed for 18.5ks on december 24 , and 21.5ks on december 27 , 2007 .
we used the acis - s instrument with the source positioned at the aimpoint on ccd s3 .
being one of the two back - illuminated ccds , s3 has the best sensitivity to soft x - rays ; the archival x - ray source close to 2007on had been observed to be quite soft .
figure [ fig : positions_chandra ] shows the pre - outburst image together with the new ddt observations .
the source appears to have become fainter , with 2 photons in a 1.0@xmath1 aperture that should contain @xmath290% of the flux , centered on the position of the x - ray source in the archival images , and counting only photons in the @xmath5kev energy range ( for rough comparison , there were 17 such photons in a 1.0@xmath1 aperture in the progenitor image ) .
given the measured average background rate of 1.3 photons per 1.0@xmath1 aperture , the observed 2 photons suggest that the source may have disappeared altogether .
a larger , 2@xmath1 aperture contains 10 counts ( versus 22 in the pre - outburst image ) ; this has a 4% probability of happening by chance for a purely poissonian background , or a measured 8% probability of happening in a random aperture between 4@xmath1 and 20@xmath1 from the supernova in the post - outburst image .
there is thus no convincing evidence that there is still a source in the follow - up observations .
as expected the wavdetect source - finding algorithm from the standard ciao data - reduction package ( version 4.0 ) does not find a source when run with the standard false - alarm probabilities of @xmath6 per pixel . in order to be able to compare both epochs , we determined the relative effective collecting area times exposure time of the individual observations . since part of the progenitor data
were taken with one of the front - illuminated acis - i ccds , which have different sensitivities especially in soft x - rays , and since furthermore the soft - x - ray sensitivity of acis has overall degraded significantly since spring 2003 when the progenitor data were taken , the relative effective areas of the pre- and post - outburst data were expected to be a function of the assumed source spectrum .
we thus worked out the relative effective areas times exposure times for the 2003 and 2007 epochs as a function of source temperature , which are shown in fig .
[ fig : probs ] . as it turned out ,
the relative sensitivities were very much independent of the assumed source spectrum : the degradation in soft x - ray sensitivity of acis conspired with the lower soft x - ray sensitivity of the acis - i ccd used for part of the progenitor observations to produce an almost identical sensitivity - versus - energy curve for both epochs .
this made comparison of both epochs considerably easier , since we did not have to worry about the source spectrum , or possible spectral changes between both epochs .
we performed a fully poissonian monte carlo simulation to determine the probability of observing , by chance , the observed lower number of photons in the post - outburst image compared to the pre - outburst image , with the same ( though unknown ) underlying source luminosity .
we did so for a range in apertures and , for completeness , also as a function of source spectral temperature , even though we had already concluded that the relative sensitivities of the observations in both epochs are quite insensitive to the assumed source temperature .
the results are also shown in fig . [
fig : probs ] . the confidence level for the x - ray source having dimmed increases with smaller aperture , to @xmath799% for apertures of radii @xmath81.25@xmath1 centered on the position of the pre - outburst x - ray source .
sn 2007on fell a distance 2.5@xmath9 and 1.7@xmath9 away from the optical axis in the pre - outburst acis - i and acis - s observations , respectively , so that a 1.0@xmath1 aperture should still contain approximately 80% of the photons .
however if one allows the source to be slightly extended or blended , a larger aperture may be appropriate , with a correspondingly lower confidence level .
the slightly smaller psf of the on - axis , post - outburst observations will make us underestimate the confidence level of the source having dimmed or disappeared .
we should mention that these results change slightly if we do the data analysis with the standard photon position randomization applied during the data reduction .
while there is no scientific justification for this randomization , it does shift several photons in the post - outburst image by 1 pixel , such that 4 instead of 2 fall within a 1.0@xmath1 aperture .
this leads to confidence levels of the source having dimmed of 97% and 91% for 1.0@xmath1 and 1.25@xmath1 apertures respectively , while the confidence levels for the larger apertures go up slightly compared to the non - randomized results .
finally , to test for variability of the source in the pre - outburst observations , insofar as the limited number of photons allows , we performed a kolmogorov
smirnov test against a uniform distribution ( more accurately a set of uniform distributions with poisson - distributed average rate ) for both the acis - s and acis - i observations .
the photon arrival times , and also those of subsets of only the soft or hard x - rays , were found to be perfectly compatible with uniform distributions , within the 1-sigma confidence intervals .
there is thus no evidence for variability in the pre - outburst images .
@xcite found an offset between the optical sn and the x - ray source of @xmath4 , entirely compatible with the x - ray source being the progenitor of the supernova . in order to verify that there is no significant offset between the two , we tried to improve on their astrometry .
we retrieved archival observations of ngc1404 obtained with the wide field imager ( wfi ) at the eso 2.2-m telescope , and of sn2007on taken november 30 , 2007 with the eso faint object spectrograph and camera ( efosc ) at the 3.6-m telescope , both located at la silla observatory .
a 5-min @xmath10-band wfi image of ngc1404 was used to astrometrically calibrate the efosc observation of the supernova .
a total of 76 stars from the two - micron all - sky survey ( 2mass ) coincided with the @xmath11 field - of - view of a single wfi chip ; 38 of these were not saturated and appeared stellar .
after iteratively removing outliers we obtained an astrometric solution using 31 2mass stars , yielding root - mean - square ( rms ) residuals of @xmath12 in both right ascension and declination .
a list of secondary astrometric standard stars was compiled by measuring the positions of stars on the wfi image .
these calibrated positions were used to transfer the astrometry onto the 20sec @xmath10-band efosc image of the supernova .
the final astrometric calibration used 71 stars common to both the wfi and the efosc image , with rms residuals of @xmath13 in right ascension and @xmath14 in declination . based on this astrometric calibration
, we determined the position of the supernova in the efosc image to be @xmath15 and @xmath16 .
the uncertainty of the position on the efosc image is about @xmath17 in both coordinates , while the uncertainty on the absolute position is @xmath18 in right ascension and @xmath19 in declination ( i.e. the quadratic sum of the uncertainties in the tie between the 2mass catalog and the wfi image , the tie between the wfi and the efosc images and the uncertainty of the supernova position in the efosc image ) . in order to compare the optical position of the supernova with the x - ray position of the candidate counterpart in the pre - outburst _ chandra _ acis - i ( obsid 4174 ) and acis - s ( obsid 2942 ) observations
, we determined a boresight correction using x - ray sources ( from wavdetect ) which , by eye , had likely optical counterparts in the efosc images . objects that appeared elongated or blended in either the optical of x - ray image were rejected , and one rejection iteration of optical sources outside the 95% confidence circles as given by wavdetect was applied . for the acis - i observation we thus found 7 suitable x - ray sources , which provided an offset to the raw x - ray positions of @xmath20 in right ascension and @xmath21 in declination . for the acis - s observation , we found 10 x - ray sources , providing an offset of @xmath22 in right ascension and @xmath23 in declination .
the boresight corrections thus appear to be very small .
[ fig : offsets ] shows the spatial distribution of the x - ray sources used for referencing the x - ray to the optical image .
arrow indicates a 0.25@xmath1 displacement .
circles and arrows drawn with thick and thin strokes show the reference sources for the acis - i and acis - s observations , respectively . ] applying these corrections to the pre - outburst x - ray observations gives @xmath24 and @xmath25 for the absolute position of the x - ray source in the combined image , with uncertainties of @xmath26 in right ascension and @xmath27 in declination .
for the relative positions of the optical sn and the x - ray source , the uncertainty in the referencing to the absolute icrs frame is irrelevant , so the error in their relative positions is smaller .
we find that the x - ray source is offset by [email protected]@xmath1 from the optical supernova , or [email protected]@xmath1 _ excluding _ the error on the position of the x - ray source as given by wavdetect ( i.e. , only including the positional uncertainty of the optical sn relative to its reference stars , and the tie between the optical and x - ray reference stars ) , where we have conservatively added the errors on the boresight corrections in quadrature for want of a better method .
the uncertainty in the position of the x - ray source , relative to its reference sources , is thus by far the dominant uncertainty , as expected . to estimate the probability of such a misalignment happening by chance , i.e. with the x - ray source near 2007on still corresponding to the optical supernova
, we performed a detailed monte carlo simulation in which we added artificial sources to the combined progenitor image from the _ chandra _ archive .
we began by running wavdetect on the image to detect the sources in the image . from the source and smoothed background images returned by wavdetect , we filtered all pixels that had an average background within 10% of the value measured around the progenitor ( 6.9 counts per 2@xmath1 aperture ) , and that had no detected source flux anywhere within five times this radius .
we randomly picked coordinates from the qualifying pixels , took a random sub - pixel sampling , and added photons according to a bivariate normal distribution , until the number of photons in a 2@xmath1 aperture equalled 21 , the number found in the pre - outburst images @xcite . since the psf width in _ chandra _
images varies quite a bit with energy , we probably gained little by modelling the psf with a more complicated function .
instead we ran the simulation for three reasonable widths of the distribution , choosing 0.8@xmath1 , 1.0@xmath1 and 1.2@xmath1 radii of 86% encircled energy , and doing 40,000 trials for each .
the resulting images were again analysed with wavdetect , and the positional offsets between the artificial sources and the reconstructed artificial sources were recorded .
[ fig : positionerror ] shows the cumulative distribution of these offsets .
the probability of finding a source a distance 1.15@xmath1or more away from the input source is low , close to 1% for the 1.0@xmath1 psf , and slightly higher and lower for the larger and smaller psfs , respectively .
an interesting feature of the probability distributions shown in fig .
[ fig : positionerror ] are the multiple components : a component that drops off quickly on roughly the pixel size scale , and a much slower component .
presumably , this slower component represents cases in which the new source combines with an undetected source or overdensity in the nearby background , pushing it above the detection threshold , and being detected as a source offset from the input source . to test this hypothesis we ran a further monte carlo simulation , this time with purely poissonian backgrounds of the same average value ( within 10% of 6.9 counts per 2@xmath1 aperture ) .
the results are shown as the dotted lines in fig .
[ fig : positionerror ] . indeed , with the artificial backgrounds the slow component vanishes and the number of cases in which an offset of 1.15@xmath1 or more is recorded drops , to about 0.3% for a 1.0@xmath1 psf .
in order to test whether the wavdetect algorithm could be responsible for some of the positional scatter in the monte carlo simulation , we repeated the analysis with a simple centroid algorithm , which we again applied only to simulated sources that were first detected with wavdetect in order to ensure that they are still representative of sources that would have been found in the original search by @xcite .
we measured centroids for the @xmath28 set of simulated images with a 1.0@xmath1 psf .
the amount of scatter in the centroid positions depends on the aperture chosen : for a 2@xmath1 aperture it is slightly smaller , while for 1@xmath1 and 3@xmath1 apertures it is slightly larger than the wavdetect scatter .
an adaptive aperture , starting from 3@xmath1 and decreasing in 10% steps to 1.5@xmath1 after positional convergence in each step , gives the same scatter as a 2@xmath1 aperture , suggesting that this is about the best one can do ( in minimizing the positional scatter ) .
for this ` optimal ' aperture , the x - ray source centroid is offset from the optical sn by 1.09@xmath1 , slightly less than the wavdetect offset of 1.15@xmath1 .
the net result is that the offset probability is very similar for centroid and wavdetect positions , at approximately 1% .
the important question now is how these numbers compare to the probabilities of a chance alignment of the optical sn with an x - ray source in the field .
for this we counted the number of x - ray sources in the progenitor image found by wavdetect .
we filtered out the region around ngc 1404 where the background is within 50% of the value measured around 2007on , which should be representative of both the position where a sn would typically be discovered , and of the probability of detecting an x - ray source there with wavdetect . in an area measuring 7.1
square arcmin wavdetect finds 17 sources , giving a probability of 0.28% for a chance alignment within 1.15@xmath1 of the optical position of 2007on .
note that this figure somewhat conservatively includes sources that are detected with a lower significance level than the possible progenitor of 2007on .
counting only equally or more significant detections gives 14 sources , or a 0.23% chance alignment probability .
given that sn 2007on was the fourth trial in their search for progenitor detections in archival _ chandra _ data @xcite , one has to conclude that the chance alignment probability for the detected source is @xmath20.9% , or @xmath21.1% for any detectable source .
, 1.0@xmath1 and 0.8@xmath1 psf widths in the progenitor x - ray images .
dotted lines are the same but for artificial , pure - poissonian backgrounds .
the dashed line shows the trial - corrected chance alignment probability , and the solid vertical line the observed offset between the optical and x - ray source . ]
based on just the astrometry , the chances of the x - ray source being related to the supernova are small but equal to the chances of a chance alignment . the fact that the x - ray source appears to have dimmed ( or disappeared ) in the _ chandra _ ddt observations suggests that at least some of the x - ray photons in the progenitor images may have come from the progenitor , although alternatively , the x - ray source might be intrinsically variable ( and unrelated ) . in this section
we try to collect further evidence that may point to a relation ( or not ) between the x - ray source and the supernova .
@xcite mentioned the softness of the x - ray source close to 2007on as further suggestive evidence of a connection between the two . in the accreting models ,
the accretion of matter by a white dwarf at a high rate is expected to create a strong , soft x - ray source ( e.g. @xcite ) , which makes the hardness of the source a potentially useful property .
kev , @xmath29kev and @xmath30kev , respectively .
the source close to 2007on is the left - most data point in both panels , shown in bold . ]
[ fig : hardness ] shows the hardness ratios of all x - ray sources in the combined pre - outburst image that are detected with at least 15 photons .
photons are binned into s(oft ) , m(edium ) and h(ard ) energy bands , corresponding to @xmath31kev , @xmath29kev and @xmath30kev photon energies , respectively , and colours are constructed following the method of @xcite .
the source close to 2007on is the softest source in terms of its soft - to - medium hardness ratio , while its medium - to - hard hardness ratio is quite hard and more comparable to the other sources in the field .
its spectrum is thus difficult to interpret , but a possible explanation could be that there are two components to the spectrum : a very soft one and a rather hard one . although a few extra photons in the m band would have pushed the source in the direction of the more ` mundane ' region occupied by the majority of sources in the upper panel of fig .
[ fig : hardness ] , the source would still have shown an overabundance of soft x - rays , as is clear from the lower panel in this figure . to further investigate the possibility of a two - component source , we show in fig . [
fig : positions_energies ] the pre - outburst x - ray image divided into soft ( @xmath31kev ) and medium / hard ( @xmath32kev ) photons , together with the position of the optical supernova .
the obvious problem is that the photon statistics of these images become even worse .
nevertheless the medium / hard photons are seen to line up with the position of the reconstructed x - ray source , while the soft photons appear more scattered .
although the photons closest to the optical sn are soft , there is no general trend by which the soft photons line up better with the optical position .
the figure does therefore not provide convincing evidence for a connection between the optical supernova and the soft x - rays in particular . given the indication that the progenitor x - ray source may ( partly ) have been an unrelated , nearby source , we reanalyse the archival _ hst _ images of the region around 2007on , while performing the same accurate astrometry .
figure [ fig : positions_hst ] shows a drizzled @xcite 1224-s image taken august 6 , 2006 with the advanced camera for surveys ( acs ) using the f814w far - red filter . there is evidence for a faint , extended source close to the position of the x - ray source , although it is close to the detection limit of @xmath33 .
nothing is visible at the optical position of the sn .
if we compare the observed x - ray flux @xmath34ergs@xmath35@xmath36 in the @xmath5kev band ( see @xcite , after a correction to the value given in @xcite ) to the @xmath37-band flux @xmath38 of the object in the acs image , and we compare this to results from the _ chandra _ deep field north & south surveys @xcite , we conclude that the observed flux ratio @xmath39 is at or beyond the upper ends of the distributions observed in those surveys .
this makes it unlikely that all of the observed x - ray photons originate from this possible background source , although it is conceivable that a few of them do .
nevertheless we conclude that an origin within ngc1404 for the x - ray source appears to be more likely . ) , although it is close to the detection limit of @xmath40 .
nothing is visible at the position of the optical supernova ( @xmath41 ) . ]
we have presented accurate astrometry of the field of type - ia supernova 2007on , and conclude that there appears to be an offset of [email protected]@xmath1 between the optical supernova and the x - ray source close to it in archival _ chandra _ images .
based on just this information , and barring astrometric errors , the probability of the x - ray source being related to the supernova is small : about 1% .
the probability for a chance alignment with an unrelated x - ray source is , unfortunately , equally small , which makes it difficult to decide between the two scenarios .
several additional pieces of information suggest that at least part of the x - ray source may have been related to the supernova , although none of them is decisive .
first there is the indication that the x - ray source has dimmed ( or disappeared ) in the post - outburst _ chandra _ images taken about six weeks after optical maximum , at the 9099% confidence level depending on the choice of aperture size .
second , the source in the pre - outburst images shows an excess of soft x - ray photons relative to the other sources in the field , and sn ia progenitors in the accreting models are expected to be prolific sources of soft x - rays .
one could argue that the probability for a chance alignment between the optical sn and a _ soft _ x - ray source is much lower than the quoted 1% probability of an alignment with _ any _ x - ray source , which would make a physical connection between the x - ray source and the optical sn _ less unlikely _ than a chance alignment . a possible extended background object close to the x - ray source position ( but away from the optical supernova ) , as seen in archival _ hst_/acs data , could conceivably be responsible for some of the observed hard x - rays . probably the only way out of the impasse will be future _ chandra _ observations of 2007on , which can decide whether the x - ray source close to 2007on really has dimmed or disappeared .
x - rays from the supernova are not expected to be detectable with _
chandra _ : @xcite consider compton scattering of gamma rays from radioactive decay which , although model - dependent , is expected to fall short of the detection limit by three orders of magnitude , while the x - ray luminosity from interaction of the sn ejecta with the local interstellar medium is expected to fall off as 1/time and should thus have decreased by an order of magnitude since our ddt observations .
a future detection would thus imply that the source is unrelated .
although the possibility of the x - ray source having been an unrelated transient source can never be ruled out completely , a future non - detection would further lower the probability of the source having been a chance alignment , since only a few per cent of all x - ray sources observed in a given extragalactic snapshot are transients @xcite .
gr is supported by nwo rubicon grant 680.50.0610 to g.h.a .
rv was supported by the dfg cluster of excellence ` origin and structure of the universe ' .
gn is supported by the netherlands organisation for scientific research .
we thank the director of the chandra x - ray observatory for allocating director s discretionary time , and the referee for suggesting to redo our analysis without the standard photon position randomization .
this research has made use of data obtained from the chandra data archive and software provided by the chandra x - ray center in the application package ciao . based on observations made with the nasa / esa hubble space telescope , obtained from the data archive at the space telescope science institute .
stsci is operated by the association of universities for research in astronomy , inc.under nasa contract nas 5 - 26555 . based on data obtained from the eso science archive facility . based on observations made with eso telescopes at the la silla observatory under programme 080.a-0516 .
alexander d. m. , brandt w. n. , hornschemeier a. e. , garmire g. p. , schneider d. p. , bauer f. e. , griffiths r. e. , 2001 , aj , 122 , 2156 fruchter a. s. & hook r. n. , 2002 , pasp , 114 , 144 gal - yam a. , simon j. , klotz a. , rosolowsky e. , 2007 , atel , 1263 giacconi r. et al .
, 2002 , apjs , 139 , 369 van den heuvel e. p. j. , bhattacharya d. , nomoto k. , rappaport s. a. , 1992 , a&a , 262 , 97 hillebrandt w. & niemeyer j. c. , 2000 , ara&a , 38 , 191 iben jr i. & tutukov a. v. , 1984 , apjs , 54 , 335 immler s. & brown p.j . , 2007 , atel , 1261 immler s. et al . , 2006 ,
apj , 648 , 119 leibundgut b. , 2000 , a&a rev . , 10 , 179 morrell n. , folatelli g. , stritzinger m. , 2007 , cbet , 1131 , 1 nelemans g. , voss r. , roelofs g. , bassa c. , 2008 , mnras , 388 , 487 nomoto k. , 1982 , apj , 253 , 798 pollas c. & klotz a. , 2007 , cbet , 1121 , 1 prestwich a. h. , irwin j. a. , kilgard r. e. , krauss m. i. , zezas a. , primini f. , kaaret p. , boroson b. , 2003 , apj , 595 , 719 tutukov a. v. , yungelson l. r. , 1981 , nauchnye informatsii , 49 , 3 voss r. & nelemans g. , 2008 , nature , 451 , 802 voss r. & gilfanov m. , 2007 , a&a , 468 , 49 webbink r. f. , 1984 , apj , 277 , 355 whelan j. & iben i. j. , 1973 , apj , 186 , 1007 [ lastpage ] | we present new _
chandra _ x - ray observations and detailed astrometry of the field of the type ia supernova 2007on , for which the detection of a likely progenitor in archival _ chandra _ data was recently reported .
no source is detected in the new _ chandra _ images , taken six weeks after optical maximum .
we calculate a 9099% probability that any x - ray source near the position of the supernova ( sn ) is fainter than in the pre - outburst images , depending on the choice of aperture , which supports the identification of the archival x - ray source with the sn .
detailed astrometry of the x - ray and new optical images , however , gives an offset between the supernova and the measured x - ray source position of [email protected]@xmath1 .
extensive simulations show that the probability of finding an offset of this magnitude is @xmath21% , equal to the ( trial - corrected ) probability of a chance alignment with any x - ray source in the field .
this casts doubt on the identification of the x - ray source with the progenitor , although the scenario in which at least some of the observed x - rays are connected to the supernova may be the _
least unlikely _ based on all available data .
after a brief review of the auxiliary evidence , we conclude that only future x - ray observations can shed further light on the proposed connection between the x - ray source and the progenitor of sn 2007on , and thus whether an accreting white dwarf scenario is truly favoured for this sn ia .
[ firstpage ] supernovae binaries : close white dwarfs x - ray : binaries |
this paper is concerned with mathematics of diffraction .
more specifically we are interested in the famous inverse problem for diffraction : given something that is putatively the diffraction of something , what are all the somethings that could have produced this diffraction .
diffraction has been a mainstay in crystallography for almost a hundred years . with the proliferation of extraordinary new materials with varying degrees of order and disorder
, the importance of diffraction in revealing internal structure continues to be central .
the precision , complexity , and variety of modern diffraction images is striking , see for instance the recent review article @xcite . nonetheless , in spite of many advances ,
the fundamental question of diffraction , the inverse problem of deducing physical structure from diffraction images , remains as challenging as ever .
the discovery of aperiodic tilings and quasicrystals revived interest in the mathematics of diffraction , particularly since the types of diffraction images generated by such structures essentially or exactly pure point diffraction together with symmetries not occurring in ordinary crystals had not been foreseen by mathematicians , crystallographers , or materials research scientists .
starting with @xcite , the resulting new techniques created in order to understand mathematical diffraction have been considerable . among them
we mention the introduction of ideas from the harmonic analysis of translation bounded measures @xcite , ergodic dynamical systems @xcite , and stochastic point processes @xcite .
the diffraction of a structure is defined as the fourier transform of its autocorrelation .
the latter is a positive - definite measure in the ambient space of the structure typically @xmath0 , but mathematically any locally compact abelian group @xmath1 and the diffraction is then a centrally symmetric positive measure in the corresponding fourier reciprocal space either @xmath2 or generally @xmath3 as the case may be . in this paper
we examine three basic questions that arise from this theory .
the first question asks whether _ every _ positive centrally symmetric measure is actually the diffraction of something .
the second asks about the classification of the all various ` somethings ' that have this given diffraction .
the third asks what kinds of structures the ` somethings ' can be . in the case of pure point diffraction ( when the diffraction measure is a pure point measure ) we can give a satisfactory answer to these problems .
we will be in the setting of locally compact abelian groups .
the difficulties of the questions , at least as we examine them here , have little to do with the generality of the setting .
they are just as hard for @xmath0 .
indeed the second question is a very old one in crystallography , called the _ homometry problem _ and is part of _ the _ fundamental problem of diffraction theory , namely how to unravel the nature of a structure that has created a given diffraction pattern .
our first question immediately raises the third question . exactly what do we mean by a structure that can diffract ?
typically diffractive structures are conceived of as discrete measures , for instance point measures describing the positions of atoms , vertices in tilings , etc . , weighted by appropriate scattering strengths , or as continuous measures describing the ( scattering ) distribution of the material structure in space .
more recently , starting with the work of gour @xcite there has been a shift to discussing the diffraction of certain point processes , which in effect are random point measures , and various distortions of these ( see e.g. @xcite ) . in this paper
we are led to introduce new structures which we call a _ spatial stationary processes_. we shall show that these always lead to autocorrelation and diffraction measures and contain the typical theories of diffraction discussed above . in terms of these
we get an attractive answer our three questions , in the case of pure point diffraction . however , the answer to the third question is interesting , because it does not describe the underlying structure explicitly , but rather in terms of the ways that it appears to compactly supported continuous functions on the ambient space . in this respect
it resembles the description of measures given by the riesz representation theorem or the theory of distributions with their spaces of test functions . in our case
, the convergence conditions of measures do not seem to be generally valid . in this respect
there is much to be learned about the full scope of which structures can diffract .
let @xmath1 be a locally compact abelian group and @xmath4 its dual group .
we treat these groups in additive notation .
we begin with an abstract notion of a stationary process for @xmath1 and then shift our attention to the more concrete notion of a spatial stationary process for @xmath1 ( [ stochasticprocesses ] and [ spatial ] ) . as for terminology
let us note that outside of the discussion of usual stochastic processes , the word ` stochastic ' is not used in our discussion of processes and stationary processes .
a spatial stationary process basically consists of a @xmath1-invariant mapping @xmath5 where @xmath6 is a probability space on which there is a measure preserving action of @xmath1 .
the idea here derives directly from the theory of stochastic point processes , and it is perhaps useful to discuss this briefly .
we follow @xcite here .
a stochastic point process , say in @xmath7 , is a random variable @xmath8 on a probability space @xmath9 with values in the space @xmath10 of point measures @xmath11 on @xmath7 for which @xmath12 for all bounded subsets @xmath13 of @xmath7 .
these are point measures and can be written in the form @xmath14 where @xmath15 can be a positive integer or infinity and the points @xmath16 are ( not necessarily distinct ) points of @xmath7 . usually the space @xmath10 is far larger than needed .
for one thing it allows points in @xmath7 to appear with multiplicities and for another it does not require any minimal separation between distinct points ( a hard core condition ) that would usually be imposed in crystallography or in tiling theories . we assume then that we are really only interested in some subset @xmath17 of permissible outcomes for the point process , and that almost surely the point process produces measures in @xmath18 .
typically we wish @xmath18 to be invariant under translation since the position of our point sets is not particularly relevant , and we assume that the law @xmath19 of the process ( the probability measure @xmath19 induced on @xmath18 by the random variable ) is invariant under translation ( i.e. the process is stationary ) .
now the point is this . if @xmath20 then we obtain a random variable on @xmath9 by @xmath21 , where @xmath22 is the measure indicated above . in effect
@xmath23 can be viewed as a mapping from @xmath24 into mappings on @xmath18 . under mild assumptions @xmath25 . as is often the case with random variables , we ignore the underlying probability space @xmath9 since we have everything we need from the probability space @xmath6 and the mapping @xmath23 .
then @xmath23 itself can be referred to as the stochastic point process . in this paper @xmath7
is replaced by any locally compact abelian group @xmath1 . also @xmath18 is not assumed to have any particular form ( e.g. a set of point sets , or a set of measures ) .
we do not wish to assume any particular mathematical structure on the outcomes of the processes we are studying .
our point of view is that we will start with a measure in @xmath4 which we would like to show is the diffraction of something .
but we do not know in advance what sort of mathematical objects could produce this diffraction ( point sets , continuous distributions of density , schwartz distributions , etc . ) .
we derive information about @xmath18 only from the behaviour of the functions @xmath26 , @xmath27 , upon it .
this is , physically , rather appropriate . in a physical situation
one determines the structure of a physical object by measurements upon it and the results of measurements are ultimately all that one can know about it .
we are particularly interested in pure point diffraction .
however , the concept of stationary processes seems interesting in its own right and we develop the theory of these a bit , independent of the question of pure point diffraction .
it is even useful initially to ignore the fact that the resulting process is _ spatial _ , that is , it is supposed to result from some structure in @xmath1 .
so at the beginning we often only require that we have a @xmath1-invariant mapping @xmath28 where @xmath29 is a hilbert space with conjugation we call these stationary processes . assuming existence of second moments , we can then , for each such process , find an ` autocorrelation measure ' on @xmath1 and a ` diffraction measure ' on @xmath4 .
the latter is always a centrally symmetric positive measure , and through it we can define continuous and pure point diffraction .
we show how pure point and continuous diffraction are tied to notions of almost periodicity of @xmath23 .
also we show how to decompose a sum of spatial processes into sub - processes which are respectively pure point and continuous . coming to the pure point part of the paper , [ purepoint ] ,
we assume given a centrally symmetric positive pure point measure @xmath30 on @xmath4 , which can be expressed as a fourier transform .
we wish to create a probability space @xmath6 on which there is an ergodic action of @xmath1 and a continuous stationary process @xmath31 whose diffraction is @xmath30 .
not surprisingly phase factors play a crucial role in this since it is phase information that is lost in the process of diffraction .
there is a natural combinatorial type of group @xmath32 generated by the positions of the bragg peaks .
this group can be seen as an abeleanized homotopy group of an associated cayley graph .
it is the dual group of @xmath32 , or the dual group of a natural factor @xmath33 of this group , that classifies all the spatial processes on @xmath1 with diffraction @xmath30 , the classification being up to isomorphism or up to translational isomorphism respectively .
we call these phase factors _ phase forms_. background on phase forms is given in section [ cycles ] . in [ cyclestostochasticprocesses ]
we show how to use @xmath30 and a phase form @xmath34 to construct a spatial process whose diffraction is @xmath30 .
uniqueness is already shown in [ uniqueness ] .
how all these ingredients give a solution to the homometry problem is discussed in [ homometry ] .
our solution to the pure point problem is in some sense an elaboration of the famous halmos - von neumann result about realizing ergodic pure point dynamical systems in terms of the character theory of compact abelian groups .
there the objective was to classify pure pointedness at the level of the spectrum . in our case
we wish to classify the diffraction , which amounts to not only knowing the spectrum but also the intensity of diffraction at each point of the spectrum .
our approach to the construction of a spatial process is to realize @xmath18 as the dual to the discrete group generated by the set of bragg peaks , but at the same time to include all the phase information which derives from the bragg intensities . in this context
we mention the recent work of robinson @xcite on how to realize systems with a given group of eigenvalues via so called cut - and - project schemes .
in [ compactg ] we take a closer look at the situation that the underlying group is compact . in this case
all processes can be realized as measure processes under some weak assumption .
we then specialize even further and in [ simpleperiodic ] discuss some results of grnbaum and moore on rational diffraction from one dimensional periodic structures .
these pertain to the simple case that @xmath35 , but even so , the results are interesting and not easy to obtain . finally in
[ altconstruction ] we sketch out a second approach to the construction of a spatial process , this time based on gelfand theory . summarizing : *
each pure point ergodic spatial process gives rise to a pure point diffraction measure @xmath30 and a phase form @xmath34 .
* spatial processes with the same pure point diffraction measure and the same phase form are naturally isomorphic up to translation .
* the set of all possible phase forms associated with a given pure point measure @xmath30 form an abelian group @xmath36 , and for each choice of @xmath37 , there is a ergodic spatial process with diffraction @xmath30 and phase form @xmath34 .
the paper has two main features , after the introduction of stationary and spatial processes : one is a development of a general theory of these types of processes and the other is a detailed study of pure point diffraction and examples of how it looks in special cases .
readers interested primarily in the pure point theory can skip sections @xmath38 on first reading if they so wish .
in this section we gather some notation used throughout . @xmath39 .
let @xmath1 be a locally compact abelian group .
we let @xmath40 denote the space of compactly supported complex valued continuous functions on @xmath1 . for compact @xmath41 , @xmath42 is the subspace of elements of @xmath40 whose support is in @xmath15 . for @xmath27 ,
@xmath43 is defined by @xmath44 .
the translation action of @xmath1 on itself is denoted by @xmath45 : @xmath46 for all @xmath47 .
this action determines , in the usual way , an action on @xmath40 . by @xmath48
we denote the ( fixed ) haar measure on @xmath1 .
the convolution of two functions @xmath49 is defined to be the function @xmath50 given by @xmath51 the dual group of @xmath1 is denoted by @xmath52 and the _ fouriert ransform _ of an @xmath53 is denoted by @xmath54 i.e. @xmath55 a measure @xmath56 on @xmath1 is called _ transformable _ if there exists a measure @xmath57 on @xmath52 with @xmath58 for all @xmath53 . in this case
@xmath57 is uniquely determined and called the _ fourier transform of @xmath56_. in such a situation we call @xmath57 _ backward transformable_. unlike all other notation introduced in this section the concept of backward transformability is not standard .
however , it is clearly very useful in our context as it is exactly the property shared by the diffraction measures we investigate .
a sequence @xmath59 of compact subsets of @xmath1 is called a _ van hove sequence _ if for every compact @xmath60 @xmath61 here , for compact @xmath62 , the `` @xmath15-boundary '' @xmath63 of @xmath64 is defined as @xmath65 the existence of van hove sequences for all @xmath57-compact locally compact abelian groups is shown in @xcite , see also section 3.3 and theorem ( 3.l ) of ( * ? ? ? * appendix ) .
in this section we introduce our concept of stationary process on the locally compact abelian group @xmath1 . to do so we first discuss some background on unitary representations .
the concept of stationary process is somewhat more general than needed in the remainder of the paper .
the reason for this is twofold : on the one hand this general treatment does not lead to more complicated proofs but rather makes proofs more transparent . on the other hand for a the study of mixed spectra this general concept may prove to be especially useful .
let @xmath1 be a locally compact abelian group and let @xmath29 be a hilbert space with inner product @xmath66 .
we assume linearity in the first variable and conjugate linearity in the second .
a strongly continuous unitary representation of the group @xmath1 on @xmath29 is a map @xmath67 from @xmath1 into the bounded operators on @xmath29 such that @xmath68 is constant on @xmath1 for any @xmath69 , @xmath70 is the identity on @xmath29 , @xmath71 holds for all @xmath72 and @xmath73 is continuous for any @xmath69 .
then , obviously each @xmath74 is unitary ( i.e. bijective and norm preserving ) .
an @xmath69 is then called an _ eigenfunction _ of @xmath67 with _ eigenvalue _
@xmath75 if @xmath76 for every @xmath77 . the closed subspace of @xmath29 spanned by all eigenfunctions is denoted by @xmath78 .
the representation @xmath67 has _ pure point spectrum _ if @xmath79 , or equivalently if @xmath29 has an orthonormal basis consisting of eigenfunctions . for eigenvalue @xmath75 . for our purposes things
work out more smoothly by formulating it with the conjugate value . ] by stone s theorem , compare ( * ? ? ?
36d ) , there exists a projection - valued measure @xmath80 with @xmath81 for all @xmath82 and @xmath69 , where @xmath83 is the measure on @xmath52 defined by @xmath84 .
then @xmath83 is called the _ spectral measure _ of @xmath85 .
it is the unique measure on @xmath4 satisfying . a mapping @xmath86 of a hilbert space @xmath87 into itself
is said to be a _ conjugation _ of @xmath29 if * @xmath86 is conjugate - linear and @xmath88 ; * for all @xmath89 , @xmath90 .
let @xmath1 be a locally compact abelian group .
a _ stationary process _ or @xmath29-_stationary process _ on @xmath1 is a triple @xmath91 consisting of a hilbert space @xmath29 with conjugation , a strongly continuous unitary representation @xmath67 of @xmath1 on @xmath29 and a linear @xmath1-map @xmath92 such that for all @xmath53 @xmath93 .
a stationary process is called _ ergodic _ if the eigenspace of @xmath67 for the eigenvalue @xmath94 is one - dimensional .
the set of all @xmath29-stationary processes on @xmath1 forms a real vector space in the obvious way by taking linear combinations of mappings .
there is also a canonical notion of isomorphism between stochastic processes : the processes @xmath95 and @xmath96 are called _ isomorphic _ if there exists an invertible unitary map @xmath97 which intertwines @xmath98 and @xmath99 and satisfies @xmath100
. we will be interested in the spectral theory of stationary processes .
an important notion is given next .
a stationary process @xmath101 is said to have _ pure point spectrum _ if the representation @xmath67 of @xmath1 on @xmath29 has pure point spectrum . a stationary process
@xmath102 is called a _ spatial stationary process _ if there is a probability space @xmath6 and an action @xmath103 such that @xmath19 is invariant under the @xmath1-action on @xmath18 through @xmath67 , @xmath104 with the usual inner product @xmath105 , and the action of @xmath67 is extended to an action on @xmath29 by @xmath106 for all @xmath82 , @xmath107 .
the conjugation here is the natural one : complex conjugation of functions on @xmath18 . we write @xmath108 .
the stationary spatial process is called _ full _ if the algebra generated by functions of the form @xmath109 , @xmath110 , @xmath27 , is dense in @xmath111 . in some sense
the assumption of fullness is only a convention .
more precisely , given any spatial stationary process we may create a full process with essentially the same properties by a variant of gelfand construction .
details are given below in theorem [ fullprocess ] .
for the sake of economy , we have used the same notation , @xmath67 , for the action of @xmath1 on @xmath18 and the corresponding action of @xmath1 on @xmath112 .
we often omit the word ` stationary ' in the sequel , but we will always understand it to be in force .
the definition of stationary processes implicitly defines them as _ real _ processes : @xmath26 is real for all real - valued functions @xmath113 on @xmath1 in the sense that @xmath114 . in the case of spatial processes ,
all real - valued functions on @xmath1 are mapped by @xmath23 to real - valued functions on @xmath18 .
one could relax the conditions to allow complex - valued processes , but we shall not do that here .
[ definition - isomorphism ] two stationary spatial processes @xmath115 and @xmath116 are _ spatially isomorphic _ if there is an invertible unitary mapping @xmath117 with * @xmath118 ; * @xmath119 for all @xmath77 ; * @xmath120 for all @xmath121 and @xmath122 for all @xmath123 .
the map @xmath8 is then called spatial isomorphism between the processes .
the isomorphism class of @xmath115 is denoted by @xmath124=[(n , x,\mu , t)]$ ] .
\(a ) the first two points of the definition just say that the processes are isomorphic .
the special requirement for spatial isomorphy is the third point .
as shown in the next proposition , the spatial isomorphism are exactly those isomorphisms which are compatible with the @xmath125 module structure of @xmath126 in the sense that @xmath120 for all @xmath127 and @xmath128 .
\(b ) our definition of spatial isomorphism parallels the notion of spectral isomorphism for measure preserving group actions on probability spaces .
our conditions on @xmath8 imply that spatial isomorphism implies conjugacy in the sense that there is a measure isomorphism between the borel sets of @xmath18 and those of @xmath129 @xcite , theorem 2.4 .
if in addition @xmath130 are complete separable metric spaces then this can be improved to being an isomorphism between @xmath18 and @xmath129 , that is a bijective intertwining mapping between two subsets of full measure in @xmath18 and @xmath129 respectively @xcite , theorem 2.6 .
let @xmath115 and @xmath116 be spatially isomorphic processes with spatial isomorphy @xmath8 .
then , @xmath8 maps @xmath131 into @xmath132 and @xmath120 holds for all @xmath133 and @xmath134 .
the corresponding statements hold for @xmath135 as well .
it suffices to consider the statements on @xmath8 . by assumption
we have that for @xmath136 the product @xmath137 belongs to @xmath126 and in fact equals @xmath138 ( and similarly for @xmath135 ) . by a simple limit argument
, we then see that for any @xmath127 the equation @xmath139 must hold for any @xmath128 .
this shows that the operator of multiplication by @xmath140 is defined on the whole of @xmath126 ( as @xmath8 is onto ) .
furthermore , the graph of every multiplication operator can easily be seen to be closed .
thus , the closed graph theorem gives that the operator of multiplication by @xmath140 is a bounded operator on @xmath126 .
this in turn yields that @xmath140 belongs to @xmath125 .
in this section we show how each stationary process with a second moment comes with a positive definite measure @xmath141 on @xmath1 called the autocorrelation , a positive measure @xmath30 on @xmath4 called the diffraction measure , and the diffraction - to - dynamics map @xmath142 .
we then discuss how the autocorrelation comes about as a limit and characterize existence of a second moment .
let a stationary process @xmath102 be given .
we say that @xmath143 has a _ second moment _ ( or is with second moment ) if there is a measure @xmath144 on @xmath145 ( necessarily unique if it exists ) satisfying @xmath146 for all real valued @xmath49 .
much of this paper is based on the assumption that the second moment measure exists .
since @xmath23 is real , so is @xmath147 , and hence it is real valued.th moment of @xmath143 will be defined by @xmath148 which is different from what we get from the inner product @xmath149 when @xmath150 unless we restrict to real - valued functions .
this is why we restrict to real - valued functions in this definition . ]
furthermore @xmath147 is evidently translation bounded since @xmath23 is @xmath1-invariant .
[ def : gamma ] let @xmath101 be a process
. then @xmath143 has a second moment if and only if there is a measure @xmath151 on @xmath1 satisfying @xmath152 for all @xmath153 . in the case that such a measure exists
, it is unique and furthermore it is real ( i.e. assigns real values to real valued functions ) , positive definite , and translation bounded .
it suffices to restrict attention to real - valued @xmath113 and @xmath154 in @xmath155 as @xmath23 is invariant under a conjugation and hence both @xmath156 and @xmath157 are linear in @xmath113 and antilinear in @xmath154 .
let @xmath158 above .
the assumption of stationarity of @xmath23 shows that @xmath159 is invariant under simultaneous translation of both of its two variables .
define the mapping @xmath160 for @xmath161 we define @xmath162 by @xmath163 .
this change of variables defines a new measure @xmath164 via @xmath165 since @xmath166 , the translation invariance of @xmath159 translates to translation invariance of the second variable of @xmath164 .
thus for measurable sets @xmath167 we have @xmath168 for all @xmath77 . for @xmath64
fixed this is leads to a translational invariant measure on @xmath1 which is hence a multiple @xmath169 of haar measure @xmath170 : @xmath171 the mapping @xmath172 is a measure on @xmath1 , and this measure is the reduction @xmath173 of @xmath159 : @xmath174 now for @xmath175 , @xmath176 so finally @xmath177 and @xmath173 is the desired measure , which we now denote by @xmath141 .
it is positive definite since @xmath178 .
all positive definite measures are translation bounded @xcite . since @xmath158 is a real measure , one sees that @xmath141 is also a real measure .
the denseness of the set of functions @xmath179 in @xmath40 shows that @xmath141 is unique .
we can reverse all the steps of this proof , and from the existence of the real positive definite measure @xmath141 satisfying derive the corresponding second moment measure .
existence of a second moment immediately implies some continuity properties of @xmath23 .
[ continuity - n ] let @xmath180 be a stationary process with second moment .
let @xmath181 be given with @xmath182 .
then , for any compact @xmath60 , there exists a @xmath183 with @xmath184 for all @xmath49 with support contained in @xmath15 .
in particular the map @xmath185 is continuous ( with respect to the @xmath126-norm on @xmath40 ) .
also @xmath186 is continuous with respect to the sup norm on @xmath42 . for @xmath49 with support contained in @xmath15
the support of @xmath187 is contained in @xmath188 and @xmath189 .
as @xmath190 , this easily gives the first statement .
now , the second statement follows by taking @xmath191 . finally , for @xmath192 , @xmath193 .
the continuity property of the preceding corollary allows one to extend the map @xmath23 .
this is discussed next .
[ extendingn ] let @xmath101 be a stationary process with second moment .
the map @xmath23 can be uniquely extended to the vector space of all measurable functions @xmath85 on @xmath1 all of which vanish outside some compact set and are square integrable with respect to the haar measure on @xmath1 .
this extension ( again denoted by @xmath23 ) is @xmath1-equivariant and satisfies @xmath194 meaning in particular that the integral exists and is finite .
it suffices to show that @xmath23 can be uniquely extended to @xmath195 for any open @xmath196 in @xmath1 with compact closure .
let such an @xmath196 be given .
let @xmath197 be given .
as , @xmath196 is open , we can then find a sequence @xmath198 in @xmath199 converging to @xmath113 in the sense of @xmath195 . by the previous corollary ,
we infer then that @xmath200 must be a cauchy - sequence , whose limit does not depend on the choice of the approximating sequence @xmath198 .
this shows that @xmath23 can be extended to @xmath195 . by construction of @xmath26 and
the definition of @xmath141 we have furthermore @xmath201 moreover , a direct application of cauchy - schwartz inequality shows that @xmath202 converges to @xmath203 with respect to the supremum norm and has support contained in @xmath204 .
this easily gives that @xmath205 this finishes the proof .
note that any bounded measurable function with compact support belongs to @xmath206 for any open @xmath196 containing the support .
thus , @xmath23 can in particular be extended to bounded measurable functions with compact support .
the above discussion shows that any stationary process with a second moment gives rise to a real positive definite measure @xmath141 .
as @xmath141 is positive definite it is transformable i.e. its fourier transform @xmath30 exists @xcite , and since @xmath141 is a real measure , @xmath30 is centrally symmetric ( @xmath207 for all measurable sets @xmath64 ) .
the measures @xmath141 and @xmath30 lie at the heart of our investigation .
let @xmath180 be a stationary process with second moment .
then , @xmath208 is called the _ autocorrelation _ of the stationary process .
its fourier transform @xmath209 is called the _ diffraction _ or _ diffraction measure _ of the stationary process . a stationary process @xmath180 with second moment is said to have _ pure point diffraction _
( resp . _ continuous diffraction _ ) if the diffraction measure @xmath30 is a pure point measure ( resp .
continuous measure ) .
the definition of @xmath30 yields that it is a centrally symmetric positive measure and is the unique measure satisfying @xmath210 or equivalently , by linearizing this , @xmath211 for all @xmath153 .
considering the hilbert space @xmath212 with the corresponding inner product being written @xmath213 and using the definition of @xmath141 , we obtain for any @xmath153 , @xmath214 this shows that the map @xmath215 is an isometry .
we define an action @xmath216 of @xmath1 on @xmath217 by defining @xmath218 for all @xmath219 through @xmath220 for all @xmath221 and all @xmath222 .
we note that if @xmath23 is a @xmath29-stationary process on @xmath1 with second moment then for all @xmath223 , @xmath224 is another such process .
the corresponding autocorrelation and diffraction measures are then scaled by @xmath225 .
any process with a second moment comes with an isometric @xmath1-map @xmath142 from @xmath226 to @xmath29 .
this allows one to transfer certain questions from @xmath29 to @xmath226 .
also , it means that for certain eigenvalues there are canonical eigenfunctions available .
these topics are studied next .
[ diffractiontodynamics ] let @xmath102 be a stationary process with second moment .
then @xmath227 is dense in @xmath226 and there exists a unique isometric embedding @xmath228 with @xmath229 for each @xmath53 .
this mapping is a @xmath1-map and furthermore , for all @xmath230 , @xmath231 we begin with the denseness statement .
let @xmath232 be the closure of @xmath227 in @xmath226 .
we wish to show that @xmath233 . as @xmath30 is a translation bounded measure ,
the space @xmath234 is dense in @xmath226 .
thus it suffices to show that @xmath234 belongs to @xmath232 .
as the convolutions of elements from @xmath235 belong to @xmath155 as well , we infer that for all @xmath236 @xmath237 for all @xmath238 .
now @xmath239 belongs to the set @xmath240 of continuous functions on @xmath4 vanishing at infinity .
an application of the stone - weierstrass theorem then shows that @xmath241 for each @xmath242 .
fixing @xmath243 we can then use another application of the stone - weierstrass theorem , to infer that @xmath244 for all @xmath243 and @xmath245 .
this yields @xmath246 as we intended to show .
the existence of @xmath142 now follows by defining it initially by @xmath247 for all @xmath248 , using the fact that it is an isometric mapping by , and then extending it to the @xmath126-closure @xmath232 which is @xmath217 . as for the second statement
, first we note that the @xmath1-action on @xmath212 is designed to make the mapping @xmath249 into a @xmath1-map , while @xmath250 is already a @xmath1-map , and second , for all @xmath27 , @xmath251 , from which @xmath252 we finish using the denseness of the first part .
the map @xmath253 associated to a stationary process @xmath102 with second moment is called the _ diffraction to dynamics map_. the relevance of @xmath142 for spectral theoretic considerations comes from the following consequence of the previous proposition .
[ theta - as - spectralmeasure ] let @xmath102 be a stationary process with second moment and @xmath142 the associated diffraction to dynamics map . then , for any @xmath254 the spectral measure @xmath255 is given by @xmath256 as @xmath142 is a @xmath1-map and an isomorphism a short calculation gives @xmath257 thus , the ( inverse ) fourier transform of @xmath258 is given by @xmath259 . by
the characterization of the spectral measure in the theorem follows . from the previous theorem it is clear that if @xmath29 has pure point spectrum then @xmath30 is discrete and the diffraction is pure point .
however , it is not clear that pure point diffraction implies pure point spectrum but as we shall see , it is so : each implies the other in the case of spatial processes .
the diffraction to dynamics map connects pure point diffraction to eigenvectors of the @xmath1 action on @xmath29 .
later , we shall see that in the case of spatial processes , this allows us to make a complete correspondence between pure point diffraction and pure point dynamics . for a pure point
diffractive stationary processes with diffraction measure @xmath30 we define the set of its _ atoms _ by @xmath260 the set @xmath261 is often called the _ bragg spectrum _ of the process , and its elements are sometimes called _ bragg peaks _ of the atom and its intensity @xmath262 . ] . in the sequel we will often omit the brackets when dealing with one element sets of the form @xmath263 . in particular , we will set @xmath264 for @xmath222 .
let @xmath265 be the characteristic function of the set @xmath266 i.e. @xmath267 is @xmath268 or @xmath94 according as @xmath269 equals @xmath270 or not .
it is easy to see that these functions are the only possible eigenfunctions for our action of @xmath1 on @xmath212 .
define @xmath271 , @xmath272 , by @xmath273 .
then each @xmath271 is an eigenfunction in @xmath29 for the eigenvalue @xmath270 , in the sense that @xmath274 .
[ fminusk ] let @xmath102 be a stationary process with second moment .
then , for all @xmath275 , @xmath276 , and @xmath277 and @xmath278 . since @xmath30 is centrally symmetric , @xmath279 , which gives the first statement .
for the second , using prop .
[ diffractiontodynamics ] we have @xmath280 . then latexmath:[$||f_{-k}||^2
= \langle f_{-k } \mid f_{-k}\rangle = \langle f_{k } \mid f_{k}\rangle = we can use the preceding considerations to compute the map @xmath23 in the case of pure point diffraction .
[ eigenexpansionofn(f ) ] let @xmath180 be a stationary process with pure point diffraction and associated bragg peaks @xmath261 .
let @xmath142 be the associated diffraction to dynamics map and @xmath282 , @xmath272 .
then @xmath283 for all @xmath53 .
by definition of @xmath142 we have @xmath284 .
obviously , @xmath285 and the claim follows .
in this section we show how the autocorrelation can be given a meaning that agrees with ` classical ' two - point correlation associated to stationary point processes .
assume that we are given an ergodic spatial stationary process @xmath115 .
as discussed in corollary [ extendingn ] we can assume that @xmath23 is defined on all measurable bounded functions with compact support .
we also assume that @xmath1 has a countable basis of topology and is hence metrizable .
let @xmath286 be a van hove sequence for @xmath1 ( as discussed in section [ notation ] ) .
we assume that this is fixed once and for all .
[ acdef ] for @xmath287 we define the _ two - point correlation or autocorrelation _ of @xmath11 at any @xmath248 as @xmath288 whenever the limit exists .
this needs some comments : the limits in the definition are taken in the order indicated : first the inner and then the outer .
@xmath13 is an open ( or measurable ) neighbourhood of @xmath94 in @xmath1 .
the statement @xmath289 means that we take a nested descending sequence @xmath290 of such neighbourhoods , all within some fixed compact set @xmath15 , and that @xmath291 .
we are using the notation @xmath292 to stand for the longer @xmath293 .
the definition requires that we give meaning to @xmath294 as a measurable function on @xmath1 .
this uses the extension of @xmath23 from @xmath155 to @xmath126-functions with compact support given in corollary [ extendingn ] . the intuition behind the definition is as follows .
the two - point correlation at @xmath287 for @xmath27 should look something like @xmath295 where the right hand side arises by using the usual trick from van hove sequences and the compactness of the support of @xmath113 .
( that is , the difference of the two sides of the equation is due to the difference between @xmath296 and @xmath297 , which by the van hove assumption is irrelevant in the limit . )
of course in our case @xmath11 is not a function of @xmath298 and the integrands do not make sense .
but the inner integral on the right - hand side is what should be @xmath299 and we can rewrite this ` autocorrelation ' as @xmath300 the term @xmath301 has no meaning .
but palm theory tells us how to go around this .
we instead average over a small neighbourhood @xmath13 of @xmath94 , and this brings us to definition [ acdef ] . [ acthm ] let @xmath1 be a locally compact group whose topology has a countable basis and let @xmath302 be an ergodic spatial stationary process on @xmath1 with second moment .
the two - point correlation of @xmath287 exists @xmath19-almost surely and its value at @xmath27 is @xmath303 .
the function @xmath304 is measurable as a function of @xmath305 and the birkhoff theorem says that almost surely @xmath306 meaning that the limit will exist and equal the right hand side .
we shall prove that @xmath307 proving that the definition of definition [ acdef ] works almost surely in @xmath11 for each @xmath248 and that @xmath141 does have the meaning of an autocorrelation .
this has to be made to work simultaneously for all @xmath113 in @xmath40 , which will be shown in the usual way from a countable basis of @xmath40 .
this will prove theorem [ acthm ] .
here are the details : it is enough to prove for real - valued @xmath113 .
using we have for all @xmath308 , @xmath309 let @xmath310 and choose @xmath311 .
let @xmath13 be a measurable subset of @xmath1 with compact closure . then by prop .
[ extendingn ] , @xmath312 is defined and is a measurable @xmath126-function on @xmath18 . by linearization
we have @xmath313 since @xmath113 is uniformly continuous on @xmath1 , for any sufficiently small neighbourhood @xmath13 of @xmath94 , @xmath314 for all @xmath315 and for all @xmath316 .
then @xmath317 for all @xmath316 , so @xmath318 ( @xmath141 is positve - definite and real , prop .
[ def : gamma ] ) , we see that as @xmath319 ( and correspondingly @xmath320 ) we have a proof of .
together with we have the desired interpretation of def . [ acdef ] as the autocorrelation at @xmath11 and its almost sure equality with @xmath303 , and prop .
[ acthm ] is proved .
this type of result was first established for certain uniformly discrete point processes on euclidean space by hof @xcite ( see @xcite for the case of general locally compact abelian groups ) .
this was then extended to rather general point processes by gour @xcite and to certain measure processes by baake / lenz @xcite .
a unified treatment was then given by lenz / strungaru @xcite .
our result contains all these results ( provided the underlying process is real ) .
the discussion above has shown that existence of second moments has strong consequences .
it implies existence of the diffraction to dynamics map by proposition [ diffractiontodynamics ] and further continuity properties by corollary [ continuity - n ] .
it turns out that a converse of sorts holds .
this is investigated in this section .
along the way , we will also show that continuity of @xmath23 implies an intertwining property of the map @xmath23 and this will be crucial for our considerations .
any process @xmath321 gives rise to two representations of @xmath155 ( and in fact even of @xmath322 ) : the representation @xmath232 lives on @xmath323 and acts by @xmath324 for @xmath49 .
the representation @xmath67 ( extending the action @xmath67 and therefore denoted by the same letter ) acts by @xmath325 for @xmath326 and @xmath69 .
continuity of @xmath23 now yields an intertwining property : [ intertwiner ] let @xmath321 be a process with a continuous @xmath23 .
then , @xmath327 for all @xmath326 .
we have @xmath328 by the linearity and continuity of @xmath23 , it commutes with taking integrals and @xmath329 holds , which finishes the proof .
we will need a somewhat stronger continuity property of @xmath23 .
let @xmath330 with @xmath182 be given .
the process @xmath321 is said to be weakly @xmath331-continuous if for any compact @xmath60 there exists a @xmath332 with @xmath333 for all @xmath49 with support contained in @xmath15 ( see cor .
[ continuity - n ] ) .
\(a ) note that continuity of @xmath23 is exactly weak @xmath334-continuity .
\(b ) by standard interpolation theory , we can conclude that weak @xmath334 continuity together with weak @xmath335 continuity implies weak @xmath331 continuity for all @xmath336 with @xmath337 and @xmath182 .
let @xmath321 be a stochastic process .
then , the following assertions are equivalent : * @xmath23 has a second moment .
* @xmath23 is weakly @xmath331-continuous for all @xmath330 with @xmath182 . in this case
@xmath23 is continuous and has the intertwining property that @xmath338 for all @xmath49 .
( i)@xmath339 ( ii ) : this follows immediately from corollary [ continuity - n ] .
( ii)@xmath339 ( i ) : as @xmath23 is weakly @xmath334 continuous , it is continuous and the intertwining property follows from the previous lemma . in the remaining part of the proof we will only consider real valued functions . by proposition [ def :
gamma ] , it suffices to show existence of a measure @xmath141 on @xmath1 with @xmath340 for all @xmath49 .
a short calculation shows that for such a @xmath141 the equality @xmath341 must hold for all @xmath49 .
the idea is now to ` reverse ' this reasoning to conclude existence of @xmath141 . the main issue is to show continuity of the object @xmath342 replacing the not yet defined @xmath343 .
this is shown using the intertwining property . here
are the details : we choose an arbitrary compact @xmath15 in @xmath154 and assume without loss of generality that @xmath344 .
let @xmath196 be an open neighborhood of @xmath94 with compact closure and set @xmath345 to be the closure of @xmath346 .
set @xmath347 . as @xmath23 is weak
@xmath335 continuous , and @xmath348 is the dual of @xmath349 , we can find to each @xmath326 with support in @xmath350 a function @xmath342 in @xmath348 with @xmath351 for all @xmath53 with support in @xmath232 and @xmath352 where @xmath353 .
as @xmath23 has the intertwining property , a short argument shows that @xmath354 for @xmath355 for all @xmath356 with support contained in @xmath15 and all @xmath357 with support contained in @xmath196 .
this gives in particular , that @xmath358 is a continuous function ( on @xmath15 ) for all such @xmath359 .
now yields @xmath360 taking an approximate unit for @xmath361 , we wee that @xmath342 can be approximated uniformly by continuous functions .
this shows continuity of @xmath342 .
moreover , by @xmath362 again the map @xmath363 is continuous for each @xmath298 .
thus , we can indeed define the measure @xmath141 with @xmath364 by construction @xmath141 has the desired properties .
almost periodicity is closely linked to nature of diffraction via the fourier transform .
this concept allows one to compare properties of the autocorrelation and the diffraction measure . in particular
, it can be used to characterize pure point and continuous diffraction .
the considerations of this section play a role in subsequent parts of the paper : they are used in section [ pc : split ] to decompose an arbitrary processes into a part with pure point diffraction and a part with continuous diffraction , and in section [ purepoint ] to prove the equivalence of pure point diffraction spectrum and pure point spectrum .
the material of this section derives from @xcite and , in a more accessible account @xcite , and from @xcite as well . for further studies of aspects of almost periodicity in our context we refer to @xcite .
almost periodicity is most commonly defined by a compactness condition : @xmath365 is almost periodic if the translation orbit @xmath366 of @xmath85 is compact . here
@xmath367 is the space of uniformly continuous @xmath368-valued functions on @xmath1 .
the key point , however , is in which topology is the translation orbit to be compact ? in the case of the sup - norm topology the concept defines _ strong almost periodicity_. in the case that the topology is defined by the family of semi - norms induced by the set of all continuous linear functionals on @xmath367 , it is called _ weak almost periodicity_. strong almost periodicity coincides with h. bohr s original concept of almost periodicity : @xmath369 is strongly periodic if and only if for every @xmath311 the set of @xmath77 for which @xmath370 is relatively dense . it is harder to get an intuitive feel for weak almost periodicity .
fortunately what one needs to know about it is fairly straightforward : * all positive definite functions on @xmath1 are weakly almost periodic ; * for every weakly almost periodic function @xmath85 the _ mean _ of @xmath85 defined as @xmath371 exists for any van hove ( more generally flner ) sequence @xmath59 in @xmath1 and any @xmath305 and its value is independent of both @xmath298 and the choice of van hove sequence .
one says that a weakly almost periodic function @xmath85 is _ null weakly almost periodic _ if its mean is @xmath94 .
these concepts lift to measures in a simple way : a borel measure @xmath372 on @xmath1 is called _ strongly _ ( resp . _ weakly _ , _ null weakly _ ) _ almost periodic _ if @xmath373 is a strongly ( resp .
weakly , null weakly ) almost periodic function for every @xmath374 .
positive definite measures on @xmath1 turn out to be weakly almost periodic and are also , as we have noted already , fourier transformable . [ thm - ga ] ( gil de lamadrid , argabright @xcite ) every weakly almost periodic measure @xmath372 is uniquely expressible as the sum of a strongly almost periodic measure and null weakly almost periodic measure .
if the measure @xmath372 is fourier transformable then so too are the strong and null weak components , and furthermore the fourier transforms of these components are the pure point and continuous parts of @xmath375 .
these considerations can be applied to strongly continuous unitary representations and hence to processes as well .
this is discussed next .
the crucial connection is given by the following ( well known ) lemma .
[ fft - is - wap ] let @xmath67 be a strongly continuous unitary representation of @xmath1 on @xmath29 then for any @xmath376 , the function @xmath377 is positive definite and hence weakly almost periodic and fourier transformable .
its fourier transform is the spectral measure @xmath83 .
as @xmath67 is unitary , we have for any @xmath378 and @xmath379 @xmath380 and @xmath381 is shown to be positive definite .
the remaining statements follow from the above discussion . combining the previous lemma and the previous theorem , we infer the following corollary .
let @xmath67 be a strongly continuous unitary representation of @xmath1 on @xmath29 then for any @xmath376 , the function @xmath382 is the sum of a strongly almost periodic function and a null weakly almost periodic function .
this strongly almost periodic function is given by the ( inverse ) fourier transform of the pure point part of the spectral measure @xmath83 and the null weakly almost periodic function is given by the ( inverse ) fourier transform of the continuous part of the spectral measure @xmath83 .
of particular relevance is the question whether the spectral measures are pure point . here , the following holds as shown in @xcite , lemma 2.1 .
[ characterizationpp ]
let @xmath67 be a strongly continuous unitary representation of @xmath1 on @xmath29 .
then , the following assertions are equivalent for @xmath69 : * the map @xmath383 , @xmath73 , is almost periodic in the sense that for any @xmath384 the set @xmath385 is relatively dense in @xmath1 . *
the hull @xmath386 is relatively compact . *
the map @xmath387 , @xmath388 , is strongly almost periodic .
* @xmath389 is a pure point measure .
* @xmath85 belongs to @xmath78 .
the preceding equivalence allows one to show some stability properties of @xmath78 .
this is discussed next ( see @xcite as well for a proof of ( a ) of the following theorem ) .
there , we say that a _ measure @xmath19 is supported on a measurable set @xmath390 _ if @xmath19 of the complement of @xmath390 is zero .
[ stabilitypp ] let @xmath67 be a strongly continuous unitary representation of @xmath1 on @xmath29 .
\(a ) let @xmath391 be continuous with @xmath392 for each @xmath393 and @xmath69 .
then , @xmath394 maps @xmath78 into @xmath78 . if @xmath85 belongs to @xmath78 and @xmath389
is supported on the subgroup @xmath390 of @xmath4 , then so is @xmath395 .
\(b ) let @xmath396 be continuous with @xmath397 .
then , @xmath398 maps @xmath399 into @xmath78 .
the statements on @xmath394 and @xmath8 are clear from the equivalence of ( v ) and ( ii ) in the lemma [ characterizationpp ] and the continuity and intertwining property of @xmath394 and @xmath8 .
the statement on the support in ( a ) is proven in @xcite .
we now further restrict attention to processes with second moment .
then , theorem [ thm - ga ] can directly be applied to give the following result ( which is well known from @xcite ) . a stationary process @xmath180 with second moment is pure point diffractive if and only if its autocorrelation is strongly almost periodic .
it has continuous diffraction if and only if the autocorrelation is null weakly almost periodic .
we can then combine the above results to obtain the following . for any stationary process @xmath180 with second moment
the mapping @xmath400 is a weakly almost periodic function on @xmath1 for all @xmath248 .
the process @xmath180 then has pure point ( resp .
continuous ) diffraction if and only if @xmath400 is a strongly ( resp .
null weakly ) almost periodic function on @xmath1 for all @xmath248 .
furthermore , the diffraction is pure point if and only if for all @xmath27 , @xmath401 is a strongly almost periodic function on @xmath1 with respect to the norm on @xmath29 . the first statement follows from lemma [ fft - is - wap ] .
we now turn to the second statement . by theorem [ thm - ga ] and the definition of the diffraction spectrum we have pure point ( resp .
continuous ) diffraction if and only if @xmath141 is strongly ( resp null weakly ) almost periodic .
now , from proposition [ def : gamma ] and a straightforward calculation we obtain for all @xmath402 and all @xmath221 , @xmath403 with @xmath154 set equal to @xmath113 we then get the required almost periodicity properties of @xmath404 from the corresponding almost periodicity properties of @xmath141 , by our definition of these concepts . in the reverse direction , upon linearizing the expressions @xmath405 we obtain . from the almost periodicity properties of @xmath406
we then obtain the desired almost periodicity properties for all convolutions of the form @xmath407 .
an approximate unit argument allows us to get the strong ( resp .
null weak ) almost periodicity of @xmath141 from this , see @xcite , corollary 9 . here , we use that these almost periodicity properties are stable under uniform convergence .
the last statement is a consequence of lemma [ characterizationpp ] .
this finishes the proof .
we finish this section with a discussion of stability properties of almost periodicity in the case of spatial processes .
we will need the following proposition on continuity of composition with a fixed function .
the proposition is not hard to prove and can be found in @xcite .
[ continuity - of - phi ] let @xmath408 be a sequence in @xmath409 converging to @xmath410 .
let @xmath372 be a continuous bounded function from the complex numbers to the complex numbers .
then , @xmath411 converges to @xmath412 .
now , we turn to the following consequence of theorem [ stabilitypp ] ( see @xcite as well ) .
[ coro - stabilitypp ] let @xmath180 be a spatial process with @xmath413 .
let @xmath414 be bounded continuous functions and @xmath415 be given .
then , the following holds : \(a ) the function @xmath416 belongs to @xmath78 and the spectral measure @xmath417 is supported in the subgroup @xmath390 of @xmath52 if @xmath83 is supported in this subgroup .
\(b ) the function
@xmath418 belongs to @xmath78 . if both @xmath83 and @xmath419 are supported in the subgroup @xmath390 of @xmath52 then
so is @xmath420 . by the previous proposition
the maps @xmath421 , @xmath422 and @xmath423 are continuous .
they obviously commute with the action of @xmath1 . both ( a ) and the first statement of ( b ) follow from theorem [ stabilitypp ] .
it remains to show the second statement of ( b ) : let @xmath424 be the subspace of @xmath78 consisting of elements whose spectral measure is supported on @xmath390 .
then , it is not hard to see that @xmath424 is a closed subspace of @xmath29 . by ( a ) and
the assumption on @xmath85 and @xmath425 , the spectral measures of both @xmath426 and @xmath427 belong to @xmath424 .
moreover , @xmath426 and @xmath427 are bounded functions .
thus , it suffices to show that @xmath428 is an algebra .
this can be shown by mimicking the proof of lemma 1 in @xcite ( see @xcite as well ) .
for the convenience of the reader we sketch a proof : note first that every eigenfunction can be approximated by bounded eigenfunctions via a simple cut - off procedure viz if @xmath429 is an eigenfunction , then for an arbitrary @xmath430 , the function @xmath431 is again an eigenfunction ( to the same eigenvalue ) .
evidently , the @xmath432 converge to @xmath429 in @xmath126 as @xmath433 .
now , let non - zero functions @xmath434 be given and choose @xmath435 arbitrarily .
as @xmath436 belongs to @xmath437 , there exists a finite linear combination @xmath438 of eigenfunctions to eigenvalues in @xmath390 with @xmath439 by the conclusion after eq .
( [ cutoff ] ) , we can assume that all @xmath440 are bounded functions .
thus , in particular , @xmath441 . similarly , we can choose another finite linear combination @xmath442 of bounded eigenfunctions to eigenvalues in @xmath390 with @xmath443 then , @xmath444 the product of bounded eigenfunctions to eigenvalues in @xmath390 is again a bounded eigenfunction to an eigenvalue in @xmath390 ( as @xmath390 is a group ) and the claim follows .
in this section we first discuss how any stationary process splits into two subprocesses , one of which has pure point spectrum and one of which has continuous spectrum .
if the original process has a second moment then so have the two subprocesses and their diffraction measure will be pure point and purely continuous respectively ( theorem [ split ] ) . if furthermore the original process is spatial , then its pure point part is canonically isomorphic to a spatial pure point process ( theorem [ prop - decomposition ] ) .
this last result is based on the gelfand construction of commutative banach algebras .
whenever @xmath67 is a strongly continuous unitary representation of @xmath1 on the hilbert space @xmath29 and @xmath196 is a @xmath67 invariant closed subspace of @xmath29 with corresponding orthogonal projection @xmath445 we can form the restriction @xmath446 of @xmath67 to @xmath196 and this will be a strongly continuous unitary representation as well .
the orthogonal complement @xmath447 of @xmath196 in @xmath29 is also @xmath67 invariant and we can form the restriction @xmath448 as well .
we will now focus on the special decomposition into the continuous and its point part .
more precisely , we can define the continuous hilbert space @xmath449 and the point hilbert space @xmath78 by @xmath450 then , both @xmath78 and @xmath449 are closed @xmath67-invariant with @xmath451 with the projections @xmath452 and @xmath453 onto @xmath78 and @xmath449 we then have @xmath454 and @xmath455 for @xmath456 .
this gives the decomposition of the representation @xmath67 as @xmath457 with @xmath458 , @xmath459 .
obviously , all spectral measures associated to @xmath460 are purely discrete and all spectral measures associated to @xmath461 are purely continuous .
accordingly , we call @xmath462 the _ pure point part of the representation @xmath67 _ and @xmath461 the _ continuous part of the representation @xmath67_. we now turn to the case that we are given a stationary process @xmath180 . of course , @xmath67 is then a strongly continuous unitary representation of @xmath1 on @xmath29 .
so , any @xmath67 invariant subspace @xmath196 gives rise to a decomposition of @xmath67 .
moreover , the @xmath1-invariance of @xmath23 then implies that @xmath23 can as well be decomposed into @xmath463 and @xmath464 .
accordingly , the process @xmath143 can be decomposed in to the two processes @xmath465 and @xmath466 and we write @xmath467 .
as just discussed , @xmath67 can be decomposed into its pure point part and its continuous part . into @xmath468 with @xmath469 , @xmath456 . hence
, any process @xmath470 can be decomposed into the two processes @xmath471 and @xmath472 which are called the _ pure point part _ and the _ continuous part _ of @xmath143 respectively .
now , assume that the original process has a second moment and associated autocorrelation measure @xmath141 and diffraction measure @xmath473 . by theorem [ thm - ga
] , this @xmath141 can be decomposed uniquely into its strongly almost periodic part @xmath474 and its null weakly almost periodic part @xmath475 and the fourier transform of @xmath474 is the pure point part @xmath476 of @xmath477 and the fourier transform of @xmath478 is the continuous part @xmath479 of @xmath477 .
it turns out that these are just the autocorrelation and diffraction measures of the pure point part and the continuous part of @xmath143 .
this is the content of the next theorem .
[ split ] let @xmath480 be a stationary process with a second moment and associated autocorrelation @xmath141 , diffraction measure @xmath477 and diffraction to dynamics map @xmath142 .
let @xmath481 and @xmath478 be the strongly almost periodic part and the null weakly almost periodic part of @xmath141 and @xmath476 and @xmath479 the respective fourier transforms .
then , the point part @xmath482 of @xmath143 has a second moment with autocorrelation given by @xmath474 and diffraction measure given by @xmath476 and the continuous part @xmath483 of @xmath143 with autocorrelation given by @xmath478 and diffraction measure given by @xmath479 . by proposition [ def : gamma ]
it suffices to show that @xmath484 for all @xmath49 . by linearity it suffices to consider the case @xmath485
let @xmath486 be the function @xmath487 and consider the function @xmath488 .
we will decompose this function in a strongly almost periodic part and a null weakly almost periodic part in two ways and then conclude the desired statement from the uniqueness of the decomposition . here
are the details : by definition of @xmath474 and @xmath478 we can decompose this function as @xmath489 with a strongly almost periodic function @xmath490 and a null weakly almost periodic function @xmath491 . on the other hand by definition of @xmath141 and the fact that @xmath452 and @xmath492
are orthogonal we also obtain the decomposition @xmath493 now , @xmath494 is strongly almost periodic as it is the inverse fourier transform of a pure point measure and @xmath495 is null weakly almost periodic as it is the inverse fourier transform of a continuous measure . from theorem [ thm - ga ]
we conclude that @xmath496 for all @xmath77 .
setting @xmath497 we obtain .
in the situation of the previous theorem we can easily see that the diffraction to dynamics map @xmath498 can be decomposed as @xmath499 here , @xmath500 and @xmath501 are the diffraction to dynamics maps of @xmath482 an @xmath483 respectively and and the natural decompositions @xmath502 and @xmath503 are implicitly used . we now restrict attention to spatial processes .
in this situation we can induce subprocesses via subalgebras of @xmath125 .
this is discussed next .
we will use the main result of gelfand theory that any commutative @xmath504-algebra @xmath505 with a unit is canonically isomorphic to the algebra @xmath506 of continuous functions on the compact space @xmath507 given by the maximal ideals of @xmath505 .
we will denote the space of all continuous bounded complex valued functions on the complex plane by @xmath508 .
[ prop - inducing ] let @xmath509 be a spatial process with hilbert space @xmath510 .
let @xmath505 be a @xmath1-invariant subalgebra of @xmath511 which is closed under complex conjugation , contains the constant functions and is closed with respect to the sup norm .
let @xmath512 be the subspace of @xmath29 generated by @xmath505 and let @xmath507 the maximal ideal space of @xmath505 .
then , @xmath507 can be equipped in a unique way with a @xmath1 action and a @xmath1-invariant measure @xmath513 such that the canonical gelfand isomorphism @xmath514 extends to an unitary @xmath1-map ( also called @xmath515 ) @xmath516 the map @xmath515 is compatible with the algebraic structure in that it satisfies * @xmath517 for all @xmath518 and @xmath519 and * @xmath520 for all @xmath521 and for all continuous bounded @xmath522 .
the subspace @xmath196 is a @xmath1-invariant subspace and @xmath143 can be decomposed as @xmath523 where @xmath524 is isomorphic via @xmath515 to the spatial process @xmath525 with @xmath526 , @xmath527 . if @xmath143 is ergodic , so is @xmath524 .
all statements are rather straightforward up to the compatibility of @xmath515 with the algebraic structure and the ergodicity .
this compatibility with algebraic structure will be discussed next : by definition , @xmath515 is an algebra isomorphism from @xmath506 and @xmath505 . in particular , @xmath528 for @xmath529 . by a simple approximation argument
, we then obtain @xmath517 for all @xmath530 and @xmath531 . here
, we use that @xmath532 converges to @xmath533 in @xmath534 whenever @xmath535 converges to @xmath425 in @xmath536 and @xmath85 is bounded .
this shows that @xmath515 is compatible with multiplication .
as @xmath515 is an isomorphism of algebras , we also have @xmath537 for any polynomial @xmath538 and any @xmath539 . by a stone / weierstrass type argument
, this gives @xmath540 for any @xmath541 and @xmath542 .
consider now the set @xmath543 then , @xmath544 by what we have just shown .
moreover , @xmath545 is closed in @xmath126 by proposition [ continuity - of - phi ] .
thus , @xmath545 must agree with @xmath546 as @xmath506 is dense in @xmath547 ( as @xmath548 converges to @xmath549 whenever @xmath408 converges to @xmath550 ) .
it remains to show the statement on ergodicity . by definition ergodicity
means that the eigenspace of @xmath67 to the eigenvalues @xmath268 is one - dimensional .
this is obviously stable under the above constructions .
the process @xmath525 constructed in the previous proposition is called the subprocess of @xmath143 induced by @xmath505 .
[ fullprocess ] let @xmath509 be a spatial process with hilbert space @xmath510 .
let @xmath505 be the closure with respect to the supremum norm of the subalgebra of @xmath511 generated by @xmath551 with @xmath552 , @xmath553 and @xmath554 continuous bounded functions from @xmath368 to @xmath368 .
let @xmath512 be the subspace of @xmath29 generated by @xmath505 .
then @xmath196 contains the range of @xmath23 and the subprocess @xmath524 is isomorphic to the full spatial process @xmath525 .
the process @xmath525 is ergodic if @xmath143 is ergodic .
moreover , @xmath555 holds for all @xmath49 .
in particular , @xmath143 has a second moment if and only if @xmath525 has a second moment and their autocorrelations agree in this case . by definition of @xmath505 , the range of @xmath23 is obviously contained in the subspace @xmath196 .
thus , @xmath556 and therefore @xmath557 holds for all @xmath49 . given this all statements of the theorem
follow immediately from the previous proposition .
the theorem says that to any ( ergodic ) stationary process @xmath143 we can find an ( ergodic ) stationary subprocess @xmath524 which is full and has the same second moment .
in the situation of the theorem fullness of the original process @xmath143 then just means that @xmath558 .
now let @xmath180 be a spatial process with @xmath510 .
let @xmath559 be the closure of the linear span of products of the form @xmath560 then , @xmath561 is a closed invariant subspace of @xmath111 .
let @xmath562 be the orthogonal projection onto @xmath561 . then , @xmath562 commutes with @xmath67 .
then , @xmath563 with @xmath564 and @xmath565 is a stationary process .
we call @xmath566 the _ augmented point part _ of the stationary process @xmath180 .
let @xmath567 be the orthogonal complement of @xmath561 in @xmath29 .
then , @xmath567 gives rise to a stationary process @xmath568 and @xmath23 is the sum of these processes .
[ prop - decomposition ] let @xmath509 be a spatial process with hilbert space @xmath510 and associated augmented point part @xmath569 . then the following holds : * @xmath570 ; * @xmath571 is isomorphic to a full spatial process ; * @xmath572 and @xmath573 hold for all @xmath49 .
if @xmath143 has a second moment with autocorrelation @xmath141 and diffraction @xmath477 furthermore the following is true : * @xmath571 has a second moment with autocorrelation @xmath474 and diffraction measure given by the pure point part of @xmath477 .
* @xmath574 has a second moment with autocorrelation @xmath478 and the diffraction measure given by the continuous part of @xmath477 .
let @xmath575 .
let the algebra @xmath505 be the closure with respect to the sup - norm of the linear span of products of the form @xmath560 then , @xmath561 is just the subspace generated by @xmath505 .
we can then apply the previous theorem to @xmath576 and the algebra @xmath505 .
this gives a decomposition @xmath577 , where @xmath578 is isomorphic to a full spatial process and @xmath579 holds for all @xmath49 .
here , the last equality follows by definition of @xmath580 as @xmath581 .
it remains to show that @xmath582 ( then all remaining statements follow easily ) .
to do so , we first note that the range of @xmath583 is contained in the range of @xmath196 by construction .
moreover , by general principles the algebra @xmath505 and hence the subspace @xmath561 is contained in the pure point part of @xmath67 @xcite .
thus , @xmath584 holds .
similar remarks apply to @xmath585 .
note that @xmath586 .
the considerations of this section and in particular the preceding theorem have the following consequence : whenever we are given a spatial process we can restrict attention to its point part and obtain a full spatial process with pure point spectrum .
in particular , the whole theory developed below for full spatial processes with pure point spectrm will apply to the point part of any spatial process .
in this section we discuss spatial processes with pure point diffraction .
it is these processes that we will be able to classify .
we show that for these processes the two notions of pure pointedness of the spectrum defined above viz pure point diffraction spectrum and pure point dynamical spectrum actually agree .
as noted in theorem [ prop - decomposition ] , any spatial process with second moment can be decomposed into a full spatial process with pure point diffraction and a general stationary process with continuous diffrction .
thus , the results of this section apply to the corresponding parts of any spatial process .
let @xmath302 be a stationary process with pure point diffraction as discussed in section [ stochasticprocesses ] .
thus , its diffraction measure @xmath30 is a pure point measure and the set of its atoms denoted by @xmath587 is given by @xmath588 let @xmath589 be the associated diffraction to dynamics map and @xmath590 , @xmath272 . while the subspace @xmath591 of @xmath409 may be small in some sense , in another sense it controls the whole hilbert space @xmath409 due to our assumption of pure point spectrum .
a precise version is given next .
[ ergodic+pure point ] let @xmath115 be a spatial stationary process , which is full and possesses a second moment .
then , the following assertions are equivalent : * the diffraction measure @xmath30 of @xmath302 is a pure point measure . *
the representation @xmath67 has pure point spectrum . in this case
, the group @xmath592 of eigenvalues of @xmath67 is generated by the set @xmath593 , and any eigenfunction is a ( multiple of a ) product of eigenfunctions of the form @xmath271 , @xmath594 .
we mimick the argument of @xcite ( see @xcite as well ) . by definition of the spectral measures @xmath389 given in we have @xmath595 for all @xmath53 . moreover , by definition of @xmath30 and @xmath141 and the fact that @xmath23 is a @xmath1-mapping we find @xmath596 first for real - valued @xmath53 and then by linearity for all @xmath53 . taking inverse fourier transforms we infer @xmath597 for all @xmath53 .
the implication @xmath598 is now clear from .
( we also saw this earlier from the properties of @xmath142 . )
we next show @xmath599 : as the diffraction measure is a pure point measure , the spectral measures associated to @xmath26 are pure point measures ( supported on @xmath261 ) by .
thus , by part ( a ) of corollary [ coro - stabilitypp ] , the spectral measures of @xmath600 for @xmath601 , @xmath53 , are then also pure point measures supported on the group generated by @xmath261 . by the fullness assumption and part ( b ) of corollary [ coro - stabilitypp ]
we then infer that @xmath67 has pure point spectrum with eigenvalues contained in the group generated by @xmath261 .
now , multiplying eigenfunctions @xmath271 , @xmath272 , we obtain eigenfunctions for arbitrary elements in the group generated by @xmath261 .
this shows that the group generated by @xmath261 is indeed the group of eigenvalues . at the same time it gives an eigenfunction for each eigenvalue . by ergodicity
the multiplicity of each eigenvalue is one and these eigenfunctions are all eigenfunctions .
\(a ) there is quite some history to theorem [ ergodic+pure point ] : the implication @xmath598 is sometimes known as dworkin argument .
it was first established by dworkin @xcite with later extensions by hof @xcite and schlottmann @xcite ( see @xcite for remarkable applications as well ) . in special situations equivalence
was then shown by lee / moody / solomyak in @xcite .
this was then generalized in various directions by gour @xcite , baake / lenz @xcite , deng / moody @xcite , and lenz / strungaru @xcite ( the latter result containing all earlier ones ) .
the above result contains all these results ( provided the underlying process is real ) .
\(b ) the equivalence between diffraction and dynamical spectrum can not hold for the complete spectrum , as discussed by van enter / mieskicz @xcite .
indeed , @xcite gives an example of mixed spectrum where the pure point part of the dynamical spectrum is completely missing in the diffraction ( up to the constant eigenfunction ) .
due to the previous theorem we do not need to distinguish between pure point diffraction and pure point dynamical spectrum when dealing with full processes with second moments .
this suggests the following definition .
[ def - pure - point - process ] a spatial stationary process @xmath602 is called _ pure point _ if it is full , possesses a second moment and its diffraction measure is a pure point measure . if @xmath603 in the situation of prop . [ eigenexpansionofn(f ) ] then by lemma [ fminusk ] @xmath604 is a real function belonging to the @xmath605-eigenspace of @xmath606 and @xmath607 . since constant function @xmath608 for all @xmath287 is an eigenfunction for @xmath605 and the dynamical system is ergodic , it spans the @xmath605-eigenspace of @xmath606 .
thus @xmath609 .
now , we have already pointed out that if @xmath23 is a stationary process then so is any non - zero real multiple of it .
in particular @xmath610 are stationary processes and one of them has the corresponding function @xmath611 .
it causes unnecessary awkwardness later on in the discussion of phase forms to deal with the case @xmath612 , so wish to normalize @xmath23 to avoid this situation : [ 0-eigenfunction assumption ]
as mentioned already , our aim is to describe all stationary processes with a given pure point diffraction measure @xmath477 .
this description will be given in terms of a set of objects called _ relators _ arising out of the bragg spectrum of @xmath477 .
while the later sections need the material of this section , it can be read independently of the previous ones .
let @xmath4 be a locally compact abelian group .
let @xmath613 be given with @xmath614 and let @xmath592 be the subgroup of @xmath4 generated by @xmath261 .
we shall give this group the discrete topology , and in order to keep this clear we shall usually write @xmath615 instead of @xmath592 .
in the sequel we often use @xmath616-tuples @xmath617 ( for various @xmath616 ) of elements of @xmath261 or of @xmath615 and we often simply denote these by bold letters @xmath618 .
if @xmath619 then @xmath620 is the concatenation @xmath621 and @xmath622 : = k_1 + \dots + k_m \in { \mathcal{e}}_d.\ ] ] in this way @xmath623 becomes a monoid under concatenation , with the empty @xmath94-tuple @xmath624 as the identity element .
a _ relator _ is any @xmath616-tuple @xmath625 with @xmath622= k_1 + \dots + k_m = 0.\ ] ] the empty tuple @xmath626 is also taken to be a relator .
thus the relators are a subset of @xmath627 .
let @xmath628 = 0 \ } \ , , n\ge 1 \,;\\ z_0 & : = & \{\emptyset\}\ , ; \\ z & : = & \bigcup_{n=0}^\infty z_n \ , .\end{aligned}\ ] ] note that @xmath629 if @xmath630 , otherwise it is empty .
concatenations of relators are relators and this makes @xmath32 a submonoid of @xmath627 .
we introduce an equivalence relation on @xmath627 by transitive extension of the three rules : * @xmath631 if @xmath632 is a permutation of the symbols of @xmath618 ; * if @xmath633 and @xmath630 then @xmath634 ( along with the first rule , this means that when @xmath630 , zeros can be dropped or added wherever they appear ) ; * @xmath631 if @xmath632 can be obtained from @xmath618 by inserting removing pairs @xmath635 , @xmath272 .
the second item here is related to our assumption [ 0-eigenfunction assumption ] , see below .
it is easy to see that @xmath636 , so multiplication descends from @xmath627 to @xmath637 , whereupon @xmath638 becomes an abelian group under this multiplication with @xmath639 as the identity element .
note that if @xmath603 then @xmath640 .
we give @xmath638 the discrete topology , so in particular it is a locally compact abelian group . since the sum @xmath641 = k_1 +
\dots + k_m$ ] of an element of @xmath642 is constant on the entire equivalence class @xmath643 , we see that @xmath32 is the union of all equivalence classes with component sum equal to @xmath94 , and we obtain @xmath644 as a subgroup of @xmath638 . in addition , we see that there is a surjective homomorphism @xmath645\ ] ] and its kernel is precisely @xmath33 .
the group @xmath646 is called the _ relator group_. the elements of its dual group @xmath647 are called _ phase forms_. thus a phase form is a group homomorphism @xmath648 ( the latter being the unit circle in @xmath368 ) .
\(a ) it is useful to think of relators in terms of the cycle structure of the cayley graph of the group of eigenvalues with respect to the generators @xmath261 .
in this way the group @xmath33 could then be seen as a a kind of abelianized homotopy group or a first cohomology group of the graph .
this works in the following way : we begin with @xmath592 and the set @xmath261 which satisfies @xmath614 and generates @xmath592 .
the cayley graph has vertices @xmath592 and edges consisting of all pairs @xmath649 where @xmath650 and @xmath275 . since for each edge
@xmath649 we have the reverse edge @xmath651 , we can think of the graph as being non - directed .
since @xmath261 generates @xmath592 , the graph is connected .
each @xmath652 can be thought of as an edge sequence . given any @xmath650
we obtain a path @xmath653 from it .
this path is a cycle if and only if @xmath618 is a relator , that is , @xmath641=0 $ ] .
there is a sort of homotopy of edge sequences , which is generated by rules corresponding to ( r1 ) , ( r2 ) , ( r3 ) above .
the first amounts to saying that for all @xmath650 and for all @xmath654 , the paths @xmath655 and @xmath656 are homotopic ; the second says that if @xmath630 then the path @xmath657 is homotopic with the empty path from @xmath298 ; and the third says that the path @xmath658 is homotopic to the empty path from @xmath298 .
the group @xmath33 may be thought of as a sort of homotopy group for cycles originating ( and terminating ) at @xmath94 .
\(b ) it should be noted that the homotopy group @xmath33 may be infinite even when the group @xmath1 , and hence the associated graph , is finite .
we shall see this in the examples below .
the phase forms play a central role in homometry problem .
we note that they depend only on @xmath261 , not on the actual values of the diffraction measure @xmath30
. we shall find it convenient to sometimes abuse the notation and write things like @xmath659 where we mean @xmath660 when using the phase form @xmath34 . of the various @xmath661 that can represent a given element in @xmath33
there is ( at least ) one of minimal length @xmath662 .
this minimal length is denoted by @xmath663 , and is called the _ reduced length _ of @xmath618 . here
, of course , the length of @xmath664 is given by @xmath616 .
define @xmath665 to be the set of elements which have reduced length less than or equal to @xmath662 .
we have @xmath666 we let @xmath667 be the abelian group ( with discrete topology ) generated by symbols @xmath668 , @xmath272 , subject only to the relations @xmath669 ( if @xmath630 ) and @xmath670 for all @xmath272 .
[ f(s ) homomorphism ] the mapping @xmath671 for all @xmath672 extends to an isomorphism @xmath673 .
the existence of @xmath674 as a surjective homomorphism is clear since @xmath638 is the abelian group freely generated by the symbols @xmath675 subject only to the relations defining @xmath667 .
these correspond precisely to @xmath676 and @xmath677 when @xmath630 which hold in @xmath638 .
a typical element @xmath678 of @xmath667 is in the kernel of @xmath674 if and only if @xmath679 .
this happens only if the non - zero components of @xmath680 cancel in pairs @xmath681 .
then also the corresponding terms in @xmath678 cancel in pairs . since also @xmath669 in @xmath667 when @xmath603 , we see that @xmath682 reduces to the identity element .
we find it convenient to identify @xmath667 and @xmath638 through prop .
[ f(s ) homomorphism ] . having done this we obtain from @xmath683 , defined above , the homomorphism @xmath684 whose kernel can be identified with @xmath33 .
thus , we have the following exact sequence of groups ( all given the discrete topology ) @xmath685 dualization then gives the exact sequence @xmath686 where @xmath687 is a compact abelian group by pontryagin duality .
note that we get surjectivity from the continuity of the mappings and the compactness of these groups .
in fact , surjectivity of the map @xmath688 is crucial for our subsequent considerations . the mapping @xmath689 satisfies the @xmath616-_moment condition _ if @xmath690 whenever @xmath691 and @xmath692 .
we shall see the reason for calling these moment conditions in [ special ] .
notice that the first moment condition is empty ( and so trivially holds ! ) unless @xmath603 , in which case it says that @xmath693 .
the second moment condition is equivalent to the statement that @xmath694 for all @xmath275 . if @xmath630 then the second moment condition alone gives @xmath695 ,
so @xmath696 .
see remark [ 0-eigenfunction assumption ] . if the first moment condition is not empty and holds then @xmath697 and then the @xmath616-moment condition includes the @xmath662-moment condition for all @xmath698 since additional slots can be filled with zeros . obviously , any mapping @xmath699 satisfying the first and second moment conditions determines a character @xmath667 , i.e. an element of @xmath700 , via @xmath701 for all @xmath272 .
conversely every character of @xmath667 clearly determines a mapping @xmath436 satisfying the first and second moment conditions .
( notice that when @xmath702 then @xmath703 is the identity element of @xmath667 and also @xmath693 . ) in the sequel we will use these two concepts interchangeably and use the symbol @xmath436 both as a mapping on @xmath261 and as the corresponding character on @xmath667 .
when thought of as elements of @xmath700 they are called _ elementary phase forms_. by restriction any elementary phase form @xmath436 determines a phase form @xmath34 , that is , an element of @xmath647 .
the exact sequence shows that every phase form arises by restriction from an elementary phase form and two elementary phase forms restrict to the same phase form if and only their ratio is in @xmath704 , that is to say , if and only if their ratio is a character on @xmath615 .
this is the essential part of determining equivalence of pure point processes .
[ charactercondition ] * any mapping @xmath699 satisfying the first and second moment conditions determines a unique elementary phase form and every elementary phase form arises in this way .
* any phase form is the restriction of an elementary phase form .
the ratio of any two elementary phase forms giving the same phase form is an element of @xmath704 .
* an elementary phase form satisfies the @xmath616-moment conditions for @xmath705 if and only it kills @xmath665 . *
an elementary phase form @xmath436 determines a character on @xmath615 , i.e. lies in @xmath704 , if and only if it satisfies all @xmath616-moment conditions , @xmath706
. parts ( i ) and ( ii ) are already done above .
@xmath665 is generated by all the @xmath652 with @xmath707 , @xmath708 . an elementary phase form @xmath436 kills the equivalence class of @xmath709 in @xmath710 iff @xmath711 this proves ( iii ) .
an elementary phase form @xmath436 kills all of @xmath33 iff @xmath34 is trivial which happens iff @xmath712 , which gives ( iv ) .
of particular interest to us is the case where the group @xmath33 is generated by @xmath665 for some @xmath713 .
this can be understood on the level of phase forms in the following way .
[ charchar ] let @xmath714 .
then the following assertions are equivalent : * @xmath715 . *
any mapping @xmath716 satisfying the @xmath616-moment conditions for @xmath717 extends to a character on @xmath615 ( i.e. lies in @xmath704 ) .
( ii)@xmath339 ( i ) : the fact that @xmath716 satisfies the @xmath616-moment conditions for @xmath717 is equivalent to saying that the corresponding character @xmath436 on @xmath710 kills @xmath665 , or equivalently , kills the subgroup @xmath718 that it generates .
but prop .
[ charactercondition ] says that @xmath436 extends to a character on @xmath615 iff @xmath436 kills @xmath33 .
suppose that fact that the mapping @xmath716 satisfies the @xmath616-moment conditions for @xmath717 is enough to sufficient to guarantee that whenever @xmath436 kills @xmath718 it must also kill @xmath33 .
then @xmath719 because @xmath720 , both are closed subgroups of @xmath667 , and the pontryagin duality theory says that the characters can distinguish distinct closed subgroups .
the implication ( i)@xmath339 ( ii ) is clear .
the preceding considerations suggest to consider @xmath721 here , the infimum of the empty set is defined to be @xmath722 .
[ 2m+1 theorem ] if @xmath723 ( with @xmath724 summands ) then any relator can be written as a product of relators of length at most @xmath725 so @xmath726 generates @xmath33 .
thus , in this case @xmath727 @xcite .
the restriction of a phase form or an elementary phase form to @xmath728 is called its @xmath616th - moment .
[ mthmomentcharacterization ] two elementary phase forms @xmath436 and @xmath729 give rise to the same @xmath616th moments if and only if their ratio @xmath730 satisfies the @xmath616th moment condition .
elementary phase forms @xmath731 give rise to equal @xmath616th moments if and only if @xmath732 whenever @xmath691 with @xmath707 .
however , being elementary phase forms , these equations can be written as @xmath733 , which leads to the equivalent statement that @xmath734 whenever @xmath691 with @xmath707 .
this is the @xmath616th moment condition .
two elementary phase forms @xmath731 give rise to the same phase form @xmath34 if and only if their ratio is an element of @xmath704 and so satisfies all the moment conditions . by prop .
[ mthmomentcharacterization ] this means that all their moments are the same , so that there is no ambiguity in speaking of the @xmath616th _ moments of phase forms_. [ momentscondition ] assume @xmath715 for some @xmath713 .
then two phase forms agree if and only if they have the same @xmath616th moments for @xmath735 .
in particular , two elementary phase forms restrict to the same phase form if and only if they have the same @xmath616th moments for @xmath736 .
the first statement is clear .
the second statement is a direct consequence of the first statement .
any of the phase forms @xmath436 that we are interested in from the point of view of diffraction satisfy the first and second moment conditions by assumption .
the relevance of the previous lemma is the following .
if two stationary processes have the same pure point diffraction and their associated group of relators satisfies @xmath715 , then the processes are isomorphic if they have the same @xmath616th moments for @xmath736 ( see cororllary [ momentscondition1 ] and corollary [ momentscondition2 ] for precise statements ) .
in this section we assume that we are given an ergodic spatial stationary process with pure point diffraction .
we will discuss how this process gives rise to a diffraction measure @xmath30 , a set @xmath261 which generates a group @xmath592 , canonical eigenfunctions @xmath271 , @xmath272 , and a phase form @xmath34 . in the next section
we will see that @xmath737 in some sense uniquely determines the process .
let @xmath302 be an ergodic stationary spatial process with pure point diffraction . from the considerations of section [ stochasticprocesses ]
we then obtain the following : the diffraction measure @xmath30 of @xmath738 is a pure point measure and the set of its atoms , which are called bragg peaks , is denoted by @xmath587 i.e. @xmath739 and satisfies @xmath740 .
we let @xmath592 be the subgroup of @xmath4 generated by @xmath261 and let @xmath615 denote this group when it is given the discrete topology .
we are now clearly in the situation of the previous section . in particular
, there is an associated relator group @xmath741 at our disposal .
let @xmath589 be the associated diffraction to dynamics map and for each @xmath742 , let @xmath265 be the characteristic function on the set @xmath263 .
let @xmath196 be the action of @xmath1 acting on @xmath212 defined by @xmath743 and observe that whenever @xmath744 , @xmath265 is a @xmath270 eigenfunction for this action , .
it is easy to see that up to scaling these are the only eigenfunctions of @xmath212 . by the orthogonality of eigenfunctions we have , @xmath745 and @xmath746 for all @xmath747 . to each element of @xmath748
there exists a corresponding canonical eigenfunction .
more specifically , we define @xmath271 , @xmath272 , by @xmath749
. then each @xmath271 is an eigenfunction in @xmath111 for the eigenvalue @xmath270 and @xmath750 .
since @xmath751 is a constant function ( due to ergodicity ) and @xmath19 is a probability measure , @xmath752 .
then , for all @xmath691 , @xmath753 at this point we invoke assumption [ 0-eigenfunction assumption ] which says that if @xmath702 then @xmath754 .
for @xmath652 with @xmath755 and @xmath641=0 $ ] the product @xmath756 is an eigenfunction to @xmath94 and hence ( by ergodicity ) a multiple of @xmath757 , which yields @xmath758 for some @xmath759 . in this way we have a mapping @xmath760 .
it clearly respects the multiplication by concatenation on @xmath32 . in view of our assumption on @xmath604 and
lemma [ fminusk ] we have @xmath277 , which shows this @xmath34 is well - defined on the equivalence classes of @xmath761 defined in [ cycles ] , and in this way we obtain a phase form
@xmath762 the preceding considerations show that any pure point stationary spatial process comes with a natural measure @xmath477 and a phase form @xmath34 as summarized in the following proposition .
[ fromsptopf ] each ergodic pure point stationary spatial process @xmath302 gives rise to a pair @xmath763
consisting of a pure point measure @xmath30 on @xmath4 characterized by @xmath764 for all @xmath49 and the phase form @xmath34 on the relator group @xmath765 satisfying @xmath766 next we show that @xmath30 and @xmath767 are a complete set of invariants for the underlying process .
recall that we have introduced a notion of isomorphism between spatial processes in [ stochasticprocesses ] . in this section
we show that two pure point ergodic spatial stationary processes are isomorphic ( in the sense of definition [ definition - isomorphism ] ) if they yield the same diffraction measure @xmath477 and the same phase form @xmath34 .
we remind the reader that our concept of a pure point spatial stationary process given in definition [ def - pure - point - process ] entails that the process in question is full and possesses a diffraction measure ( which is pure point ) .
[ uniquenesstheorem ] two pure point ergodic stationary spatial processes based on the same group @xmath1 are isomorphic if and only if their associated diffraction measures and phase forms are the same .
assume that the full processes @xmath302 and @xmath768 are pure point ergodic and have the same diffraction measure @xmath30 and the same phase form @xmath34 .
set @xmath769 .
let @xmath770 and @xmath771 , @xmath272 , be the corresponding natural eigenfunctions .
thus , @xmath772 and @xmath773 for all @xmath27 .
moreover , by theorem [ ergodic+pure point ] products of the @xmath271 and the @xmath774 respectively provide orthonormal bases consisting of eigenfunctions of @xmath409 and @xmath775 respectively . we construct an isomorphism @xmath8 as follows : we define @xmath8 to be the unique linear map with @xmath776 this mapping is well - defined , for suppose that @xmath756 and @xmath777 are linearly dependent . then @xmath778 , so @xmath779 and @xmath780 using lemma [ fminusk ] we obtain @xmath781 by our assumptions we obtain the same relation when the @xmath85s are replaced with @xmath782s , so the same linear dependence occurs in @xmath783 . as products of the @xmath271 and @xmath774
respectively give an orthogonal basis of @xmath409 and @xmath775 , and as @xmath784 our mapping is a unitary map .
in particular it is continuous .
the definition of @xmath8 easily shows that it intertwines @xmath67 and @xmath785 and maps @xmath26 to @xmath786 for all @xmath53 .
moreover , by definition @xmath8 satisfies @xmath787 for all @xmath788 and @xmath789 .
note that both @xmath790 and @xmath791 are bounded functions .
thus , we can take linear combinations ( in @xmath126 ) and their limits to obtain @xmath792 for all @xmath134 and @xmath793 .
in particular , @xmath792 holds for all @xmath794 and @xmath793 . another approximation argument
now yields @xmath120 for all @xmath121 .
this proves the isomorphism .
conversely suppose that @xmath8 is an isomorphism between the pure point ergodic stationary spatial processes @xmath302 and @xmath768 on @xmath1 . as @xmath8 is a unitary mapping , one sees from directly from the definitions that their correlation measures , and hence their diffraction measures are the same ( if second moments exist ) , and then that the diffraction to dynamics mappings are related by @xmath795 .
then looking at the eigenfunctions we have @xmath796 . since the @xmath271
are bounded functions , the multiplicative property of @xmath8 gives @xmath791 for all @xmath691 .
then the definition of the phase form and shows that both processes have the same phase form .
[ momentscondition1 ] let @xmath477 be a positive symmetric pure point measure on @xmath4 and @xmath33 the associated group of relators . if @xmath797 , then two stationary spatial processes with diffraction @xmath477 are spatially isomorphic if and only if the first @xmath662 moments of their phase forms agree .
this is a direct consequence of the previous theorem and lemma [ momentscondition ] .
[ automorphism ] ( a ) the above theorem deals with isomorphism between two processes .
we can also characterize the automorphisms of a given pure point stationary spatial point processes with bragg peaks @xmath261 and eigenfunctions @xmath271 , @xmath272 .
these automorphisms are in one to one correspondence with elements from @xmath704 in the following way : any character @xmath798 yields an automorphism of @xmath302 via @xmath799 .
conversely any automorphism of @xmath302 must necessarily map map @xmath271 to a multiple @xmath800 with @xmath801 .
the arising function @xmath372 must then belong to @xmath704 as gives @xmath802 whenever @xmath803 .
\(b ) in the previous theorem , the isomorphism @xmath8 between isomorphic pure point stationary spatial processes is not unique .
however , if @xmath804 are two such isomorphisms then @xmath805 is an automorphism of the first , say @xmath302 . by ( a ) of this remark , we conclude that the isomorphisms are unique up to elements of @xmath704 .
in this section we show how to construct an ergodic full stationary process with a given pure point measure as its diffraction measure @xmath477 and a given phase form @xmath34 as its associated phase form .
this yields a canonical model realizing a given diffraction measure .
we assume that we are given a a locally compact abelian group @xmath1 , a pure point positive symmetric backward transformable measure measure @xmath30 on @xmath4 .
these data give rise to @xmath806 and @xmath807 .
there is an associated group of relators @xmath33 as discussed in section [ cycles ] .
we are also given @xmath808 .
we are going to construct a stationary process corresponding to @xmath763 . from prop .
[ charactercondition ] we know that there is an elementary phase form @xmath809 ( i.e. a character of @xmath667 ) which restricts to @xmath34 : in other words , we can find @xmath436 so that @xmath810 whenever @xmath641=0 $ ] .
( this @xmath436 is not unique , but the ratio of any two of such elementary phase forms is a character on @xmath615 . )
thus , we can assume without loss of generality that @xmath34 comes from an elementary phase form @xmath811 . since @xmath30 is backward transformable
there is a measure @xmath141 on @xmath1 so that for all @xmath27 we have @xmath812 we consider @xmath592 to have the induced topology from @xmath4 and , as usual , let @xmath615 denote the same group @xmath592 but with the discrete topology .
the dual of @xmath615 is a compact abelian group @xmath704 .
we denote the haar measure of total volume @xmath268 on @xmath704 by @xmath813 and the counting measure on @xmath615 by @xmath814 .
then , can be reformulated as saying that @xmath815 for @xmath248 , define a function @xmath816 on @xmath817 by @xmath818 where we take the positive square roots and @xmath265 is the function on @xmath615 whose value at @xmath270 is @xmath268 and which takes the value @xmath94 everywhere else .
we have @xmath819 which implies in particular that @xmath820 .
the fourier transform provides a fundamental isomorphism between @xmath821 and @xmath822 taking @xmath823 to the character defined by @xmath270 on @xmath704 .
we usually denote this character by @xmath824 .
thus we obtain , by applying the fourier transform to @xmath825 , @xmath826 with @xmath827 this provides us with the mapping @xmath828 and shows that for any compact subset @xmath41 we have @xmath829 for all @xmath310 with support inside @xmath15 , which shows that @xmath830 is continuous .
the continuous embedding of @xmath831 defined by inclusion leads to a dense homomorphism @xmath832 .
then each function on @xmath704 restricts " to one on @xmath1 and in particular we have an obvious meaning to the functions @xmath833 for all @xmath77 and for all @xmath834 .
we also obtain a natural ergodic action of @xmath1 on @xmath704 . for @xmath77 and @xmath835 , @xmath836 for all @xmath834 . for @xmath248 and @xmath837 @xmath838
( here we use the invariance of the measure @xmath170 . )
this leads by a straightforward calculation that @xmath839 , which shows that @xmath830 is a @xmath1-equivariant map .
thus given @xmath840 we obtain an ergodic spatial stationary process @xmath841 .
the above considerations also show that @xmath30 is the diffraction measure associated to @xmath830 .
we will show next that @xmath830 is full and has associated phase form @xmath34 .
to do so we consider the map @xmath842 then , @xmath142 is an isometry with @xmath843 for all @xmath27 . by uniqueness of the dynamics - to - diffraction map
we infer that @xmath142 _ is _ the dynamics - to - diffraction map .
this gives in particular , that @xmath844 for all @xmath272 .
these are the functions @xmath271 of the previous sections . from this equality and the fact that @xmath436 is a character on @xmath667 it follows easily that @xmath34 is the phase form associated to @xmath830 ( see and ) .
moreover , proposition [ diffractiontodynamics ] now shows that @xmath845 belongs to the closure of the linear span of the @xmath26 , @xmath53 , for any @xmath672 . as products of these @xmath824 form an orthonormal basis of @xmath846 we obtain fullness of the stationary process @xmath830 .
we summarize the conclusions of this section .
[ inverseproblem ] let a pure point positive symmetric backward transformable measure @xmath30 on @xmath1 be given and @xmath34 be a phase form . then for any choice of elementary phase form
@xmath436 restricting to @xmath34 , @xmath847 is an ergodic full stationary process on @xmath1 with diffraction measure @xmath30 and phase form @xmath34 . the pure point process constructed in prop .
[ inverseproblem ] depends on the choice of the elementary phase form @xmath436 that we choose to represent the phase form @xmath34 .
what happens if we choose another representing elementary phase form , @xmath848 ?
we know already by thm .
[ uniquenesstheorem ] that the resulting process will be isomorphic , and since both use the same constructed dynamical system @xmath849 , this difference between the two processes is in effect an automorphism .
also , proposition [ charactercondition ] ( ii ) says that the ratio @xmath850 is an element of @xmath704 .
from the perspective of @xmath830 and @xmath851 , hten @xmath851 is in effect a translation of @xmath830 ba @xmath852 i.e. @xmath853 [ continuouscase ] if @xmath854 above is actually continuous with respect to the original topology on @xmath592 then @xmath855 , being a character on @xmath592 , lifts to a character on @xmath856 and then by general character theory to all of @xmath4 .
thus @xmath855 can be identified with an element of @xmath1 .
now becomes @xmath857 this time the difference between @xmath830 and @xmath851 is a translation on @xmath1 .
this situation occurs in the periodic situation , when @xmath1 is compact and @xmath4 is discrete .
[ momentscondition2 ] let @xmath477 be a positive backward transformable pure point measure on @xmath4 and @xmath33 the associated group of relators .
if @xmath797 , then two mappings @xmath858 define isomorphic stationary pure point processes if and only if their first through @xmath662th moments are equal .
this is a direct consequence of corollary [ momentscondition1 ] .
in this section we discuss a main result of the paper viz our solution to the homometry problem for pure point diffraction .
[ diffractionclass ] given @xmath1 and a measure @xmath30 on @xmath4 , we let @xmath859 denote the set of all spatial isomorphism classes of ergodic full stationary spatial processes with diffraction @xmath30 satisfying assumption [ 0-eigenfunction assumption ] . the considerations of the previous sections rather directly prove the following result , which is in effect a solution of the homometry problem for pure point diffraction .
[ homometrytheorem ] let @xmath1 be a locally compact abelian group .
let @xmath30 be a positive backward transformable pure point measure on @xmath4 and @xmath741 the associated group of relators .
then , the map @xmath860 @xmath861,\ ] ] where @xmath436 is any elementary phase form representing @xmath34 , is a bijection . by proposition [ inverseproblem ] , @xmath862 is an ergodic spatial stationary process with diffraction @xmath477 and phase @xmath767 .
moreover , by theorem [ uniquenesstheorem ] , the map is well - defined ( i.e. independent of the choice of the elementary phase form representing @xmath767 ) and one - to - one .
it is surjective by proposition [ fromsptopf ] and theorem [ uniquenesstheorem ] .
we note the following immediate consequence of the previous theorem and corollary [ momentscondition1 ] ( see corollary [ momentscondition2 ] as well ) .
[ finite - coro ] assume the situation of the theorem .
if @xmath797 , then two ergodic spatial stationary processes over @xmath1 with diffraction measure @xmath30 are spatially isomorphic if and only if their phase forms have the same @xmath616-th moments for @xmath863 .
although it is obvious from the theory of pure point spatial processes that we have developed , it is worthwhile noting explicitly that the problem of classification of pure point spatial processes with a given @xmath30 depends on only the positions of the bragg peaks , namely @xmath261 .
once one has @xmath261 , one has the group @xmath33 , which depends only on the relationship of @xmath261 to the subgroup @xmath592 of @xmath4 that it generates . and
in fact this purely algebraic problem is in reality the entire homometry problem !
[ compactg].2 for a @xmath864 case when @xmath1 is finite .
cases when @xmath1 is compact ( or even finite ) are important since they arise in the study of periodic and limit periodic cases of diffractive structures , and their relative simplicity allows us to see more clearly the nature of the spatial processes that we have introduced . in this section
we first look at what happens in general when we are given that the group @xmath1 is compact .
we then move to some examples .
these examples revolve around the case that @xmath1 is finite and/or around one of the most basic of all diffraction measures , namely the diffraction of @xmath865 , @xmath866 .
we also look at an interesting result of grnbaum and moore that involves rational periodic diffraction on the line .
the fact that there is anything to say in this remarkably simple case and that it seems to be very difficult to extend their results to a two dimensional setting is a bit of a testimony to the difficulties that are inherent in the homometry problem .
we now look at the method of [ cyclestostochasticprocesses ] in this compact case .
we assume that @xmath30 is positive , pure point , centrally symmetric ( @xmath867 , for all @xmath272 ) , and backward transformable .
let @xmath261 be the support of @xmath30 .
it is convenient to assume that @xmath261 generates @xmath4 as a group , for otherwise we may use @xmath868 and @xmath869 instead
. then @xmath870 and the carrying space for our spatial process is @xmath871 when we wish to distinguish two roles of @xmath1 , first as a group of translations and second as the set of states of dynamical system , we shall use notation like @xmath872 and @xmath873 respectively .
moreover , for any character @xmath270 on @xmath704 ( i.e. @xmath222 ) we will denote the map @xmath874 by @xmath270 as well .
when we come to the transformability into the autocorrelation measure @xmath141 , we require that for all @xmath875 , @xmath876 thus the autocorrelation fourier dual to @xmath30 is given by @xmath877 provided that this is indeed a measure .
since @xmath1 is compact , and @xmath141 is supposed to be a borel measure , this measure must be finite .
assuming transformability we can , for any character @xmath809 , form the associated stationary process @xmath878 namely , @xmath879 for all @xmath880 .
note that @xmath881 belongs indeed to @xmath409 due to @xmath882 .
our interpretation is that @xmath830 represents some sort of density on the space @xmath1 .
each point @xmath883 represents an instance of the yet unspecified structure which can be paired with elements of @xmath155 to give @xmath884 where the left hand side needs to be given an interpretation .
it turns out that such an interpretation can be given easily if we make the finiteness assumption that @xmath885 note that this assumption immediately implies that @xmath477 is backward transformable as it yields that @xmath141 given in @xmath886 is indeed a measure .
given @xmath887 , we can define the function @xmath888 by @xmath889 by standard theory of fourier series .
then , a short calculation invoking unitarity of the fourier transform ( see e.g. as well ) gives @xmath890 note that we have indeed pointwise ( in t ) existence of @xmath891 as both @xmath892 and @xmath477 are square summable i.e. belong to @xmath893 . comparing with @xmath894
, we find that we can identify @xmath895 with the @xmath126 function @xmath896 and the pairing between @xmath897 and @xmath155 with ordinary integration . when can this ` density ' be interpreted as a measure @xmath898 on @xmath899 ?
the idea is that for some @xmath900 ( which we can take to be @xmath901 ) the structure is @xmath898 and that as we translate around the translate @xmath902 represents the structure at @xmath895 .
usually one would have to assume by ergodicity the denseness of this orbit , but in the compact case as we have it here , the orbit is the entire space @xmath18 .
this entails that @xmath903 where and then @xmath904 .
this gives when we treat @xmath898 as a measure .
as @xmath1 is compact the @xmath126-function @xmath896 belongs to @xmath905 as well and we can consider it to be the measure @xmath906 thus , in this situation we can realize the process as a measure process . if @xmath907 , then @xmath898 will even be a continuous function . if @xmath908 , the situation becomes different .
such a situation can easily arise : it suffices to find @xmath436 and @xmath477 such that @xmath898 does not belong to @xmath909 .
let for example @xmath1 be compact admitting a function @xmath910 .
as @xmath911 belongs to @xmath126 , we can expand @xmath911 in a fourier series @xmath912 with @xmath913 and @xmath914 with @xmath915 .
the diffraction of @xmath911 is seen directly to be equal to @xmath477 . for more on interpreting @xmath830
as some sort of measure induced density on @xmath1 see [ ssp - measures ] .
in general the number of solutions to the inverse problem is vast , even for the most basic diffraction patterns . in this section
we consider the most famous example of a diffraction pattern , the diffraction of the integers .
more properly this is the diffraction of the dirac comb @xmath917 which represents a point density of one at each of the integers .
its diffraction is also @xmath916 .
crystallographers have long relied on additional information ( like the periodicity of crystals and their known constituents ) to find solutions to the inverse problem .
we shall see that by imposing periodicity and/or imposing arithmetic conditions on the solutions , the inverse problem we can do the same .
one of the advantages of our approach , that applies to all locally compact abelian groups , is that it also applies to compact and to finite groups .
although our object of attention is @xmath866 , much of this section is devoted to cases in which @xmath154 is finite and to showing what the theory then looks like and tells us .
we also deal with the case @xmath918 . in the case of diffraction from a periodic set or periodic crystal , it is customary to compute the diffraction directly from the originating measure describing the density @xmath911 , by taking the square - absolute value of the fourier transform of @xmath911 . in the case of @xmath866
this approach gives the answer directly as a consequence of the poisson summation formula .
this also works directly for finite groups .
we start with the pair of dual groups @xmath919 and @xmath920 where @xmath8 is a positive integer .
the duality is most conveniently written down in the form @xmath921 for @xmath922 and @xmath923 . in the context of this paper ,
one of these groups plays the role of @xmath1 and the other of @xmath4 .
most of the time it will be @xmath924 and @xmath925 . to help clarify things , elements of @xmath4 will be written in the form @xmath926 or @xmath824 accordingly .
the idea here is to use the fact that the diffraction @xmath866 is periodic and then represent it in @xmath920 and look for all solutions for an originating distribution @xmath911 in @xmath919 .
the resulting distribution will then be interpreted as a periodic ( modulo @xmath865 ) measure on @xmath927 . in this way
we pick up periodic solutions to the inverse problem .
alternatively we could do things the other way around and consider @xmath928 represented in @xmath919 and find solutions to the inverse problem in @xmath920 which we could interpret as periodic solutions in @xmath865 .
however , we shall see that does not lead to new solutions to the inverse problem .
we note that for finite @xmath1 and @xmath4 the haar measure we want to use is counting measure normalized to total measure @xmath268 by dividing by the order of the group . for @xmath929 , simple calculation gives the expected result that @xmath930 ( note that @xmath824 is a function on @xmath1 and its fourier transform is a function on @xmath4 ) .
since the groups are finite , there is no complication in computing diffraction . for a density or distribution of density @xmath911 on @xmath1 ,
the autocorrelation is @xmath931 and the diffraction is @xmath932 .
we should note that in general the approach via the square - absolute value of the fourier transform is not applicable to aperiodic structures .
as a. hof has shown @xcite and @xcite , in the study of aperiodic crystals one usually meets distributions of density that are not fourier transformable .
our approach is to follow the lines laid out above and for given diffraction @xmath30 and elementary phase form @xmath436 define the associated density via : @xmath933 this is a finite sum and there are no questions about convergence or ambiguities about how to interpret it .
this seems too trivial to consider , but it is interesting nonetheless .
@xmath935 , @xmath936 , @xmath937 . here
@xmath938 and @xmath939 is generated by @xmath940 .
so @xmath939 is the trivial group and its only character is the trivial character .
also @xmath941 .
the density measure @xmath911 defined by the trivial character @xmath436 is @xmath942 which is simply the identity function on @xmath1 .
thus we obtain the expected solution to the inverse problem : @xmath943 whose absolute square is again @xmath937 . the function @xmath944 is the identity function on @xmath945 , which is also the measure @xmath946 on @xmath1 .
\(1 ) @xmath948 , @xmath949 , @xmath950 .
the two elements of @xmath4 are @xmath944 and @xmath951 .
we have @xmath952 and @xmath939 is generated by @xmath953 and @xmath954 , with @xmath955 since @xmath956 .
thus @xmath939 is cyclic of order @xmath957 .
there are only two characters on @xmath939 , i.e. elementary phase forms , @xmath958 and @xmath959 .
the exact sequence @xmath960 comes from the mapping @xmath961 which here reads @xmath962 and @xmath963 .
thus the kernel @xmath33 is trivial .
the significance of this is that there is only one phase form ( trivial ! ) and thus the two solutions we derive from the two elementary phase forms will be the same up to translation . in detail , we have , using @xmath964 @xmath965 note that this is a function on @xmath966 .
its values are @xmath967 .
this is a nice transformable measure and its fourier transform is the measure is @xmath968 on @xmath969 .
its absolute square is @xmath30 , as we expected .
similarly , using @xmath970 we obtain @xmath971 its values are @xmath972 , so the density distribution @xmath973 on @xmath1 is just translation by @xmath974 of @xmath975 found in the case of @xmath964 .
the fourier transform is @xmath976 and again the absolute square of this is @xmath30 .
\(2 ) next we reverse the roles of @xmath1 and @xmath4 : @xmath977 , @xmath978 , @xmath937 .
notice that @xmath4 represents the reduction of the integers and the half - integers modulo @xmath865 .
the reduction of the diffraction @xmath866 modulo @xmath865 is thus supported only on the @xmath865 part which is represented by the class of @xmath94 modulo @xmath865 . following the general discussion of compact groups we should ,
at this point , replace @xmath4 by the subgroup @xmath979 .
this then reduces us to the case @xmath934 already discussed .
\(3 ) it is interesting to consider @xmath977 and @xmath978 with the diffraction @xmath980 .
this may be considered as the reduction modulo @xmath865 of @xmath981 .
the two elements of @xmath4 are @xmath944 and @xmath982 .
the discussion parallels that in ( 1 ) .
we have @xmath983 and @xmath939 is generated by @xmath953 and @xmath984 with @xmath985 since @xmath986 .
thus @xmath939 is cyclic of order @xmath957 .
there are only two characters on @xmath939 , i.e. elementary phase forms , @xmath987 and @xmath988 .
the exact sequence @xmath960 comes from the mappinng @xmath961 which here reads @xmath962 and @xmath989 .
thus again the kernel @xmath33 is trivial and the two solutions we derive will be the same up to translation . using @xmath964 @xmath990 thus the distribution on @xmath991
is @xmath992 .
its fourier transform is the measure @xmath993 , whose absolute square is @xmath30 .
going back to @xmath865 we have @xmath994 on @xmath865 whose diffraction is @xmath995 , which agrees with what we obtained after removing all periodicity . similarly , using @xmath970
we obtain @xmath996 and the distribution on @xmath991 is @xmath997 .
then @xmath998 . taking absolute squares we again
obtain the diffraction @xmath30 .
again going back to @xmath865 , this time we have @xmath999 on @xmath865 whose diffraction is @xmath1000 , which agrees with what we obtained after removing all periodicity . here
something more interesting happens .
@xmath1002 , @xmath1003 .
the three elements of @xmath4 are @xmath1004 , with @xmath1005 , where @xmath1006 .
we have @xmath1007 and @xmath939 is generated by @xmath1008 , with @xmath1009 since @xmath1010 .
thus @xmath1011 with @xmath954 as a ( multiplicative ) generator .
there are infinitely many characters on @xmath939 , i.e. elementary phase forms , one for each @xmath1012 : @xmath1013 and @xmath1014 for all @xmath1015 .
the exact sequence @xmath960 comes from the homomorphism that maps @xmath1016 ( which includes @xmath1017 ) .
the kernel @xmath1018 via the isomorphism @xmath1011 . for each @xmath1019
we obtain @xmath1020 thus the distribution on @xmath1021 is @xmath1022 , @xmath1023 , and @xmath1024 since @xmath930 we obtain @xmath1025 .
its absolute square is @xmath30 as required .
we note that replacing @xmath855 by @xmath1026 or @xmath1027 changes the distribution @xmath911 by translation , which corresponds to the fact that the elementary phase forms map @xmath1028 to phase forms .
thus there is a one parameter family of solutions to the inverse problem for @xmath30 , which are supported on @xmath865 and periodic of period @xmath1029 .
if we look at the inverse problem to @xmath1030 in @xmath919 the situation evolves in a natural way . at @xmath1031 the value @xmath957 leads to @xmath1032 , and @xmath1033 is arbitrary , so we end up with solutions @xmath1034 . at @xmath1035
we are get two free parameters @xmath1036 , one relating to the pair @xmath1037 and one to the pair @xmath1038 ( integers taken modulo @xmath1039 ) , and the solutions are @xmath1040 . the results for higher @xmath8 follow a similar pattern .
we again consider the diffraction @xmath866 , but now use @xmath865 as @xmath4 and @xmath1042 as @xmath1 .
then @xmath1043 and @xmath939 is generated by @xmath1044 with the single set of relations @xmath1045 ( along with @xmath940 ) .
then we have elementary phase forms @xmath1046 with @xmath1047 arbitrary except that @xmath1048 .
each such @xmath436 leads to a formal solution @xmath1049 this is simply a formal fourier series on @xmath865 .
however , given that for all @xmath270 , @xmath1050 , there is no convergence as a fourier series .
this is expected since in the simplest case when @xmath1051 for all @xmath270 its fourier transform is the unbounded measure @xmath866 . however , the solution for @xmath830 : @xmath1052 for all @xmath1053 is well - defined as an @xmath126-function , and we can learn something from it . in the case that @xmath436 is the trivial elementary phase form we obtain @xmath1054 thus the associated density is @xmath946 , which is what we expect .
recall [ the compact setting ] here .
the compact group @xmath704 , which is @xmath1042 here , is viewed both as a dynamical system with states @xmath895 and as a group of translations @xmath1055 which translate the states about .
@xmath1056 is something that gives the effect of the process on the function @xmath113 at the state @xmath895 .
now let @xmath1057 be a finite symmetric subset of @xmath1058 and let @xmath436 be an elementary phase form which satisfies @xmath1059 for all @xmath1060 and @xmath1061 arbitrary for all @xmath1062 .
now for @xmath1053 we have @xmath1063 for all @xmath1055 .
following the same reasoning as in we arrive at @xmath1064 where @xmath1065 remarkably the density @xmath1066 on @xmath1042 is indeed a solution to the inverse problem for @xmath946 , as can be seen by directly computing its autocorrelation and taking its fourier transform .
this shows the ubiquity of quite reasonable looking solutions to the problem .
in this section we consider the very simple situation of the diffraction from the integers when they are weighted periodically with period @xmath8 by a set @xmath1067 of real numbers . in this case
the diffraction is also periodic , and we will assume that the ambient space is @xmath1068 and the diffraction occurs in its dual , @xmath1069 , say @xmath1070 as we noted above , making the assumption that @xmath1068 is not entirely innocent since it puts considerable limitations on the kinds of solutions available namely that they are periodic of period @xmath8 and the distribution of intensity is entirely confined to integers , neither of which is mandatory . taking a larger ambient space
will in general produce more solutions ( many more ! ) .
equation defines the density on @xmath1 arising from @xmath30 and the elementary phase form @xmath436 .
since its diffraction is to be @xmath30 , this leads to the @xmath8 equations @xmath1071 where @xmath1072 is the set of non - vanishing points of @xmath30 and @xmath1073 $ ] , where @xmath1074 ( the values of @xmath270 are of the form @xmath1075 where @xmath1076 ) .
notice that for finite groups , the situation we are in here , the densities @xmath898 are functions and also measures , since the distinction between them disappears .
grnbaum and moore @xcite consider the case in which @xmath898 is rational valued , i.e. all of the weights @xmath1077 are rational numbers . in this case , solving the system of equations for the ` unknowns ' @xmath1078 we see that they must all lie in @xmath1079 $ ] , and then @xmath1080 $ ] also .
let @xmath1081/{\mathbb{q } } ) \simeq ( { \mathbb{z}}/m{\mathbb{z}})^\times$ ] , the group of units of the ring @xmath920 .
we can view @xmath1082 as both a group of automorphisms permuting the @xmath8 roots of unity , or as a group of automorphisms of @xmath919 ; namely @xmath1083 acts as @xmath1084 and as @xmath1085 respectively . in the rational case
we are now in , it is shown in @xcite that @xmath1086 for all @xmath1087 .
we show next why this happens : we take @xmath1088 if @xmath1089 . if @xmath1090 then @xmath1091 which is zero if @xmath1092 .
let @xmath1093 . since @xmath898 is rational valued , @xmath1094 . applying the process of summation of equation to the element @xmath1095 where @xmath1096 we obtain @xmath1097 .
but applying @xmath1098 dircectly to we obtain @xmath1099 .
these two are equal , and this shows , in particular , that @xmath1100 , and hence that @xmath1101 , which is what we wished to prove . in short ,
if we identify @xmath261 as a subset of @xmath919 , then @xmath275 and @xmath1083 implies @xmath1102 . in effect
@xmath261 is a union of orbits of @xmath919 under the action of the multiplicative group @xmath1103 .
this condition appears in an equivalent form in the following key result of @xcite . for @xmath1104 ,
let @xmath1105 denote its order as a group element modulo @xmath8 ( under addition ) .
[ gmthm]if the density @xmath898 of is rational valued then for all @xmath272 and for all @xmath1106 with @xmath1107 one has @xmath1108 . in this case
any mapping @xmath1109 satisfying the @xmath616-moment condition for all @xmath1110 if @xmath8 is even ( respectively @xmath1111 if @xmath8 is odd ) extends to a character on @xmath919 .
using proposition [ charchar ] , and corollary [ finite - coro ] and the corresponding notation , we obtain under the conditions of proposition [ gmthm ] , the set of distributions @xmath898 whose diffraction is @xmath30 of is classified by the first @xmath1112 moments ( resp .
first @xmath1113 moments ) if @xmath8 is even ( respectively if @xmath8 is odd ) .
we do not know of any extensions of this result into higher dimensional periodic situations , and indeed it seems very difficult to do .
here we give an example based on an example given in @xcite ( see also @xcite ) .
we begin with @xmath1114 and the two weighted dirac combs @xmath1115 on @xmath1 , which represent two density distributions for which we would like to find all possible homometric density distributions .
what is remarkable is that not only are these two density distributions homometric , they actually have all of their first @xmath1039 moments in common .
let us begin with the first of these two densities .
the fourier transform of this density is a function on @xmath1116 .
the latter is most conveniently thought of as the @xmath1112th roots of unity .
we write the fourier transform in the form @xmath1117 where @xmath1118 varies over the @xmath1112th roots of @xmath268 , @xmath1119 .
note that @xmath1120 , so the coefficient of @xmath1121 is @xmath1122 as it is supposed to be .
the diffraction at @xmath742 is @xmath1123 , which works out to give the following values for @xmath1124 @xmath1125 , @xmath1126 , @xmath1127 , @xmath94 , @xmath94 , @xmath1128 at @xmath1129 .
thus in this example @xmath1130 .
in fact grnbaum and moore constructed the polynomial @xmath1131 precisely to make the bragg spectrum have extinctions at @xmath1132 . in the rest of this subsection
we shall make the notation a little simpler by suppressing the ubiquitous denominators @xmath1112 which occur in the values of @xmath270 .
alternatively we multiply everything by @xmath1112 and work in @xmath1133 instead of @xmath1134 . in any @xmath616-tuple @xmath1135 in @xmath642 , we can , using the equivalence relation @xmath1136 , drop any occurrences of @xmath94 and drop any pair @xmath1137 that occurs amongst the @xmath1138 .
this leads us to conclude the only equivalence classes of @xmath627 are of the form @xmath1139 , where there are any positive integer @xmath616 entries in the two latter types .
since @xmath1140 and @xmath1141 with @xmath616 entries are inverses of each other in @xmath627 , we conclude that @xmath1142 . since the mapping @xmath1143 , we have @xmath1144 ( prop . [
f(s ) homomorphism ] ) . using the identification of @xmath638 and @xmath667 we have @xmath1145 and we can put together the exact sequence :
@xmath1146 dualization then gives the exact sequence @xmath1147 each element of @xmath1042 can be viewed as a character of @xmath1148 , and when restricted to @xmath865 characters @xmath1149 and @xmath1150 , @xmath1151 , are the same .
we now write down the densities @xmath1152 here @xmath1153 .
the character @xmath436 of @xmath667 is a character of @xmath865 and hence is an element of @xmath1042 .
recall that @xmath667 is generated by the tuples @xmath1154 and their inverses @xmath1155 where @xmath1137 represent classes modulo @xmath1112 in @xmath865 .
there are no further relations .
thus @xmath1156 here really means @xmath1157 , that is , the value of @xmath436 on the the equivalence class of @xmath1158 under @xmath1136 .
thus by @xmath1159 we mean @xmath1160 and by @xmath1161 we mean @xmath1162 . there is no restriction on what @xmath1159 might be in @xmath1042 , so simply writing it as @xmath436 we may write @xmath911 as @xmath1163 where @xmath1164 is arbitrary .
each value of @xmath436 gives a solution to the diffraction problem .
the following figure illustrates the graphs of @xmath1165 .
the situation that we began with occurs when @xmath1166 .
another solution indicated in @xcite is @xmath1167 which occurs at @xmath1168 . as @xmath436 varies for each of the six values @xmath1169 of @xmath298 . here
@xmath436 varies over the unit circle @xmath1170 , indicated here by the interval @xmath1171 and the six graphs give the corresponding values of the coefficients or weights of the deltas that make up the distribution on @xmath1133 .
the original weighting distribution @xmath1172 occurs when @xmath1166
. the second density of with coefficients @xmath1167 also occurs , when @xmath1173 .
, width=264 ] now we are going to show that in this example , @xmath1174 , but @xmath1175 .
the argument is quite simple .
first note that @xmath1176 but can not be reduced , so its reduced length is @xmath1112 ( note that denominators are still being suppressed ) .
so @xmath1177 .
now consider any @xmath1178 with @xmath1179 , which is of minimal length and can not be split into a product of two ( non trivial ) elements from @xmath33 .
there can be no occurrences of @xmath94 since those can be discarded .
since @xmath1180 and @xmath1181 are annihilating pairs modulo @xmath1112 , we can not have both @xmath268 and @xmath1039 in @xmath618 and neither can we have both @xmath957 and @xmath1113 .
neither can we have @xmath957 appearing three or more times since @xmath1182 and similarly for @xmath1113 .
likewise @xmath1029 can appear at most once .
suppose @xmath957 or @xmath1029 appears in @xmath618 .
the possibilities are @xmath1183 .
the remaining symbols are just @xmath268s or @xmath1039s . but
@xmath1184 then show that @xmath618 is not reduced .
the argument with @xmath1113 is much the same .
so we can assume that none of @xmath1185 appear and not both @xmath268 and @xmath1039 appear .
this leaves just @xmath1186 and @xmath1187 as the possibilities for reduced words , and in either case the number of entries must be a multiple of @xmath1112 .
the consequence of this is that any phase form is uniquely determined by its first through sixth moments . for
if two have these moments in common then taking corresponding elementary phase forms @xmath731 we find that @xmath1188 and satisfies the first @xmath1112 moment conditions .
since @xmath1174 , @xmath1189 lies in @xmath704 and so the corresponding phase forms were equal .
it is a surprising fact , pointed out in @xcite , that the two densities of , @xmath1190 and @xmath1191 indicated above , have all of their first through fifth moments in common , and it is only at the sixth moments that they look different .
in [ cyclestostochasticprocesses ] we offered a direct method of constructing a spatial process from the original data consisting of a positive backward transformable point measure @xmath30 on @xmath52 and a phase form @xmath34 ( or equivalently an elementary phase form @xmath436 ) .
this approach can in a certain sense be understood as an elaboration of the theorem of halmos - von neumann which tells us in advance that since the required process will have an associated dynamical system with pure point spectrum , it can be modelled ( measure theoretically ) on a compact abelian group . in [
pc : split ] we have used the gelfand method to produce pure point spatial processes in the splitting of spatial processes into their pure and continuous parts . in this present section we sketch out how we can also use the gelfand method to produce a pure point spatial process out of the given pure point measure @xmath30 and the elementary phase form @xmath436 .
the method ultimately reverses many of the steps that we saw earlier in which , starting from a process we constructed the diffraction and the diffraction - to - dynamics map . for this reason we do not give proofs but rather provide an outline of intermediate steps .
we use the notation of [ cycles ] . for simplicity of presentation
we shall assume that @xmath630 .
define @xmath1192 by @xmath1193 or more compactly , @xmath1194 , where for each square root we take the non - negative root .
we define @xmath1195 .
note that @xmath1196 .
let @xmath1197 $ ] be the free associative algebra in the variables @xmath1198 .
for elements @xmath1199 where the @xmath755 , we will write @xmath1200 .
we define an algebra @xmath592-grading on @xmath1197 $ ] by assigning degree @xmath270 to each @xmath1198 , so @xmath1201 gets degree @xmath1202 $ ] .
let @xmath1203 denote the ideal of @xmath1197 $ ] generated by all the elements of the form @xmath1204 for all @xmath1205 , i.e. for all @xmath618 with @xmath641 = 0 $ ] ( this includes @xmath1206 ) and let @xmath1207/i \,.\ ] ] we let @xmath1208 \longrightarrow { \mathcal{p}}\ ] ] be the natural homomorphism and set @xmath1209 for @xmath1210 .
since @xmath1211 is independent of the order of the @xmath1138 , we see that @xmath1212 is commutative .
furthermore , since the relations imposed on @xmath1213 $ ] are homogeneous of degree @xmath94 , the resulting algebra @xmath1214 is also graded by @xmath592 .
we denote its space of elements of degree @xmath1215 by @xmath1216 .
the space @xmath1213_\kappa$ ] of elements of degree @xmath1215 is the linear span of all the elements @xmath1217 with @xmath1202 = \kappa$ ] . if @xmath1202 = [ { { \bf{l } } } ] = \kappa$ ] then @xmath1218 and the relations defining @xmath1219 lead to @xmath1220 from this we see first that @xmath1221 where @xmath1222 is any one of the elements @xmath1217 with @xmath1202 = \kappa$ ] .
we shall always assume that we are working with a phase form @xmath436 for which @xmath1214 is not reduced to the trivial ring @xmath1223 .
such cycles always exist , for example the trivial cycle , which is @xmath268 everywhere .
since the algebra is @xmath592-graded , non - zero elements of different degrees are linearly independent and @xmath1224 define an conjugate linear form on @xmath1213 $ ] by sesqui - linear extension ( with conjugation on the second variable ) of @xmath1225 = [ { \bf l}]$ } \ , , \\
0 \quad \mbox{otherwise } \ , .
\end{array } \right.\ ] ] it is a straightforward exercise to show the identity @xmath1226 for all @xmath1227 $ ] .
we shall write @xmath1228 for @xmath1229 .
let @xmath1230 be tuples of elements of @xmath261 .
* @xmath1231 ; * @xmath1232 ; @xmath1233 @xmath1234 = 0 $ ] ; * @xmath1235 ; * @xmath1236 when @xmath1202 = 0 $ ] ; * the kernel of @xmath1237 is the ideal of relations that we have imposed on @xmath1213 $ ] ; * @xmath1237 induces an inner product ( for which we use the same notation ) on @xmath1212 .
now @xmath1212 is an inner product space .
we wish next to consider @xmath1212 acting on itself by multiplication and to make it into a normed algebra by use of the operator norm .
the details are as follows .
a typical element of @xmath1212 can be written as a finite sum @xmath1238 with @xmath1239 , where there is only one summand of any particular degree ( i.e. @xmath1240 \ne [ { { \bf{l } } } ' ] $ ] if @xmath1241 )
. then we have @xmath1242 let @xmath1243 .
then @xmath1244 with degrees @xmath1202 + [ { { \bf{l}}}]$ ] and @xmath1245 thus @xmath1246 , left multiplication by @xmath1247 is a linear operator of norm @xmath1248 .
now we see that the entire left representation of @xmath1212 on itself is a representation by bounded linear operators , and this provides us with an operator norm @xmath56 on @xmath1212 with @xmath1249 and @xmath1250 in general . in view of we
always have @xmath1251 being an operator norm , @xmath56 automatically satisfies @xmath1252 we let @xmath505 denote the completion of @xmath1212 with respect to @xmath56 .
this is a commutative banach algebra .
the norm will still be called @xmath56 .
we identify @xmath1212 with its image in @xmath505 .
since @xmath1253 for @xmath1254 , the embedding of @xmath1212 in @xmath505 is faithful .
we define an @xmath1 action on @xmath1214 by @xmath1255 or in abbreviated form @xmath1256(t ) f_{{\bf{k}}}$ ] , for all @xmath77 .
we note that @xmath1257 acts as a unitary transformation relative to the inner product and we obtain a unitary representation @xmath67 of @xmath1 on @xmath1212 . each @xmath1257 is an automorphism of @xmath505 as a banach algebra . at this point
we can apply gelfand theory to produce the space @xmath1258 . which is the set of algebra homomorphisms @xmath1259 .
then @xmath18 is non - empty and compact in the gelfand topology @xcite .
this is the weakest topology on @xmath18 making continuous the evaluation mappings @xmath1260 for each @xmath1261 .
this means that the open sets are generated from the open sets @xmath1262 where @xmath1263 , @xmath561 open in @xmath368 .
this is the way in which @xmath505 may be viewed as a space of continuous mappings on @xmath18 , namely @xmath1264 , and this is the point of view that we take from now on .
one has for all @xmath1265 , @xmath1261 , @xmath1266 ( see @xcite , ch .
18 , prop .
1 ) from which for the sup norm we have @xmath1267 . for all tuples
@xmath1268 from @xmath261 and for all @xmath1265 , @xmath1269 since @xmath1257 acts as an algebra automorphism of @xmath505 , @xmath1270 is another homomorphism of @xmath505 into @xmath368 , and so is in @xmath18 .
so @xmath1 acts on @xmath18 .
since @xmath1257 just permutes the defining open sets of the topology of @xmath18 , this action is continuous .
[ continuity ] for each @xmath1271 the mapping @xmath1272 is uniformly continuous . [ bicontinuity ] the mapping @xmath1273 defined by the action of @xmath1 on @xmath18 is continuous .
@xmath1274 is a topological dynamical system . since @xmath1 acts on @xmath18
, it acts naturally on @xmath1275 too .
@xmath1212 already has an action of @xmath1 and acquires another one when treated as a subalgbra of @xmath1275 ; however , not surprisingly , the two actions of @xmath1 on @xmath1212 the same .
@xmath1212 is dense in the space @xmath1275 of all continuous functions under the sup - norm .
the next objective is to create an ergodic invariant measure for @xmath1274 .
we define a linear functional @xmath1276 on @xmath1212 by linear extension of @xmath1277 = 0$}\,;\\ 0 & \text{otherwise } \end{cases}\ ] ] for all tuples @xmath1268 from @xmath592 .
note that this is well - defined the relations amongst the @xmath1247 for equal values of @xmath1202 $ ] are consistent with the definition and one can see immediately that @xmath19 is @xmath1-invariant since it picks up only the @xmath94-eigenspace of @xmath1212 .
we need to extend this to a linear functional defined on all of @xmath1275 .
let @xmath286 be a van hove sequence for @xmath1 .
for each @xmath1278 define a new function @xmath140 on @xmath18 by @xmath1279 where integration is with some prefixed haar measure on @xmath1 and @xmath1280 is the value of this measure on @xmath1281 .
[ mean1 ] for all @xmath1282 , @xmath140 is a constant function .
moreover , @xmath1283 = 0 $ } \,;\\ 0 \ & \text{otherwise } \ , .
\end{cases}\ ] ] this shows that @xmath19 is the same as the mean @xmath8 on @xmath1212 .
but , as we now see , @xmath8 is defined on all of @xmath1275 and is a bounded linear operator on @xmath1275 : [ mean2 ] for all @xmath1284 , @xmath1285 .
for all @xmath1284 , @xmath1286 is a constant function . in this way
, @xmath8 produces a measure on @xmath18 , which coincides on @xmath1212 with @xmath19 defined above .
we also call this measure @xmath19 and note the fact that we have established , namely for all @xmath1287 , and for all @xmath1265 , @xmath1288 which is a form of ergodic theorem .
[ cc ] for each @xmath1215 , the space of @xmath1289-eigenfuctions in @xmath1275 ( for the action of @xmath67 on @xmath18 ) is one dimensional and it is spanned by any of the functions @xmath1247 for which @xmath1202 = \kappa$ ] .
in particular the @xmath94-eigenfunctions are constant functions .
the complex conjugate of @xmath1247 as a function on @xmath18 is @xmath1290 .
[ mu = measure ] @xmath19 is @xmath1-invariant probability measure on @xmath18 and @xmath1291 is a pure point ergodic dynamical system .
we now wish to define the ergodic spatial process @xmath23 .
to do this we first define the diffraction - to - dynamics map . for each @xmath275 ,
let @xmath1292 be the function which is @xmath268 at @xmath270 and @xmath94 everywhere else .
since the measure @xmath30 is only non - zero on the points of @xmath261 , it follows that every function in @xmath212 can be represented in the form @xmath1293 the diffraction - to - dynamics embedding is the mapping @xmath1294 comparing these two @xmath126-spaces we have @xmath1295 which shows that @xmath142 is an isometric embedding .
define an action @xmath196 of @xmath1 on @xmath212 by setting , for each @xmath1296 , @xmath1297 to be the function @xmath1298 under this action , @xmath1299 is a @xmath270-eigenfunction for the action and @xmath142 becomes @xmath1-equivariant . since @xmath30 is a positive
, translation bounded , and backward transformable measure on @xmath4 there is a unique positive definite measure @xmath141 on @xmath1 for which @xmath1300 .
this measure is also translation bounded ( and hence automatically transformable ( with fourier transform equal to @xmath30 ) , @xcite ) .
the basic relationship between @xmath141 and @xmath30 is expressed by for all @xmath248 , see @xcite , ch .
1.4 , which in particular gives @xmath1301 .
direct calculation shows that for @xmath27 , @xmath1302 .
we define the process by @xmath1303 also @xmath23 is * real * , linear , continuous , real , and stationary ( @xmath1-equivariant ) .
it is also full .
[ stochasticprocess ] @xmath115 is a pure point ergodic spatial process with diffraction @xmath30 and associated phase form @xmath34 .
thus , based on the positive measure @xmath30 on @xmath4 and a phase form @xmath34 we have constructed a pure point spacial process on @xmath1 .
in this section we collect together various small items that are of interest .
a spatial stationary process @xmath115 is said to have an @xmath616th moment if @xmath1304 belongs to @xmath1305 for any @xmath1306 . in this case , the @xmath616th moment is defined to be the unique linear map @xmath1307 on @xmath1308 ( @xmath616 factors ) with @xmath1309 it it not hard to see that the measure @xmath19 determined by its moments @xmath1307 , @xmath1310 , if all these moments exist and the process if full .
in fact , this is more or less the definition of fullness . in order to have a meaningful diffraction theory
our running assumption is that the second moment exists and is a measure .
then , also the first moment is a measure ( as can easily be seen ) .
thus , let @xmath738 be an bounded , ergodic full stationary process with pure point spectrum and elementary phase form @xmath436 .
then , the @xmath616th moment of @xmath19 and the @xmath616-moment of @xmath436 uniquely determine each other .
more precisely , the following holds in this situation .
\(b ) for each @xmath1316 , let @xmath1317 be a sequence in @xmath155 whose fourier transforms converge to @xmath1318 in @xmath1319 . then
, @xmath1320 where the limits are taken in @xmath126 and one after the other . in particular , @xmath1321 let two bounded spatial processes with pure point diffraction and the same diffraction measure @xmath477 be given
. then , these processes have the same @xmath616-moment if and only if their associated phase forms have the same @xmath616 moment . in particular , if furthermore @xmath1322 , then the two processes are isomorphic if and only if their @xmath616 moments agree for @xmath1323 .
if every element of @xmath592 can be expressed as a sum of @xmath616 or less elements of @xmath261 , then @xmath1324 , and so @xmath1325 moments suffice to determine the isomorphism class of a process , see remark [ 2m+1 theorem ] . in this section we look at a pure point stationary process @xmath31 and assume that @xmath18 is a topological space and all eigenfunctions are continuous . here
, continuity means that for each eigenvektor @xmath270 , we have an continuous nontrivial function @xmath1326 with @xmath1327 for all @xmath1328 and @xmath77 . without loss of generality
we can then normalize the @xmath1326 to satisfy @xmath1329 for some fixed @xmath1330 .
these normalized @xmath1326 will then obey @xmath1331 of course , as discussed in section [ stochasticprocessestodiffraction ] , the pure point process comes with a diffraction measure @xmath30 a set of eigenvalues @xmath592 and a phase form @xmath436 .
set @xmath1332 .
then the considerations of section [ stochasticprocessestodiffraction ] provide a torus system @xmath1333 .
this system can be directly calculated as follows : due to the normalization and , @xmath1336 belongs indeed to @xmath704 .
obviously , @xmath1337 is continuous ( as @xmath592 is given the discrete topology ) .
moreover , @xmath8 can easily be seen to map the character @xmath1338 of @xmath704 to the eigenfunction @xmath1326 .
thus , @xmath8 is unitary ( as it maps an orthonormal basis to an orthonormal basis ) .
as @xmath8 is an isometry , @xmath1337 must have dense image .
as it it a @xmath1 map it is then onto .
moreover , a short calculation invoking again shows that @xmath1339 by usual limiting arguments , this gives that @xmath1340 for all @xmath136 .
as @xmath8 maps the character @xmath1338 to the eigenfunction @xmath1326 , its inverse @xmath135 must map the eigenfunction @xmath1326 to the character @xmath270 .
combined with the usual limiting arguments this shows that @xmath135 also satisfies @xmath1341 for bounded @xmath1342 .
so , if we now define the @xmath23-function @xmath1343 on the torus by @xmath1344 we have indeed an isomorphism of processes .
thus , the phase form belonging to @xmath1343 must then be given by @xmath436 as well . by uniqueness then @xmath1345 = [ n_{{\mathbb{t}}}]$ ] .
another issue around continuity is the possible continuity of the mappings @xmath1109 in the topology on @xmath261 induced from @xmath4 . for
these one uses this same topology to make @xmath667 into a topological group .
we do not know if continuity here is physically relevant but we do note that in this case the classification results for elementary phase forms compared to actual phase forms change from @xmath704 to @xmath1 ( see the discussion around and including remark [ continuouscase ] ) .
our underlying understanding of a spatial process @xmath108 is that it represents some sort of ` density ' on the space @xmath1 .
each point @xmath287 represents an instance of the yet unspecified structure @xmath1346 which can be paired with elements of @xmath155 to give thus @xmath40 provides a set of test functions for some sort of distribution @xmath1346 at @xmath287 .
of course @xmath26 is not necessarily defined at all points of @xmath18 ( it is only an @xmath126-function on @xmath18 ) , and even if it is defined it is not clear whether or not @xmath1346 could be interpreted as any particular kind of distribution .
we do note however , that from @xmath1-invariance we have @xmath1348 , so @xmath1349 for all @xmath77 . an important question , and one for which we can not give a satisfactory answer , is when can this ` density ' be interpreted as a measure @xmath1346 on @xmath899 ? in the theory of delone point sets @xmath1350 this actually happens ; typically @xmath18 is the orbit closure of @xmath1350 under the local topology and @xmath23 is the process that arises from the point measures @xmath1351 for all @xmath287 .
thus for @xmath248 , @xmath1352
. we have already seen one way in which we can see from general principles that measures arise in [ compactg ] . but this is unnecessarily restrictive since in many basic situations @xmath30 is not summable .
for example , in
[ g = u(1 ) ] we looked at the diffraction of the point set @xmath866 , when it is modeled on @xmath1042 as @xmath946 .
its diffraction is @xmath1353 on @xmath865 .
this is not @xmath905 with respect to the haar measure of @xmath865 ( counting measure ) .
here @xmath18 can be identified with @xmath1042 ( since it is compact ) and for all @xmath1354 , @xmath1355 is just the measure @xmath1356 .
so in this case there is a suitable measure in spite of the lack of summability .
the exact conditions that @xmath1346 defined by should be a measure at @xmath11 are that @xmath26 actually be defined at @xmath11 for all @xmath27 and that for each compact subset @xmath15 of @xmath1 there is a constant @xmath1357 so that for all @xmath27 with support inside @xmath15 we have @xmath1358 .
in fact this is precisely the definition of measure via the riesz representation theorem . if for some @xmath1359
, @xmath1360 is defined , _ positive _ , and finite for each positive @xmath27 , then the right - hand side of provides a positive linear functional on @xmath40 and so defines a positive measure .
* are there simple process theoretical conditions on pure point measures @xmath30 that allow one to conclude that the processes atttached to @xmath30 can be interpreted as measure valued processes ? *
our homometry theorem shows that the stationary processes with fixed pure point diffraction form a group .
does this have any deeper meaning ?
can we directly realize multiplication / inversion of stationary processes ? *
when there are moments of all orders , these moments characterize the corresponding processes .
there are various scenarios under which not all these moments are necessary to solve the inverse problem .
for instance in @xcite it is seen that real model sets with real internal spaces are recoverable from their second and third moments . in @xcite
it is seen that if @xmath1361 ( @xmath616 summands ) then one needs only the first @xmath1325 moments to solve the inverse problem .
this suggests that solutions to the inverse problem can be organized into hierarchies depending on the number of moments required to specify them .
* the sequence of moments suggests that there is some sort of cohomology theory that would allow one to organize the increasing information that is added as successive moments are included in the picture . such a cohomology theory would be a most welcome addition to this study .
robinson , _ halmos - von neumann theorem for model sets , and almost automorphic dynamical systems _ , in : dynamics , ergodic theory , and geometry , 243272 , math . sci .
, 54 , cambridge univ . press , cambridge , ( 2007 ) .
b. solomyak , _ spectrum of dynamical systems arising from delone sets _ , in : _ quasicrystals and discrete geometry _ , ed .
j. patera , fields institute monographs , vol .
10 , ams , providence , ri ( 1998 ) 265275 . | we consider the construction and classification of some new mathematical objects , called ergodic spatial stationary processes , on locally compact abelian groups , which provide a natural and very general setting for studying diffraction and the famous inverse problems associated with it . in particular we can construct complete families of solutions to the inverse problem from any given pure point measure that is chosen to be the diffraction . in this case these processes can be classified by the dual of the group of relators based on the set of bragg peaks , and this gives a solution to the homometry problem for pure point diffraction .
an ergodic spatial stationary process consists of a measure theoretical dynamical system and a mapping linking it with the ambient space in which diffracting density is supposed to exist . after introducing these processes
we study their general properties and link pure point diffraction to almost periodicity . given a pure point measure we show how to construct from it and a given set of phases a corresponding ergodic spatial stationary process .
in fact we do this in two separate ways , each of which sheds its own light on the nature of the problem .
the first construction can be seen as an elaboration of the halmos
von neumann theorem , lifted from the domain of dynamical systems to that of stationary processes .
the second is a gelfand construction obtained by defining a suitable banach algebra out of the putative eigenfunctions of the desired dynamics . |
mathematical models are now commonly used in the study of growth of cell tissue . for instance , a wide literature is now available on the study of the tumor growth through mathematical modeling and numerical simulations @xcite . in such models , we may distinguish two kinds of description : either they describe the dynamics of cell population density ( see e.g. @xcite ) , or they consider the geometric motion of the tissue through a free boundary problem of hele - shaw type ( see e.g. @xcite ) .
recently the link between both descriptions has been investigated from a mathematical point of view thanks to an incompressible limit @xcite . in this paper , we depart from the simplest cell population model as proposed in @xcite . in this model the dynamics of the cell density is driven by pressure forces and cell multiplication .
more precisely , let us denote by @xmath0 the cell density depending on time @xmath1 and position @xmath2 , and by @xmath3 the mechanical pressure .
the mechanical pressure depends only on the cell density and is given by a state law @xmath4 .
cell proliferation is modelled by a pressure - limited growth function denoted @xmath5 .
mechanical pressure generates cells displacement with a velocity whose field @xmath6 is computed thanks to the darcy s law .
after normalizing all coefficients , the model reads @xmath7 the choice @xmath8 has been taken in @xcite .
this choice allows to recover the well - known porous medium equation for which a lot of nice mathematical properties are now well - established ( see e.g. @xcite ) .
the incompressible limit is then obtained by letting @xmath9 going to @xmath10 .
however , this state law does not prevent cells to overlap .
in fact , it is not possible with this choice to avoid the cell density to take value above @xmath11 ( which corresponds here to the maximal packing density after normalization ) . a convenient way to avoid cells
overlapping is to consider a pressure law which becomes singular when the cell density approaches @xmath11 .
such type of singularity is encountered , for instance , in the kinetic theory of dense gases where the interaction between molecules is strongly repulsive at very short distance @xcite .
similar singular pressure laws have been also considered in @xcite to model collective motion , in @xcite to model the traffic flow , and in @xcite to model crowd motion ( see also the review article @xcite ) .
then , in order to fit this non - overlapping constraint , we consider the following simple model of pressure law given by @xmath12 finally , the model under study in this paper reads , for @xmath13 , @xmath14 this system is complemented by an initial data denoted @xmath15 .
the aim of this paper is to investigate the incompressible limit of this model , which consists in letting @xmath16 going to @xmath17 in the latter system . at this stage
, it is of great importance to observe that from , we may deduce an equation for the pressure by simply multiplying by @xmath18 and using the relation @xmath19 from , @xmath20 formally , we deduce from that when @xmath21 , we expect to have the relation @xmath22 moreover , passing formally to the limit into , it appears clearly that @xmath23 .
we deduce from this relation that if we introduce the set @xmath24 , then we obtain a free boundary problem of hele - shaw type : on @xmath25 , we have @xmath26 and @xmath27 , whereas @xmath28 on @xmath29 .
thus although the pressure law is different , we expect to recover the same free boundary hele - shaw model as in @xcite .
the incompressible limit of the above cell mechanical model for tumor growth with a pressure law given by @xmath30 has been investigated in @xcite and in @xcite when taking into account active motion of cells . in @xcite , the case with viscosity , where the darcy s law is replaced by the brinkman s law ,
is studied .
we mention also the recent works @xcite where the incompressible limit with more general assumptions on the initial data has been investigated .
however , in all these mentionned works the pressure law do not prevent the non - overlapping of cells .
up to our knowledge , this work is the first attempt to extend the previous result with this constraint , i.e. with a singular pressure law as given by .
the outline of the paper is the following .
in the next section we give the statement of the main result in theorem [ th1 ] , which is the convergence when @xmath16 goes to @xmath17 of the mechanical model towards the hele - shaw free boundary system .
the rest of the paper is devoted to the proof of this result .
first , in section [ sec : estim ] we establish some a priori estimate allowing to obtain space compactness .
then , section [ sec : tcompact ] is devoted to the study of the time compactness .
thanks to compactness results , we can pass to the limit @xmath21 in system in section [ sec : conv ] , up to the extraction of a subsequence .
finally the proof of the complementary relation is performed in section [ sec : compl ] .
the aim of this paper is to establish the incompressible limit @xmath31 of the cell mechanical model with non - overlapping constraint . before stating our main result , we list the set of assumptions that we use on the growth fonction and on the initial data . for the growth function ,
we assume @xmath32 } |g'| = \gamma,\\ & \exists\ , p_m>0 , \quad g(p_m)=0 . \end{aligned } \right.\ ] ] the quantity @xmath33 , for which the growth stops , is commonly called the homeostatic pressure @xcite .
this set of assumptions on the growth function is quite similar to the one in @xcite , except for the bound on the growth term which is needed here due to the singularity in the pressure law . for the initial data ,
we assume that there exists @xmath34 such that for all @xmath35 , @xmath36 notice that this set of assumptions imply that @xmath15 is uniformly bounded in @xmath37 .
we are now in position to state our main result .
[ th1 ] let @xmath38 , @xmath39 .
let @xmath5 and @xmath40 satisfy assumptions and respectively .
after extraction of subsequences , both the density @xmath41 and the pressure @xmath42 converge strongly in @xmath43 as @xmath44 to the limit @xmath45;l^1(\rr^d))\cap bv(q_t)$ ] and @xmath46;h^1(\rr^d))$ ] , which satisfy @xmath47 and @xmath48 moreover , we have the relation @xmath49 and the complementary relation @xmath50 this result extends the one in @xcite to singular pressure laws with non - overlapping constraint .
we notice that we recover the same limit model whose uniqueness has already been stated in ( * ? ? ?
* theorem 2.4 ) .
although our proof follows the idea in @xcite , several technical difficulties must be overcome due to the singularity of the pressure law .
indeed , we first recall that with the choice @xmath30 , equation may be rewritten as the porous medium equation @xmath51 .
a lot of estimates are known and well established for this equation ( see @xcite ) , in particular a semiconvexity estimate is used in @xcite which allows to obtain estimate on the time derivative and thus compactness . with our choice of pressure law , should be consider as a fast diffusion equation .
thus we have first to state a comparison principle to obtain a priori estimates ( see lemma [ lem : estim ] ) . unlike in @xcite
, we may not use a semiconvexity estimate to obtain estimate on the time derivative . to do so
, we use regularizing effects ( see section [ sec : tcompact ] )
. then the convergence proof has to be adapted for these new estimates . solving , the dashed line correspond to the constant value @xmath11 . on the left ,
the pressure law is @xmath52 . on the right ,
the pressure law is @xmath53 with @xmath54.,title="fig : " ] solving , the dashed line correspond to the constant value @xmath11 . on the left ,
the pressure law is @xmath52 . on the right ,
the pressure law is @xmath53 with @xmath54.,title="fig : " ] finally , we illustrate the comparison between the two pressure laws @xmath55 and @xmath56 by a numerical simulation .
we display in figure [ fig1 ] the density computed thanks to a discretization with an upwind scheme of . in figure
[ fig1]-left , the pressure law is @xmath57 as in with @xmath58 . in figure [ fig1]-right ,
the pressure law is @xmath53 with @xmath54 .
we take @xmath59 as growth function ( which satisfies obviously assumption with @xmath60 ) .
the dashed lines in these plots correspond to the constant value @xmath11 . as expected , we observe that the density @xmath61 is bounded by @xmath11 in the case of the pressure law @xmath55 whereas it takes values greater than @xmath11 for the pressure law @xmath56 .
this observation illustrates the fact that the choice of the pressure law @xmath56 does not prevent from overlapping .
the following lemma establishes the nonnegativity of the density .
[ lem : nonneg ] let @xmath62 be a solution to such that @xmath63 and @xmath64 . then , for all @xmath1 , @xmath65 .
we have the equation @xmath66 we use the stampaccchia method .
we multiply by @xmath67 , then using the notation @xmath68 for the negative part , we get @xmath69 we integrate in space , using assumption , we deduce @xmath70 so , after a time integration @xmath71 with the initial condition @xmath72 , we deduce @xmath73 . in order to use compactness results ,
we need first to find a priori estimates on the pressure and the density .
we first observe that we may rewrite system as , by using , @xmath74 with @xmath75 .
[ lem : estim ] let us assume that and hold .
let @xmath62 be a solution to . then , for all @xmath38 , we have the uniform bounds in @xmath76 , @xmath77;l^1\cap l^\infty(\rr^d ) ) ; \\ & 0\leq p_\epsilon \leq p_m , \qquad 0\leq n_\epsilon \leq \frac{p_m}{p_m+\epsilon } \leq 1.\end{aligned}\ ] ] more generally , we have the * comparison principle * : if @xmath78 , @xmath79 are respectively subsolution and supersolution to , with initial data @xmath15 , @xmath80 as in and satisfying @xmath81 . then for all @xmath82 , @xmath83 . finally , we have that @xmath84
is uniformly bounded in @xmath85,w^{1,1}(\rr^d))$ ] and @xmath86 is uniformly bounded in @xmath87,w^{1,1}(\rr^d))$ ] . * comparison principle .
* let @xmath41 be a subsolution and @xmath88 a supersolution of , we have @xmath89 notice that , since the function @xmath90 is nondecreasing , the sign of @xmath91 is the same as the sign of @xmath92 .
moreover , @xmath93 so for @xmath94 and @xmath95 is the positive part , the so - called kato inequality reads @xmath96 . thus multiplying the latter equation by @xmath97
, we obtain @xmath98 from assumption , we have that @xmath5 is nonincreasing .
thus , since @xmath99 is increasing , we deduce that the last term of the right hand side is nonpositive .
since @xmath5 is uniformly bounded we obtain @xmath100 after an integration over @xmath101 , @xmath102 then , integrating in time , we deduce @xmath103 since we have @xmath104 , we deduce that for all @xmath82 , @xmath105 . * @xmath106 bounds . *
we define @xmath107 , such that @xmath108 , then applying the comparison principle with @xmath109 , we deduce , using also the assumption on the initial data that for all @xmath110 , @xmath111 moreover , since @xmath17 is clearly a subsolution to , we also have by the comparison priniciple @xmath112 . since @xmath113 , we have @xmath114 which implies @xmath115 * @xmath116 bound of @xmath117 . * by nonnegativity , after a simple integration in space of equation , we deduce @xmath118 where we use . integrating in time
give the @xmath116 bound , @xmath119 then , using @xmath120 by , we get from the bound @xmath121 , which has been proved above , @xmath122 * estimates on the @xmath123 derivative . * we derive equation with respect to @xmath124 for @xmath125 , @xmath126 multiplying by sign@xmath127 , we get @xmath128 we can remark that @xmath129 , so , by the same token as above , we have @xmath130 moreover , @xmath131 , thus @xmath132 . by assumption , we know that @xmath133 we deduce @xmath134 after an integration in time and space , @xmath135 this latter inequality provides us with a uniform bound on the space derivative of @xmath78 in @xmath116 .
then @xmath136 we split the integral in two : either @xmath137 and then @xmath138 ; or @xmath139 .
@xmath140 where we have used the estimate for the last inequality . then , integrating in time , we deduce , using again the estimate @xmath141 it concludes the proof .
the following lemma proves that assuming that the initial data is compactly supported , then the pressure is compactly supported for any time with a control of the growth of the support .
[ lem : supp ] under the same assumptions as in theorem [ th1 ] , we have that @xmath142 with @xmath143 , where @xmath144 is the ball of center @xmath17 and radius @xmath145 .
using the equation on @xmath42 , @xmath146 let us introduce for @xmath147 , @xmath148 with @xmath149 .
then @xmath150 is compactly supported in @xmath151 with @xmath152 we have @xmath153 then , for all @xmath154 $ ] , @xmath155 in other words , @xmath150 is a supersolution for the equation for the pressure .
let us show that it implies that @xmath156 .
we define @xmath157 . we know that @xmath158 then , on the one hand , multiplying with by @xmath159 we get @xmath160 on the other hand , from , @xmath161 by the comparison principle ( see lemma [ lem : estim ] )
, we have @xmath162 thus , for all @xmath154 $ ] , @xmath163 and @xmath164 is compactly supported in @xmath151 provided we choose @xmath165 large enough such that @xmath166 , which can be done thanks to our assumption on the initial data . since @xmath167 is uniformly bounded in @xmath168 , we may iterate the process on @xmath169 $ ] .
after several iterations , we reach the time @xmath170 and prove the result on @xmath171 $ ] . in the following lemma , we state a uniform @xmath172 estimate on the gradient of the pressure .
[ lem : l2dp ] under the same assumptions as in theorem [ th1 ] , we have a uniform bound on @xmath174 in @xmath175 . for a given function @xmath176 we have , multiplying by @xmath177 , @xmath178
let @xmath179 be an antiderivative of @xmath176 , we have thanks to an integration by parts @xmath180 we choose @xmath176 such as @xmath181 , i.e. @xmath182 . after straightforward computations , we find @xmath183 and @xmath184 .
it gives @xmath185 we integrate in time , using also the expression of @xmath167 in , @xmath186 then , to have a bound on the @xmath172-norm of @xmath174 , it suffices to prove a uniform control on @xmath187 .
we have @xmath188 the second term of the right hand side is small when @xmath16 is small thanks to the @xmath116 bound on @xmath78 , thus it is uniformly bounded .
using the expression of @xmath167 in , we get @xmath189 then , since @xmath190 and since @xmath191 is uniformly bounded on @xmath192 $ ] , we get @xmath193 we conclude thanks to lemma [ lem : supp ] , which provides a uniform control on the support of @xmath167 .
as already noticed in @xcite , regularizing effects , similar to the ones observed for the heat equation @xcite , allow to deduce estimates on the time derivatives .
[ lem : regul ] under the assumptions and , the weak solution @xmath194 satisfies the estimates @xmath195 for a large enough ( independent of @xmath16 ) constant @xmath196 .
let us denote @xmath197 , the equation on the pressure reads @xmath198 the proof is divided into several steps .
we first find a lower bound for @xmath199 by using the comparison principle .
then we deduce estimates on the density and on the pressure . +
_ 1st step .
_ thanks to , we deduce an equation satisfied by @xmath199 . on the one hand , by multiplying by @xmath200 , we deduce , since @xmath5 is decreasing from @xmath201 on the other hand , we have @xmath202 thus , with , we deduce that @xmath203 satisfies @xmath204 by definition of @xmath199 , we have @xmath205 .
thus we deduce that @xmath206 where we have used the notation @xmath207 following an idea of @xcite which has been generalized in @xcite , we introduce the function @xmath208 where the function @xmath209 will be defined later such that @xmath210 is a subsolution for .
we compute @xmath211 using again equation , we have @xmath212 by definition of @xmath213 in , we deduce with , @xmath214 we may rearrange it into @xmath215 let us choose @xmath216 where @xmath217 is chosen large enough ( independent of @xmath16 ) such that @xmath218 thanks to this choice , we have @xmath219 and @xmath220 finally , we obtain from @xmath221 where we use the fact that by definition we have @xmath222 ( recalling also that @xmath5 is decreasing by assumption ) .
thus , by the sub- and super - solution technique , we deduce , using also that @xmath223 + _ 2nd step . _ using again equation , we get from @xmath224 which is the first inequality of lemma [ lem : regul ] .
finally , by definition , we have also @xmath225 .
thus @xmath226 where we use the definition for the last identity .
we conclude easily the proof .
thanks to this latter lemma , we may deduce uniform estimates on the time derivative of @xmath78 and @xmath167 .
[ estimdtn ] for any @xmath227 , we have that @xmath228 is uniformly bounded in @xmath229;l^1(\rr^d))$ ] and @xmath230 is uniformly bounded in @xmath231\times\rr^d)$ ] .
we use the equality @xmath232 , where we recall that @xmath233 denotes the negative part .
thus @xmath234 where we have used equation to bound the first term and lemma [ lem : regul ] for the second term . by the same token , we have @xmath235\times\rr^d ) } & = \int_\tau^t \frac{d}{dt } \int_{\rr^d } p_\epsilon \,dx + 2 \int_\tau^t\int_{\rr^d } |\partial_t p_\epsilon|_- \,dx \\ & \leq \|p_\epsilon(t)\|_{l^1(\rr^d ) } + \|p_\epsilon\|_{l^\infty([\tau , t];l^1(\rr^d ) ) } 2 \kappa \ln(t/\tau).\end{aligned}\ ] ] we conclude the proof thanks to the estimates on @xmath78 and @xmath167 in @xmath236 obtained in lemma [ lem : estim ] .
this section is devoted to the proof of theorem [ th1 ] apart from the complementary relation which is postponed to the next section .
since the sequences @xmath237 and @xmath238 are bounded in @xmath239 , due to lemma [ lem : estim ] and [ estimdtn ] , we may apply the helly theorem and recover strong convergence in @xmath240 , up to an extraction . if we want to extend this local convergence to a global convergence in @xmath43 we need to prove that we can control the mass in an initial strip and in the tail . indeed , let @xmath241 , @xmath242 , @xmath243 @xmath244 since we have strong convergence of @xmath41 in @xmath240 , @xmath245 then we have to control the two other terms in the right hand side .
the control of the initial strip comes from the @xmath116 estimate of @xmath61 , @xmath246 for the control of the tail we consider @xmath247 such that @xmath248 , @xmath249 for @xmath250 and @xmath251 for @xmath252 .
we define @xmath253 .
then @xmath254 where the notation @xmath165 stand for a generic nonnegative constant . moreover ,
using equation , we deduce @xmath255 then , integrating on @xmath171 $ ] , we get @xmath256 by assumption , since the initial data is uniformly compactly supported , we deduce that the right hand side tends to @xmath17 as @xmath257 goes to @xmath10 and @xmath258 goes to @xmath17 .
then @xmath259 is a cauchy sequence in @xmath43 .
it implies its convergence in @xmath43 .
the convergence of the pressure follows from the same kind of computation .
the only difference is for the control of the tail and which is directly given by the estimate @xmath260 therefore , we can extract subsequences and pass to the limit in the equation @xmath261 which implies @xmath262 this is the relation .
we can also pass to the limit in the uniform estimate of lemma [ lem : estim ] which provides and @xmath263 .
[ [ limit - model . ] ] limit model .
+ + + + + + + + + + + + we first recall that from , we have @xmath264 we get , @xmath265 thus , the term in the laplacien converges strongly to @xmath266 as @xmath16 goes to @xmath17 .
then , thanks to the strong convergence of @xmath78 and @xmath167 , we deduce that in the sense of distribution @xmath267 satisfies .
moreover , due to the uniform estimate on @xmath173 in @xmath175 of lemma [ lem : l2dp ] , we can show , by passing into the limit in a product of a weak - strong convergence , that in the sense of distribution @xmath267 satisfies .
[ [ time - continuity . ] ] time continuity .
+ + + + + + + + + + + + + + + + let us define @xmath268 , @xmath269 .
for a given @xmath270 , we consider a smooth function @xmath271 on @xmath101 such that @xmath272 , @xmath273 for @xmath250 and @xmath274 for @xmath252 .
we have @xmath275 we have @xmath276 with @xmath277 a function which is zero on @xmath278 . thus , as for the control of the tail , for @xmath257 large enough , we have , uniformly for @xmath268 , @xmath279 in addition , we know from lemma [ lem : regul ] ( and the @xmath168 bound on @xmath280 ) that @xmath281 , so @xmath282 .
then , since @xmath283 , @xmath284 then , using equation and an integration by parts , we obtain @xmath285 then we can choose @xmath286 close enough such that @xmath287 we conclude that @xmath288 .
[ [ initial - trace ] ] initial trace + + + + + + + + + + + + + for any test function @xmath289 , we have from , @xmath290 letting @xmath16 going to @xmath17 , we obtain with , @xmath291 letting @xmath292 we can conclude that @xmath293 .
in this section we prove the complementary relation @xmath294 in the weak sense , this identity reads , for any test function @xmath295 , @xmath296 the proof is divided into two steps .
+ _ 1st step .
_ in this first step we prove the inequality @xmath297 in .
we start with the pressure equation that we multiply by @xmath298 @xmath299 we multiply by a test function @xmath300 and integrate , @xmath301 where we use an integration by parts for the last identity . from the estimates in lemma
[ lem : estim ] , we have @xmath302 we deduce that for any test function @xmath300 , @xmath303 since we have strong convergence of @xmath304 and weak convergence of @xmath305 , we can pass into the limit in the last two term in , @xmath306 now we are looking for the limit of the first term in .
we have @xmath307 . by weak convergence of @xmath308 and with jensen inequality ( since @xmath309 is convex ) , @xmath310 thus , we conclude from that @xmath311 which is a first inequality for . + _
_ now we want to show the reverse inequality , i.e. @xmath312 we know that @xmath313 with @xmath314 thanks to the inequality @xmath315 , and the strong convergence @xmath316 , we know that @xmath317 as @xmath21 . because @xmath318 we deduce from lemma [ lem : estim ] that @xmath319;l^{1}(\rr^d ) ) $ ] .
it gives us compactness in space but not in time .
thus , following the idea of @xcite , we use a regularization process la steklov. let introduce a time regularizing kernel @xmath320 such that @xmath321 . then with the notations @xmath322 , @xmath323 , where the convolution holds only in the time variable , @xmath324 we denote @xmath325 , then @xmath326 since @xmath327 and @xmath328 are uniformly bounded in @xmath329 from lemma [ lem : estim ] ,
@xmath330 is bounded in @xmath329 and we can extract a converging subsequence , still denoted @xmath331 , converging towards @xmath332 in @xmath333 for @xmath334 fixed . moreover @xmath335 for the second term , we have @xmath344 and @xmath345 .
let @xmath346 and @xmath227 the smallest time in its support , we then have for @xmath347 @xmath348 so integrating on @xmath349 @xmath350 then @xmath351 where we use the bound on @xmath352 in lemma [ estimdtn ] .
for the third term , since @xmath353 , for any test function @xmath354 as above , @xmath355 dx dt \\ & \qquad = \underset{\eta\to 0}{o}(1 ) .
\end{aligned}\ ] ] so for all test function @xmath354 as above , and all @xmath269 , @xmath356 now it remain to pass to the limit @xmath357 in the regularization process . thanks to an integration by parts , @xmath358 from the @xmath172 estimate on @xmath359 ( lemma [ lem : l2dp ] ) and the @xmath236 estimate on @xmath266 ( lemma [ lem : estim ] ) , we deduce that we can pass to the limit @xmath360 and get @xmath361 p. ciarletta , l. foret , m. ben amar , _ the radial growth phase of malignant melanoma : multiphase modelling , numerical simulations and linear stability analysis _ ,
j. r. soc .
interface * 8 * ( 2011 ) , 345368 . j. s. lowengrub , h. b. frieboes , f. jin , y .- l . chuang , x. li , p. macklin , s. m. wise , v. cristini , _ nonlinear modelling of cancer : bridging the gap between cells and tumours _ , nonlinearity ( 2010 ) * 23 * ( 1 ) , r1r91 . c. perrin , e. zatorska , _
free / congested two - phase model from weak solutions to multi - dimensional compressible navier - stokes equations _ , communications in partial differential equations , ( 2015 ) 40:8 , 15581589 | a mathematical model for tissue growth is considered .
this model describes the dynamics of the density of cells due to pressure forces and proliferation .
it is known that such cell population model converges at the incompressible limit towards a hele - shaw type free boundary problem .
the novelty of this work is to impose a non - overlapping constraint .
this constraint is important to be satisfied in many applications .
one way to guarantee this non - overlapping constraint is to choose a singular pressure law .
the aim of this paper is to prove that , although the pressure law has a singularity , the incompressible limit leads to the same hele - shaw free boundary problem .
nonlinear parabolic equation ; incompressible limit ; free boundary problem ; tissue growth modelling . 35k55 ; 76d27 ; 92c50 . |
the usual test of whether a deterministic dynamical system is chaotic or nonchaotic is the calculation of the largest lyapunov exponent @xmath12 .
a positive largest lyapunov exponent indicates chaos : if @xmath13 , then nearby trajectories separate exponentially and if @xmath14 , then nearby trajectories stay close to each other .
this approach has been widely used for dynamical systems whose equations are known ( abarbanel _ et al .
_ 1993 ; eckmann _ et al .
_ 1986 ; parker & chua 1989 ) .
if the equations are not known or one wishes to examine experimental data , this approach is not directly applicable .
however lyapunov exponents may be estimated ( wolf _ et al . _ 1985 ; sana & sawada 1985 ; eckmann _ et al .
_ 1986 ; abarbanel _ et al .
_ 1993 ) by using the embedding theory of takens ( 1981 ) or by approximating the linearisation of the evolution operator .
nevertheless , the computation of lyapunov exponents is greatly facilitated if the underlying equations are known and are low - dimensional . in this article
, we propose a new @xmath0@xmath1 test for chaos which does not rely on knowing the underlying equations , and for which the dimension of the equations is irrelevant .
the input is the time series data and the output is either a @xmath0 or a @xmath1 depending on whether the dynamics is nonchaotic or chaotic . since our method
is applied directly to the time series data , the only difference in difficulty between analysing a system of partial differential equations or a low - dimensional system of ordinary differential equations is the effort required to generate sufficient data .
( as with all approaches , our method is impracticable if there are extremely long transients or once the dimension of the attractor becomes too large . ) with experimental data , there is the additional effect of noise to be taken into consideration .
we briefly discuss this important issue at at the end of this paper
. however , our aim in this paper is to present our findings in the situation of noise - free deterministic data .
to describe the new test for chaos , we concentrate on the continuous time case and denote a solution of the underlying system by @xmath4 .
the discrete time case is handled analogously with the obvious modifications .
consider an observable @xmath15 of the underlying dynamics .
the method is essentially independent of the actual form of @xmath3 almost any choice of @xmath3 will suffice . for example if @xmath16 then @xmath17 is a possible and simple choice .
choose @xmath6 arbitrarily and define @xmath18 ( throughout the examples in [ sec : example ] and [ sec : morex ] we fix @xmath19 once and for all . )
we claim that * @xmath8 is bounded if the underlying dynamics is nonchaotic and * @xmath8 behaves asymptotically like a brownian motion if the underlying dynamics is chaotic .
the definition of @xmath20 in , which involves only the observable @xmath21 , highlights the universality of the test .
the origin and nature of the data which is fed into the system is irrelevant for the test , and so is the dimension of the underlying dynamics .
later on , we briefly explain the justification behind the claims ( i ) and ( ii ) . for the moment
, we suppose that the claims are correct and show how to proceed .
to characterise the growth of the function @xmath8 defined in , it is natural to look at the mean square displacement ( msd ) of @xmath8 , defined to be @xmath22 if the behaviour of @xmath8 is brownian , i.e. the underlying dynamics is chaotic , then @xmath23 grows linearly in time ; if the behaviour is bounded , i.e. the underlying dynamics is nonchaotic , then @xmath23 is bounded .
we define @xmath9 as the asymptotic growth rate of the msd .
the growth rate @xmath24 can be numerically determined by means of linear regression of @xmath25 versus @xmath26 ( press _ et al .
( to avoid negative logarithms we actually calculate @xmath27 which obviously does not change the slope @xmath24 . )
this allows for a clear distinction of a nonchaotic and a chaotic system as @xmath24 may only take values @xmath10 or @xmath11 .
we have lost though the possibility of quantifying the chaos by the magnitude of the largest lyapunov exponent @xmath12 .
numerically one has to make sure that initial transients have died out so that the trajectories are on ( or close to ) the attractor at time zero , and that the integration time @xmath28 is long enough to ensure @xmath29 .
we illustrate the @xmath0@xmath1 test for chaos with the help of a concrete example , the forced van der pol system , @xmath30 which has been widely studied in nonlinear dynamics ( van der pol 1927 ; guckenheimer & holmes 1990 ) . for fixed @xmath31 and @xmath32
, the dynamics may be chaotic or nonchaotic depending on the parameter @xmath33 .
following parlitz & lauterborn ( 1987 ) , we take @xmath34 and let @xmath33 vary from @xmath35 to @xmath36 in increments of @xmath37
. choose @xmath38 and @xmath19 .
we stress that the results are independent of these choices for all practical purposes . as described below in [ sec : just ] , almost all choices will work .
( deliberately poor choices such as @xmath39 , or @xmath40 or @xmath41 obviously fail ; sensible choices that fail are virtually impossible to find . ) in figure [ fig : skew ] we show a plot of @xmath24 versus @xmath33 .
the periodic windows are clearly seen . as a comparison
we show in figure [ fig : lyap ] the largest lyapunov exponent @xmath12 versus @xmath33 . since the onset of chaos does not occur until after @xmath42 we display the results only for the range @xmath43 in figures [ fig : skew ] and [ fig : lyap ] .
( both methods easily indicate regular dynamics for @xmath44 . )
naturally we do not obtain the values @xmath10 and @xmath11 exactly the mathematical results that underpin our method predict these values in the limit of infinite integration time .
( the same caveat applies equally to the lyapunov exponent method . ) in producing the data for figures [ fig : skew ] and [ fig : lyap ] , we allowed for a transient of 200,000 units of time and then integrated up to time @xmath45 .
as can be seen in figure [ fig : cutoff ] , for most of the @xmath46 data points in the range of @xmath33 , we obtain @xmath47 or @xmath48 .
next , we carry out the @xmath0@xmath1 test for the forced van der pol system in the situation of a more limited quantity of data .
the results are shown in figure [ fig : skew4 ] for @xmath49 .
we again allow for a transient time 200,000 but then integrate only for @xmath50 .
the transitions between chaotic dynamics and periodic windows are almost as clear with @xmath50 as they are with @xmath45 even though the convergence of @xmath24 to @xmath0 or @xmath1 is better with @xmath45 .
to test the method on high - dimensional systems we investigated the driven and damped kortweg - de vries ( kdv ) equation ( kawahara & toh 1988 ) @xmath51 on the interval @xmath52 $ ] with periodic boundary conditions .
this partial differential equation has non - chaotic solutions if the dispersion @xmath53 is large and exhibits spatio - temporal chaos for sufficiently small @xmath53 .
note that equation reduces to the kdv equation when @xmath54 , and reduces to the kuramoto - sivashinksy equation when @xmath55 .
we fix @xmath56 , @xmath57 and vary @xmath53 . for @xmath55
, it is expected that the dynamics of the kuramoto - sivashinksy equation are chaotic for these parameter values . as an observable we used @xmath58 where @xmath59 is an arbitrarily fixed position , and we iterated until time @xmath60 .
the @xmath0@xmath1 test confirms that the dynamics is chaotic at @xmath55 ( with @xmath61 ) .
also , we obtain @xmath62 at @xmath63 and @xmath64 at @xmath65 , indicating chaotic and regular dynamics respectively at these two parameter values .
finally , for discrete dynamical systems , we tried out the test with an ecological model whose chaotic component is coupled with strong periodicity ( brahona & poon 1996 ; cazelles & ferriere 1992 ) .
the model @xmath66 has a non - connected chaotic attractor consisting of seven connected components .
our test yields @xmath67 with only @xmath68 data points and clearly shows that the dynamics is chaotic .
the function @xmath8 can be viewed as a component of the solution to the skew product system @xmath69 driven by the dynamics of the observable @xmath70 . here
@xmath71 represent coordinates on the euclidean group @xmath72 of rotations @xmath73 and translations @xmath74 in the plane .
we note that inspection of the dynamics of the @xmath74-trajectories of the group extension provides very quickly ( for small @xmath28 ) a simple visual test of whether the underlying dynamics is chaotic or nonchaotic as can be seen from figure [ fig : dynam ] .
( a ) ( b ) in nicol _ et al . _
( 2001 ) it has been shown that typically the dynamics on the group extension is sublinear and furthermore that typically the dynamics is bounded if the underlying dynamics is nonchaotic and unbounded ( but sublinear ) if the underlying dynamics is chaotic .
moreover , the @xmath20 and @xmath75 components each behave like a brownian motion on the line if the chaotic attractor is uniformly hyperbolic ( field _ et al .
a nondegeneracy result of nicol _ et al . _
( 2001 ) ensures that the variance of the brownian motion is nonzero for almost all choices of observable @xmath3 .
recent work of melbourne & nicol ( 2002 ) indicates that these statements remain valid for large classes of nonuniformly hyperbolic systems , such as hnon - like attractors .
there is a weaker condition on the ` chaoticity ' of @xmath76 that guarantees the desired growth rate @xmath11 for the mean square displacement : namely that the autocorrelation function of @xmath77 decays at a rate that is better than quadratic .
more precisely , let @xmath4 and @xmath78 denote solutions to the skew product equations with initial conditions @xmath79 and @xmath80 respectively .
if there are constants @xmath81 and @xmath82 such that @xmath83 for all @xmath84 , then @xmath11 as desired ( biktashev & holden 1998 ; ashwin _ et al .
_ 2001 ; field _ et al .
( again , results of nicol _ et al . _
( 2001 ) ; field _ et al . _
( 2002 ) ; and melbourne & nicol ( 2002 ) ensure that the appropriate nondegeneracy condition holds for almost all choices of @xmath3 . )
there is a vast literature on proving decay of correlations ( baladi 1999 ) and this has been generalised to the equivariant setting for discrete time by field _
( 2002 ) and melbourne & nicol ( 2002 ) and for continuous time by melbourne & trk ( 2002 ) .
it follows from these references that @xmath11 , for large classes of chaotic dynamical systems .
one might ask why it is not better to work , instead of the @xmath72-extension , with the simpler @xmath85-extension @xmath86 in principle , @xmath8 can again be used as a detector for chaos .
however , by the ergodic theorem @xmath8 will typically grow linearly with rate equal to the space average of @xmath3 .
this would lead to @xmath87 irrespective of whether the dynamics is regular or chaotic .
hence , it is necessary to subtract off the linear term of @xmath8 in order to observe the bounded / diffusive growth that distinguishes between regular / chaotic dynamics .
subtracting this linear term is a highly nontrivial numerical obstruction .
the inclusion of the @xmath73 variable in the definition of @xmath8 and in the skew product kills off the linear term .
we have established a simple , inexpensive , and novel @xmath0@xmath1 test for chaos .
the computational effort is of low cost , both in terms of programming efforts and in terms of actual computation time .
the test is a binary test in the sense that it can only distinguish between nonchaotic and chaotic dynamics .
this distinction is extremely clear by means of the diagnostic variable @xmath24 which has values either close to @xmath0 or close to @xmath1 .
the most powerful aspect of our method is that it is independent of the nature of the vector field ( or data ) under consideration .
in particular the equations of the underlying dynamical system do not need to be known , and there is no practical restriction on the dimension of the underlying vector field .
in addition , our method applies to the observable @xmath70 rather than the full trajectory @xmath4 .
related ideas ( though not with the aim to detect chaos ) have been used for pde s in the context of demonstrating hypermeander of spirals in excitable media ( biktashev & holden 1998 ) where the spiral tip appears to undergo a planar brownian motion ( see also ashwin _ et al .
we note also the work of coullet & emilsson ( 1996 ) who studied the dynamics of ising walls on a line , where the motion is the superposition of a linear drift and brownian motion .
( this is an example of an @xmath85-extension which we mentioned briefly in [ sec : just ] . as we pointed out then
, the linear drift is an obstruction to using an @xmath85-extension to detect chaos . ) from a purely computational point of view , the method presented here has a number of advantages over the conventional methods using lyapunov exponents . at a more technical level , we note that the computation of lyapunov exponents can be thought of abstractly as the study of the @xmath88-extension @xmath89 where @xmath88 is the space of @xmath90 matrices , @xmath91 , and @xmath92 is the size of the system .
note that the extension involves @xmath93 additional equations and is defined using the linearisation of the dynamical system . to compute the dominant exponent , it is still necessary to add @xmath92 additional equations , again governed by the linearised system .
in contrast , our method requires the addition of two equations . in this paper , we have concentrated primarily on the idealised situation where there is an in principle unlimited amount of noise - free data .
however , in
[ sec : example ] we also demonstrated the effectiveness of our method for limited data sets . an issue which will be pursued in further work is the effect of noise which is inevitably present in all experimental time series .
preliminary results show that our test can cope with small amounts of noise without difficulty .
a careful study of this capability , and comparison with other methods , is presently in progress .
we are grateful to philip aston , charlie macaskill and trevor sweeting for helpful discussions and suggestions .
the research of gg was supported in part by the european commission funding for the research training network `` mechanics and symmetry in europe '' ( masie ) . | chaos , deterministic dynamical systems , lyapunov exponents , mean square displacement , euclidean extension we describe a new test for determining whether a given deterministic dynamical system is chaotic or nonchaotic .
in contrast to the usual method of computing the maximal lyapunov exponent , our method is applied directly to the time series data and does not require phase space reconstruction . moreover
, the dimension of the dynamical system and the form of the underlying equations is irrelevant .
the input is the time series data and the output is @xmath0 or @xmath1 depending on whether the dynamics is non - chaotic or chaotic .
the test is universally applicable to any deterministic dynamical system , in particular to ordinary and partial differential equations , and to maps .
our diagnostic is the real valued function @xmath2 where @xmath3 is an observable on the underlying dynamics @xmath4 and @xmath5 .
the constant @xmath6 is fixed arbitrarily .
we define the mean - square - displacement @xmath7 for @xmath8 and set @xmath9 . using recent developments in ergodic theory
, we argue that typically @xmath10 signifying nonchaotic dynamics or @xmath11 signifying chaotic dynamics .
[ firstpage ] |
the soft x - ray telescope ( sxt ; * ? ? ? * ) on board yohkoh @xcite revealed that a soft x - ray emitting plasma ejection , or plasmoid ejection , is sometimes observed in solar flares ( e.g. * ? ? ? * ) .
it was also found that the plasmoids show blob - like or loop - like shapes and that the strong acceleration of the plasmoid ejection occurs during the peak time of the hard x - ray emission @xcite .
their ejection velocities are typically several hundred km s@xmath0 and the ejected plasma is heated to more than 10 mk before the onset of the ejection @xcite .
they often start to rise up gradually a few tens of minutes before the onset of a hard x - ray burst and are then strongly accelerated just before or at the impulsive phase of the flare .
a similar kinetic evolution is also seen in the case of coronal mass ejections ( cmes ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
similarly , slowly drifting radio structures , observed at the beginning of the eruptive solar flares in the 0.6 - 1.5 ghz frequency range , have been interpreted as the radio signatures of plasmoid ejection @xcite .
hudson et al . ( 2001 ) , moreover , identified a rapidly moving hard x - ray source associated with a moving microwave source and an x - ray plasmoid ejection .
kundu et al .
( 2001 ) also identified moving soft x - ray ejecta associated with moving decimetric / metric radio sources observed by the nanay radioheliograph .
sui et al .
( 2004 ) also found a plasmoid ejection in hard x - ray images with rhessi satellite . in the standard model of solar flares , so - called cshkp model @xcite , a filament / plasmoid ejection is included .
however , it does not necessarily stress the importance of the role of plasmoid ejection explicitly .
shibata et al . ( 1995 ) and shibata & tanuma ( 2001 ) extended the cshkp model by unifying reconnection and plasmoid ejection and stressed the importance of the plasmoid ejection in a reconnection process .
the model is called the `` plasmoid - induced - reconnection '' model . in that model ,
the plasmoid inhibits reconnection and stores magnetic energy in a current sheet .
then , once it is ejected , inflow is induced because of the mass conservation , resulting in the enhancement of reconnection rate and the acceleration of the plasmoid due to the faster reconnection outflow .
moreover , reconnection theories predict several plasmoids of various scales are generated .
the dynamics of plasmoid formation in the solar flare and their subsequent plasmoid ejection affect the reconnection rate in the nonlinear evolution .
therefore , plasmoid ejections are observational evidence of magnetic reconnection of solar flares . since plasmoid ejections have been observed in both long duration events and compact flares @xcite , it is shown that the magnetic reconnection model may be applicable even for the compact flares that do not show the other typical features of the magnetic reconnection .
on the basis of the results of magnetohydrodynamic ( mhd ) numerical simulations , kliem et al .
( 2000 ) suggested that each individual burst in the slowly pulsating structure is generated by suprathermal electrons , accelerated in the peak of the electric field in the quasi - periodic and bursting regime of the magnetic field reconnection .
this is the so - called `` impulsive bursty '' reconnection @xcite . in that regime , several plasmoids
can be formed successively as a result of the tearing and coalescence instabilities @xcite .
the repeated formation of magnetic islands can induce magnetic reconnection and their subsequent coalescence @xcite .
these processes even have a cascading form : secondary tearing , tertiary tearing , and so on , always on smaller and smaller spatial scales @xcite .
furthermore the formed plasmoids can merge into larger plasmoids .
tanuma et al .
( 2001 ) also showed that an increase in the velocity of the plasmoid ejection leads to an increase in the reconnection rate , and brta et al .
( 2008 ) analyzed the dynamics of plasmoids formed by the current sheet tearing .
the unsteady reconnection mentioned above can release a large amount of energy in a quasi - periodic way .
the energy released in the upward direction can be observed as plasmoid ejections or coronal mass ejections , while that in the downward direction as downflows @xcite and impulsive bursts at the footpoints of the coronal loops .
bursty energy release in solar flares has been observed as highly time variable hard x - ray bursts and microwave bursts @xcite .
benz & aschwanden ( 1992 ) and aschwanden ( 2002 ) argued that these impulsive bursts suggest the existence of highly fragmented particle acceleration regions .
this fragmented structure of solar flares indicates that a flare is an ensemble of a vast amount of small - scale energy releases and the fractal / turbulent structure of the current sheet can be expected ( see also * ? ? ? * ) .
recently the kinematics of multiple plasmoids have been studied by full - particle simulations and how the particles interact with their surroundings has been explained ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
it is interesting to note that the stochastic acceleration mechanism ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) may be related to particle acceleration in fractal / turbulent current sheet ( see also * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
karlick et al .
( 2004 ) showed a unique series of slowly drifting structures during one flare , from which the authors proposed that it indirectly maps a formation of several plasmoids and their interactions .
however in most of the previous studies , only one plasmoid or one drifting structure was reported during the solar flare . in this paper
, we present for the first time the direct observations of multiple x - ray emitting plasmoid ejections associated with a single solar flare observed by yohkoh / sxt ( firstly reported by takasaki 2006 ) . in section 2 , we describe the multiple plasmoid ejection events .
then we analyzed the data in section 3 by examining in detail the relationship between the multiple plasmoid ejections and the nonthermal hard x - ray emissions using yohkoh data .
finally we discuss the dynamic features of magnetic reconnection and the roll of plasmoid ejections in the particle acceleration in a solar flare in section 4 .
a series of homologous flare - cme events occurred in noaa active region 9236 from 2000 november 24 to november 26 .
these events have been reported by several researchers .
nitta & hudson ( 2001 ) found that the cme - flare events of the homologous flares show quite similar characteristics in both their coronal / photospheric magnetic structures and their cme properties .
zhang & wang ( 2002 ) compared the homologous flares in detail through the use of multiwavelength observations .
wang et al .
( 2002 ) reported that the activities of these flares was driven or triggered by newly emerging magnetic flux , which appeared on the western side of the leading sunspot in this active region .
figure 1a-1c show snapshot images of the preceding sunspot in noaa 9236 and an associated two - ribbon flare , which occurred on 2000 november 24 observed in white light , ultraviolet and soft x - ray emission .
takasaki et al .
( 2004 ) performed a comparison of the physical parameters between the individual flares and from this they could confirm that the plasmoid - induced - reconnection model is reasonable .
they then showed that the interaction between the new emerging magnetic flux loops and the pre - existing magnetic field was essential for producing the homologous flares and plasmoid ejections in the active region .
these ejections were followed by a single halo - cme which occurred at 15:30 ut on 2000 november 24 .
figure 2 shows a cme image observed with the large angle spectroscopic coronagraph ( lasco ; * ? ? ?
* ) that occurred following the flare ( e.g. see * ? ? ?
* ; * ? ? ?
the core of the cme was observed traveling in the northwest direction .
moon et al .
( 2003 ) found a good correlation between cme speed and the goes x - ray peak flux of the associated flares in this series of homologous flare - cme events .
impulsive hard x - ray bursts were also observed in this flare with the hard x - ray telescope ( hxt ; * ? ? ?
* ) on board yohkoh ( see fig .
a pair of hard x - ray sources was located at the footpoints of the coronal arcade .
we used the hard x - ray emission data observed with the h - band ( 52.7 - 92.8 kev ) of hxt , whose temporal resolution was 0.5 s. we focus on a goes x2.3 class flare that occurred in noaa 9236 ( n19@xmath1 , w06@xmath1 ) at 14:51 ut on 2000 november 24 .
this flare was one of the homologous flares which were described previously .
in this flare , we observed seven plasmoid ejections in the soft x - ray images of the flare taken with yohkoh / sxt .
we mainly used the partial frame images with half- and quarter - resolution for the analysis .
the spatial resolutions are about 5 and 10 , respectively .
we used the sandwich ( almg ) filter images which were taken with 20 second cadence .
figure 3 and supplement movie 1 show the temporal evolution of the flare , which is made of full- , half- and quarter - resolution images taken with yohkoh / sxt almg filter ( negative images ) .
full- , half- and quarter - resolution images are different in their spatial resolutions , field of views and exposure times .
full - resolution images are of short exposure time and focus on the brightest region of the active region , such as the two ribbon structure . on the other hand , quarter - resolution images are of longer exposure time so that they are applicable for the detection of large - scale and faint phenomena such as plasmoid ejections . the black vertical line in the middle of figure 3 shows the saturation of the quarter - resolution images .
we identified seven major ejections which we named p1-p7 . in figure 4 ( supplement movie 2 ) ,
figure 5 and figure 6 , we marked each plasmoid ejection with a circle .
each plasmoid ejection can be seen more clearly in quarter - resolution images of figure 5 , while half - resolution images of figure 6 are convenient to see plasmoids just after ejections .
the fields of view of quarter- and half - resolution images are also shown in figure 3 .
figure 7 shows the temporal evolution of soft x - ray emission observed with sxt .
a full resolution image is inset on figure 7a .
figure 7a-7c shows three of the plasmoid ejections denoted as p1 , p4 and p7 . in figure
7d-7f we overlaid contour images of the soft x - ray emission , which show the time evolution of the plasmoid ejection .
the directions of the ejections are indicated by the arrows in the panels . to make clear the traveling of the plasmoids
, we also overlaid the contours of the sxr images taken at different times ( e.g. 15:09:19 ut , 15:09:39 ut and 15:09:59 ut for fig .
7d ) on figures 7d , 7e , and 7f . from these contour images
, we can roughly outline the position and the size of the bright cores of the plasmoids .
we can also measure the velocity @xmath2 by taking the time difference of these contour images .
those size and apparent velocity of the plasmoids are listed in table 1 . here
we note that these are the `` apparent '' velocities , and the motions in the line of sight are ignored . therefore , the actual velocity of the plasmoid @xmath3 will be greater than the plane - of - the - sky value , where @xmath4 representing the expected deprojection over a reasonable range of angles @xmath5 to the plane of the sky .
the trajectory of the plasmoid is not necessarily in the straight lines as shown with the arrows in figure 7 . in order to take into account those non - straight motions , we measured the apparent velocity of each plasmoid by averaging the velocities derived from each time differences . in the following discussions
, we consider the apparent velocity as the actual velocity .
we can see several plasmoids were ejected in the northwest direction , and one plasmoid ejected in the southwest direction in figure 4 ( supplement movie 2 ) and figure 5 , which is marked as p6 in this paper .
each plasmoid has a unique velocity , brightness and size .
the first ejection was that of p1 , which was followed by p2-p4 which were successively ejected as a group in the same direction .
p1 shows blob - like structure , while p2-p4 seem to be a part of an expanding loop .
we are not able to clearly define the structure of p5 and p6 due to the faint emission , although we can surely identify p5 and p6 traveling outward from the active region .
p7 is the brightest ejection of blob - like structure .
it starts to rise up gradually at 15:12 ut and ejected / accelerated upwards at 15:14 ut . in the lasco cme images we can no longer identify the fine structure corresponding to the individual plasmoid ejections , though some complicated structures can be observed ( see fig.2 )
this is probably because the ejected plasmoids merge into a single cme .
the average velocity of the cme listed in the cme event catalog is about 1245 km s@xmath0 .
this apparent velocity of the cme is faster than those of the plasmoid ejections p1-p7 summarized in table 1 .
this observational fact qualitatively suggests that the merged plasmoids are continuously accelerated as they are ejected into interplanetary space , as shown by cheng et al .
( 2003 ) , although we can hardly identify the one - to - one relation between them .
here we focus on the time evolution of the plasmoid ejections .
we used a time slice image of the plasmoid ejections as we show in figure 8 .
the horizontal axis is the time from 15:06 ut to 15:18 ut , and the vertical axis is the 1d image ( negative images ) using a slit line placed along the direction of the several plasmoid ejections ( p1 , p3 , p4 , p5 and p7 ) .
this time slice image is made of half - resolution images .
the position of the slit line is shown in figure 6 and 7a .
those ejections seen in the time slice image are marked with signs @xmath6 , @xmath7 @xmath8 in figure 8b .
we can also identify further additional faint ejections in the time slice image .
some ejections travel along the slit lines , while others travel on a path which is slightly different from the slit line . as a result
, the visibility of each plasmoid ejection is different .
initially p1 , p2 and p3 are slowly accelerated , then strongly accelerated during the initial impulsive phase of the hard x - ray emission ( 15:07:40 - 15:08:40 ut ) followed closely by the ejection of p4 .
a group of plasmoids gradually rise up 15:09:40- 15:12:20 ut followed by the faint ejection p5 .
p6 is ejected in a different direction ( southwest ) and does not cross the slit line , so it does not appear in figures 8a , b .
p7 is the brightest ejection and the most clearly visible in figures 8a , b .
the apparent velocities of the plasmoids along the slit can also be derived from the slopes of the fitted lines in figure 8b .
we note that these are the apparent velocities measured from the time slice images of figure 8b and different from the velocities measured from contour plot images in figure 7 .
this is because the former shows the front velocity of thinner density plasma , while the latter shows the velocity of the thick core part of the plasmoids . in figure 8a and 8b ,
the plasmoid ejections roughly start to rise at the apparent height of approximately 50
( @xmath935 mm ) and propagate into the upper atmosphere .
this probably means that reconnection occurs at around or just below the height of @xmath935 mm . in this paper , we set the start position of plasmoid ejections as the height ( @xmath935 mm ) and define the start times of ejections as the time when each plasmoid crosses the height ( which is shown with dotted line in fig .
it is noted that figure 6 and figure 8a , b are drawn with half - resolution images , but figure 5 is a series of quarter - resolution images .
therefore we see more of the earlier phase of plasmoid ejection in figure 6 and 8 than in figure 5 , and so figure 8 is more appropriate to determine the time of plasmoid ejection .
we confirmed that the start times are comparable to those defined above . in figure 8c , we show the light curve of the hard x - ray emission obtained with the h - band ( 52.7 - 92.8 kev ) of yohkoh / hxt and the goes soft x - ray light curve .
we can distinguish the hard x - ray bursts into three separate periods : the first period , a ( 15:08 - 15:09 ut ) , is the brightest phase of the hard x - ray emission , the second period , b ( 15:09 - 15:11 ut ) , show gradual enhancement and the last period , c ( @xmath915:14 ut ) , is an isolated hard x - ray peak .
the plasmoid ejections of p1-p3 seem to be ejected during the peak time of period a. p1 seems to be ejected just before the hard x - ray peaks in figure 8c .
this is consistent with the report by ohyama & shibata ( 1997 ) who showed that plasmoids are ejected at or just before the hard x - ray peak . during period b ,
a group of plasmoids gradually rise up and are followed by the faint ejection of p4 and p5 , while the brightest plasmoid p7 is ejected during period c. the correspondence of the plasmoid ejections and hard x - ray peaks are shown by the arrows in figure 8b . since the bursts in period a are superposed and very complex , it is difficult to identify the exact correspondence between the plasmoid ejections ( p1-p3 ) and the hard x - ray bursts .
we made a correlation plot of the times of the hard x - ray bursts and those of the plasmoid ejections ( figure 9a ) to make the relation clearer .
both the horizontal and vertical axes show the times ( ut ) . the horizontal ( _ light gray _ ) and the vertical ( _ dark gray _ )
lines illustrate the times of the plasmoid ejections and those of the hard x - ray bursts , respectively .
the thickness of the lines shows an estimation of the error .
figure 9b shows hard x - ray ( 52.7 - 92.8 kev ) light curves obtained with h - band of yohkoh / hxt .
the thin solid line shows the points where the times of the plasmoid ejections correspond to those of the hard x - ray bursts .
these results appear to show that several plasmoid ejections coincide with hard x - ray bursts . here
we note that , although the intense hxr emissions indicate that strong energy releases occur at those times , it does not necessarily mean that certain amount of plasmoids , that is , notable plasmoids are ejected .
we have already noticed that there are many fainter ejections , which also tend to appear at hxr bursts , while some of the hard x - ray intensity fluctuations have no associated ejections in figures 8 and 9 . on the other hand ,
the hxr burst for p5 and p6 does not show a sharp summit but a gentle hump .
this is probably because ejections p5 and p6 are parts of continuous outflows from the active region during this time range .
this may also suggest a milder energy release compared with the others , resulting into a gradual enhancement of hard x - ray emission between 15:09:40 ut and 15:11:00 ut .
we also studied temperature diagnostics on the soft x - ray emitting plasma using the sxt filter ratio method @xcite .
we used the half - resolution images taken with the be and thick al filters of sxt for the analysis of plasmoid ejections .
we subtracted the background photon flux of plasmoids .
the temperature and emission measure are determined using the two filter data of the be and thick al filters .
the size of the plasmoid @xmath10 was measured from the contour plot images in figure 7b , which shows the lower limit of the observable size ( summarized in table 1 ) .
we assume that the x - ray emitting plasma , measured by the filter ratio method , fills the plasmoid with a filling factor of 1 .
we assumed that the volume of the plasmoid is @xmath11 , such that the line of sight width of a plasmoid is equal to the square - root of the size @xmath10 . since the observation times of the be and thick al filters are not exactly the same , we used two images from the thick al filter which were taken just before and just after the be filter observation . the error shown in table 1 mostly results from combining these two images
this error is much greater than the background photon noise ( see * ? ? ? * in more detail ) . as for plasmoid ejections p2 and p3 , there are no be and thick al filter images , because p2 and p3 are out of field of view of half - resolution images as shown in figure 6 . in these cases
we assumed the temperatures of p2 and p3 as @xmath12 k. table 1 summarizes the physical parameters , such as temperature , emission measure , density , mass , thermal energy and kinetic energy of each plasmoid , of the seven plasmoid ejections identified in figure 5 . since the size of the plasmoids that we derived from the images is just the lower limit , the mass , thermal and kinetic energies calculated with the size are also the lower limit . in table 1 , each plasmoid shows a typical temperature of 10@xmath13 k , a density of 10@xmath14 @xmath15 and an apparent velocity of 200 - 1400 km s@xmath0 , which is similar to results of previous studies ( i.e. * ? ? ?
* ; * ? ? ?
the kinetic energy of each plasmoid ejection seems to be comparable to or twice as large as their thermal energy .
we also estimate the total flare energy from the full - resolution images ( spatial resolution 2.5 " ) of the same filters at the peak time of goes soft x - ray emission . since the only part of the total flare energy is converted into plasmoid ejections , the total energy of plasmoid ejections is smaller than the total energy of solar flare .
we analyzed the goes x2.3 class flare which occurred at 14:51 ut on 2000 november 24 .
we found multiple plasmoid ejections from a single flare .
furthermore , each plasmoid ejection seems to be associated with a peak in the hard x - ray emission . in figure 8 ,
the plasmoid ejections seem to occur at the height of approximately 50 " ( @xmath9 35 mm ) .
this tells us that reconnection occurs at the height of approximately 35 mm .
the horizontal light gray lines in the top panel of figure 9 show that the times of plasmoid ejections , when they reach the height of 35 mm , seems to be well correlated to the peak time of hard x - ray emission , which is consistent with previous studies ( e.g. * ? ? ? * ) .
since the peak in hard x - ray emission indicates strong energy release , we have demonstrated that each plasmoid ejection occurs during a period of strong energy release , suggesting a series of impulsive energy releases in a single flare .
we also performed a temperature and emission measure analysis and investigated the physical parameters of the plasmoid ejections shown in table 1 .
figure 8 and table 1 show that the hard x - ray bursts in period a have large intensities in correlation with the large kinetic energy of plasmoid ejections p1-p4 .
conversely , the hard x - ray bursts in period b and c have small intensities in correlation with the kinetic energy of plasmoid ejections p5-p7 .
figure 10a shows the relation between the kinetic energy of plasmoid ejection and the intensity of the corresponding hard x - ray peak emission .
we can see a rough tendency that the larger kinetic energy of plasmoid ejections is associated with the brighter hard x - ray peak emission , and vice versa .
the hard x - ray emission is known to show energy release rate ( e.g. neupert effect ; * ? ? ?
* ; * ? ? ?
* ) , which leads us to the following equation : @xmath16 where we assumed that the released thermal energy of a solar flare in a short period is comparable to the kinetic energy of the plasmoid ejection .
this means that a plasmoid ejection with large hard x - ray emission , and therefore with large energy release , can be accelerated strongly .
a similar kinetic evolution is also seen in the case of cmes @xcite .
figure 10b shows the relation between the apparent velocity and the intensity of the corresponding hard x - ray peak emissions .
there seems to be a correlation between the plasmoid velocities and the hard x - ray emissions , although it is difficult to measure the velocities of the plasmoids precisely due to the faint emission .
similar to the above equations , we can derive the following relation between hard x - ray emission and the kinetic energy of a plasmoid ejection : @xmath17 here @xmath18 is the typical magnetic field in a current sheet , @xmath19 is the inflow velocity and @xmath20 is the characteristic length of the inflow region .
the inflow velocity @xmath19 is thought to be about 0.01@xmath21 from direct observations @xcite .
it has been also known that inflow can be controlled by the plasmoid ejection . as a plasmoid
is ejected out of the current sheet , the density in the current sheet decreases , and the inflow is enhanced to conserve the total mass under the condition that incompressibility is approximately satisfied . here , we assume that @xmath19 is proportional to the plasmoid velocity @xmath22 .
then , we find the relation @xmath23
. moreover , if we can further assume the @xmath22 is proportional to cme velocity @xmath25 , then @xmath23 @xmath24 @xmath25 .
this is consistent with the result of yashiro & gopalswamy ( 2009 ) .
a correlation between the energy release rate , which is represented by @xmath23 , and the plasmoid velocity @xmath22 is also successfully reproduced by a magnetohydrodynamic simulation @xcite . in the simulation
, they clearly showed that the plasmoid velocity controls the energy release rate ( i.e. reconnection rate ) in the nonlinear evolution .
observations of plasmoid ejections have been paid attention to as evidence of magnetic reconnection , though they found only one plasmoid ejection per one solar flare .
however , magnetic reconnection theory suggests that the impulsive bursty regime of reconnection or fractal reconnection is associated with a series of plasmoids of various scales .
it is known that magnetic reconnection is an effective mechanism for energy release in a solar flare .
once the current sheet becomes thin enough for the tearing instability to occur , repetitive formation of magnetic islands and their subsequent coalescence drives the `` impulsive bursty '' regime of reconnection @xcite .
furthermore , such reconnection can produce a fractal structure in the current sheet , which is not only theoretically predicted but has also been observed ( i.e. * ? ? ?
* ; * ? ? ?
the plasmoids generated in this cycle control the energy release by inhibiting magnetic reconnection in the current sheet and/or by inflow driven by the ejection
. the plasmoid ejection enhances reconnection and promotes further plasmoid ejections from the current sheet .
similar intermittent energy release from a solar flare has been reported as multiple downflows associated with hard x - ray bursts in the impulsive phase @xcite .
mckenzie et al . ( 2009 ) estimated the total amount of energy of multiple downflows and showed that it is comparable to the total amount of energy released from the magnetic field , which is consistent with the magnetic reconnection model .
tanuma et al .
( 2001 ) showed through numerical simulations that plasmoid ejections are closely coupled with the reconnection process and the greatest energy release occurs when the largest plasmoid is ejected .
if the strong energy release corresponds to magnetic reconnection , we may conclude that this is evidence of unsteady magnetic reconnection in a solar flare and that plasmoids have a key role in energy release and particle acceleration .
we first acknowledge an anonymous referee for his / her useful comments and suggestions .
we also thank a. hillier for his careful reading and correction of this paper .
we also thank j. kiyohara for her help in making the movies and t. morimoto for his help in finding the multiple plasmoid ejection events .
we made extensive use of yohkoh / sxt and hxt data .
this work was supported in part by the grant - in - aid for creative scientific research `` the basic study of space weather prediction '' ( head investigator : k.shibata ) from the ministry of education , culture , sports , science , and technology of japan , and in part by the grand - in - aid for the global coe program
the next generation of physics , spun from universality and emergence from the ministry of education , culture , sports , science , and technology ( mext ) of japan .
+ asai , a. , yokoyama , t. , shimojo , m. , & shibata , k. , 2004a , , 605 , l77 asai , a. , yokoyama , t. , shimojo , m. , masuda , s. , kurokawa , h. & shibata , k. , 2004b , , 611 , 557 aschwanden , m. j. 2002 , particle acceleration and kinematics in solar flares ( dordrecht : kluwer ) brta , m. , vrnak , b. , & karlick , m. 2008 , a&a , 477 , 649 benz , a. o. , & smith , d. f. , 1987 , sol .
phys . , 107 , 299 benz , a. o. & aschwanden , m. j. 1992 , in lect .
notes in physics , vol . 399 : eruptive solar flares , eds .
z. svestka , b. v. jackson , & m. e. machado ( new york : springer ) , 106 brueckner , g. e. et al .
, 1995 , , 162 , 357 brown , j. c. , & loran , j. m. , 1985 , , 212 , 245 carmichael , h. , 1964 , in the physics of solar flares , ed .
w. n. hess ( nasa sp-50 ; washington , dc : nasa ) , 451 cheng , c. z. , ren , y. , choe , g. s. , & moon , y. -j . , , 2003 , 596 , 1341 daughton , w. , roytershteyn , v. , albright , b. j. , karimabadi , h. , yin , l. , bowers , k. j. , 2009 , , 103 , 065004 dennis , b. r. , 1985 , sol .
phys . , 100 , 465 drake , j. f. , swisdak , m. , che h. , & shay , m. a. , 2006 , nature , 433 , 553 drake , j. f. , cassak , p. a. , shay , m. a. , swisdak , m. , & quataert , e. , 2009 , , 700 , l16 finn , j. m. & kaw , p. k. 1977 , phys .
fluids , 20 , 72 frost , k. j. & dennis , b. r. , 1971 , , 165 , 655 handy , b. n. , et al . 1999 , , 187 , 229 hara , h. , tsuneta , s. , lemen , j. r. , acton , l. w. , mctiernan , j. m. , 1992 , pasj , 44 , l135 hirayama , t. , 1974 , , 34 , 323 hudson , h. s. , kosugi , t. , nitta , n. v. , & shimojo , m. , 2001 , , 561 , l211 kane , s. r. , & anderson , k. a. , 1970 , , 162 , 1003 karlick , m. , frnk , f. , & mszrosov , h. 2002 , a&a , 395 , 677 karlick , m. , frnk , f. , & kurucker , s. , 2004 , a&a , 419 , 365 khan , j. i. , vilmer , n. , saint - hilaire , p. , & benz , a. o. 2002 , a&a , 388 , 363 kim , y .- h . , moon , y .- j . ,
cho , k .- s . , kim , k .- s . , & park , y. d. , 2005a , , 622 , 1240 kim , y .- h . , moon , y .-
cho , k .- s . , bong , s .- c . ,
& park , y. d. , 2005b , , 635 , 1291 kiplinger , a. , l. , dennis , b. r. , emslie , a. g. , frost , k. j. , & orwig , l. e. , 1983 , , 265 , l99 kliem , b. , karlick , m. & benz , a. o. 2000 , a&a , 360 , 715 kopp , r. a. , & pneuman , g. w. , 1976 , , 50 , 85 kosugi , t. et al .
1991 , sol.phys . , 136 , 17 kowal , g. , lazarian , a. , vishniac , e. t. , & otmianowska - mazur , k. , 2009 , , 700 , 63 kundu , m. r. , nindos , a. , vilmer , n. , klein , k. -l . , shibata , k. , ohyama , m. , 2001 , , 559 , 443 liu , s. , petrosian , v. , & mason , g. m. , 2006 , , 636 , 462 mckenzie , d. e. & hudson , h. s. , 1999 , 519 , l93 mckenzie , d. e. , & savage , s. l. , 2009 , , 697 , 1569 miller , j. a. , larosa , t. n. , & moore , r. l. 1996 , , 461 , 445 moon , y .- j . ,
choe , g. s. , wang , h. , park , y. d. , & cheng , c. z. , 2003 , korean astron .
soc . , 36 , 61 narukage , n. , & shibata , k. , 2006 , , 637 , 1122 neupert , w. m. , 1968 , , 153 , l59 nitta , n. v. & hudson , h. s. , 2001 , geophys . res .
lett . , 28 , 3801 nishida , k. , shimizu , m. , shiota , d. , takasaki , h. , magara , t. , & shibata , k. , 2009 , , 690 , 748 nishizuka , n. , asai , a. , takasaki , h. , kurokawa , h. , & shibata , k. , 2009 , , 694 , l74 ogawara , y. , takano , t. , kato , t. , kosugi , t. , tsuneta , s. , watanabe , t. , kondo , i. , uchida , y. , 1991 , , 136 , 1 ohyama , m. & shibata , k. 1997 , publ .
japan , 49 , 249 ohyama , m. & shibata , k. 1998 , , 499 , 934 priest , e. r. , 1985 , rep .
48 , 955 pritchett , p. l. 2008 , phys .
plasma , 15 , 102105 samtaney , r. , loureiro , n. f. , uzdensky , d. a. , schekochihin , a. a. , & cowley , s. c. , 2009 , , 103 , 105004 shibata , k. , masuda , s. , shimojo , m. , hara , h. , yokoyama , t. , tsuneta , s. , kosugi , t. , ogawara , y. , 1995 , , 451 , l83 shibata , k. & tanuma , s. , 2001 , earth planets space , 53 , 473 strauss , h. r. , 1988 , , 326 , 412 sturrock , p. a. , 1966 , nature , 211 , 695 sui , l. , holman , g. d. , & dennis , b. r. , 2004 , , 612 , 546 tajima , t. , sakai , j. , nakajima , h. , kosugi , t. , brunel , f. , kundu , m. r. , 1987 , , 321 , 1031 takasaki , h. , asai , a. , kiyohara , j. , shimojo , m. , & terasawa , t. , takei , y. , shibata , k. , 2004 , , 613 , 592 takasaki , h. , 2006 , phd thesis , kyoto university tanuma , s. , yokoyama , t. , kudoh , t. & shibata , k. 2001 , , 551 , 312 tsuneta , s. et al .
1991 , , 136 , 37 tsuneta , s. , 1997 , , 483 , 507 wang , h. , gallagher , p. , yurchyshyn , v. , yang , g. , & goode , p. r. , 2002 , , 569 , 1026 yashiro , s. & gopalswamy , n. , 2009 , in proc iau symp .
257 , universal heliophysical processes , eds .
n. gopalswamy & d. f. webb ( cambridge : cambridge univ . press ) , 233 yokoyama , t. , akita , k. , morimoto , t. , inoue , k. , newmark , j. , 2001 , , 546 , l69 zhang , j. , dere , k. p. , howard , r. a. , kundu , m. r. , & white , s. m. , 2001 , , 559 , 452 zhang , j. & wang , j. , 2002 , , 566 , l117 ccccccccccccrl & ( ut ) & ( k ) & ( @xmath15 ) & ( 10@xmath26@xmath27 ) & ( 10@xmath14@xmath15 ) & ( 10@xmath28 g ) & ( km / s ) & ( 10@xmath29erg ) & ( 10@xmath29erg ) & & ( cts / sec / sc ) + p1 & 15:09:19 - 15:09:59 & 6.98 - 7.09 & 45.32 - 45.33 & @xmath303.9 - 4.7 & 2.76 - 2.79 & @xmath302.63 - 2.66 & 654 - 818 & @xmath305.63 - 8.91 & @xmath302.51 - 3.28 & northwest & 1000 + p2 & 15:09:59 - 15:10:59 & ( 7.00 ) & ( 45.34 ) & @xmath304.6 - 5.8 & ( 2.65 ) & @xmath30(3.69 ) & 573 - 654 & ( @xmath306.05 - 7.88 ) & ( @xmath303.69 ) & northwest & 1300 + p3 & 15:10:19 - 15:10:59 & ( 7.00 ) & ( 45.32 ) & @xmath304.7 - 8.3 & ( 2.59 ) & @xmath30(3.60 ) & 982 - 1470 & ( @xmath3017.4 - 38.9 ) & ( @xmath303.60 ) & northwest & 1500 + p4 & 15:10:39 - 15:11:19 & 7.02 - 7.07 & 45.80 - 45.80 & @xmath303.1 - 7.1 & 4.85 & @xmath304.30 & 654 - 828 & @xmath309.19 - 14.7 & @xmath304.50 - 5.05 & northwest & 950 + p5 & 15:13:33 - 15:14:27 & 7.02 - 7.08 & 45.73 - 45.74 & @xmath301.5 - 3.1 & 5.76 - 5.83 & @xmath301.12 - 1.13 & 360 - 844 & @xmath300.72 - 4.03 & @xmath301.17 - 1.36 & northwest & 120 + p6 & 15:13:33 - 15:13:51 & 7.07 - 7.11 & 45.49 - 45.50 & @xmath301.6 - 5.2 & 4.37 - 4.42 & @xmath300.85 - 0.86 & 360 - 720 & @xmath300.55 - 2.22 & @xmath300.99 - 1.11 & southwest & 130 + p7 & 15:16:27 - 15:18:43 & 6.97 - 7.03 & 46.05 - 46.06 & @xmath303.2 - 7.8 & 6.95 - 7.03 & @xmath303.99 - 4.04 & 254 - 409 & @xmath301.29 - 3.38 & @xmath303.72 - 4.33 & northwest & 150 + total & - & - & - & - & - & - & - & 43.5 - 84.7 & 20.2 - 22.4 & & & & + flare & 15:18:33 - 15:18:53 & 6.73 - 7.08 & 48.27 - 48.96 & 9.33 & 137 - 304 & 391 - 865 & - & - & 210 - 1040 & & & & + | the soft x - ray telescope ( sxt ) on board yohkoh revealed that the ejection of x - ray emitting plasmoid is sometimes observed in a solar flare .
it was found that the ejected plasmoid is strongly accelerated during a peak in the hard x - ray emission of the flare . in this paper
we present an examination of the goes x 2.3 class flare that occurred at 14:51 ut on 2000 november 24 . in the sxt images we found `` multiple '' plasmoid ejections with velocities in the range of 250 - 1500 km s@xmath0 , which showed blob - like or loop - like structures .
furthermore , we also found that each plasmoid ejection is associated with an impulsive burst of hard x - ray emission .
although some correlation between plasmoid ejection and hard x - ray emission has been discussed previously , our observation shows similar behavior for multiple plasmoid ejection such that each plasmoid ejection occurs during the strong energy release of the solar flare . as a result of temperature - emission measure analysis of such plasmoids
, it was revealed that the apparent velocities and kinetic energies of the plasmoid ejections show a correlation with the peak intensities in the hard x - ray emissions . |
a wide variety of technical processes involve the storage and transport of granular materials .
on the other hand , granular materials have very unusual properties which have intrigued researches in physics over recent years @xcite . of great importance
is the characterization of stress fluctuations in dry granular materials .
optical measurements in two and three dimensional @xcite arrays of granular materials have showed that the stress in packed granular materials is not distributed uniformly inside the medium , but is concentrated along `` chains '' .
further experiments dealing with force fluctuations in the bottom of relative small containers have shown an exponential distribution of vertical forces ( weight ) @xmath0 @xcite .
another interesting phenomena associated with weight fluctuations is the problem of arching .
arching refers to the observation that the average weight at the bottom of a column of grains saturates at a value @xmath1 as the depth @xmath2 measured from the top of the column is increased , where @xmath1 shows large fluctuations when repeating this procedure with the same amount of grain .
about one century ago janssen @xcite showed , using a simple argument , that @xmath3 $ ] .
@xmath4 is the saturation depth which scales as @xmath5 , where @xmath6 is the linear horizontal size of the container .
more recently , liu _
et al _ @xcite proposed a simple discrete model ( @xmath7-model ) which describe the essential features of force fluctuations in granular materials .
for instance , numerical simulations of the model in @xmath8 dimensions give rise to an exponential distribution @xmath9 for the normalized weight @xmath10 , in agreement with experimental observations and numerical simulations of sphere packings @xcite .
later peralta - fabi _ et al _ @xcite showed through numerical simulations in @xmath11 dimensions and mean field ( mf ) analysis that the same model , but with different boundary conditions , explains the process of arching , obtaining the janssen law but with @xmath12 and @xmath13 .
generalizations of the @xmath7-model has been considered @xcite .
some of them including a local slip condition which leads to anisotropies in the stress transmission @xcite and introduces correlations @xcite .
these correlations , nevertheless , do not alter the asymptotic exponential decay of the normalized weight distribution @xcite .
on the other hand , other authors have introduced more realistic models which takes into account the stress fluctuations in the horizontal cross - section @xcite .
some of their predictions are in agreement with those obtained for the @xmath7-model , such as the exponential asymptotic decay of the normalized weight distribution characteristic of the @xmath7-model .
we also count with some coarse - grained equations for the @xmath7-model @xcite .
et al _ @xcite obtained a diffusion equation for the weight with a multiplicative noise and derived some analytical results for the weight correlation function .
however , their analysis was limited to pile configurations and silos with periodic boundary conditions . on the contrary , peralta - fabi _ et al _ @xcite the case when
part of the weight is supported by the walls of the silo , which lead to a saturation of the average weight profile .
nevertheless , they neglected fluctuations in the local stress transmission . in the present work we focus our attention in obtaining a continuum model to describe weight fluctuations and the process of arching in granular materials , for both periodic and wall boundary conditions .
we take as starting point the @xmath7-model introduced by liu _
et al _ @xcite .
we obtain a coarse - grained continuum equation of this model which is similar to the edwards - wilkinson equation , where the grain weight acts as an external force and noise is added to consider fluctuations in shape and orientation .
boundary and initial conditions are imposed following the definition of the discrete model .
two particular cases are analyzed , periodic and absorbing boundaries . in the case of periodic boundary conditions
the average weight increases linearly with depth and is independent of the lattice size , while in the case of absorbing boundary conditions it saturates for @xmath14 to @xmath12 .
moreover , in both cases fluctuations around the average are of the order of @xmath15 , with @xmath16 .
the paper is organized as follows . in section [ sec : introduction ] we derive the coarse - grained equation for the @xmath7-model and solve it for the case of periodic and absorbing boundary conditions . in section [ sec : discussion ]
our results are compared with numerical simulations in one dimension , and with other works in the literature .
finally the summary and conclusions are given in section [ sec : summary ] .
liu _ et al _ @xcite introduce a simple model ( @xmath7-model ) that describe some of the experimental observations in bead packs . the model assumes that the dominant physical mechanism leading to force fluctuations is the inhomogeneity of the packing , which causes an unequal distribution of weight on the beads supporting a given grain .
spatial correlations in these fractions and variations in the coordination numbers of the grains are ignored .
only the vertical components of the forces are considered explicitly .
consider a @xmath17 dimensional array of grains in which each layer has @xmath18 beads . the total weight on a given bead is transmitted unevenly to @xmath19 adjacent beads in the layer underneath .
specifically , a fraction @xmath20 of the total weight supported by the bead @xmath21 in layer @xmath2 is transmitted to bead @xmath22 in layer @xmath23 .
thus , a site at depth @xmath23 has weight @xmath24 , due to its own weight @xmath25 and to weights of neighbors at depth @xmath2 , according to @xmath26 where the sum runs over the @xmath19 neighbors , in layer @xmath2 , of the site @xmath22 , in layer @xmath23 .
the fractions @xmath27 are taken to be random variables , independent except for the constraint @xmath28 where the sum runs over the @xmath19 neighbors , in layer @xmath23 , of site @xmath21 , in layer @xmath2 . to complete model definition ,
boundary and initial conditions must be specified . in the top layer ( @xmath29 )
the grains have no neighbors above and , therefore , they only support their own weight , i.e. @xmath30 . on the other hand , we can take different boundary conditions .
we consider two cases , periodic and absorbing boundary conditions .
absorbing boundary conditions are more appropriate to describe the force fluctuations in silos where part of the weight is supported ( `` absorbed '' ) by the wall containing the grains .
the absorbing boundary conditions are implemented by simply imposing that boundary sites have zero weight .
if a site near the boundary transmit part of its weight to a boundary site ( i.e. to the wall ) this weight fraction is `` dissipated '' , while the boundary site never transmit weight to its neighbors because by definition its weight is zero .
this boundary conditions are different but equivalent to the one used by peralta - fabi _ et al _ @xcite . on the other hand ,
for periodic boundary conditions we recover the original @xmath7-model @xcite . to derive a coarse - grained continuum equation of eq .
( [ eq:1 ] ) let us introduce the new random fraction @xmath31 , such that @xmath32 which following ( [ eq:2 ] ) satisfies the constraint @xmath33 substituting @xmath27 in eq .
( [ eq:1 ] ) by eq .
( [ eq:3 ] ) and substrating @xmath34 to both sides it results that @xmath35+w_0+\eta_i(t ) , \label{eq:5}\ ] ] where @xmath36 @xmath37 is a noise term associated with the random fractions @xmath31 and with weights at neighbor sites . in average the contribution of @xmath37
should be zero because of the constraint in eq .
( [ eq:4 ] ) .
( [ eq:5 ] ) can be coarse - grained to obtain a continuum equation for the effective @xmath38 . in the left hand side
we have the discrete depth derivative , while the first term in the right hand side is the discrete laplacian in the horizontal direction .
after coarse - graining it result that @xmath39 @xmath40 and @xmath41 are coarse - grained coeficients . @xmath42 and @xmath43 , where @xmath44 and @xmath45 are characteristic lengths in the vertical and horizontal direction , respectively . in @xmath46-dimensions
( [ eq:7 ] ) is actually very similar to the one derived in @xcite .
et al _ @xcite considered the explicit multiplicative nature of the noise .
however , it is not clear how their analysis can be extended to ( @xmath8)-dimensions , in particular the precise form of the coarse - grained noise is in this case not clear for us . instead , we assume that @xmath47 is a gaussian noise with zero mean and uncorrelated in space , with noise correlator @xmath48 @xmath49 is a monotonically decreasing even function .
we do not know the precise form of @xmath49 , but we assume it decays to zero beyond depth @xmath50 .
depending on the value of @xmath50 we can obtain different behaviors .
for @xmath51 the noise will be also uncorrelated in the vertical direction , i.e. @xmath52 . on the contrary , if @xmath50 is much larger than any characteristic depth then @xmath49 may be considered constant .
we expect the noise @xmath47 to be strongly correlated in the vertical direction . according to eq .
( [ eq:6 ] ) the noise actually depends on weights at the different sizes , which should be strongly correlated from layer to layer .
we have avoided the multiplicative nature of the noise but in compensation we must keep the correlations in the vertical direction . in any case
we are going to consider different choices of @xmath49 and compare the results with numerical simulations .
( [ eq:7 ] ) can be interpreted as the equation of motion of an interface profile @xmath38 , where depth plays the role of time , under an external force @xmath25 and annealed noise @xmath47 .
this observation is very important because it shows that force fluctuations in granular media can be described through a more general framework , that if interface dynamics , which has been extensively studied in the literature @xcite . in this context ,
( [ eq:7 ] ) is known as the edwards - wilkinson ( ew ) equation after @xcite .
a central quantity of interest is the width @xmath53 of the fluctuating `` interface '' , given by @xmath54 ^ 2\rangle$ ] , where @xmath55 is the mean height of the interface ( mean weight ) . in general the `` surface roughness ''
@xmath53 has the following asymptotic behavior @xmath56 where @xmath57 and @xmath58 are the roughness and dynamic exponent , respectively . in the thermodynamic limit @xmath59 , being linear , eq .
( [ eq:7 ] ) is readily solved via fourier methods .
direct integration shows that , if the noise is spatially and temporally uncorrelated , the roughness exponent is given by @xmath60 @xcite . in this section
we investigate the solution of eq .
( [ eq:7 ] ) for different boundary conditions and noise correlators @xmath49 .
we are interested in the stationary solution and , therefore , the initial condition is irrelevant .
let first analyze the case of periodic boundary conditions . in this case
the solution can be written as @xmath61 after substituting this expression in eq .
( [ eq:7 ] ) we obtain the following equation for @xmath62 @xmath63 the solution of this equation clearly has zero average , i.e. @xmath64 , and therefore @xmath65 the average weight thus increases linearly with @xmath2 and no saturation is observed . on the contrary , for absorbing boundary conditions
the solution can not be proposed as in eq .
( [ eq : b1 ] ) . in this case weight dissipation at the boundary leads to a stationary state where the average weight saturates . in the language of interface dynamics this situation correspond with an elastic interface pinned at the boundary under an external force @xmath25 . in this case
is better to look for a solution of the form @xmath66 where @xmath67 is the solution of the stationary problem @xmath68 with homogeneous boundary conditions ( @xmath69 ) , and @xmath62 satisfies eq .
( [ eq : b2 ] ) but with homogeneous boundary conditions .
again , @xmath62 has zero average and , therefore , the mean weight in the stationary state is @xmath67 .
the solution of eq .
( [ eq : b5 ] ) can be easily obtained for certain geometries . in @xmath70
we can look for the solution in the interval @xmath71 with @xmath72 , obtaining @xmath73 for @xmath74 , @xmath75 and @xmath76 ( only the case @xmath74 has a physical realization ) we can look for the solution in a circle , sphere and hyper - sphere of radius @xmath6 such that @xmath77 , with @xmath78 and the boundary condition @xmath79 , where @xmath80 is the distance to the center .
the solution is in this case given by @xmath81 thus , in any dimension , @xmath6 is the characteristic length of the system and the saturation weight scales as @xmath82 .
now we are going to analyze the fluctuations around the average , described by eq .
( [ eq : b2 ] ) with the corresponding boundary conditions and noise correlator . a formal solution of eq .
( [ eq : b2 ] ) is given by @xmath83 where @xmath84 is the green function of corresponding boundary problem , which depends on the dimension of the horizontal space @xmath85 and on the linear size of the system @xmath6 .
the precise form of the green function can be obtained only for some suitable geometries , however , in general it satisfies the scaling relation @xmath86 where @xmath87 depends on the spatial dimension and boundary conditions and @xmath6 is a characteristic linear dimension of the system .
@xmath88 can be obtained exactly , for instance , in one dimension and in a systems with radial symmetry .
using equations ( [ eq : b8 ] ) , ( [ eq : b9 ] ) , and ( [ eq:8 ] ) we compute the `` surface roughness '' @xmath53 , obtaining the scaling relation in eq .
( [ eq : b0 ] ) with @xmath89 and a roughness exponent @xmath57 which depends on the choice of @xmath49 .
if the noise is uncorrelated in the vertical direction @xmath2 ( @xmath90 ) then @xmath91 which corresponds with the ew universality class . on the contrary ,
if it is strongly correlated in the vertical direction ( @xmath92 ) then @xmath93 in both cases the exponent @xmath57 is independent of the choice of boundary conditions , and they are identical to the exponents obtained from the fourier analysis for a @xmath59 system .
thus , while the average profile is strongly dependent on the boundary conditions , the `` surface roughness '' only depends on the noise correlator .
to test results obtained in the previous section we have performed numerical simulations of the discrete model in one dimension ( @xmath94 ) with a uniform @xmath7- distribution .
we have used @xmath95 and lattice sizes @xmath96 , @xmath97 , @xmath98 and @xmath99 .
average were taken over 1000 realizations for the three smallest lattice sizes and over 100 realizations for the largest one .
different magnitudes associated with force fluctuations were computed .
our analytical approach reveals that the existence of absorbing boundaries is necessary to obtain a saturation weight , otherwise the weight increases linearly with depth . in fig .
[ fig:1 ] we have plotted the global average weight @xmath100 at depth @xmath2 , for periodic and absorbing boundary conditions .
the agreement with our prediction becomes evident .
moreover , in the case of absorbing boundary conditions , we have computed the average weight profile after saturation which in one dimension is given by eq .
( [ eq : b6 ] ) . the average weight after saturation as a function of lattice position @xmath101
is shown in fig .
[ fig:2 ] , for different lattice sizes .
the scaled variables @xmath102 vs. @xmath103 have been used as suggested by eq .
( [ eq : b6 ] ) .
all the curves collapse in a single plot showing that our choice of scaled variables is correct .
moreover , the scaled plot was fitted to the quadratic dependency @xmath104 , with @xmath105 , @xmath106 and @xmath107 .
these fitting parameters are in very good agreement with eq .
( [ eq : b6 ] ) since they satisfy @xmath108 and @xmath109 .
we have thus obtained the correct profile for the saturation weight . to investigate the transient region before saturation
we have computed the global average weight as a function of depth .
the result for different lattice sizes is shown in fig .
[ fig:3 ] .
the scaled variables @xmath110 vs.@xmath111 have been used . for small depths
@xmath112 scales linearly with @xmath2 , as for periodic boundary conditions . in this region
the force chains starting from bulk sites have not reached the boundary and , therefore , the weight dissipation at the boundary is negligible .
this transient behavior was previously observed in numerical simulations by peralta - fabi _ et al _ in one dimension @xcite .
the average squared fluctuations ( interface roughness ) @xmath113 as a function of @xmath2 are shown in fig .
[ fig:4 ] . using the scaled variables @xmath114 , with @xmath115 , vs. @xmath111
we obtain a good data collapse .
this scaled plot is very sensitive to the choice of @xmath57 , a good data collapse is only obtained for @xmath116 .
if the noise is uncorrelated in the vertical direction then @xmath117 , which is much smaller than our numerical estimate . on the contrary , if the noise is strongly correlated in the vertical direction then @xmath115 , in very good agreement with our numerical estimate .
when we avoid the multiplicative character of the noise we guess that , in compensation , strong correlations in the vertical direction should be considered .
this supposition is now supported by numerical simulations .
the simplicity of the continuum model we have proposed in the previous section allowed us to obtain some analytical results .
we have avoided the complicated form of the noise , associated with the fluctuations in the fractions @xmath118 , introducing an `` annealed '' noise with correlations in the vertical direction . within this approximation
we have obtained the average weight in the stationary state and characterized the fluctuations around the average , for both periodic and absorbing boundary conditions . in earlier numerical simulations by liu _
et al _ @xcite , they considered periodic boundary conditions and compute the distribution of normalized weight @xmath10 .
they observe that the distribution of normalized weight @xmath119 was independent of @xmath2 for large @xmath2 .
this numerical result was later corroborated by coppersmith _
et al _ using a mf theory for the discrete model @xcite .
according to our approach , for periodic boundary conditions the average weight is given by @xmath120 and , therefore , @xmath121 .
since the average normalized weight is independent of @xmath2 so should be its distribution , in agreement with the numerical sim@xmath122 ulations and mf theory of the discrete model .
the robustness of the asymptotic exponential decay of the normalized weight distribution @xmath119 is one of the main features of the @xmath7-model .
we thus check if this behavior is still observed when absorbing boundary conditions are considered . before going to the numerical results we should remember that in this case for @xmath123 the average weight does not increases linearly with @xmath2 but scales as @xmath82 .
hence , the appropriate choice of normalized weight is in this case @xmath124 . with this remark , in fig .
( [ fig:5 ] ) we show the distribution of normalized weight for different lattice sizes .
first we notice that @xmath119 is independent of lattice size , supporting our choice of normalized weight . moreover
, the resulting universal curve clearly show an exponential decay for large @xmath125 .
however , for small @xmath125 the data does not display the linear behavior predicted by the @xmath7-model for a uniform distribution @xcite .
this discrepancy for small @xmath125 is just a consequence of the different boundary conditions .
on the other hand , in the case of absorbing boundary conditions our analytical approach is in agreement with previous mf theory by peralta - fabi _ et al _ @xcite in one dimension .
peralta - fabi _ et al _ analyzed the case @xmath126 , obtaining the scaling dependencies @xmath12 and @xmath127 .
we have shown that these scaling relations are valid in larger dimensions and are independent of the noise .
however , within the mf theory by peralta - fabi _ et al _ one can not determine the scale dependency of the average squared fluctuations @xmath128 , which constitutes one of our main results . the scaling dependencies @xmath12 and @xmath13 are , nevertheless , in contradiction with the linear scaling obtained in classical theories @xcite , some generalizations of the @xmath7-model @xcite and even in experimental observations @xcite .
however , this discrepancy can not be attributed to our continuum analysis , which shows a very good agreement with the numerical simulations of the @xmath7-model .
we have obtained a coarse - grained equation of the discrete model introduced by liu _ et al _ to describe force fluctuations in granular media .
the multiplicative nature of the noise has been considered assuming strong correlations in the vertical direction .
the stationary and transient behavior were obtained analytically for periodic and absorbing boundary conditions . in this way we have demonstrated that the existence of walls , modeled by the absorbing boundary conditions , are a necessary condition to obtain an stationary state
were the average weight is independent of depth .
we have also shown that the the scaling of the saturation weight @xmath12 and depth @xmath13 with lattice size are valid in any dimension and are independent of the noise , generalizing previous mf calculations in one dimension .
for the first time we have obtained the scale dependency of the average squared fluctuations .
the comparison of this scale dependence with the one obtained in the numerical simulations support our guess about the existence of strong correlations of the noise in the vertical direction .
h. m. jaeger and s. r. nagel , science * 255 * , 1523 ( 1992 ) ; d. bideau and a. hansen , _ disorder and granular media _ ( north holland , amsterdam , 1993 ) ; c. thornton , _ powders and grains _ ( balkema , rotterdam , netherlands , 1993 ) ; a. metha , granular matter ( springer , berlin , 1993 ) .
p. dantu , ann .
ponts chausses * 4 * , 144 ( 1967 ) ; a. drescher and g. de josselin de jong , j. mech .
solids * 20 * , 337 ( 1972 ) ; t. travers , m. ammi , d. bideau , a. gervois , j. c. messager , and j. p. troadec , europhys .
* 4 * , 329 ( 1987 ) . | force fluctuations in granular materials are investigated .
a continuum equation is derived starting from a discrete model proposed in the literature . the influence of boundary conditions is investigated . for periodic boundary conditions the average weight
is found to increase linearly with depth while it saturates to a constant value for absorbing boundary conditions , which models the existence of walls .
the scale dependencies of the saturation weight , the saturation depth and the average squared fluctuations are obtained .
the analytical results are compared with previous works and with numerical simulations in one dimension . 2 |
photoemission lineshape contains information on the many - body interactions in the solid under study .
one notable example is the kink structure in the angle resolved photoemission(arpes ) spectral function from a metal at the bosonic mode energy due to electron - bosonic mode coupling@xcite . in this regard ,
the kink structures at near 70 mev found in the arpes spectral functions from cuprate high temperature superconductors ( htscs ) have been a hot topic during the last several years as it may provide a clue on what mediates the pairing in the htsc@xcite .
it is still under debate whether the kink is caused by phonons or magnetic excitations .
in addition to the kink near 70mev , an anomalous high energy kink structure from insulating cuprate ca@xmath2cuo@xmath2cl@xmath2 ( ccoc ) was recently reported@xcite .
subsequently , it was found that the high energy kink also exists in doped cases , making it omnipresent at all dopings@xcite .
existence of the high energy kink reveals that there may be another energy scale from the many - body interactions in htscs .
it is thus important to find the physics in this high energy scale .
one can extract from the data the information on the many - body interactions through the self energy analysis .
conventionally , the self energy analysis is done by extracting the dispersion and peak width from the momentum distribution curves ( mdcs)@xcite . however , there are underlying assumptions for the analysis to be valid .
first of all , arpes peaks should be sharp enough to be quasiparticle - like .
secondly , the bare band dispersion should be quite linear within the energy range of interest so that one can guessthe bare band dispersion .
unfortunately , none of these are valid assumptions in the analysis of arpes spectral function from ccoc : the peaks are too broad to be considered as quasi - particle peaks@xcite and there is an inherent confusion regarding what should be considered as the original bare band as ccoc is a mott insulator
. similar limitations also apply to the recently found high energy kinks in htscs@xcite .
however , the lack of quasiparticles should not be confused with a lack of physics .
for example , in luttinger liquids , where fermi liquid quasiparticles do not exist , arpes has provided detailed information on the excitation spectrum.@xcite .
it is therefore necessary to develop a method to extract the self energy from non - quasi particle arpes spectral functions .
in fact , there was an earlier attempt to obtain the self energy through a hilbert transformation with the assumption of particle - hole symmetry@xcite .
this method however is not applicable to systems without particle - hole symmetry ( for example , ccoc@xcite ) .
here we propose to use the self energy defined for the electron _ removal _ green s function .
even though different , it bares similar characters of the usual self energy and its imaginary part is found to appropriately represent the width of the spectral feature .
moreover , we developed a method to extract the self energy from a general arpes spectrum from a single band system .
it can thus be used to for spectral functions with broad spectral features .
we believe our analysis method provides a new way to examine the underlying physics from photoemission spectral functions .
the full green s function for a single band system at 0 k can be written as @xmath4\nonumber\\ & = & \frac{1}{\omega-\varepsilon_k-\sigma^{full}(\textbf{k},\omega)}\end{aligned}\ ] ] where @xmath5 and @xmath6(*k*,@xmath1 ) are the bare band and self energy , respectively .
the chemical potential @xmath7 is set to zero for convenience .
note that due to the sum rule @xmath8=1 , the asymptotic behavior ( @xmath9 ) of @xmath10 is 1/@xmath1 .
this makes the self energy non - divergent when @xmath11 .
in addition , the self energy is analytic in the positive complex plane , and as a consequence the real and imaginary parts of the self energy are hilbert transforms of each other@xcite . the green s function defined in this way ( full green s function )
is thought to represent electron addition or removal processes and the self energy holds the information on the interaction .
the full green s function in the above equation however includes not only electron removal but also electron addition process for which we do not have information . our goal is to obtain the self energy from a general spectral function with as few assumptions as possible .
therefore , we attempt to look at the self energy for the electron removal green s function which should best describe the arpes process and is defined as @xmath12 or @xmath13 where @xmath14 and @xmath15 is the electron removal spectral function ( the superscript er stands for electron removal ) .
another difference is that @xmath5 @xmath16 @xmath17 be the bare band dispersion as will be discussed later .
the asymptotic behavior of @xmath18 is now @xmath19 and with the definition given in equation ( 2 ) , the real and imaginary parts of the electron removal self energy ( re@xmath0 and im@xmath0 , respectively ) are again related through hilbert transforms . to obtain the electron removal self energy @xmath0
, we use the fact that the measured arpes spectrum is proportional to the electron removal spectral function , that is , @xmath20 where _ _ i__@xmath21 and m(*k * ) are the measured arpes spectrum and matrix element , respectively . here , the @xmath1 dependence of the matrix element m(*k * ) is ignored as usual . note that the fermi function is missing as we deal with the electron @xmath22 green s and spectral functions .
the real and imaginary parts of the electron removal green s function are , as is the case for @xmath10 , hilbert transforms of each other , re@xmath23=h(im@xmath23 ) where h ( ) represents the hilbert transformation . then the combination of equations ( 3 ) and ( 4 ) gives@xcite , @xmath24 @xmath25 where @xmath5 and @xmath26 are the bare band and -@xmath27 , respectively . the only remaining task to get
@xmath3 is to determine @xmath26 and @xmath5 .
we can obtain @xmath26 by integrating both sides of equation ( 4 ) .
@xmath28 to define @xmath5 , we enforce the standard relationship re@xmath0=h(im@xmath0 ) .
for this relationship to hold , @xmath0(*k*,@xmath1 ) @xmath29 0 as @xmath30 . by expanding the last term in equation ( 3 ) in terms of 1/@xmath1 for @xmath31
, we get @xmath32 using @xmath33 and @xmath34 , @xmath0 becomes @xmath35 since @xmath0(*k*,@xmath1 ) @xmath29 0 as @xmath31 , we set @xmath36 and thus we have @xmath37 that is , @xmath5 is the center of gravity of the spectral function but may not be the true bare band .
we rationalize this definition of @xmath5 from the knowledge that @xmath38 , the bare band value , independent of the strength of interactions in the limit @xmath39 1@xcite .
alternatively , @xmath5 could have been chosen differently such as the true bare band value of the original non - interacting problem but it will only shift re@xmath0 determined here by a constant . in that case , the relationship between re@xmath0 and im@xmath0 would be modified to re@xmath0=h(im@xmath0)+@xmath40-@xmath5 .
we note then that re@xmath0 goes to a finite value of @xmath40-@xmath5 as @xmath11 ( similarly re@xmath6 asymptotically goes to the hartree - fock value ) . using this expression
we can see that the parameter @xmath5 drops out and that the imaginary part of the self energy remains uniquely determined independent of our choice of @xmath5 .
in applying the newly developed method to the arpes data , one has to consider other aspects of the experimental data such as experimental resolution and thermal broadening . in that respect
, ccoc can be an ideal system to test the method : the experimental resolution and thermal broadening are much smaller than the width of the characteristic features of ccoc and thus can be ignored .
this provides an extra motivation for this study . for the experimental data , we use the published ccoc data along the ( 0,0 ) to ( @xmath41,@xmath41 )
cut shown in fig . 1(a)@xcite .
before we apply the procedure to extract @xmath0 , we also have to subtract the background intensity from the data .
this is because the background intensity is from a process that can not be described by the green s function in equation ( 2 ) . in our case , subtracting the background so that @xmath42(*k*,@xmath1 ) @xmath29 0 as @xmath31 is a necessary condition for re@xmath0 and im@xmath0 to be hilbert transforms of each other .
the background intensity is also casually subtracted in other methods , either naturally by fitting the mdcs@xcite or enforcing it@xcite .
there are a number of potential ways to subtract the background , among which we may consider the following two .
the first one is subtracting the shirley background@xcite .
the other is subtracting the intensity of the unoccupied * k * states ( approximately beyond @xmath43 in fig .
1(a ) ) , that is , the mdc background assuming that it is from the momentum independent background intensity .
we show the result extracted by using the latter method in fig .
however , the final results are similar independent of the background subtraction method .
= 8.5 cm in panels ( a ) and ( b ) of fig .
2 , we show re@xmath0 and im@xmath0 obtained by applying the procedure discussed above . note that while re@xmath0 can be negative , -im@xmath0 is always positive .
the most notable aspect of the data is that overall @xmath44 dependence for both re@xmath0 and im@xmath0 is weak .
this is very striking considering the fact that the dispersion of the spectral features is very strong .
the momentum independence of the self energy which has been casually assumed@xcite is indeed seen in the data .
the momentum independence could further be tested for our systems . yet
, we still see some momentum dependence of the self energy , especially at the low energy side .
therefore , for the accurate measure of the self energy , it is important to have the self energy in the full * k*-@xmath1 space as is the case for our method . the potential importance of the above observation can not be understated .
the largest shortcoming of dynamical mean field theory ( dmft ) is that momentum independence of the self energy must be assumed ( some momentum dependence can be added back in with cluster calculations , but this is computationally very expensive)@xcite .
while experimental limitations forced us to make a few assumptions such as the background subtraction , our results are encouraging that the momentum independent self energy assumption made in dmft calculations of the cuprates are valid . to our knowledge , this is the first experimental attempt to prove the validity of this assumption .
looking at im@xmath0 plotted in panel ( b ) , one may wonder about the validity of the result near the gap region as im@xmath0 is zero while it is known that im@xmath6 diverges near the gap region@xcite .
the divergence of im@xmath6 near the gap is due to the fact that multiple poles of @xmath10 ( or two peak structure in @xmath45(*k*,@xmath1 ) ) are attempted to be described by a single pole with a self energy .
however , by using only the occupied part , _ a@xmath46 _ effectively becomes a single peak spectral function .
as im@xmath6 does not diverge for a single peak spectral function , im@xmath0 calculated from an effective single peak does not necessarily diverge even though ccoc is an insulator .
this issue will be discussed again with simulations presented in fig .
3 . in the conventional method of fitting mdcs and converting the mdc widths to edc widths@xcite ,
one estimates the width , thus im@xmath0 , at the peak positions .
we present the result in a similar fashion by plotting im@xmath0 along the dispersion .
this provides the most accurate estimate of @xmath0 of the feature because of the possible momentum dependence discussed above . in fig .
2(c ) , we plot -im@xmath0 along the mdc dispersion shown in panel ( b ) . at this point
, it is worth checking the validity of the new result .
one way of doing so is to compare the im@xmath0 obtained by the new method with the value from a conventional method ( if such a method can be applied ) . in the case of ccoc
, a conventional way of measuring the half width of a quasiparticle - like peak may be applicable only to the features with the lowest binding energy . in the inset
, we plot the edc of the feature that has the lowest binding energy of 0.4 ev , hence is sharpest .
the half width of the feature is about 150 mev .
im@xmath0 at 0.4 ev indeed shows a similar value .
this fact gives us confidence in the new method .
looking at im@xmath0 curve in panel ( c ) , one sees bendsat about 0.3 and 1 ev binding energies .
0.3 ev is where the spectral weight of the lowest binding energy feature starts as can be seen from fig .
therefore , the bend at 0.3 ev may not mean any coupling energy .
meanwhile the bend at 1 ev is in the middle of the dispersion .
interestingly , 1 ev binding energy is where the kink - like dispersion appears to have bosonic modecoupling .
when the kink - like feature from ccoc was first reported@xcite , the most obvious problem with the interpretation of it in terms of bosonic mode coupling was how an insulator can have bosonic mode coupling .
strictly speaking , an insulator can have bosonic mode coupling and should display kink - like features even though the effect would be small@xcite .
even though it does not show the origin , our analysis firmly shows that im@xmath0 behaves as if there is a bosonic mode coupling at 1 ev ( or 0.6 ev from the top of the band ) .
this value is comparable to the energy found in doped htscs@xcite . in fig .
2(d ) , we plot the mdc dispersion of the high energy features and its extrapolation as the dashed line .
this has a strong resemblance to the lda band and may serve as the bare band@xcite . also plotted in the panel is the @xmath5 obtained by calculating the center of the gravity of the spectral function . in comparing the two
, one sees very little resemblance between them .
the deviation of @xmath5 from the bare band near @xmath47 ( where @xmath48 ) is from the reason discussed above .
meanwhile , we believe the difference near @xmath49 , where we expect @xmath50 to be closer to 1 , is from the fact that we cut out the spectral weight at high binding energies under the main valence band@xcite during the background subtraction .
= 8.5 cm an important question is whether the self energy defined only for the electron removal process is meaningful .
we argue that it still bares useful information .
first of all , the self energy defined in this way is identical to the usual @xmath6 when @xmath51 .
this is important because for most of the cases one analyzes the self energy for the occupied parts of the momentum space where @xmath52 .
in addition , we find that im@xmath0 measures the widthof spectral function fairly well in various cases as illustrated below .
lastly , on the practical side , the method can also be applied when the dispersion is flat and therefore mdcs can not be obtained .
this is very useful when the dispersion does not cross the fermi energy as in insulators . to discuss
if im@xmath0 indeed measures the width of the occupied part of a spectral function well , we use numerical simulations as illustrated in fig .
the first case , shown in panel ( a ) , is when the spectral weight is truncated by the fermi function . from the results plotted in panel ( b ) ,
we find that im@xmath0 still measures the width of the peak fairly well below the fermi energy . we may also discuss cases where there is a gap at the fermi energy and the spectral function splits into two due to , for example , charge density wave or mott transition .
to simulate such cases , we use two lorentzian peaks as shown in panel ( c ) . from the self energies plotted in panel ( d ) , we find that im@xmath0 is very close to the lorentzian peak width while im@xmath6 has a large value near @xmath1=0 .
these numerical simulation results support the assertion that im@xmath0 measures the width of the occupied part of a spectral feature .
the case illustrated in panels ( c ) and ( d ) is also relevant to the discussion on the behavior of the self energies in the gap region . while im@xmath6 has a large value at @xmath1=0 , im@xmath0 stays at a finite value close to the peak width . our method ,
as it is , may not be used for multi - band systems .
this is due to the fact that states from different bands may have different matrix element @xmath53 .
however , our scheme in principle can be extended to multi - band systems even though it is more difficult .
if it can be extended to multi - band system , our method will find far more applicability and usefulness in the analysis of self energies of many - body systems .
this is at the present left for the future work .
in conclusion , we propose using the electron removal self energy which is found to represent the spectral width very well .
the method has the advantage that one can easily extract the self energy from a spectral function with arbitrary shape , allowing us to avoid making inappropriate assumptions .
application of the new method to experimental data from ccoc reveals momentum independence of @xmath3(*k*,@xmath1 ) which has been an assumption in the conventional analysis .
we expect this new method to be useful in strongly correlated systems where spectral features are often quite broad .
authors would like to thank b. j. kim for a key suggestions in programming and w. meevasana , jun won rhim and jung hoon han for helpful discussions .
this work is supported by the kosef through cscmr and by the kicos through a grant provided by most in m60602000008 - 06e0200 - 00800 .
work at los alamos was performed under the auspices of us doe .
a. lanzara , p. v. bogdanov , x. j. zhoud , s. a. kellar , d. l. feng , e. d. lu , t. yoshida , h. eisaki , a. fujimori , k. kishio , j .-
shinoyama , t. noda , s. uchida , z. hussain , and z .- x .
shen , nature ( london ) * 412 * , 510 ( 2001 ) .
w. meevasana , n. j. c. ingle , d. h. lu , j. r. shi , f. baumberger , k. m. shen , w. s. lee , t. cuk , h. eisaki , t. p. devereaux , n. nagaosa , j. zaanen , and z .- x .
shen , phys .
lett . * 96 * , 157003 ( 2006 ) .
b. p. xie , k. yang , d. w. shen , j. f. zhao , h. w. ou , j. wei , s. y. gu , m. arita , s. qiao , h. namatame , m. tanaguchi , n. kaneko , h. eisaki , z. q. yang , and d. l. feng , phys .
. lett . * 98 * , 147001 ( 2007 ) .
a. a. kordyuk , s.v .
borisenko , v. b. zabolotnyy , j. geck , m. knupfer , j. fink , b. buchner , c. t. lin , b. keimer , h. berger , a.v .
pan , seiki komiya , and yoichi ando , phys .
* 97 * , 017002 ( 2006 ) . | we propose using the self energy defined for the electron removal green s function . starting from the electron removal
green s function , we obtained expressions for the removal self energy @xmath0(*k*,@xmath1 ) that are applicable for non - quasiparticle photoemission spectral functions from a single band system .
our method does not assume momentum independence and produces the self energy in the full * k*-@xmath1 space .
the method is applied to the angle resolved photoemission from ca@xmath2cuo@xmath2cl@xmath2 and the result is found to be compatible with the self energy value from the peak width of sharp features .
the self energy is found to be only weakly * k*-dependent .
in addition , the im@xmath3 shows a maximum at around 1 ev where the high energy kink is located . |
nodal structures on continuous manifolds have been investigated ever since the days of chladni .
his work was experimental and involved the observation of nodal lines on vibrating plates .
his research was resumed on a more rigorous footing by the pioneering works of sturm @xcite , courant @xcite and pleijel @xcite .
+ in recent years a surge of research has begun on inverse nodal problems , i.e. learning about the geometry of a system by observing its nodal features @xcite .
this research follows what is already known for many years in the regime of inverse spectral problems : one can deduce geometrical information about a system by observing its spectrum . + a key question in the framework of inverse spectral theory
was posed by mark kac who asked ( 1966):can one hear the shape of a drum ?
" @xcite . generally speaking
, this question raises the issue of whether this information is unique . in other words , are there non - congruent systems with the exact same spectrum ? ( these are called isospectral systems ) .
it turns out that the answer to this question is positive .
milnor was the first to show that there are isospectral systems in the case of flat tori in 16 dimensions @xcite .
after him we should mark a few names who contributed significantly to the study of the subject : sunada @xcite ( riemannian manifolds ) , gordon , webb and wolpert @xcite as well as buser _
@xcite ( domains in @xmath0 ) , band _ et al .
_ @xcite ( quantum graphs ) and godsil and mckay @xcite as well as brooks @xcite ( discrete graphs ) . + as a matter of fact , in the context of graphs , gnthard and primas @xcite preceded kac , raising the same question regarding the spectra of graphs with relation to hckel s theory ( 1956 ) .
a year later collatz and sinogowitz presented the first pair of isospectral trees @xcite .
as mentioned , aside from the spectrum , one can also try to mine information from the eigenfunctions of a given system .
today , it is known that there exists geometrical information in the nodal structures and nodal domains of eigenfunctions of manifolds , billiards and graphs @xcite .
furthermore , it is known that this geometrical information is different from the information one can deduce solely by observing the spectrum .
the pioneering work began with gnutzmann _
@xcite , and continued with many other papers , such as @xcite for example .
in particular , gnutzmann _ et al . _
@xcite conjectured that isospectral systems could be differentiated by their nodal domain counts ( we shall refer to it simply as the ` conjecture ' throughout the paper ) .
this conjecture has proven to be quite a strong one with many numerical and analytical evidence to back it up .
in particular , in the case of graphs , both quantum and discrete there exist much numerical evidence as well as rigorous proofs for the validity of the aforementioned conjecture , see for example @xcite .
in addition , the conjecture was proven to hold for a family of isospectral four dimensional tori @xcite .
however , it was found recently that for a different method of counting , there exist a family of isospectral pairs of flat tori , sharing the same nodal domain counts @xcite .
this serves as a first counter - example to the conjecture . in this paper
, we would like to focus on the conjecture within the context of discrete graphs .
we shall first demonstrate its strength and present some known results .
our main topic , however , is to display the first counter - example to the conjecture . to this end
we will need to broaden our view from the usual operators defined on graphs , to the more general setting of weighted graphs .
in addition we would like to report a peculiarity which involves the discrete graphs of the counter - example .
it turns out that this pair of isospectral ( discrete ) graphs are also isospectral as quantum graphs .
this is intriguing since we have not been able to underatsnd this phenomena , nor could we build this isospectral pair using any of the ( many ) known methods which produce isospectral quantum graphs .
sturm @xcite , and courant @xcite after him , were the first to give analytical results about nodal domain counts on continuous systems . denoting the nodal count sequence by @xmath1 , courant s nodal domain theorem
can be generally phrased as @xmath2 .
+ in 1950 gantmacher and krein @xcite investigated the sign patterns of eigenvectors of tridiagonal graphs , and in the 1970 s fiedler wrote a couple of papers about the sign pattern of eigenvectors of acyclic matrices ( matrices which are defined on trees ) @xcite .
both gantmacher and krein , as well as fiedler did not formulate their findings in the language of nodal domains .
it took almost thirty years for the discrete counterpart of the courant nodal domain theorem to appear .
_ @xcite and davies _ et al .
_ @xcite were the first to discover this analogue , and soon afterwards they were followed by biyikoglu @xcite ( who formulated a nodal domain theorem for trees ) .
recently a lower bound for the nodal count was derived by berkolaiko @xcite .
this bound is given explicitly by @xmath3 , where @xmath4 is the number of independent cycles of the graph .
trees are an extremal class of graphs in the sense that for a given number of vertices , they are the smallest connected graph ( least number of edges ) . for trees , assuming some generic conditions ( which are manifested by the fact that the eigenvectors do not vanish on any of the vertices ) , it was proven that the nodal domain count of the @xmath5 eigenvector of the laplacian matrix has exactly @xmath6 nodal domains @xcite .
therefore all trees ( with the same number of vertices ) share the same nodal domain count sequence .
furthermore , it is known that almost all tree graphs are isospectral @xcite ( meaning that almost any tree has a isospectral mate ) .
this means that we can not resolve the isospectrality using nodal domain counts , when it comes to trees .
this shortcoming of the conjecture is well known , and to the best of our knowledge , occurs only for trees .
+ if we introduce weighted graphs , then there exist two more trivial counter - examples : complete graphs and polygon graphs ( connected graphs in which all vertices have degree 2 ) . in the case of complete weighted graphs ,
the first eigenvector has only one nodal domain and all other eigenvectors have exactly 2 nodal domains .
hence , they are an obvious counter - example .
it should be noted that complete graphs are also extremal in the sense that for a given number of vertices , they are the largest connected graph ( largest number of edges ) . for polygons , it can be shown ( using the courant bound @xcite and berkolaiko s bound @xcite ) that polygons always have the same nodal count . as far as the authors know , these three cases are the only counter - examples to the conjecture . aside from this extreme cases , in all isospectral graph pairs which were compared ( analytically and numerically ) , different nodal domain sequences were observed @xcite .
in addition we have a proof for the conjecture for a certain class of discrete graphs @xcite . up until now , we only discussed isospectrality of the traditional matrices defined on graphs , most notably the adjacency matrix and the laplacian .
additional work was done on less studied matrices such as the signless laplacian and the normalized laplacian .
+ however , since nodal domain theorems were proven for a more general class of matrices ( generalized laplacians ) , it is natural to test the conjecture for this class as well .
the paper is organized as follows .
we will begin with some background and necessary definitions .
the following section will describe the method of construction of isospectral weighted graphs .
then we will present the counter - example to the conjecture and finally prove the isospectrality of the quantum analogue of our discrete graphs .
a _ graph _ @xmath7 is a set @xmath8 of vertices connected by a set @xmath9 of edges .
the number of vertices is denoted by @xmath10 and the number of edges is @xmath11 .
the _ degree _ ( _ valency _ ) of a vertex is the number of edges which are connected to it .
the number of independent cycles of a graph is denoted by @xmath4 and is given by @xmath12 , where @xmath13 is the number of connected components of the graph .
the _ weighted adjacency matrix _ ( _ connectivity _ ) of @xmath7 is the symmetric @xmath14 matrix @xmath15 whose entries are given by : @xmath16 the @xmath17 s values are called weights and are usually taken to be positive . for non - weighted graphs , all the weights are equal to unity .
a diagonal element in @xmath18 corresponds to a loop , which is an edge connecting a vertex to itself .
we shall only discuss graphs without loops .
+ a _ generalized laplacian _
, @xmath19 , also known as a _ schrdinger operator _ of @xmath7 , is a matrix @xmath20 where @xmath21 is an arbitrary on - site potential which can assume any real value and @xmath22 . the _ combinatorial laplacian _ results by taking all weights to be unity , and @xmath23 , where @xmath24 is the degree of the vertex @xmath25 .
this way , the sum of each row , or column is equal to zero .
+ the eigenvalues of @xmath19 together with their multiplicities , are known as the _ spectrum _ of @xmath7 . to the @xmath5 eigenvalue , @xmath26 ,
corresponds ( at least one ) eigenvector whose entries are labeled by the vertex indices , i.e. , @xmath27 .
nodal domain _ is a maximally connected subgraph of @xmath7 such that all vertices have the same sign with respect to @xmath28 .
the number of nodal domains of an eigenvector @xmath28 is called a _ nodal domain count _ , and will be denoted by @xmath29 .
the _ nodal count sequence _ of a graph is the number of nodal domains of eigenvectors of the laplacian , arranged by increasing eigenvalues .
this sequence will be denoted by @xmath30 .
+ we recall that the known bounds for the nodal count @xcite are @xmath31 to define _ quantum graphs _ a metric is associated to @xmath7 .
that is , each edge is assigned a positive length : @xmath32 .
the total length of the graph will be denoted by @xmath33 .
this enables to define the metric laplacian ( or schrdinger ) operator on the graph as the laplacian in 1-d @xmath34 on each bond .
the domain of the schrdinger operator on the graph is the space of functions which belong to the sobolev space @xmath35 on each edge @xmath36 and satisfy certain vertex conditions .
these vertex conditions involve vertex values of functions and their derivatives , and they are imposed to render the operator self adjoint .
we shall consider in this paper only the so - called neumann vertex conditions : @xmath37 where @xmath38 is the set of all edges connected to the vertex @xmath39 .
the derivatives in ( [ eq : boundary_cond ] ) are directed out of the vertex @xmath39 .
the eigenfunctions are the solutions of the edge schrdinger equations : @xmath40 which satisfy at each vertex the neumann conditions ( [ eq : boundary_cond ] ) .
the spectrum @xmath41 is discrete , non - negative and unbounded .
one can generalize the schrdinger operator by including potential and magnetic flux defined on the bonds .
other forms of vertex conditions can also be used . however
, these generalizations will not be addressed here , and the interested reader is referred to two recent reviews @xcite . finally , two graphs , @xmath42 and @xmath43 , are said to be _ isospectral _ if they posses the same spectrum ( same eigenvalues with the same multiplicities ) . in perfect analogy ,
two graphs with the same nodal domain sequence will be referred to as _ isonodal_. these two definitions hold both for discrete and quantum graphs .
our method for constructing isospectral graphs is a variation of a method described in @xcite , called the _ line graph construction_. this method uses the gallery of isospectral billiards of buser _ et al .
_ @xcite in order to build isopectral discrete graphs .
a similar idea was used by gutkin _
@xcite to construct isospectral discrete and quantum graphs . + a line graph is built from a parent " graph in the following way : each edge becomes a vertex , and two vertices in the line graph are adjacent if and only if their corresponding edges shared a vertex in the parent graph . in @xcite an example is given , based on the first family of isospectral domains in @xcite called the @xmath44 family .
our method is simpler than the one in @xcite .
it results with graphs with the same topology as in @xcite , but with different laplacian matrices . instead of using the gallery of billiards
as it appears in @xcite , we use a graph representation of them as it is described in @xcite .
in particular , the @xmath45 family is shown in figure [ fig:7 - 1 pair as graphs ] .
we consider the two graphs in figure [ fig:7 - 1 pair as graphs ] as parent graphs and apply the line graph construction on them .
we still have to specify how we assign weights in the resulting line graphs .
we start by assigning three different weights : @xmath46 to each of the three types of edges in the parent graphs .
suppose that in the parent graph , an edge of weight @xmath47 shared a vertex with an edge of weight @xmath48 ( @xmath49 ) .
then , in the line graph , the corresponding vertices would be connected by an edge of the remaining weight @xmath50 .
+ the two resulting weighted line graphs are shown in figure [ fig:7 - 1 line graphs ] .
let us denote the left graph by @xmath42 and the right one by @xmath43 .
the generalized laplacians of the two graphs are given explicitly by following matrices : @xmath51 it is not hard to check that for any @xmath52 , the characteristic polynomials of @xmath53 and @xmath54 are identical and hence the graphs are isospectral .
another way to prove the isospectrality is to construct the transplantation matrix @xmath55 such that @xmath56 .
then it is clear that the two matrices are similar and therefore isospectral .
the transplantation matrix between @xmath53 and @xmath54 is @xmath57 the same construction can be carried out for any graph in the gallery of @xcite .
+ we can construct many more isospectral graphs by using polynomials in @xmath53 and @xmath54 .
namely , for any polynomial @xmath58 , we will consider @xmath59 and @xmath60 as the laplacian matrices of two new weighted graphs ( assuming that @xmath59 and @xmath60 are indeed generalized laplacians as defined in section [ subsec : discrete graphs ] ) .
these two graphs might be topologically different than the original @xmath42 and @xmath43 . since we have a transplantation matrix ,
it is clear that @xmath59 and @xmath60 are similar matrices and therefore the resulting graphs are also isospectral .
we have introduced the conjecture that the isospectrality between graphs can be resolved by counting nodal domains .
we have also said that three known cases ( trees , polygons and complete graphs ) are exceptions to this conjecture .
we now prove that @xmath42 and @xmath43 can not be resolved by counting nodal domains .
this is a non - trivial exception to the conjecture .
we define the vertices with degree larger than one as the _ interior vertices _ ( vertices @xmath61 ) , and the rest as _ boundary vertices _ ( vertices @xmath62 ) . + we begin by checking the relations between the interior and boundary vertices . + let @xmath63 be the @xmath5 eigenfunction of @xmath53 , and @xmath64 be the @xmath5 eigenfunction of @xmath54 .
for the first graph , we get the following relations : @xmath65 for the second graph , we get the same relations with @xmath48 and @xmath66 replaced . therefore , since the weights are positive , if @xmath67 then each boundary vertex has the same sign as the interior vertex connected to it .
this means that for @xmath67 , the boundary vertices will not contribute to the nodal domain count . on the other hand , if @xmath68 , each boundary vertex has an opposite sign than the interior vertex connected to it .
this means that for @xmath68 , the boundary vertices will contribute three to the nodal domain count .
the most important point is that the contribution of the boundary vertices to the nodal count depends on the spectrum , and since the two graphs are isospectral , it is the same for both graphs . as a result ,
it is enough to compare only the nodal count sequence of the interior vertices .
+ the interior vertices form a triangle
. therefore the nodal domain count of any vector , on the subgraph induced by the interior vertices , is either one or two . by computing the rest of the equations , and with the aid of ( [ eq : relation between interior and boundary vertices ] )
, we can formulate the conditions for having one or two nodal domains .
+ the interior nodal domain count of a vector @xmath69 ( of any of the graphs ) is one if and only if one of the following is true : @xmath70 in any other case , the nodal domain count is two .
+ in other words , a nodal domain count of a specific vector , is determined uniquely by the corresponding eigenvalue .
therefore , the entire nodal domain counts of the graphs are determined by the spectrum , and since the two graphs are isospectral , the nodal count sequence does not resolve the isospectrality .
+ as we have shown in subsection [ subsec : iso ] , for any polynomial @xmath58 , the two graphs represented by @xmath59 and @xmath60 are isospectral .
we will now show that these graphs are also isonodal , thus extending our family of counter - examples to the conjecture .
+ assuming that the weights @xmath52 are rationally incommensurate , the following observations can be easily proven : * if the polynomial consists only of a second degree term ( @xmath71 ) , then the obtained graphs @xmath59 and @xmath60 are given by figure [ fig : second order graph ] .
* if @xmath72 , then the obtained graphs @xmath59 and @xmath60 are given by figure [ fig : first and second order graph ] .
* polynomials of third degree or larger represent weighted , complete graphs , which are trivial counter - examples to the conjecture ( see section [ subsec : nd theorems ] ) . for these reasons , we only need to check the first two cases above .
+ the resulting graphs , in both cases , clearly have the same eigenvectors as @xmath42 and @xmath43 .
then , using and , it can be easily shown that in both types of graphs , the nodal count is determined by the spectrum , precisely as is the case with @xmath42 and @xmath43 .
+ we conclude that both types of graphs are isonodal and as a consequence , they are also non - trivial counter - examples to the conjecture . _
remark : _ applying the line graph construction to other families from the gallery in buser _
@xcite , one can build many pairs of isospectral graphs .
some of these pairs are isonodal ( such as the @xmath73 and @xmath74 families ) and some are not ( such as the @xmath75 family ) .
when we come to discuss quantum graphs , we need to define the lengths of the different edges .
the weights we put on the weighted discrete graphs can be viewed as coupling constants .
thus , the most intuitive notion is to associate a length which is inversely proportional to the weights . if we also specify the vertex conditions , we go from the realm of discrete graphs into the realm of quantum graphs .
+ we then come to ask the following interesting question : is the isospectrality preserved when we enter the world of quantum graphs ? this question is only a small part of a much broader subject - the spectral relations between quantum graphs and the underlying discrete graphs .
this subject was addressed by several authors in the past , see for example @xcite .
however , most of these references have a complete analysis only for equilateral quantum graphs , with neumann vertex conditions .
the graphs @xmath42 and @xmath43 are _ not _ equilateral , and therefore we can not make an a - priori prediction whether or not the isospectrality is preserved .
nevertheless , we will show by direct computation that @xmath42 and @xmath43 are indeed isospectral as quantum graphs , once neumann vertex conditions are considered at all vertices . +
a function @xmath76 on the graph , which is continuous on the vertices , can be written as : @xmath77 , \label{wave function}\ ] ] where @xmath78 is the value of the function on the vertex @xmath25 and @xmath79 is the length of the edge @xmath80 .
note , that we still use the notations @xmath52 to denote the lengths of the edges .
+ the neumann vertex conditions on the boundary vertices for @xmath42 , dictate these relations : @xmath81 the neumann conditions on the interior vertices are ( we make use of ( [ neumann graph 1 ] ) ) : @xmath82+\frac{1}{\sin{kc}}\left[-\phi(4)\cos{kc}+\phi(5)\right ] + \frac{1}{\sin{kb}}\left[-\phi(4)\cos{kb}+\phi(6)\right]=0 \label{kirchoff graph 1:1}\ ] ] @xmath83+\frac{1}{\sin{kc}}\left[\phi(5)\cos{kc}-\phi(4)\right ] -\frac{1}{\sin{ka}}\left[-\phi(5)\cos{ka}+\phi(6)\right]=0 \label{kirchoff graph 1:2}\ ] ] @xmath84+\frac{1}{\sin{kb}}\left[\phi(6)\cos{kb}-\phi(4)\right ] + \frac{1}{\sin{ka}}\left[\phi(6)\cos{ka}-\phi(5)\right]=0 \label{kirchoff graph 1:3}\ ] ] this can be written more conveniently as a matrix - vector product : @xmath85 where @xmath86 comes to represent that this is the matrix corresponding to @xmath42 , and @xmath87 . + the matrix @xmath88 is : @xmath89 ( [ matrix of secular equation ] ) has a solution if and only if @xmath90 @xmath91 is called the _ secular function _ , and equation ( [ secular equation ] ) is called the _ secular equation_. it is fulfilled at the values @xmath92 which are in the spectrum of the laplacian of the graph .
we can get @xmath93 by switching the lengths @xmath48 and @xmath66 in @xmath88 .
it can be easily checked that @xmath94 , hence the graphs are isospectral .
+ although we have proven that @xmath42 and @xmath43 are isospectral as quantum graphs , the profound reason for this is still a riddle for us .
the recent papers on isospectrality @xcite generalize former seminal papers such as those of sunada @xcite and buser _ et al . _
@xcite , and can produce many of the known examples of isopectral quantum graphs . however , we were not able to build the two graphs @xmath42 and @xmath43 using the constructions described in @xcite . furthermore , we were unable to build a transplantation matrix for the quantum graphs ( although there is a transplantation matrix for the discrete case - see ( [ transplantation matrix 7_1 discrete ] ) ) .
it should be emphasized that all isospectral quantum graphs which are built using any of the methods in @xcite posses a transplantation matrix between the two graphs . in @xcite ,
the authors consider the two graphs in the present paper and turn them into scattering systems .
they prove that there is no transplantation which involves the values of the eigenfunctions on the vertices .
they do not , however , eliminate the possibility of having any other form of transplantation .
all these pieces of evidence suggest that @xmath42 and @xmath43 might belong to a new class of isospectral quantum graphs . _
remark : _ unlike the graphs @xmath42 and @xmath43 which correspond the the @xmath45 family , the isospectrality is not preserved in the @xmath73 and @xmath74 graphs ( i.e. , the corresponding quantum graphs are not isospectral ) .
this leads us to contemplate the issue of converting isospectral weighted discrete graphs into their isospectral quantum analogues .
how to do so , or whether at all it is possible , remains an open problem .
the conjecture that isospectrality can be lifted by comparing nodal domain counts was originally stated for flat tori of dimension larger than three @xcite . later on ,
this conjecture was proven for four - dimensional flat tori @xcite . however , using a different counting method , a family of both isosepctral and isonodal pairs of flat tori was discovered @xcite .
the conjecture was imported into the realm of graphs where it was proven for some quantum and discrete graphs @xcite .
in addition , there exist much numerical evidence for the validity of the conjecture in discrete graphs @xcite ( using a construction by godsil and mckay @xcite ) .
+ in this paper we show that for discrete graphs , the conjecture is not true in its most general form .
what we demonstrate is that if we use generalized laplacians , the conjecture ceases to be valid even for graphs which are not extremal .
one should keep in mind that if we restrict ourselves only to the traditional matrices - the adjacency and laplacian matrices - then the only known counter - examples to the conjecture are trees .
the paper also presents an intriguing connection between isospectral discrete and quantum graphs .
the fact that both the discrete graphs , and their quantum analogues are isospectral , calls for more study on the relation between these two regimes .
the authors warmly thank uzy smilansky for his continuous support , significant encouragement and for many invaluable discussions .
the work was supported by isf grant 169/09 .
rb is supported by epsrc , grant number ep / h028803/1 .
10 f. r. gantmacher and m. g. krein , _ oscillation matrices and kernels and small vibrations of mechanical systems _ , state publishing house of technical - theoretical literature , moscow , leningrad , 1950 .
english translation by us atomic energy commission , washington d. c. 1961 .
p. kuchment , _ quantum graphs : an introduction and a brief survey _ , in analysis on graphs and its applications , volume * 77 * of proc .
sympos . pure math . ,
soc . , providence , ri , 291 - 312 , 2008 .
p. mcdonald and r. meyers , _ isospectral polygons , planar graphs and heat content _ , proceedings of the american mathematical society , vol . 131 , number 11 , pages 3589 - 3599 s 0002 - 9939(03)07123 - 5 article electronically published on june 18 , 2003 . f.a .
mehmeti , j. von below , and s. nicaise ( editors ) , _ partial differential equations on multistructures _ , proceedings of the international conference held in luminy , april 1924 , 1999 , lecture notes in pure and applied mathematics , 219 .
marcel dekker , inc .
, new york , 2001 . | according to a recent conjecture , isospectral objects have different nodal count sequences @xcite .
we study generalized laplacians on discrete graphs , and use them to construct the first non - trivial counter - examples to this conjecture .
+ in addition , these examples demonstrate a surprising connection between isospectral discrete and quantum graphs . |
the measurement of electron - positron pairs ( dielectrons ) in the low invariant mass region allows studying the vacuum and in - medium properties of light vector mesons .
additionally , low - mass dielectrons are produced by internal conversion of virtual direct photons .
they are excellent probes to study all collision stages , since they pass through the created medium almost unaffected . to quantify modifications of the dielectron production in heavy - ion collisions , measurements in pp collisions serve as a reference , while the analysis of p - a collisions allows disentangling cold from hot nuclear matter effects . in alice @xcite at the lhc
, dielectron measurements are performed using the central barrel detectors around mid - rapidity .
electrons can be identified via their specific energy loss in the inner tracking system ( its ) and the time projection chamber ( tpc ) , combined with time - of - flight information from the tof detector @xcite . in this proceedings
we present a virtual direct photon measurement in pp collisions and the invariant mass analyses in p pb and pb pb collisions .
prospects of a future measurement in pb pb after the alice upgrade for lhc run 3 are also discussed .
the analysis presented here is based on @xmath1 minimum bias pp collisions at @xmath0 , recorded in 2010 .
fiducial cuts on transverse momentum ( @xmath2 ) and pseudorapidity ( @xmath3 ) are applied to electron candidates .
a clean electron sample is achieved by cuts on the time - of - flight and on the @xmath4 in the tpc .
dielectron spectra are created for unlike - sign and like - sign combinations of these particles .
the unlike - sign distribution contains a superposition of signal and combinatorial background , while the like - sign is used to provide an estimate of such a background .
the acceptance may be different for unlike- and like - sign pairs due to detector effects .
therefore , the signal yield is obtained as @xmath5 , where @xmath6 is built via a mixed - event technique as @xmath7 .
+ figure [ pp - fit ] shows the differential dielectron cross - section in a kinematic region @xmath8 which is useful to study direct photon production .
also shown in fig .
[ pp - fit ] are a hadronic cocktail and its components as well as the expected mass distribution of dielectrons coming from virtual direct photons , after having applied the single - electron fiducial cuts on @xmath9 and @xmath10 .
a fit to the data is performed to determine the virtual photon fraction @xmath11 by using the function @xmath12 in the range @xmath13 .
this fit is done in four pair-@xmath9 regions and the extracted virtual photon fraction is then multiplied by the inclusive photon cross section , also measured by alice via the photon conversion method ( pcm ) @xcite . in figure
[ pp - photons ] the resulting direct photon cross section is compared to nlo pqcd calculations .
reasonable agreement is found , within uncertainties , between data and model .
details on this analysis can be found in @xcite .
direct photon cross section compared to nlo pqcd calculations , and the inclusive photon cross section extracted from pcm . ]
direct photon cross section compared to nlo pqcd calculations , and the inclusive photon cross section extracted from pcm . ]
the analyzed data sample consists of @xmath15 minimum bias p pb collisions .
the analysis is carried out in @xmath3 and two different track selections @xmath2 or @xmath16 are performed , to enable comparisons with both the pp and pb pb data .
the pid scheme was modified to increase the electron efficiency : a tof signal is not required , yet used if available to purify the electron sample .
kaons and protons are instead rejected by an electron inclusion cut on the @xmath4 measured by the silicon drift and silicon pixel detectors of the its .
in addition , electrons from photon conversions are identified and rejected either by their displaced vertex or by the orientation of the pair plane relative to the magnetic field .
more in detail , since a conversion pair does not have an intrinsic opening angle , an apparent opening after reconstruction may only appear perpendicular to the magnetic field vector , while true finite - mass signal pairs have a uniform azimuthal opening angle .
therefore , an exclusion cut on the pair plane normal vector @xmath17 is applied for @xmath18 .
+ the signal extraction is performed as described in section [ sect - pp ] .
the signal - to - background ratio is smaller than in pp data by a factor two to four , and range from @xmath19 in the @xmath20 mass region to @xmath21 at @xmath22 and to @xmath23 in the @xmath24 region .
the corrected dielectron yield for @xmath2 and integrated over pair-@xmath9 , is shown in fig .
[ p - pb - mee ] .
the data points and systematic uncertainties are extracted from the mean values and the spread of results obtained with @xmath25 different combinations of analysis cut settings ( inspired by @xcite ) .
the data are compared to the hadronic cocktail , which uses the charged pion measurement @xcite as @xmath20 input and @xmath26-scaling for the other light mesons .
open heavy - flavour and @xmath24 contributions are calculated from pythia , tuned to independent alice measurements in pp and p pb . uncertainties on the input of the cocktail sources are propagated to the shapes shown in fig .
[ p - pb - mee ] .
data and cocktail are in reasonable agreement within their uncertainties over the full mass range . dielectron raw subtracted yield in central and semi - central pb
pb collisions after background subtraction . ]
dielectron raw subtracted yield in central and semi - central pb
pb collisions after background subtraction . ]
from the 2011 pb pb run , @xmath28 and @xmath29 events in central ( 0 - 10% ) and semi - central ( 20 - 50% ) pb pb collisions , respectively , are selected for this analysis .
the pid scheme is as described for the p pb analysis . a track selection cut @xmath30 ensures exclusion of charged pions and reduces the contribution from soft conversion and @xmath20-dalitz electrons to the combinatorial background .
a residual hadron contamination at the @xmath31 level also contributes to the background pairs but monte - carlo studies show that it does not distort the signal . a conversion rejection cut @xmath32 is applied for @xmath33 .
the signal - to - background ratio for @xmath34 decreases from @xmath35 for 20 - 50% to @xmath36 for 0 - 10% centrality .
the combinatorial - background - subtracted uncorrected yields ( _ raw subtracted yield _ ) are shown in fig .
[ pb - pb - mee ] in the low - mass region for @xmath37 for both centralities .
efficiency correction and cocktail comparison are in progress .
a fine - binning differential low - mass dielectron measurement in pb
pb collisions is a major physics case for the alice upgrade for lhc run 3 @xcite .
we evaluated to which precision a possible thermal excess yield @xcite could be reconstructed in central and peripheral pb
pb collisions at full lhc energy ( @xmath38 5.5 tev ) .
electron tracking and pid efficiencies from a monte - carlo simulation using the current alice setup , operated at reduced solenoid magnetic field , are used and modified according to the expected performance after the upgrade .
a statistical significance is calculated from realistically constructed signal and background spectra , and is then used for poisson sampling of data points around the model input .
figures [ excess - curr ] and [ excess - upg ] show the expected spectra after subtraction of hadronic sources , for @xmath39 and @xmath40 events using the current and upgraded setup , respectively . while the measurement is dominated by uncertainties in the first case , the upgrade is expected to allow for a temperature measurement in the intermediate mass region ( imr ) within @xmath41 statistical and @xmath42 systematic uncertainty @xcite .
same as fig .
[ excess - curr ] , but with upgraded its and tpc .
reduced statistical and systematic uncertainties make a temperature measurement in the imr feasible . ]
same as fig .
[ excess - curr ] , but with upgraded its and tpc .
reduced statistical and systematic uncertainties make a temperature measurement in the imr feasible . ]
virtual direct photons are observed in pp collisions at @xmath8 .
the extracted direct photon cross section agrees with nlo pqcd calculations within uncertainties . the pair-@xmath9-integrated invariant mass spectrum measured in p
pb collisions is in agreement with the expected hadronic sources . from pb
pb collisions , raw subtracted yields have been extracted in two centrality classes and a further analysis is ongoing .
a feasibility study indicates that the alice upgrade for lhc run 3 will facilitate a measurement of the early effective temperature .
9 k. aamodt et al .
( alice collaboration ) , _ jinst _ * 3 * ( 2008 ) s08002 .
b. abelev et al .
( alice collaboration ) , _ int .
phys . _ a * 29 * ( 2014 ) 1430044 .
m. wilde ( for the alice collaboration ) , _ arxiv:_1210.5958 ( 2012 ) .
m. k. khler , doctoral thesis , tu darmstadt ( 2014 ) .
r. barlow , _
arxiv:_hep - ex/0207026 ( 2002 ) .
b. abelev et al .
( alice collaboration ) , _ phys .
_ b * 728 * ( 2014 ) 25 .
b. abelev et al .
( alice collaboration ) , cern - lhcc-2012 - 012 / lhcc - i-022 ( 2012 ) .
r. rapp and j. wambach , _ eur .
j. _ a * 6 * ( 1999 ) 415 and private communication .
b. abelev et al .
( alice collaboration ) , _ j. phys .
_ g * 41 * ( 2014 ) 087002 . | the alice collaboration measures the production of low - mass dielectrons in pp , p pb and pb pb collisions at the lhc .
the main detectors used in the analyses are the inner tracking system , time projection chamber and time - of - flight detector , all located around mid - rapidity .
the production of virtual photons relative to the inclusive yield in pp collisions is determined by analyzing the dielectron excess with respect to the expected hadronic sources .
the direct photon cross section is then calculated and found to be in agreement with nlo pqcd calculations .
results from the invariant mass analysis in p
pb collisions show an overall agreement between data and hadronic cocktail . in pb
pb collisions , uncorrected background - subtracted yields have been extracted in two centrality classes . a feasibility study for lhc
run 3 after the alice upgrade indicates the possibility for a future measurement of the early effective temperature . |
the eclipsing binary systems containing classical cepheids are very rare .
although we know tens of binary cepheids in our galaxy ( e.g. evans @xcite ) , none is an eclipsing system .
however extensive search for variability in microlensing surveys in magellanic clouds resulted in reporting of reliable eclipsing binary candidates : a double cepheid system macho 05:21:54.8 - 69:21:50 ( alcock @xcite ) , first overtone cepheid ogle lmc sc16 119952 ( udalski @xcite ) and two fundamental mode cepheids ogle - lmc - cep-0227 and ogle - lmc - cep-1812 ( soszyski @xcite ) .
deriving fundamental physical parameters of classical cepheids is an important task regarding their great significance for extragalactic distance scale . as part of the araucaria project we made
follow up of candidate eclipsing binaries in the large magellanic cloud ( lmc ) .
this allowed us to confirm first known eclipsing binary cepheid ( pietrzyski @xcite ) and to resolve so called the mass discrepancy problem in a favour of cepheid pulsation models .
our goals are summarised as follow : 1 ) determination of accurate dynamical masses of classical cepheids ; 2 ) verification of the mean cepheid radius - pulsation period relation ; 3 ) developing of a consistent scheme for solving light and radial velocity curves of radially pulsating star in an eclipsing binary system ; 4 ) determination of distance independent the projection factor for cepheids ; 5 ) analysis of the limb darkening dependence on a pulsation phase .
below we presented some results and details of our cepheids research .
the sections are devoted to particular eclipsing binary systems .
the figure [ fig1 ] presents position of our target eclipsing binary cepheids within the lmc body .
this relatively bright star ( v = 15.2 mag ) was discovered to be a classical fundamental mode cepheid by soszyski ( @xcite ) . in the same paper
the cepheid was mentioned as possible eclipsing binary candidate .
we confirmed it to be an eccentric double lined binary with orbital period of 309 days and deep eclipses ( pietrzyski @xcite ) .
we derived physical parameters of both components and , especially , their precise masses .
subsequently we performed extensive optical and near infrared photometric and spectroscopic monitoring of this star to improve our solution . to this end
we developed a novel method of light curve and radial velocity curve analysis .
the new code and data allowed us to put stringent limits onto cepheid physical parameters ( pilecki @xcite ) .
the quality of the fit is illustrated in fig .
[ fig2 ] . derived masses and radii are fully consistent with a solution reported by pietrzyski ( @xcite ) but by a factor of two more precise .
we detected anomalously large limb darkening in the pulsating component in disagreement with predictions from static atmosphere models .
for more we calculated the projection factor of this cepheid which is an important parameter in baade - wesselink type methods of distance determination to radially pulsating stars .
this direct determination is a distance and a limb darkening independent what differs our result from those obtained for galactic cepheids from interferometry ( e.g. mrand @xcite ) .
the value we obtained remain in agreement with some theoretical model predictions ( e.g. nardetto @xcite ) .
fundamental parameters of the cepheid component are given in table [ tab1 ] .
the companion star has almost the same mass and luminosity like the cepheid .
the star was confirmed to be an eccentric double lined eclipsing binary with an orbital period of 552 days by pietrzyski ( @xcite ) .
the pulsating component is a fundamental mode classical cepheid having shorter pulsation period than cep-0227 - see table [ tab1 ] . however during analysis of this system we detected non - negligible third light .
first hint of it came from spectroscopy as we observed an additional peak in the broadening function power distribution - fig .
the peak is not observed on all spectra suggesting that the third light comes from an optical blend . in our analysis
we assumed that the light contribution from this blend is about 10% of the total flux in i - band .
the presence of the optical blend(s ) was confirmed later through hst multicolour imaging .
as the orbital inclination of the system is very close to 90 degrees the presence of the third light does not affect significantly masses reported by pietrzyski ( @xcite ) .
thus the cepheid component is only the second one with mass known with accuracy better than 3% .
however radii and luminosities of both component are a subject of some uncertainty up to 5 - 10% .
this system is somewhat puzzling because the cepheid has much larger mass ( and luminosity ) than the companion star , however we meet both these stars in a short evolutionary phase of giant evolution .
currently we work on improving our previous solution and deriving the p - factor for this cepheid .
this star contains a first overtone cepheid in a eccentric double lined eclipsing system having an orbital period of 800 days .
we confirm here that both stars are gravitationally bounded by detecting their mutual orbital motion - see fig [ fig4 ] . because of the eccentricity and a position of the orbit we observe only one minimum in this system when a red giant companion is eclipsing the cepheid star .
our photometric campaign to detect secondary minimum around expected orbital phase 0.55 gave null result .
it is interesting that the companion star is in fact more massive and more luminous than the cepheid .
star lmc562.05.9009 is reported by soszyski @xcite as candidate eclipsing binary cepheid based on results from the gaia south ecliptic pole search by ogle - iv survey .
the cepheid star pulsates in fundamental mode .
we monitor this star extensively with spectroscopy and photometry . up to now
we observed only two deep eclipses of comparable depth ( primary and secondary minimum ) and thus we can not calculate orbital period .
we collected some spectra which show two strong peaks in the broadening function power ( fig .
[ fig5 ] ) suggesting that this is a true binary .
star lmc506.14.8910 was found to be a candidate eclipsing binary cepheid during search for variable stars in ogle - iv survey .
the cepheid also pulsates in fundamental mode , however eclipses are shallow .
especially we see only a trace of a possible secondary eclipse at orbital phase 0.53 - see fig .
[ fig5 ] . at the moment
we have no information about a presence of a companion star in spectra .
.physical parameters of the lmc classical cepheids in eclipsing binaries [ cols="^,>,>,>,>,>",options="header " , ] _ notes : _
+ @xmath0 assuming the orbital inclination of 90 deg .
+ @xmath1 ? - not directly determined yet .
during a few last years we were able to significantly improve the knowledge about physical parameters of classical cepheids by observing and analysing two eclipsing binary cepheids in the lmc .
at least one more eclipsing binary system ogle-2532 is confirmed .
our future plans consist of 1 ) performing similar analysis for cep-1812 like we did in the case of cep-0227 , 2 ) follow up of two new candidate eclipsing binaries , 3 ) confirmation of binary nature of double eclipsing cepheid ogle - lmc - cep-1718 and 4 ) improving our understanding of the baade - wesselink method applied for cepheids .
we gratefully acknowledge financial support for this work from the polish national science center grant maestro dec-2012/06/a / st9/00269 . 99 alcock , c. , ( macho ) 1995 , aj , 109 , 1653 evans , r. n. , bond , h. e. , schaefer , g. h. , mason , b. d. , karovska , m. and tingle , e. 2013 , aj , 146 , 93 mrand , a. , kervella , p. , coud de foresto , v. 2005 , a&a , 439 , l9 nardetto , n. , gieren , w. , kervella , p. , 2009 ,
a&a , 502 , 951 pietrzyski , g. , thompson , i. b. , gieren , w. , graczyk , d. , bono , g. , 2010 , nature , 468 , 542 pietrzyski , g. , thompson , i. b. , graczyk , d. , gieren , w. , pilecki , b. , udalski , a. , soszyski , i. , bono , g. , konorski , p. , nardetto , n. and storm , j. 2011 , apj , 742 , 20 pilecki , b. , graczyk , d. , pietrzyski , g. , gieren , w. , thompson , i. b. , 2013 , to appear in mnras , arxiv:1308.5023 pojmaski , g. 2002 , acta astron . , 52 , 397 soszyski , i. , udalski , a. , szymaski , m. k. , kubiak , m. , pietrzyski , g. , wyrzykowski , . , szewczyk , o. , ulaczyk , k. & poleski , r. 2008 , acta astron . , 58 , 293 soszyski , i. , udalski , a. , poleski , r. , et al .
2012 , acta astron . , 62 , 219 udalski , a. , soszyski , i. , szymaski , m. , kubiak , m. , pietrzyski , g. , woniak , p. , and ebru , k. 1999 , acta astron . | the status of our work on binary classical cepheid systems in the large magellanic cloud is presented .
we report on results from our follow up of two eclipsing binary cepheids ogle - lmc - cep-0227 and ogle - lmc - cep-1812 . here
we presented for the first time confirmation that a third cepheid ogle - lmc - cep-2532 is a true eclipsing binary cepheid with a period of 800 days .
two other very good candidates for eclipsing binaries detected during ogle - iv survey are also discussed . |
there are many reasons to study the spin of the nucleon .
an ingenuous reason is that it is interesting in itself : we are curious to know how the spin of the nucleon emerges from its constituents , i.e. , how the spins and orbital angular momenta of the quarks and gluons combine together into the spin 1/2 of the nucleon .
we have also more practical reasons : spin degrees of freedom provide additional tests to theories .
hence , studying spin observables provides a more complete study of the fundamental force ruling the nucleon structure , the strong force , of which the accepted gauge theory is quantum chromodynamics ( qcd ) .
we can also learn about important emergent properties of qcd , such as the mechanism of confinement , how the effective hadronic degrees of freedom stem from the fundamental partonic ones , or how well non - perturbative methods , e.g. , chiral perturbation theory ( @xmath3pt ) , lattice gauge theory or the gravity - gauge ( ads / cft ) correspondence , describe the strong force in its low energy domain .
finally , a precise knowledge of polarized parton distributions is necessary for investigations of physics beyond the standard model or high precision atomic physics .
we have thus many good reasons to be interested in the nucleon spin . in this document
, we will focus on what spin structure studies teach us about confinement , the emergence of the hadronic degrees of freedom and tests of non - perturbative approaches to qcd .
due to space limits , we will discuss results based on inclusive lepton scattering , which are only a fraction of the polarized observables necessary for a full description of the nucleon spin structure .
in fact , we will focus on the first moment of the spin structure function @xmath4 at low to moderate @xmath0 .
these results were obtained with the 6 gev jlab accelerator .
the future spin program with the new 12 gev jlab beam is discussed in v. burkert s contribution to these proceedings .
inclusive polarized lepton scattering data , obtained by experiments in which the only particle of the reaction detected is the scattered lepton , is parametrized ( neglecting the weak interaction ) by four structure functions : two unpolarized ones : @xmath5 and @xmath6 , and two polarized ones : @xmath7 and @xmath8 ( @xmath9 is the bjorken scaling variable and @xmath10 , with @xmath11 the 4-momentum exchanged between the beam particle and the target particle ) . in the bjorken limit , @xmath12 while @xmath13 stays finite , @xmath14 and @xmath15 are constructed from the quark densities @xmath16(@xmath17 ) , @xmath18(@xmath19 ) , @xmath20(@xmath21 ) ... which provide the amount of up , down , strange ... quarks ( antiquarks ) carrying a nucleon momentum fraction @xmath9 , respectively . @xmath14 and @xmath15 are simple sums of these densities : @xmath22+q_d^2[d + \overline d]+ ... $ ] and @xmath23 , with @xmath24 the electric charge of the quark of flavor @xmath25 .
likewise , @xmath26 is constructed from the quark polarizations @xmath27 ... which are , for a given @xmath9 , the difference between the amount of quarks with their spins parallel to the nucleon spin and those with their spins antiparallel : @xmath28+q_d^2[\delta d + \delta \overline d]+ ... $ ] in the bjorken limit the other spin structure function has no role : @xmath29 .
this relatively simple picture is only valid at @xmath30 where the qcd coupling @xmath1 should vanish , i.e. in the asymptotic case where quarks are free . at finite ( but still large ) @xmath0
, quarks start to interact , which leads to gluon corrections .
furthermore , the mass of the target @xmath31 and effects of transverse momentum @xmath32 or transverse spin are less suppressed : @xmath33 , @xmath34 and thus , they start to play a role .
this breaks down the above relations : @xmath35 and @xmath36 . at smaller @xmath0 , ( @xmath37 gev@xmath38 )
, @xmath1 becomes large and the significantly stronger quark interaction induces correlations , known as higher twists ( ht ) effects , which invalidate the simple description summarized above : @xmath39+q_d^2[d + \overline d]+ ... $ ] and @xmath40+q_d^2[\delta d + \delta \overline d]+ ... $ ] . some of the ht effects that contribute to inclusive polarized scattering have been interpreted recently as the transverse confining force acting on the quarks @xcite .
ht have been measured in a number of jefferson lab ( jlab ) experiments in halls a @xcite , b @xcite and c @xcite .
in particular , the ht were extracted in the high-@xmath9 domain in @xcite .
the sane experiment in hall c will further add to these data ( see w. armstrong s contribution to these proceedings ) .
an example of ht extraction specially relevant to this document is the determination of the twist-4 element @xmath41 @xcite .
it is extracted from bjorken sum measurements ( see next section ) at jlab .
its absolute value at @xmath42 gev@xmath38 , @xmath43 , is large ( the scale is set by the leading twist value of the bjorken sum , @xmath44 ) .
such large ht agrees with the intuition that non - perturbative effects should be large at moderate @xmath0 .
however , the @xmath45 ht correction @xmath46 is small and @xmath47 is of similar magnitude as @xmath48 but of opposite sign .
thus , overall the sum of ht is small near @xmath42 gev@xmath38 .
this explains why ht have been so elusive , as well as the onset of hadron - parton duality @xcite .
if ht are overall small at these @xmath0 , what are their actual effects and what is their practical connection to confinement ? to answer these questions , we first need to discuss the subject of sum rules " .
a sum rule is a rule ( i.e an equality ) that relates a sum ( i.e. , an integral such as a moment of a structure function ) to a quantity characterizing the studied particle .
a relevant sum rule to this document is the gerasimov - drell - hearn ( gdh ) sum rule @xcite that relates the photo - absorption cross - sections @xmath49 and @xmath50 ( 3/2 ( 1/2 ) indicates that the photon helicity is aligned ( anti - aligned ) with the target polarization ) to the anomalous magnetic moment @xmath51 of the target : @xmath52 with @xmath53 the photon energy , @xmath54 the photo - production threshold , @xmath55 the qed coupling constant and @xmath31 the target mass .
originally derived for photo - production ( @xmath56 ) , the gdh sum rule has been latter generalized to @xmath57 @xcite : @xmath58 where @xmath59 is a double dvcs spin
dependent amplitude .
the famous bjorken sum rule for spin - dependent deep inelastic scattering ( dis ) @xcite then appears as the isovector part of the generalized gdh sum rule in the large @xmath0 limit : @xmath60 where the series coefficients and @xmath1 are expressed in the @xmath61 renormalization scheme ( rs ) , and where the @xmath62 are non - perturbative ht corrections discussed in the preceding section that become important only at low @xmath0 .
data on @xmath63 , @xmath64 and @xmath65 , including preliminary results from experiments eg1 , eg4 and e97110 are shown in fig .
the high precision of the mapping at intermediate and moderately high @xmath0 done for the proton and the neutron at jlab can be seen on fig .
[ fig1 ] ( top plots ) .
the @xmath66 moments display a strong @xmath0variation while transiting from high - to - low @xmath0 .
the mapping is being finalized with the analysis of more recent experiments covering the @xmath67 domain at very low @xmath0 ( preliminary results shown on the bottom plots ) .
again , the data have high precision and should test well the @xmath67 predictions . the preliminary results on low-@xmath0 measurement of the generalized gdh sum rule on the neutron(@xmath68he )
are discussed in c. peng s contribution to these proceedings .
we see that the recent @xmath67 results agree with the preliminary data .
we can also notice the similar negative slopes for @xmath63 and @xmath64 at very low @xmath0 .
this is explained by the similar absolute values of the anomalous magnetic moments of the proton and neutron ( assuming the gdh sum rule is not violated , or at least not strongly ) .
consequently , these slopes mostly cancel in the bjorken sum @xmath65 , which results for this observable in a restoration of the approximate conformal behavior ( no @xmath0-dependence ) of qcd already seen at large @xmath0 .
this observation is important in the context of using the ads / cft ( anti - de sitter/_conformal _ field theory ) approach to non - perturbative qcd , as we will discuss latter .
other spin - dependent sum rules exist , such as the burkhardt - cottingham sum rule on @xmath69 @xcite or polarizabilities sum rules that involve higher moments of @xmath4 and @xmath69 , see c. peng , j. zhang and k. slifer contributions to these proceedings . the high @xmath0 data have historically provided a check of qcd in its spin sector , as well as the measurement of the quark spin contribution to the nucleon spin and _via _ evolution equations a first assessment of the gluon spin contribution @xmath70 g ( see e.g. @xcite ) . the moderate @xmath0 data provide as discussed in the previous section a handle on ht .
the new low-@xmath0 data test recent @xmath3pt calculations @xcite , see c. peng and k. slifer contributions to these proceedings .
we will discuss now what we learn in practice from these measurements .
exploring the connection between hadronic and partonic degrees of freedom was the main goal of the jlab experiments that covered the moderate @xmath0values that bridge the hadronic and partonic domains .
that is , their goal was to provide data to help developing / improving non - perturbative theoretical frameworks , and to help explain how these various theoretical frameworks are connected .
we will show here how this goal was achieved .
the precise mapping of low and moderate @xmath0 regions motivated : 1 .
improvement at low @xmath0 on @xmath3pt calculations @xcite , in particular to address the @xmath71 puzzle , see c. peng , j. zhang and k. slifer contributions to these proceedings ; 2 .
improvement at higher @xmath0 of pqcd techniques , such as the analytic or massive perturbation theories @xcite .
these approaches essentially fold partially the ht contributions into the definition of @xmath1 @xcite in order to extend the domain of applicability of pqcd series to lower @xmath0 ; 3 . the calculation of @xmath1 using an ads / cft duality approach @xcite .
this is this last item that we will discuss to illustrate how the jlab spin data have pushed fundamental advances in our understanding of non - perturbative qcd .
the light front holographic qcd ( lfhqcd ) approximation to qcd @xcite is based on light - front quantization , which allows a rigorous and exact formulation of non - perturbative qcd @xcite . in particular
, light - front quantization yields a relativistic schrdinger - like equation describing hadrons as quark bound - states . in principle , all elements of this equation can be determined from the qcd lagrangian ; but , in practice , it is only in ( 1 + 1 ) dimensions that we know how to compute the effective confining potential entering the equation @xcite . due to the overwhelming complexity , in ( 3 + 1 ) dimensions ,
the potential must be determined from other means than first - principle light - front calculations .
one such method is the correspondence between gravity in anti - de sitter ( ads ) space and qcd on the light - front @xcite .
this correspondence originates from the fact that the group of isometries of a 5-dimensional ads space is the conformal invariance of the dual field theory in minkowsky space , and thus encodes the scale invariance of the classical qcd lagrangian .
the calculations are tractable if short - distance quantum fluctuations are neglected and if the quark masses are set to zero ( chiral limit ) .
the 5-dimensional ads calculations are projected on the 4-dimensional boundary of the ads space which is identified with the physical minkowski spacetime ( hence the holographic " denomination ) .
thus , light - front holography provides a semiclassical approximation to qcd in its long distance regime which incorporates fundamental aspects of qcd . enforcing the conformal symmetry of qcd ,
i.e that the lagrangian has no energy scale in the chiral limit , without explicitly breaking the symmetry , leads to a unique choice of potential @xcite .
it has the form of a harmonic oscillator on the light front .
the uniqueness of this form is confirmed by the facts that 1 ) only this form yields a zero - mass pion in the chiral limit @xcite ; 2 ) it explains the remarkable mass symmetry between mesons and baryons @xcite ; 3 ) it is equivalent to the well - established linear potential for static quarks in the usual instant - form front @xcite .
hence , fundamental aspects of nonperturbative qcd are incorporated in lfhqcd which is based on the embedding of light - front relativistic bound state equations in ads space . this semiclassical approximation to qcd in its large distance domain
is completely determined by the constraints imposed by superconformal algebra , a symmetry which can originate from the dynamics of color su(3 ) @xcite .
the only free parameter is the scale @xmath72 .
in fact , chiral qcd is independent of conventional units of mass such as mev . in other words ,
a theory or model of the strong force can only predict dimensionless ratios such as the proton to @xmath73meson mass ratio @xmath74 , or @xmath75 . for lfhqcd this single parameter
is denoted @xmath72 . for standard qcd ,
it is @xmath2 .
the relation between the two is known analytically and numerically @xcite and was obtained _ via _ the qcd coupling @xmath1 , which we discuss next .
a quantity of central interest for this article that can be determined by lfhqcd is the strong coupling @xmath1 @xcite . a particularly useful way to define
it in the long distance regime is _ via _ the concept of effective charge @xcite , which is defined from the perturbative series of an osbservable , truncated to first order in @xmath1 .
for example , with such coupling , eq . ( [ bjorken sr ] ) becomes : @xmath76 where the subscript @xmath4 reminds us of this particular choice of definition .
this choice can be viewed as equivalent to a rs choice @xcite .
one sees that with this definition , both the short distance pqcd effects ( the higher order terms in the leading twists pqcd series ) and long distance confinement effects ( the ht terms ) are now folded into the definition of @xmath1 .
this is in analogy with the original coupling constant becoming a running ( effective ) coupling when short distance quantum effects are folded into its definition @xcite .
the folding of the long distance confinement effects into the coupling definition regularizes it , removing the unphysical landau pole at @xmath77 @xcite .
we already mentioned how the folding of ht lead to optimized perturbative series .
the definition of @xmath1 as an effective charge can be viewed as a generalization of this procedure : the role of ht or more generally the long distance confining effects is to regularize @xmath1 .
the bjorken sum measurement can be used to obtain @xmath78 @xcite .
the most important data are from jlab at moderate and low @xmath0 . at
lower and higher @xmath0 , the gdh and bjorken sum rules , respectively , can supplement the data .
the experimental data yield the coupling shown in fig . [
fig : alpha_g1 ] . from experimental data ( symbols ) .
the data are supplemented at low @xmath0 by the gdh sum rule ( dashed line ) and at high @xmath0 by the bjorken sum rule ( band ) .
the lfhqcd prediction is given by the continuous line .
[ fig : alpha_g1 ] + , scaledwidth=40.0% ] an important observation provided by the measurement of @xmath78 at low @xmath0 is that it freezes , i.e. , qcd s approximate conformal behavior is restored .
( qcd is also conformal at large @xmath0 , a phenomenon called bjorken scaling ) .
this allows us to apply lfhqcd at low @xmath0 .
it predicts that @xcite : @xmath79 this prediction agrees well with the data , see fig .
[ fig : alpha_g1 ] .
it is important to remember that there are no free parameters in eq .
( [ alpha_s from lfhqcd ] ) , @xmath72 being determined by hadron masses ( in fig .
[ fig : alpha_g1 ] , @xmath80 @xcite ) .
the lfhqcd prediction is drawn in fig .
[ fig : alpha_g1 ] up to @xmath81 gev@xmath38 , its expected domain of validity . at
higher @xmath0 , short distance quantum effects , which are not included in the usual semiclassical approximation in lfhqcd become important and void the prediction .
there , however , pqcd is available to calculate @xmath78 . in fact
, the respective domains of applicability of lfhqcd and pqcd appear to overlap around @xmath81 gev@xmath38 @xcite .
this allows us to match the @xmath78 computation from pqcd to that from lfhqcd , thereby relating hadronic masses to the fundamental qcd parameter @xmath2 @xcite .
for example , the leading order relation between @xmath82 and @xmath83 is : @xmath84 where @xmath85 , ( @xmath86 is the first coefficient of the qcd s @xmath87-series ) .
numerically , @xmath88 for @xmath89 quark flavors .
at n@xmath68lo the numerical relation is @xmath90 .
the @xmath73 meson is the ground state solution of the lfhqcd schrdinger equation , i.e. , with radial excitation @xmath91 and internal orbital angular momentum @xmath92 .
higher mass states are solutions with @xmath93 and @xmath94 and are shown on fig .
[ fig : masses ] , together with a similar prediction for strange mesons .
baryon masses can be obtained similarly or thanks to the mass symmetry between mesons and baryons @xcite .
conversely , the known value of @xmath72 can be used with the same matching procedure to predict @xmath2 .
this leads to @xmath95 gev , which agrees well with the world average value of 0.339(17 ) gev @xcite .
we have reported on the low and moderate @xmath0 ( few gev@xmath38 ) nucleon spin structure measurements done at jlab , focusing on the inclusive program on the experimental side . on the phenomenology side
, we focused on what this teaches us about quark confinement and the emergence of effective degrees of freedom ( hadrons ) from the fundamental ones ( quark and gluons ) .
the jlab data cover an extensive kinematic range with high precision thanks to the high luminosity of the experimental equipment .
they complement the data from the cern , slac and desy high energy facilities .
most of the data at moderate @xmath0 are now available .
remaining data from the clas eg1b and hall c sane experiments will be available soon .
the analysis of the lower @xmath0 data is being finalized for the neutron , both from polarized deuteron and @xmath68he targets , and should be available early 2017 .
more work remains for the low @xmath0 proton data . finally , some very low @xmath0 data ( @xmath96 gev@xmath38 ) from n ( @xmath68he ) should become available in the upcoming years .
we argued that the jlab spin sum rule program has reached its goal : it has provided a precise mapping at low , intermediate and moderately high @xmath0 and triggered advances on the theoretical front .
in particular , there is good progress in the description of the strong force over the full @xmath0-range , with improvements in the chiral perturbation theory calculations ( low @xmath0 ) and pqcd series ( high @xmath0 ) .
a goal for these data was to help understanding how hadronic degrees of freedom connect to partonic ones .
better measurements and understanding of higher twists , as well as new approaches to the confinement problem , e.g. , from light
front holographic qcd , represent advances on this front . as examples of recent developments directly connected to the jlab spin data , we showed the extraction of the qcd coupling @xmath1 and its determination from light
front holographic qcd , whose applicability domain overlaps with that of pqcd near @xmath97 1 gev .
this overlap allows the analytical determination of the hadron spectrum with @xmath2 as the only input .
one parameter e.g. , @xmath2 allows one to express quantities in terms of conventional gev units . since @xmath2 is well determined , the mass spectrum prediction from light
front holographic qcd has no free parameters .
obtaining such result has been the long - thought goal of the strong force studies .
arguably light - front holographic qcd is not qcd , but it is a semiclassical approximation which incorporates basics aspects of nonperturbative qcd dynamics which are not apparent from the qcd lagrangian , such as the emergence of a mass scale and confinement , the existence of a zero mass particle in the chiral limit and universal regge trajectories @xcite .
hence the derivation of the strong coupling in the infrared domain and the hadron spectrum represents an important advance toward reaching this long - thought goal , and the jefferson lab spin data has played a major role in it .
we thank s. j. brodsky , e. chudakov , v. sulkosky , g. f. de tramond and n. ton for their useful comments on the manuscript .
this material is based upon work supported by the u.s .
department of energy , office of science , office of nuclear physics under contract de ac0506or23177 .
m. burkardt , phys .
d * 88 * , 114502 ( 2013 ) x. zheng _ et al .
_ , phys .
c * 70 * , 065207 ( 2004 ) ; k. kramer _
et al . _ ,
lett , * 95 * , 142002 ( 2005 ) ; z .- e .
et al . _ ,
b 613 , 148 ( 2005 ) ; m. posik _
_ , phys .
lett , * 13 * , 022002 ( 2014 ) x. d. ji and j. osborne , j. phys .
g * 27 * , 127 ( 2001 ) j. d. bjorken , phys .
rev . * 148 * , 1467 ( 1966 ) ; phys .
d * 1 * , 1376 ( 1970 ) h. burkhardt and w. n. cottingham , annals phys . * 56 * , 453 ( 1970 ) v. bernard _ et al .
d 87 054032 ( 2013 ) ; v. lensky , j. m. alarcon and v. pascalutsa , phys .
c 90 055202 ( 2014 ) s. j. brodsky , g. f. de tramond , h. g. dosch and j. erlich , phys . rep .
05 001 ( 2015 ) s. j. brodsky , h. c. pauli and s. s. pinsky , phys .
rept . * 301 * , 299 ( 1998 ) k. hornbostel , s. j. brodsky and h. c. pauli , phys .
d * 41 * , 3814 ( 1990 ) s. j. brodsky and g. f. de tramond , phys .
96 , 201601 ( 2006 ) , phys .
102 , 081601 ( 2009 ) h. g. dosch , g. f. de tramond and s. j. brodsky , phys .
d * 91 * , no .
8 , 085016 ( 2015 ) g. f. de tramond , h. g. dosch and s. j. brodsky , phys
. rev .
d * 91 * , no . 4 , 045040 ( 2015 ) a. p. trawinski , s. d. glazek , s. j. brodsky , g. f. de tramond and h. g. dosch , phys .
d * 90 * , no . 7 , 074017 ( 2014 ) g. f. de tramond , s. j. brodsky , a. deur , h. g. dosch and r. s. sufian , arxiv:1611.03763 [ hep - ph ] a. deur , v. burkert , j. p. chen and w. korsch , phys .
b * 650 * , 244 ( 2007 ) ; phys .
b * 665 * , 349 ( 2008 ) c. patrignani _ et al .
_ [ particle data group collaboration ] , chin . phys .
c * 40 * , no .
10 , 100001 ( 2016 ) | we report on the low and moderate @xmath0 nucleon spin structure measurements done at jefferson lab , examining specifically the inclusive program .
we discuss what the data teach us about quark confinement and the emergence of the effective hadronic degrees of freedom from the fundamental partonic ones .
we show how this experimental program has reached its goal by providing a precise mapping at low , intermediate and moderately high @xmath0 which has followed in many advances , e.g. , with chiral perturbation theory .
another example of a recent advance imputable to the jlab spin data is the improved understanding of @xmath1 at low energy , which allowed light
front holographic qcd , an approximation to non - perturbative qcd , to derive the hadron spectrum from @xmath2 . |
after years of study to understand how the nucleon spin originates from the constituent partons , an exhaustive answer is still missing .
moreover , while a lot of information has been gathered concerning the longitudinal structure of a fast moving nucleon ( with respect to its direction of motion ) , very little is known about the transverse structure . in recent years these aspects have raised a lot of interest and after important theoretical developments and experimental findings , transverse spin and transverse momentum @xmath0 of the quarks are by now considered as fundamental ingredients in the description of the hadron structure .
spin-@xmath0 correlations give rise to various observables in hard hadronic processes such as the azimuthal asymmetries seen both in unpolarized or transversely polarized semi - inclusive deep - inelastic scattering ( sidis ) , and led to the introduction of transverse momentum dependent ( tmd ) parton distribution functions ( pdf ) and fragmentation functions ( ff ) . among these functions of special interest
are the sivers functions @xmath3
@xcite , which describe an azimuthal asymmetry in the parton distributions inside a transversely polarized nucleon , and by their chirally - odd partner @xmath4 , the boer - mulders functions @xcite , describing the transverse parton polarization inside an unpolarized hadron , and generating azimuthal asymmetries in unpolarized sidis .
compass results on the sivers asymmetries are given elsewhere @xcite while here the unpolarized azimuthal asymmetries are presented .
the cross - section for hadron production in lepton - nucleon dis @xmath5 for unpolarized targets and an unpolarized or longitudinally polarized beam is the following @xcite : @xmath6 & & \left .
+ \varepsilon_2 \cos(2\phi_h ) f^{\cos\ ; 2\phi_h}_{uu } + \lambda_\mu \varepsilon_3 \sin
\phi_h f^{\sin \phi_h}_{lu } \right ] \end{array}\ ] ] where @xmath7 is the fine structure constant , @xmath8 , @xmath9 and @xmath10 are the inclusive dis variables , @xmath11 is the fraction of the virtual photon energy carried by the detected hadron , @xmath12 is the azimuthal angle of the outgoing hadron in the @xmath13-nucleon system .
@xmath14 , @xmath15 , @xmath16 , @xmath17 and @xmath18 are structure functions , with the first and second subscripts which indicate the beam and target polarization respectively , and the last subscript which , if present , indicates the polarization of the virtual photon .
finally @xmath19 is the beam longitudinal polarization and : @xmath20 \varepsilon_2 & = & \dfrac{2(1-y)}{1+(1-y)^2 } \\[2ex ] \varepsilon_3 & = & \dfrac{2 y \sqrt{1-y}}{1+(1-y)^2 } \end{array}\ ] ] are depolarization factors .
the boer - mulders pdfs contribute to both the @xmath21 and the @xmath22 structure functions , together with the so called cahn effect @xcite which arises from the fact that the kinematics is non collinear when the @xmath0 is taken into account ( i.e. a kinematical higher twist ) , and with the perturbative gluon radiation , resulting in order @xmath23 qcd processes .
pqcd effects are becoming important for high transverse momenta @xmath24 of the produced hadrons , while are small for @xmath24 up to 1 gev@xmath25 .
the @xmath26 modulation , which arises from the natural polarization of the muon beam , does not have a clear interpretation in the parton model . in the past
, azimuthal asymmetries have been measured by the emc collaboration @xcite , with a liquid hydrogen target and a muon beam at a slightly higher energy , but without separating hadrons of different charge .
these data have been used @xcite to extract the average @xmath27 .
azimuthal asymmetries have been also measured by e665 @xcite and at higher energies by zeus @xcite .
more recent are the compass results first presented at @xcite , and the measurements done by hermes , first shown at @xcite .
the compass experiment has been set up at the cern sps m2 beam line .
it combines high rate beams with a modern two stage magnetic spectrometer@xcite .
compass has collected data with a 160 gev positive muon beam impinging on a polarized solid target .
the beam is naturally polarized by the @xmath28decay mechanism , and the beam polarization is estimated to be @xmath29 with a @xmath30 relative error .
the beam intensity is @xmath31 muons per spill . up to
2006 the experiment has used @xmath1lid as deuteron target because its favorable dilution factor of @xmath320.4 , particularly important for the measurement of . in 2007
an ammonia nh@xmath33 target has been used as proton target .
polarizations of 50% and 90% have been reached , respectively for the two target materials .
the event selection requires standard dis cuts , i.e. @xmath34 , mass of the final hadronic state @xmath35 , @xmath36 , and the detection of at least one hadron in the final state .
moreover only events with a vertex in the forward target cell are used in order to minimize nuclear interactions and for a better data / monte carlo agreement .
finally , for the detected hadrons it is also required that : r0.50 * the fraction of the virtual photon energy carried by the hadron be @xmath37 to select hadrons from the current fragmentation region ; * @xmath38 , where @xmath24 is the hadron transverse momentum with respect to the virtual photon direction , for a better determination of the azimuthal angle @xmath12 . data taken both with a longitudinally polarized and a transversely polarized target have been used , mixing positive or negative orientations in order to cancel effects depending on the polarization of the nucleons .
this statistics corresponds to almost 1 month of the 2004 data taking and after all the cuts consists of @xmath39 positive hadrons and @xmath40 negative hadrons entering the asymmetry calculations . in the measurement of unpolarized asymmetries all the methods used to cancel the experimental acceptance
can not be adopted and the correction for this effect is mandatory before fitting the azimuthal modulation .
this is done by using a full monte carlo chain , which starts from the sidis event generation performed by lepto@xcite , simulate the experimental setup and the particle interactions in the passive and active material of the detectors ( also including the detector response done by comgeant ) , and ends with the reconstruction of the generated events by the same program ( coral ) used to analyze the real data .
the quality of this chain is evaluated by comparing distributions of real data and of generated events both for the dis variables and for the hadronic variables . the experimental acceptance as a function of the azimuthal angle @xmath41
is then calculated as the ratio of reconstructed over generated events for each bin of @xmath8 , @xmath11 and @xmath24 on which the asymmetries are measured .
the overall @xmath9 , @xmath11 and @xmath24 acceptances are quite constant over the range used in the analysis , so that the effect coming by the integration over the unlooked variables when the asymmetries in one of the variables are extracted is well within the systematic error . the measured distribution , corrected for acceptance
is fitted with the following functional form : @xmath42 an example of a measured azimuthal distribution , acceptance corrections and corrected azimuthal distribution is shown in fig .
[ azidist ] , together with the resulting fit .
the contribution of the acceptance corrections to the systematic error has been studied with care . as the asymmetries were extracted from data taken
both with longitudinal and transverse target configurations , comparing the two results gives the effect of the acceptance changes due to the different configuration ( solenoid vs. dipole ) of the target magnet and to the different direction of the incoming beam ( for the transverse setup the beam is bent in order to leave the target with the same direction as in the longitudinal case ) . in order to check the effect of the simulation parameters the acceptances have been calculated using two different sets of lepto parameters .
all the resulting asymmetries were compared in order to quantify the systematic error in each kinematical bin .
further systematic tests , like splitting the data sample according to the event topology and to the time of the measurement , gave no significant contributions .
r0.65 the @xmath43 asymmetries , not shown here , measured by compass are compatible with zero , at the present level of statistical and systematic errors , over the full range of @xmath8 , @xmath11 and @xmath24 covered by the data .
the @xmath21 asymmetries extracted from compass deuteron data are shown in fig .
[ f : cosphi ] for positive ( upper row ) and negative ( lower row ) hadrons , as a function of @xmath8 , @xmath11 and @xmath24 .
the bands indicate the size of the systematic error .
the asymmetries show the same trend for positive and negative hadrons with a slightly larger values for the positive one .
values as large as 30@xmath4440% are reached in the last point of the @xmath11 range .
the theoretical predictions @xcite in fig .
[ f : cosphi ] takes into account the cahn effect only , which does not depend on the hadron charge .
the boer - mulders pdfs are not taken into account in this case .
the @xmath45 asymmetries are shown in fig .
[ f : cos2phi ] together with the theoretical predictions of @xcite , which take into account the kinematical contribution given by the cahn effect , first order pqcd ( which , as expected , is negligible in the low @xmath24 region ) , and the boer - mulders pdfs ( coupled to the collins ff ) , which give a different contribution to positive and negative hadrons . in @xcite the boer - mulders pdfs
are assumed to be proportional to the sivers function as extracted from the preliminary hermes data .
the compass data show different amplitude for positive and negative hadrons , a trend which confirms the theoretical predictions .
there is a satisfactory agreement between the data points and the model calculations , which hints to a non zero boer - mulders pdfs .
99 slides : + ` http://indico.cern.ch/contributiondisplay.py?contribid=308&sessionid=4&confid=53294 ` d.w .
sivers , phys .
* d41 * ( 1991 ) 83 .
d. boer and p.j .
mulders , phys . rev .
* d57 * ( 1998 ) 5780 .
a. bressan , `` compass results on collins and sivers asymmetries '' , these proceedings .
bacchetta _ et al .
_ , jhep * 0702 * ( 2007 ) 93 .
cahn , phys .
* b78 * ( 1978 ) 269 .
the emc collaboration , m. arneodo _
et al . _ , z. phys . * c34 * ( 1987 ) 277 . the emc collaboration , j. ashman _ et al . _ , z. phys .
* c52 * ( 1991 ) 361 . m. anselmino _
et al . _ ,
phys . rev .
* d71 * ( 2005 ) 74006 the e665 collaboration , m.r .
et al . _ ,
phys . rev .
* d48 * ( 1993 ) 5057 .
the zeus collaboration , j. breitweg _
et al . _ ,
b481 * ( 2000 ) 199 .
w. kfer , `` measurements of unpolarized azimuthal asymmetries at compass '' , in the proceedings of ` transversity 2008 ' and arxiv:0808.0140 . f. giordano , `` measurements of azimuthal asymmetries of the unpolarized cross - section at hermes '' , in the proceedings of spin2008 and arxiv:0901.2438 . the compass collaboration , p. abbon _ et al .
* a577 * ( 2007 ) 455 .
g. ingelman , a. edin and j. rathsman , comp .
commun . 101
( 1997 ) 108 .
m. anselmino _ et al .
_ , eur .
j. * a31 * ( 2007 ) 373 . v. barone , a. prokudin and b.q .
ma , phys . rev . *
d78 * ( 2008 ) 45022 . | azimuthal asymmetries measured in unpolarized semi - inclusive deep inelastic scattering bring important information on the inner structure of the nucleons , and can be used both to estimate the average quark transverse momentum @xmath0 and to access the so - far unmeasured boer - mulders functions .
compass results using part of the 2004 data collected with a @xmath1lid target and a 160 gev @xmath2 beam are presented separately for positive and negative hadrons . |
real quantum systems are inevitably open because of interaction with their ambient environment . as a result ,
information , in principle , can leak into the environment , and may be fed back later in a different form into the system @xcite .
this scenario makes accurate description of the related dynamics difficult @xcite , because one often has no much information / control about / over the environment unless , e.g. , the environment can be controllably engineered @xcite .
nevertheless , under some specific ( ideal ) conditions , such as weak coupling with a memoryless environment and born - markov approximation , the underlying system evolution can be well described by a quantum _ markovian _ dynamical equation @xcite ( see also ref .
@xcite for an alternative derivation ) .
this approximation has been proved useful in various situations , such as devising quantum error - correction schemes @xcite . in this case , the dynamics of the system for any _ interval _ @xmath0regardless of the start time @xmath1is given by a completely - positive ( cp ) map @xmath2 , having the dynamical semigroup " property @xmath3 , for any @xmath4 @xcite ; or equivalently , the dynamical equation @xmath5 $ ] governs the system density matrix , where the time - independent @xmath6 has the following lindblad form : @xmath7= -i[h,\cdot]+\sum_{m } r_{m}\bigl(2 f_m \cdot f_m^{\dag}-\{f_m^{\dag}f_m,\cdot\}_+\bigr ) , \label{lf}\end{aligned}\ ] ] leading to @xmath8 .
here , @xmath9 , @xmath10 is a hermitian operator , and @xmath11s are some operators acting on the system s hilbert space @xcite .
note that the case of time - dependent @xmath12s is also referred to as a time - dependent markovian evolution when @xmath13 @xcite .
this simplified picture ( even its beyond - markovian version
redfield theory ) , although fairly well applicable to various situations , often fails to capture dynamics of complex quantum manybody systems in condensed matter or in some biological complexes ( which may feature quantum effects ) . in such cases , at least one of the assumptions of the born - markov ( or redfield ) theory breaks down @xcite .
these difficulties have spurred introducing various formalisms to somehow incorporate non - markovian effects into open system s dynamics , e.g. , through adding memory kernels @xcite or even devising specific dynamical equations valid for non - markovian regimes ( see , e.g. , refs .
@xcite ) .
although deciding whether a given quantum channel / dynamics is markovian ( or not ) is in complex - theoretic sense hard @xcite , there exist a number of measures of markovianity based on the lindblad form @xcite .
in addition , some experimentally measurable criteria have been introduced for detecting non - markovianity of an evolution by using , e.g. , increasing ( decreasing ) the distance ( fidelity ) of two density matrices in time @xcite , or increasing the entanglement between the system and an isolated ancilla in some instances of time @xcite .
these measures provide sufficient conditions for a channel to be non - markovian , which leaves open search for more delicate measures for a critical study and comparison of some of these measures , see ref .
@xcite . on a related note ,
since correlations and how they are distributed between a system and its embedding environment play a crucial role in the markovianity property of the resultant system dynamics , it seems natural that tracking correlations may provide a hint for ( at least partially ) identifying markovianity property of the dynamics . along this line , here we employ a recently introduced quantum correlation measurediscord " @xcite and its associated properties to propose a ( non-)markovianity measure .
we illustrate the main idea through two examples .
quantum discord ( qd ) is a measure for quantumness of correlations in a bipartite state @xmath14 @xcite , and is defined as @xmath15=\mathcal{s}[\varrho_s]+\min_{\{\pi_j^s\}}\mathcal{s}\left[e|\{\pi_j^{s}\}\right ] -\mathcal{s}[\varrho_{se}],\end{aligned}\ ] ] where @xmath16 is the von neumann entropy , and @xmath17=\sum_j p_j \mathcal{s}\bigl[\varrho_{e|\pi_j^s}\bigr]$ ] is the average entropy of the state of @xmath18 provided that some ( rank-@xmath19 ) projective measurements @xmath20 have been performed on @xmath21 ( whence @xmath21-discord " @xmath22 ) and have given the result @xmath23 with probability @xmath24 . ) and
environment ( @xmath18 ) is a generally mixed state with vanishing discord .
the state of the ancilla ( @xmath25 ) is chosen such that it purifies @xmath26 ; i.e , @xmath27 .
the state of the combined system ( @xmath28 ) hence will remain pure in time .
after state tomography to determine @xmath29 , we are able to conclude whether the map @xmath30 ( acting only on the system ) is markovian or non - markovian.,width=325,height=177 ] qd has recently attracted much attention due to its importance in distinguishing classical and quantum correlations @xcite , and in particular in showing how a one - qubit quantum computer may work @xcite .
one can also find operational interpretations for qd , e.g. , by exploiting the concept of quantum state merging @xcite . vanishing and time - invariant discord in quantum states
have been investigated in , respectively , refs .
@xcite and ref .
@xcite . of particular importance to our goal in this paper
, we remind that in ref .
@xcite , it has been shown that , except for a zero - measure " set of quantum states , vanishing quantum @xmath21-discord of the initial state of the system and the environment is a _ necessary and sufficient _
condition for the dynamics of the system to be cp . given a system - environment interaction @xmath31 , whereby the evolution @xmath32 $ ]
, one can define as below a _ cp map _
@xmath33 on @xmath34 if(f ) @xmath35 has vanishing @xmath21-discord ( up to the limitations stated in ref .
@xcite ) , @xmath36\nonumber\\ & \equiv & \widehat{\mathcal{e}}(t+\delta , t)\varrho_s(t),\end{aligned}\ ] ] for a given @xmath37 .
one can constitute a concatenation of these maps as @xmath38 , which by construction is an
_ infinitesimal divisible _ " cp map .
such maps have been proven to be always represented by a markovian dynamics @xcite .
it is evident that @xmath39 .
thus in summary , we have the following result : * proposition 1 : * the dynamics of a system @xmath21 interacting with an environment @xmath18 is markovian if(f ) the instantaneous combined state of the system and the environment has vanishing qd , i.e. , @xmath40=0~\forall t\geq0 $ ] .
this result , as is , provides a method which is hardly experimentally measurable because it requires the knowledge of @xmath35 at all times , which is too demanding . to overcome this problem , we relax its strict condition and
just focus on the sufficiency part . specifically
, we use a lower bound for qd which is independent of the information of the environment . to this end
, we use a monogamy relation for qd in a _
pure _ tripartite state @xmath41 as @xmath42=\mathcal{d}_{s}\bigl[\varrho_{es}\bigr]+\mathcal{s}_{e|s } , \label{discord - conservation}\end{aligned}\ ] ] where @xmath43 $ ] is the entanglement of formation of the state of @xmath44 @xcite , and @xmath45-\mathcal{s}\bigl[\varrho_{s}\bigr]$ ] is the conditional entropy @xcite . since the total state @xmath41 is assumed to be pure , from the schmidt decomposition we have @xmath46=\mathcal{s}\bigl[\varrho_{a}\bigr]$ ] , whence @xmath47=\mathcal{d}_{s}\bigl[\varrho_{es}\bigr ] + \mathcal{s}\bigl[\varrho_{a}\bigr]-\mathcal{s}\bigl[\varrho_{s}\bigr ] .
\label{discord - conservation-1}\end{aligned}\ ] ] now let us purify the system - environment initial state @xmath26 by attaching an ancilla @xmath25 which is not interacting with @xmath21 and @xmath18 .
the total state @xmath48 will remain pure in time because @xmath49 evolves unitarily in time , while @xmath25 does not evolve . in this case , noting that entanglement of formation is always nonnegative , eq .
( [ discord - conservation-1 ] ) yields @xmath50\geq \mathcal{s}\bigl[\varrho_{s}(t)\bigr ] - \mathcal{s}\bigl[\varrho_{a}(t)\bigr]=:\delta_{sa}(t).\end{aligned}\ ] ] this relation implies that to recognize non - markovianity of a quantum dynamics , it suffices to measure @xmath51 ; an strictly positive value for @xmath52 at some time @xmath53 is a signature of non - markovianity of the dynamics at any later time .
proposition 1 can also be used to define a measure for non - markovianity power of a quantum dynamics in an interval @xmath54 as follows : @xmath55 dt.\end{aligned}\ ] ] a nonzero value for @xmath56 is a necessary and sufficient condition for non - markovianity of the associated quantum dynamics . since again computing @xmath57 $ ]
is difficult , we propose @xmath58 as a lower bound for non - markovianity power .
_ remark 1 .
_ it should be noted that our qd measure is not a trivial extension of the entanglement measure for non - markovianity power of a quantum dynamics .
one of the characteristics of qd is that it can increase under local operation and classical communication ( locc ) while entanglement can not .
separable states @xmath59 ( @xmath60 , @xmath61 ) can be generated by locc from some initial product state , say , @xmath62 .
since there exist separable states with nonvanishing qd @xcite , locc may increase qd . as an explicit example , consider the preparation of the system and ancilla in the vanishing qd state @xmath63 , where @xmath64 and @xmath65 are two orthogonal density matrices .
a hadamard channel @xmath66\varrho+p(\tau)h \varrho h,\end{aligned}\ ] ] with @xmath67 , acting on the system yields @xmath68{|0\rangle\langle 0|}+p{|+\rangle\langle + |}\}\otimes w_1+\nonumber\\&&~\frac{1}2 \{[1-p(\tau)]{|1\rangle\langle 1|}+p(\tau){|-\rangle\langle -|}\}\otimes w_2,\end{aligned}\ ] ] in which @xmath69 is the eigenvector of @xmath70 .
since @xmath71\neq 0 $ ] , @xmath72 is a state with nonvanishing qd @xcite , while @xmath73=0 $ ] .
_ remark 2 .
_ the positivity of the associated choi - jamilkowski state of a quantum process has been used as a measure for its markovianity @xcite . to use this criterion , one needs to find @xmath74 .
but the problem is that in general @xmath74 is not a well - defined quantum process / map .
when the evolution is non - markovian , @xmath74 is not necessarily a map or identifiable by a physically conceivable tomography scheme .
in fact , in the middle of the process at some time @xmath53 , it may happen that @xmath75 has nonzero quantum correlation .
hence , the evolution of the system can not be interpreted as the application of a quantum _ map _ on @xmath76 .
note also that the number of independent parameters to be evaluated to identify the process between a @xmath77 and @xmath78 thoroughly is @xmath79 , where @xmath80 is the dimension of the hilbert space of the system ; while a complete set of quantum states has only @xmath81 independent members .
this implies that process tomography for identification of @xmath74 is in general infeasible . in our approach
, however , all one needs is just state tomography of @xmath35 in every point of time , which is well - defined ( although of course difficult ) .
besides , if one use the lower bound proposed , our measure seems experimentally feasible .
_ remark 3 .
_ it should be clear that , although we use the result of ref .
@xcite , here we have set different goals and strategy .
we deal with the problem of non - markovianity and provide a measure thereof by using the concept of qd , while ref .
@xcite deals with a condition for cpt - ness of quantum maps
. is decreasing in every arbitrary time interval [ @xmath77 is in @xmath82 units ] . hence , deciding about markovianity of the dynamics is not possible through this entanglement measure . on the other hand ,
@xmath52 [ eq . ( [ discord - lowerbound ] ) ] ( the red curve ) , is positive in @xmath83 $ ] , which shows that @xmath57 $ ] is nonzero in this interval ; hence , the process is non - markovian.,title="fig:",width=340,height=234 ] -9 mm
in the following , we illustrate the main idea of this paper through two examples .
* example i. * consider the jaynes - cummings model , in which the system of interest is a two - level atom interacting with a cavity mode through the hamiltonian @xmath84 where @xmath85 and @xmath86 are , respectively , the raising operators of the system and environment .
we assume the initial state of the atom ( @xmath21 ) and cavity ( @xmath18 ) to be a separable state @xmath87 , in which @xmath88 is a coherent state , and @xmath89 .
for the case of @xmath90 , the purification with a qubit ancilla yields @xmath91 .
when @xmath92 and @xmath93 , the concurrence of the system and ancilla decreases in time .
hence , the entanglement measure of non - markovianity does not imply a reliable clue about markovianity / non - markovianity of the dynamics .
but since @xmath94 , we are sure that the evolution is non - markovian ( see fig . [ entropyvsnegativity ] ) .
calculation of the lower bound on non - markovianity power of the dynamics gives @xmath95 .
-4 mm -discord ( red ) of example ii vs. time [ in @xmath82 units ] for @xmath96 , where @xmath97 and @xmath98 .
@xmath21-discord has been obtained from eq .
( [ discord - conservation-1 ] ) in which @xmath99 $ ] .
the blue curve shows concurrence between the system and ancilla vs. time .
since concurrence increases only after @xmath100 , the non - markovianity property of the dynamics for @xmath101 is not captured by using the entanglement measure of non - markovianity.,title="fig:",width=415,height=275 ] -16 mm * example ii . *
consider two qubits ( system - environment ) interacting through the hamiltonian @xmath102 leading to the evolution operator @xmath103 in which @xmath104 are the identity and the pauli matrices .
we show that if the initial state of the environment is diagonal in the computational basis , the state of the environment does not change at any time , @xmath105 , and the system and the environment remain unentangled in time
see , e.g. , ref .
@xcite for a note on lack of back - action in revival of quantum correlations .
one may infer that the evolution of the system should then be described by a quantum dynamical semigroup .
however , we show that it is a non - markovian transformation because a nonzero qd is produced during the evolution due to the interaction of the system and the environment . assuming @xmath106 , the kraus operators of the dynamics of the system are @xmath107 and @xmath108 , where @xmath109 . to find whether the obtained time - dependent map @xmath110 is markovian or not , one needs to calculate the @xmath21-discord of @xmath35 .
it is straightforward to show that @xmath111 which is a separable state .
given the environment initially in the state @xmath112 , it is evident that @xmath105 .
now we use a result of ref .
@xcite that if a bipartite quantum state has zero @xmath21-discord , then @xmath113= 0 $ ] . since in this example
@xmath114\neq 0 $ ] unless @xmath115=0 $ ] , thus except for these special states , @xmath21-discord is nonzero , and hence the dynamics of the system is non - markovian ( see fig . [ discord - phase - flip - channel ] ) .
an interesting question in open quantum system theory is that given the interaction hamiltonian of a system and its environment whether the induced dynamics on the system is markovian or not .
several methods have been suggested to assess non - markovianity of a dynamics . here ,
based on the idea that underlying correlations play key role in a reduced dynamics , we employ quantum discord ( qd ) to define a measure for non - markovianity . to alleviate the need for the state of the environment in calculating qd in our scheme , we have considered an experimentally measurable lower bound for qd by using a monogamy property for tripartite quantum systems .
this bound being nonvanishing provides a sufficient condition for non - markovianity of a quantum dynamics .
we also have discussed two examples to illustrate the result of this study . in the first example
, it has been shown that qd can signal non - markovianity in a process which eludes the entanglement measure of non - markovianity .
the second example has featured an explicit quantum dynamics whose non - markovianity is strictly due to the production of quantum correlation not in the form of entanglement . here , the initial state of the system and environment was a product state and remained separable in time , while the state of the environment did not vary in time either .
having powerful non - markovianity measures at hand , one can think of various applications in quantum information theory .
for example , an intriguing question is that how one should engineer an environment so that the generated sub - dynamics of the attached system becomes markovian .
this is important because , for example , it has been argued @xcite that one can perform adiabatic quantum computation in an indirect fashion in an open system , provided that the effective system dynamics is a markovian evolution with fixed points encoding ground states of some given hamiltonian .
such scenarios require ability to decide whether an environment induces markovian sub - dynamics .
our method provides a way to go to answer this question .
given an interaction hamiltonian with unknown coupling constants , @xmath116 , and a preparation of the system and environment ( assuming being feasible ) , @xmath26 , we can find the state of the system and the environment as a function of @xmath116 , @xmath117 , and check for which values of @xmath116 , @xmath21-discord remains zero in time . as a result , we are in principle able to manipulate coupling constants such that the desired map becomes markovian or non - markovian on demand . _
acknowledgements. _ the authors acknowledge v. karimipour , d. a. lidar , and a. shabani for useful discussions . an earlier version of this work was presented in the school and workshop of new trends in quantum dynamics and entanglement ( the abdus salam international center for theoretical physics , trieste , italy , 2011 ) ; we appreciate all comments by the participants of this school / workshop .
this research is partially supported by sharif university of technology s office of vice - chair for research .
breuer and f. petruccione , _ the theory of open quantum systems _
( oxford university press , new york , 2002 ) ; h. carmichael , _ an open system approach to quantum optics _ ( springer , berlin , 1994 ) ; a. rivas and s. f. huelga , arxiv:1104.5242 .
t. s. cubitt , j. eisert , and m. m. wolf , arxiv:1005.0005 .
f. verstraete , m. m. wolf , and j. i. cirac , nature phys .
* 5 * , 633 ( 2009 ) .
d. a. lidar , z. bihary , and k. b. whaley , chem . phys . * 268 * , 35 ( 2001 )
. m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _
( cambridge university press , cambridge , england , 2000 ) .
m. m. wolf , j. eisert , t. s. cubitt , and j. i. cirac , phys .
* 101 * , 150402 ( 2008 ) .
e. andersson , j. d. cresser , and m. j. w. hall , arxiv:1009.0845 .
m. m. wolf and j. i. cirac , commun .
phys . * 279 * , 147 ( 2008 ) .
a. rivas , a. d. k. plato , s. f. huelga , and m. b. plenio , new j. phys .
* 12 * , 113032 ( 2010 ) .
a. ishizaki and g. r. fleming , j. chem . phys . *
130 * , 234110 ( 2009 ) ; _ ibid . _
* 130 * , 234111 ( 2009 ) .
c . cheng and g. r. fleming , annu .
phys . chem . *
60 * , 241 ( 2009 ) .
a. shabani and d. a. lidar , phys .
a * 71 * , 020101(r ) ( 2005 ) ; h .-
breuer and b. vacchini , phys .
* 101 * , 140402 ( 2008 ) .
breuer , j. gemmer , and m. michel , phys .
e * 73 * , 016139 ( 2006 ) ; h .-
breuer , phys .
a * 75 * , 022103 ( 2007 ) .
t. s. cubitt , j. eisert , and m. m. wolf , commun . math
* 310 * , 383 ( 2012 ) .
breuer , e .-
laine , and j. piilo , phys .
lett . * 103 * , 210401 ( 2009 ) .
a. k. rajagopal , a. r. usha devi , and r. w. rendell , phys .
a * 82 * , 042107 ( 2010 ) .
a. rivas , s. f. huelga , and m. b. plenio , phys .
lett . * 105 * , 050403 ( 2010 ) .
b. vacchini , a. smirne , e .-
laine , j. piilo , and h .-
breuer , new j. phys .
* 13 * , 093004 ( 2011 ) h. ollivier and w. h. zurek , phys .
* 88 * , 017901 ( 2002 ) ; l. henderson and v. vedral , j. phys . a : math
. gen . * 34 * , 6899 ( 2001 ) .
k. modi , t. paterek , w. son , v. vedral , and m. williamson , phys .
lett . * 104 * , 080501 ( 2010 ) .
k. modi , a. brodutch , h. cable , t. paterek , and v. vedral , arxiv:1112.6238 .
a. ferraro , l. aolita , d. cavalcanti , f. m. cucchietti , and a. acn , phys .
a * 81 * , 052318 ( 2010 ) . c. zhang , s. yu , q. chen , and c. h. oh , phys . rev .
a * 84 * , 032122 ( 2011 ) ; _ ibid . _ 052112 ( 2011 ) ; q. chen , c. zhang , s. yu , x. x. yi , and c. h. oh , _ ibid . _ 042313 ( 2011 ) . m. arsenijevi , j. jekni - dugi , and m. dugi , arxiv:1203.4612 .
a. datta , a. shaji , and c. m. caves , phys .
lett . * 100 * , 050502 ( 2008 ) .
v. madhok and a. datta , phys .
a * 83 * , 032323 ( 2011 ) ; d. cavalcanti , l. aolita , s. boixo , k. modi , m. piani , and a. winter , _ ibid .
_ , 032324 ( 2011 ) .
b. daki , v. vedral , and .
brukner , phys .
105 * , 190502 ( 2010 ) ; b. bylicka and d. chruciski , phys .
a * 81 * , 062102 ( 2010 ) ; b. bylicka and d. chruciski , arxiv:1104.1804 . l. mazzola , j. piilo , and s. maniscalco , intl .
j. quantum inf . * 9 * , 981 ( 2011 ) .
a. shabani and d. a. lidar , phys .
lett . * 102 * , 100402 ( 2009 ) .
r. horodecki , p. horodecki , m. horodecki , and k. horodecki , rev .
phys . * 81 * , 865 ( 2009 ) .
f. f. fanchini , m. f. cornelio , m. c. de oliveira , and a. o. caldeira , phys .
a * 84 * , 012313 ( 2011 ) .
r. lo franco , b. bellomo , e. andersson , and g. compagno , phys .
a * 85 * , 032318 ( 2012 ) . | the problem of recognizing ( non-)markovianity of a quantum dynamics is revisited through analyzing quantum correlations .
we argue that instantaneously - vanishing quantum discord provides a necessary and sufficient condition for markovianity of a quantum map .
this is used to introduce a measure of non - markovianity .
this measure , however , requires demanding knowledge about the system and the environment . by using a quantum correlation monogamy property and an ancillary system ,
we propose a simplified measure with less requirements .
non - markovianity is thereby decided by quantum state tomography of the system and the ancilla . |
consider a wide area computer network such as the one in figure [ geant ] .
a _ flow _ is defined as all traffic with common origin and destination nodes .
monitoring flow volumes plays an important role in network management tasks , such as capacity planning by tracking demands and forecasting traffic , identifying failures together with their causes and impact , detecting malicious activity and configuring routing protocols [ @xcite , ] .
these flow volumes have been observed to exhibit complicated structure , as seen in figure [ flowvols ] .
for example , the highly aggregated flows usually have diurnal patterns [ figure [ flowvols](a ) ] , while lighter flows can be extremely noisy [ figure [ flowvols](b ) ] . network traffic is carried on packets that can be observed ( and sampled ) at router interfaces , henceforth called _
observation points_. however , during the measurement process , sampling is employed due to high flow volumes and resource constraints at routers . ) and the corresponding logical topology . ]
it is increasingly common for such measurement infrastructure to be deployed in computer networks [ @xcite ] .
each packet from the aggregate flow at an observation point is sampled independently with a certain probability ( sampling rate ) [ @xcite ] .
typical sampling rates range between 0.0010.01 . for every packet sampled ,
its header information is recorded which allows one to reconstruct objects of interest , such as volumes of flows with a particular source and destination traversing the network .
an important issue is how to _ select _ ( design ) the sampling rates across the network subject to resource constraints , in order to collect the maximum amount of information on the underlying source - destination flows .
obviously low sampling rates result in large sampling noise .
one way of achieving lower estimation error with the same sampling rate is through _ filtering _ ; that is , combining the present measurement with past measurements to track the time - series of flow volumes . in designing a sampling scheme for this situation
one needs to take into account measurement noise and process noise ( innovation noise ) .
while modeling the dynamics of flow volumes is a challenging task in itself [ @xcite ] , we use a simple random walk model for this purpose .
this is a robust enough model to be useful in a large range of applications and leads to scalable filters .
we consider the problem of minimizing the ( running ) estimation error through optimal design of measurement scheme in the filtering context . in this paper
we take an optimal design of experiment approach to the above problem and demonstrate its application to computer network monitoring using sampled data . the related research on optimal design has focused on one of the following scenarios .
there is a large body of work on optimal input design for dynamical systems [ @xcite , @xcite ] .
there the focus is on parameter estimation ( system identification ) rather than filtering , as in this paper .
another related area is sequential design for nonlinear systems [ @xcite , @xcite ] , where the optimal design depends on values of unknown parameters .
while there are some commonalities , the design problem in a filtering context is unique in that the design at any time affects not just the current estimation error but also future ones .
the problem of optimal sensor placement in control system literature looks at an equivalent problem [ @xcite , @xcite ] .
however , the formulation is not in terms of information matrices and the special case of random walks has not been analyzed to our knowledge .
more details are provided in section [ sec4 ] .
the remainder of the paper is organized as follows : in section [ sec2 ] we formulate and investigate the idealized problem of optimal design in the context of filtering for multiple random walks . in section [ sec3 ]
we study its application to tracking flow volumes using sampled data .
we end with discussion of a possible generalization and some comments in section [ sec4 ] .
let us first briefly review the concept of e - optimality from classical design - of - experiment literature for a simple setting .
assume we have _ independent _ observations @xmath0 for @xmath1 .
the natural estimate for @xmath2 is @xmath3 for all @xmath4 .
it is standard to assume that the inverse variance of observation noise is roughly proportional to design variables .
the inverse variance , @xmath5 , can be thought of as the information collected on parameter @xmath2 .
specifically , we assume that the relation between an @xmath6 information vector @xmath7 and an @xmath8 vector of design variables @xmath9 is @xmath10 for example , suppose there is a library of measurements @xmath11 , each of which is independently distributed as @xmath12},\sigma _
i^2 i)$ ] , where @xmath13}$ ] is a subset of elements of @xmath14 .
let @xmath15 be equal to ( or proportional to ) the number of independent measurement of type @xmath4 ( replications of @xmath16 ) collected during the experiment .
then , the weighted least squares estimate @xmath17 of @xmath14 can be shown to have distribution given by ( [ eqny ] ) and ( [ eqnm ] ) .
the matrix @xmath18 depends on the the membership of subsets @xmath13}$ ] and variances @xmath19 ( assumed known ) , for @xmath20 .
we assume that the design variables are constrained to be positive and , in addition , satisfy @xmath21 linear inequality constraints .
these can be written as @xmath22 , where @xmath23 is an @xmath24 matrix and @xmath25 is @xmath26 vector .
we think of this type of constraint as a budgetary one , that specifies upper limits on weighted sums of the design variables . now the e - optimal design problem is given by @xmath27
the objective function , @xmath28 , is the minimum information over all flows .
note that this corresponds to minimizing the maximum mean squared error ( mse ) since @xmath29 is the mse in the estimate of @xmath2 .
using maximum mse as the objective function corresponds to aiming for the best possible worst case performance .
as an example consider the situation where @xmath30 and @xmath31 .
further assume the constraint @xmath32 figure [ objcont1 ] shows the contours of the objective function , that is , @xmath28 .
the region below the thick line is the constraint space . as usual
, the optimal solution corresponds to the point where the contour of the objective function is `` tangent '' to the boundary of the constraint space .
it is clear that the optimal design would be @xmath33 , which is also reasonable from the symmetry of the setup .
we now extend the above criteria to the optimal design for random walks in a steady state .
consider a collection of _ independent _ random walks @xmath34 for @xmath35 and @xmath36 the term @xmath37 is called the innovation noise and we assume that @xmath38 , which is referred to as the innovation variance . further , suppose we have noisy observations @xmath39 let @xmath40 .
as before , we assume the relation between observed information and design variables to be @xmath41 , with @xmath42 matrix @xmath18 assumed known .
the estimates of interest in this case are the ones obtained through filtering @xmath43.\ ] ] let @xmath44 .
further , let @xmath45 when it exists .
we will refer to this as the _ steady state information_. when the innovation and measurement noise , @xmath37 and @xmath46 respectively , are gaussian , the optimal filter corresponds to a kalman filter and in this case the steady state always exists [ @xcite ] . for the remainder of the paper
we will assume that @xmath37 and @xmath46 are independent mean 0 gaussian random variables .
if @xmath47 , then the kalman filter update equations give us @xmath48 and @xmath49 note that , given @xmath50 , @xmath19 and @xmath5 , one can calculate @xmath51 at any time @xmath52 by iterating the above equations .
further , the choice of @xmath5 impacts @xmath51 not just for a specific @xmath52 but for all @xmath52 .
thus , @xmath53 or @xmath54 hence , @xmath55 we define the steady state e - optimal design problem as @xmath56 as an example consider the same setting as above , with @xmath30 and @xmath31 .
further , let the innovation noise be characterized by @xmath57 and @xmath58 .
as before , we assume the design constraint @xmath32 notice in figure [ objcont2 ] that even though there is symmetry in the measured information , the first random walk is smoother than the second one and , hence , less measurement resources need to be allocated to it .
we establish next the main technical result of the paper , that the steady state e - optimal design problem is a second order cone program .
first , we introduce a new variable @xmath59 as the lower bound for the steady state information over all flows . to solve the steady state e - optimal design problem
, we have to maximize @xmath59 subject to @xmath60 for @xmath35 and @xmath61 equation ( [ origcons ] ) can be equivalently written as @xmath62 which is a hyperbolic constraint [ @xcite ] .
thus , this problem can be cast as a second order cone program ( see the for a review of second order cone programs and the representation of the above optimization in canonical form ) .
such optimization programs can be solved efficiently through interior point methods [ @xcite ] , software implementations of which are commonly available [ @xcite , @xcite ] . in the following , we present a greedy alternative to the steady state optimal design .
as before , assume @xmath63 .
further , we assume that @xmath64 ; that is , we allow for time varying design variables @xmath65 with @xmath66 . as before , @xmath44 .
define the information at time @xmath52 to be given by @xmath67 .
note that @xmath68 is a function of @xmath69 the myopic e - optimal design at time @xmath52 is defined as @xmath70 note that the objective function only involves @xmath68 , that is , the information at time @xmath52 . however , the choice of @xmath65 impacts not just @xmath68 but also @xmath71 due to the iterative nature of kalman filtering .
since it ignores this `` long term impact , '' we refer to this scheme as myopic .
equation ( [ eqnkf2 ] ) implies that @xmath72 as before , a new variable @xmath59 can be introduced to lower bound @xmath68 which gives a new set of constraints @xmath73 in addition to the original constraint @xmath74 .
now the objective is to maximize @xmath59 with the optimization variables being @xmath59 and @xmath65 . since both the objective function and the constraints are linear in @xmath65 and @xmath59 , the above optimization is a linear program .
not surprisingly , the myopic optimal design is a much easier problem than the steady - state optimal design even in more general settings as noted in section [ sec4 ] . note that since the sampling rates are allowed to vary with time , the myopic optimal design may have an objective function larger than the steady state optimal case . however , as the objective of optimization is to maximize present information with no regard to impact on future information , such a scheme can not be guaranteed to perform well in the long run .
the ideas developed above can be used for designing the sampling rate in a computer network for tracking flow volumes . as mentioned in the introduction
, we will use the random walk model for flow volumes due to its simplicity and robustness .
suppose there are @xmath75 origin - destination flows in a network .
let @xmath76 be the volume of the @xmath4th flow in the @xmath52-the time interval , for @xmath35 .
these flow volumes are tracked using sampled data which are noisy .
recall that flows are sampled at router interfaces , which we refer to as observation points . in the past ,
a systematic sampling scheme was the dominant technology , but truly random sampling technologies have recently become available and are commonly deployed [ @xcite ] .
all flows traversing an observation point ( router interface ) experience the same sampling rate .
each incoming edge at a node in figure [ geant](b ) is an interface of the corresponding router .
each router typically has multiple interfaces and each flow may traverse multiple observation points due to multi - hop paths and multi - path routing .
suppose there are @xmath77 observation points on the network where sampled data on flows can be collected .
further , assume that sampling rates of @xmath78 are used at observation points @xmath79 , respectively .
any given observation point @xmath80 generates estimates for @xmath81 elements of @xmath82 , that is , the number of flows that go through that node .
thus , a total of @xmath83 measurements are available in each time interval @xmath52 , say , @xmath84 , which need to be optimally combined to get the required estimates .
consider the router in figure [ schema ] .
assume that @xmath85 is the observation point at which the @xmath4th measurement is collected , @xmath86 , and @xmath87 the corresponding flow .
thus , @xmath88 and @xmath89 .
further , let @xmath90=x_{l(i)}(t)\ ] ] and for the moment assume @xmath91 where @xmath92 $ ] . the exact sampling mechanism and approximation involved in the above relation are described in section [ sec3.2 ] .
thus , in vector notation we get @xmath93=lx(t ) , \label{v1}\ ] ] where @xmath94 is a @xmath95 matrix with @xmath96 only if @xmath97 ( i.e. , @xmath4th measurement corresponds to @xmath98 flow ) and @xmath99 otherwise and @xmath100 where @xmath101 is a @xmath102 diagonal matrix , with @xmath103_{ii}=\mu_{l(i)}/\xi_{k(i)}$ ] . using ( [ approx ] ) , the inverse of @xmath101 is given by @xmath104 , where @xmath105 , @xmath106 , are @xmath102 diagonal matrices with their @xmath4th element given by @xmath107_{ii}=\cases { 1/\mu_{l(i)},&\quad $ \mbox{if } k = k(i),$\cr 0,&\quad $ \mbox{otherwise}.$}\ ] ] let @xmath108 be the general least squares estimate of @xmath82 , under equation ( [ v1 ] ) and ( [ v2 ] ) .
thus , @xmath109 from the definition of @xmath94 , it follows that the @xmath98th elements of any two columns of @xmath94 can not be nonzero simultaneously .
thus , the matrix in ( [ mscov ] ) is diagonal .
further , @xmath110 where @xmath111_{ik}=l_{\cdot , i}'\psi_kl_{\cdot , i } .\ ] ] we will refer to the above as the linear model . sampling is employed in network flow measurements because measurement resources like cpu time and available storage are limited .
typically , all observation points ( router interfaces ) belonging to a particular router share these resources .
we assume that the sampling rates are constrained to lie in a convex polygon @xmath112 .
this includes the case where the sum of sampling rates on the interfaces of a router is bounded above by the budget for that router .
we will focus on this constraint for the rest of the paper . in this case
, the constraints are given as one linear inequality for each router .
for the available data , we set up the performance evaluation as follows .
we use the geant network topology , which has @xmath113 nodes ( routers ) and @xmath114 bidirectional edges .
the available data [ @xcite ] correspond to flow volumes over time .
each time interval is equal to 15 minutes .
the original data set spans 4 months , but we focus on the first 200 time intervals to avoid severe non - stationarities inherent in an evolving network .
further , we focus on the top 25% of measured flows by volumes since one is typically interested in tracking heavy flows .
this corresponds to @xmath115 flows .
we assume that sampled data can be collected at each incoming edge of a router and , thus , we have @xmath116 observation points .
we assume that these flows are routed through minimum distance paths , which is a common routing mechanism in wide area networks [ ] .
this leads to @xmath117 and the routing information gives us the mapping @xmath118 and hence the matrix @xmath94 .
matrix @xmath94 is @xmath119 , matrices @xmath120 are all @xmath121 .
we assume that the sum of sampling rates on all interfaces of a router is bounded above by 0.01 , that is , @xmath122 and @xmath123 if observation point @xmath98 is an interface of router @xmath4 and 0 otherwise .
thus , matrix @xmath23 is @xmath124 .
finally , we estimate the @xmath19 and @xmath125 parameters associated with the flow volume processes , and assume they are available for filtering purposes and measurement design . as we have argued ,
both the steady state optimal and myopic design problems are standard optimization programs and once they are written as such , any standard optimization package [ @xcite , ] can be used to solve them numerically . for the purpose of comparison , we also define a naive sampling scheme as follows . for any given router , an equal sampling rate is allocated to every interface that carries any of the 76 flows of interest .
this allocation is done so as to make the corresponding budget constraint tight .
for example , suppose the @xmath4th router has 5 interfaces , but only 4 of them are traversed by one of the 76 flows of interest .
in this case , each of the latter 4 interfaces will be allocated a sampling rate of @xmath126 , while the remaining interface will be allocated a sampling rate of 0 .
figure [ linperf](a ) shows the value of the maximum mse as a function of time .
note that as information accumulates over time , we obtain an improvement in performance under all three sampling mechanisms , myopic , naive and steady state optimal . here
performance is measured as the maximum of @xmath51 over all flows , calculated using equations ( [ eqnkf1 ] ) and ( [ eqnkf2 ] ) .
surprisingly , both the myopic and steady state optimal sampling mechanisms perform equally well in the steady state and achieve a 42% improvement over the naive sampling in the steady state .
figure [ linperf](b ) shows that the myopic optimal sampling rates at all observation points reach a steady state .
figure [ spatial ] shows the value of steady state sampling rates at various router interfaces in the network topology .
even though the myopic scheme has the flexibility of time varying sampling rates , if the sampling rates do reach a steady state , its performance can clearly be no better than the steady - state optimal scheme .
however , as figure [ linperf ] shows , in this case , the additional flexibility permits the myopic scheme to reach steady state performance faster than the steady state optimal one . a more detailed model for flow volumes and sampled measurements would have to include significant departures from the linear model assumed above .
first , the true flow volumes clearly have more structure than independent random walks , as seen in figure [ flowvols ] . in applying the above ideas to the geant data
, we will investigate their robustness to the independent random walk assumption . and sampling rates at various interfaces under a myopic scheme [ panel ] .
] a more serious departure is the following .
suppose that a flow with volume @xmath127 in a certain time interval is sampled at a rate @xmath9 . if the number of sampled packets is @xmath128 , then the usual ( approximate maximum likelihood ) estimate of flow volume is @xmath129 .
the variance of measurement noise can be shown to be @xmath130 [ @xcite ] .
thus , @xmath125 in ( [ approx ] ) is actually equal to the unknown @xmath76 .
the observation above implies that in applying the presented techniques to sampled network data , one would have to rely on an approximate model for measurements @xmath131 .
we will follow an approach similar to batch sequential design [ @xcite ] .
assume that the sampling rates are to be held constant for a batch of contiguous time intervals .
at the beginning of each batch , we use the most recent estimate @xmath132 in place of @xmath125 in ( [ lin ] ) for sampling rate design . for filtering purposes
, we employ a kalman filter with @xmath132 in place of @xmath125 in equation ( [ lin ] ) at each time @xmath52 .
we replace the budget constraint inequalities @xmath22 with the corresponding equalities @xmath133 to force full utilization of available resources . for routers that are traversed by at least one of the 76 flows of interest ,
we introduce additional equality constraints as follows .
design variable @xmath134 for an interface @xmath135 not traversed by one of the 76 flows of interest is constrained to be identically 0 .
figure [ batch ] shows the performance of different sampling schemes averaged over 200 realizations of sampled data .
the sampled data emulate the exact sampling mechanism described above ( with respective sampling rates ) with the geant data treated as the underlying ( unobserved ) flow volumes .
sampling rates were adjusted only at the beginning of a 40 time period block and were held constant over each block . in the first block , the sampling rates were forced to be the same as the naive scheme irrespective of the sampling mechanism under study .
notice that for low values of the objective function ( maximum mean squared error ) the myopic and steady state allocations perform better than the naive allocation . on the other hand , when the maximum mean squared error spikes , the naive allocation performs better , indicating robustness to model departures .
the median ( over time periods 41 to 200 ) of maximum mse for myopic , naive and steady - state optimal sampling is @xmath136 , @xmath137 and @xmath138 , respectively .
thus , the myopic scheme performs better than the steady state optimal scheme , which in turn performs better than the naive scheme .
finally , we look at the performance of the myopic allocation when the above scheme is employed with a block size of just one time interval ; that is , sampling rates were adjusted at the beginning of each time period using the myopic scheme .
the results are displayed in figure [ eachp ] .
as before , the current estimate of flow volumes is used in place of @xmath125 in equation ( [ lin ] ) for both filtering and myopic sampling scheme design .
the myopic sampling scheme can be seen to perform better than the naive version in most time periods .
the median ( over time periods 1 to 200 ) of maximum mse is @xmath139 and @xmath140 for myopic and naive sampling , respectively .
the _ specification _ of the steady state optimal design problem can be easily generalized to linear dynamical systems .
such systems are described by a pair of equations [ @xcite ] .
the state transition equation can be written as @xmath141 where @xmath142 .
the observation equation can be written as @xmath143 assume @xmath144 , where @xmath145 is a linear function and @xmath9 is the value of design variables . for the above dynamical system a kalman filter can be used to iteratively compute @xmath146 $ ] .
let the steady state estimation error covariance be @xmath147 ( assuming the system is observable [ @xcite ] ) .
then , @xmath148 satisfies the algebraic riccati equation : @xmath149 such equations have no analytic solution in general .
the steady state optimal design problem can now be defined as @xmath150 where @xmath151 is an appropriate scalarization of the information matrix [ @xcite ] .
an interesting open problem is to solve the above optimization efficiently in the absence of an analytic solution to ( [ eqnare ] ) .
the sensor placement problem in control system literature [ @xcite ] is equivalent , though not identical .
the newton - type algorithm proposed in [ @xcite ] for this problem requires the solution of the algebraic riccati equation at each iteration of the algorithm .
it would be desirable to develop more efficient algorithms . in summary
, we have shown that steady state e - optimal design for random walks is a second order cone program .
we have illustrated numerically that the performance of the kalman filter can be significantly improved by incorporating an optimal experimental design .
the linear state space model is of general interest and one would like to investigate the steady state optimal design problem described above .
finally , from a practical point of view , it would be useful to extend these ideas to nonlinear filtering .
in this section we summarize the concepts of second order cone programs and hyperbolic constraints from @xcite .
we also present the steady state optimal design problem in the canonical form .
a second order cone program is defined as @xmath152 here , @xmath153 is the optimization variable , and the problem parameters are @xmath154 , @xmath155 , @xmath156 , @xmath157 and @xmath158 .
the norm in the constraints is the standard euclidean norm .
a second order cone program is a standard convex program and algorithms to numerically solve it are well studied and implemented in computational software .
a constraint of the form @xmath159 is called hyperbolic .
the above can be shown to be equivalent to @xmath160 using the above representations , we can write the steady state optimal design problem as a canonical second order cone program as follows . equation ( [ hyperbolic ] ) can be equivalently written as @xmath161 thus , the steady state optimal problem is a second order cone program with @xmath162 , @xmath163 , @xmath164 , @xmath165 for @xmath35 and @xmath166 , for @xmath167 , @xmath168 , for @xmath35 and @xmath169 , for @xmath167 , @xmath170 , for @xmath35 and @xmath171 , for @xmath167 , @xmath172 , for @xmath35 and @xmath173 , for @xmath167 .
the authors would like to thank the editor steve fienberg , the associate editor and two referees for helpful comments and suggestions .
barford , p. , kline , j. , plonka , d. and ron , a. ( 2002 ) . a signal analysis of network traffic anomalies . in _
imw02 : proceedings of the 2nd acm sigcomm workshop on internet measurment _ 7182 .
acm press , new york .
duffield , n. , lund , c. and thorup , m. ( 2002 ) .
properties and prediction of flow statistics from sampled packet streams . in _
imw02 : proceedings of the 2nd acm sigcomm workshop on internet measurment _ 159171 .
acm press , new york .
duffield , n. , lund , c. and thorup , m. ( 2004 ) .
flow sampling under hard resource constraints . in _
sigmetrics04/performance04 : proceedings of the joint international conference on measurement and modeling of computer systems _ 8596 .
acm press , new york .
gautier , r. and pronzato , l. ( 1998 ) .
sequential design and active control . in _
new developments and applications in experimental design _ ( n. flournoy , w. f. rosenberger and w. k. wong , eds . ) 138151 .
ims , hayward , ca .
grant , m. and boyd , s. ( 2009 ) .
cvx : matlab software for disciplined convex programming ( web page and software ) .
available at http://stanford.edu/\textasciitilde boyd / cvx[http://stanford.edu/\textasciitilde boyd / cvx ] .
harvey , a. c. ( 1990 ) .
_ forecasting , structural time series models and the kalman filter_. cambridge univ .
press , cambridge .
soule , a. , lakhina , a. , taft , n. , papagiannaki , k. , salamatian , k. , nucci , a. , crovella , m. and diot , c. ( 2005 ) .
traffic matrices : balancing measurements , inference and modeling . in _ proceedings of the joint international conference on measurement and modeling of computer systems_. acm press , new york . | we examine the problem of optimal design in the context of filtering multiple random walks . specifically , we define the steady state e - optimal design criterion and show that the underlying optimization problem leads to a second order cone program .
the developed methodology is applied to tracking network flow volumes using sampled data , where the design variable corresponds to controlling the sampling rate .
the optimal design is numerically compared to a myopic and a naive strategy .
finally , we relate our work to the general problem of steady state optimal design for state space models . . |
in this note , we announce the classification in @xcite of spectral stability of roll - wave solutions of the `` viscous '' st . venant equations for inclined shallow - water flow , taking into account drag and viscosity . written in nondimensional eulerian form , the shallow water equations for a thin film down an incline are @xmath4 where @xmath1 is the froude number and @xmath5 is the inverse of the reynolds number .
here @xmath6 denotes the fluid height whereas @xmath7 is the fluid velocity averaged with respect to height .
the terms @xmath8 and @xmath9 on the right hand side of the second equation model , respectively , gravitational force and turbulent friction along the bottom .
roll - waves are well - known hydrodynamic instabilities of ( [ swe1 ] ) , arising in the region @xmath10 for which constant solutions , corresponding to parallel flow , are unstable .
they are commonly found in man - made conduits such as aqueducts and spillways , and have been reproduced in laboratory flumes @xcite .
however , up until now , there has been no complete rigorous stability analysis of viscous st .
venant roll - waves either at the linear ( spectral ) or nonlinear level .
roll - waves may be modeled as periodic wave train solutions of ( [ swe1 ] ) . in @xcite , it was proved for a large class of viscous conservation laws and under suitable spectral assumptions that periodic wave trains are nonlinearly stable ( in a spatially - modulated sense ) . in @xcite
this nonlinear analysis has been extended to encompass all periodic wave train solutions of the shallow water system ( [ swe1 ] ) that satisfy those spectral assumptions .
the main issue then is the verification of such assumptions . here
, we provide a complete description of the set of stable roll - waves of ( [ swe1 ] ) : for each froude number @xmath10 , we exhibit ( either theoretically or numerically ) the range of spatial periods where stable roll - waves are found . to our knowledge
, this is the first complete result of stability in the case of shallow water equations .
however , let us mention the study in @xcite : there , the authors studied the _ modulational stability _ of dressler _ inviscid _ roll - waves .
a set of modulation equations is derived by assuming that the parameters which encode the roll - waves slowly vary in time and space : lack of hyperbolicity of the modulation equations is expected to provide a sufficient criterion for spectral instability of roll - waves under special kinds of large scale perturbations . in section [ sec2 ]
, we introduce the spectral problem and recall the spectral assumptions that have to be verified in order to obtain nonlinear stability of periodic waves . in section [ sec3 ] ,
we consider the intermediate froude number regime @xmath11 .
we find a dramatic transition around @xmath12 from the small - f description of stability to a remarkably simple power - law description of surfaces bounding from above and below regions in parameter space corresponding to stable waves .
these surfaces eventually intersect , yielding instability for all sufficiently large @xmath1 . in section [ sec4 ] , we focus on two asymptotic regimes : @xmath2 ( onset ) and @xmath3 . as @xmath2 ,
the shallow water equations reduce to a generalized kuramoto - sivashinsky equation and we obtain asymptotic analytic formula for the stability boundaries . as @xmath3 , we exhibit a non - trivial regime and an asymptotic model which admits only unstable roll - waves , indicating the instability of roll - waves for sufficiently large @xmath1 .
as the full nonlinear theory is given in lagrangian coordinates of mass @xcite , for the sake of consistency we rewrite the viscous shallow water system ( [ swe1 ] ) as @xmath13 where @xmath14 and @xmath15 denotes now a lagrangian marker rather than a physical location @xmath16 , satisfying the relations @xmath17 and @xmath18 .
there is a one - to - one correspondence between periodic waves of the lagrangian and eulerian forms .
it also holds for the spectral problem in its floquet - by - floquet description ; see @xcite .
thus there is no loss of information in choosing to work with the lagrangian form .
now we introduce the spectral problem .
denote by @xmath19 a particular periodic traveling ( roll - wave ) solution of ( [ swe2 ] ) of period @xmath20 .
linearizing ( [ swe2 ] ) about @xmath21 in the co - moving frame @xmath22 and seeking modes of the form @xmath23 , one obtains @xmath24 where primes denote differentiation with respect to @xmath15 . setting @xmath25
, the spectral problem ( [ spec1 ] ) may be written as @xmath26 where @xmath27 is a differential operator with _ periodic coefficients_. by floquet theory , one has that @xmath28 ( the spectrum of @xmath27 acting on @xmath29 ) if and only if there are @xmath30 and @xmath31)$ ] ( a function of period @xmath20 ) such that @xmath32 , where @xmath33 is the corresponding bloch operator defined via @xmath34(x ) .
$ ] consequently , the spectrum may be decomposed into countably many curves @xmath35 of @xmath36)$]-eigenvalues of the operators @xmath33 .
roll - waves are proved to be nonlinearly stable under the following _ diffusive spectral stability conditions _ :
1 . ( d1 ) @xmath37 .
d2 ) there exists a @xmath38 such that for all @xmath39 , @xmath40)}(l_\xi)\subset\{\lambda\,|\,\re(\lambda)\leq -\theta\xi^2\}$ ] .
( d3 ) @xmath41 is an eigenvalue of @xmath42 with generalized eigenspace @xmath43)$ ] of dimension @xmath44 . for a discussion of the significance of these conditions ,
see @xcite . in order to locate the spectrum
, we introduce the evans function @xmath45 .
write ( [ spec1 ] ) as a first order differential system by setting @xmath46 : @xmath47 . denoting the resolvent matrix @xmath48 associated to this system
, it follows that @xmath49)}(l_\xi)$ ] if and only if @xmath50 satisfies @xmath51
in this section , we report on numerical investigations of ( d1 ) , ( d2 ) and ( d3 ) in the regime @xmath11 ( see figure [ fig476 ] ) that is relevant for hydraulic engineering applications @xcite . we exhibit a simple description of the stability region and find that for sufficiently large froude numbers , stable roll - waves do not exist .
our investigation roughly consists of two steps . to determine the global picture of spectrum of a linear @xmath20-periodic operator @xmath27 , we use hill s method , a galerkin based truncation procedure which is implemented into stablab @xcite . however , this method is not sufficient to study the spectrum near the origin and thus to verify hypothesis ( d1 ) , ( d2 ) and ( d3 ) . for that purpose
, we used the evaluation of the evans function and its derivatives on contours to determine the coefficients and estimate the error terms in the expansion of @xmath52 .
a suitable parameterization , available for all froude numbers , is given by @xmath53 , where @xmath54 is the _ total outflow _ and @xmath20 is the period . in @xcite we have gathered numerous pieces of evidence leading to the clear picture that from @xmath1 near @xmath55 and onward , stability is determined by simple relations @xmath56 with higher and higher accuracy as @xmath1 increases , for some universal constants @xmath57 and @xmath58 .
constants providing the lower and upper stability boundary are given approximately by @xmath59 and @xmath60 and @xmath61 respectively . in figure [ fig476 ]
we illustrate this simple rule by providing one slice of the stability diagram obtained by enforcing the arbitrary constraint @xmath62 .
@xmath63{bjnrz5fig87 } & ( b ) \includegraphics[scale=0.25]{bjnrz5fig86 } & ( c ) \includegraphics[scale=0.25]{bjnrz5fig88 } \end{array } $ ]
we now consider the two asymptotic regimes @xmath64 and @xmath3 .
we provide an analytical description of the stability region in the limit @xmath64 .
the limit @xmath3 is studied by a combination of asymptotic expansion and numerical simulations on the limit problems ; roll - waves are always unstable there .
we focus on the onset of roll - waves at @xmath65 .
various weakly nonlinear models have been derived depending crucially on the scaling between time and space and mildly on the precise form of the diffusion term and on whether or not a vanishing viscosity regime is under consideration . in the korteweg - de vries regime @xmath66 and in the small amplitude limit @xmath67 , one obtains the generalized kuramoto - sivashinsky equation @xcite ( up to additional rescaling ) : @xmath68 the spectral and nonlinear stability of periodic traveling waves of ( [ gks ] ) in the limit @xmath69 is fully described in @xcite and a companion paper @xcite .
the classification of stable periodic wave can be extended to the shallow water equations ( [ swe1 ] ) as follows .
+ for @xmath70 sufficiently small , uniformly for @xmath71 on compact sets , periodic traveling waves of ( [ swe2 ] ) are stable for ( lagrangian ) periods @xmath72 and unstable for @xmath73 and @xmath74 $ ] where @xmath75 , @xmath76 , @xmath77 and @xmath78 . 0.3
cm we do not expect other stability regimes when @xmath79 is sufficiently small .
indeed , both in the regime @xmath80 and in the regime @xmath81 amplitude equations have been derived from the shallow water equations indicating that periodic waves are always unstable @xcite , @xcite .
numerical observations support this expectation .
to consider now the infinite - froude number limit @xmath3 , we introduce a suitable rescaling in the equations and profiles with the requirements that ( i ) the limiting system ( @xmath3 ) be nontrivial and ( ii ) the limit be a regular perturbation .
this results in a one - parameter family of rescalings indexed by @xmath82 , given by @xmath83 and @xmath84 where @xmath85 and @xmath86 are real constants . under this rescaling , we find that @xmath20-periodic traveling wave solutions of ( [ swe2 ] ) correspond to @xmath87-periodic solutions to the rescaled profile equation @xmath88 where @xmath89 . noting that the behavior of @xmath90 as @xmath3 depends on whether @xmath91 or @xmath92 , one obtains two classes of limiting profile equations as @xmath3 . an additional rescaling @xmath93 and @xmath94 yields the associated spectral problem @xmath95 where @xmath96 denotes the perturbation of the underlying state @xmath97 .
observe that for @xmath92 the limiting profile equations , selection principles , and spectral problems are independent of the specific value of @xmath98 .
noting that ( [ e : specf->infty ] ) is , again by design , a _
regular perturbation _ of the appropriate limiting spectral problem as @xmath3 , the following sufficient _ instability condition _ is obtained using standard perturbation techniques .
+ for all @xmath82 , the profiles of ( [ rprof ] ) converging as @xmath3 to solutions of the appropriate limiting profile equation , are _ spectrally unstable _ if the appropriate limiting spectral problem about the limiting profiles admit @xmath29-spectrum in @xmath99 with positive real part .
0.3 cm we have investigated the stability of the limiting spectral problems numerically in both the cases @xmath91 and @xmath100 ; recall that the results for @xmath100 in fact hold for all @xmath92 .
this numerical study strongly indicates that , in both cases , all periodic solutions of the appropriate limiting profile equations are _ spectrally unstable _ and hence spectrally stable periodic traveling wave solutions of the viscous st .
venant system ( [ swe2 ] ) do not exist for sufficiently large froude numbers ; see figure [ comparison ] .
@xmath101{bjprz1fig473}&\quad&&(b ) \includegraphics[scale=0.22]{bjprz1fig303 } \end{array } $ ]
we have provided a complete stability diagram in the plane @xmath102 , @xmath1 being the froude number ,
@xmath103 the total discharge rate , @xmath20 the period and @xmath104 the reynolds number . for various parametrizations of the problem
, we found that for each @xmath105 ( for some @xmath106 ) , @xmath107 and @xmath103 fixed in some @xmath108-dependent interval , there exist @xmath109 and @xmath110 such that @xmath20-periodic roll - waves are stable if @xmath111 .
generically , the transition to instability for @xmath112 is due to a loss of hyperbolicity of the whitham modulation equations . on the other hand , the transition to instability for @xmath113 is due to the crossing of a pair of eigenvalues far from the origin and is thus undetectable by similar criteria . up to now ,
we have considered only viscous shallow water equations with turbulent friction terms .
it is an interesting and physically relevant problem to extend our results to more realistic turbulent shallow water models such as ( a viscous version of ) the one derived in @xcite which accurately reproduces brock s experiments on turbulent roll - waves @xcite .
another physically relevant problem is to consider laminar roll - waves as found e.g. in @xcite . in this case
, we would have to take into account surface tension effects as they play there an important role .
14 , stability of viscous st .
venant roll - waves : from onset to infinite - froude number limit , submitted ( 2015 ) arxiv:1503.01154 . | in this note , we announce a complete classification of stability of periodic roll - wave solutions of the viscous shallow - water equations , from their onset at froude number @xmath0 up to the infinite - froude limit . for intermediate froude numbers ,
we obtain numerically a particularly simple power - law relation between @xmath1 and the boundaries of the region of stable periods , that appears potentially useful in hydraulic engineering applications . in the asymptotic regime @xmath2 ( onset ) , we provide an analytic expression of the stability boundaries whereas in the limit @xmath3 , we show that roll - waves are always unstable . |
infrared observations of radio galaxies are a useful tool to explore the physics of their nuclei in the framework of the unified models for active galactic nuclei ( agn ) as in this band the impact of dust absorption is strongly reduced with respect to the optical . in particular , the study of the infrared nuclear sources allows us to investigate the ` scheme ' unifying different classes of agn , by exploring the properties of their accretion disks and the presence of any absorbing material that can account for the complex agn taxonomy . in the `` zeroth - order approximation '' of the agn unification scheme for radio - loud sources ( e.g. @xcite ) , powerful radio galaxies with fr ii edge - brightened morphology @xcite
are believed to be misaligned quasars , while lower power , edge - darkened fr is are associated with bl lac objects .
however , it is clear that this zeroth order approximation picture based on statistical comparison of the properties of radio - loud agn is over - simplified .
inconsistencies between the characteristics of the `` parent '' and `` beamed ''
populations are well reported in the literature : for example , several bl lacs show broad emission lines ( e.g. @xcite ) ; and there are inconsistent environments ( e.g. @xcite ) and inconsistent radio morphologies ( e.g. @xcite ) ( see also @xcite and @xcite for reviews on this subject ) .
furthermore , how the large - scale radio properties relate to different nuclear properties is still not clear .
in fact , there is mixed evidence for a correlation between the morphological fr i / fr ii dichotomy and different levels of nuclear activity @xcite .
another classification , based on the optical narrow emission lines ratios @xcite introduces two groups of radio - loud agn : the low excitation galaxies ( leg ) and high excitation galaxies ( heg ) .
radio galaxies with fr i radio morphology almost always belong to leg class , while nearly all hegs are fr ii radio sources .
however , there is a group of fr ii legs whose place in the unifying scenario is still largely not understood . @xcite and
@xcite proposed that such objects constitute a single population of radio galaxies together with fr is . in line with this idea ,
the two spectroscopic classes , leg and heg , show significant differences in several properties : star formation ( e.g. @xcite ) , environment ( e.g. @xcite ) , triggering mechanism of nuclear activity ( gas - rich or gas - poor merger ) ( e.g. @xcite ) ; there are indications that they differ also in terms of accretion rate and radiative efficiency of disk ( e.g. @xcite ) ; contrarily to what is seen in heg , leg do not show evidence for large absorbing column densities in the x - ray spectra ( e.g. @xcite ) , for warm dusty tori in mid - ir images @xcite , and for broad lines in their optical spectra ( e.g. @xcite )
. taken together these differences suggest that high and low excitation galaxies belong to different classes of radio - loud nuclear activity which are associated with distinct environmental conditions .
furthermore , it has been suggested that in the framework of the unification model , heg and leg might represent the parent population of qsos and bl lacs respectively ( e.g. @xcite ) . a number of recent papers in the literature address the topic of the infrared view of the active nucleus of radio galaxies .
@xcite show that the infrared emission at both 24 and 70 micron can be explained in terms of dust heated predominantly by the agn .
more comprehensive mid - ir and far - ir spectrophotometry for the z @xmath1 1 3cr objects has been obtained from iso ( @xcite ) as well as from spitzer ( e.g. , @xcite ) .
for high - redshift 3cr sources near infrared spectrophotometry using spitzer has been presented by @xcite and @xcite .
the mid - infrared observations are sensitive to thermal emission from warm dust and have revealed rather complex properties within the different radio - galaxy classifications .
the majority of fris are faint at those wavelengths , implying that the bulk kinetic power of the radio jets is the energetically dominant mechanism in most of these sources @xcite . even in fr
is this is not a universal conclusion as @xcite with iso data and @xcite with spitzer data show that in some fr is a dust emission bump is detected . on the other hand
, @xcite found that about 50% of the fr ii narrow - line radio galaxies have mid - infrared photometric and spectroscopic properties comparable to quasars of matched radio luminosity , while there is a subsample of mid - ir weak fr ii radio galaxies which may constitute a separate population of non - thermal , jet - dominated sources with low accretion power . @xcite present evidence that the mir data do in fact support the orientation - dependent unification scheme of the powerful fr ii galaxies with quasars .
this has been subsequently confirmed by further work , e.g. @xcite .
this paper has the aim of investigating agn activity by focusing on the nuclear 1.6 @xmath0 m near infrared emission , identifying its physical origin and relating it to both the radio morphological and optical spectroscopic classes .
@xcite studied the hst optical images of the nuclear regions of radio galaxies belonging to 3cr sample with the same purpose .
the nir snapshot survey of 3cr objects with @xmath2 provided us with a larger coverage of the sample than that in the optical band .
furthermore , the fraction of 3cr sources with spectroscopic classification is now substantially larger @xcite , allowing us to explore in greater detail the behavior of the nuclei of each class .
the resolution of hst images is suitable for this purpose and allow us to distinguish the nir emission of stellar host galaxy background from the genuine agn radiation .
the structure of the paper is as follows . in sect . 2
we describe the sample and in sect .
3 we derive the surface brightness profiles which we fit with either srsic or core - srsic models .
the multiwavelength properties of our sample are discussed in sect . 4 in order to identify the origin of the nir nuclear emission tackled in sect .
5 . in sect .
6 we summarize and draw conclusions . in appendix
we compare the fr i radio core flux observed with vla and vlbi .
we adopt a hubble constant of h@xmath3 = 75 km s@xmath4 mpc@xmath4 and q@xmath3 = 0.5 .
we used this cosmology in order to be in accordance with the similar results obtained by @xcite , @xcite , @xcite . assuming a different cosmology would not affect our results since the redshift of the sources considered in this paper is limited to @xmath5 .
the sample of galaxies we analyzed belongs to the revised third cambridge catalog 3cr ( @xcite , @xcite , @xcite ) .
since it is selected based on the low - frequency radio flux ( s@xmath6 @xmath7 10 jy ) , the sample is essentially free from orientation bias .
we chose all 116 3cr sources with z @xmath1 0.3 with the primary goal of characterizing the radio galaxy hosts mostly free from the effects of dust .
one hundred objects were observed as part of hst / nicmos 3cr snapshot program go 10173 ( pi : sparks ) .
the nir images are published in @xcite and @xcite .
only eleven galaxies were observed as a part of other programs and details of these are given in table [ log ] .
we also included in the sample the fr i radio galaxy ngc 6251 because , as pointed out by @xcite it should be included in 3cr catalog .
all of the objects of our sample were observed with hst nicmos camera 2 ( nic2 ) , which has a field of view of 19@xmath82 @xmath9 19@xmath82 and a projected pixel size of 0.075 , using the f160w filter ( similar to h band ) , which is centered at 1.60 @xmath0 m , covering a wavelength range from 1.4 to 1.8 @xmath0 m and which includes the pa@xmath10 line for objects with z @xmath7 0.1 . using the broad and narrow line h@xmath11 fluxes measured for 3cr galaxies @xcite over a 2@xmath12 aperture
, we have estimate the contamination from line emission in the h band . in the case of the narrow lines we assumed ratio pa@xmath10/h@xmath11 = 0.06 from atomic physics @xcite .
since the broad line galaxies are know not to conform to case b we have used the observed near infrared broad line region ratios observed in type 1 agn by @xcite and @xcite to estimate the contamination in the f160w filter .
this analysis shows the contamination to be @xmath13 15% for broad line objects and
@xmath13 4% for narrow line objects and in both cases that is small enough not to affect our conclusions regarding the origin of the nuclear nir emission .
due to technical difficulties , 3c 410 and 3c 442 were only observed for 575.9 s , while all the remaining galaxies were observed for a total exposure time of 1151.8 s. a detailed description of the observations and data reduction can be found in @xcite .
ccc name & date & t@xmath14 + & yy - mm - dd & s + 3c 084 & 1998 - 03 - 16 & 639.9 + 3c 264 & 1998 - 05 - 12 & 447.7 + 3c 270 & 1998 - 04 - 23 & 767.8 + 3c 272.1 & 1998 - 07 - 13 & 767.8 + 3c 274 & 1997 - 11 - 20 & 127.9 + 3c 293 & 1998 - 08 - 19 & 2751.8 + 3c 305 & 1998 - 07 - 19 & 2783.8 + 3c 317 & 1998 - 08 - 26 & 639.8 + 3c 338 & 1997 - 12 - 18 & 4159.8 + 3c 405 & 1997 - 12 - 16 & 1343.9 + ngc 6251 & 1998 - 07 - 06 &
543.5 + [ log ] each galaxy was classified in two ways .
firstly , into fr i and fr ii based on radio morphology @xcite : the radio morphological classification , mostly being adopted from @xcite . secondly ,
based on the optical broad line strengths and narrow emission lines ratios ( from @xcite ) we classified them into the stand groupings of high and low ionization galaxies ( hig and lig ) and broad line objects ( blo ) .
note that the blo objects belong to hig class because they show high - ionization narrow emission lines .
we adopt the hig / lig nomenclature to better represent the physical condition of the narrow line region gas in these objects .
this classification is however entirely consistent with the heg / leg scheme widely adopted in the literature ( e.g. , @xcite ) .
following @xcite we introduced a 4th class of extremely - low excitation galaxies ( eleg ) , characterized by an extremely low value of [ o iii]/h@xmath10 @xmath15 0.5 , @xmath156 times lower than usually observed in 3cr / lig . in this work
we consider elegs as a separate class of objects .
the objects for which we do not have new spectral data are classified according to @xcite . in the sample
there is also one galaxy , 3c 198 , characterized by a star forming spectrum ( marked as `` sf '' in table [ table1 ] ) and a bl lac object ( 3c 371 ) which are not considered in the following statistical evaluations because of their unique properties among the objects of our sample . for one galaxy ( 3c 410 ) we could not find any reliable spectral classification in the literature .
the aim of this paper is to measure the nir nuclear sources . in order to disentangle such a component from the underlying stellar emission of the host galaxy , it is necessary to model the distribution of the galactic light . to this purpose
we derived the surface brightness profile of all our sources .
radial profiles can be derived using a number of different methods : ellipse fitting to the isophotes of the galaxies ( e.g. , @xcite ) obtaining a one - dimensional analysis ( e.g. , @xcite ) , and full two - dimensional analysis ( e.g. , @xcite ) . in this paper
, we adopted the first approach .
the `` ellipse '' task in iraf stsdas @xcite is suitable to analyzing the axisymmetric structure , evident in most objects .
this task allows us to trace the radial profile and radial changes in ellipticity , position angle , and coordinates of the center of the isophotes .
most of the host galaxies of 3cr sources appear as smooth ellipticals in infrared . only in few cases
are distortions due to dust lanes , jets , or plumes . before deriving the galaxy profile ,
we mask out all spurious objects such as stars , jets , dust lanes , bright , galaxy companions and residual diffraction patterns associated with bright stars . when dust features are present in the nuclear region we adopted a multi - stage strategy .
to stabilize the fit in the radial range affected by the dust we started by fixing some or all fit - parameters to the values of the last isophote that is not affected by the dust .
we then produce an initial model image and we subtracted it from the original data to unveil the presence of further faint sources and/or dust features , that were then also masked out and so on until no further feature were found . in fig .
[ esempio ] we show the stages of the procedure to derive and to fit the radial brightness profile .
the surface brightness profile can be fitted using a variety of functions .
we used two different models , srsic and core - srsic .
the former @xcite is a law of the form : @xmath16}\ ] ] where @xmath10 is the concentration parameter and @xmath10 = 1/n , the inverse of the srsic index n. i@xmath17 is the intensity at the effective radius r@xmath18 . for a srsic model with 0.5 @xmath13 n @xmath13 10 , b@xmath19 @xmath20 1.9992n - 0.3271 ( e.g. , @xcite )
. the latter is composed of an outer srsic model and a inner power - law profile , suitable to describe the innermost part of the profile when the object shows a deficit toward the center with respect to the srsic law .
@xcite argued that it is convenient to restrict the core - srsic model to an infinitely sharp transition between the srsic and power - law profiles at the break radius r@xmath21 , assuming this form : @xmath22\ ] ] where r@xmath21 separates the inner power law of logarithmic slope @xmath23 from the outer part of the profile described by a srsic model .
@xmath24(r@xmath21-r ) is the heaviside step function .
we fit the brightness profiles with both models by minimizing the residuals , leaving free to vary the 5 ( or 7 ) variables : the background level , the three parameters describing the srsic ( or five for core - srsic ) model and the nuclear component . fig . [ all ] illustrates four examples of fits to the nir brightness profiles . in sect .
[ serocore ] we describe the method used to select the appropriate law for each galaxy . to take into account the psf effects , we convolved the models prior to the comparison with the data @xcite .
the psf of nicmos - nic2 at the center of the chip was derived using the tinytim program written by j. krist .
the psf image was then convolved with a model galaxy with a power law profile , which represents the innermost region of a srsic / core - srsic model .
this was done for a set of galaxy models with different profile slopes @xmath25 , all of them having zero ellipticity .
we then derived the radial profile of the convolved model galaxies using circular isophotes , and we calculated the intensity ratio between the convolved model galaxy and the original model galaxy for different radii . with this method we obtained a set of radial ratios for different values of the innermost slope that can be applied to the models to `` simulate '' the two - dimensional convolution .
the appropriate value of @xmath25 was measured as part of the fitting procedure . in the case of a srsic model
, @xmath25 is the value of the logarithmic slope of the profile at 0@xmath81 , which approximately corresponds to the resolution of nicmos - nic2 . in the core - srsic model
, @xmath25 is the logarithmic slope of the inner power - law profile .
@xcite already performed the fitting of the radial brightness profiles of 3cr sample with srsic law , focusing on their large scale behavior , thus not including the central region of the galaxies .
for the objects in which @xcite found a disk - like component in the galaxy profile , we exclude from the fit the regions dominated by the disk .
the parameter values found are mostly consistent with those of @xcite for the objects in which the nuclear component is not dominant with respect to the galaxy stellar emission .
the resulting models produce residuals in the form of large scale fluctuations with a typical amplitude of @xmath1 5% , indicative of the goodness of the fit .
besides the parameters describing the host galaxy profile , the fitting procedure also yields the flux of the nuclear source . to assess the reliability of the detection of a nuclear source we adopt the following operative approach .
we extract the radial brightness profile using the iraf `` radprof '' task and measure the nuclear fwhm , setting the background level at the intensity measured just beyond the nuclear point source at a radius of @xmath150.3 .
if the fwhm is consistent with that of the psf ( @xmath15 0.14 arcsec ) , we took this as an indication of the presence of an unresolved source . alternatively ,
if there was no evidence for a nuclear point source , we proceed to fit the profile with a core - srsic model leaving all parameters free to vary , but fixing @xmath26 .
this corresponds to the most conservative choice for @xmath23 as it minimizes the galaxy contribution and maximizes the nuclear source .
we consider the nuclear flux thus derived as the upper limit to the intensity of the central source .
the choice between srsic and core - srsic model for a given galaxy is related to the corresponding minimized @xmath27 value obtained after the fitting procedure .
since the core - srsic model includes more free parameters , the best core - srsic model usually will reproduce better the galaxy profile than the srsic model .
therefore , to determine the goodness of the fit , we use the `` reduced '' @xmath27 ( @xmath28 , where @xmath29 is the number of fitted data points and @xmath30 is the number of fitted parameters , which can be either 3 or 5 for srsic or core - srsic models , respectively ) .
we consider the best model that corresponds to the @xmath31 closest to the unity .
qualitatively , the choice can be justified by the form of the residuals . in fact , when fitting a core - srsic galaxy with a pure srsic model , the residuals in the central region show the characteristic s - shape , already discussed by @xcite .
nevertheless , since our objects are at distances of up to z = 0.3 , we can misclassify a genuine core galaxy as a pure srsic due to an insufficient physical resolution of the image .
the typical core size of an elliptical galaxy is @xmath15200 pc ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) which corresponds to 0@xmath814 at z = 0.15 .
considering the objects at z @xmath7 0.15 with detected nuclei ( excluding the blos , for which the nucleus outshines the host galaxy emission in the central region ) , for which there is ambiguity on the presence of a core , the typical discrepancy between the estimates of the nuclear luminosity obtained adopting either a srsic law or a core - srsic model is @xmath32 % , and at most a factor of 2 .
the sources potentially involved in this ambiguity are only @xmath1515% of the sample and thus they do not alter the main results of this paper .
the presence of a bright point source may also limit our ability to see a possible shallow core and to discriminate between srsic and core - srsic models .
however , in those cases ( e.g. in the case of the blos , fig.[sersiccores ] ) , the difference between the nuclear fluxes obtained adopting the two different models is irrelevant , being smaller than @xmath33% , thus well below the typical photometric error .
in table [ table1 ] we report the 1.6 @xmath0 m nir nuclear fluxes and luminosities for the sample measured with the procedure explained in sect .
[ method ] .
we also give the optical ( r band ) , radio ( nuclear at 5 ghz and large scale at 178 mhz ) fluxes , and [ o iii ] emission line luminosities for all the objects of the sample .
l|ccc|ccc|ccc|cc|c name & z & fr - e.l . & ref & log f@xmath6 & log f@xmath34 & ref & log f@xmath35 & log f@xmath36 & log l@xmath36 & log l@xmath37}$ ] & ref & profile + 3c 15 & 0.073 & lig & b09 & -21.80 & -23.42 & z95 & @xmath1 - 28.05 & -27.42 & 27.55 & 40.46 & b09 & s + 3c 17 & 0.2197 & blo & b09 & -21.70 & -23.08 & z95 & -26.91 & -26.84 & 29.00 & 41.68 & b09 & s + 3c 20 & 0.174 & hig & j97 & -21.37 & -25.48 & f97 & - & @xmath1 - 27.72 & @xmath127.94 & 41.47 & l96 & cs + 3c 28 & 0.1953 & eleg & b09 & -21.79 & @xmath1 - 26.70 & g88 & @xmath1 - 29.28 & @xmath1 - 29.64 & @xmath126.10 & 40.67 & b09 & cs + 3c 29 & 0.0488 & fr i & b09 & -21.82 & -24.04 & m93 & -28.02 & -27.56 & 27.03 & 40.06 & b09 & s + 3c 31 & 0.017 & fr i & b09 & -21.77 & -24.04 & g88 & -27.61 & -27.44 & 26.29 & 39.41 & b09 & cs + 3c 33 & 0.059 & hig & b09 & -21.28 & -24.52 & z95 & - & @xmath1 - 27.14 & @xmath127.64 & 42.05 & b09 & cs + 3c 33.1 & 0.1809 & blo & b09 & -21.89 & -24.66 & z95 & -27.5 & -27.12 & 28.57 & 42.03 & b09 & s + 3c 35 & 0.067 & lig & b09 & -21.98 & -24.62 & z95 & @xmath1 - 28.41 & @xmath1 - 28.30 & @xmath126.59 & 39.87 & b09 & cs + 3c 52 & 0.28540 & lig & b09 & -21.87 & -25.11 & c08 & - & @xmath1 - 28.85 & @xmath127.18 & @xmath140.84 & b09 & cs + 3c 61.1 & 0.184 & hig & b09 & -21.51 & -25.43 & z95 & - & @xmath1 - 28.57 & @xmath127.13 & 42.20 & b09 & cs + 3c 66b & 0.0215 & fr i & b09 & -21.61 & -23.74 & g88 & -27.09 & -27.04 & 26.91 & 39.97 & b09 & cs + 3c 75n & 0.02215 & fr i & b09 & -21.59 & -24.29 & z95 & - & @xmath1 - 28.48 & @xmath125.48 & @xmath139.80 & b09 & cs + 3c 76.1 & 0.03249 & fr i & b09 & -21.88 & -24.61 & g88 & - & @xmath1 - 28.29 & @xmath125.99 & @xmath139.75 & b09 & cs + 3c 79 & 0.256 & hig & b09 & -21.52 & -24.76 & z95 & -27.62 & -27.44 & 28.51 & 42.51 & b09 & s + 3c 83.1 & 0.0251 & fr i & b09 & -21.58 & -24.67 & r75 & -28.64 & -27.65 & 26.43 & @xmath139.33 & b09 & cs + 3c 84 & 0.01756 & fr i & b09 & -21.39 & -21.37 & t96 & -25.61 & -25.47 & 28.28 & 41.53 & b09 & s + 3c 88 & 0.03022 & lig & b09 & -21.82 & -23.72 & z95 & -27.62 & -27.46 & 26.76 & 40.05 & b09 & s + 3c 98 & 0.03 & hig & b09 & -21.45 & -24.97 & z95 & @xmath1 - 28.55 & @xmath1 - 27.64 & @xmath126.58 & 40.90 & b09 & cs + 3c 105 & 0.089 & hig & b09 & -21.75 & -24.76 & z95 & - & -27.82 & 27.3 & 41.29 & b09 & s + 3c 111 & 0.0485 & blo & b09 & -21.19 & -22.93 & z95 & -26.48 & -25.30 & 29.33 & 42.33 & b09 & s + 3c 123 & 0.2177 & lig & b09 & -20.72 & -24.01 & h98b & @xmath1 - 29.39 & @xmath1 - 28.16 & @xmath127.67 & 41.69 & b09 & cs + 3c 129 & 0.0208 & fr
i & b09 & -21.33 & -24.99 & v82 & - & -27.48 & 26.43 & @xmath139.75 & b09 & s + 3c 129.1 & 0.02220 & fr i & b09 & -21.98 & -25.51 & c08 & - & @xmath1 - 28.52 & @xmath125.44 & @xmath140.39 & b09 & cs + 3c 130 & 0.032 & fr i & h98a&-21.81 & -23.48 & h99 & - & @xmath1 - 28.04 & @xmath126.24 & @xmath140.59 & b09 & cs + 3c 132 & 0.214 & lig & j97 & -21.91 & @xmath1 - 24.42 & z95 & @xmath1 - 28.99 & @xmath1 - 28.03 & @xmath127.78 & - & - & cs + 3c 133 & 0.2775 & hig & b09 & -21.65 & -23.69 & n98 & -28.01 & -26.82 & 29.18 & 42.39 & b09 & s + 3c 135 & 0.1253 & hig & b09 & -21.76 & -25.22 & z95 & -28.04 & -27.60 & 27.80 & 41.85 & b09 & s + 3c 153 & 0.277 & lig & b09 & -21.78 & @xmath1 - 26.28 & h98b & @xmath1 - 29.35 & @xmath1 - 27.94 & @xmath128.07 & 41.26 & b09 & cs + 3c 165 & 0.2957 & lig & b09 & -21.87 & -24.97 & h98b & -28.77 & -28.39 & 27.67 & 41.28 & b09 & s + 3c 166 & 0.245 & lig & b09 & -21.83 & -23.19 & z95 & -27.91 & -27.28 & 28.63 & 41.32 & b09 & s + 3c 171 & 0.2384 & hig & b09 & -21.71 & -25.54 & z95 & -29.40 & @xmath1 - 28.05 & @xmath127.84 & 42.55 & b09 & cs + 3c 173.1 & 0.2921 & lig & b09 & -21.81 & -24.86 & z95 & - & -28.18 & 27.86 & 40.47 & b09 & cs + 3c 180 & 0.2200 & hig & b09 & -21.82 & - & - & - & @xmath1 - 28.50 & @xmath127.34 & 42.03 & b09 & cs + 3c 184.1 & 0.1182 & blo & b09 & -21.89 & -25.10 & z95 & -27.18 & -26.60 & 28.76 & 42.02 & b09 & s + 3c 192 & 0.0598 & hig & b09 & -21.64 & -25.07 & z95 & @xmath1 - 27.81 & @xmath1 - 27.92 & @xmath126.88 & 41.22 & b09 & cs + 3c 196.1 & 0.198 & lig & b09 & -21.73 & -24.11 & b89 & - & @xmath1 - 28.57 & @xmath127.18 & 41.23 & b09 & cs + 3c 197.1 & 0.1301 & blo & b09 & -22.09 & -25.13 & z95 & -27.33 & -27.32 & 28.11 & 40.73 & b09 & s + 3c 198 & 0.0815 & sfg & b09 & -22.01 & @xmath1 - 25.77 & f78 & -27.02 & -26.96 & 28.09 & 40.82 & b09 & s + 3c 213.1 & 0.1937 & lig & b09 & -22.18 & -23.22 & a95 & - & -27.52 & 28.22 & 40.78 & b09 & s + 3c 219 & 0.1744 & blo & b09 & -21.39 & -24.12 & z95 & -27.02 & -26.95 & 28.71 & 41.51 & b09 & s + 3c 223 & 0.1368 & hig & b09 & -21.83 & -24.90 & z95 & @xmath1 - 28.20 & -27.56 & 27.91 & 41.95 & b09 & s + 3c 223.1 & 0.107 & hig & b09 & -22.22 & -25.03 & h98b & @xmath1 - 28.09 & -26.65 & 28.62 & 41.39 & b09 & s + 3c 227 & 0.0861 & blo & b09 & -21.52 & -24.62 & z95 & -26.26 & -25.70 & 29.40 & 41.60 & b09 & cs + 3c 234 & 0.185 & hig & b09 & -21.58 & -23.83 & z95 & -26.61 & -26.04 & 29.66 & 42.84 & b09 & s + 3c 236 & 0.1005 & lig & b09 & -21.84 & -23.71 & z95 & @xmath1 - 28.18 & -27.50 & 27.72 & 40.72 & b09 & s + 3c 258 & 0.1650 & lig & b09 & -22.01 & - & - & - & @xmath1 - 28.38 & @xmath127.24 & 39.94 & b09 & cs + 3c 264 & 0.0206 & fr i & b09 & -21.58 & -23.70 & g88 & -26.62 & -26.64 & 27.28 & 39.07 & b09 & cs + 3c 270 & 0.0074 & fr i & b09 & -21.26 & -23.51 & g88 & -27.97 & -27.05 & 25.96 & 37.89 & b09 & cs + 3c 272.1 & 0.0037 & fr i & b09 & -21.71 & -23.75 & g88 & -26.88 & -27.07 & 25.35 & 38.13 & b09 & cs + 3c 274 & 0.0037 & fr i & b09 & -19.98 & -22.40 & g88 & -26.06 & -25.52 & 26.90 & 38.80 & b09 & cs + 3c 277.3 & 0.0857 & hig & b09 & -22.05 & -24.91 & g88 & -28.61 & @xmath1 - 27.08 & @xmath128.09 & 40.78 & b09 & cs + 3c 284 & 0.239 & hig & b09 & -22.21 & -25.65 & z95 & - & @xmath1 - 28.53 & @xmath127.37 & 41.26 & b09 & cs + 3c 285 & 0.0794 & hig & b09 & -21.95 & -25.10 & z95 & -29.38 & -27.93 & 27.11 & 40.40 & b09 & s + 3c 287.1 & 0.2159 & blo & b09 & -22.09 & -23.29 & z95 & -27.11 & -26.41 & 29.42 & 41.43 & b09 & s + 3c 288 & 0.246 & fr i & z95 & -21.72 & -24.53 & g88 & -28.94 & -28.71 & 27.40 & - & - & cs + 3c 293 & 0.0452 & lig & b09 & -21.90 & -24.81 & g88 & - & -28.09 & 26.47 & 39.70 & b09 & cs + 3c 296 & 0.0237 & fr i & b09 & -21.89 & -24.10 & g88 & -28.25 & -28.13 & 25.92 & 39.65 & b09 & cs + 3c 300 & 0.27 & hig & b09 & -21.75 & -24.92 & z95 & -28.20 & -27.83 & 28.16 & 41.64 & b09 & s + 3c 303 & 0.141 & blo & b09 & -21.95 & -23.69 & z95 & -26.83 & -26.49 & 29.01 & 41.51 & b09 & s + 3c 305 & 0.04164 & hig & b09 & -21.80 & -24.50 & g88 & - & @xmath1 - 28.10 & @xmath126.40 & 40.93 & b09 & cs + 3c 310 & 0.054 & fr i & b09 & -21.26 & -24.10 & g88 & -28.24 & @xmath1 - 27.07 & @xmath127.69 & 39.95 & b09 & cs + 3c 314.1 & 0.1197 & eleg & b09 & -21.98 & @xmath1 - 26.00 & g88 & @xmath1 - 28.81 & @xmath1 - 27.37 & @xmath128.09 & 39.49 & b09 & cs + 3c 315 & 0.1083 & fr i & b09 & -21.75 & -23.75 & g88 & - & -27.76 & 27.52 & 40.69 & b09 & cs + 3c 317 & 0.0342 & fr i & b09 & -21.33 & -23.41 & m93 & -27.67 & -27.05 & 27.31 & 40.25 & b09 & cs + 3c 319 & 0.192 & lig & b09 & -21.82 & @xmath1 - 25.80 & z95 & @xmath1 - 28.33 & @xmath1 - 27.77 & @xmath127.96 & @xmath139.91 & b09 & cs + 3c 321 & 0.096 & hig & b09 & -21.87 & -24.41 & z95 & - & -27.64 & 27.55 & 40.73 & b09 & cs + 3c 323.1 & 0.264 & blo & b09 & -22.00 & -24.28 & z95 & -25.83 & -25.77 & 30.20 & 42.45 & b09 & s + 3c 326 & 0.089 & lig & b09 & -22.11 & -24.78 & z95 & @xmath1 - 28.04 & -27.87 & 27.25 & 40.23 & b09 & cs + 3c 332 & 0.1515 & blo & b09 & -22.02 & -24.90 & z95 & -26.76 & -26.08 & 29.47 & 41.57 & b09 & s + 3c 338 & 0.03035 & fr i & b09 & -21.33 & -23.99 & g88 & -27.78 & -28.17 & 26.08 & 39.44 & b09 & cs + + [ cols="<,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] description of the table : col.(1 ) : name ; col.(2 ) : redshift ; col.(3 ) : fanaroff & riley and optical spectral classifications .
for simplicity , we mark as fr i all radio galaxies with fr i - like radio morphology and lig - like optical spectrum , as lig all radio galaxies with fr ii - like radio morphology and lig - like optical spectrum , as hig ( blo ) all radio galaxies with fr ii - like radio morphology and hig - like optical spectrum with the absence ( presence ) of broad emission lines .
references ( col .
4 ) : b09 @xcite , j97 @xcite , h98a @xcite , z95 from @xcite , s91 @xcite , o75 @xcite , d96 @xcite , p84 @xcite ; col.(5 ) : flux density at 178 mhz in erg s@xmath4 @xmath38 hz@xmath4 ( in these logarithmic units -23 corresponds to 1 jy ) ; col.(6 ) : nuclear radio flux density at 5 ghz in erg s@xmath4 @xmath38 hz@xmath4 . references ( col .
7 ) : b92 @xcite ( 8.3 ghz ) , g88 : @xcite , l91 : @xcite , m93 : @xcite , r75 @xcite , t96 @xcite , z95 @xcite , n98 @xcite , h98b @xcite ( 8.4 ghz ) f97 : @xcite , v82 : @xcite , h99 : @xcite , b89 : @xcite , f78 : @xcite , a95 : @xcite , p81 : @xcite , s85 : @xcite ( 1.5ghz ) , s84 : @xcite , j86 : @xcite , c08 : a. capetti , priv . com .. radio core data at frequencies different from 5
ghz were converted to 5 ghz using a flat ( @xmath11 = 0 ) spectral index ; col.(8 ) : nuclear optical flux density in erg s@xmath4 @xmath38 hz@xmath4 ( @xcite , @xcite and @xcite ) in r band ; col.(9 ) : nuclear 1.6 @xmath0 m near infrared flux density in erg s@xmath4 @xmath38 hz@xmath4 ; col.(10 ) : nuclear 1.6 @xmath0 m near infrared luminosity in erg s@xmath4 hz@xmath4 ; col.(11 ) : [ o iii]@xmath395007 luminosity in erg s@xmath4 . references ( col.12 ) : b09 @xcite , o75 @xcite , s81 @xcite , l96 @xcite , r89 @xcite .
col.(13 ) : type of model used to fit the nir surface brightness profile . here
we examine the detection rate of the nir central sources ( table [ detect ] ) , considering each radio - morphological and optical - spectral class separately .
overall , we detect nuclei in 64%@xmath405% of the entire 3cr sample . we detect nuclei in 81%@xmath408% and 58%@xmath406% of the fr
is and fr iis , respectively . among the fr
iis , all of the blos have an unresolved nuclear point source , while the detection rate of nuclei in ligs and higs is 44%@xmath4010% and 54%@xmath409% , respectively .
the only one bl lac included in the sample has a bright nuclear source .
the nucleus of 3c 410 , the only spectroscopically unclassified fr ii , is detected , while neither the elegs nor the star forming galaxy 3c 198 have an nir nucleus .
the different detection rates in the fr is and fr iis are not a consequence of the redshift distribution of the 3cr catalog .
all fr ii sources ( hig , lig , and blo ) in our sample share a similar median redshift , and only the fr is have a lower median @xmath41 , because of their lower total power . due to the presence of upper limits
, we use the kaplan - meier product - limit estimator @xcite , which provides the mean value for a distribution with censored data , where the censoring is random , to measure the average value of nir nuclear luminosities of each class ( reported in table [ detect ] ) .
the blos have , on average , the brightest near infrared nuclear sources in the sample in luminosity ( @xmath1l@xmath36@xmath7 = 1.6@xmath910@xmath42 erg s@xmath4 hz@xmath4 ) . the fr i nuclei are on average the faintest ( @xmath1l@xmath36@xmath7 = 4.0@xmath910@xmath43 erg s@xmath4 hz@xmath4 ) .
fr ii ligs and higs share a similar range of nir luminosity and appear to bridge the gap between blos and fr is .
undetected nir nuclei are present for f@xmath44 10@xmath45 erg @xmath38 s@xmath4 hz@xmath4 and l@xmath44 2.5@xmath910@xmath46 erg s@xmath4 hz@xmath4
. however about one half of the detected nuclei have fluxes and luminosities below these values .
c|cccc |c class & nuclei & upper lim . & tot .
& @xmath47 & log l@xmath36 + fr i & 21 & 5 & 26 & 81 & 26.7 @xmath40 0.2 + fr ii & 42 & 31 & 73 & 58 & @xmath48 + lig & 11 & 14 & 25 & 44 & 26.9 @xmath40 0.2 + hig & 15 & 13 & 28 & 54 & 27.5 @xmath40 0.2 + blo & 15 & 0 & 15 & 100 & 29.2 @xmath40 0.1 + eleg & 0 & 3 & 3 & 0 & @xmath48 + sfg & 0 & 1 & 1 & 0 & @xmath48 + undef & 1 & 0 & 1 & 100 & @xmath48 + bllac & 1 & 0 & 1 & 100 & @xmath48 + tot . &
64 & 36 & 100 & 64 & + [ detect ] it is instructive to compare the nir properties with those of the optical nuclei . in fig .
[ oir ] , left panel , we plot f@xmath49-f@xmath36 and l@xmath49-l@xmath36 for all objects in the sample .
not surprisingly , because of the relative proximity of the two bands , most of the points occupy a narrow strip parallel to the bisectrix of the planes
. however , we have to be cautious in taking this plot at face value as it is sensitive to the presence of nuclear absorption and also to nuclear variability , since infrared and optical data were not obtained simultaneously .
the fr i galaxies ( empty yellow circles ) cover the whole range in optical and nir fluxes ( @xmath155 dex ) .
the fr ii ligs ( filled blue circles ) are clustered in small optical and nir flux range ( @xmath152 dex ) in the lower - left slide of the plane .
the higs ( filled green squares ) are also clustered , but with a slight tail towards higher flux , covering a wider range in both bands ( @xmath153 dex ) with respect to fr ii ligs .
the blos ( filled red triangles ) show the highest optical and nir fluxes of the sample . in the l@xmath50-l@xmath36 plane ( fig .
[ oir ] , middle panel ) we find that fris exhibit the lowest luminosities .
fr ii ligs and higs are present at increasing optical and nir luminosities . finally , the blos are by far the most powerful objects . to highlight the presence of outliers in the l@xmath50-l@xmath36 plane , in fig .
[ oir ] ( right panel ) we show the ratio between nir and optical nuclear luminosities plotted vs. the nir nuclear luminosity .
most objects lie within @xmath150.5 dex of the dashed line representing @xmath51l@xmath36/@xmath51l@xmath52 .
note that most of the nuclei that are found above this line ( i.e. showing an nir excess with respect to the optical emission ) belong to the hig class .
we discuss the origin of this behavior of the higs in more detail in sect .
[ discushig ] . in order to investigate the origin of the nir nuclei , we now explore possible correlations between the properties of the infrared nuclear emission and those of the radio cores .
[ rir ] , left panel , shows the ir nuclear flux f@xmath36 plotted vs. the radio core flux f@xmath34 .
the fr is ( empty yellow circles ) cover the whole range of flux spanned by the sample , both in the radio and in the nir .
higs ( filled green squares ) and fr ii ligs ( filled blue circles ) are clustered in the central region of the plane .
the blos ( filled red triangles ) are located in the top - right quadrant , at higher radio and nir fluxes .
a similar distribution of the points is evident in the plane defined by the radio and the nir core luminosities ( fig .
[ rir ] , right panel ) .
however in the l@xmath34-l@xmath36 plane the distributions are stretched because of the common dependence of the two quantities on distance .
the fr i galaxies have radio core luminosities that are on average lower than those of fr iis .
this is expected because the fanaroff - riley morphological radio classification in fr i and fr ii corresponds to different total luminosities at 178 mhz , and a positive trend links total and core radio luminosity @xcite .
a clear trend is visible for fr is both in the f@xmath34-f@xmath36 and l@xmath34-l@xmath36 planes ( fig .
[ rir ] ) .
fr ii ligs nuclei also show a trend in both planes , even if they cover a smaller range in radio core flux ( 2 dex for the nir detected nuclei ) with respect to fr is .
conversely , there is no evident link between radio and nir emission for higs and blos . in order to assess the presence of correlations
, we performed a statistical analysis using the astronomy survival analysis ( asurv ) package @xcite .
this package is available under iraf / stsdas and provides us a tool to deal with censored data ( @xcite , @xcite ) .
note that the method we use is based on the assumption that the censored data are uniformly distributed . in order to determine whether this approach is viable , we apply the censoring analysis only when the above assumption is satisfied .
we initially check the location of the non - detected data points with respect to the entire range of available data .
we used the
schmittbin task @xcite to calculate the associated linear regression coefficients for two set of variables .
operatively , we carried out this procedure twice , obtaining two linear regressions : firstly considering the former quantity as the independent variable and the latter as the dependent one , secondly switching the roles of the variables .
the best fit is represented by the bisector of these two regression lines .
this followed the suggestion of @xcite that consider such a method preferable for problems that require a symmetrical treatment of the two variables . in order to estimate the quality of the linear regression , for small data sets ( n @xmath1 30 ) we also derived the generalized kendall s
@xmath53 @xcite between the two variables , using the `` bhkmethod '' task . otherwise , we used spearman s rank order correlation coefficient , using the `` spearman '' task @xcite only for large data sets ( n @xmath7 30 , for the cases of higs and the whole sample ) .
the statistical parameters of all the linear regressions are reported in table [ statist1 ] .
the trend shown by fr is ( fig .
[ rir ] , empty yellow circle points ) in the f@xmath34-f@xmath36 plane corresponds to a linear correlation coefficient of r = 0.79 .
the probability to obtain this level of correlation from two independent variables is p = 2.2@xmath910@xmath54 .
the statistical parameters are even better in the radio - infrared luminosity plane , where the linear correlation coefficient is r = 0.85 , with an associated probability p = 8.9@xmath910@xmath55 .
the dispersions of these correlations are @xmath150.51 and @xmath150.48 dex , for fluxes and luminosities , respectively .
the generalized kendall s @xmath53 coefficient , which allows us to include censored data , is higher for the luminosities ( @xmath53 = 1.18 ) than that for the fluxes ( @xmath53 = 1.00 ) .
the associated probabilities of no correlation , are 0.0002 and @xmath10.0001 , respectively .
note that the slopes of both linear regressions are consistent with unity .
we now turn our attention to the possible presence of correlations for fr ii galaxies .
fr ii ligs show a linear correlation in both planes of fig .
[ rir ] . considering only the detected nuclei , the linear correlation coefficient is r = 0.89 in the f@xmath34-f@xmath36 plane , with a probability of a no correlation of p = 1.0@xmath910@xmath56 .
the censoring statistics gives a generalized kendall s coefficient @xmath53 = 0.77 with a probability , p = 0.0012 , that a fortuitous correlation appears at a level measured by our test statistic .
in l@xmath34-l@xmath36 plane for fr ii ligs a linear relation ( dotted line in fig .
[ rir ] , right panel ) is also statistically supported : the linear correlation coefficient is r = 0.93 ( p = 1.1@xmath910@xmath54 ) .
the radio - infrared core luminosity relation for fr ii ligs has a slope of [email protected] .
therefore , an important result of our analysis is that the correlations found for fr is and fr ii ligs nuclei are consistent with each other within the errors .
this result is shown in fig .
[ pippo ] where for more clarity we have marked these two classes with different symbols . for higs and blos the absence of a trend is confirmed by the results of the statistical analysis and in particular by the high probabilities that their distribution in the l@xmath34-l@xmath36 plane is random ( see the statistical parameters in tab .
[ statist1 ] ) .
hz@xmath4 ) for fr is ( empty yellow points ) and fr ii ligs ( blue points ) .
we show the linear correlations : the solid line is obtained for fr is and fr ii ligs , the dotted line for fr is only , and the dashed line for fr ii ligs only . ] in order to derive the multi - wavelength properties of the nuclei , it is useful to analyze the broad - band radio - infrared spectral index @xmath57 . in fig .
[ alfaisto ] we show the empirical distribution in @xmath57 for the four different classes ( fr i , fr ii lig , hig and blo ) .
note that the average value of the radio - ir spectral index distributions shifts towards lower values from fr ii ligs to the blos , with fris and higs having intermediate values .
this indicates that with respect to the ir flux , the fr ii ligs are the most radio - powerful objects .
another intriguing aspect worth noting is that even though fr ii ligs and fr is behave similarly in the radio - ir plane , as they define the same correlations , their distribution of @xmath57 are slightly different .
in fact , fr ii ligs have on average a slightly higher @xmath57 than the fr i galaxies .
on the other hand , higs and blos show an near infrared excess with respect to the radio emission and therefore their @xmath57 spectral indices are on average lower than for the other two classes . to assess the presence of any significant differences between the spectral index distribution of the different classes
, we can make use of the twosampt task , available in the asurv package .
the task computes several nonparametric two - sample tests , giving a variety of ways to test whether two censored samples are drawn from the same parent population .
we utilize the peto - prentice statistic @xcite , which is a generalization of standard test for uncensored data , such as the wilcoxon @xcite and kolmogorov - smirnov @xcite test .
this statistical test quantifies a distance between the empirical distribution functions of two samples .
the null distribution of this statistic is calculated under the null hypothesis that the samples are drawn from the same distribution .
we statistically measure the different distributions of spectral indices between fr is and fr ii ligs and between higs and blos .
we can assert that fr is and fr ii ligs are not drawn from the same parent population at the 97% confidence level .
the same holds for the hig and blo classes at the 99% confidence level .
@xcite measured the optical nuclear fluxes for most fr i radio galaxies in the 3cr catalog . the optical detection rate is similar to that found for the fr is of our sample . for the sub - sample of objects for which the total radio luminosity at 178 mhz is lower than the fr i / fr ii break ( corresponding to l@xmath58 @xmath15 2@xmath910@xmath59 erg s@xmath4 hz@xmath4 ) , they found a tight linear correlation between the optical and radio nuclear fluxes with a dispersion of @xmath150.4 dex . in order to compare the radio - infrared and radio - optical correlations we selected the
15 fr is in common with the sample of @xcite ( namely 3c 29 , 3c 31 , 3c 66b , 3c 83.1 , 3c 84 , 3c 264 , 3c 270 , 3c 272.1 , 3c 274 , 3c 296 , 3c 317 , 3c 338 , 3c 442 , 3c 449 , and 3c 465 ) .
we estimate the linear regression between radio and near infrared fluxes for this sub - sample : the correlation coefficient is r = 0.82 ( p = 2.2@xmath910@xmath56 ) and the dispersion is @xmath150.49 dex .
we tested that the dispersions of the two relations , f@xmath34-f@xmath36 and f@xmath34-f@xmath35 , are not statistically different at a significance level of 95% .
cccccccccccccc relation & class & n@xmath60 & n@xmath61 & r@xmath60 & p@xmath62 & @xmath63 or @xmath64 & p@xmath65 or p@xmath66 & @xmath67 or @xmath53 & p@xmath68 or p@xmath69 & rms@xmath70 & rms@xmath71 & m & q + f@xmath34-f@xmath36 & fr i & 21 & 5 & 0.788 & 2.2@xmath910@xmath54 & 1.181 & 0.0002 & 0.997 & 0.0002 & 0.477 & 0.543 & [email protected] & -3@xmath405 + f@xmath34-f@xmath36 & fr i @xmath72 & 15 & 0 & 0.834 & 0.00011 & 1.238 & 0.0013 & / & / & 0.442 & 0.542 & [email protected] & -8@xmath406 + f@xmath34-f@xmath36 & fr i @xmath73 & 15 & 1 & 0.855 & 4.9@xmath910@xmath54 & 1.257 & 0.0011 & 1.233 & 0.0006 & 0.394 & 0.407 & [email protected] & -3@xmath406 + f@xmath74-f@xmath36 & fr i & 15 & 1 & 0.866 & 3.0@xmath910@xmath54 & 1.239 & 0.0012 & 1.217 & 0.0007 & 0.380 & 0.291 & [email protected] & 5@xmath405 + f@xmath34-f@xmath36 & lig & 11 & 12 & 0.890 & 0.00010 & 1.490 & 0.0014 & 0.775 & 0.0012 & 0.168 & 0.298 & [email protected] & -14@xmath4015 + f@xmath34-f@xmath36 & fr i+lig & 32 & 17 & 0.781 & 4.1@xmath910@xmath75 & 0.743 & @xmath10.0001 & 0.645 & @xmath10.0001 & 0.445 & 0.511 & [email protected] & -7@xmath4014 + f@xmath34-f@xmath36 & hig & 15 & 12 & 0.150 & 0.59 & -0.095 & 0.804 & 0.296 & 0.218 & 0.542 & 0.544 & [email protected] & 3@xmath4024 + f@xmath34-f@xmath36 & blo & 15 & 0 & 0.348 & 0.20 & 0.514 & 0.181 & / & / & 0.680 & 0.727 & [email protected] & -3@xmath409 + l@xmath34-l@xmath36 & fr i & 21 & 5 & 0.854 & 8.9@xmath910@xmath55 & 1.324 & @xmath10.0001 & 1.077 & @xmath10.0001 & 0.420 & 0.532 & [email protected] & -3@xmath406 + l@xmath34-l@xmath36 & fr i@xmath73 & 15 & 1 & 0.898 & 5.4@xmath910@xmath76 & 1.276 & 0.0009 & 1.167 & 0.0008 & 0.392 & 0.417 & [email protected] & -3@xmath405 + l@xmath74-l@xmath36 & fr
i & 15 & 1 & 0.903 & 4.1@xmath910@xmath76 & 1.333 & 0.0005 & 1.183 & 0.0007 & 0.384 & 0.347 & [email protected] & -7@xmath405 + l@xmath34-l@xmath36 & lig & 11 & 12 & 0.931 & 1.1@xmath910@xmath54 & 1.491 & 0.0014 & 0.696 & 0.0011 & 0.211 & 0.306 & [email protected] & 0@xmath4012 + l@xmath34-l@xmath36 & fr i+lig & 32 & 17 & 0.890 & 1.8@xmath910@xmath75 & 0.900 & @xmath10.0001 & 0.688 & @xmath10.0001 & 0.363 & 0.487 & [email protected] & 1@xmath4018 + l@xmath34-l@xmath36 & hig & 15 & 12 & 0.503 & 0.056 & 0.514 & 0.181 & 0.461 & 0.054 & 0.637 & 0.606 & [email protected] & -13@xmath4027 + l@xmath34-l@xmath36 & blo & 15 & 0 & 0.2398 & 0.39 & 0.152 & 0.692 & / & / & 0.488 & 0.802 & [email protected] & 4@xmath4014 + [ statist1 ]
in the following we consider the properties of each group of 3cr radio galaxies separately .
the presence of a linear correlation between fr i radio and nir cores , both in flux and luminosity , argues for a real physical connection between the two variables .
this is indeed not the result of a common dependence on redshift of the two quantities , as shown by the fact that the statistical parameters of both correlations are similar .
following @xcite , that interpreted the optical - radio core correlation as the indication that synchrotron emission dominates the nuclear emission at both wavelengths , we argue that the correlations between the radio and the nir nuclei described here can be explained by the same physical mechanism .
more precisely , the presence of such a tight correlation between the nir and radio cores indicates that non thermal ( synchrotron ) radiation dominates their emission at all considered wavelengths , including the x - ray as shown by @xcite .
@xcite found that the optical hst nuclei of 9 3cr / fr i , which also belongs to our sample , are highly polarized .
this strengthens the synchrotron origin of their nuclear emission .
the correlations among nir , vla and vlbi cores and the fact that the fluxes of the vla and vlbi cores ( see appendix ) are essentially equal suggest that the nir cores are typically produced on scales consistent with or even smaller than the size of the vlbi cores .
in fact , according to various jet models ( e.g. , * ? ? ?
* ) , it is likely that the radio and the ir / optical radiation is produced in the innermost region of the jet , close to the black hole ( @xmath13 10@xmath77 cm ) , while the radio is emitted on larger scales . the dominance of synchrotron radiation even in the
nir indirectly confirms that accretion in most fr i nuclei occurs at low rates and/or with low radiative efficiency ( e.g. , @xcite , @xcite , @xcite , @xcite )
. however , there are exceptions to this result : the clearest example is possibly 3c 120 ( a radio galaxy which is formally not part of 3cr sample ) which displays a radio morphology typical of fr is ( e.g. @xcite ) while its nuclear emission shows the clear signature of broad lines and radiatively efficient accretion @xcite . on the other hand , exceptions are also present among bl lacs , the putative `` beamed '' counterpart of lig population .
a number of bl lacs also show ( faint ) narrow and/or broad emission lines ( e.g. @xcite ) and others have fr ii - like extended radio emission ( e.g. @xcite ) .
@xcite pointed out that the dispersion of the radio - optical correlation ( @xmath15 0.4 dex ) is consistent with dust absorption , provided that the @xmath78 is randomly distributed among the sources and does not exceed @xmath15 2 mag .
since the extinction in h - band is 4.5 times lower than in the optical r - band , the observations presented in this work are far less sensitive to the presence of dust along the line of sight than those in @xcite .
therefore , one would expect the dispersion of the infrared - radio correlation to be smaller than that of the radio - optical , if that is due to absorption .
our results instead show that the rms of the two correlations are similar .
thus , the most likely explanation for the scatter is that variability of the nuclei plays a fundamental role , while the effect of absorption is less significant .
indeed , this is supported by the variability observed for the nuclear flux of 3c 317 @xcite and of 3c 274 ( @xcite , @xcite ) .
conversely , the presence of dust appears to play a role in dimming the nuclear flux on the uv @xcite . note that data in the literature indicate that the radio core variability is typically at 10% over a time scale of several years @xcite
therefore the main source of the scatter in the radio - nir correlation is mostly likely due to a variable nir nuclear component . furthermore , the region of the jet base that produces the optical / nir emission in fr is is likely to be mildly relativistic ( e.g. , @xcite ) .
therefore , we consider the effects of doppler boosting on our results .
firstly , the range in nir nuclear luminosity might be enhanced by relativistic beaming , which would boost ( or de - boost ) the emission of fr i nuclei , depending on the viewing angle .
objects seen under a small angle between the line - of - sight and the jet axis would appear boosted , while objects observed closer to the plane of the sky would be de - boosted .
secondly , nuclear variability might be enhanced by relativistic beaming when the flux variation is associated with 1 ) a change in the jet bulk lorentz factor , or 2 ) a change in the jet direction , which would result in a different angle between the jet direction and the line - of - sight . however , as noted by @xcite , the bulk lorentz factor of the component responsible for the nuclear optical / nir emission is most likely around @xmath79 and the source is possibly a `` slower '' shear layer in the jet .
therefore , the effects of relativistic beaming should not dramatically affect our results .
are the nir data telling us something new about the presence of optically and geometrically thick obscuring tori in fr is with respect to the optical observations ? the results derived considering the uv , optical , and nir bands are only apparently in contradiction .
in fact , the median absorption toward fr i nuclei derived from uv observations is @xmath80 .
this corresponds to only 0.2 mag in h band , negligible with respect to a dispersion of @xmath150.5 dex , and to @xmath81 mag at 2500 .
the detection rate of fr i nuclei in the nir ( 81% ) is similar to that found in the r band by @xcite .
all sources with an optical nucleus also show a nir counterpart , except 3c 310 . instead , in 3c 75n a faint dust lane prevents the study of its nuclear regions in the optical , while its nir core is clearly detected .
furthermore , the absence of fr i outliers in the l@xmath36/l@xmath50 ratio ( see right panel of fig . [ oir ] ) can be interpreted as a hint that the _ detected _ fr i nuclei are not affected by significant absorption ( @xmath82 mag ) .
the reason for the non - detection of some of the fr i nir nuclei might instead be due to a small contrast between the nucleus and the host galaxy stellar emission .
assuming that the radio and the near infrared emission of fr i nuclei are intrinsically tightly correlated , the intensity of their radio cores provides a robust prediction of their infrared nuclear flux . following the analysis on the b2 radio - galaxies by @xcite
, we plot the radio core fluxes of all fr is against the galaxies central surface brightness ( fig .
[ contrast ] ) .
four out of the five undetected nuclei are indeed among those with the lowest predicted contrast against the galaxy emission .
3c 310 is the only object that does not have an nir detected nucleus and apparently shows an rather larger contrast with galaxy emission than that of the other four non - detected nuclei .
however , since its optical nucleus is detected , its nir non - detection is likely not due to a obscuration . summarizing
, we argue that in most cases we have a direct view of fr i nuclei and that the few undetected nir central sources are below the detection threshold set by the contrast with the underlying host galaxy emission .
note that the detection of massive dusty structures in the central regions of some fr is ( e.g. @xcite , @xcite ) does not contrast with our result that the nir nuclei are mostly unobscured .
in fact dust is distributed in large kpc - scale geometrically thin disks which do not appear to ` hide ' the nuclei to our line - of - sight .
( erg s@xmath4 @xmath38 hz@xmath4 ) , versus central ( on a circular region of r=0@xmath815 ) surface brightness of the host galaxy , log f@xmath83 ( erg s@xmath4 @xmath38 @xmath4 arcsec@xmath84 ) for all fr i. objects with detected near infrared nuclei are marked with empty yellow circles , while those with non - detection by filled yellow circles .
the dotted line marks a constant ratio between the radio core and galaxy emission . ]
@xcite found that the nuclei of fr ii ligs lie on the fri radio - optical correlation .
the sample of fr ii ligs available in this work is substantially larger than the sample those authors considered , both because of the increasing number of fr ii lig identifications and of the larger sample size .
this allows us to establish the presence of a correlation between the radio and the nir emission within the sample of fr ii ligs itself .
an important result is that the correlations for fr is and fr ii ligs are statistically indistinguishable .
this strengthens the hypothesis that the ligs are fr ii galaxies with fr i - like nuclei .
such a nuclear similarity implies that non - thermal synchrotron radiation ( e.g. , from a relativistic jet ) dominates their nuclear emission in nir band and that accretion may occur with a radiatively inefficient process and/or at low accretion rates in fr ii ligs as well .
the large fraction of fr ii ligs in our sample ( @xmath1533% of the fr iis ) proves that these objects are not rare exceptions in the fr ii class .
this class of objects show nuclear properties similar to those of fr is , whereas their radio morphologies span the range from classical double to x - shaped , and from compact source to fat double ( fd ) .
the fd morphology @xcite consists of a fr ii radio structure with elongated and diffuse lobes and sometimes with prominent jets , and may be considered as an intermediate class between fr is and fr iis .
at least half of the fr ii lig sample show a fd radio morphology .
the large variety of extended radio morphologies associated with a central lig - like agn makes it difficult to understand how the properties of the core and extended radio structures are related ( e.g. @xcite ) . the existence of fr ii ligs as a class shows that radiatively inefficient accretion disks are able to produce jets that give rise to the typical large scale fr ii morphology .
another possibility is that fr ii ligs constitute an evolutionary ( or transient ) state of the life of the fr ii class .
however , the environment of fr ii ligs is different from that of other classes of fr iis , as fr ii ligs are predominantly found in dense groups or clusters @xcite .
that seems to rule out the evolutionary scenario
. the lower detection rate of fr ii lig nuclei with respect to that of fr is is most likely due to the reduced contrast between their faint ir nuclei and the surrounding stellar emission of the host galaxy , because of the higher redshift of fr ii ligs as compared to that of fr is .
in fact , the fraction of detected fr ii lig nuclei decreases as redshift increases , from 55 % for [email protected] to 29 % for [email protected] . however , at this stage we can not rule out that the presence of significant obscuration along the line of sight to the fr ii ligs with non - detected nuclei might also contribute to the reduce the fraction of detected fr ii lig nuclei .
the best way to probe obscuration in fr ii ligs is through x - ray spectral analysis .
fr ii ligs indeed seem to lack significant absorbing column densities , @xmath85 @xmath38 @xcite .
the fact that the central engines of fr ii ligs and fr is are similar ( i.e. , they are both unobscured and lack a significant broad line region ) has important bearings for the unification schemes .
as pointed out by @xcite , the broad distribution in core - dominance is consistent with fr ii ligs being a randomly oriented sample . since fr ii
ligs also appear to lack significant broad emission lines , when observed along the jet axis , these objects would appear as bl lacs , in agreement with @xcite in light of their findings in the optical band .
such a unification scenario , which was previously explored by @xcite based on the optical spectral properties of fr ii ligs , is in apparent contrast with zeroth order unification scheme which associates all fr ii with quasars .
however , besides being in agreement with our results , it may also account for the presence of bl lacs with an fr ii morphology .
nevertheless , fr ii ligs do show some differences with respect to fr is .
in fact , as shown in fig .
[ alfaisto ] , fr ii ligs and fr is have slightly different radio - ir broad - band spectral indices . if fr ii ligs are indeed the counterparts of bl lacs with fr ii radio morphology , the spectral index discrepancy can be explained in the framework of the so - called `` blazar sequence '' @xcite .
such a scenario predicts a smooth transition in the properties of blazars spectral energy distribution ( sed ) . in blazars ,
the synchrotron emission peak is observed to shift towards lower frequencies as the radio luminosity at 5 ghz increases .
the luminosity of the synchrotron emission peak also increases following the same trend .
this gives rise to a sequence having high energy peaked bl lacs at the lowest luminosities , low energy peaked bl lacs at intermediate luminosities , and flat spectrum radio quasars at the highest bolometric powers . in fig .
[ sed ] we schematically show what the effect of such a scenario on the broad band spectral indices would be . as the peak in the sed shifts towards lower frequencies , @xmath57 steepens .
therefore , the steeper spectral indices we found in fr ii ligs might be explained in terms of a slightly higher 5ghz core luminosity with respect to that of fr is .
fr ii ligs would thus correspond to intermediate power blazars such as the low - energy peaked bl lacs .
the possibility that , conversely , the change in @xmath57 is due to the redshift of the sed for the more distant objects can be excluded using the formula presented by @xcite .
the formula , originally derived for estimating the effect of relativistic beaming , can be used to to account for the effect of redshift on the broad - band spectral indices , by substituting r@xmath86 , the ratio of the beaming factors , with @xmath87 .
the variation in the radio - infrared spectral due to the frequency shift for an object at z@xmath150 moved to z=0.3 is only @xmath88 , a correction that has a negligible effect on the results
. steepens . ] according to the standard unification scheme , the blos are objects where the nuclear region is unobscured along our line of sight ( e.g. * ? ? ?
* ; * ? ? ?
this idea is supported by the fact that all the nir nuclei are detected for blos .
the blos show a strong nir excess , of up to two orders of magnitude in luminosity , with respect to the correlation defined by the fr i ( and fr ii lig ) radio - infrared nuclei ( fig .
[ rir ] , right panel ) .
a similar excess was already found in the optical band by @xcite interpreted that as thermal emission from the accretion disk .
let us now explore the origin of their nir nuclear light .
the infrared - optical spectral index for blos covers the range @xmath89 @xmath15 0 - 2 ( with the exception of 3c 111 for which @xmath89 = 3.3 ) .
such a large range might result from a wide distribution of spectral indices ( see fig .
11 in @xcite ) .
in addition , variability may smear the intrinsic distribution .
therefore we consider the median value @xmath90 .
this value indicates that the spectral energy distribution ( sed ) of blos is essentially flat in a @xmath51l@xmath91 representation . to better investigate the origin of nir emission in the blos of our sample we compare their properties with those of the radio - loud qsos ( rlqso ) from @xcite .
as demonstrated by those authors , the averaged sed of qsos shows a minimum at @xmath151@xmath0 m . for wavelengths shorter than 1@xmath0
m the light is likely to be dominated by the accretion disk ( i.e. , the big blue bump ) , while for longer wavelengths a broad peak is observed , which is interpreted as emission from hot dust .
the flat spectral index of blos is an indication that , on average , the optical and nir hst observations are located on opposite sides of the sed minimum . therefore , by analogy with rlqsos , we then conclude that in most 3cr / blos the h band light is mainly produced by the high temperature tail of the hot dust emission , located at the inner face of the obscuring material .
we estimate a spectral index of @xmath89 = 0.35 from the averaged sed of rlqsos after adopting the mean redshift of blos ( z = 0.14 ) .
the difference of the spectral indices between rlqsos and blos can be accounted by an nir / optical flux ratio of the blos larger by a factor @xmath152 than that of rlqsos .
this is an indication of a relatively larger contribution from the dust in infrared emission of blos than that observed in rlqsos .
furthermore , reminding the reader that the rlqso have optical and nir luminosities a factor of 100 - 1000 higher than those of the 3cr / blos , there is a dependence of the optical - infrared spectral index on the luminosity of the object . at odds with the results found for fr ii ligs and
fr is , the nir emission of hig nuclei does not correlate with the radio cores .
furthermore , most of the detected nuclei exceeds the emission of fr is and fr ii ligs at a given radio - core luminosity , suggesting an additional component dominating over the jet emission , similar to what we found for the blos
. however , hig nuclei are substantially fainter than those of the blos , on average by a factor of 50 .
this is expected in a standard unification scheme , since the higs are considered to be the obscured counter parts of blos .
this result indicates that substantial obscuration is still present in the nir , despite the reduced effects of dust in the h band . in order to investigate the origin of nir emission in higs ,
we follow the method described by @xcite who studied a complete sub - sample of the 3cr radio catalog , which includes all higs with z @xmath1 0.3 , observed in k@xmath92 band from the ground .
they first estimated the ionizing luminosity , l@xmath93 = l@xmath94 c@xmath4 = 0.0675 @xmath9 l@xmath95}$ ] ( e.g. total luminosity in narrow lines , l@xmath94 , estimated from [ o ii ] and [ o iii ] , @xcite ) for higs and blos assuming the same covering factor for the two classes , c = 0.01 .
then , for each source , @xcite plotted l@xmath93/l@xmath36 against l@xmath35/l@xmath36 , which are both sensitive to nuclear obscuration .
they found that higs do not follow the ` reddening line ' ( see caption of fig .
[ marchesini ] ) in such a plane , as one might expect if they were simply `` partially obscured '' blo nuclei .
this lead them to conclude that a substantial role is played by scattered nuclear light .
with respect to @xcite our analysis can rely on _ direct _ measurements of the nir nuclei and these are available for a larger number of hig , due to better sensitivity of the hst images to faint ir nuclei .
furthermore , we also have nir data for the nuclei of blo , thus providing a more robust estimate of the location of unreddened sources . in agreement with the results
@xcite we found that only a few higs are located along the reddening line in fig .
[ marchesini ] but this is clearly not the case for the whole population .
we then include the effects of scattered nuclear light . in this scenario
a dusty torus attenuates the direct nuclear light by an extinction a@xmath96 and a fraction @xmath97 of the total nuclear light l@xmath36 is scattered into our line of sight ( see eq . 3 and in @xcite ) .
the effects of varying a@xmath96 is as follows : an increase of the nuclear obscuration simply corresponds to an increase of l@xmath93/@xmath51l@xmath36 .
the behavior of l@xmath35/l@xmath36 is more complex : initially , an increase of the absorption moves the point ( starting from the average location of blo ) to the right , along the reddening line . at larger values of a@xmath96 the direct ( transmitted )
optical light becomes fainter than the scattered component , while the infrared nucleus is still substantially unaltered . at even higher values of a@xmath96 ,
the infrared transmitted component starts to decrease ; since the scattered optical component remains constant this causes an increase of l@xmath50/l@xmath36 , producing a turnover in the source path .
the track ends at the same ratio of l@xmath35/l@xmath36 found for blos , but shifted downward by a factor linked to a@xmath96 .
the location of the turnover instead depends on the fraction of scattered light @xmath97 , moving to smaller l@xmath35/l@xmath36 and to larger l@xmath93/@xmath51l@xmath36 with decreasing @xmath97 . and @xmath51l@xmath36 versus the ratio between l@xmath35 and l@xmath36 .
the blos are the red triangles and the higs the green squares .
the solid oblique line is the reddening line that purely obscured objects should follow .
the origin of this line is determined by the ratios measured for blos , i.e. log(l@xmath93/@xmath51l@xmath36 ) = [email protected] and log(l@xmath50/l@xmath36 ) = [email protected] .
the dashed region reproduces the uncertainty of its origin .
the dotted vertical line corresponds to the average value of log(l@xmath50/l@xmath36 ) measured for blos .
the grid of dotted curves represent the effects of varying the nuclear a@xmath98 and the fraction of scattered light @xmath97 ( see the eq.3 in @xcite ) . ]
because both the ir and the optical emission is most likely variable , therefore non - simultaneous measurements may lead to large errors in the estimate of the intrinsic l@xmath50/l@xmath36 .
nevertheless , this does not compromise the analysis : while a handful higs can be interpreted with being purely obscured blos , most of them ( approximately 3/4 of the sample ) require obscuration as well as scattered light , in the range of 10@xmath1a@xmath96@xmath130 and 0.02%@xmath9910% respectively .
this agrees with the general results of @xcite ( they derived a fraction of @xmath1570% ) and also considering higs on a object - by - object basis .
our analysis of the brightness profiles of the 3cr sources always assumed that the nuclear nir components are unresolved . while for the other classes of 3cr galaxies the nuclear sources
are directly associated with the agn ( and thus effectively point - like at our resolution ) , for higs our results suggest that they are in general dominated by scattered light , a spatially resolved component .
this is confirmed by several observations of imaging polarimetry and spectropolarimetry of radio - galaxies ( e.g. @xcite ) .
we then re - modeled the observed nir brightness profiles taking into account the possibility that the central hig sources are actually resolved and can be described by a gaussian light distribution .
we considered in detail the three objects ( 3c 234 in polarized flux ( @xcite ) . ] , 3c 300 , and 3c 349 ) that should have the larger scattered light fraction , being located closer to the grid origin in fig .
[ marchesini ] . for sizes larger than @xmath15
@xmath100 significant residuals start to appear in the model fitting and we consider this value as an upper limit to the size of the scattering region .
this corresponds to a range of 300 - 400 pc for these three objects , consistent with the size of the seyfert polar scattering zone .
this does not mean that there is no scattered light outside this region ( as in e.g. 3c 321 , @xcite ) , but it must be taken as an indication that the bulk of scattering occurs within this scale .
the presence of low surface brightness scattered components ( in contrast to the host galaxy ) can only be probed with imaging polarimetry .
we conclude that hig , unlike fr ii ligs , are consistent with the interpretation that they host a hidden quasar , but they are not simply partially obscured blo .
scattered light must provide a substantial contribution to their nuclei in both the optical and nir bands .
we have analyzed 1.6 @xmath0 m near infrared images of 100 3cr radio galaxies from hst nicmos in the f160w band with z @xmath1 0.3 , @xmath1590% of the whole 3cr sample .
on the basis of the radio morphology we divided the sample between fr i and fr ii and from the point of view of their optical spectra we also classified them as low ionization galaxies ( lig ) , high ionization galaxies ( hig ) , and broad line objects ( blo ) .
we measured the nuclear nir luminosity of all the objects , by subtracting the 1-d radial brightness profile ( modeled with a srsic or core - srsic law representing the galaxy emission ) from the total profile derived from the isophote fitting of the images .
the nir nuclear emission is found to be strongly correlated with radio cores for fr is .
the linear correlation extends over @xmath15 4 dex in luminosity with a rms of 0.4 dex and it has slope consistent with unity , as already found for the relations with optical and x - ray nuclei .
this correlation provides further support to the identification of their nuclear emission as the synchrotron radiation produced in the inner part of the relativistic jet . in turn
this implies a low contribution from thermal emission , an indication for the presence of radiatively inefficient disks .
the similar dispersions for the correlations in nir and in optical with the radio cores ascribes to the nuclear variability the dominant role of this scatter , rather than to obscuration .
the large number of fr ii ligs present in this sample made possible for the first time a detailed analysis their nuclear properties in radio - infrared plane .
we found a strong linear correlation between radio and nir emission .
the multiwavelength properties of fr ii lig nuclei are statistically indistinguishable from those of fr is , indicating a common structure of the central engine , despite the difference in radio morphology and radio power .
the similar nuclear properties suggest that the nuclei of fr ii ligs belong to the same class of those of fr is , i.e. , the nucleus is dominated by non - thermal emission from a relativistic jet .
all blos have detected nir nuclei and show an nir excess with respect to the fr i - fr ii lig radio - infrared correlation .
this excess does not correlate with the radio cores , suggesting the presence of an additional infrared emitting component . in analogy with the properties of radio - quiet qso , their optical nuclei
most likely originate from the thermal emission of the accretion disk .
since their optical - infrared spectral indices are essentially flat , this suggests that thermal infrared emission from hot dust dominates in the nir .
the origin of the nir nuclei of higs is more complex .
although a few hig nuclei lie on the fr i - fr ii lig radio - infrared correlation , most of them show a large nir excess which does not correlate with the radio cores .
this means that synchrotron emission from the jet base is not sufficient to explain their nuclear luminosities .
nonetheless , the hig nir nuclei are less luminous than those of blos , by a factor of 50 on average , suggesting partial obscuration , as expected by the unified model . yet
the infrared - optical luminosity ratios for higs are not consistent with a model of higs as the purely - obscured counterpart of the blos .
we found that this inconsistency can be solved by including a significant contribution from nuclear light reflected in a compact circumnuclear scattering region .
the sources which show the largest scattered light contribution , have a scatter region of 300 - 400 pc ( @xmath101 ) .
the analysis of the properties of the nir nuclear sources of 3cr radio - galaxies presented here broadly confirms the indications derived from the study of the optical nuclei . in particular , the presence of two classes of radio - loud agn , lig and hig , with different nuclear properties is strengthened .
fr i radio - galaxies as well as the sub - population of sources with fr ii morphology characterized by a low ionization optical spectrum belong to the same class from the nuclear point of view . in these sources
the mechanism of emission related to the presence of non - thermal plasma from their radio - jets dominates over the contribution from thermal radiation related to the accretion process .
moving from optical to infrared observations has no substantial effect despite the reduced absorption at longer observing wavelengths .
however the nir band is not the optimal band to investigate the issue of obscuring tori .
the data preseneted here leave the results based on optical observations unchanged , bearing in mind that for most of fr is we have a clear view to their optical / infrared nucleus . while mir spitzer spectra @xcite obtained with a large slit size ( 10.7 ) do show an infrared excess with respect to the jet emission in three fr i sources ( and in one fr ii lig )
, our results on 3cr sample do not directly require the presence of dusty torus on the line of sight to the fr i nucleus .
furthermore , although @xcite argue that the dust emission is powered the nucleus , the comparison between mir spectra and hst data is extremely difficult because of the different spatial resolution of the two instruments . on the other hand , at this stage we can not rule out the presence of a dusty torus in these objects because of the higher distance of the fr ii ligs than that of fr is . despite the lack of such information , the nuclear properties of fr ii lig nuclei are found to be statistically indistinguishable from those of fr is .
how all the sources of the fr i / fr ii lig group can be associated with such a large range of radio power , while still maintaining their disks in radiatively inefficient state , and what causes the transition to the fr ii morphology in ligs remain open questions .
the nir data provide new insights also for the second class of objects , formed by higs and blos . the blo nir nuclei are likely to be dominated by the emission from hot dust located in the immediate vicinity of the active nucleus .
in hig we found many examples of nuclei in which the nir emission is substantially larger than in the optical . due to the proximity of the two bands , the only mechanism that can explain this effect is the presence of nuclear structures covered by absorbing material , optically thin at least in the infrared .
scattered light must provide a substantial contribution to their nuclei in both the optical and nir bands .
the lack of broad emission line in higs is universally associated with a substantially decreased luminosity of their nuclei in the infrared band and to a different angle of sight to the nucleus .
this implies that obscuration still plays a substantial role .
the detailed comparison of radio , optical , and infrared data for the 3cr sample also shows that the differences in the relative contribution in these bands , on a object - by - object basis , is dominated by variability and not by absorption .
thus , only simultaneous nuclear data would allow us to explore intrinsic differences among the sources of the sample .
the lack of such data effectively hampers the use of the nuclear spectral energy distribution to enlighten how they vary with e.g. , source power and viewing angle .
r.d.b . acknowledges the financial support ( hst - go-11219.07-a grant ) from space telescope science institute , baltimore .
we are grateful to the referee , r. antonucci for the extremely careful and detailed report which stimulated us and significantly contribuited to greatly improve the paper .
we also thank g. k. miley for useful comments .
it is interesting to investigate the radio and nir correlation for fr is by considering radio data of higher angular resolution . to this aim , we take from the literature the fluxes at 5 ghz obtained with vlbi observations .
the fr is for which the vlbi radio data are available are listed in table 5 . in fig .
[ vlbiir ] ( left panel ) we show the correlations in the l@xmath102-l@xmath36 planes .
the linear correlation coefficient is r = 0.90 ( p = 4.1@xmath910@xmath76 ) .
these values as well as the slope and the intercept of the derived correlations are essentially identical to those obtained with the vla data .
the reason of these similarities can be explained by looking at fig .
[ vlbiir ] ( right panel ) , which shows the vla core flux of fr is plotted vs. their vlbi flux .
the figure clearly shows that the two measurements lead to very similar flux values .
the dispersion is extremely small , with the exception of only two outliers ( 3c 84 and ngc 6251 ) , for which the vla flux is higher than that of vlbi by a factor of @xmath103 and @xmath104 , respectively .
however , since these two objects have a relatively high core dominance , they are likely to be observed at a small angle with respect to jet axis and thus they are more likely to be affected by flux variability .
that might explain the larger discrepancy between the vla and vlbi fluxes for these two sources with respect to the other fr is .
summarizing , the correlation between the vlbi and the ir nuclear luminosity is statistically indistinguishable from that obtained using the vla cores simply because the core fluxes measured with vla and vlbi are only marginally different .
c|ccc name & f@xmath102 & l@xmath102 & ref + 3c 31 & -24.04 & 29.69 & g05 + 3c 66 & -23.74 & 30.19 & g05 + 3c 83.1 & -24.57 & 29.50 & x99 + 3c 84 & -22.44 & 31.32 & k05 + 3c 264 & -23.70 & 30.20 & g05 + 3c 270 & -23.74 & 29.27 & j97 + 3c 272.1 & -23.74 & 28.67 & g05 + 3c 274 & -22.40 & 30.02 & g05 + 3c 296 & -24.19 & 29.83 & g05 + 3c 310 & -24.06 & 30.66 & g05 + 3c 317 & -23.53 & 30.80 & v00 + 3c 338 & -23.98 & 30.25 & g05 + 3c 346 & -23.66 & 31.95 & g05 + 3c 449 & -24.43 & 29.35 & g05 + 3c 465 & -23.61 & 30.61 & g05 + ngc 6251 & -23.44 & 30.61 & j86 + , d. l. , unwin , s. c. , readhead , a. c. s. , sargent , w. l. w. , seielstad , g. a. , simon , r. s. , walker , r. c. , benson , j. m. , perley , r. a. , bridle , a. h. , pauliny - toth , i. i. k. , romney , j. , witzel , a. , wilkinson , p. n. , baath , l. b. , booth , r. s. , fort , d. n. , galt , j. a. , mutel , r. l. , & linfield , r. p. 1986
, , 305 , 684 , y. y. , kellermann , k. i. , lister , m. l. , homan , d. c. , vermeulen , r. c. , cohen , m. h. , ros , e. , kadler , m. , lobanov , a. p. , zensus , j. a. , kardashev , n. s. , gurvits , l. i. , aller , m. f. , & aller , h. d. 2005 , , 130 , 2473 , r. a. , jenkins , c. r. , wall , j. v. , & unger , s. w. 1994 , in the first stromlo symposium : the physics of active galaxies .
asp conference series , vol .
54 , 1994 , g.v .
bicknell , m.a .
dopita , and p.j .
quinn , eds . ,
p.201 , 201+ , m. , isobe , t. , & feigelson , e. 1992 , in astronomical society of the pacific conference series , vol . 25 ,
astronomical data analysis software and systems i , ed .
d. m. worrall , c. biemesderfer , & j. barnes , 245+ , d. e. 1989 , astrophysics of gaseous nebulae and active galactic nuclei ( research supported by the university of california , john simon guggenheim memorial foundation , university of minnesota , et al .
mill valley , ca , university science books , 1989 , 422 p. ) | using hst nicmos 2 observations we have measured 1.6 @xmath0 m near infrared nuclear luminosities of 100 3cr radio galaxies with z @xmath1 0.3 , by modeling and subtracting the extended emission from the host galaxy .
we performed a multi - wavelength statistical analysis ( including optical and radio data ) of the properties of the nuclei following classification of the objects into fr i and fr ii , and lig ( low - ionization galaxies ) , hig ( high - ionization galaxies ) and blo ( broad - lined objects ) using the radio morphology and optical spectra , respectively .
the correlations among near infrared , optical , and radio nuclear luminosity support the idea that the near infrared nuclear emission of fr is has a non - thermal origin . despite the difference in radio morphology ,
the multi - wavelength properties of fr ii lig nuclei are statistically indistinguishable from those of fr is , an indication of a common structure of the central engine .
all blos show an unresolved near infrared nucleus and a large near infrared excess with respect to fr ii ligs and fr is of equal radio core luminosity .
this requires the presence of an additional ( and dominant ) component other than the non - thermal light .
considering the shape of their spectral energy distribution , we ascribe the origin of their near infrared light to hot circumnuclear dust .
a near infrared excess is also found in higs , but their nuclei are substantially fainter than those of blo .
this result indicates that substantial obscuration along the line - of - sight to the nuclei is still present at 1.6 @xmath0 m .
nonetheless , higs nuclei can not simply be explained in terms of dust obscuration : a significant contribution from light reflected in a circumnuclear scattering region is needed to account for their multiwavelength properties .
based on observations made with the nasa / esa hubble space telescope , obtained at the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .. ] |
inference on population genetic models such as coalescent trees is one representative example of cases when statistical analyses like bayesian inference can not operate because the likelihood function associated with the data is not completely known , i.e. can not be computed in a manageable time @xcite .
the fundamental reason for this impossibility is that the statistical model associated with coalescent data needs to integrate over trees of extreme complexity . in such settings , traditional approximation tools based on monte carlo simulation @xcite from the bayesian posterior distribution
are unavailable for all practical purposes . indeed , due to the complexity of the latent structures defining the likelihood ( such as the coalescent tree ) , simulation of those structures is too unstable to be trusted to bring a reliable approximation in a manageable time .
such complex models call for a practical if cruder approximation method , the abc methodology being a serious contender , where abc stands for _ approximate bayesian computation_. @xcite and @xcite introduced abc methods as a rejection technique bypassing the computation of the likelihood function via a simulation from the corresponding distribution . for recent reviews on abc , see @xcite and @xcite .
the wide and successful array of applications based on implementations of abc in genomics and ecology is covered by @xcite , while the number of publications relying on this technique runs in the hundreds .
@xcite describe the use of model choice based on abc for distinguishing between different mutation models .
the intuition behind the method is that the average abc acceptance rate associated with a given model is proportional to the marginal likelihood corresponding to this approximative model , when identical summary statistics , distance , and tolerance level are used for all models . in practice ,
an estimate of the ratio of marginal likelihoods is given by the ratio of observed acceptance rates .
using bayes formula , estimates of the posterior probabilities are straightforward to derive .
this approach has been widely used in the literature ( see , e.g. , @xcite , @xcite , and @xcite , @xcite ) . note that @xcite is particularly influencial for the conclusion it derives from the abc analysis : the focus of this _ science _ paper is the european invasion of the western corn rootworm , which is north america s most destructive corn pest . because this pest was initially introduced in central europe , it was believed that subsequent outbreaks in western europe originated from this area .
based on this abc model choice analysis of the genetic variability of the rootworm , the authors conclude that this belief is false : there have been at least three independent introductions from north america during the past two decades .
an improvement to the above estimate is due to @xcite , thanks to a regression regularisation . in this approach .
model indices are processed as categorical variables in a formal multinomial ( polychotomous ) regression .
for instance , when comparing two models , this leads to a standard logistic regression .
rejection - based approaches were lately introduced by @xcite , @xcite and @xcite , in a monte carlo perspective simulating model indices as well as model parameters .
those more recent extensions are already widely in use by the population genetics community , as exemplified by @xcite , or @xcite .
another illustration of the popularity of this approach is given by the availability of three three softwares implementing an abc model choice methodology : * abc - sysbio , developped by the theoretical systems biology group at imperial college london , which implements a smc - based abc for inference in system biology , including model - choice @xcite .
* diyabc , developped by the centre de biologie et de gestion des populations , at inra montpellier , which implements a regularised abc - mc algorithm on population history using molecular markers @xcite . * popabc , developped by the school of biological sciences at the university of reading , which implements a regular abc - mc algorithm for genealogical simulation @xcite .
@xcite process via abc the specific case of gibbs random fields with missing normalising constants .
they establish that exact bayesian model selection can be implemented in this setting , deriving this result from the property that the concatenation of the sufficient statistics across models is also sufficient for model comparison . in a subsequent paper , @xcite
advocate the role of abc approximations in general bayesian model choice .
the issue of sufficiency is covered in this paper , with a generic cross - model sufficiency completion leading the authors to validate the method in full generality , including in - sufficient cases . in this paper
, we argue that abc is a valid approximation method for conducting bayesian inference in complex stochastic models , barring the limitation that it can not discriminate between those complex stochastic models when based on summary statistics .
in essence , we highlight the fact that , since abc is conducting model choice based on in - sufficient statistics , the resulting inference is flawed in that the loss of information is severe to the point of inconsistency , namely that the abc model selection can not recover the proper model , even with an infinite amount of observation and computation . we demonstrate this inconsistency in the limiting ( and more favourable ) case of sufficient statistics .
the conclusion of the current paper are thus quite negative in that we consider that conducting testing or model comparison using abc does not carry any reliable weight of evidence and therefore should not be trusted .
more empirical measures such as those proposed in @xcite and @xcite seem to be the only possibility at the current time for conducting model comparison .
we are therefore at odds with the positive conclusion found in @xcite , as discussed below .
we stress here that , while @xcite repeatedly expressed reservations about the formal validity of the abc approach in statistical testing , those criticisms were addressed at the bayesian paradigm _ per se _ rather than at the approximation method .
quite clearly , templeton s criticisms got rebutted in @xcite and are not relevant for the current paper .
the plan of the paper is as follows : in section [ genesis ] , we recall the basics of abc as well as its justification ; section [ theramones ] exposes why a bayes factor based on an abc approximation is not converging to the true bayes factor as the computational effort increases ; section [ theclash ] explains the specificity of mrfs in this regard , while section [ siouxie ] illustrates the potential for divergence in examples .
sectoion [ rollingstones ] concludes the paper .
the setting in which abc operates is the approximation of the simulation from the posterior distribution @xmath0 when both distributions associated with @xmath1 and @xmath2 can be simulated .
the first abc algorithm was introduced by @xcite in a genetic setting , as follows : given a sample @xmath3 from a sample space @xmath4 , generate @xmath5 from the prior distribution @xmath6 generate @xmath7 from the likelihood @xmath8 set @xmath9 , the parameters of the abc algorithm are the statistic @xmath10 , the distance @xmath11 , and the tolerance level @xmath12 .
the approximation of the posterior distribution provided by the algorithm is that it samples from the marginal in @xmath13 of the joint distribution @xmath14 where @xmath15 denotes the indicator function of the set @xmath16 and where @xmath17 the basic justification of the abc approximation is that , when using a sufficient statistic @xmath10 and a small ( enough ) tolerance @xmath18 , we have @xmath19 the ( correct ) posterior distribution @xmath20 being the limit as @xmath18 goes to zero of @xmath21 . in practice ,
the statistic @xmath10 is not sufficient and the approximation then converges to @xmath22 .
this fact is appreciated by users in the field who see this loss of information as an unvoidable price to pay for the access to computable quantities . while acknowledging the gain brought by abc in handling bayesian inference in complex models , we will demonstrate below that the loss due to the abc approximation may be arbitrary in the specific setting of bayesian model choice and testing , whether or not @xmath10 is sufficient .
testing and model choice constitute a highly specific domain of bayesian analysis that involves conceptual and computational complexification since several models are simultaneously considered @xcite . given that both inferential problems are processed the same way in a bayesian perspective , we will only mention model choice in the remainder of the paper , but the reader must bear in mind that we cover testing as a particular case .
the standard tool on which a bayesian approach relies is the evidence @xcite , also called the marginal likelihood , @xmath23 that leads to the bayes factor for comparing the evidences brought by the data on models with likelihoods @xmath24 and @xmath25 , @xmath26 as detailed in the bayesian literature @xcite , this ratio provides an absolute criterion for model comparison that is naturally penalised for model complexity @xcite and whose first order approximation is the bayesian information criterion ( bic ) .
given that this issue is fundamental to our point , we recall that bayesian model choice proceeds by creating a probability structure across models ( or likelihoods ) .
namely , in addition to the parameters associated with each model , a bayesian inference introduces the model index @xmath27 as an extra parameter .
it is associated with its own prior distribution , @xmath28 ( @xmath29 ) , while the prior distribution on the parameter is conditional on the value @xmath30 of the model index , denoted by @xmath31 and defined on the parameter space @xmath32 .
the choice between those models is then driven by the posterior distribution of @xmath27 , @xmath33 where @xmath34 denotes the marginal likelihood of @xmath3 for model @xmath35 . while this distribution is well - defined and straightforward to interpret , it offers a challenging computational conundrum in bayesian analysis .
moreover , the solutions found in the literature @xcite do not handle the case when the likelihood is not available and abc represents the almost unique alternative .
as exposed in e.g. @xcite , @xcite , and @xcite , once @xmath27 is incorporated within the parameters , the abc approximation to the posterior follows from the same principles as regular abc .
the corresponding implementation is as follows , using for the tolerance region a statistic @xmath36 that is the concatenation of the summary statistics used for all models ( with an obvious elimination of duplicates ) .
generate @xmath30 from the prior @xmath28 generate @xmath37 from the prior @xmath38 generate @xmath7 from the model @xmath39 set @xmath40 and @xmath41 the abc estimate of the posterior probability @xmath42 is then the frequency of acceptances from model @xmath30 in the above simulation @xmath43 this also corresponds to the frequency of simulated pseudo - dataset from model @xmath30 that are closer to the data @xmath3 than the tolerance @xmath18 . in order to improve the estimation by smoothing , @xcite follow the rationale that motivated the use of a local linear regression in @xcite and
rely on a weighted polychotomous logistic regression to estimate @xmath42 .
this modelling is implemented in the diyabc software .
most perspectives on abc do not question the role of the abc distance nor of the statistic @xmath44 in model choice settings .
there is however a much stronger discrepancy between the genuine bayes factor / posterior probability and the approximations resulting from abc .
the abc approximation to a bayes factor , @xmath45 say , resulting from algorithm [ algo : abcmoc ] is @xmath46 an alternative representation is given by @xmath47 where the pairs @xmath48 are simulated from the ( joint ) prior and @xmath49 is the total number of simulations that are necessary for @xmath50 acceptances in algorithm [ algo : abcmoc ] . in order to study the limiting behaviour of this approximation , we first let @xmath49 go to infinity .
( for simplification purposes and without loss of generality , we choose a uniform prior on the model index . )
the limit of @xmath51 is then @xmath52 } { \mathbb{p}[\mathcal{m}=2,\rho\{{\boldsymbol{\eta}}({\mathbf{z}}),{\boldsymbol{\eta}}({\mathbf{y}})\ } \le \epsilon]}\\ & = & \dfrac{\int \mathbb{i}_{\rho\{{\boldsymbol{\eta}}({\mathbf{z}}),{\boldsymbol{\eta}}({\mathbf{y}})\ } \le \epsilon } \pi_1({\boldsymbol{\theta}}_1)f_1({\mathbf{z}}|{\boldsymbol{\theta}}_1)\,\text{d}{\mathbf{z}}\,\text{d}{\boldsymbol{\theta}}_1 } { \int \mathbb{i}_{\rho\{{\boldsymbol{\eta}}({\mathbf{z}}),{\boldsymbol{\eta}}({\mathbf{y}})\ } \le \epsilon } \pi_2({\boldsymbol{\theta}}_2)f_2({\mathbf{z}}|{\boldsymbol{\theta}}_2)\,\text{d}{\mathbf{z}}\,\text{d}{\boldsymbol{\theta}}_2}\\ & = & \dfrac{\int \mathbb{i}_{\rho\{{\boldsymbol{\eta}},{\boldsymbol{\eta}}({\mathbf{y}})\ } \le \epsilon } \pi_1({\boldsymbol{\theta}}_1)f_1^{{\boldsymbol{\eta}}}({\boldsymbol{\eta}}|{\boldsymbol{\theta}}_1)\,\text{d}{\boldsymbol{\eta}}\,\text{d}{\boldsymbol{\theta}}_1 } { \int \mathbb{i}_{\rho\{{\boldsymbol{\eta}},{\boldsymbol{\eta}}({\mathbf{y}})\ } \le \epsilon } \pi_2({\boldsymbol{\theta}}_2)f_2^{{\boldsymbol{\eta}}}({\boldsymbol{\eta}}|{\boldsymbol{\theta}}_2)\,\text{d}{\boldsymbol{\eta}}\,\text{d}{\boldsymbol{\theta}}_2}\,,\end{aligned}\ ] ] where @xmath53 and @xmath54 denote the distributions of @xmath55 when @xmath56 and @xmath57 , respectively . by lhospital formula ,
if we let @xmath18 go to zero , the above converges to @xmath58 which is precisely and exactly the bayes factor for testing model @xmath59 versus model @xmath60 based on the sole observation of @xmath61 .
this result is completely coherent with the current perspective on abc , namely that the inference derived from the ideal abc output when @xmath62 only uses the information contained in @xmath61 .
thus , in the limiting case , i.e. when the abc algorithm uses an infinite computing power , the abc odds ratio does not take into account the features of the data besides the value of @xmath61 , which is why the limiting bayes factor only depends on the distributions of @xmath44 under both models .
in contrast with point estimation where using a sufficient statistic has no impact on the inference in the limiting case , the loss of information resulting from considering solely @xmath44 seriously impacts the resulting inference on which model is best supported by the data .
indeed , as exhibited in a special case by @xcite , the information contained in @xmath61 is almost always smaller than the information contained in @xmath3 and this even in the case @xmath61 is a sufficient statistic for _ both models_. in other words , _
@xmath61 being sufficient for both @xmath63 and @xmath64 does not usually imply that @xmath61 is sufficient for @xmath65 . _ to see why this is the case , consider the most favourable case , namely when @xmath61 is a sufficient statistic for both models .
we then have by the factorisation theorem @xcite that @xmath66 , therefore that @xmath67 therefore , unless @xmath68 , the two bayes factors differ by this ratio , @xmath69 , which is only equal to one in a very small number of known cases .
this decomposition is a straightforward proof that a model - wise sufficient statistic is usually not sufficient across models , i.e. for model comparison .
an immediate corollary is that the abc - mc approximation does not converge to the exact bayes factor .
the discrepancy between the limiting abc inference and the genuine bayesian inference does not completely come as a surprise , because abc is indeed an approximation method .
users of abc algorithms are therefore prepared for some degree of imprecision in their final answer , a point stressed by @xcite or @xcite when they qualify abc as exact inference on a wrong model .
however , the magnitude of the difference between @xmath70 and @xmath71 expressed by is such that there is no direct connection between both answers . in a general
setting , if @xmath44 has the same dimension as one component of the @xmath72 components of @xmath3 , the ratio @xmath69 is equivalent to a density ratio for a sample of size @xmath73 , hence it can be arbitrarily small or arbitrarily large when @xmath72 grows . on the opposite
, the bayes factor @xmath74 is based on what is equivalent to a single observation , hence does not necessarily converge with @xmath72 , as shown by the poisson and normal examples below .
the conclusion derived from one bayes factor may therefore completely differ from the conclusion derived from other one and there is no possibility of a generic agreement between both , or even of a manageable correction factor . for this reason ,
we conclude that the abc approach can not be used for testing nor for model choice , with the exception of gibbs random fields as explained in the next section . in all cases when @xmath69 is different from one and impossible to approximate , no inference on the true bayes factor
can be made based on the abc - mc approximation without further information on the ratio @xmath69 , which is most often unavailable .
we note that @xcite also derived this relation between both bayes factors in their formula ( 18 ) but surprisingly concluded on advocating the use of abc in complex models , where there are no sufficient statistics .
we disagree with this perspective for reasons that will be made clear in the following sections .
@xcite showed that , for gibbs random fields and in particular for potts models , when the goal is to compare several neighbourhood structures , the computation of the posterior probabilities of the models / structures under competition can be operated by likelihood - free simulation techniques , in the sense that there exists a converging approximation to the true bayes factor .
the reason for this property is that , in the above ratio , @xmath68 in this special model . indeed , if we consider a gibbs random field given by the likelihood function @xmath75where @xmath3 is a vector of dimension @xmath72 taking values over the finite set @xmath76 ( possibly a lattice ) , @xmath77 is the potential function defining the random field , taking values in @xmath78 , @xmath79 is the associated parameter , and @xmath80 is the corresponding normalising constant , the potential function @xmath10 is a sufficient statistic for the model . for instance , in potts models , the sufficient statistic is the number of neighbours , @xmath81 associated with a neighbourhood structure denoted by @xmath82 ( meaning that @xmath83 and @xmath84 are neighbours ) .
the property that validates an abc resolution for the comparison of gibbs random fields is that , due to their specific structure , there exists a sufficient statistic vector that runs across models and which allows for an exact ( when @xmath62 ) simulation from the posterior probabilities of the models .
more specifically , consider @xmath85 gibbs random fields in competition , each one being associated with a potential function @xmath86 @xmath87 , i.e. with corresponding likelihood @xmath88 where @xmath89 and @xmath90 is the unknown normalising constant .
a bayesian analysis operates on the extended parameter space @xmath91 that includes both the model index @xmath27 and the corresponding parameter space @xmath32 .
the inferential target is thus the model posterior probability @xmath92 i.e. the marginal in @xmath27 of the posterior distribution on @xmath93 given @xmath3 .
each model has its own sufficient statistic @xmath94 .
then , for _ each _ model , the vector of statistics @xmath95 is clearly sufficient ; furthermore @xcite exposed the fact that @xmath44 is also sufficient for the joint parameter @xmath93 . that this concatenation of sufficient statistics is jointly sufficient across models
is a property that is rather specific to gibbs random field models , at least from a practical perspective ( see below ) .
figure [ figgrrr ] shows an experiment from @xcite concluding rightly at the agreement between the exact bayes factor and an abc approximation .
comparison between the true bayes factor and the abc approximation in a markov model selection of @xcite , based on @xmath96 simulated sequences and @xmath97 proposals from the prior .
the solid / red line is the diagonal . _
( source : @xcite . ) _ ] @xcite point out that this specific property of gibbs random fields can be extended to any exponential family ( hence to any setting enjoying sufficient statistics , see e.g. @xcite ) .
their argument is an encompassing property : by including all sufficient statistics and all dominating measure statistics in an encompassing model , models under comparison become submodels of the encompassing model .
they then conclude that the concatenation of those statistics is jointly sufficient across models . while this encompassing principle holds in full generality , in particular when comparing models that are already embedded , we think it leads to a biased perspective about the merits of abc for model choice : in practice , complex models do not enjoy sufficient statistics ( if only because they are not exponential families ) . as demonstrated in the next section
, there is more than a mere loss of information due to the use of insufficient statistics and looking at what happens in the limiting case when one is relying on a common sufficient statistic is a formal study that brings light on the potentially huge discrepancy between the abc - based bayes factor and the true bayes factor . to study a solution to the problem in the formal case of the exponential families does not help in the understanding of the discrepancy in non - exponential models .
the difficulty with the arbitrary discrepancy between @xmath70 and @xmath74 is that it is impossible to evaluate in a general setting , while there is no reason to expect a reasonable agreement between both quantities .
a first illustration was produced by @xcite in the setting of @xmath98 time series : a simulation experiment showed that , when comparing an @xmath99 with an @xmath100 model , the abc approximation to the bayes factor was stable ( around @xmath101 ) as @xmath18 decreases , remaining far from the true bayes factor @xmath102 for an @xmath99 simulated sample , while the approximation was @xmath103 against a true value of @xmath104 in the case of a simulated @xmath100 sample .
as a first illustration of the discrepancy due to the use of a sufficient statistic , consider the simple case when a sample @xmath105 could come from either a poisson @xmath106 distribution or from a geometric @xmath107 distribution , already introduced in @xcite as a counter - example to gibbs random fields and later reprocessed in @xcite to support their sufficiency argument . in this setting ,
the sum @xmath108 is a sufficient statistic for both models but not across models .
the distribution of the sample given @xmath109 is a multinomial @xmath110 distribution when the data is poisson , since @xmath109 is then a poisson @xmath111 variable , while it is the uniform distribution with constant probability @xmath112 in the geometric case , since @xmath109 is then a negative binomial @xmath113 variable .
the discrepancy ratio is therefore @xmath114 when simulating @xmath72 poisson or geometric variables and using prior distributions @xmath115 on the respective models , the exact bayes factor can be evaluated and the range and distribution of the discrepancy are therefore available .
figure [ fig : poisneg ] gives the range of @xmath70 versus @xmath74 , showing that @xmath74 is in this case absolutely un - related with @xmath70 : the values produced by both approaches simply have nothing in common .
as noted above , the approximation @xmath74 based on the sufficient statistic @xmath109 is producing figures of the magnitude of a _ single _ observation , while the true bayes factor is of the order of the sample size .
comparison between the true log - bayes factor _
( first axis ) _ for the comparison of a poisson model versus a negative binomial model and of the log - bayes factor based on the sufficient statistic @xmath116 _
( second axis ) _
, for poisson _ ( left ) _ and negative binomial _ ( left ) _ samples of size @xmath117 , based on @xmath118 replications . ]
the discrepancy between both bayes factors is in fact increasing with the sample size , as shown by the following result : consider performing model selection between model 1 : @xmath106 with prior distribution @xmath119 equal to an @xmath120 distribution and model 2 : @xmath107 with a uniform prior distribution @xmath121 when the observed data @xmath3 consists of iid observations with @xmath122 = \theta_0 > 0 $ ] .
then @xmath123 is the minimal sufficient statistic for both models and the bayes factor based on the sufficient statistic @xmath124 , @xmath74 , satisfies @xmath125 therefore , the bayes factor based on the sufficient statistic @xmath124 is _ not _ consistent ; it converges to a non - zero , finite value almost surely . under model 1
, we have @xmath126 , with corresponding likelihood @xmath127 the marginal likelihood of @xmath109 under the prior @xmath128 is then @xmath129 under model 2 , the sufficient statistic has a negative binomial distribution , @xmath130 and thus @xmath131 the corresponding marginal likelihood under the prior @xmath121 is @xmath132 therefore from and , the bayes factor based on the sufficient statistic is given by @xmath133 since the @xmath134 s are iid with mean @xmath135 , the law of large numbers implies that @xmath136 almost surely , thus @xmath137 since @xmath138 .
furthermore , @xmath139 thus from we deduce that @xmath140 proving the result . in this specific setting , @xcite show that adding @xmath141 to the sufficient statistic @xmath109 induces a statistic @xmath142 that is sufficient across both models .
while this is a mathematically correct observation , we think it is not helpful for the understanding of the behaviour of abc - model choice in realistic settings : outside toy examples as the one above and well - structured although complex exponential families like gibbs random fields , it is not possible to come up with completion mechanisms that ensure sufficiency across models and it is therefore more fruitful to consider the diverging behaviour of the abc approximation as given , rather than attempting at solving the problem . first , note that , given a one - dimensional sufficient statistic @xmath143 , the functions @xmath144 and @xmath145 can on principle be anything .
for instance , @xmath146 and @xmath147 is a possible model . in other words , by a reparameterisation of the models , we could observe @xmath148 with @xmath149 this independently of the distributions of @xmath109 under both models .
( this means that we can find two competing models where the distributions of @xmath109 are not connected with @xmath150 nor with @xmath151 . ) because they depend on the choice of those distributions , the true bayes factor and the abc - bayes factor are unrelated and may as well diverge from one another .
admitedly , this construct is artificial in that there is no clear statistical setting when this could occur , but the construct is both mathematically valid and informative about the lack of control over the diverging factor @xmath69 .
if we look at a fully normal @xmath152 setting , we have @xmath153 hence @xmath154 if we reparameterise the observations into @xmath155 , we do get @xmath156 ^ 2 \big/2 \right\}\end{aligned}\ ] ] since the jacobian is @xmath59 .
hence @xmath157 ^ 2 /2 \right\ } \sigma^{-n}\ ] ] considering both models @xmath158 the discrepancy ratio is then given by @xmath159 ^ 2 /2 \right\ } \sigma_1^{-n+1 } } { \exp\left\{-\sigma_2^{-2 } \sum_{i=1}^{n-1 } ( y_i-\bar y)^2/2 - \sigma_2^{-2 } \left [ \sum_{i=1}^{n-1 } ( y_i-\bar y ) \right]^2 /2
\right\ } \sigma_2^{-n+1 } } \\ & \quad= \dfrac{\sigma_2^{n-1}}{\sigma_1^{n-1}}\ , \exp\left\{\dfrac{\sigma_2^{-2}-\sigma_1^{-2}}{2}\left(\sum_{i=1}^{n-1 } ( y_i-\bar y)^2 + \left [ \sum_{i=1}^{n-1 } ( y_i-\bar y ) \right]^2
\right)\right\ } \end{aligned}\ ] ] and is connected with the lack of consistency of the bayes factor : consider performing model selection between model 1 : @xmath160 and model 2 : @xmath161 , @xmath150 and @xmath151 being given , with prior distributions @xmath162 equal to a @xmath163 distribution and when the observed data @xmath3 consists of iid observations with finite mean and variance .
then @xmath123 is the minimal sufficient statistic for both models and the bayes factor based on the sufficient statistic @xmath124 , @xmath74 , satisfies @xmath164 the marginal likelihood associated with @xmath124 and the prior @xmath165 is @xmath166 hence leading to the bayes factor @xmath167 which indeeds converges to @xmath59 as @xmath72 goes to infinity .
figure [ fig : twonormal ] illustrates the behaviour of the discrepancy ratio when @xmath168 and @xmath169 , for datasets of size @xmath170 simulated according to both models .
the discrepancy ( expressed on a log scale ) is once again dramatic , in concordance with the above lemma .
empirical distributions of the log discrepancy @xmath171 for datasets of size @xmath170 simulated from @xmath160 _
( left ) _ and @xmath161 _ ( right ) _ distributions when @xmath168 and @xmath169 , based on @xmath172 replications and a flat prior . ]
if we now turn to an alternative choice of sufficient statistic , using the pair @xmath173 with @xmath174 we follow the solution of @xcite . using a conjugate prior @xmath165 , the true bayes factor is given by @xmath175 and it is equal to the bayes factor based on the corresponding distributions of the pair @xmath173 in the respective models .
again , we do not think this coincidence brings the proper light on the behaviour of the abc approximations in realistic settings .
since its introduction by @xcite and @xcite , abc has been extensively used in several areas involving complex likelihoods , primarily in population genetics . in those domains ,
abc has been used both for point estimation and testing of hypotheses . in realistic settings , with the exception of gibbs random fields that satisfy a resilience property with respect to their sufficient statistics , the conclusions drawn on model comparison can not alas be trusted _ per se _ but require further analyses as to the pertinence of the ( abc ) bayes factor based on the summary statistics .
this paper has only examined in details the case when the summary statistics are sufficient for both models , while practical situations imply the use of in - sufficient statistics , and further research is needed for the latter case . however
, this practical situation implies a wider loss of information compared with the exact inferential approach , hence a wider discrepancy between the exact bayes factor and the quantity produced by an abc approximation .
it thus appears to us an urgent duty to warn the community about the dangers of this approximation , especially when considering the rapidly increasing number of applications using abc for conducting model choice and hypothesis testing . as a final ( and negative ) point ,
we unfortunately do not see an immediate and generic alternative for the approximation of bayes factors because importance sampling techniques are suffering from the same difficulty , namely they only depend on the summary statistics . as a final remark
, we note that @xcite advocate the use of full allelic distributions in an abc framework , instead of resorting to summary statistics .
they show that it is possible to apply abc using allele frequencies to draw inferences in cases where it is difficult to select a set of suitable summary statistics ( and when the complexity of the model or the size of dataset makes it computationally prohibitive to use full - likelihood methods ) . in such settings , were we to consider a model choice problem , the divergence exhibited in the current paper would not occur because the measure of distance does not rely on a reduction of the sample .
the first two authors work has been partly supported by the agence nationale de la recherche ( anr , 212 , rue de bercy 75012 paris ) through the 2009 - 2012 project emile , directed by jean - marie cornuet . | approximate bayesian computation ( abc ) , also known as likelihood - free methods , have become a favourite tool for the analysis of complex stochastic models , primarily in population genetics but also in financial analyses .
we advocated in @xcite the use of abc for bayesian model choice in the specific case of gibbs random fields ( grf ) , relying on a sufficiency property mainly enjoyed by grfs to show that the approach was legitimate . despite having
previously suggested the use of abc for model choice in a wider range of models in the diy abc software @xcite , we present theoretical evidence that the general use of abc for model choice is fraught with danger in the sense that no amount of computation , however large , can guarantee a proper approximation of the posterior probabilities of the models under comparison . * keywords * : likelihood - free methods , bayes factor , diyabc , bayesian model choice , sufficiency . |
the probabilistic nature of quantum mechanics imposes fundamental limits to hypothesis testing and parameter estimation @xcite .
such limits are relevant to many metrological applications , such as optical interferometry , optomechanical sensing , gravitational - wave detection @xcite , optical imaging @xcite , magnetometry , gyroscopy , and atomic clocks @xcite .
the ultimate quantum limits to parameter estimation have been studied extensively in recent years , as they imply that a minimum amount of resource , such as the average photon number for optical phase estimation , is needed to achieve a desired precision , regardless of the measurement method . for the measurement of a single optical phase parameter ,
the ultimate quantum limit to the mean - square error scales as @xmath4 , where @xmath5 is the average photon number of the field which undergoes that phase shift .
this scaling is often called the heisenberg limit .
after years of speculation and debate @xcite , the heisenberg limit for single - parameter linear phase estimation has only recently been proven @xcite .
although decoherence , such as optical loss and dephasing , can impose stricter limitations @xcite , the heisenberg limit is a more fundamental bound and will be increasingly relevant as quantum technologies continue to improve and decoherence effects are further reduced .
many real - world tasks , such as optical imaging @xcite , quantum tomography and system identification @xcite , and waveform estimation ( e.g. estimating a signal that varies continuously in time ) @xcite , require the estimation of multiple parameters .
multiparameter quantum cramr - rao bounds have been known since the 1970s @xcite , but efforts to derive multiparameter heisenberg limits from these bounds have not been successful . this is not surprising , since even in the case of single - parameter phase estimation , it is not possible , without additional assumptions on the state , to derive the heisenberg limit from the quantum cramr - rao bound .
( the latter gives a lower bound on the mean - square error of @xmath6 , which does not imply the heisenberg limit of @xmath4 , as can be seen from the state @xmath7 which has @xmath8 but divergent @xmath9 . ) in ref .
@xcite , some of us recently proposed a heisenberg - style limit for the estimation of an optical phase waveform with stationary gaussian prior statistics and a power - law spectrum .
however , that limit , being derived from a quantum cramr - rao bound , requires additional assumptions : it applies only to the specific class of optical beams described by gaussian fields , with statistics that are both stationary and time - symmetric .
a very different approach was that of ref .
@xcite , which derives a multiparameter heisenberg limit for independent parameters by applying the single - parameter heisenberg limit to each parameter . in practice ,
multiple parameters often have nontrivial prior correlations , particularly in the case of continuous waveform estimation , where the correlations are crucial to pose the problem @xcite .
thus the existence of general heisenberg limits for such cases has remained an open question . in this paper , we derive new quantum bounds on multiparameter estimation by developing quantum versions of the bell - ziv - zakai bounds .
the bell - ziv - zakai bounds were proposed in 1997 by bell _
et al . _
@xcite , building upon the ziv - zakai bound @xcite , and futher generalized by basu and bresler @xcite .
we then apply our bounds to the notable task of quantum optical phase waveform estimation . here , the waveform to be estimated is a time - varying phase shift signal , @xmath10 , applied to an optical beam . for a waveform @xmath10 with stationary gaussian prior statistics and a power - law spectrum ( @xmath11 ) , we prove a lower bound on the mean - square error with a @xmath2 scaling , where @xmath3 is the mean photon flux .
this proof confirms that the scaling previously proposed in ref .
@xcite is valid for arbitrary quantum states .
moreover , we show that this scaling is achievable for all @xmath1 .
previously , achievability has been shown only numerically , and only for @xmath12 @xcite .
by contrast the results in the current paper are completely rigorous heisenberg bounds , being both applicable to arbitrary field states , and achievable , for all @xmath1 .
this paper is separated into two main parts . the first part , sec .
[ qzzb ] to sec .
[ sec : ach ] , assumes unbounded parameters and focuses on the mean - square error as the distortion measure ( i.e. the figure of merit for the accuracy of the estimation ) .
the second part , sec .
[ sec : per ] and sec .
[ sec : ach2 ] , focuses on periodic distortion functions , which are more appropriate for periodic parameters such as phase or orientation angles for gyroscopy .
they are also insensitive to phase - wrap errors and enable us to rigorously prove that our bounds are achievable .
first we summarize known results for the classical estimation problem , then present quantum versions of the bounds in sec .
[ sec : qbzzb ] .
let @xmath13 be a column vector of unknown real parameters , @xmath14 be the prior probability density , and @xmath15 be the likelihood function with observation @xmath16 . both @xmath13 and @xmath16 are random variables .
note that @xmath16 need not be the same dimension as @xmath13 .
further , let @xmath17 be the estimator of @xmath13 from @xmath16 .
( we use @xmath18 rather than @xmath19 , as is common in statistics , to avoid possible confusion with quantum operators . )
we also define the error vector as @xmath20 to characterise the performance of the estimate , we consider a distortion function of the form @xmath21 , where @xmath22 is a given but arbitrary real column vector that defines the error components of interest , and @xmath23 denotes the transpose .
for example , the mean - square error for a particular component @xmath24 is the expected value of a distortion function @xmath25 with @xmath26 , so @xmath27}^2 .
\label{point}\ ] ] suppose that the distortion function is symmetric [ that is , @xmath28 , nondecreasing on @xmath29 , differentiable , and has @xmath30
. then the expected distortion is , from eq .
( 44 ) of ref .
@xcite , @xmath31 } & = \frac{1}{2 } \int_0^\infty d\tau \ , \dot d{\left(\frac{\tau}{2}\right ) } \textrm{pr}{\left(|{\boldsymbol{u}}^\top{\boldsymbol{\epsilon}}| \ge \frac{\tau}{2}\right ) } , \label{d_bound}\end{aligned}\ ] ] where @xmath32 is the derivative of @xmath33 , @xmath34 denotes expectation over @xmath13 and @xmath16 , and pr is the probability for the boolean function of these random variables to be true . for a general mean - square error criterion
, the expected distortion can be expressed in terms of the error covariance matrix @xmath35 as @xmath31 } & = { \mathbb e}{\left[({\boldsymbol{u}}^\top{\boldsymbol{\epsilon}})^2\right ] } = { \boldsymbol{u}}^\top{\boldsymbol{\sigma}}{\boldsymbol{u } } , { \nonumber \\}{\boldsymbol{\sigma } } & : = { \mathbb e}{\left({\boldsymbol{\epsilon}}{\boldsymbol{\epsilon}}^\top\right)}. \label{quad_d}\end{aligned}\ ] ] since @xmath36 is assumed to be nonnegative , a lower bound on the expected distortion can be obtained by lower - bounding the probability @xmath37 . using eqs .
( 31 ) and ( 35 ) of @xcite to bound @xmath37 and noting property 1 of @xcite , yields the bell - ziv - zakai bounds @xcite : @xmath38 } \ge \frac{1}{2 } \int_0^\infty d\tau \ , \dot d{\left(\frac{\tau}{2}\right ) } \mathcal v\bigg\ { \max_{{\boldsymbol{v}}:{\boldsymbol{u}}^\top{\boldsymbol{v } } = 1 } \int d{\boldsymbol{x } } { \nonumber \\}&\quad\times { \left[p_{{\boldsymbol{x}}}({\boldsymbol{x}})+p_{{\boldsymbol{x}}}({\boldsymbol{x}}+{\boldsymbol{v}}\tau)\right ] } p_e({\boldsymbol{x}},{\boldsymbol{x}}+{\boldsymbol{v}}\tau)\bigg\ } , \label{bzzb1 } \\ & \ge \int_0^\infty d\tau \ , \dot d{\left(\frac{\tau}{2}\right ) } \mathcal v\bigg\{\max_{{\boldsymbol{v } } : { \boldsymbol{u}}^\top { \boldsymbol{v } } = 1 } \int d{\boldsymbol{x } } { \nonumber \\}&\quad\times \min{\left[p_{{\boldsymbol{x}}}({\boldsymbol{x}}),p_{{\boldsymbol{x}}}({\boldsymbol{x } } + { \boldsymbol{v}}\tau)\right ] } p_e^{\rm el}({\boldsymbol{x}},{\boldsymbol{x}}+{\boldsymbol{v}}\tau)\bigg\ } .
\label{bzzb2}\end{aligned}\ ] ] here @xmath39 is the valley - filling function defined as @xmath40 and @xmath41 is the minimum error probability for the bayesian binary hypothesis testing problem with hypotheses defined as @xmath42 and @xmath43 , observation probability densities given by @xmath44 and @xmath45 , and prior probabilities given by @xmath46 to be explicit @xcite , @xmath47 @xmath48 is defined in the same way as @xmath49 except that the prior probabilities are equal ( @xmath50 ) . for the quantum parameter estimation problem ,
let @xmath51 be the density operator that describes the state of a quantum probe as a function of the unknown parameter @xmath52 , and @xmath53 be the positive operator - valued measure ( povm ) that describes the measurement with outcome @xmath54 @xcite .
the likelihood function becomes @xmath55,\end{aligned}\ ] ] with @xmath56 denoting the operator trace .
it is known @xcite that , for any povm , @xmath57 } , \label{f_bound}\end{aligned}\ ] ] where @xmath58 is the uhlmann fidelity .
equations and , together with the first bell - ziv - zakai bound given by eq .
( [ bzzb1 ] ) , then give quantum lower bounds on the estimation error : @xmath38 } \ge \frac{1}{4 } \int_0^\infty d\tau \
, \dot d{\left(\frac{\tau}{2}\right ) } { \nonumber \\}&\quad\times \mathcal v\bigg\ { \max_{{\boldsymbol{v}}:{\boldsymbol{u}}^\top{\boldsymbol{v } } = 1 } \int d{\boldsymbol{x}}{\left[p_{{\boldsymbol{x}}}({\boldsymbol{x}})+p_{{\boldsymbol{x}}}({\boldsymbol{x}}+{\boldsymbol{v}}\tau)\right ] } { \nonumber \\}&\quad\times \left(1 - \big{|}\big{|}\pi_0\rho_{{\boldsymbol{x}}}- \pi_1\rho_{{\boldsymbol{x}}+{\boldsymbol{v}}\tau}\big{|}\big{|}_1\right ) \bigg\ } \label{qbzzb1 } \\ & \ge \frac{1}{4 } \int_0^\infty d\tau \ , \dot d{\left(\frac{\tau}{2}\right ) } { \nonumber \\}&\quad\times \mathcal v\bigg\ { \max_{{\boldsymbol{v}}:{\boldsymbol{u}}^\top{\boldsymbol{v } } = 1 } \int d{\boldsymbol{x}}{\left[p_{{\boldsymbol{x}}}({\boldsymbol{x}})+p_{{\boldsymbol{x}}}({\boldsymbol{x}}+{\boldsymbol{v}}\tau)\right ] } { \nonumber \\}&\quad\times { \left[1 - \sqrt{1 - 4\pi_0\pi_1f({\boldsymbol{x}},{\boldsymbol{x}}+{\boldsymbol{v}}\tau)}\right ] } \bigg\}.\end{aligned}\ ] ] similarly , eq . ( [ bzzb2 ] ) and quantum bounds on @xmath48 via eqs .
( [ trace_bound ] ) and ( [ f_bound ] ) lead to the lower bounds @xmath38 } \ge \frac{1}{2 } \int_0^\infty d\tau \
, \dot d{\left(\frac{\tau}{2}\right ) } { \nonumber \\}&\quad\times \mathcal v\bigg\{\max_{{\boldsymbol{v } } : { \boldsymbol{u}}^\top { \boldsymbol{v } } = 1 } \int d{\boldsymbol{x } } \min{\left[p_{{\boldsymbol{x}}}({\boldsymbol{x}}),p_{{\boldsymbol{x}}}({\boldsymbol{x } } + { \boldsymbol{v}}\tau)\right ] } { \nonumber \\}&\quad\times \left(1 - \frac{1}{2}\big{|}\big{|}\rho_{{\boldsymbol{x}}}- \rho_{{\boldsymbol{x}}+{\boldsymbol{v}}\tau}\big{|}\big{|}_1\right ) \bigg\ } \\ & \ge \frac{1}{2 } \int_0^\infty d\tau \ , \dot d{\left(\frac{\tau}{2}\right ) } { \nonumber \\}&\quad\times \mathcal v\bigg\{\max_{{\boldsymbol{v } } : { \boldsymbol{u}}^\top { \boldsymbol{v } } = 1 } \int d{\boldsymbol{x } } \min{\left[p_{{\boldsymbol{x}}}({\boldsymbol{x}}),p_{{\boldsymbol{x}}}({\boldsymbol{x } } + { \boldsymbol{v}}\tau)\right ] } { \nonumber \\}&\quad\times { \left[1 - \sqrt{1-f({\boldsymbol{x}},{\boldsymbol{x}}+{\boldsymbol{v}}\tau)}\right ] } \bigg\}. \label{qbzzb4}\end{aligned}\ ] ] we call eqs .
( [ qbzzb1])([qbzzb4 ] ) the quantum bell - ziv - zakai bounds . to derive further analytic results , we focus on the fidelity bound given by eq .
( [ qbzzb4 ] ) .
it can be further simplified if @xmath59 does not depend on @xmath52 and the prior @xmath14 is a multivariate gaussian distribution .
the integral with respect to @xmath52 then becomes , using eqs .
( a2 ) and ( a10 ) of ref .
@xcite , @xmath60 } = { \operatorname{erfc}}{\left(\frac{\tau}{\tau_0}\right ) } , \label{erfc}\end{aligned}\ ] ] where @xmath61 and @xmath62 is the prior covariance matrix .
a convenient lower bound on the erfc function is @xmath63 where @xmath64 is the triangle function @xmath65 , as shown in fig .
[ fig : erfc ] .
may be entangled between the modes , and each mode passes through a phase shift @xmath66 .
the output may be measured via some general joint measurement @xmath53.,scaledwidth=48.0% ] we now consider the problem of phase estimation from the measurement of quantum optical modes , as illustrated in fig .
[ phasemod ] .
the output quantum state is @xmath67 where @xmath68 is the initial quantum state and @xmath69 is a column vector of photon number operators for the optical modes .
we use a hat to distinguish the number operator from other uses of @xmath70 as an integer .
we will not otherwise use a hat to indicate operators .
purifying @xmath51 to @xmath71 , and taking the purification of @xmath72 to be @xmath73 , uhlmann s theorem @xcite yields a lower bound on @xmath74 given by @xcite @xmath75},\end{aligned}\ ] ] where we have defined @xmath76 , and @xmath77 is the photon - number distribution of the initial quantum state , with @xmath78 an eigenstate of @xmath69 . to derive a bound on @xmath59 in terms of the average photon numbers ,
the following bound on cosine is useful : @xmath79 where @xmath80 is a solution of @xmath81 , as shown in fig .
[ cosine_bound ] .
this leads to @xmath82 } { \nonumber \\}&\ge \sum_{{\boldsymbol{n}},{\boldsymbol{m}}}c({\boldsymbol{n}})c({\boldsymbol{m } } ) { \left[1-\lambda\tau{\left(|{\boldsymbol{v}}^\top{\boldsymbol{n}}|+|{\boldsymbol{v}}^\top{\boldsymbol{m}}|\right)}\right ] } { \nonumber \\}&\geq \sum_{{\boldsymbol{n}},{\boldsymbol{m}}}c({\boldsymbol{n}})c({\boldsymbol{m } } ) { \left[1-\lambda\tau{\left(|{\boldsymbol{v}}|^\top{\boldsymbol{n}}+|{\boldsymbol{v}}|^\top{\boldsymbol{m}}\right)}\right ] } { \nonumber \\ } & = 1 - 2\lambda\tau |{\boldsymbol{v}}|^\top{\left\langle\hat{{\boldsymbol{n}}}\right\rangle},\end{aligned}\ ] ] where @xmath83 means taking the absolute value of each element of @xmath84 .
since @xmath85 , a tighter bound is @xmath86 a slightly tighter bound may be obtained using the method in refs .
@xcite , but the scaling would remain the same . focusing on the mean - square error , putting eqs .
( [ quad_d ] ) , ( [ qbzzb4 ] ) , ( [ erfc ] ) , ( [ erfc_bound ] ) , and ( [ speed_limit ] ) together , and using @xmath87 , @xmath88 the maximization of @xmath89 , subject to the constraint @xmath90 , gives the tightest bound , but it is difficult to perform analytically . in the next section ,
we shall focus on waveform estimation and discover that an appropriate choice of @xmath84 , though suboptimal , can still lead to a reasonably tight bound .
we now consider phase modulation that varies in time , as illustrated in fig .
[ waveform ] .
define discrete time as @xmath91 @xmath92 being an integer .
each parameter @xmath93 corresponds to a phase at time @xmath94 : @xmath95 and each photon - number operator @xmath96 is related to the photon - flux operator @xmath97 by @xmath98 other quantities are redefined as follows : @xmath99 in the continuous - time limit @xmath100 , the mean - square error becomes @xmath101 and the constraint in eq .
( [ qbzzb ] ) becomes @xmath102 to evaluate the bound @xmath89 given by eq .
( [ z ] ) in this limit , we need to compute @xmath103 given by eq .
( [ tau0 ] ) and @xmath104 given by eq .
( [ speed_limit ] ) .
they depend on the following : @xmath105 is defined by @xmath106 assume now that the prior statistics of @xmath10 are stationary .
this means that we can define a prior power spectral density @xmath107 such that @xmath108,\end{aligned}\ ] ] and the inverse of @xmath109 is given by @xmath110.\end{aligned}\ ] ] we then obtain @xmath111 we are particularly interested in the estimation error at a particular time @xmath112 , in which case @xmath113 and from eq .
( [ conuv ] ) , @xmath114 .
we will see below that the choice @xmath115 , so @xmath116 where @xmath117 is a convenient one for deriving a lower bound , for a suitable choice of characteristic time @xmath118 .
it gives latexmath:[\ ] ] where @xmath437 is the riemann zeta function .
this means that the sum is exponentially small in @xmath3 , which is why it can be omitted in eq . .
67ifxundefined [ 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 10.1126/science.1104149 [ * * , ( ) ] http://dx.doi.org/10.1038/nphoton.2011.35 [ * * , ( ) ] http://www.degruyter.com/view/product/180320[__ ] ( , , ) @noop _ _ ( , , ) link:\doibase 10.1103/physrevlett.106.090401 [ * * , ( ) ] link:\doibase 10.1103/physreva.86.042115 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/15/7/073005 [ * * , ( ) ] link:\doibase 10.1126/science.1086489 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.253601 [ * * , ( ) ] http://dx.doi.org/10.1038/nphoton.2012.346 [ * * , ( ) ] link:\doibase 10.1103/physreva.54.r4649 [ * * , ( ) ] link:\doibase 10.1103/physreva.33.4033 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.75.2944 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.77.2352 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.180402 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.107.059904 [ * * , ( ) ] http://stacks.iop.org/1367-2630/14/i=9/a=093052 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.034101 [ * * , ( ) ] link:\doibase 10.1016/j.aop.2012.12.004 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.103602 [ * * , ( ) ] http://stacks.iop.org/1751-8121/46/i=3/a=035302 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.230401 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.260405 [ * * , ( ) ] link:\doibase 10.1103/physreva.86.053813 [ * * , ( ) ] link:\doibase 10.1103/physreva.85.041802 [ * * , ( ) ] @noop ( ) , link:\doibase 10.1103/physrevlett.108.210404 [ * * , ( ) ] link:\doibase 10.1103/physrevx.2.041006 [ * * , ( ) ] link:\doibase 10.1103/physreva.83.021804 [ * * , ( ) ] link:\doibase 10.1038/nphys1958 [ * * , ( ) ] link:\doibase 10.1007/s13538 - 011 - 0037-y [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.190404 [ * * , ( ) ] @noop ( ) , link:\doibase 10.1038/ncomms2067 [ * * , ( ) ] , @noop ( ) , link:\doibase 10.1103/revmodphys.71.1539 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.070403 [ * * , ( ) ] and , eds . ,
@noop _ _ ( , , ) link:\doibase 10.1103/physreva.79.062301 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.113601 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.093601 [ * * , ( ) ] link:\doibase 10.1126/science.1225258 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.163602 [ * * , ( ) ] link:\doibase 10.1109/tit.1973.1055103 [ * * , ( ) ] link:\doibase 10.1109/tit.1974.1055173 [ * * , ( ) ] link:\doibase 10.1142/s0219749909004839 [ * * , ( ) ] @noop ( ) , http://onlinelibrary.wiley.com/book/10.1002/0471221082[__ ] ( , , ) link:\doibase 10.1109/18.556118 [ * * , ( ) ] link:\doibase 10.1109/tit.1969.1054301 [ * * , ( ) ] link:\doibase 10.1109/18.841198 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.019901 [ * * , ( ) ] and , eds .
, @noop _ _ ( , , ) link:\doibase 10.1109/tcom.1967.1089532 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1016/0034 - 4877(76)90060 - 4 [ * * , ( ) ] link:\doibase 10.1103/physreva.67.052109 [ * * , ( ) ] http://stacks.iop.org/0954-8998/3/i=5/a=003 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.75.4587 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.89.133602 [ * * , ( ) ] link:\doibase 10.1103/physreva.73.063824 [ * * , ( ) ] link:\doibase 10.1103/physreva.73.063824 [ * * , ( ) ] link:\doibase 10.1103/physreva.78.053820 [ * * , ( ) ] link:\doibase 10.1103/physreva.79.053843 [ * * , ( ) ] http://books.google.com/books?id=4htgaaaayaaj[__ ] ( , , ) | we propose quantum versions of the bell - ziv - zakai lower bounds on the error in multiparameter estimation . as an application we consider measurement of a time - varying optical phase signal with stationary gaussian prior statistics and a power law spectrum @xmath0 , with @xmath1 . with no other assumptions ,
we show that the mean - square error has a lower bound scaling as @xmath2 , where @xmath3 is the time - averaged mean photon flux .
moreover , we show that this accuracy is achievable by sampling and interpolation , for any @xmath1 .
this bound is thus a rigorous generalization of the heisenberg limit , for measurement of a single unknown optical phase , to a stochastically varying optical phase . |
advection - dominated accretion flow ( adaf ) is an important type of solution for black hole accretion .
a prominent feature of an adaf compared to the standard thin disk is its low radiative efficient at low accretion rates ( narayan & yi 1994 , hereafter ny94 ; narayan & yi 1995 ; see narayan , mahadevan & quataert 1998 and kato , fukue & mineshige 1998 for reviews ) .
the adaf solution has received much attention in the past years because it successfully explains why some nearby galaxies are so dim even though their accretion rates are not very small ( see narayan 2005 , yuan 2007 , and ho 2008 for reviews ) .
the best evidence comes from the supermassive black hole in our galactic center , sgr a * ( narayan , mahadevan & yi 1995 ; manmoto , kusunoze & mineshige 1997 ; yuan , quataert & narayan 2003 ) . from _
observations and bondi theory , we can estimate the mass accretion rate of sgr a*. if the accretion flow were not an adaf but a standard thin disk , the luminosity would be five orders of magnitude larger than observed ( yuan , quataert & narayan 2003 ) . another attractive feature of an adaf is that it can partly solve the problem of the origin of x - ray emission from accretion flows .
the temperature of a standard thin disk at the inner disk is only @xmath1k for a supermassive black hole , too low to produce x - ray emission ( frank , king & raine 2002 ) .
a hot corona has been thought to be responsible for the x - ray emission , but recent mhd simulations of disks show that they have hardly any coronae ( hirose , krolik & stone 2006 ) . on the other hand ,
the temperature of an adaf is high enough to produce x - ray emission .
of course , a canonical adaf exists only below a critical accretion rate @xmath2 with @xmath3 and @xmath4 is the viscous parameter , which corresponds to @xmath5 at most ( esin , mcclintock & narayan 1997 ) .
therefore it can not explain luminous x - ray sources such as quasars .
the luminous hot accretion flow ( lhaf ; yuan 2001 ) , which is an extension of an adaf to higher accreton rates , is promising , but the details of this model have not been fully worked out ( see yuan et al .
2007 for an example of application to luminous black hole x - ray binaries ) . in spite of the great success of adafs ,
more work is required to test the model . on the one hand , it would useful to expand the application of adafs to more sources , and on the another hand , such modeling is expected to help us understand some important microphysical issues which are still unclear .
one example of the latter is the value of @xmath6 ( defined in eqs . 4 and 5 ) , which measures the amount of direct electron heating through viscous dissipation in a hot accretion flow ( sharma et al .
another is the potential importance of collective plasma effects which will determine how realistic the two - temperature assumption is ( begelman & chiueh 1988 ; yuan et al .
2006 ) . for such work ,
the global solution rather than the self - similar solution of the adaf equations is required .
this is because most of the radiation of an adaf comes from its innermost region where the self - similar solution breaks down .
however , it is technically very difficult to calculate the global solution of an adaf .
an adaf is transonic , and thus its global solution should satisfy the sonic - point condition in addition to the outer boundary condition .
mathematically , it is a two point boundary value problem and not easy to deal with .
this is an obstacle to the wide application of the adaf model . in this paper
we propose a simplified global adaf solution .
we adopt a simple algebraic relation to replace the radial momentum equation , thus avoiding the two point boundary value problem .
we present our approach in 2 and show some examples in 3 .
the final section is devoted to a short summary . watarai ( 2007 ) has recently presented related work .
the basic equations of an adaf describe the conservation of mass , radial and aximuthal components of the momentum , and energy ( e.g. , narayan , mahadevan & quataert 1998 ) : @xmath7 @xmath8 @xmath9 @xmath10 @xmath11 all the quantities have their usual meaning . in the present paper
we do not include outflows from the adaf , but it is easy to extend our calculation to that case by simply using a radius - dependent mass accretion rate , @xmath12 with @xmath13 being a constant ( e.g. , yuan , quataert & narayan 2003 ) .
the quantity @xmath6 in equations ( 4 ) and ( 5 ) describes the fraction of the turbulent dissipation rate @xmath14 which directly heats electrons ; we set @xmath15 .
the quantity @xmath16 describes the energy transfer rate from ions to electrons by coulomb collision , and @xmath17 is the radiative cooling rate .
we consider synchrotron and bremsstrahlung emissions and their comptonization .
the details of the calculation of the spectrum can be found in yuan , quataert & narayan ( 2003 ) .
we consider a schwarzschild black hole and adopt the paczyski & wiita ( 1980 ) potential to mimic its geometry .
the most difficult part of solving the global solution is the radial momentum equation ( 2 ) .
our key idea of simplifying the global adaf solution is to replace this differential equation by the following simple algebraic relation : @xmath18 with @xmath19 where @xmath20 is the innermost stable circular orbit .
the above simplification is based on the following physical consideration .
the immediate idea we think of to simplify the radial momentum equation is to use the self - similar solution obtained by ny94 .
consider eqs .
( 7)-(9 ) in ny94
. we can use eq .
( 9 ) to solve for @xmath21 in terms of the sound speed @xmath22 : 5 + 2^=2 , ^=-
. we can then substitute this in eqs .
( 8) and ( 7 ) in ny94 to obtain for a fully advection - dominated flow ( @xmath23 ) = ( 1-)^1/2 = = const .
here @xmath24 is the adiabatic index .
we therefore in principle could set @xmath25 in eq .
( 6 ) to this constant . however , we find that the simplfied solution is very sensitive to the value of @xmath25 .
the reason is that , as we will see , the radial velocity is sensitive to the value of @xmath25 ( ref .
eq . 9 below ) .
the velocity determines the density , and also the temperature via the energy equations , two quantities that determine the emitted spectrum .
we therefore set @xmath25 as a free parameter which we adjust for different accretion parameters @xmath26 and @xmath4 to get the best approximation . because the angular momentum @xmath27 in a global solution keeps decreasing with decreasing radius , while the keplerian angular momentum @xmath28 begins to increase when @xmath29 ( fig .
1 ) , @xmath25 can not be a constant when @xmath29 .
instead we require the angular momentum to be continuous at @xmath30 and assume that it is proportional to @xmath31 . after some tests we set @xmath32 , independent of the values of @xmath26 and @xmath4 .
thus @xmath33 is not a free parameter in our model . substituting eq.(6 ) into eq.(3 ) we have @xmath34 the quantity @xmath35 is the specific angular momentum of the accretion gas when it falls into the black hole and it is the eigenvalue of the exact global solution . in our simplified model , we set @xmath35 as the second free parameter and adjust its value to get the best approximate solution for @xmath36 .
substituting eqs .
( 1 ) , ( 6 ) and ( 9 ) into the energy equations for ions and electrons , eqs . ( 4 ) and ( 5 ) , we have two differential equations with two unknown variables , @xmath37 and @xmath38 .
all other quantites such as @xmath36 , @xmath39 , @xmath22 and @xmath40 can be expressed as simple functions of @xmath37 and @xmath38 for a given @xmath26 and @xmath4 and assumed values of the free parameters @xmath41 and @xmath35 .
when @xmath37 and @xmath38 are given at the outer boundary , we can easily integrate the differential equations inwards to get the approximate global solution .
we adjust the values of @xmath41 and @xmath35 for different @xmath26 and @xmath4 to get the best simplified global solution . here
`` best '' means that the profiles of all quantities such as @xmath42 , and most importantly , the emitted spectrum , are very close to the exact global solution .
because our main purpose is to model the continuum spectrum of black hole sources ( agns and black hole x - ray binaries ) , our first priority will be the closeness of the spectrum when we judge how good a simplified solution is . in the calculations presented here
, we set @xmath26 of the approximate solution equal to @xmath26 of the global solution it is meant to fit .
however , in real applications , we only know the spectrum rather than @xmath26 .
so it might be more realistic to adjust @xmath43 of the simplified solution to fit the spectrum produced by the exact global solution with a given @xmath44 rather than setting @xmath45 .
fortunately we find that @xmath46 and @xmath44 are very close , typically eviating by no more than @xmath47 .
when modeling black hole sources with an adaf model , the accretion rate @xmath26 spans a wide range , say from @xmath48 to @xmath49 . but
the value of @xmath4 adopted in adaf modeling ( e.g. , narayan , mahadevan & quataert 1998 ) is usually within a very narrow range , @xmath0 .
this is also supported by mhd numerical simulations of accretion flows ( e.g. , hawley & krolik 2001 ) .
we adjust the values of @xmath35 and @xmath41 to obtain the `` best '' simplified adaf solutions for @xmath26 and @xmath4 within the above ranges . as state below we find that the same set of ( @xmath50 ) often holds for quite a wide range of @xmath26
. we first present results for @xmath51 .
we find that in this case the simplified global solution with @xmath52 gives a satisfactory spectrum for any @xmath26 .
1 & 2 give two examples with @xmath53 and @xmath54 , respectively .
the dashed lines in the figure denote the exact global solution while the solid lines are for the simplified global solution .
the plots in each figure show the emitted spectrum , mach number , electron and ion temperature , density , and the angular momentum , respectively . for fig .
1 , the outer boundary is at @xmath55 and the outer boundary condition is @xmath56 with the virial temperature @xmath57 . for fig .
2 , the outer boundary is at @xmath58 and the outer boundary condition is @xmath59 . when @xmath60 , it is hard for a single set of ( @xmath63 ) to give a good solution for all @xmath26 .
when @xmath26 is relatively low , @xmath62 , we find @xmath61 gives a satisfactory solution . fig .
3 shows an example with @xmath64 .
the outer boundary conditions are @xmath65 at @xmath66 .
when @xmath26 is relatively high , @xmath69 , we find @xmath67 gives a satisfactory solution .
4 shows an example of @xmath70 .
the outer boundary conditions are @xmath71 at @xmath72 .
note this solution is in the regime of lhaf , because @xmath49 is well above the critical accretion rate of an adaf @xmath73 . for other values of @xmath4
, we find that simply using the `` linear fit '' values of @xmath74 between those for @xmath51 and @xmath75 gives a good solution .
for example , for @xmath76 and @xmath64 , the values of @xmath35 and @xmath41 are just @xmath77 . for @xmath76 and @xmath78 , @xmath79 .
the global solution of adafs is difficult to calculate because it is mathematically a two point boundary value problem .
this hampers wide application of the adaf model .
we propose a simplifed global solution to overcome this difficulty .
prompted by the self - similar solution of adafs , we replace the radial momentum equation , which is the most difficult to handle , with a simple algebraic relation ( eq .
6 ) , and then solve the remaining two diferential equations ( eqs . 4 & 5 ) .
we adjust the two free parameters ( @xmath35 and @xmath41 in eqs . 6 and 9 ) so that we obtain the best approximation compared to the exact global adaf solution .
the spectra of the simplified solutions are impressively good , as shown in figs
. 1 - 4 for various @xmath26 and @xmath4 .
we have been unable to identify a single set of values of the two adjustable constants @xmath35 and @xmath41 which works for all adaf models . however , when @xmath4 is large , say @xmath80 , we find that @xmath81 , @xmath82 gives very good results for all accretion rates @xmath26 for which an adaf solution is possible .
recent work by sharma et al .
( 2006 ) suggests that the viscosity parameter in the collisionless plasma in a hot accretion flow will be larger than in a standard thin disk .
therefore , @xmath83 is probably not unrealistic for an adaf
. it would thus be reasonable to use a single set of parameters , @xmath51 , @xmath81 , @xmath82 , for practical applications of the approximate global model described here .
the success of the present work encourages us to extend our approach to the case of a slim disk , which is an extension of the standard thin disk to accretion rates above the eddington rate ( abramowicz et al .
it potentially has important application in ulxs and narrow - line seyfert 1 galaxies ( mineshige et al .
2000 ; watarai et al .
we hope to report the results in a future paper .
this work was supported in part by the natural science foundation of china ( grant 10773024 ) , shanghai pujiang program , bairen program of cas ( f.y . ) , the knowledge innovation program of cas , and shanghai postdoctoral scientific program ( r.m . ) . | when we model black hole accretion sources such as active galactic nuclei and black hole x - ray binaries as advection - dominated accretion flows ( adafs ) , it is neccesary to use the global solution to the equations rather than the simpler self - similar solution , since the latter is inaccurate in the region near the black hole where most of the radiation is emitted .
however , technically , it is a difficult task to calculate the global solution because of the transonic nature of the flow , which makes it a two - point boundary value problem . in this paper
we propose a simplified approach for calculating the global adaf solution .
we replace the radial momentum equation by a simple algebraic relation between the angular velocity of the gas and the keplerian angular velocity , while keeping all other equations unchanged .
it is then easy to solve the differential energy equations to obtain an approximate global solution . by adjusting the free parameters
, we find that for almost any accretion rate and for @xmath0 we can get good simplified global solutions .
the predicted spectra from the approximate solutions are very close to the spectra obtained from the true global solutions . |
the influence of initial conditions on decaying magnetohydrodynamic ( mhd ) turbulence received considerable interest in the 1980 s , because of its relevance to explain solar - wind data @xcite .
indeed , in magnetohydrodynamics the behavior of decaying turbulent flow depends strongly on the initial conditions , and different initial values and ratios of integral quantities can lead to a wide variety of distinct behaviors .
the first systematic study of the different possible types of decay was performed by ting , matthaeus and montgomery @xcite , who identified four classes of possible decay behavior , corresponding roughly to a magnetically dominated , a hydrodynamically dominated , a magnetically - hydrodynamically equipartitioned and an erratic transition regime .
their study considered the two - dimensional case , which is not only relevant in applications in which an externally imposed field renders the flows quasi two - dimensional , but also from a general physical understanding of mhd turbulence , which behaves quite similar in two and three dimensions , due to the equivalent role of the ideal invariants @xcite . whereas the influence of the initial conditions on decaying mhd turbulence has been studied and understood to some extend , studies on the effect of boundary conditions have been limited to low resolutions @xcite , imposed by the numerical methods used to account for boundaries .
even though these investigations highlighted interesting physics , higher resolution simulations are needed to obtain a better understanding of wall - bounded mhd , which plays a dominant role in geophysical flows in the core of planets such as the earth and industrial processes involving liquid metals . for the hydrodynamic case it was found that
, boundary conditions have a significant influence on two - dimensional turbulence @xcite .
in contrast to the periodic domain , where generally a long lasting state is found with a functional @xmath0 relationship between the vorticity and the stream function @xcite , corresponding to two counterrotating vortices , in a bounded domain with no - slip wall conditions the final state yields an axisymmetric vorticity distribution , for which a linear relation between vorticity and stream function is observed @xcite .
= 0.02 cm ( 200,100 ) ( -150,0 ) ( 200,1 ) .
the fluid domain @xmath1 is a circular container with radius @xmath2 , surrounded by the solid domain @xmath3.[domain],title="fig : " ] ( 260,-8)@xmath4 ( 260,50)@xmath1 ( 210,15 ) @xmath3 in the present work we propose an extension of the volume penalization method @xcite to two - dimensional mhd to compute decaying flows in bounded domains using an efficient fourier pseudo - spectral method .
we address the following questions : what is the influence of confinement by fixed solid boundaries on decaying two - dimensional mhd turbulence ?
do the four regimes found by ting @xcite continue to exist in the presence of boundaries ?
what are the final ( viscously decaying ) states ?
we consider resistive mhd , formulated in usual dimensionless variables @xmath5 and @xmath6 which are respectively the velocity and the magnetic field . the flow is considered to be two - dimensional , incompressible and we assume the mass density to be constant .
+ the governing equations are the following : @xmath7 here @xmath8 and @xmath9 are respectively the kinematic viscosity and the magnetic diffusivity .
@xmath10 is the vorticity , @xmath11 is the current density .
furthermore we define the vector potential @xmath12 as @xmath13 and the stream function @xmath14 as @xmath15 .
an originality in our approach is the way in which the boundary conditions are imposed : we use volume ( or surface in 2d ) penalization @xcite to include the boundary conditions .
this method has the advantage that arbitrary basis - functions can be used . in our case
a fourier pseudo - spectral code is employed . the advantage with respect to a method based on a decompostion in terms of chandrasekhar - kendall eigenfunctions
@xcite is , that fast fourier transforms can be used , allowing for high resolution computations of low computational cost .
also , its application to three dimensional flows is conceptually straightforward and will be addressed in a future work .
the additional terms on the right hand side of equation ( [ qte_mvt ] ) and ( [ eq_magnetic ] ) correspond to this penalization - method .
the quantities @xmath16 and @xmath17 correspond to the values imposed in the solid part of the numerical domain @xmath3 , illustrated in figure [ domain ] . here
we choose @xmath18 and @xmath19 ( where @xmath20 is the tangential component of @xmath21 at the wall ) , corresponding to vanishing velocity and no penetration of magnetic field into the solid domain which is hence considered as a perfect conductor , coated inside with a thin layer of insulant , which guarantees that the current density can not penetrate into the solid @xcite . the mask function @xmath22 is equal to 0 inside @xmath1 ( where the penalization terms thereby dissappear ) and equal to 1 inside @xmath3 .
the physical idea is to model the solid part as a porous medium whose permeability @xmath23 tends to zero @xcite . for @xmath23
@xmath24 0 , where the obstacle is present , the velocity @xmath25 tends to @xmath26 and the magnetic field @xmath27 tends to @xmath28 .
the nature of the boundary condition for the velocity is thus no - slip at the wall .
the equations are discretized with a classical fourier pseudo - spectral method imposing periodic boundary conditions on the square domain of size @xmath4 , using @xmath30 grid points . at each iteration
the fields are dealiased by spherical truncation following the @xmath31 rule .
the penalization parameter @xmath23 , corresponding to the permeability of the solid domain , is taken equal to @xmath32 , a value validated by a systematic study of the sensitivity of the results to this parameter @xcite .
the fluid viscosity @xmath8 and magnetic diffusivity @xmath9 were taken equal to @xmath32 , the timestep @xmath33 equals to @xmath34 .
the initial kinetic and magnetic reynolds number are defined as @xmath35 and @xmath36 , where @xmath37 is the radius of the domain and , @xmath38 and @xmath39 are the kinetic and magnetic energies , respectively ( see table [ init_val ] ) .
both vorticity and current density fields are initialized with gaussian random initial conditions . therefore
, their fourier transforms @xmath40 and @xmath41 , where @xmath42 , are initialized with random phases and , their amplitudes give the energy spectra : @xmath43 with @xmath44 and , where @xmath45 and @xmath46 .
this energy spectrum follows a power law proportional to @xmath47 at large wavenumbers and was chosen to compare with simulations performed in the periodic case .
both fields are statistically identical .
the corresponding fields @xmath25 and @xmath27 are calculated from @xmath48 and @xmath49 using the biot - savart law . for vanishing viscosity and resistivity ,
two - dimensional mhd has three conserved invariants .
the total energy is @xmath50 , defined as sum of the kinetic energy @xmath38 and the magnetic energy @xmath39 : @xmath51 @xmath52 is the cross helicity : @xmath53 which mesures the global correlations between @xmath25 and @xmath27 and @xmath54 is the integral of the squared vector potential : @xmath55 as was shown by ting @xcite for periodic boundary conditions , the dynamics of decaying mhd turbulence depend strongly on the initial values of these invariants . because of its interest for the present study we recall briefly the four distinct decay regimes discerned by ting @xcite depending on the initial values and ratios of the invariants .
first , in the case of small initial @xmath52 and @xmath56 , a magnetically dominated regime is obtained .
selective decay is observed in this regime which corresponds to the decay of @xmath38 relative to @xmath39 .
second , in the case of vanishingly small initial magnetic energy , the lorentz force acting in the vorticity equation can not become strong enough , so that the vector potential is advected like a passive scalar .
following biskamp and welter @xcite , the magnetic field may however be amplified even if the initial ratio @xmath57 is very small , given that @xmath9 is sufficiently small .
they found that @xmath58 is necessary such that the magnetic field can be intensified .
this is a regime which essentially corresponds to the navier - stokes limit .
third , in the case of substantial initial cross - helicity , the turbulence tends towards an alfvnic state in which @xmath59 and @xmath21 are aligned or anti - aligned and approximately equipartitioned .
this process is called dynamic alignment and the ratio @xmath60 tends to two .
this is a state free from nonlinear interactions , inhibiting cascade processes ( even though this depletion of nonlinearity is rather slow with increasing cross - helicity @xcite ) .
the fourth and final regime is an erratic regime which might tend to different final states and which could be related to various competing subregions with unequal sign of cross - helicity .
this regime can be found if the flow is initialized with small cross - helicity and comparable kinetic and magnetic energies .
whether these regimes persist in the presence of solid boundaries is one of the main questions we want to answer in the present work . to obtain the desired initial conditions corresponding to the four regimes we proceed as follows :
starting from random initial conditions in fourier space , we renormalize @xmath59 and @xmath21 in physical space by varying the coefficient @xmath61 : @xmath62 this generally yields initial conditions with vanishingly small cross - helicity , and initial conditions for regime i , ii and iv can hereby be created . in the case of regime iii ,
a non - zero cross - helicity needs to be imposed .
we achieve this by creating a random initial condition for @xmath59 and a perpendicular field @xmath63 by rotating @xmath59 by @xmath64 . the magnetic field
is then obtained by a linear combination of the 2 fields : @xmath65 hereby any given cross helicity can be imposed .
table [ init_val ] summarizes the initial values of @xmath66 , @xmath67 and @xmath52 for the four different regimes , together with the reynolds number .
.[init_val]initial values of the four different regimes .
[ cols="<,^,^,^,^,^",options="header " , ]
center : ratio of total energy and integral of the squared vector potential , @xmath66 .
bottom : ratio of total energy and magnetic helicity , @xmath60 .
[ e_sur_a],title="fig : " ] .
center : ratio of total energy and integral of the squared vector potential , @xmath66 .
bottom : ratio of total energy and magnetic helicity , @xmath60 .
[ e_sur_a],title="fig : " ] .
center : ratio of total energy and integral of the squared vector potential , @xmath66 .
bottom : ratio of total energy and magnetic helicity , @xmath60 .
[ e_sur_a],title="fig : " ] in a periodic domain , starting from similar initial conditions as in figure [ e_sur_a ] .
[ energy_per ] ] in figure [ e_sur_a ] we show the time evolution of several integral quantities for the 4 different sets of initial conditions .
the main observation is that the 4 different regimes , discerned by ting @xcite are robust enough to survive within a bounded domain .
we now discuss the results in more detail . regarding the ratio of kinetic and magnetic energy ( fig .
[ e_sur_a ] , top ) , it is observed that in the absence of initial cross - helicity ( case i , ii and iv ) the magnetic energy finally dominates , unless it is very small initially ( navier stokes limit ) .
however , if the initial cross - helicity is initially large and @xmath68 is of order unity , the flow energy will remain approximately equipartitioned between the velocity and magnetic field . this picture is confirmed by the time evolution of the ratio @xmath66 ( fig .
[ e_sur_a ] , center ) . in this representation
it is however emphasized that in the navier - stokes limit ( case ii ) , the character of the magnetic field has changed : in the ideal system ( vanishing viscosity and magnetic diffusivity ) , @xmath54 is a quantity that cascades towards the small wavenumbers . in
a non - ideal system an inverse cascade generally slows down the dissipation rate of the quantity .
however , in the limit of small lorentz force , the equations of the vorticity and vector potential become equivalent to the equations that describe a passive scalar advected by a two - dimensional velocity field .
the passive scalar being a quantity which cascades towards higher wavenumbers , the vector potential gets dissipated faster in this case than in the case where the lorentz force is significant .
this results in a rapid increase of the quantity @xmath66 in case ii . the ratio @xmath60 ( fig .
[ e_sur_a ] , bottom ) attains its minimum absolute value @xmath69 for case iii .
this corresponds to dynamic alignment : the velocity field is equal in magnitude and perfectly aligned , or anti - aligned with the magnetic field .
the erratic regime is clearly represented by case iv , in which the cross - helicity approaches a value close to zero .
as we will see in the following , this is caused by different subregions with oppositely valued @xmath52 . for comparison ,
we show in figure [ energy_per ] @xmath68 in a periodic domain , starting from similar initial conditions as in the bounded case using the same numerical parameters . even though the trend is similar
, we see that a more oscillatory behavior for case i and ii is observed than in the case of the bounded domain .
this oscillatory behavior is related to energy exchange between the magnetic field and the velocity field by means of alfvn waves @xcite .
whereas in a periodic domain these waves can freely propagate , in a bounded domain they might be more rapidly suppressed , explaining the less oscillatory behavior of @xmath67 in a bounded domain .
further research is needed to clarify this .
the quantity @xmath70 gives a measure for the dynamic alignment , which corresponds to measuring both the equipartitioning of energy and the alignment properties .
if we are exclusively interested in the alignment properties , the relative cross helicity , which corresponds to the cosine of the angle @xmath71 between the velocity and magnetic field vector , @xmath72 should be considered . at @xmath73 , @xmath74 and @xmath75 in the regime iv.[pdf_theta ] ] at @xmath73 , @xmath74 and @xmath75 in the regime iv.[pdf_theta ] ] in figure [ cost ] , @xmath76 is plotted as a function of time .
it can be observed that in case i and iii , the velocity field tends to a nearly aligned state . in case ii and iv ,
this quantity remains close to zero , however for a different reason . in case
ii , the alignment is small , because the vector potential is advected as a nearly passive scalar . in case
iv the local alignment is large but different aligned or anti - aligned regions cancel out the contributions , yielding a net - global alignment close to @xmath77
. this can be observed in the corresponding probability distribution function of @xmath76 at @xmath73 and @xmath74 , shown in figure [ pdf_theta ] .
nevertheless , for long time ( @xmath75 ) we observe an anti - alignment .
the decay of total energy is shown in figure [ eloglin ] . at intermediate times , the energy in case i and ii decays following a powerlaw with exponents varying for the different sets of initial conditions ( fig .
[ eloglin ] , top ) .
the exponents of these powerlaws are approximately @xmath78 ( dotted line ) for regime i and @xmath79 ( solid line ) for regime iv .
it is seen that these powerlaws are observed only after an initial period of rapid decay . in the other cases
no clear powerlaw behavior can be identified .
this can be compared to previous studies @xcite in which values around @xmath80 and @xmath81 were found for the decaying periodic case . in case iii , in which dynamic alignment is observed , no clear power - law behavior is observed . in this case
the nonlinear interactions are progressively damped by the alignment process , so that no selfsimilar period is observed in the energy decay . at late times
[ eloglin ] , bottom ) all cases show an exponential viscous decay of the form @xmath82 with @xmath83 in case i and @xmath84 in cases ii , iii and iv , a value related to the largest stokes eigenmode of the circle ( @xmath85 ) , which contains most of the energy , as found in @xcite for the hydrodynamical case .
and the dotted line ( top ) corresponds to @xmath86 .
[ eloglin],title="fig : " ] and the dotted line ( top ) corresponds to @xmath86 .
[ eloglin],title="fig : " ] figures [ visuvort ] and [ visucurr ] show the vorticity and the current density field , respectively . for each of the cases i - iv , three typical time instants are visualized .
these instants are @xmath87 , showing the self - organization of the flow at early times , @xmath73 , when nonlinear processes are dominating and @xmath88 ( regime i ) and @xmath75 ( regimes ii , iii and iv ) , corresponding to the final , viscously decaying state .
+ + + + + + one flagrant feature of the visualizations is the local alignment of the magnetic and velocity field . indeed in most regimes
the vorticity and current density fields are rather similar .
we also observe the coincidence of the maxima of @xmath48 and of @xmath49 which may have some effect on the stabilization of vorticity and current filaments . in case
i an almost perfect axi - symmetrical state is achieved at @xmath88 .
case ii is the only case in which the formation of circular vortices is well pronounced , leading to a roll up of the current sheets .
apparently in the other regimes the lorentz force suppresses the generation of circular vortices .
case iii shows almost identical magnetic and velocity fields , as expected in this case of dynamic alignment , in which @xmath59 and @xmath21 are aligned ( or anti - aligned ) and in which kinetic and magnetic energies are in equipartition .
case iv is a typical example of the erratic regime : at the intermediate time , four dominant flow stuctures are observed , with both positive and negative cross - helicity . locally the flow is close to an aligned or anti - aligned state , but globally the cross - helicity is weak because the different regions with opposite contributions cancel each other out . a supplementary information on the final states is given by scatter - plots .
it was shown by joyce and montgomery @xcite that in hydrodynamic unbounded two - dimensional flows a long lasting final state is reached , depleted from nonlinearity .
this state is characterized by a functional relation between the vorticity and the streamfunction of the form @xmath89 . that a functional relation leads to a state , depleted from nonlinearity
is easily shown from the equation for the vorticity : @xmath90,\end{aligned}\ ] ] with the poisson bracket defined as @xmath91=(\partial a/\partial x)(\partial b/\partial y)-(\partial a/\partial y)(\partial b/\partial x)$ ] . a functional relation
@xmath92 leads to a vanishing poisson bracket . if we consider now the equations for incompressible mhd : @xmath93-[a , j]\\ ( \partial_t-\eta\delta)a&=&[a,\psi],\end{aligned}\ ] ] we see that two nonlinearities play a role : @xmath94 $ ] and @xmath95 $ ] .
the term @xmath96 $ ] can be considered as a pseudo - nonlinearity if @xmath14 is regarded as given . although important theoretical progress has been made in the comprehension of final states @xcite no analytical nontrivial solution is presently known for the case of decaying mhd turbulence .
it was however shown in kinney @xcite that close to functional relations do exist in homogeneous two - dimensional mhd turbulence . in figure [ scatter1 ] we show for the cases i - iv these scatter plots corresponding to the three nonlinearities . in case i we see a well defined nonlinear functional relation @xmath97 .
clearly , we have a non trivial final state .
the plot @xmath98 vs. @xmath49 shows a straight line , which corresponds to a vanishing lorentz - force : the magnetic field does not interact with the velocity field at this final period of decay .
the plot @xmath98 vs. @xmath14 also shows a clear functional relation . in case
ii , the scatter plots do not show such clear functional relations which is due to the fact that the flow is not yet sufficiently relaxed .
the plot @xmath48 vs. @xmath14 is perhaps closest to a functional relation . in case
iii we see as expected a vanishing nonlinearity : for dynamic alignment it can be shown that nonlinearities vanish in the perfectly aligned case , when the equations are stated in elssser variables ( see for example @xcite ) . in case iv it
is expected that eventually the same behavior is observed as in case i. if the initial reynolds number is initially too low this behavior will however not be observed .
preliminary computations were performed at lower resolution , which showed that non - trivial final states are only observed if the initial reynolds number is sufficiently high .
otherwise linear relations are obtained for all different scatter plots .
+ + +
we have investigated the influence of non - periodic boundary conditions on decaying two - dimensional magnetohydrodynamic turbulence .
the use of a penalization method in combination with a classical fourier pseudo - spectral method allows for efficient resolution of mhd flows in bounded domains .
a main result is the observation of the robustness of the four different regimes discerned by ting @xcite . the same trends are found as in their pioneering work , depending on the initial values of the kinetic energy , magnetic energy , vector potential and cross - helicity .
a detailed description was given of the relaxation - process which leads to the final states .
in the case of a magnetically dominant , cross - helicity free case , a clear nontrivial functional relation was observed describing the magnetic and velocity fields .
functional relationships were also observed in regimes iii and iv , while in regime ii this functional relation was less clear .
+ future work will address the influence of other types of boundary conditions for the magnetic field and also other geometries will be studied .
+ * acknowledgments : * we gratefully acknowledge professor david montgomery for his valuable remarks and we acknowledge financial support from the agence nationale de la recherche , project m2tfp " . | the effect of non periodic boundary conditions on decaying two - dimensional magnetohydrodynamic turbulence is investigated .
we consider a circular domain with no - slip boundary conditions for the velocity and where the normal component of the magnetic field vanishes at the wall .
different flow regimes are obtained by starting from random initial velocity and magnetic fields with varying integral quantities .
these regimes , equivalent to the ones observed by ting , matthaeus and montgomery [ phys .
fluids * 29 * , 3261 , ( 1986 ) ] in periodic domains , are found to subsist in confined domains .
we examine the effect of solid boundaries on the energy decay and alignment properties .
the final states are characterized by functional relationships between velocity and magnetic field . |
one of the attractive solutions of the hierarchy problem is the consideration of our 4-dimensional universe as a sub - manifold that is embedded in a higher dimensional space - time .
the models of type add ( arkani - hamed , dimopoulos and dvali ) @xcite allow compactification radius of extra dimensions to be of macroscopic size @xcite .
black holes whose horizon radius is much smaller than this characteristic length can be well described by solution for higher - dimensional static black hole @xcite .
such black holes could be produced at the next - generation particle colliders , probably at energies of very low order @xmath5 @xcite .
thus study of their properties is strongly motivated .
the very important black hole characteristic is its quasi - normal ( qn ) spectrum : the set of complex frequencies ( see @xcite for review ) .
the real parts of the qn modes are the frequency of oscillations , while the imaginary parts are their damping rates .
these oscillations dominate on the intermediately late times of a perturbation evolution near a black hole .
they do not depend on the initial perturbation , being defined by the black hole parameters only . while the qn spectrum of different objects in four - dimensional @xcite and higher dimensional @xcite cases was extensively studied for massless fields of different spin and for the perturbation of the metric itself , our knowledge about the massive field qn spectrum still has gaps .
massive scalar field qn modes were studied only for four - dimensional black holes within wkb approximation @xcite , and later with continued fraction method of leaver @xcite .
massive vector qnms were investigated for schwarschild-(anti ) de sitter black holes @xcite .
massive dirac qnms were studied in @xcite .
it was found that the behaviour of massive fields is quite different from that of the massless ones : qn spectrum of massive fields has infinitely long - living modes , called quasi - resonances @xcite .
the late time behaviour of the the massive fields in black hole background is the same as in pure minkowski space - time .
asymptotically late time behaviour of them does not depend on their spin , mass and multipole number @xcite .
also massive fields demonstrate the so - called superradiant instability @xcite , which is absent for massless fields .
this paper is dedicated to the study of the qn spectrum of the massive scalar field in a @xmath0-dimensional black hole background .
note , that the scalar field is widely considered as a good model for physical fields when the effect of their spin is negligibly small .
it appears also in cosmological models , and , as one of the fundamental states in string theories .
the scalar field with the mass term can be also interpreted as a self - interacting scalar field within regime of small perturbations @xcite . also , the scalar field gains large effective mass when considered in models with extra dimensions of rs - type @xcite .
we study the qn spectrum in this regime as well .
this paper is organized as follows : in sec .
[ sec.basic ] we present basic formulae and the method of numerical calculation of the quasi - normal modes , sec . [ sec.numerical ] is devoted to our numerical results and their comparison with those obtained in other works .
finally , in sec .
[ sec.conclusions ] we summarize the obtained results .
the metric of the schwarzshild black hole in @xmath0-dimensions has the form @xcite : @xmath6 where @xmath7 here we used the quantities @xmath8 after separation of angular and time variables , the radial part of the massive scalar field equation can be reduced to the form : @xmath9 where @xmath10 @xmath11 is the overtone number , @xmath3 is the field mass . by definition ,
the quasi - normal modes are eigenvalues of @xmath12 with the boundary conditions which correspond to the outgoing wave at spatial infinity and the ingoing wave at the black hole horizon . in 4-dimensional case @xcite
, these boundary conditions look like @xmath13 & r & \simeq c_- ( r - r_0)^{-\imo\omega } , \quad { \rm as } \quad r \rightarrow r_h\ , , \label{horizonbc4d}\end{aligned}\ ] ] where @xmath14 .
the sign of @xmath15 should be chosen in order to remain in the same complex quadrant as @xmath12 .
it turns out that the behaviour of the wave function in a higher dimensional background differs qualitatively from that in the 4-dimensional one . for @xmath16 the leading term of @xmath17 at infinity
is of order @xmath18 , that is why the appropriate boundary conditions for @xmath16 have the form : @xmath19 & r & \simeq c_- ( r - r_0)^{-\imo\omega/(d-3 ) } , \quad { \rm as } \quad r \rightarrow r_h\ , .
\label{horizonbc}\end{aligned}\ ] ] one should note , that these boundary conditions do not change the possibility of quasi - resonance existing ( see @xcite ) . to solve the equation ( [ radial ] ) , we use the continued fraction method @xcite .
equation ( [ radial ] ) has an irregular singularity at spatial infinity @xmath20 and @xmath21 regular singular points : @xmath22 and the roots of the equation @xmath23 , one of them is the horizon at @xmath24 .
thus the appropriate frobenius series have the following form : @xmath25 where @xmath26 is regular in the circle @xmath27 for @xmath28 . for higher
@xmath0 , some of the singularities appear in the unit circle and one has to continue the frobenius series through some midpoints .
this technique was recently developed in @xcite , but in the present work we were limited by @xmath28 . substituting ( [ singularity ] ) into ( [ radial ] )
, one can obtain a @xmath29-term recurrence relation for the coefficients @xmath30 @xmath31 each of the functions @xmath32 has an analytical form in terms of @xmath33 , @xmath34 and @xmath12 , for a specific dimensionality @xmath0 and each integer @xmath35 .
we now decrease the number of terms in the recurrence relation @xmath36 by one , i. e. we find the @xmath37 , which satisfy the equation @xmath38 for @xmath39 , we can rewrite the above expression as @xmath40 subtracting ( [ subsre ] ) from ( [ srcre ] ) we find the relation ( [ finre ] ) explicitly . thus we obtain : @xmath41 & & c_{j , i}^{(k)}(\omega ) = c_{j , i}^{(k+1)}(\omega)-\frac{c_{k , i}^{(k+1)}(\omega)\ , c_{j-1,i-1}^{(k)}(\omega)}{c_{k-1,i-1}^{(k)}(\omega}\ , .
\nonumber%\label{gausselimination}\end{aligned}\ ] ] this procedure is called _ gaussian eliminations _ , and allows us to determine the coefficients in the three - term recurrence relation @xmath42 & & c_{0,1}^{(3)}\,b_1+c_{1,1}^{(3)}\,b_0=0,\end{aligned}\ ] ] numerically for a given @xmath12 up to any finite @xmath34 .
the complexity of the procedure is _ linear _ with respect to @xmath34 and @xmath43 .
the requirement that the frobenius series be convergent at spatial infinity implies that @xmath44 which can be inverted @xmath45 times to give @xmath46 the equation ( [ invcf ] ) with _ infinite continued fraction _ on the right - hand side can be solved numerically by minimising the absolute value of the difference between the left- and right - hand sides .
the equation has an infinite number of roots ( corresponding to the qn spectrum ) , but for each @xmath45 , the most stable root is different . in general
, we have to use the @xmath45 times inverted equation to find the @xmath45-th qn mode . the requirement that the continued fraction be itself convergent allows us to limit its depth by some large value , always ensuring that an increase in this value does not change the final results within the desired precision .
it turns out , that the convergence of the infinite continued fraction becomes worse if the imaginary part of @xmath12 increases with respect to the real part .
this problem was circumvented by h .-
nollert @xcite . to improve the convergence of the infinite continued fraction
one can consider initial approximation for @xmath47 that for large @xmath45 can be expanded as @xmath48 since for the four - dimensional case the coefficients @xmath49 , @xmath50 and @xmath51 are known analytically , one can find @xmath52 from the equation @xmath53 the found expansion ( [ nollertexp ] ) could be used as an initial approximation for the `` remaining '' infinite continued fraction .
in order to ensure the convergence of ( [ nollertexp ] ) for a given value of @xmath12 , one has to start from the found approximation deeply enough inside the continued fraction ( [ cfeq ] ) .
it turns out , that the required depth is less than it would be if we start from some arbitrary value as the initial approximation . as mentioned in @xcite ,
quasi - resonances correspond to purely imaginary @xmath15 .
thus the convergence of the continued fraction decreases dramatically , and one needs to use the described method for calculation of those frequencies .
even though @xmath49 , @xmath50 and @xmath51 are not known analytically in the case under consideration , we are still able to find the expansion ( [ nollertexp ] ) directly from ( [ rrelation ] ) . after dividing the equation ( [ rrelation ] ) by @xmath54 and using the definition of @xmath55
, one can find @xmath56 the important is that the expression ( [ nollertexp ] ) should be defined without any indeterminations .
since the infinite continued fraction is convergent , it is clear that @xmath57 is itself well - defined .
our task is just to recover it from the equation ( [ nollerthd ] ) .
it turns out that only determination of @xmath58 and @xmath59 could cause ambiguities : the first appears due to a lot of roots of the equation ( [ nollerthd ] ) and the second is connected with two possible signs of @xmath60 in ( [ nollertexp ] ) .
fortunately , both ambiguities can be eliminated by using our knowledge about the properties of the series ( [ frobenius ] ) .
let us consider them more precisely . for large @xmath45 @xmath61
, thus @xmath58 satisfies @xmath62 the equation ( [ c0eq ] ) has in general @xmath63 roots ( in fact there are repeated roots ) .
one of the roots is _ always _
@xmath64 ( it is also repeated ) , implying the unit radius of convergence of the series ( [ frobenius ] ) .
other roots appear due to the existing of additional singular points of the equation ( [ radial ] ) .
thus we choose @xmath64 .
after fixing @xmath64 one can find that @xmath65 the sign of @xmath66 can be recovered using the convergence of the series ( [ frobenius ] ) at spatial infinity ( @xmath67 ) .
therefore : @xmath68 i. e. @xmath69 .
since for large @xmath45 @xmath70 we find out that the real part of @xmath66 can not be negative .
after the sign of @xmath66 is fixed , the other coefficients in ( [ nollertexp ] ) can be found without encountering indeterminations @xcite .
thus we have now possibility to use ( [ nollertexp ] ) as an initial approximation for @xmath57 when @xmath71 .
this allows us to improve the convergence of the infinite continued fraction near quasi - resonances .
let us start from an important notion .
since the black hole mass in @xmath0-dimensional space - time can not be measured in units of length , as in four - dimensional space - time , we measure all the values in units of the black hole horizon radius that is an observable quantity . in order to do this
we multiply ( [ radial ] ) by @xmath72 and consider dimensionless quantities @xmath73 and @xmath74 . making use of the technique described in the previous section ,
we have found dependence of the quasi - normal modes on the field mass . from the figure [ fig.l=0 ]
we see that for four- and five - dimensional cases increasing of @xmath73 gives rise to decreasing of the imaginary part of the fundamental quasi - normal frequency until reaching zero , while the real part changes insignificantly .
as discussed in @xcite , when some threshold value of @xmath73 is exceeded , the particular quasi - normal mode disappears .
the higher threshold value corresponds to the higher overtone number .
since we reached purely real frequency numerically with some finite accuracy , we can not state that purely real frequencies exist in the spectrum .
we are on position , that at least while @xmath73 tends to some threshold values , the quasi - normal ringing lifetime increases infinitely . for @xmath1
we observe that the imaginary part of the fundamental mode tends to zero asymptotically while the real part approaches @xmath3 .
we observed the same behaviour for different multipole numbers .
for @xmath75 it is shown on the figure [ fig.d=6 ] .
the picture is similar for higher dimensional cases , but since massless field quasi - normal frequency has higher imaginary part , one needs to reach higher @xmath73 to see this tendency .
even though the behaviour of the fundamental overtone is different for @xmath1 , the dependence on @xmath73 of higher overtones is qualitatively the same ( see figure [ fig.d=6.high ] ) as for @xmath76 and @xmath77 .
also because all the equations depend on @xmath78 we observe that the contribution of mass for higher overtones is of order @xmath79 as in four - dimensional case @xcite .
thus the asympotical behaviour of the high overtones does not depend on @xmath3 . from the figure [ fig.highqnms ]
one can see that for 7-dimensional case the difference between massless and massive @xmath80 cases is insignificant already for 20th overtone . according to the analytical formula @xcite
, the spectrum is asymptotically equidistant for all @xmath0 : @xmath81 and the real part approaches @xmath82 the difference in the mass dependence of the fundamental overtone for @xmath1 leads to another remarkable fact . since for high field mass the imaginary part of the fundamental overtone tends to zero only asymptotically ( see figure [ fig.l=0 ] ) , and the imaginary part of higher overtones reaches zero for some finite value of @xmath73 ( see figure [ fig.d=6 ] ) , there are some values of @xmath73 for which the imaginary parts of two overtones _ are the same_. after one of these values is reached the overtones can be distinguished only by their real part and the quasi - normal ringing has two dominant frequencies in its spectrum .
it turns out that this could happen only for relatively large field mass , thus the imaginary part of those frequencies is very small : for example , for the field configuration of spherical symmetry ( @xmath83 ) in @xmath75 , the smallest value of @xmath84 , for which there are dominant modes with @xmath85 and @xmath86 .
their imaginary parts are equivalently small , thus their lifetime is extremely high .
thus one could observe the superposition of that two frequencies at late times of the quasi - normal ringing .
one should note that this exotic behaviour was found for relatively large field mass for which the linear approximation ( i. e. without taking into consideration the back reaction of the field upon the black hole ) could not be good .
the fundamental modes of massless scalar field for different @xmath0 are presented in table [ fundamentalmodes ] .
we see that as @xmath0 increases , the real part of the fundamental overtone increases faster than the imaginary part .
thus fields that live in the bulk are better oscillators than brane - localised ones @xcite , considered in the context of large extra dimension scenario ( see @xcite for review ) .
.fundamental quasi - normal frequencies of massless scalar field ( measured in the units of the black hole horizon radius ) . [ cols="<,^,^,^",options="header " , ] we checked the correctness of our results by comparing them with those @xcite obtained with the help of the wkb method @xcite for higher values of the multipole number .
the wkb method is known to give very accurate results in this regime @xcite .
one should note also , that massless scalar field quasi - normal spectrum coincides with that of the metric perturbation of the tensor type for @xmath87 , which was recently studied @xcite .
our results were obtained independently and demonstrate complete agreement .
in the present paper we studied quasi - normal spectrum of the massive scalar field near the higher dimensional black hole within continued fraction method of leaver with improvement of nollert that was generalised for the case of @xmath43-term recurrence relation for frobenius series coefficients .
we found that the fundamental modes for @xmath1 have qualitatively different dependence on the field mass : their real part approaches the field mass while the imaginary part tends to zero for high values of @xmath73 . at the same time , the behaviour of higher modes is qualitatively the same as for @xmath76 and @xmath77 cases : the imaginary part decreases quickly and the particular mode disappears after reaching some threshold value of @xmath73 .
n. arkani - hamed , s. dimopoulos and g. r. dvali , phys .
b * 429 * , 263 ( 1998 ) [ hep - ph/9803315 ] ; phys .
d * 59 * , 086004 ( 1999 ) [ hep - ph/9807344 ] ; + i. antoniadis , n. arkani - hamed , s. dimopoulos and g. r. dvali , phys .
b * 436 * , 257 ( 1998 ) [ hep - ph/9804398 ] .
d. cremades , l. e. ibanez and f. marchesano , nucl .
* b * 643 ( 2002 ) 93 , [ hep - th/0205074 ] ; + c. kokorelis , nucl . phys .
* b * 677 ( 2004 ) 115 , [ hep - th/0207234 ] ; + e. floratos , c. kokorelis , [ hep - th/0607217 ] .
p. c. argyres , s. dimopoulos and j. march - russell , phys .
b * 441 * , 96 ( 1998 ) [ hep - th/9808138 ] ; + t. banks and w. fischler , [ hep - th/9906038 ] ; + s. b. giddings and s. thomas , phys .
d * 65 * , 056010 ( 2002 ) [ hep - ph/0106219 ] ; + s. dimopoulos and g. landsberg , phys .
lett . * 87 * , 161602 ( 2001 ) [ hep - ph/0106295 ] .
p. r. brady , c. m. chambers , w. g. laarakkers and e. poisson , phys .
d * 60 * , 064003 ( 1999 ) [ gr - qc/ 9902010 ] ; + b. wang , e. abdalla and r. b. mann , phys .
d * 65 * , 084006 ( 2002 ) [ hep - th/0107243 ] ; + v. suneeta , phys .
d * 68 * , 024020 ( 2003 ) [ gr - qc/ 0303114 ] ; + r. a. konoplya , phys .
d * 66 * , 044009 ( 2002 ) ; + a. zhidenko class .
* 23 * 3155 - 3164 ( 2006 ) [ gr - qc/0510039 ] ; +
j. x. tian , y. x. gui , g. h. guo , y. lv , s. h. zhang and w. wang , gen . rel .
grav . * 35 * , 1473 ( 2003 ) [ gr - qc/0304009 ] ; + c. molina , d. giugno , e. abdalla and a. saa , phys .
d * 69 * , 104013 ( 2004 ) [ gr - qc/0309079 ] ; + e. abdalla , r. a. konoplya , c. molina , phys .
d * 72 * , 084006 ( 2005 ) [ hep - th/0507100 ] ; + t. r. choudhury and t. padmanabhan , phys .
d * 69 * , 064033 ( 2004 ) [ gr - qc/0311064 ] ; + m. giammatteo , i. g. moss , class .
* 22 * , 1803 ( 2005 ) [ gr - qc/0502046 ] ; + v. cardoso , r. a. konoplya , j. p. s. lemos , phys .
d**68**,044024 ( 2003 ) ; + j. jing , phys . rev .
* d * 69 084009 ( 2004 ) [ gr - qc/0312079 ] ; phys . rev . *
d * 71 124006 ( 2005 ) [ gr - qc/0502023 ] ; jhep 0512 ( 2005 ) 005 [ gr - qc/0512015 ] ; + j. jing and q. pan , phys . rev . *
d * 71 124011 ( 2005 ) [ gr - qc/0502011 ] ; nucl . phys .
* b*728 109 - 120 ( 2005 ) [ gr - qc/0506098 ] ; + v. cardoso , j. natario and r. schiappa , j. math .
* 45 * , 4698 ( 2004 ) [ hep - th/0403132 ] ; + c .- g .
shao , b. wang , e. abdalla , r .- k .
. rev . *
d * 71 ( 2005 ) 044003 [ gr - qc/0410025 ] . g. t. horowitz and v. e. hubeny , phys .
d * 62 * , 024027 ( 2000 ) ; + a. nunez , a. o. starinets , phys .
d * 67 * , 124013 ( 2003 ) ; + v. cardoso , o. j. c. dias and j. p. s. lemos , phys . rev .
d * 67 * , 064026 ( 2003 ) ; + r. a. konoplya , phys .
d * 68 * , 124017 ( 2003 ) ; + c. molina , phys .
rev . * d68
* , 064007 ( 2003 ) ; + v. cardoso , s. yoshida , o. j. c. dias and j. p. s. lemos , phys . rev . * d68 * , 061503 ( 2003 ) ; + v. cardoso , j. p. s. lemos and s. yoshida , jhep * 0312 * , 041 ( 2003 ) ; phys .
d * 69 * , 044004 ( 2004 ) ; + g. siopsis , phys .
b590 * , 105 ( 2004 ) ; + v. cardoso , o. j. c. dias and j. p. s. lemos , phys .
d * 70 * , 024002 ( 2004 ) [ hep - th/0401192 ] ; + r.a .
konoplya , e. abdalla , phys .
d71 , 084015 ( 2005 ) [ hep - th/0503029 ] ; + l. vanzo and s. zerbini , phys .
d * 70 * , 044030 ( 2004 ) [ hep - th/0402103 ] ; + j. natario and r. schiappa , adv .
* 8 * , 1001 ( 2004 ) [ hep - th/0411267 ] ; + y. morisawa , d. ida , phys .
d * 71 044022 ( 2005 ) [ gr - qc/0412070 ] ; + r. g. daghigh , g. kunstatter , d. ostapchuk , v. bagnulo [ gr - qc/0604073 ] ; + a. lopez - ortega , [ gr - qc/0605027 ] ; [ gr - qc/0605034 ] .
h. t. cho , phys . rev .
* d * 68 024003 ( 2003 ) [ gr - qc/0303078 ] ; + j. jing , phys . rev . *
d * 70 065004 ( 2004 ) [ gr - qc/0405122 ] ; phys . rev . * d * 72 027501 ( 2005 ) [ gr - qc/0408090 ] ; + a. lpez - ortega , gen .
. grav . * 38 * 743 - 771 ( 2006 ) .
h. koyama , a. tomimatsu , phys . rev .
* d * 65 , 084031 ( 2002 ) ; phys . rev .
* d * 64 , 044014 ; + r. moderski and m. rogatko , phys . rev .
* d * 64 , 044024 ( 2001 ) ; + l. m. burko , g. khanna , phys . rev . * d * 70 , 044018 ( 2004 ) ; +
r. moderski , m. rogatko , phys . rev .
* d * 72 ( 2005 ) 044027 [ hep - th/0508175 ] ; + j. jing , phys . rev .
* d * 72 , 027501 ( 2005 ) ; + r. a. konoplya and c. molina , [ gr - qc/0602047 ] .
k. d. kokkotas , b. f. schutz , phys .
d * 37 * 3378 ( 1988 ) ; + v. ferrari , m. pauri , f. piazza , phys .
d * 63 * , 064009 ( 2001 ) ; + r. a. konoplya , gen .
rel . grav . *
34 * , 329 ( 2002 ) [ gr - qc/0109096 ] ; phys .
d * 66 * 084007 ( 2002 ) [ gr - qc/0207028 ] ; + s.
fernando , c. holbrook , [ hep - th/0501138 ] ; + a. zhidenko , class .
quant . grav .
* 21 * , 273 ( 2004 ) [ gr - qc/0307012 ] ; + r. a. konoplya and a. zhidenko , jhep * 0406 * , 037 ( 2004 ) [ hep - th/0402080 ] . | we study quasi - normal spectrum of the massive scalar field in the @xmath0-dimensional black hole background . we found the qualitatively different dependence of the fundamental modes on the field mass for @xmath1 .
the behaviour of higher modes is qualitatively the same for all @xmath0 .
thus for some particular values of @xmath2 , where @xmath3 is the field mass and @xmath4 is the black hole mass , the spectrum has two dominating oscillations with a very long lifetime .
we show that the asymptotically high overtones do not depend on the field mass .
in addition , we present the generalisation of the nollert improvement of the continued fraction technique for the numerical calculation of quasi - normal frequencies of @xmath0-dimensional black holes . |
the vast amount of data collected at the relativistic heavy ion collider at brookhaven national laboratory along with those recently collected from the collision of lead nuclei at large hadron collider at cern have led to the momentous discovery of quark gluon plasma ( qgp ) .
a large part of the effort in the coming years will be devoted to the determination of the precise values of the transport properties of the qgp . in this context , the energy loss suffered by quarks of different flavours and gluons as they traverse the qgp , undergoing collisions and radiating gluons , is a subject matter of considerable topical interest .
it is most simply demonstrated by a suppressed production of hadrons having large transverse momenta in nucleus - nucleus collisions when compared to appropriately scaled productions in @xmath3 collisions at the same centre of mass energy per nucleon .
it is often suggested that heavy quarks may lose a smaller amount of energy per unit length during their passage through qgp compared to light quarks , due to the dead cone effect @xcite .
the experimental results , however , show similar suppression for light and heavy mesons @xcite . a recent calculation by abir et al @xcite incorporating
the generalized distribution of gluons in @xmath4 and @xmath5 processes with a proper accounting of mass of the heavy quarks , leads to a @xmath6 for charm quarks which is quite similar to those for light quarks at energies @xmath7 10 - 15 gev .
the study of charm quark energy loss provides several unique advantages over those of light partons .
the charm mesons easily stand out in the multitude of light mesons .
most of the charm quarks are produced in initial fusion of quarks and anti - quarks ( @xmath8 ) and gluons ( @xmath9 ) , though a small additional production is expected @xcite from multiple scattering between jets , jets and thermalized partons , and thermalized partons .
it is not yet clearly established if the charm quarks thermalize in the qgp @xcite , though it is expected that due to their small numbers their impact on the bulk properties of the qgp would be negligible .
one also expects that due to their large mass , charm quarks will not change their direction as they traverse the plasma , though they will slow down , making them excellent probes of azimuthal dependence of the conditions of the interacting system . as charm quarks have large mass
, pqcd remains reasonably valid down to lowest @xmath2 .
there is one additional trait which should help us in getting flavour dependence of energy loss of heavy quarks . while a @xmath10 or a @xmath11 or a @xmath12 quark or a gluon can fragment into one of many mesons or baryons , a charm or a bottom quark would mostly fragment into only a @xmath13 or a @xmath14 meson or a charm or bottom baryon , respectively .
thus a charm quark after losing energy will appear as a d - meson having lower energy .
we shall see that this would lead to a characteristic enhancement of charm mesons or single electrons at low @xmath2 .
the paper is organized as follows .
2 contains expressions for calculations of nuclear modification factor- @xmath15 , azimuthal anisotropy coefficient- @xmath16 , charm production , and our model prescription for energy loss .
contains discussion of our results , followed by a summary in sec .
the energy loss of quarks and gluons is most easily seen via the suppressed production of hadrons measured using nuclear modification factor , @xmath15 : @xmath17 where @xmath18 is the hadron production for the nucleus - nucleus system at a given impact parameter , @xmath19 is the corresponding nuclear thickness , and @xmath20 is the cross - section for the production of hadrons at the corresponding centre of mass energy / nucleon in @xmath3 collisions
. one can now calculate @xmath21 /(1+\delta_{ij})~
, \label{sigma}\end{aligned}\ ] ] where @xmath22 and @xmath23 are the momenta and rapidities of produced charm and anti - charm and @xmath24 and @xmath25 are the fractions of the momenta carried by the partons from their interacting parent hadrons .
these are given by @xmath26 where @xmath27 is the transverse mass , @xmath28 , of the produced heavy quark .
the subscripts @xmath29 and @xmath30 denote the interacting partons , and @xmath31 are the partonic distribution functions for the nucleons .
the fundamental processes included for lo calculations are : @xmath32 we recall that the above lo pqcd expression reproduces the nlo results @xcite when supplemented with a @xmath33-factor @xmath34 2 ( see ref .
@xcite ) .
we have used @xmath19= 225 @xmath35 for 0 - 10@xmath36 centrality for au+au collisions at rhic , as calculated from glauber formalism . for
pb+pb collisions at lhc , @xmath19 = 195 @xmath35 for 0 - 20@xmath36 centrality has been used .
we use cteq5 m structure function along with eks98 @xcite shadowing function .
the factorization , renormalization , and fragmentation scales are chosen as @xmath37 and the charm quark mass has been taken as 1.5 gev .
we propose an empirical model for the energy loss for charm quarks which is inspired by a multiple scattering model used earlier by @xcite supplemented with considerations of baier et al @xcite for partonic energy loss @xcite .
a straight - forward empirical implementation of a similar energy loss procedure for light parton has been found to provide a satisfactory dependence of energy loss on the centralitiy of the collision @xcite .
we recall once again that the nuclear modification of heavy meson production is similar to those for light mesons .
we perform a monte carlo implementation of our model calculations and estimate the momentum loss of charm quarks and nuclear modification of d - meson and single electron production .
we assume that the energy loss of heavy quarks proceeds via multiple collisions and that the momentum loss per collision is given by , ( see for example ref .
@xcite ) @xmath38 so that one can write @xmath39 where @xmath1 and @xmath40 are parameters to be determined and @xmath41 is the mean free path of the charm quark , taken as 1 fm , in these initial studies . in what follows
, we shall consider charm quarks at central rapidities and thus @xmath42 .
thus the momentum of the charm quark after @xmath43 collisions will be given by @xmath44 the charm quark can continue to lose energy in collisions as long as the resulting momentum remains positive .
we estimate the probability for the charm quark to have @xmath43 collisions , while covering the path length @xmath45 from a poisson distribution @xmath46 taking a value for the coefficient @xmath1 and the exponent @xmath40 , we estimate the largest number of collisions- @xmath47 , which the charm quark having momentum @xmath2 can undergo .
next we sample the number of collisions @xmath43 , which the charm undergoes from the distribution @xmath48 to get the final momentum of the charm quark .
finally we fragment the charm quarks into d - mesons .
thus we have , @xmath49 where the fragmentation of the heavy quark @xmath50 into the heavy - meson @xmath51 is described by the fragmentation function @xmath13 .
we have assumed that the shape of @xmath52 , where @xmath53 , is identical for all the @xmath13-mesons , @xcite and @xmath54 ^ 2}~ , \label{frag}\ ] ] where @xmath55 is the peterson parameter and @xmath56 we have kept it fixed at @xmath55=0.13 .
here @xmath57 denotes semileptonic decay of @xmath13-mesons and the electron distribution is taken from ref .
@xcite .
non - central collisions of identical nuclei will lead to an oval overlap zone , whose length in and out of the reaction plain would be different .
thus , charm quarks traversing the qgp in and out of the plain will cover different path lengths and lose differing amount of energy .
this would lead to an azimuthal dependence in the distribution of resulting charm mesons , whose azimuthal anisotropy could be measured in terms of the @xmath16 coefficient defined by @xmath58 we have approximated the colliding nuclei as having a uniform density with radius @xmath59 in these calculations and obtained average path - length for the charm quarks along a given @xmath60 using eq .
9 of ref .
let us now discuss our results for nuclear modification factor and azimuthal anisotropy .
we consider three values for the exponent @xmath40 ; 0 , 0.5 , 1.0 appearing in eq .
[ deltapt ] , inspired by the three energy loss mechanisms , namely those applicable in the so - called bethe - heitler regime , lpm regime , and complete coherence regimes considered by baier et al .
@xcite which lead to energy loss per unit length as proportional to energy , square - root of the energy , and independent of the energy for light partons .
kampfer et al @xcite had earlier used this approach to study the effect of charm quark energy loss on the correlated charm decay .
next we vary @xmath1 to get a description of the @xmath15 for single electrons at rhic ( fig .
[ fig1 ] ) and for @xmath13-mesons at lhc ( fig .
[ fig2 ] ) . in fig .
[ fig1 ] we show our results for nuclear modification factor @xmath15 and azimuthal anisotropy @xmath16 for non photonic electrons at top rhic energy @xmath61 = 200 a gev @xcite . comparing the results of figs .
[ fig1 ] , we see that the model assuming momentum loss per collision as proportional to the momentum closely follows the shape of the experimentally determined @xmath15 for single electrons almost over the entire range of @xmath2 under consideration .
we add that our theoretical calculations have not included the @xmath62 contribution which is contained in the experimental results which can modify the @xmath15 for larger @xmath2 by up to 10% as the produced @xmath63 quarks
are much less in number and also lose much smaller energy @xcite .
the scenario , where @xmath64 , is only moderately successful in describing the data over a limited @xmath2 range of 24 gev/@xmath65 ( fig .
[ fig1 ] ) .
while the assumption of a constant momentum loss per collision may bracket the @xmath15 over the very limited range of 23 gev/@xmath65 , it does not follow the shape of the @xmath2 dependence ( fig .
[ fig1 ] ) .
the best values of the @xmath1 determined from the results in fig .
[ fig1 ] are used to estimate @xmath16 for the single electrons .
we see that our calculations provide a reasonable description of @xmath66 for @xmath67 2 gev/@xmath65 and overestimate the results for lower @xmath2 .
a relaxation of our assumption of a static medium at a constant temperature and a uniform density of the nuclei may improve this agreement .
similar results are obtained when we apply the model to the @xmath15 measured @xcite for the @xmath13-mesons at the lhc ( fig .
[ fig2 ] ) .
we again see that the model using @xmath68 provides a good description of the data over the entire @xmath2 range , while that using @xmath64 seems to describe the data for @xmath67 4 gev/@xmath65 .
the constant momentum transfer collision misses the shape of the @xmath2 distribution completely though it is able to bracket the numerical values over a very narrow @xmath2 range of 35 gev/@xmath65 ( fig .
[ fig2 ] ) .
the predictions for @xmath16 are given for a ready reference .
several factors could affect the value of the energy loss coefficient @xmath1 for a given mechanism .
we have verified that increasing ( decreasing ) the mean free path , @xmath41 , by 0.5 fm results in a decrease ( increase ) of the coefficient , @xmath69 such that @xmath70 remains unaltered .
we have kept @xmath71 fixed at 0.3 while estimating the initial charm distribution .
taking the renormalization scale as @xmath72 , with @xmath73= 1 or 2 leads to a decrease in the value of @xmath15 by 7 - 10 % , which can then be offset by decreasing @xmath69 by about 12 % .
it is interesting to recall that a focker - planck equation given by @xmath74~ , \label{focker}\ ] ] describes the evolution of the distribution @xmath75 , of charm quarks propagating in quark gluon plasma @xcite and losing energy due to multiple soft scatterings with light quarks and gluons .
this leads to a drag , @xmath76 , and a diffusion @xmath77 on the momentum of the charm quark . assuming that @xmath76 depends on momentum only , we have @xmath78 and the energy loss @xmath6 can be related to drag coefficient @xmath79 by @xmath80 where @xmath81 is the energy of the charm quark , and @xmath82 its momentum . considering the average temperature of the plasma attained at rhic as @xmath34 220 mev @xcite
, we can read the drag coefficient from fig .
3(a ) of ref .
@xcite as @xmath34 0.02 @xmath83 for @xmath2 up to 5 gev/@xmath65 .
we can re - write the above equation as @xmath84 comparing this with one of the ansatzes used for energy loss per collision in the present work , namely @xmath85 , we can write that @xmath86 thus the effective drag coefficient @xmath87 , can be written as @xmath88 which reduces to @xmath70 for large values of @xmath82 we thus note that the effective drag at rhic energies is about 0.04 @xmath83 compared to 0.02 @xmath83 estimated for soft multiple collisions by authors of ref .
thus we conclude that at rhic energies only half of the energy loss could be due to collisions , while the other half could attributed to radiations of gluons . similarly estimating the average temperature at lhc as about 270 mev and using the results for drag due to collisions as @xmath340.04 @xmath83 from ref .
@xcite at high momentum , we note that the collisions account for only one - third of energy loss at 2.76 tev / nucleon .
it is of interest to compare our results with other studies on medium modification of charm propagation reported in the literature .
thus moore and teaney @xcite have calculated the diffusion @xmath13 and drag coefficient @xmath89 , ( denoted by @xmath90 here ) using lo pqcd as @xmath91 and @xmath92 taking @xmath930.3 , this provides @xmath940.06 @xmath83 at 220 mev and @xmath940.09 @xmath83 at 270 mev , which are larger than our values at rhic and smaller than those at 2.76 tev / nucleon at lhc .
recall however , the results of bass et al . @xcite that a description of medium modification as well as @xmath16 for single electrons at rhic brackets the diffusion coefficient @xmath13 between @xmath95 and @xmath96 , when flow and contributions of bottom electrons is included .
this provides a large value for the drag coefficient between 0.17 and 0.68 @xmath83 . at first these large values
may look surprising .
however from eq .
[ dmoore ] we see that these would correspond to values of @xmath71 between @xmath341.0 and @xmath340.5 , which are rather large and make the use of perturbative qcd questionable . in this connection the studies of gossiaux
et al @xcite , are also of considerable interest which suggest that the collisional energy loss could be substantially larger if the debye mass is replaced by hard thermal loop calculation and a running coupling constant is used .
we further recall the work of authors of ref .
@xcite , which suggests that a considerable drag could be produced by the resonant heavy quark - light quark interaction , beyond that determined by lo pqcd interactions .
we have already mentioned that @xmath65 quarks will materialize mostly as @xmath13-mesons .
this can have an interesting consequence which is already apparent in the figs .
[ fig1 ] and [ fig2 ] .
the @xmath65 quarks after losing energy will pile up at lower energies and this would result in a characteristic increase in the production of @xmath13-mesons as well as single electrons having lower transverse momenta ( fig . [ fig3 ] ) .
we note that while there is suppression by almost a factor of 24 depending on the incident energy ( for single electrons ) at larger @xmath2 , there is essentially no suppression at lower @xmath2 at rhic and even an increase by a few percent at lhc where the energy loss is higher and the momentum spectra of @xmath65 quarks have less steeper slopes .
the increase in the case of @xmath13-mesons at lhc is rather spectacular .
this is different from the normal enhancement of mesons having low @xmath2 due to the so - called cronin effect , which is expected to be less important at higher incident energies and heavy mesons ( see e.g. , ref .
we recall that a similar enhancement in d - meson production at low @xmath2 has been predicted by considering the drag suffered the heavy quarks in the plasma and during their hadronization @xcite . in a forthcoming paper
we shall show that this can lead to a slight increase in the production of low mass dileptons due to correlated charm decay @xcite . at the very minimum
the present work describes a simple procedure to implement energy loss of heavy quarks in relativistic collision of heavy nuclei .
it will be of interest to explore the energy and centrality dependence of the momentum loss coefficient @xmath1 .
we have used a simple model where charm quarks traversing qgp lose energy in multiple collisions . exploring different parametric forms of energy loss we find that the form where the momentum loss per collision is proportional to the momentum gives a good description of nuclear modification of single electron production at rhic and d - meson production at lhc . comparing our results with drag coefficients due to collisional energy loss calculated earlier , we find that only a part of the energy loss could be due to collisions .
we note a characteristic increase in production of single electrons and d - mesons having low momenta due to energy loss of charms having larger energy .
one of us ( my ) acknowledges financial support of the department of atomic energy , government of india during the course of these studies .
110 a. adare _ et al . _
[ phenix collaboration ] , phys .
rev . lett . * 98 * , 172301 ( 2007 ) ; b. i. abelev _ et al .
lett . * 98 * , 192301 ( 2007 ) ; b. abelev _ et al . _
[ alice collaboration ] , arxiv:1203.2160 [ nucl - ex ] , 2012 . | we propose an empirical model to determine the form of energy loss of charm quarks due to multiple scatterings in quark gluon plasma by demanding a good description of production of d mesons and non - photonic electrons in relativistic collision of heavy nuclei at rhic and lhc energies .
best results are obtained when we approximate the momentum loss per collision @xmath0 , where @xmath1 is a constant depending on the centrality and the centre of mass energy .
comparing our results with those obtained earlier for drag coefficients estimated using langevin equation for heavy quarks we find that up to half of the energy loss of charm quarks at top rhic energy could be due to collisions while that at lhc energy at 2760 gev / a the collisional energy loss could be about one third of the total .
estimates are obtained for azimuthal anisotropy in momentum spectra of heavy mesons , due to this energy loss .
we further suggest that energy loss of charm quarks may lead to an enhanced production of d - mesons and single electrons at low @xmath2 in aa collisions .
key - words : charm quark , d - mesons , non - photonic electrons , qgp , nuclear modification , azimuthal anisotropy , relativistic heavy ion collisions pacs indices : 14.65.dw , 14.40.lb , 13.30.ce , 12.38.mh , 25.75.dw |
james clerk maxwell s early work on the kinetic theory of gases was a major step - stone in the introduction of probabilistic methods into physics .
proposition iv of maxwell s @xcite _ illustrations of the dynamical theory of gases _ , his first derivation of the velocity distribution law , is frequently cited as `` one of the most important passages in physics '' @xcite and as such had been thoroughly investigated in the works of s. g. brush @xcite , m. c. dias @xcite , c. w. f. everitt @xcite , e. garber @xcite , c. c. gillispie @xcite , p. m. heimann @xcite , t. m. porter @xcite , c. truesdell @xcite , j. uffink @xcite and other historians of science .
proposition iv shows that , given certain assumptions , the three velocity components of molecules of a box of gas follow the normal distribution and the speed ( the magnitude of velocity ) follows what nowadays is called the maxwell - boltzmann distribution .
this paper provides a new conceptual and historical analysis of maxwell s derivation .
we make four contributions to the literature .
first , the paper sheds new light on the logical structure of proposition iv and gives the first detailed analysis of how maxwell attempted to physically justify the three main mathematical assumptions on which it rests .
this allows us to show , second , that contrary to the common historical wisdom molecular collisions did play an essential role in establishing the conclusion of proposition iv : one of its three mathematical assumptions requires a prior lemma showing that collisions among particles of the same mass bring about an equilibrium velocity distribution , and we argue that maxwell indeed made an attempt to lend credence to such an approach to equilibrium on the basis of his propositions i iii .
to substantiate this claim we take a look at maxwell s remarks preceding proposition iv as well as at his attempt to arrive at a proof of tendency towards equilibrium in his discredited and widely ignored proposition vi .
we present a surprisingly simple proof of tendency towards equilibrium that rests on propositions i iii , both in the relevant special case when all particles have the same mass and in the general case when masses differ .
although this proof of equalization of temperatures is interesting in its own right and is not known by experts , we argue that it is not novel since its general case is nothing but a charitable reconstruction of maxwell s proposition vi . since the charitable reading of proposition vi provides a proof of tendency towards equilibrium in the general case , and since the special case of this proof is really simple , it is reasonable to assume that maxwell s brief remarks preceding proposition iv are at least partially motivated by this special case . additionally , if the charitable reading is tenable then maxwell preceded ludwig boltzmann by at least 6 years in providing a mechanical argument for tendency towards equilibrium , and this priority of maxwell should be more widely recognized .
fourth , the paper contributes to the scholarly discussion of the crucial and arguably unjustified probabilistic independence assumption of proposition iv .
the conflation of different interpretations of probability have already been pointed out in the literature as a potential source of maxwell s mistake in accepting this independence assumption ; besides furthering this analysis we also provide a novel interpretation according to which maxwell , instead of conflating different interpretations of probability , might have conflated different notions of independence .
we distinguish two notions
probabilistic independence and value ( or parameter ) independence , argue that their difference was not clear in the relevant works of clausius and herschel , and point out that in fact herschel mistakenly emphasized that it is the satisfaction of value independence that is crucial for the proof that maxwell allegedly have adapted as his proposition iv . thus if maxwell indeed adapted the proof from herschel s review article then he also likely to have accepted herschel s assessment of the assumption on which `` the whole force of the proof turns . ''
this in turn could explain why maxwell was content with the physical justification of the independence assumption since value independence of the velocity components would have appeared immediately clear .
although there seems to be no conclusive evidence deciding which of these two conflations were committed by maxwell ( maybe both ) , we point out that in his prior and contemporaneous work maxwell means value independence when he is discussing `` independence '' . .linear structure of the beginning of the _ illustrations of the dynamical theory of gases ( @xcite ) _ with the sections where they are analyzed in this paper . [ cols="<,<,<",options="header " , ] the paper is organized as follows .
section [ sect_proposition_iv ] starts with a reconstruction of maxwell s proof of proposition iv and identifies the three mathematical assumptions ( of existence , independence , and symmetry ) that are required for the proof to work .
we work backwards from here , first historically , then conceptually .
we recall maxwell s own wording of his proof of proposition iv and contrast it with a proof of herschel that has been identified by historians as its likely source .
section [ sect_historicaltreatment ] also briefly reviews the existing historical literature regarding maxwell s attempt to justify the mathematical assumptions of proposition iv .
sect_propi - iii ] reviews the particle collision model of maxwell s proposition i iii and identifies a crucial assumption ( condition m ) that is needed for proposition ii to work .
section [ sect_propvi ] shows how proposition i iii and condition m can be used to argue for the equalization of mean kinetic energies and contrasts the argument with maxwell s proposition vi .
section [ sect_twonotionsofindependence ] distinguishes between two independence notions and calls attention to the conflation of these two notions in the works of herschel , clausius , and maxwell .
section [ sect_justification ] puts the pieces back together by analyzing the extent to which the three aforementioned mathematical assumptions of proposition iv that of existence ( section [ subsect_existence ] ) , independence ( section [ subsect_independence ] ) , and symmetry ( section [ subsect_rotationalsymmetry ] ) could be physically justified by maxwell s collision model .
as we know from a 1859 letter to george gabriel stokes @xcite , maxwell s interest in the kinetic theory of gases was aroused by the papers of rudolf clausius .
clausius was mainly interested in explaining heat in terms of molecular motion , and in his 1857 article he used an elastic sphere model to establish a connection between the average kinetic energy and the pressure of the gas .
this treatment met the criticism of a dutch meteorologist , c. h. d. buys
ballot @xcite : if , as calculations suggest , the molecules of a gas move at speeds of several hundreds meters per second , odors released in one corner of a room should almost instantly be noticed in the other corner . to answer the objection , clausius @xcite attempted to show that repeated collisions prevent molecules from traveling for great distances in straight lines .
these considerations led him to introduce the notion of the mean free path ( for further historical details see @xcite ) .
maxwell developed the approach further in his _ illustrations of the dynamical theory of gases _ @xcite .
the main goal of herapath , waterston , krnig and clausius was to establish the kinetic theory as a bridge between thermodynamics and atomic theory .
however , the focus of the _ illustrations _ is on problems of viscosity , diffusion and heat conduction : maxwell attempted to treat these as special cases of a general process in which momentum or energy is transported by molecular motion . to investigate these transport properties he relied on results he derived about the velocity distribution of a gas in his proposition iv .
proposition iv aims to show that in a box of gas the number of particles whose velocity component in a particular direction lies between @xmath0 and @xmath1 is proportional to @xmath2 and that the number of particles whose speed lies between @xmath3 and @xmath4 is proportional to @xmath5 , where @xmath6 is a parameter which gets determined later .
in essence the proof relies on the fact that the only solution of the functional equation @xmath7 is the gaussian @xmath8 equation ( [ functionalequation ] ) arises from three assumptions : that * a stationary velocity distribution exists ; * the components of velocity in an orthogonal coordinate system are independent ; * the velocity distribution only depends on the magnitude of the velocity . if we denote the stationary velocity distribution of ( a1 ) with @xmath9 and if by `` independence '' we mean probabilistic independence ( treating the velocity components as random variables ) then ( a2 ) translates to @xmath10 and ( a3 ) translates to rotational symmetry of @xmath11 , namely that there exists a function @xmath12 for which @xmath13 . noting that @xmath14 , and @xmath15 are distributions it is easy to show that @xmath16 and that @xmath17 , leading to equation ( [ functionalequation ] ) .
let us quote maxwell s own proof ( with a minor change of notation ) .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ prop .
iv . to find the average number of particles whose velocities lie between given limits , after a great number of collisions among a great number of equal particles .
let @xmath18 be the whole number of particles .
let @xmath0 , @xmath19 , @xmath20 be the components of the velocity of each particle in three rectangular directions , and let the number of particles for which @xmath0 lies between @xmath0 and @xmath1 , be @xmath21 , where @xmath22 is a function of @xmath0 to be determined
. the number of particles for which @xmath19 lies between @xmath19 and @xmath23 will be @xmath24 ; and the number for which @xmath20 lies between @xmath20 and @xmath25 will be @xmath26 , where @xmath27 always stands for the same function . now
the existence of the velocity @xmath0 does not in any way affect that of the velocities @xmath19 or @xmath20 , since these are all at right angles to each other and independent , so that the number of particles whose velocity lies between @xmath0 and @xmath1 , and also between @xmath19 and @xmath23 , and also between @xmath20 and @xmath28 , is @xmath29 if we suppose the @xmath18 particles to start from the origin at the same instant , then this will be the number in the element of volume @xmath30 after unit of time , and the number referred to unit of volume will be @xmath31 but the directions of the coordinates are perfectly arbitrary , and therefore this number must depend on the distance from the origin alone , that is @xmath32 solving this functional equation , we find @xmath33 @xcite _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ by applying the fact that the total number of particles is finite and fixing the constants , @xmath22 turns out to be the normal distribution , @xmath34 .
thus , maxwell concludes , the number of particles whose velocity in a given direction lies between @xmath0 and @xmath1 is @xmath35 and the number whose speed lies between @xmath3 and @xmath4 is @xmath36 according to proposition v we can obtain similar results for the distribution of relative velocities of pairs of randomly chosen particles : the components of relative velocities also follow the normal distribution .
proposition vi intends to show that collisions lead to the equalization of the mean kinetic energies of systems of particles of different masses . from these results , mean values of various functions of velocities used in determining the viscosity and diffusion coefficients
could be expressed in terms of @xmath6 , the standard deviation of the velocities .
but it s not just the derivation of mean values that had significance , for many physical properties of gases depend on the actual distribution of molecular velocity .
although the qualitative idea that even at constant temperature different molecules have different velocities was often assumed in verbal formulations
e.g. in clausius @xcite explanation of evaporation equations ( [ xesdxkozott ] ) and ( [ vesdvkozott ] ) are the first quantitative , experimentally verifiable results of the kinetic theory of heat .
on the basis of indirect evidence , several authors @xcite have suggested that maxwell simply adapted the derivation of proposition iv from a work in statistics , most probably relying on a july 1850 _ edinburgh review _
article by sir william herschel on quetelet s collection of essays on probability @xcite .
the review was also reprinted in herschel s _ essays _ in 1857 , and we know from an early 1858 letter @xcite that maxwell read and `` liked '' these essays . in this 57 pages long , typical victorian - style review , herschel first gives his own analysis of the subject and praises quetelet s work on the derivation of the law of least squares .
but , to make the subject more palpable to non - mathematical readers , herschel comes up with a new , simple , general , and perfectly elementary proof of the principle in question , " based on the following three assumptions : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ we set out from three postulates .
1st , that the probability of a compound event , or of the concurrence of two or more independent simple events , is the product of the probabilities of its constituents considered singly ; 2nd , that there exists a relation or numerical law of connexion ( at present unknown ) between the amount of error committed in any numerical determination and the probability of committing it , such that the greater the error the less its probability , according to some regular law of progression , _ which must necessarily be general and apply alike to all cases , since the causes of error are supposed alike unknown in all ; _ and it is on this ignorance , and not upon any peculiarity on cases , that the idea of probability in the abstract is founded ; 3dly , that errors are equally probable if equal in numerical amount , whether in excess , or in defect of , or in any way beside the truth .
this latter postulate necessitates our assuming the function of probability to be what is called in mathematical language _
an even function , _ or a function of the square of the error , so as to be alike for positive and negative values ; and the postulate itself is nothing more than the expression of our state of _ complete _ ignorance of the causes of error , and their mode of action .
@xcite _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in the course of the derivation herschel adds that the independence assumption applies to deviations in a rectangular direction : the observed oblique deviation is equivalent to the two rectangular ones , supposed concurrent , and which are essentially independent from one another " @xcite . together with this later addition herschel s three main assumptions existence of a distribution of deviations , probabilistic independence of the deviations in a rectangular direction , and
spherical symmetry of the distribution are similar to those of maxwell .
herschel s derivation of the functional form , albeit in a verbalized form , proceeds in the same way as maxwell s , the main difference being that herschel considers two rather than three dimensions and that he shows no interest in calculating the distribution of the magnitude of the errors .
even if maxwell adopted herschel s derivation it is natural to ask whether maxwell tried to justify the mathematical assumptions of proposition iv . according to maxwell the conclusion of proposition iv
is intended to apply `` after a great number of collisions among a great number of equal particles '' takes place .
proposition iv of the _ illustrations _ is preceded by three other propositions that describe the nature and effects of particle collisions .
yet we find that the historical literature pays very little to no attention to these prior findings of maxwell ; we typically find half - sentence - remarks about the role propositions i iii played in the argumentation , with as many different half - sentence - remarks as there are historians making them . the only pertaining remark @xcite who explicitly deals with the history of statistics and probability theory that lies in the background of proposition iv makes is that maxwell s `` kinetic gas model [ ... ] was subject to a nearly infinite number of minute causes of deviation , and it displayed perfect statistical regularity , '' and this somehow `` led maxwell to the insight that the distribution of molecular velocities probably conformed to [ the error ] curve . '' it is rather unclear from this remark how maxwell is supposed to have been led to his insight or even just what the insight of maxwell was ; understood strictly porter s description is misleading since maxwell did not make a probabilistic statement about the conformity of the actual velocity distribution to the normal distribution .
@xcite mentions that maxwell s work
begins [ ... ] with consideration of the impact of two spheres " but he does nt attribute any importance to this part of maxwell s paper .
other historians generally assume that maxwell s analysis of molecular collisions has something to do with underpinning proposition iv , but we find no clear consensus about the motivations .
the longest treatment , referred to by many subsequent historical articles , comes from @xcite ( repeated almost verbatim in @xcite ) whose length still permits it to be quoted in full : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ maxwell ] began by pointing out that , if two elastic spheres collide , all directions of rebound are equally likely .
he apparently believed that this fact would ensure , not only that all directions of motion were equally probable in the gas , but also that the probability distribution for each component of the velocity was independent of the values of other components .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in other words brush suggests that maxwell s analysis of particle collisions was aimed at justifying the probabilistic independence assumption ( a2 ) and the rotational symmetry assumption ( a3 ) but he provides no details about how maxwell s justification is supposed to have proceeded .
it is also not clear what brush means by `` pointing out '' ; as we will see later in this respect @xcite is slightly more accurate when he says in his half sentence remark that `` maxwell assumed that all motions [ of the particles ] were equally probable '' ( although , as we will see , maxwell s _ assumptions _ only _ implied _ that after rebound all _ directions _ of motion are equally probable ) .
according to @xcite maxwell claimed that `` when a moving sphere collides with with another fixed sphere randomly placed in its path , the direction in which it rebounds is distributed with equal probability over each element of solid angle '' ( this is incorrect since maxwell did not assume any of the particles to be fixed ) , and they further state without providing any further details or evidence that maxwell used this claim to argue that `` the distribution law has an explicit functional form '' .
@xcite mentions en passant that maxwell realized that the `` velocities [ of the particles ] will suffer incessant change due to mutual collisions '' but he implies that maxwell s reference to collisions is , with maxwell s own remark made 20 years later about certain derivations in the works of others , `` rather for the sake of enabling the reader to form a mental image of the material system than as a condition for the demonstration '' ( ibid .
p. 948 ) .
uffink also emphasizes that the assumptions of maxwell s proposition iv , despite appearances and in contrast to maxwell s subsequent proof of 1867 , `` do not refer to collisions at all '' ( ibid .
p. 948 ; see also his contrast with maxwell s 1867 proof on ibid .
p. 950 ) .
@xcite is the only exception in giving an extensive treatment of maxwell s collision model but the focus of the paper is on exploring the connections between maxwell s proposition ii and later debates of the 1890s on the stosszahlansatz ; dias does not address what role , if any , maxwell s collision model played in justifying the assumptions of proposition iv .
the scant attention paid to the justification question may stem from historians unanimously following brush s @xcite reconstruction of proposition iv as a proof resting on _ two _ assumptions , that of probabilistic independence ( a2 ) and of rotational symmetry ( a3 ) , and they are at best only tacit about the need for a third assumption ( a1 ) that a stationary velocity distribution exists ( see i.e. @xcite , @xcite , @xcite , @xcite ) .
they do so despite acknowledging the analogy with herschel s derivation pointed out by gillispie , five years after brush s @xcite reconstruction , in @xcite . as we have seen hershel
explicitly emphasized a third assumption , that there must `` exists a relation or numerical [ ... ] regular law of progression , _ which must necessarily be general and apply alike to all cases _ '' ( * ? ? ?
* ; emphasis in original ) , which adapted to the kinetic theory translates as the assumption of the existence of a stationary velocity distribution , our ( a1 ) . by only explicitly acknowledging two assumptions instead of three the natural guess
might have been that what maxwell tried to do with his collision model was to physically justify one or both of these two assumptions . since without a detailed analysis of maxwell s collision model
only superficial analogies are apparent between maxwell s prior results and assumptions ( a2)(a3 ) these prior results of maxwell were supposedly aimed at justifying , and since maxwell himself later abandoned the independence assumption ( a2 ) describing it as being `` precarious '' @xcite , it could have been tempting to glance over the opening part of maxwell s paper as being non - consequential to the argument presented in proposition iv .
we are going to argue that the most plausible role the collision model played in maxwell s thinking is in justifying assumption ( a1 ) , although our analysis of maxwell s collision model also sheds light on the extent to which his prior results are capable to support assumptions ( a2 ) and ( a3 ) beyond the superficial analogies . to do this
we start with a brief reconstruction of his propositions i iii .
in the beginning of the first chapter , `` on the motion and collision of perfectly elastic spheres , '' maxwell presents us with three propositions concerning the interaction of particles .
proposition i and proposition iii are straightforward exercises in newtonian mechanics that establish the motion of two particles modeled by perfectly elastic spheres of non - zero radii @xmath37 , @xmath38 and of mass @xmath39 , @xmath40 after they collide .
it is assumed that the particles do not interact aside such pairwise collisions . in modern notation maxwell
calculates the encounter operator @xmath41 where @xmath42 are the velocities of the particles before the impact , @xmath43 are the velocities after impact , and @xmath44 is a further parameter describing the impact in the following way .
let @xmath45 be the velocity of the first particle relative to the velocity of the center of mass @xmath46 before impact and let @xmath47 be a circle of radius @xmath48 that lies in the plane perpendicular to @xmath49 , @xmath47 being centered on the center of the second particle
. for a collision to take place the line of motion of the first particle needs to intersect @xmath47 ; @xmath44 is the position of this intersection within @xmath47 .
( see figure [ propiifigure ] . ) : velocities relative to the center of mass .
@xmath47 : circle of radii @xmath48 centered on the second particle in the plane perpendicular to @xmath49 .
@xmath44 : point where the line of motion of the first particle intersects circle @xmath47.,height=170 ] due to conservation of momentum the collision does not affect the velocity of the center of mass : @xmath50 .
maxwell shows that a perfectly elastic collision alters the direction of the relative velocity but it keeps its magnitude intact : @xmath51 and therefore the velocity after impact can be expressed as @xmath52 where @xmath53 and @xmath54 is a direction of unit length depending on the position @xmath44 within the circle @xmath47 .
( mutatis mutandis for the @xmath55 velocity of the second particle relative to the center of mass . )
maxwell also claims to have proven that whenever a collision takes place all directions of rebound are equally likely in the center of mass frame .
what his proposition ii actually shows is that if the assumption that condition m : : the impact parameter @xmath44 is distributed uniformly within the circle @xmath47 holds then condition a : : the @xmath54 directions of rebound relative to the center of mass are distributed uniformly on the surface of the unit sphere .
the name condition a is due to s. h. burbury @xcite who invoked it in a debate on the pages of _ nature _ with g. h. bryan about boltzmann s h - theorem .
condition m is the terminology we introduce here to distinguish condition a from the assumption maxwell actually made during the proof of proposition ii . since in maxwell s particle model condition m and condition
a imply each other they tend to be identified ( for such an exchange of conditions see i.e. @xcite ) .
maxwell himself neither emphasizes nor attempts to justify condition m ; he merely asserts that `` within this circle [ @xmath47 ] every position [ @xmath44 ] is equally probable '' @xcite , and in the coming years he keeps referring to the conclusion of proposition ii as if it were unconditional ( see e.g. @xcite ; for a further analysis of condition a and the _ nature _ debate see @xcite ) .
the conclusion drawn from propositions i iii is that the @xmath58 velocity of a particle after impact consists of two components : the first is the @xmath59 velocity of the center of mass , which is determined by the @xmath60 and @xmath61 incoming velocities , and the second is the @xmath62 velocity relative to the center of mass , whose magnitude is determined by the incoming velocities , but whose direction is not determined by the incoming velocities , but may with equal probability be in any direction .
after being done with propositions i iii , but before turning his attention to proposition iv , maxwell makes the following remark about the effect of many collisions : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ if a great many equal spherical particles were in motion in a perfectly elastic vessel , collisions would take place among the particles , and their velocities would be altered at every collision ; so that after a certain time the _ vis viva _ will be divided among the particles according to some regular law , the average number of particles whose velocity lies between certain limits being ascertainable , though the velocity of each particle changes at every collision .
@xcite _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ thus maxwell claims that after sufficient number of collisions the kinetic energies become distributed along a non - changing distribution .
this implies that some sort of process that brings about this distribution of kinetic energies takes place , apparently due to the collisions these particles go through . on the surface of it
maxwell s remark strikes as an expression of hope rather than an argument .
is there any evidence that maxwell had good reasons to hold that propositions i iii are capable of supporting a claim about tendency towards equilibrium ?
the answer is affirmative , and this becomes clear if we jump ahead to proposition vi in which maxwell argues that when two gases mix their temperatures equalizes .
this argument is unknown in the historical and in the physics literature and since besides being relevant to the question of justification of an assumption of proposition iv it is also interesting and historically significant on its own right , we now briefly reconstruct it .
when two particles , modeled as perfectly elastic spheres of mass @xmath39 and @xmath40 , collide the difference between their kinetic energies after the collision can be readily expressed as a function of the difference between their kinetic energies before the collision : @xmath63 where @xmath64 is the angle enclosed by the incoming velocities whose magnitude is @xmath3 and @xmath65 and @xmath66 is the angle enclosed by the velocity of the center of mass @xmath59 with the relative rebound velocity @xmath62 ( see figure [ propvifigure ] ; for calculation of the coefficients see the appendix ) .
propositions i iii together with condition m entail that @xmath62 may point to all directions with equal probability , which means there is an equal chance that the angle of rebound @xmath66 falls between @xmath67 and @xmath68 or between @xmath69 and @xmath70 . : incoming velocities ; @xmath59 : velocity of center of mass ; @xmath45 : velocity of the first particle relative to the center of mass ; @xmath58 : velocity of the first particle after rebound ; @xmath71 : angle between incoming velocities ; @xmath72 : angle of rebound.,height=170 ] consider now collisions between many particles during a time interval @xmath73 , and let us write @xmath74 whenever @xmath75 is a particle of mass @xmath39 , @xmath76 is a particle of mass @xmath40 , and @xmath75 and @xmath76 collide during @xmath73 with respective incoming speeds @xmath3 , @xmath65 and angle of income @xmath64 .
we assume that @xmath73 is long enough for @xmath77 to be large but short enough to allow for at most one collision for each particle .
let us sum up equations ( [ eq_differenceofkineticenergiesafterrebound ] ) for all collision pairs in @xmath78 by assuming that condition m holds for all these particle collisions independently .
since then all directions of rebound occur with equal frequency and since @xmath79 only depends on @xmath80 , and @xmath64 which are now all fixed , the @xmath81 terms approximately cancel and we get @xmath82 let us further assume that all directions of incoming velocities for the colliding particles occur with approximately equal frequency an assumption that maxwell took for granted in his later derivations , i.e. in @xcite and that the number of collisions is independent of the enclosed angle of income ( that @xmath77 is the same for all @xmath64 , or at least that @xmath83 for all @xmath64 given a fixed @xmath3 and @xmath65 ) and sum up for all possible angles of income . since @xmath84 only depends on @xmath85 , and @xmath65 which are now all fixed , the @xmath86 terms approximately cancel and we get @xmath87 by further summing up for all possible magnitudes of incoming velocities @xmath88 on both sides of the equation we account for all collisions between particles of mass @xmath39 and @xmath40 during @xmath73 exactly once , so a division with the total number of such collisions and substituting for @xmath89 yields @xmath90 meaning that the difference between the mean kinetic energies of the colliding particles decrease after the collisions ! any change in the difference between the mean kinetic energies ( between the temperatures ) of _ all _ @xmath39-particles and of _ all _ @xmath40-particles during @xmath73 is due to collisions between particles of different mass ( expressed by equation ( [ eq_kineticenergydifferencedecreases ] ) ) since the total kinetic energy of particles which do not collide and the total kinetic energy of those particles that collide with another particle of the same mass does not change during @xmath73 . if the particles that undergo collision during @xmath73 are randomly drawn from the set of all particles then the larger the initial temperature difference between @xmath39-particles and @xmath40-particles the larger the chance that the drawn @xmath78 samples of colliding pairs also exhibit a mean kinetic energy difference of the same sign . if the sign is the same then since due to ( [ eq_kineticenergydifferencedecreases ] ) the mean kinetic energy difference between the colliding pairs decreases during @xmath73 it follows that the temperature difference of all @xmath39-particles and @xmath40-particles also decreases .
( if the sign is different then the temperature difference between the gases _ increases _ , underlining that the present argument for tendency towards equilibrium is statistical . ) by repeating the argument for successive @xmath73 time intervals it follows that the difference between temperatures of @xmath39-particles and @xmath40-particles tends to vanish .
when the temperature difference vanishes or becomes small the collisions continue to produce small temperature fluctuations ( due to the mean kinetic energy differences that may still occur in the random sample of colliding particle pairs ) .
note that the proof becomes extremely simple when we mix gases of different temperature whose particles have the same mass . in this case equation ( [ eq_differenceofkineticenergiesafterrebound ] ) simplifies to @xmath91 and one only needs to invoke the first cancellation argument provided by condition m to conclude that the temperature difference decreases as collisions take place .
what we thus have is a simple , elegant , and elementary mechanical argument for a statistical tendency towards equilibrium that is based on the assumptions that * particles can be modeled as perfectly elastic spheres , * the number , size , speed etc . of these particles
is such that the time evolution can be split into successive time intervals during which each particle has an approximately equal independent chance to participate in at most one , pairwise collision , and * in this interval all directions of rebound occur with approximately the same relative frequency ( that for any pair of incoming velocity and for the colliding particle pairs that have these incoming velocities the relative frequency of the impact parameter is roughly the same for all positions on the collision circle ) .
although we know well from later developments that assumptions ( ii ) and ( iii ) are difficult to square with the strict dynamical evolution entailed by ( i ) , its simplicity and its geometrically intuitive probabilistic assumptions could very well allow the argument to fit in an undergraduate course on classical mechanics
. how can it be that this simple argument for tendency towards thermal equilibrium is unknown by historians and physics educators alike ?
the conclusion that collisions bring about an equalization of temperatures is what maxwell draws in his proposition vi after he derives a formula he interprets as ( [ eq_kineticenergydifferencedecreases ] ) by invoking his earlier results about collisions .
prima facie , however , maxwell s proof seems rather baffling .
what he seems to have calculated in his proposition vi is the difference of kinetic energies of particles of mass @xmath39 and @xmath40 after they collide in a specially arranged manner where ( using the notation of the previous subsection ) * @xmath92 , * @xmath93 , * @xmath94 , @xmath95 . by substituting assumptions ( p1)-(p3 ) to equation ( [ eq_differenceofkineticenergiesafterrebound ] )
one gets @xmath96 which is the equation that maxwell s own derivation would have really established .
maxwell curiously assumes that @xmath97 and he also interprets ( [ eq_maxwellreally ] ) as if he had just derived ( [ eq_kineticenergydifferencedecreases ] ) and proceeds to draw the conclusion about the long term equalization of mean kinetic energies on the basis of it .
thus the calculations of proposition vi make maxwell s argument seem to be dependent on a particular arrangement of incoming velocities and rebound angle of a pair of particles , and it seems entirely unclear how he thought that an argument about a particular collision of a pair of particles could allow for a valid inference about averages over all collisions , or how the magnitude of velocity after impact that initially equals the mean velocity could be substituted for the mean of velocities after impact , or how the kinetic energy calculated from a mean velocity could be substituted for a mean kinetic energy . on this basis maxwell s proof ,
if mentioned at all , is dismissed in the historical literature as a `` rather lame argument '' and the few historians who do devote more than a footnote to it find it `` amazing '' that the otherwise brilliant maxwell or `` any of his contemporaries who bothered to examine the argument in detail would have accepted it '' @xcite . in the light of the simple argument for equalization of kinetic energies we outlined in the previous subsection
an argument that rest purely on maxwell s propositions i iii and invokes the same sort of calculations that maxwell performs during his proof of proposition vi it seems plausible that maxwell was being too terse instead of lame .
it is clear from the geometry of collisions and from the argument we gave above that for equation ( [ eq_differenceofkineticenergiesafterrebound ] ) averaging over all directions of rebound yields the same result as making assumption ( p1 ) , as a result of which the third term cancels .
this then justifies focusing on collisions with perpendicular rebound angle as being sufficiently typical for the purposes of the proof .
it is also geometrically clear that averaging over all directions of incoming velocities yield the same result as making assumption ( p2 ) , as a result of which the second term cancels .
this then would justify focusing on collisions with perpendicular income angle as being sufficiently typical for the purposes of the proof .
propositions i iii motivates assumption ( p1 ) but assumption ( p2 ) is in need of justification . instead of invoking a further assumption that all directions of incoming velocities are equally likely ( like we did above , and as maxwell did in his later works ) , maxwell sets the magnitude of the incoming velocities @xmath3 , @xmath65 to equal their mean values given by proposition iv ( assumption ( p3 ) ) , and he further assumes ( p4 ) that the magnitude of the relative velocity @xmath98 equals its mean value given by proposition v. since proposition v shows that the square of the mean relative velocity equals the sum of squares of the mean velocities , and therefore that the two incoming velocities must be perpendicular , ( p3 ) and ( p4 ) entail ( p2 ) .
assumptions ( p3 ) and ( p4 ) are also assumptions about typical velocities , but they are not appropriate for the purposes of the proof . albeit the proof that proceeds directly along ( p3 ) and ( p4 ) is clearly suspect , ( p3 ) and ( p4 ) might have been intended as mere short - cut for a more detailed argument that rests on an averaging procedure over the incoming velocities , along the lines we sketched above .
albeit this reconstruction still renders the proof of proposition vi messy and misleading , it allows us to see the arrangement choices maxwell makes during his derivation as geometrically intuitive shortcuts to an averaging procedure that rests on an application of the results of his propositions i iii .
if we make the charitable assumption that maxwell simply did nt properly clean up his proof , as opposed to making assumptions whose fallaciousness should have been obvious , then his reconstructed proposition vi becomes a convincing demonstration of a tendency towards equilibrium that lives up to the standards of its own time .
although the proof does depend on unjustified probabilistic assumptions such as condition m , in this regard it does not fare worse than boltzmann s later attempts ( starting with @xcite ) to arrive at his celebrated h - theorem , and thus it should reclaim its historical precedence over boltzmann s demonstrations as being the first good `` bad '' mechanical proof of tendency towards equilibrium . the only historical side - remark i m aware of that also suggests a direct connection between maxwell s propositions i iii and boltzmann s h - theorem is a side remark of @xcite who says that _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ maxwell ( 1860 ) invoked proposition ii to show how collisions randomized the directions of motion of the molecules , bringing about isotropy of pressure , as demanded by the equation of state for perfect gases .
in this use , proposition ii becomes a kind of `` proof '' of how equilibrium of pressure ( transmission of linear momentum ) sets in , and thus is , metaphorically speaking , a sort of h - theorem .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ although the connection between proposition ii and the randomization of direction of motions of molecules is clear , i find no evidence for dias claim that maxwell made any use of proposition ii to argue for the onset of the isotropy of pressure ( dias provides none ) . from dias reference to linear momentum i suspect that his remark was based on a misinterpretation of what proposition vi aimed to show .
maxwell never seem to have revisited proposition vi . around 1863
he became dissatisfied with clausius mean free path method and started to develop a theory valid for any kind of force laws acting between particles , which he came to treat as centers of force rather than colliding spheres with determinate radii
. with his new approach serious problems emerge concerning proposition i and iii : as maxwell s @xcite calculations show , the motion of particles subject to a never - vanishing repulsive force law in general depends on their incoming relative velocity , so the angle of rebound also depends on this incoming relative velocity .
although in his _ on the dynamical theory of gases _ @xcite he chooses a particular inverse @xmath99th - power force law for which the angle of reflection is independent of the incoming relative velocity , this choice is poorly motivated and gets abandoned later .
his writings after 1867 indicate that he grew skeptical of the possibility of justifying the second law on a mechanical basis @xcite .
before analyzing the extent to which proposition i iii , condition m , and the argument for an approach to equilibrium can be used to justify the mathematical assumptions of proposition iv it is worthwhile to take a look at how maxwell , herschel , and clausius understood what these mathematical assumptions , in particular the independence assumption ( a2 ) , amounted to .
let us start with herschel s understanding of the independence assumption of his own proof .
we note first that there is a small but important difference between herschel s original @xcite review article and its @xcite reprint , which is the version that maxwell have likely read . in a footnote added to the @xcite
reprint herschel attempts to clarify what he means by independence of the deviations of errors in different directions and emphasizes the importance of this independence assumption : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ that is , _ the increase or diminution [ of the deviation ] in one or [ sic ! ] which may take place without increasing or diminishing the other .
_ on this , the whole force of the proof turns .
@xcite , @xcite .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ although maxwell s likely adaptation of herschel s derivation became common wisdom in the historical literature , so far it has gone unnoticed by historians that in the quoted emphatic footnote herschel actually mischaracterizes the crucial necessary independence assumption of the derivation . a mathematically correct derivation requires _ probabilistic independence _ of deviations in rectangular directions ; according to herschel , however , the whole force of the proof turns on the assumption that a deviation in one direction does not impose a restriction on the values a deviation in another direction may take , which is not the same as probabilistic independence of the deviations .
let us briefly distinguish these notions of independence .
let @xmath100 and @xmath101 denote physical quantities ( i.e. velocity components ) that can take values in the set @xmath102 and @xmath103 respectively .
suppose that we may not know what other quantities serve as inputs that determine the momentary values of @xmath100 and @xmath101 , but let us assume that we might know about certain constraints that connect the values @xmath100 and @xmath101 can take in the following way .
let @xmath104 assign to each value @xmath105 the set of values @xmath106 that variable @xmath101 may take if we condition upon @xmath100 taking the value @xmath107 .
if @xmath108 is single valued for all @xmath105 then if @xmath100 takes a value @xmath109 then @xmath101 must take the only possible value in @xmath110 : we may thus say that _
@xmath100 value determines @xmath101_. if @xmath110 has more than one element for at least one @xmath105 then _ @xmath101 is partially value independent from @xmath100 _ : for some value variable @xmath100 may take there are several possibilities that @xmath101 may take that are not determined by the value of @xmath100 .
if @xmath111 for all @xmath112 then _ @xmath101 is value independent from @xmath100 _ : no matter what value variable @xmath100 takes that does not influence the possible values @xmath101 may take . finally when @xmath100 and @xmath101 are random variables then _
@xmath101 is probabilistically independent from @xmath100 _ if for all @xmath113 : @xmath114 , or , alternatively , when @xmath115 for the respective density functions .
these independence notions are not equivalent with each other
. value independence does not imply probabilistic independence : indeed the domains of two value independent variables do not even need to be endowed with a probability measure .
probabilistic independence also does not imply value independence : a value @xmath109 of @xmath100 may prevent @xmath101 taking a value @xmath116 without ruining probabilistic independence of @xmath100 and @xmath101 if the probability of the occurrence of @xmath109 and @xmath116 is zero .
value independence is symmetric , and value independent variables are also partially value independent but the converse is not the case .
it is clear both from herschel s derivation and from maxwell s proof of proposition iv that their `` independence '' assumption is meant to entail the factorization of the probability distribution function .
if by `` independence '' one means probabilistic independence then factorization does follow and the proof is correct ; factorization however does not follow from value independence , which herschel mistakenly emphasized to be the crucial assumption of his derivation .
now even if maxwell adapted the derivation from herschel s article it is natural to assume that he had reasons to believe that the required mathematical assumptions can be motivated by physical considerations about gases . without such physical considerations
the application of the mere mathematical proposition to gases would remain unjustified .
but maxwell s attempt at physically justifying independence of the velocity components would have clearly depended on his understanding of the sense of independence he had believed to guarantee the success of the proof .
we are going to return to this point in section [ subsect_independence ] .
let us now turn to clausius understanding of the meaning of independence . in his summary of maxwell s propositions
i iii clausius ( @xcite ) claims that the velocity of center of mass @xmath59 quantifies the extent to which the velocity after impact is `` dependent '' upon the velocities before impact while the velocity relative to the center of mass @xmath62 quantifies the extent to which the velocity after impact is `` independent '' from the velocities before impact .
it is not clear what clausius meant here by `` independence '' since what he writes is compatible with several different readings . given condition m @xmath62 may point in any direction no matter what the incoming velocities @xmath60 and @xmath61 were , so it is value independent since setting values for @xmath60 and @xmath61 does not restrict the range of possible values for the direction of @xmath62 .
in the same vein @xmath58 is partially value independent from @xmath60 and @xmath61 since setting values for @xmath60 and @xmath61 does narrow the range of possible values of @xmath58 but @xmath58 may still take several values , not just one .
clausius could also have meant that the direction of @xmath62 is probabilistically independent from @xmath60 and @xmath61 : conditioning upon any value for @xmath60 and @xmath61 does not alter the _ probability _ that @xmath62 points in a certain direction . upon a closer
look what clausius emphasizes in the text and in the accompanying footnote is that the incoming velocities do not determine the direction of the velocities relative to the center of mass after impact and in turn they do not fully determine the velocities after impact either , as these latter may take several ( indeed , infinitely many ) different possible values for any fixed values of the incoming velocities .
these remarks of clausius are suggestive that he intended to make a point about the partial value independence of the velocities before and after impact , but there is no evidence that clausius even distinguished between value independence and probabilistic independence . finally let us make a brief remark about maxwell s own usage of the term `` independent '' .
the term appears once more in the context of a generalization of proposition iv and proposition v ( ibid .
p. 315 ) but the appearance does not shed any more clarity on its intended interpretation than the proof of proposition iv .
in all other contexts in which the term `` independent '' appears in the _ illustrations _ ( see p. 288
, 298 , 300 ) it is clear that maxwell understands it as referring to value independence .
taking a systematic look at maxwell s prior and contemporary work on color vision ( see esp .
@xcite ) i also find that maxwell systematically uses the term independent in the sense of value independence .
indeed i find that ambiguous usage of the term `` independent '' is pervasive in and before the 1860s .
that different concepts are named by the same term would not in itself be a problem if users of the term were conscious of their non - interchangeability .
unfortunately it seems that this consciousness was missing in the works of those , such as clausius and herschel , that most likely have influenced maxwell s thinking , and there is no evidence that maxwell was aware of such distinction in and before his 1860 work on the kinetic theory .
let us now return to maxwell s proposition iv .
the paragraph preceding the statement of proposition iv , which we have already quoted , claims that after large number of collisions the kinetic energy `` will be divided among the particles according to some regular law , the average number of particles whose velocity lies between certain limits being ascertainable '' @xcite , meaning that this number is not going to be further altered by subsequent collisions , that is , that the velocity distribution becomes stationary . indeed without the assumption of stationarity it would be hard to understand what enabled the presumption of the time - independent velocity distribution @xmath27 right in the beginning of the proof of proposition iv .
but if maxwell succeeded to show that particle collisions lead to stationarity prior to proposition iv then it would have been sufficient for him to assume , for the purposes of the proof of proposition iv , the existence of the right kind of particle collisions . and
this is what maxwell actually says he does , to wit : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ prop .
iv . to find the average number of particles whose velocities lie between given limits , _ after a great number of collisions among a great number of equal particles . _
( emphasis added . ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ that his analysis of particle collisions was in part intended to establish the existence of a stationary velocity distribution is also clear from maxwell s later remarks about his 1860 proof : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the only case in which i have determined the form of this [ velocity distribution ] function [ in the 1860 paper ] is that of one or more kinds of _ molecules _ which have _ by their continual encounters brought about a distribution of velocity _ such _ that _ the number of molecules whose velocity lies within given limits _ remains constant_. @xcite ( emphasis added . ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ in addition we have also seen that the idea that molecular collisions could lead to a stationary velocity distribution was more than baseless wishful thinking .
we showed that propositions i iii are indeed sufficient for a proof of tendency towards equilibrium , which is what maxwell intended to do with his proposition vi . in the special case when the masses are equal , which is the relevant case for proposition iv , we have also seen that the proof of equalization of kinetic energies becomes so simple as to potentially warrant only verbal referencing .
the availability of a proof of reaching equalization of temperatures lends credence for the availability of a proof of reaching a stationary velocity distributions .
thus we can conclude that maxwell s argument consists of two steps : a ` lemma ' showing that particle collisions lead to stationarity , and a ` theorem ' that makes use of the lemma .
it is then of course no surprise that reference to particle collisions only appears in the statement of the theorem and is absent from the proof following the statement , since all the work having to do with particle collisions was previously delegated to the lemma ; the proof of the theorem itself only makes use of the assumption of stationarity that follows from the lemma and its assumption about collisions .
establishing the conclusion of course requires both establishing the lemma and establishing the theorem , and thus maxwell s 1860 argument does crucially depend on particle collisions , contrary to what historians of science ( i.e. @xcite , @xcite ) , who only focus on the proof that follows the statement of proposition iv , claim .
the extent to which maxwell succeeded in establishing the lemma is , of course , another question . establishing that a certain tendency towards thermal equilibrium does follow from the model assumptions is clearly significant , but it is unclear to me how equalization of kinetic energies would strictly imply the onset of a stationary velocity distribution .
if we conceptually divide the particles of the gas into two large groups
say , red and blue then maxwell s results imply that after collisions continue for a long time the mean velocity of all red particles will equal the mean velocity of all blue particles .
this puts a constraint on how the velocity distribution of all particles may change after collisions .
since the division of particles to red and blue is arbitrary we end up with many such similar constraints , and jointly all these constraints _ may _ be sufficient to force the velocity distribution to become stationary .
so much is clear , but whether this is so requires an explicit argument rather than a hopeful gesture .
although this attempt stays close to the line of reasoning maxwell himself utilizes in the end of his proof of proposition vi he does not elaborate on the details of such or similar musings .
he apparently thought that the evidence provided by propositions i
iii on the nature of particle collisions and his scattered remarks preceding the statement of proposition iv are sufficient to establish that collisions lead to a stationary velocity distribution straightforwardly .
some followers of maxwell shared this view , although it progressively became more clear that a proof of it , if possible at all , may not be easily befitted to a margin . yet
as late as 1890 we find proofs to the effect that propositions i iii and condition a implies stationarity : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ proposition ] 6 .
( a ) every distribution of velocities among the molecules which satisfies the condition that for given [ velocity of the center of mass of a pair of colliding particles ] @xmath65 all directions of [ the velocity of rebound relative to the center of mass ] @xmath117 are equally probable is [ ... ] stationary . @xcite _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ one of the main difficulties with these proofs is that they sneak in a probabilistic independence assumption in one form or another without appropriate justification .
the most crucial assumption of the proof of proposition iv is the factorization of the velocity distribution to its marginals , which is a way to state the probabilistic independence of the velocity components . let us re - quote , in its entirety , what maxwell had to say about establishing this latter : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ ... ] the existence of the velocity @xmath0 does not in any way affect that of the velocities @xmath19 or @xmath20 , since these are all at right angles to each other and independent[. ] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the obvious reading of this claim is that setting any value for the velocity component @xmath0 does not affect what values @xmath19 and @xmath20 could possibly take ; in other words , that @xmath0 is value independent from @xmath19 and @xmath20 .
this obvious reading renders the claim obviously true and in no need of further justification . as we have seen before
, herschel emphasizes that value independence is the crucial assumption that is needed to justify the factorization of the distribution to its marginals .
given that maxwell himself consistently used the term `` independent '' in the sense of value independence in his other works prior to 1860 it is then possible that he simply took what herschel said for granted , and only bothered to point out that value independence of the velocity components in different directions clearly holds .
value independence of course does not imply probabilistic independence , and hence the assumption that would have been truly required for the mathematical proof remained unjustified . according to
another reading maxwell was merely sloppy stating his assumption in the quoted claim , and what he really meant was that setting any value for the velocity component @xmath0 does not affect the _ probability _ of @xmath19 and @xmath20 taking their values .
this reading seemingly finds support in what he says in his 1867 recollection that the 1860 proof was _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ ... ] founded on the assumption that the probability of a molecule having a velocity resolved parallel to @xmath109 lying between given limits is not in any way affected by the knowledge that the molecule has a given velocity resolved parallel to @xmath116 .
@xcite _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ aside from the curious allusion to ` knowledge '
this recollection correctly states that the proof of proposition iv requires the assumption of probabilistic independence of @xmath0 and @xmath19 .
what the recollection does not make clear , however , is whether maxwell also intended to rely on the assumption of probabilistic independence ( as opposed to the assumption of value independence ) _ back in 1860 _ , and so this later reflection of maxwell in itself can not be taken as evidence for an awareness of a distinction between probabilistic independence and value independence at the time when he produced his first proof .
all we know is that in 1867 maxwell found this reformulated independence requirement sufficiently `` precarious '' to warrant the construction of a different proof . in the rest of this section
we follow the latter reading , namely that maxwell intended to have assumed a sort of probabilistic independence .
it is still not entirely clear what could have been the intended carrier of the assumption of probabilistic independence according to maxwell s 1867 recollection ( see figure [ flowchartfigure ] for some possibilities ) .
although maxwell seemingly talks about the velocity components of an individual particle , the velocity components of individual particles have a definite value at any given time , rendering any talk about a particle having a velocity component with some probability obscure ( unless we either interpret probability as ignorance
the above mentioned appearance of the term ` knowledge ' may indicate that or unless with a great dose of anachronism we give up definiteness of velocity values along the lines of an objective interpretation of the collision probabilities ) .
it is more in line with the frequentist allures of the 1860 proof to take the probabilistic independence assumption to be a statement about a collection of particles , namely that the distribution of velocities along the @xmath116 axis for all particles is the same as the distribution of velocities along the @xmath116 axis for the particles whose velocity component along the @xmath109 axis is set to some value .
does the _ illustrations _ provide a justification for the assumption of probabilistic independence of velocity components of a collection of particles , or does at least have the resources to do so ? with our reconstruction of propositions i iii at hand the obvious guess is to trace back the assumption of probabilistic independence to condition m. recall that condition m states that whenever two particles collide the impact parameter @xmath44 is distributed uniformly within a circle @xmath47 .
the ` whenever ' part of this assumption presumably states that @xmath44 is distributed uniformly regardless what are the incoming velocities of the pair of colliding particles or what is the configuration of other particles that are not participating in the collision . in a precise formulation
this would presumably translate to conditional probabilistic independence of @xmath44 from the incoming velocities of the pair and from other physical quantities characterizing the remaining particles .
if so then condition m already furnishes a sort of probabilistic independence assumption regarding the velocities of a particle pair , and it is not inconceivable that some clever aggregation procedure would then lend credence to the assumption of probabilistic independence of velocity components of a collection of particles .
maxwell himself does not hint at such argument in the _ illustrations _ , but associating condition m ( and its consequence , condition a ) with a sense of probabilistic independence that is required for a derivation of the velocity distribution was so deep that even in 1894 , during an ignited debate on the pages of _ nature _ about the validity of boltzmann s h - theorem , condition a and the stosszahlansatz ( the crucial independence assumption about factorization of the velocity probability distribution that appears in maxwell s 1867 proof ) gets identified and interchanged without justification or warning ( for a detailed account see @xcite ) . if the non - interchangeability of these assumptions did not occur to prominent physicists who had the advantage of many years of developments that followed maxwell s 1860 proof , it seems reasonable to assume that it might have not occurred to maxwell either . hence maxwell might have thought that condition m , whose geometrical meaning is at least apparent , and which ( as @xcite ventures ) may be loosely motivated on the basis of clausius mean free path approach , justifies the probabilistic independence of @xmath0 and @xmath19 .
it would be interesting to see , but i do not know of a formally satisfying investigation of the relationship of condition m and the assumption of probabilistic independence of velocity components of a collection of particles .
clearly , any such investigation would require that first someone says exactly what condition m states .
maxwell s justification of the assumption of rotational symmetry of the velocity distribution is physically intuitive : nothing in the argument he gave for the factorization of the velocity distribution depends on the particular choice of the oblique coordinate system , and the physical behavior of the colliding particles should clearly not depend on how we choose to describe their behavior . requiring that the factorization holds for other oblique coordinate systems entails that the distribution only depends from the distance from the origin .
this result also follows when we only resort to non - dependence from choice of orthogonal coordinate systems , in which case it can be conveniently mathematically expressed as @xmath118 which is what maxwell writes .
( note that the orthogonality assumption is not essential , only convenient . with any oblique coordinate system the functional equation ( [ functionalequation ] ) still has a unique solution , a generalized normal distribution . )
to sum up , among the three mathematical assumptions of maxwell s proposition iv ( a1 ) stationarity was intended to be justified on the basis of prior assumptions about particle collisions , ( a2 ) probabilistic independence of velocity components was either misunderstood as value independence or was intended to be justified on the basis of prior assumptions about particle collisions , and ( a3 ) rotational symmetry was supposed to follow from the non - dependence of the argument for the assumption of probabilistic independence of velocity components from the choice of coordinate system .
maxwell s analysis and assumptions regarding particle collisions hence played a crucial motivational role for the mathematical assumptions of proposition iv .
finally note that if the number of particles is finite then the actual distribution of velocities at any time must be discrete and have a bounded support .
however the only measurable solution of the functional equation ( [ functionalequation ] ) , the gaussian ( [ gaussian ] ) , is a continuous function with infinite support . to reconcile this tension one would need to give up at least one of the assumptions
: one could insist that stationarity , rotational symmetry , or probabilistic independence only holds approximately , or that maxwell s velocity distribution @xmath27 merely provides an approximate description .
this is one of a number of issues relating to proposition iv that we leave here undiscussed as they have been thoroughly addressed in the literature ; for an overview see @xcite .
the mathematical essence of the proof of proposition iv is that a velocity distribution @xmath9 follows the maxwell velocity distribution law if it satisfies the mathematical assumptions of probabilistic independence @xmath10 and rotational symmetry @xmath119 .
the mathematical essence of a proof may however mislead about what in it needs to be endowed with physical meaning and justification .
we argued that emphasizing the importance of the implicit assumption of the existence of a stationary velocity distribution @xmath9 make the need for an argument for an approach to equilibrium clear and this in turn may shed light on the role maxwell s widely ignored molecular collision model played in his _ illustrations of the dynamical theory of gases_. in particular we showed that for perfectly elastic particles maxwell s proposition i iii under the assumptions that their time evolution satisfies the constraint that it can be split into successive time intervals during which each particle has an approximately equal independent chance to participate in at most one , pairwise collision , and that in this interval the impact parameter is approximately equally distributed on the collision circle ( condition m ) yields a good `` bad '' mechanical proof of approach to equilibrium temperature , a precursor to boltzmann s h - theorem , which also lends partial credence to the availability of a proof of reaching a stationary velocity distribution .
we argued that condition m of proposition ii furnishes a sort of probabilistic independence assumption regarding velocities that could have motivated the assumption of probabilistic independence , but we also suggested another potential explanation for the apparent lack of effort on the part of maxwell to justify probabilistic independence , namely that maxwell might have conflated the notions of probabilistic independence and value independence .
we argued that distinction of meaning of these two notions was not yet available in the 1850 s and we presented circumstantial evidence for these claims on the basis of relevant prior works of clausius , herschel , and maxwell .
section [ sect_justification ] explored the ways how propositions i iii , condition m , and the approach to equilibrium furnished by a reconstructed version of proposition vi could have played a role in maxwell s physical justification of the three mathematical assumptions of proposition iv ; we have seen that there are multiple interpretative possibilities . in closing
let me choose the interpretation which seems to me most plausible . on a physical basis
maxwell knew that after a sufficient amount of time a box of gas settles into a thermal equilibrium , and he made the important realization that supplementing a standard newtonian collision model with the seemingly innocuous condition m allows the proof of such an approach to thermal equilibrium if the particles are modeled as perfectly elastic spheres .
maxwell implicitly assumed that the onset of this equilibrium also entails that the velocity distribution function becomes stationary . thinking that the existence of a stationary velocity distribution can be proved on a physical basis he looked around for mathematical proofs that could be used to characterize properties of such stationary distributions , and he recalled herschel s characterization of the distribution of errors from reading his _ essays _ as such .
unfortunately herschel mischaracterized the crucial independence assumption of his own proof , and maxwell implemented herschel s proof together with this mischaracterization to the context of the kinetic theory . in turn the physical justification of the so - mischaracterized independence assumption ( and as a consequence also
the assumption of rotational symmetry ) seemed straightforward .
maxwell later realized that this independence assumption is `` precarious '' and since other considerations also led him to abandon clausius mean free path approach that lies behind the collision model , he also abandoned the attempt to establish stationarity of the velocity distribution via condition m. it took several decades until the scientific community came to a clear grasp of the distinctions between various interpretations of probability and various notions of independence , and thus maxwell s early mistake in his first paper on the kinetic theory was by no means an obvious one .
this paper is based on a substantive revision of the ( unpublished ) work @xcite .
i thank john norton , john earman , jos uffink , michael strevens , michael reed , and mrton gmri for feedback .
( i also thank daniel margcsy and audiences of the philosophy of science reading group of the institute of philosophy , has rch and of the theoretical philosophy forum , elte for feedback on briefly mentioned parts of this work that got removed at the request of a reviewer . )
this work was supported by the national research , development and innovation office , k 115593 grant .
let @xmath60 and @xmath61 be the incoming velocities and @xmath58 and @xmath120 be the rebound velocities of two spheres of masses @xmath39 and @xmath40 that undergo a perfectly elastic collision . due to conservation of momentum
the velocity of the center of mass is unchanged by the collision : @xmath121 let @xmath49 , @xmath122 ( and @xmath62 , @xmath123 ) denote the incoming ( and rebound ) velocities relative to this center of mass , that is @xmath124 and hence noting that @xmath125 , @xmath126 , @xmath127 etc .
we get @xmath128 with @xmath129 . from ( [ eq_appendix_4 ] ) ,
( [ eq_appendix_5 ] ) , and from @xmath130 , @xmath131 ( which we get from ( [ eq_r_is_unchanging ] ) where we utilized that the collision is perfectly elastic ) we get @xmath132 with @xmath133 and noting that @xmath134 .
multiplying ( [ eq_appendix_14 ] ) with @xmath135 , ( [ eq_appendix_16 ] ) with @xmath136 , subtracting them from each other , and making judicious substitutions for @xmath137 ( eq . ( [ eq_appendix_7 ] ) ) , @xmath138 ( eq . ( [ eq_appendix_9 ] ) ) , and @xmath139 ( eq . ( [ eq_appendix_12 ] ) ) yields @xmath140 with @xmath141 which proves equation ( [ eq_differenceofkineticenergiesafterrebound ] ) .
clearly , @xmath142 .
32 natexlab#1#1[1]`#1 ` [ 2]#2 [ 1]#1 [ 1]http://dx.doi.org/#1 [ ] [ 1]pmid:#1 [ ] [ 2]#2 ( ) . . , _ _ , .
( ) . . , _
( ) . . , _
( ) . volume of _ _ . : .
princeton series in physics . : .
( ) . . : . , , & ( ) . .
in @xcite .
, , & ( eds . )
( ) . . : .
( ) . . , _
( ) . . , _
( ) . . , _
_ , . , & ( ) .
: ( ) . . ,
( ) . . , _
( ) . . , _
( ) . . , _
( ) . . , _
( ) . . , _
( ) . . , _
( ) . . in ( ed . ) ,
_ _ ( pp . ) . : .
( ) . . , _
( ) . . , _
( ) . . : .
( ) . . in ( ed . ) ,
_ _ ( pp . ) . :
in @xcite .
( ) . . , _
in @xcite .
( ) . . , _
( ) . . , _
( ) . . , _
( ) . . in ( ed . ) ,
_ _ handbook of the philosophy of science . : . | we investigate maxwell s attempt to justify the mathematical assumptions behind his 1860 proposition iv according to which the velocity components of colliding particles follow the normal distribution .
contrary to the commonly held view we find that his molecular collision model plays a crucial role in reaching this conclusion , and that his model assumptions also permit inference to equalization of mean kinetic energies ( temperatures ) , which is what he intended to prove in his discredited and widely ignored proposition vi .
if we take a charitable reading of his own proof of proposition vi then it was maxwell , and not boltzmann , who gave the first proof of a tendency towards equilibrium , a sort of h - theorem .
we also call attention to a potential conflation of notions of probabilistic and value independence in relevant prior works of his contemporaries and of his own , and argue that this conflation might have impacted his adoption of the suspect independence assumption of proposition iv . maxwell , kinetic theory , statistical mechanics , normal distribution , independence , h - theorem , second law of thermodynamics , condition a |
we know what a cluster is when we see one , but there is a lot more to clustering than meets the eye . embedded clusters often contain sub - clusters , for example , and clusters generally cluster together themselves into double clusters or star complexes . taken together , there is a hierarchy of young stellar structures , with the objects commonly called clusters representing only the inner and denser parts of the hierarchy .
presumably the main difference between the clusters and the rest of the hierarchy is that the clusters have had sufficient time and gravitational self - attraction to get mixed by stellar orbital motions .
the rest of the hierarchy could partially mix later , by cluster coalescence , for example , or it could disburse through tidal forces . the subclustering of clusters is evident when the clusters are still young in terms of their dynamical time .
examples are ngc 2264 , which has four subclusters with slightly different ages ( @xmath0 my ; dahm & simon 2005 ) , rho oph ( smith et al . 2005 ) , and serpens ( testi et al . 2000 ) .
not all clusters will be born with clear sub - clustering .
if a cloud core has a strong radial density gradient , then primordial gas subclustering can be erased or mixed by tidal forces ( for the same reason that clouds with strong density gradients will not fragment much during free fall collapse ) .
however , cloud cores with strong density gradients are already fairly old when measured in gas dynamical times ; i.e. , they have had at least a crossing time to respond to self - gravitational forces . in these cases ,
the hierarchy of the gas is smoothed over somewhat by dynamical motions even before star formation begins .
nevertheless , the cluster that results is the mixed inner region of the hierarchy , with the mixing occurring mostly in the gas phase rather than the stellar phase in this case with a strong gradient . on larger scales , young stars cluster together in unbound structures , like ob associations .
efremov ( 1995 ) has studied such large scale clustering for 30 years .
most recently , ivanov ( 2005 ) mapped the star complexes in m33 and bastian et al .
( 2005 ) studied cluster complexes in m51 .
there is no characteristic length or mass scale for these larger structures ; the distribution functions for their luminosity and size are power laws ( elmegreen et al .
2006 , and references therein ) .
an ob association has a characteristic size of about 80 pc ( efremov 1995 ) , but this is only because of a selection effect : there is a general correlation between the duration of star formation and the size of the region , and ob associations are selected as concentrations of ob stars .
this limits their age to @xmath1 my and thereby limits their size to the observed 80 pc ( efremov & elmegreen 1998 ) . collections of older stars , such as red supergiants or cepheid variables , are larger , 650 pc according to efremov ( 1995 ) .
collections of younger stars , such as pre - main sequence stars , are smaller ( in embedded clusters ) .
the origin of the correlation is probably turbulence , because the ratio of the size to the age , which is a velocity , correlates with the size like the linewidth - size relation for molecular clouds , having about the same slope and intercept ( elmegreen 2000 ) .
clusters have been defined _ historically _ as small , gravitationally bound , isolated collections of stars ; their density exceeds the local tidal density . by the time the objects appear as `` clusters , ''
the stars in them are mixed along with most previous subclusterings , and the peripheral stars are dispersed .
thus they tend to look isolated and unique .
in fact they are not born that way , they are born as part of a hierarchy of young stellar structures without much distinction other than their location in the high density parts .
simulations of collapsing clouds by bonnell , bate & vine ( 2003 ) show hierarchical star formation with clusters in several dense cores .
these clusters mix together over time making larger clusters .
the observations do not yet clearly distinguish between clouds like these that are collapsing freely and clouds that have some level of radial stability resulting from magnetic forces or isotropic turbulent motions .
quasi - stability seems favored by observations of cloud core self - absorption where the absorption line redshift is only @xmath2% of the linewidth ( myers et al .
such small redshifts suggest that cloud cores are contracting relatively slowly , on a time scale of several radial - crossing times .
however , recent observations of global infall in ngc 1333 ( walsh , bourke , & myers 2006 ) suggest dynamical processes on shorter times , similar to those envisioned by bonnell , bate & vine ( 2003 ) . in either case
, the stars should form in a hierarchical way ( sect .
[ sect : hier ] ) .
interstellar gas has a similar hierarchy of structures .
power spectra of gas emission show power laws , indicating no characteristic scale ( dickey et al .
the power spectra of optical light in a galaxy is about the same as the power spectrum of interstellar gas ( elmegreen et al . 2003 ) .
these results suggest that the stars follow the gas when they form , although extinction may contribute to the optical power spectrum too . in terms of cloud - like objects ,
the highest level in the hierarchy , on the largest scale , consists of giant atomic clouds that contain @xmath3 m@xmath4 of gas .
these are evident in the milky way ( mcgee & milton 1964 ; elmegreen & elmegreen 1987 ) and in local galaxies where hi maps have the resolution needed to see them ( e.g. , boulanger & viallefond 1992 ) .
molecular clouds are the next denser level in the hierarchy of structures , and they often collect together inside the atomic clouds if the ambient pressure is large enough to allow molecules to form ( elmegreen 1993 ; blitz & rosolowsky 2006 ) .
maps in grabelski et al .
( 1987 ) show this gmc collection clearly for the milky way . inside giant molecular clouds
are cloud cores , and so on down to the smallest scales that can be observed .
it is worth making the cautionary note that this hierarchical structure for both gas and stars does not mean every small object is inside a larger one .
small objects can be to the side of large objects , with no obvious larger scale structures around them .
most objects do contain substructures , however , down to or below the scale where thermal motions dominate ( or random stellar motions dominate in the case of stellar clustering ) .
an example of this type of structure is illustrated in figure 1 , below .
individual stars form at the bottom of the hierarchy , where the gas is in the form of dense clumps . as we consider lower and lower levels in the hierarchy , i.e.
, as the average gas density gets higher , the mass fraction in the form of the dense star - forming clumps goes up . on the scale of the clumps themselves , the mass fraction is unity . on the scale of the gmc ,
the mass fraction is small because there is a large volume of low - density gas between the cores and even more between the clumps .
this is the nature of the hierarchy , which is fractal .
stars form inside each clump at some average efficiency , which may be about constant , perhaps one - third or one - half ( e.g. , matzner & mckee 2000 ) .
thus the overall efficiency of star formation , which is the mass fraction of a cloud that turns into stars , increases towards higher gas density inside the cloud .
this is why the efficiency in a cloud core may be 20% or 30% , but the efficiency for a whole ob association is only 5% .
the ob association is the result of star formation inside the densest clumps of a gmc , but there is a lot of low - density intercore and interclump gas in a gmc that does not form stars .
we see now another aspect of cluster formation : in the dense regions of the hierarchy where clusters form ( gmc cores ) , the mass fraction in the form of individual star - forming clumps is automatically high , so the efficiency of star formation is automatically high .
this means that the stars have a high probably of ending up gravitationally bound together .
high average density and boundedness go together ; one follows from the other in a hierarchical interstellar medium .
the high efficiency needed for bound cluster formation is not the result of special circumstances related to the stars themselves , such as feedback processes , but only the result of hierarchical gas structure . at the density that is high enough to form a compact region of stars , one that stands out to the eye as a star cluster , the efficiency of star formation is automatically high . putting these concepts together , we can get most of the way toward explaining the origin of star clusters : they are the inner mixed regions of the young - star hierarchy , where the dynamical time is short and the age is comparable to or longer than this dynamical time ( to ensure mixing ) . because most star formation occurs in only a few dynamical times , a high fraction of the clumps will have formed stars by the time mixing has occurred , and because the clumps represent a high fraction of the local gas mass in a cloud core , the efficiency of star formation is generally high there .
then the mixed stellar regions remain gravitationally bound after the gas leaves .
thus bound clusters are an inevitable result of star formation in hierarchically structured gas for ism properties like those in the solar neighborhood .
this initial binding does not mean that young clusters stay bound for long .
ninety percent lose a high fraction of their stars within the first 10 my ( lada & lada 2003 ; fall , chandar & whitmore 2005 ) , leaving only small bound cores ( e.g. , kroupa , aarseth & hurley 2001 ) or perhaps no bound cores at all .
some giant star forming regions , such as ngc 604 in m33 , appear to have no clusters or cluster remnants , as if all star formation were isolated or initially unclustered .
maz - apellniz ( 2001 ) studied several other super - ob associations of this type .
such diversity in initial clustering follows from the hierarchical model if the average cloud density varies .
stars presumably form in similar cores everywhere , with pre - collapse densities of @xmath5 @xmath6 , but where the average density is low because of a low pressure ism , for example , then the mass fraction of star - forming clumps can be low even in the densest regions of molecular clouds ( which are not very dense in this case ) .
stars will still formed clustered , and the efficiency will still be highest in the core regions , but this peak efficiency may be only @xmath1% rather than @xmath7% .
in that case there is not enough gravitational binding from stars alone to make a bound cluster when the gas leaves .
thus the hierarchical model predicts that the fraction of stars forming in initially bound clusters should depend on the local ism pressure or density .
this prediction is consistent with the observation by larsen & richtler ( 2000 ) , who found that the fraction of uv light from stars in clusters increases with star formation rate , considering that star formation rate increases with gas column density and therefore ism pressure .
the origin of hierarchical structure in interstellar gas is probably a combination of gravitational fragmentation and turbulence compression .
both produce scale - free density structures in regions where the total energy density is much larger than the thermal pressure .
this inequality holds for most of the diffuse interstellar medium and much of the self - gravitating ism because collisional cooling rates are high enough , and background heating rates are low enough , that the temperatures and thermal pressures are usually very low .
motions from various kinematic energy sources are then supersonic , and the mixture of these motions , particularly with the velocity - size correlation that results , produces correlated density structures . in the high density parts of these structures , where self - gravity exceeds the background tidal force and the high column density
provides some shielding from both radiative and kinematic energies , the gas has time and freedom to collapse into stars .
the correlated density ensures that most of these collapses will be clustered together , in patches of various size , and it also provides the self - gravity that forces orbital mixing and self - binding to make a homogeneous embedded young cluster . as mentioned in section [
sect : hier2 ] , not every cloud is surrounded by a larger cloud in a russian doll pattern .
some small clouds are to the side of larger clouds and others may be between large clouds , with no obvious connection to one or the other . when clouds get pushed around by directed stellar pressures , the structures that form by gravity and turbulence get modified and cloud pieces can scatter anywhere . in this sense , it is not uncommon for stars to form in isolation or in small unbound groups that are far from clusters and associations
. it appears for our galaxy that the fraction of stars forming in this isolated fashion is small , perhaps 10% or less ( carpenter 2000 ) .
recent spitzer space telescope observations of the perseus cloud suggest that @xmath8% of class i protostars are outside the massive cores ; many of these could still be in low - mass subclusters ( jorgensen et al .
2006 ) .
figure 1 shows a fractal brownian motion model of a face - on galaxy compared to observations of the cluster size function in the galaxy ngc 628 . on the left , a grayscale image of a model galaxy
is shown , with four gray levels corresponding to different ratios of the density to the peak density ( this figure is explained more completely and shown in color in elmegreen et al .
the darkest regions around the edges have the lowest projected densities , between 0.3 and 0.4 of the peak , the light gray regions inside of these have projected densities between 0.4 and 0.5 of the peak , whitish regions are between 0.5 and 0.6 , and the dark gray regions inside the white regions have the highest densities , greater than 0.6 times the peak .
the model is initialized in a 3d cube where the density probability distribution function is log - normal and the 3d power spectrum of density is approximately a power law with a slope of @xmath9 .
this power spectrum is comparable to that for velocity and passive scalar density in kolmogorov incompressible turbulence ( it is also the best fit to the data compared to other power spectra that vary by @xmath0 in slope ) .
the cube is then multiplied by a gaussian on the line of sight in order to simulate a projected galaxy disk .
the panel on the right of figure 1 shows cumulative size distribution functions for connected regions , or `` clouds '' that were found objectively , for the same 4 levels of density threshold relative to the peak .
the lowest curve ( i.e. , having the smallest count ) is for the highest density threshold , greater than 0.6 times the peak .
the crosses are b - band observations of the size distribution function for star - forming regions in an hst image of ngc 628 ( with arbitrary shifts in each axis to fit the model ) .
the physical scale ranges from 5 pc to 155 pc .
the size distribution is a power law with a slope of @xmath10 on this cumulative plot .
the models match the observations .
this cumulative size distribution corresponds to the differential function @xmath11 .
the size distribution for loose stellar groupings in ngc 628 is approximately a power law from 2.5 pc to 150 pc ( elmegreen et al .
this size range is larger than that for bound clusters , so it is interesting to ask whether the size distribution for loose groupings is a continuation of the size distribution for dense clusters .
there is preliminary evidence for this from the cluster size distribution in bastian et al .
( 2005 ) , who studied m51 . they derived a size function @xmath12 , which is similar to that on larger scales .
the analogy between unbound groupings and dense clusters is not straightforward , however , because the unbound groupings have a mass that increases with size approximately as @xmath13 ( elmegreen et al .
this mass - size relation , along with the size distribution function , is consistent with a hierarchical star distribution with projected fractal dimension @xmath14 .
that is , for a fractal , we expect @xmath15 and @xmath16 ( mandelbrot 1983 ) .
the problem with this is that the mass depends very little on radius in the dense clusters studied by bastian et al .
the lack of a correlation between mass and radius is also present in the data in testi , palla & natta ( 1999 ) and larsen ( 2004 ) .
there are only a few observations of the mass distribution functions of clusters .
there are many observations of luminosity distribution functions , but mass distribution functions are more difficult to obtain because they require the additional knowledge of cluster ages .
the complete mass function also requires the conversion between luminosity and mass as a function of age .
here we define the negative slope of the cluster mass function to be @xmath17 , so @xmath18 is proportional to the number of clusters with masses between @xmath19 and @xmath20 .
a histogram of the mass function plotted in equal intervals of @xmath21 , which is the most common way to display the mass function , would then have a negative slope of @xmath22 .
the most complete cluster sample for a nearby galaxy is in the large magellanic clouds .
elson & fall ( 1985 ) measured the cluster luminosity function in the lmc and got @xmath23 for a mixture of ages .
elmegreen & efremov ( 1996 ) determined the luminosity functions of lmc clusters in four age intervals and found them all to be about @xmath24 . as the age was constant for each function , mass is proportional to luminosity .
this means the mass distribution function is also a power law with the same slope , @xmath25 , giving @xmath26 .
a similar study of lmc clusters by hunter et al . ( 2003 ) found @xmath27 , while de grijs & anders ( 2006 ) obtained @xmath28 using the same lmc data as in hunter et al .
but different age calibrations .
cluster luminosity functions in the milky way have been determined for a long time , but not mass functions .
one of the first modern determinations was by battinelli et al .
( 1994 ) , who derived a mass - luminosity relation for nearby milky way clusters and used this to get a mass function slope of @xmath29 for the high mass end of the cluster mass function .
battinelli et al .
also obtained @xmath30 for clusters in lynga s catalogue . for the antenna galaxy , mass function slopes were found by zhang & fall ( 1999 ) to be @xmath31 for young clusters and @xmath32 for old clusters . for m51 , cluster mass functions with @xmath26 and various upper mass cutoffs
were fitted to the observed luminosity functions in 3 hst passbands . similarly in ngc 3310 , @xmath33 and in ngc 6745 , @xmath34 ( de grijs et al .
these observations all imply that the cluster mass function is a power law with a slope @xmath17 within several tenths of the value @xmath35 . the same slope @xmath26 is implied by galaxy - wide imfs .
summed imfs from clusters can produce a global imf that is the same as the individual cluster imf only if the cluster mass function slope is @xmath26 or shallower .
observations suggest that galaxy - wide imfs are indeed the salpeter imf , not much steeper , based on metallicity , color , h@xmath36 equivalent width , and color - magnitude diagrams ( elmegreen 2006a ) .
this is the case for the milky way bulge , dwarf galaxies , large galaxies , galaxy clusters , and the whole universe .
because cluster imf slopes average the salpeter value too ( scalo 1998 ) , the summed imf equals the cluster imf and the cluster mass function has to be close to @xmath37 .
even a slightly steeper value of @xmath38 makes the summed imf from clusters significantly steeper than the observed imf of composite populations ( kroupa & weidner 2003 ; weidner & kroupa 2005 ) .
the cluster mass function is consistent with the theory of cluster formation discussed in the previous section .
a hierarchical gas distribution , forming clusters at arbitrary levels in the hierarchy with a constant efficiency of star formation , has a mass function slope of exactly @xmath26 ( fleck 1996 ) .
this can be seen from a simple tree model where there is one trunk with a mass of n=16 , two branches off the trunk with masses of 8 each , two branches off each branch with masses of 4 each , two more branches off each of the previous with a mass of 2 each , and 16 total branches at the top of the tree with a mass of 1 each .
the product of the number of branches times the mass of each branch is 16 for all levels in the hierarchy .
there are @xmath39 possible branching points , each of which may be viewed as a possible cluster with all of the smaller branches inside of that cluster .
if these 15 branching points are randomly sampled , then the probability that the chosen point is the trunk , giving a `` cluster '' mass of 16 , is @xmath40 .
the probability the chosen mass is 8 is @xmath41 , the probability it is 4 is @xmath42 , and so on .
considering the log of the mass in base 2 , the probability that the log is 4 , so the mass is @xmath43 , is @xmath40 .
the probability the log is 3 is @xmath41 , etc .. in general , the probability that the log of the mass is @xmath44 is @xmath45 , or , more generally , @xmath46 .
this probability is proportional to the count of branching points , and for the log mass expressions , the counting is for equal intervals of log mass ( @xmath47 ) .
thus we have a probability function @xmath48 converting @xmath44 on the right - hand side back to mass , we get @xmath49 this equation shows that the mass distribution function of tree branches is proportional to @xmath50 for equal log intervals of mass , in which case @xmath51 so @xmath26 .
snchez , alfaro , & prez ( 2006 ) show that the slope gets slightly shallower if blending effects are considered .
a similar result has been obtained for a smooth gas density distribution made by the fractal brownian motion technique ( sttzke et al .
1998 ; elmegreen 2002a ; elmegreen et al . 2006 ) .
the mass function of three - dimensional clumps depends on the contour level used to define the clump and is steeper for denser levels .
it is also steeper for steeper intrinsic power spectra ( sttzke et al 1998 ) .
it varies from @xmath52 for low density to @xmath53 for high density when the fractal has a power spectrum with a power law slope equal to that of a passive scalar in incompressible kolmogorov turbulence , and when the density has a log - normal probability distribution function .
this variation in mass function slope with density is consistent with the observation that the mass function for giant molecular clouds is a shallow @xmath54 to 1.8 , and the mass function for clusters is steeper , @xmath55 to 2.1 .
the difference is presumably because whole gmcs sample a lower density in the ism than the star clusters they produce in their cores .
an important question is whether massive stars can form in isolation , either in the remote field or on the peripheral regions of clusters .
if massive stars need the cluster environment to accrete dense gas in a certain way , or to coalesce with other protostars , then there should not be many forming in low density regions ( testi , palla , & natta 1999 ) .
of course , it is understood that all stars form in dense clumps , so the local environment is never low density , but the question is whether each massive star has a full complement of lower mass stars in the immediate neighborhood , filling out the imf toward lower mass .
an isolated dense clump could , in principle , form an isolated massive star without the thousands of other stars expected from the usual imf .
the discussion in the previous section suggested that each logarithmic interval of cluster mass produces the same total stellar imf , regardless of the cluster mass itself .
this is how the summed imf can equal the individual cluster imf : low mass clusters do not tip the balance toward exclusively low mass stars , for example ( elmegreen 2006a ) .
the implication is that stars of any mass can form in clusters of any mass ( provided the cluster mass exceeds the stellar mass ) .
one condition for this to be true is that there is a universal imf .
if two or more star formation processes operate differently in different regions and if one tends to produce more high mass stars , then there are effectively two or more imfs that have various contributions to the total depending on the environment ( elmegreen 2004 ) .
if the important environmental variable is density , for example , then low density clusters can have a slightly different imf than high density clusters . with such dichotomy
, it would no longer be true that stars of any mass form with equal probability in clusters of any mass ( or density ) .
a test of these two possibilities is the `` 100-taurus '' test .
for a universal imf and @xmath27 , the summed imf from 100 separate regions like the low mass taurus clouds should have a slope that is only @xmath56 steeper than the imf from one large region like orion , which contains 100 times the mass of taurus .
observations of imfs in many low mass or low density regions like taurus could settle this issue directly .
if a given number of stars in taurus - like regions have an imf that is significantly steeper than the imf from the same number of stars in orion , or in the orion trapezium cluster , then there would seem to be at least two distinct imfs , and massive stars would be favoring the higher mass and denser clusters . at the moment
, stars of any mass appear to form in clusters of any mass .
one can then think of a cluster as `` randomly '' sampling a universal imf . for purely random sampling , there is a very small chance that a massive star will form in a low - mass cluster , but most massive stars form in high mass clusters because most stars of all types form in high mass clusters .
for the power - law part of the imf , the maximum likely star mass out of n clusters of mass m equals the maximum likely star mass in one cluster of mass nm ( elmegreen 2006a ) .
these statements are consistent with the observations by oey , king & parker ( 2004 ) , who found the distribution of o - star counts per cluster in the lmc to be a power law with @xmath27 down to a single o star .
it is also consistent with a stronger statement by de wit et al .
( 2005 ) , who found that the total star mass in o - star containing clusters in the solar neighborhood is a power law with @xmath57 down to a single o star ( not just an o - star cluster containing other stars ) .
this power law is similar to that for whole clusters , suggesting that sometimes a single o star can form in place of a cluster that has many smaller stars but the same total mass .
the de wit et al .
observation is consistent with 4% of o - type stars forming in isolation or in peripheral regions of clusters .
there are actually many examples where o - type stars form along the periphery of massive dense clusters .
sequential triggering has this effect , and in the 30 dor region of the lmc it is particularly clear ( walborn et al .
1999 ; see plot of massive stars in elmegreen 2006b ) .
there are several reasons why clusters eventually come apart .
gas dispersal in the first 1 to 3 my leads to decreased gravitational binding for the initial stellar speed , and this causes some stars to escape directly , in a crossing time or less ( lada , margulis , & dearborn 1984 ) .
many ob associations and loose stellar groups could have been collections of embedded clusters several million years earlier .
kroupa , aarseth & hurley ( 2001 ) studied cluster expansion after rapid gas loss with a star formation efficiency of 30% .
they found that as the cluster expands to a new equilibrium radius , some stars are lost quickly and some remain in a bound core .
they suggested that a cluster like the trapezium cluster in orion can turn into an open cluster like the pleiades after an amount of time has passed that is comparable to the age of the pleiades .
mass is also lost from a cluster in the form of stellar winds and supernovae during stellar evolution .
it takes much longer for a significant mass to be lost by stellar evolution than by cloud core disruption .
after a few tens of millions of years , the decreasing cluster mass produces an overall expansion ( terlevich 1987 ) .
eventually , the cluster density gets so low that self - gravitational forces become comparable to or less than the background tidal force
. then the cluster disperses .
this is the third mechanism of cluster destruction .
tidal interactions with dense clouds , spiral arms , the bulge , and the galactic disk all give a lower limit to cluster density for survival .
the outer parts of the cluster are shed into a tidal tail at the radius of this density threshold . as the cluster density decreases , the tidal radius decreases too , so the bound part of the cluster shrinks while the stars in the outer parts expand .
detailed models of cluster disruption including evolution and tidal effects are in baumgardt & makino ( 2003 ) .
other talks at this conference consider cluster destruction in greater detail .
in particular , the destruction of clusters by giant molecular clouds has recently been shown to be important by gieles et al .
( 2006 ) . for molecular clouds ,
the important quantity is the volume filling factor of molecular material with a density comparable to or greater than the cluster density .
when a cluster enters this volume , it becomes tidally unbound for a time .
movement near these dense regions can energize the stellar orbits , leading to eventual destruction .
more distant encounters with molecular clouds are less important than those with impact parameters comparable to the cloud radius .
intermediate mass clusters may be disrupted by only a few gmc encounters , while low and high mass clusters require many encounters .
the destruction time for a cluster of mass @xmath19 through multiple encounters was found to be @xmath58 where @xmath59 and @xmath60 is the power in the cluster mass - radius relation , @xmath61 .
the typical mass column density of a molecular cloud is @xmath62 and the ism - averaged density of gmcs is @xmath63 .
gieles et al .
( 2006 ) suggest @xmath64 and @xmath65 .
thus @xmath66 .
a similar mass dependence for the destruction time of a cluster has been found in other studies .
boutloukos & lamers ( 2003 ) found the mass - dependent disruption time in four galaxies to be @xmath67 .
gieles et al .
( 2004 ) showed that models by baumgardt & makino ( 2003 ) were consistent with @xmath68 .
de la fuente marcos & de la fuente marcos ( 2004 ) used dynamical models to suggest @xmath69 .
most recently , lamers et al .
( 2005 ) compared @xmath70 from numerical experiments using this power law form for @xmath71 and the simple model @xmath72 .
the model and analytical results were in excellent agreement for @xmath73 .
if @xmath74 depends on cluster mass , then the mass function of clusters should get shallower at low mass over time because the lowest mass clusters get destroyed soonest . after a while
, the slope of the cluster mass function should approach @xmath75 at low mass ( lamers et al .
recall that @xmath17 is defined to be the negative value of the slope on a plot of the cluster mass spectrum in linear mass intervals , and @xmath22 is the slope on a spectrum in log mass intervals .
a value of @xmath76 implies that a histogram of cluster number in equal intervals of log mass should increase with @xmath21 at low @xmath19 , reach a peak , and then decrease with @xmath21 as @xmath77 for sufficiently large mass where destruction has not been significant yet .
most cluster mass functions plotted with @xmath21 intervals are indeed peaked like this , with a rising part at low @xmath21 , but this is always attributed to magnitude limits in the observations .
that is , the turnover toward low mass is the result of missing clusters that are present but too faint to discern .
for example , de grijs & anders ( 2006 ) plotted histograms of cluster counts in @xmath21 intervals for the lmc and found continuous power law functions down to and below @xmath78 m@xmath4 for ages up to @xmath79 years ( this was also the case in elmegreen & efremov [ 1996 ] and hunter , et al . [
there is no evidence for a mass - dependent destruction in these data .
in contrast , fall , chandar & whitmore ( 2005 ) suggest that the destruction rate is independent of cluster mass , and that the mass function stays constant over time , which is consistent with the observation by de grijs & anders ( 2006 ) and also with observations of the antennae galaxy by fall et al .
fall et al .
find that the number of clusters in intervals of equal age decreases inversely with the age , independent of cluster mass .
this decrease is somewhat continuous over time all the way from @xmath80 years to @xmath81 years .
this is a surprising result because there is not even a feature in this trend where the mechanism of cluster disruption is expected to change from gas expulsion to stellar evolution . in the fall
model , we can write the number of clusters more massive than @xmath82 as @xmath83 . if the observations also suggest that the form of the mass function does not change over time and is @xmath84 for linear intervals of @xmath19 , then @xmath85 for constant lower detection limit @xmath82 .
thus @xmath86 , the maximum cluster mass is @xmath87 , and each cluster has its mass decrease as @xmath88 .
we also obtain that @xmath89 from the relations @xmath90 , or @xmath91 .
why is there a difference between the fall et al .
( 2006 ) model and the lamers et al .
( 2005 ) model ?
the key observation to distinguish between these two models is the slope of the cluster mass function versus time .
the ism and star birth positions appear scale - free from the scale of the jeans length in the ambient ism ( on a kpc scale ) to below the star formation scale .
this scale - free distribution is not likely to be continuous in any one region , but still present on average , and present in a piecewise sense .
stars form in dense clumps whose local mass fraction compared to the surrounding molecular material is high and whose mixing time is relatively short .
the mixed young stars are the `` cluster . ''
gas removal stops star formation in the dense core but not everywhere .
it often continues on the periphery of the cluster and in the remaining molecular cloud as a result of sequential triggering .
this basic model explains the cluster mass spectrum , the galactic star formation rate and star formation timescale ( not discussed here ; see elmegreen 2002b ) , and the observed stellar grouping structures .
it does not give the independence between cluster mass and size , however , which presumably involves additional physical processes .
cluster mass loss is dominated by gas disruption and stellar evolution at first , by tidal shredding from giant molecular clouds and spiral density waves after a while , and by thermal disruption on the longest time scale .
the disruption time varies with cluster mass , but the exact scaling relationship is subject to debate .
observations suggest that the power - law form of the cluster mass distribution function is preserved for at least @xmath81 years in some galaxy disks , in which case the disruption time is either longer than this or it does not scale positively with mass in a noticeable way .
gieles , m. , baumgardt , h. , bastian , n. , lamers , h. j. g. l. m. 2004 , in the formation and evolution of massive young star clusters , asp conference series , vol . 322 .
lamers , l.j .
smith , & a. nota , san francisco : astronomical society of the pacific , p. 481 | clusters are the dense inner regions of a wide - spread hierarchy of young stellar structures .
they often reveal a continuation of this hierarchy inside of them , to smaller scales , when they are young , but orbital mixing eventually erases these subparts and a only smooth cluster or smooth unbound group remains .
the stellar hierarchy follows a similar structure in the interstellar gas , which is presumably scale - free because of supersonic motions in the presence of turbulence and self - gravity .
the efficiency of star formation increases automatically with density in a hierarchical ism , causing most dense stellar groups to be initially bound for local conditions . in lower pressure environments ,
the infant mortality rates should be higher . also following from hierarchical structure
is the cluster mass distribution function and perhaps also the cluster size distribution function , although the predicted mass - size relation is not observed .
cluster destruction is from a variety of causes .
the destruction time should depend on cluster mass , but the various groups who have studied this dependence have gotten significantly different results so far . |
_ stellar stratification _ is a characteristic phenomenon of star clusters , which has been well documented for more than 50 years .
it is directly linked to the dynamical evolution of star clusters @xcite , and specifically to the central concentration of the massive stars as the cluster relaxes dynamically , a phenomenon known as _ mass segregation _ ( for detailed reviews see * ? ? ?
* ; * ? ? ?
this behavior in massive galactic globular clusters ( ggcs ) is well understood and the observed mass segregation is explained as a result of their dynamical relaxation in a timescale , @xmath1 , much longer than their ages , @xmath2 , as derived from stellar evolutionary models . this _ dynamical _ mass segregation , thus , forces more massive stars to sink inwards to the center of a cluster through weak two - body interactions .
however , segregation of massive stars is observed also in young star clusters , such as the @xmath3 30 myr - old cluster ngc 330 in the small magellanic cloud @xcite , or even the @xmath3 1 myr - old galactic starburst ngc 3603 @xcite .
such systems , supposedly not sufficiently old to be dynamically relaxed , can not be included in the `` traditional '' picture of dynamical mass segregation as it occurs in ggcs .
for these cases the phenomenon has been rather tentatively characterized as _ primordial _ mass segregation , based on the suggestion that this stellar stratification is due to the formation of the stars in , or near , the central part of the clusters rather than being dynamically driven mass segregation at early stages of clustered star formation is indeed predicted by theoretical studies , which suggest that the positions of massive stars for rich young clusters of @xmath4 can not be the result of dynamical evolution @xcite .
moreover , the appearance of gas in such clusters reduces further the efficiency of any dynamically driven mass segregation . according to theory ,
massive protostars can be formed _ at the central part _ of the protocluster through cohesive collisions and dissipative merging of cloudlets , which lead to extensive mass segregation and energy equipartition among them @xcite , or through accretion from a distributed gaseous component in the cluster , which leads to mass segregation since stars located near the gravitational center of the cluster benefit from the attraction of the full potential and accrete at higher rates than other stars @xcite .
both of the aforementioned massive star formation mechanisms predict primordial mass segregation through early dynamical processes at the central part of the protocluster .
the significant difference of these processes from the subsequent energy equipartition of the stars in the formed cluster is that the latter , and consequently dynamical mass segregation , is independent of the initial cluster conditions , while the former , and therefore primordial mass segregation , is directly connected to these conditions . in order to achieve a complete understanding of the phenomenon of primordial mass segregation ,
many related studies are focused on clusters at very early stages of their formation , including a variety of types of stellar systems from loose stellar associations and open clusters to compact massive starbursts . the magellanic clouds ( mcs ) , being extremely rich in young star clusters ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , offer an outstanding sample of intermediate type of stellar systems , specifically compact young star clusters , which are not available in the galaxy . in such clusters
primordial mass segregation can be sufficiently observed from ground - based observations ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and they are also suitable for more detailed studies of this phenomenon from space with the high - resolving efficiency of the _ hubble space telescope _ ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
such studies contribute significantly to the debate about the phenomenon of primordial mass segregation and its origin which which is still ongoing ( see , e.g. , * ? ? ?
it is interesting to note that while the existence of the phenomenon itself is being questioned @xcite , new methods are still being designed for its comprehensive identification and quantification ( e.g. , * ? ? ?
* ) . in the first part of our study of stellar stratification we developed and tested a new robust diagnostic method aimed at establishing the presence ( or not ) of this phenomenon in a cluster ( * ?
* from here on paper i ) .
this method is based on the calculation of the mean - square radius , the so - called _ effective radius _ , of the stars in a cluster for different magnitude ranges , and the investigation of the dependence of the effective radii of each stellar species on magnitude as indication of stellar stratification in the cluster . , scaledwidth=45.0% ] stellar
stratification affects the presently observable physical properties of the star cluster , and thus diagnostic tools for the identification of the phenomenon are based on the detection of mass- or brightness - dependent changes of such properties .
previously established methods involve ( i ) the study of the projected stellar density distribution of stars of different magnitudes ; here , the differences in the exponents of the slopes fitted are seen as proof of mass segregation ( e.g. , * ? ? ?
* ) , ( ii ) the radial dependence of the luminosity or mass function ( lf , mf ) slope ; a gradient of this slope outwards from the center of the cluster indicates stratification ( e.g. , * ? ? ?
* ) , and ( iii ) the variance of the core radius of the cluster as estimated for stars in specific magnitude ( or mass ) groups ; a trend of this variance with brightness ( or mass ) signifies stratification ( e.g. , * ? ? ?
while these methods can provide useful information about the segregation in a cluster , all of them are sensitive to model assumptions , since they strongly depend on data fitting for parametrizing stellar stratification ; i.e. , functional expressions for the surface density profiles , or the mfs and lfs are required for the derivation of their slopes and core radii .
our method , developed in paper i for assessment of stellar stratification in star clusters , is based on the notion of the dynamically stable _ spitzer radius _ of a star cluster , defined as the mean - square distance of the stars from the center of the cluster ( e.g. , * ? ? ?
the observable counterpart of this radius , the effective radius , is given by the expression : @xmath5 where @xmath6 is the projected radial distance of the @xmath7th stellar member of the cluster in a specific brightness range and @xmath8 the corresponding total number of stars in the same brightness range .
different calculations of the effective radius are performed , each for stars in different brightness ranges that cover the whole observed stellar luminosity function of the cluster .
the simple application of this method can yield direct information about the spatial extent ( radial distribution ) of stellar groups in different magnitude ( mass ) ranges . in a star cluster , where stratification occurs
, the segregated brighter stars are expected to be more centrally concentrated and the corresponding effective radius should be shorter than that of the non - segregated fainter stars . as a consequence , stellar stratification can be observed from the dependence of the effective radius of stars in specific magnitude ( mass ) ranges on the corresponding mean magnitude ( mass ) .
this method is sufficiently described and tested in paper i , where we show that it performs efficiently in the detection of stellar stratification , provided that the incompleteness of the photometry has been accurately measured and the contamination by the field population has been thoroughly removed .
our diagnosis method is also independent of any model or theoretical prediction , in contrast to those used thus far for the detection of mass segregation .
ccccccc [ t : obs ] ngc 1983 & 05@xmath927@xmath104492 & @xmath116859070 & f555w & j8ne68req & 20 & october 7 , 2003 + & & & f814w & j8ne68riq & 20 & october 7 , 2003 + ngc 2002 & 05@xmath930@xmath102080 & @xmath116655023 & f555w & j8ne70b5q & 20 & august 23 , 2003 + & & & f814w & j8ne70b7s & 20 & august 23 , 2003 + ngc 2010 & 05@xmath930@xmath103489 & @xmath117049083 & f555w & j8ne71r5q & 20 &
october 7 , 2003 + & & & f814w & j8ne71r9q & 20 & october 7 , 2003 + in this second part of our study of stellar stratification we apply the effective radius method on deep imaging data obtained with the wide - field channel ( wfc ) of the _ advanced camera for surveys _ ( acs ) onboard the hubble space telescope ( hst ) of three young star clusters in the large magellanic cloud ( lmc ) for the investigation of primordial mass segregation in them .
young clusters in the magellanic clouds are quite different to each other , so that their study requires the detailed treatment of each cluster individually ( see , e.g. ,
[ s : dynamics ] , [ s : starcont ] ) , and therefore a statistical investigation of the phenomenon of mass segregation in a large sample of such clusters is not yet meaningful .
the purpose of the present study is the detailed investigation with the use of a consistent method of individual selected clusters , in order to establish a comprehensive comparative framework for the assessment of mass segregation in young lmc clusters .
the objects of interest , clusters ngc 1983 , ngc 2002 and ngc 2010 , are selected among others also observed with acs due to their youthfulness and the variety in their characteristics .
this selection is made in order to include in this comparative study clusters , which are very different from each other . in
[ s : clussampl ] we describe the considered sample of lmc clusters and the corresponding data sets , as well as their reduction and photometry .
we study the dynamical status of the clusters and discuss their structural parameters in
[ s : dynamics ] , and in [ s : starcont ] we present the observed stellar populations , the decontamination of the stellar samples from the contribution of the local background field of the lmc and the stellar content of the clusters .
the application of the diagnostic method is performed for all three clusters in [ s : methappl ] , where we also develop a comparative scheme for the quantification of stellar stratification among different clusters , and we discuss our results .
we conclude on them in [ s : concl ] . , scaledwidth=45.0% ]
the present study deals with stellar stratification in rich lmc star clusters .
such clusters are known to differ from each other in terms of , e.g. , their structural parameters , ages , luminosities and masses @xcite . since we focus on the phenomenon of primordial mass segregation we limit our sample to young lmc clusters , and in order to use the most complete available data we select clusters observed with acs / wfc .
the advantages introduced to crowded - field photometry of mcs star clusters by the use of acs , and specifically its wide - field channel , have been documented extensively by studies based on such observations ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
consequently , we select three young lmc clusters , ngc 1983 , ngc 2002 and ngc 2010 , observed with acs / wfc .
these clusters are different from each other , as far as their appearance , structure and stellar content is concerned . considering that we also wish to investigate the identification and quantification of the phenomenon of stratification in various cluster - hosting environments , we based the selection of these specific clusters on the on the differences between their local surrounding field star populations .
we discuss in detail the differences of the clusters themselves in their structure and dynamical behavior in
[ s : dynamics ] , and in their stellar content in [ s : starcont ] .
the decontamination of the true cluster populations from the contribution of the general background field of the lmc , which determines the actual stellar content of the clusters and the stellar sample to be used in our method for assessment of stellar stratification in the clusters is performed in [ ss : fldsub ] .
our observations come from hst program 9891 , a snapshot survey of @xmath12 rich star clusters in the magellanic clouds .
imaging was obtained in cycle 12 using acs / wfc . as snapshot targets ,
each cluster was observed for one orbit only , resulting in a single exposure through the f555w filter and another through the f814w filter .
details of the observations for the clusters considered in this paper may be found in table [ t : obs ] .
the acs wfc consists of a mosaic of two @xmath13 pixel ccds covering an area of approximately @xmath14 at a plate scale of @xmath15 arcsec per pixel .
the two ccds are separated by a gap roughly @xmath16 pixels wide .
the core of each cluster was placed at the center of wfc chip 1 .
this ensured that the inter - chip gap did not interfere with the main part of the cluster , and meant that each image reached a maximum radius of approximately @xmath17 from the cluster center .
a small dither was made between the two exposures of a given target , to help facilitate the removal of cosmic rays and hot pixels ; however with only two images per cluster it was not possible to eliminate the gap in spatial coverage due to the inter - chip separation .
the data products produced by the standard stsci reduction pipeline , which we retrieved from the public archive , have had bias and dark - current frames subtracted and are divided by a flat - field image . in addition , known hot - pixels and other defects are masked , and the photometric keywords in the image headers are calculated .
we also obtained distortion - corrected ( drizzled ) images from the archive , produced using the pyraf task multidrizzle .
we used the dolphot photometry software ( e.g. , * ? ? ?
* ) , specifically the acs module , to photometer the flatfielded ( but not drizzled ) f555w and f814w images .
dolphot performs point - spread function ( psf ) fitting using psfs especially tailored to the acs camera . before performing the photometry
, we first prepared the images using the dolphot packages acsmask and splitgroups . respectively , these two packages apply the image defect mask and then split the multi - image stsci fits files into a single fits file per chip .
we then used the main dolphot routine to simultaneously make photometric measurements on the pre - processed images , relative to the coordinate system of the drizzled f814w image .
we chose to fit the sky locally around each detected source ( important due to the crowded nature of the targets ) , and keep only objects with a signal greater than @xmath18 times the standard deviation of the background .
the output photometry from dolphot is on the calibrated vegamag scale of @xcite , and corrected for charge - transfer efficiency ( cte ) degradation . to obtain a clean list of stellar detections with high quality photometry
, we applied a filter employing the sharpness and `` crowding '' parameters calculated by dolphot .
the sharpness is a measure of the broadness of a detected object relative to the psf for a perfectly - fit star this parameter is zero , while it is negative for an object which is too sharp ( perhaps a cosmic - ray ) and positive for an object which is too broad ( e.g. , a background galaxy ) .
the crowding parameter measures how much brighter a detected object would have been measured had nearby objects not been fit simultaneously .
we selected only objects with @xmath19 in both frames , and @xmath20 mag in both frames .
we also only kept objects classified by dolphot as good stars , as opposed to elongated or extended objects which have object types @xmath21 . ] with formal errors from the psf fitting less than @xmath22 mag in both f555w and f814w . in figure
[ f : phterr ] typical uncertainties in the derived magnitudes of the stars detected in all three clusters are shown for both f555w and f814w filters . in order to assign a detection completeness fraction to each star , we used dolphot to perform a large number of artificial star tests . for each star in our final cleaned catalogue
we first generated a list of @xmath23 associated fake stars using the dolphot utility acsfakelist .
we constrained these fake stars to lie close to the real star on the ( f555w , f814w ) color - magnitude diagram ( cmd ) the fake stars were randomly and uniformly distributed in a box of size @xmath24 centered on the real star , where @xmath25 was the maximum of @xmath26 mag and @xmath27 times the uncertainty in the @xmath28 magnitude of the real star , and @xmath29 was the maximum of @xmath26 mag and @xmath27 times the uncertainty in the ( @xmath30 ) color of the real star .
we were also careful to constrain the spatial positions of the fake stars .
often this is done so that the fake stars lie at similar cluster - centric radii to the real star under consideration , but with unconstrained position angle .
this saves on computation time because the calculations are then valid for groups of real stars at a time , rather than simply on an individual basis .
however , because we are dealing with very young clusters in the present work , this implicit assumption of spherical symmetry is not necessarily valid .
these objects could potentially still be embedded in gas or dust and thus exhibit a strongly varying background .
they are also not very dynamically evolved , so may yet possess azimuthal asymmetries in stellar density as well as numerous randomly placed bright stars .
therefore , we adopted the more computationally expensive method of generating the list of fake stars locally around each real star on the ccd in this case randomly and uniformly within a radius of @xmath31 . in doing this
we were careful to account for the edges of the chips , including the inter - chip gap .
we then ran dolphot in artificial stars mode on each list of fake stars .
when operating in this way dolphot takes the next fake star from the list , adds it to the input images , and then solves for the position and photometric properties of the star using exactly the same parameters as for the original ( real star ) analysis .
once complete , we took the output photometry and ran it through our quality filters , just as we did for the list of real stars .
finally , we calculated the completeness fraction for a given real star by determining how many of its associated fake stars had been successfully ` found ' .
we considered a fake star to be found if it was recovered in both f555w and f814w and passed successfully through our quality filter , and provided its output position lay within @xmath32 pixels of its input position .
figure [ f : compl ] presents the completeness function of our photometry for all considered clusters in both f555w ( solid line ) and f814w ( dashed line ) filters . .
[ f : prof ] ] ccccc [ t : param ] ngc 1983 & 7.3 & 0.72 & 28 & 0.17 + ngc 2002 & 9.4 & 0.65 & 18 & 0.19 + ngc 2010 & 12.5 & 0.54 & 159 & 0.17 +
the stellar charts of the observed acs / wfc fields of the clusters are shown in figure [ f : maps ] ( left panel ) .
our observations show clear differences in the appearance of the three clusters . in order to reveal the spatial distribution of the detected stars and the morphology of the clusters , we performed star counts on our photometric catalogs .
the constructed ` iso - density ' contour maps for each region are also shown in figure [ f : maps ] ( right panel ) .
the coordinate system used for these maps is that of the original acs / wfc pixel coordinates .
the star counts were performed in a quadrilateral grid divided in elements with sizes @xmath3 190 @xmath33 190 wfc pix@xmath34 each , which correspond to about 10 @xmath33 10 ( or @xmath35 2.5 pc @xmath33 2.5 pc at the distance of the lmc ) .
the first isopleth in the contour map indicates the level of 1@xmath36 above the mean background density , with @xmath36 being the standard deviation of the background density .
the subsequent levels are drawn in steps of 1@xmath36 .
all contours with density equal or higher than the 3@xmath36 level , considered as the statistically significant density threshold , are drawn with thick lines .
all three star clusters are revealed in these maps as the dominant stellar concentrations in the observed regions .
these maps demonstrate that while ngc 1983 is a centrally concentrated compact cluster , ngc 2010 is a quite large and loose cluster , with ngc 2002 being an intermediate case of a large compact concentration .
moreover , ngc 2002 appears to be the most spherical cluster in the sample , since ngc 1983 is more elliptical and ngc 2010 is more amorphous , for which , as indicated by the isopleths of density @xmath37 , its most compact part is off - center .
the sizes of the clusters as derived from circular annuli considered to encircle the @xmath38 isopleth for each cluster are given in table [ t : param ] .
the corresponding surface stellar density of each cluster is also given in this table .
the observed fields cover quite large areas around the clusters in our sample , allowing the construction of the stellar surface density profiles of the clusters at relatively large distances from their centers .
we divided the area of each cluster in concentric annuli of increasing steps outwards and we counted the number of stars within each annulus .
we corrected the counted stellar numbers for incompleteness according to the corresponding completeness factors on a star - by - star basis .
completeness is a function of both distance from the center of the cluster and magnitude .
let @xmath39 be the completeness corrected stellar number within the @xmath7th annulus .
we obtained the stellar surface density , @xmath40 , by normalizing this number to the area of the corresponding annulus as @xmath41 , with @xmath42 being the area of the annulus .
the observed fields do not always fully cover the complete extent of the clusters , and , therefore , for the truncated annuli , we considered only the available area for the estimation of the corresponding surface density .
the `` raw '' radial stellar density profiles , meaning the surface stellar density as a function of distance from the center of each cluster , @xmath43 , show a smooth drop outwards away from the centers of the clusters .
they drop to a uniform level , which represents the stellar density of the lmc field in the vicinity of the clusters , and it is measured by fitting the models of @xcite to the stellar surface density profiles , @xmath44 , ( figure [ f : prof ] , left ) as described below . in the case of ngc 2002
there is a small increase in the outskirts of the profile due to another neighboring small stellar concentration ( see figure [ f : maps ] ) .
we do not consider this increase in our subsequent analysis on the structure of this cluster .
the errors in the profiles reflect the counting uncertainties , representing the poisson statistics .
ccccccc [ t : structpar ] ngc 1983 & 1.68 @xmath45 0.28 & 1.69 @xmath45 0.32 & 0.23 @xmath45 0.02 & 1.89 @xmath45 0.43 & 21.69 & 0.8 + ngc 2002 & 3.92 @xmath45 0.68 & 3.91 @xmath45 1.00 & 0.09 @xmath45 0.01 & 2.56 @xmath45 0.61 & 22.33 & 0.9 + ngc 2010 &
13.32 @xmath45 6.03 & 9.81 @xmath45 8.04 & 0.21 @xmath45 0.01 & 5.19 @xmath45 3.34 & 28.13 & 0.7 + we applied both the empirical model by @xcite and the model of @xcite ( from here - on eff ) to the stellar surface density profiles of the clusters in order to obtain their structural parameters .
the eff model is suited for clusters which are not tidally limited , while king s empirical model represents tidally truncated clusters .
both models provide the opportunity to derive accurate characteristic radii for the clusters .
the core radius ( @xmath46 ) describes the distance from the center of the cluster where the stellar density drops to the half of its central value and the tidal radius ( @xmath47 ) is the limit where the stellar density of the cluster drops to zero .
the best - fitting eff and king profiles are found by performing the levenberg - marquardt least - squares fit to the considered functions .
this fit was performed with idl ( interactive data language ) with the use of the specially developed procedure mpfitfun @xcite . for the application of king s model the stellar surface density of the cluster alone ( with no contamination from the field )
is necessary , while the eff model does not require any field subtraction .
we first apply the latter in order to estimate the core radius and the uniform background level for each cluster .
the subtraction of this density level from the measured surface density at each annulus gives the surface density profile of the cluster alone , from which we derive the core and tidal radii . from studies of young lmc clusters by eff
it appears that these clusters are not tidally truncated .
these authors developed a model more suitable to describe the stellar surface density profile of such clusters : @xmath48 where @xmath49 is the central stellar surface density , @xmath50 is a measure of the core radius and @xmath51 is the power - law slope which describes the decrease of surface density of the cluster at large radii ; @xmath52 for @xmath53 .
the uniform background density level is represented in eq .
( [ eq - elson ] ) by @xmath54 .
the measured parameters @xmath55 , @xmath51 and @xmath54 , derived from the fitting procedure on our data are given for each cluster in table [ t : structpar ] . according to eff model , the core radius , @xmath46 ,
is given from eq .
( [ eq - elson ] ) assuming no contribution from the field as : @xmath56 the estimated core radii of the clusters derived from eq .
( [ eq - rc ] ) are also given in table [ t : structpar ] .
the best - fitting eff models are shown superimposed ( blue lines ) on the stellar surface density profiles of figure [ f : prof ] . in order to construct the density profiles of the clusters alone with no contribution from the field
we subtract the background density level , @xmath54 , from the raw profiles of figure [ f : prof ] ( left ) .
we then use the field - subtracted profile to derive an indicative tidal radius of the clusters as described by @xcite . according to this model the density profile of a tidally truncated cluster
is given as : @xmath57^\frac{1}{2}}-\frac{1 } { [ 1+(\frac{r_{\rm t}}{r_{\rm c}})^2]^\frac{1}{2}}\bigr)^{2 } , \label{eq - king}\ ] ] where @xmath58 is the stellar surface density , @xmath46 and @xmath47 the core and tidal radius , respectively and @xmath59 is the distance from the center .
the tidal radius , which represents the outskirts of the cluster , is found from the formula @xmath60 and the core radius , which describes the inner region of the cluster , is given from @xmath61 @xmath62 describes again the central surface density of the cluster and @xmath63 is a constant .
the best - fitting king profiles apart from core and tidal radii for the clusters , deliver also the concentration parameters , @xmath64 , defined as the logarithmic ratio of tidal to core radius @xmath65 , which refers to the compactness of the cluster .
the derived @xmath47 and @xmath64 for our clusters are given in table [ t : structpar ] , while @xmath46 derived from the best - fitting king profiles are omitted , as they are found in excellent agreement with those derived above from the eff profiles .
the best - fitting king models of our clusters are shown with green dashed lines superimposed on the radial surface density profiles of figure [ f : prof ] ( right ) . in general ,
both king and eff models are in good agreement to each other in the inner and intermediate regions of the clusters , which are equally well fitted by both models .
however , based on the fact that young compact clusters in the lmc are not tidally truncated , the eff model should be considered as the best representative for them . on the other hand ,
the decontamination of the observed stellar samples from the populations of the local lmc field requires the determination of a limiting radius , @xmath66 , for the clusters ( [ ss : fldsub ] ) .
since eff models do not predict such a radius , we treat the tidal radii derived from the application of king s theory as this radius for each cluster
. it should be noted , though , that our clusters are not expected to be dynamically relaxed and therefore , the results on their tidal radii based on the application of king profile fitting should _ not _ be taken literally but only tentatively for the determination of an indicative _ limit _ radius for the clusters , and for reasons of comparison .
for example , as seen in the charts of figure [ f : maps ] , ngc 1983 can not be directly distinguished from the rich stellar field in which it is embedded , as opposed , e.g. , to ngc 2002 , which shows up as a compact concentration easily separated from its surroundings .
in the stellar charts of the three observed regions , shown in figure [ f : maps ] ( left ) , the star clusters immediately appear as the most prominent stellar concentrations .
these maps are constructed only from stars detected with photometric uncertainties @xmath67 mag in both f555w and f814w filters . in our subsequent analysis
we consider only these stars for each observed field .
the acs / wfc field - of - view of ngc 1983 covers 9 641 stars with good photometry , the one of ngc 2002 4 340 , and that of ngc 2010 9 006 such stars .
the @xmath68 , @xmath69 color - magnitude diagrams ( cmds ) of the total observed samples of stars for all three clusters are shown in figure [ f : totcmds ] ( top panel ) .
these cmds of the _ whole areas _ demonstrate the existence of a significant mixture of different stellar populations in each region ; from young stars , which cover the main sequence ( ms ) to the turn - off , belonging to the clusters and their surrounding field , to the evolved red giant branch ( rgb ) and red clump ( rc ) stars , representing mostly the local background lmc field .
the cmds of figure [ f : totcmds ] indicate that our photometry provides accurate magnitudes for stars with @xmath69 -2.truept 22 mag .
it is interesting to note that as seen from the cmds of figure [ f : totcmds ] ( top ) , as well as from the stellar charts of figure [ f : maps ] , the general field of the lmc in the observed regions shows to vary in stellar density .
specifically , both ngc 1983 and ngc 2010 are located in much denser stellar fields than ngc 2002 , with the local field of ngc 1983 being particularly rich in red field stars .
this is already demonstrated by the measurements of the surface stellar density , @xmath70 , of the general field of all three clusters in [ s : param ] ( table [ t : structpar ] ; column 4 ) .
the tidal radii , @xmath47 , of the clusters , as they are derived in
[ s : param ] , basically define the spatial extent of each cluster , and therefore we treat the stars that are located at radial distances from the center of the clusters larger than @xmath47 as the best representative stellar populations of the surrounding fields of the clusters .
the corresponding cmds , for which we use the term _ field cmds _ , are shown in figure [ f : totcmds ] ( middle panel ) . in these cmds
the features of both rgb and rc of the general field of the lmc are well apparent in all three observed regions .
these populations are well documented as members of the lmc field in various hst studies of the star formation history ( sfh ) of this galaxy ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
specifically , the sfh of the lmc is found to be continuous with almost constant star formation rate ( sfr ) for over @xmath3 10 gyr , with an increase of the sfr at around 1 - 4 gyr ago ( for a summary see * ? ? ?
* their 3.1 ) .
this old lmc field population is present in all the field cmds of figure [ f : totcmds ] . since it covers the same parts in all three of them , we define the age - limits of these stars only on the cmd of the field of ngc 2010 . with the use of the padova grid of evolutionary models @xcite for insignificant extinction and metallicity and distance typical for the lmc these limits
are found to be @xmath71 .
moreover , the field cmds of figure [ f : totcmds ] include also young ms stars , clearly suggesting that the general regions , where all three clusters are located , include young stellar populations that are not members of the clusters .
this suggestion does not contradict previous studies on such clusters and their surroundings . indeed
, young compact clusters in the lmc are known to form together in larger structures of young stellar populations , the _ stellar aggregates _ ( e.g. , * ? ? ?
* ) , the size of which seems to correlate with the duration of star formation in them ( e.g. , * ? ? ?
such clusters are not isolated as they may be formed in binary or multiple cluster systems @xcite .
an inspection of the general regions where the clusters of our sample are located verifies that none of them is an exception to this general rule .
wide - field images of the regions around the clusters and the whole extent of the lmc in @xmath72 , @xmath73 and @xmath74 were retrieved from the _ supercosmos sky survey _ @xcite , available at the _ aladin sky atlas _ @xcite .
these images show that indeed all three clusters are members of larger young structures .
specifically , ngc 1983 is located at the central part of the star forming _ shapley constellation ii _
@xcite that coincides with the supergiant shell ( sgs ) _ lmc 3 _ @xcite , ngc 2002 belongs to the northern part of _ constellation iii _ or sgs _ lmc 4 _ , and ngc 2010 is located in one of the eastern nebular filaments of sgs _ lmc 9 _ in _
constellation ix_. in order to compare these ms field populations with the actual stellar populations of the clusters we select the stars contained in the core radii of the clusters ( see [ s : param ] ; table [ t : structpar ] , column 5 ) , as the most probable stellar members , and we construct the corresponding cmds , which are shown in figure [ f : totcmds ] ( bottom panel ) .
a comparison of the field cmds with the cmds of the clusters within their core radii shows that there is a confusion between the young field populations and those of the clusters , especially for ngc 2002 and ngc 1983 .
the case of ngc 2010 is more straightforward , as it can be seen that while the cluster is somewhat evolved ( see discussion below ) , it seems that it is ` embedded ' in a younger field with brighter ms stars . on the other hand ,
ngc 1983 and ngc 2002 are more complicated cases as both field and clusters seem to comprise the same type of ms stars .
taking into account the similarities of field stars to cluster members in the cmds , the use of our method for the _ statistical _ decontamination of the observed stellar samples from the field populations , discussed in [ ss : fldsub ] , becomes quite important . as we show in [ ss : clusmem ] this method allows us an accurate statistical determination of the most probable stellar members of each cluster .
we use the cmds within the core radii of the clusters to derive an estimation of their ages . in the corresponding cmds of figure
[ f : totcmds ] ( bottom panel ) , the best - fitting isochrones from the padova grid of evolutionary models @xcite are also plotted .
the process of isochrone fitting of the observed cmd results in an accurate measurement of the age of the system , providing that there are distinguished features in the cmd to be fitted .
moreover , a good estimate of the distance and the interstellar reddening of the cluster is required . for clusters in the lmc
, visual extinction is known to vary around very low values ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , while the distance of the galaxy is also well defined ( e.g. , * ?
* ; * ? ? ?
* ) . indeed , for our age estimation we had to vary little both of these parameters around typical values .
since the acs photometric system is slightly different from the standard johnson - cousins system , we estimate the mean extinction in the f555w band , @xmath75 , which we arbitrarily symbolize as @xmath76 , using the computations by @xcite .
these authors derive a conversion @xmath77 and @xmath78 assuming the extinction law of @xcite with @xmath79 . in all cases
we assume the typical metallicity of the lmc of @xmath80 - @xmath81 @xmath82 ( e.g. , * ? ? ?
* ) , and therefore we use the models designed for @xmath83 . the derived ages and indicative visual extinction for all three clusters are given in table [ t : param ] .
let us consider the results for each cluster individually : _ ( i ) ngc 1983 _ : the cmd of the most prominent member stars of the cluster consists only of the ms , indicating that the cluster is quite young , but making the accurate determination of its age difficult .
indicatively , we apply the youngest isochrone that fits the brighter ms stars with @xmath84 14 mag , which shows that the cluster has an age about 28 myr .
the selected isochrone fits best the blue part of the ms for a nominal distance modulus @xmath85 mag and reddening of @xmath86 mag , which corresponds to @xmath87 mag . _
( ii ) ngc 2002 _ : the cmd of ngc 2002 within the @xmath46 of the cluster includes the significant number of five red supergiants ( rsg ) located at @xmath88 -2.truept 2.6 mag and 11 -2.truept @xmath69 -2.truept 12 mag , which helps to constrain accurately the age of the system . however , the isochrone for metallicity @xmath83 that seems to fit best this cluster with an age of @xmath89 18 myr , can not reach the rsg stars , which thus require the consideration of models of higher metallicity . indeed
the next available models from the padova grid are those of almost solar metallicity of @xmath90 . the corresponding isochrone for the same age ( plotted with a green line in figure [ f : totcmds ] ) does fit the rsg , implying that either there is a significant change in the metallicity of the cluster , or that the evolutionary models need to be fine - tuned for rsgs in low metallicities . in any case ,
our data give a fixed age for ngc 2002 of around 18 myr .
it should be noted that while the high metallicity model requires a modest @xmath91 mag to fit the cmd , the low metallicity model predicts a higher extinction of @xmath92 mag . _
( iii ) ngc 2010 _ : the cmd of figure [ f : totcmds ] ( bottom ) for ngc 2010 clearly suggests that this is the oldest cluster in our sample .
the existence within @xmath46 of bright stars of @xmath93 mag , evolved away from the ms allows us an accurate determination of the cluster age of @xmath94 myr for @xmath95 mag and @xmath87 mag .
however , the appearance in this spatially constrained cmd of a prominent rgb population , clearly suggests that the cluster may comprise a mixture of different stellar populations .
as we discuss earlier and seen from the cmds of figure [ f : totcmds ] , all three observed regions suffer from significant contamination by field stars . with no information about cluster - membership probabilities for the observed stars , obtained , e.g. , from radial velocities and/or proper motions ,
the quantitative decontamination of the clusters from the field stars on a statistical basis becomes a fundamental method to obtain the cmd of the complete sample of true stellar members of each cluster .
we decontaminate , thus , the observed cmds from the contribution of the stellar members of the surrounding local field of the lmc with the use of a sophisticated random - subtraction technique .
this field - subtraction algorithm , originally designed by @xcite , is already developed in paper i , where it is also thoroughly described . according to this technique , the successful decontamination of the clusters from the field is based on the definition of the limiting radius , @xmath66 , for each cluster . for the clusters in our sample , @xmath66 is given by the tidal radii , @xmath47 , estimated in [ ss : bestfitking ] with the use of king s theory .
all stars located at distances @xmath96 from the center of the cluster are treated as field stars , while the rest ( with @xmath97 ) are considered as the most probable cluster - member stars . assuming a homogeneous field - star distribution , the number density of field stars is calculated in the field cmd ( figure [ f : totcmds ] ; middle ) constructed for stars with @xmath96 , considering also the observational uncertainties in the photometry .
the decontamination process is then applied in three steps for each cluster : ( i ) division of the cmds of the stars with @xmath98 ( cluster region ) and of those with @xmath99 ( offset region ) , respectively , in 2d cells with the same axes along the m@xmath100 and ( m@xmath101m@xmath100 ) directions .
( ii ) calculation of the expected number density of field stars in each cell in the cmd of the offset region .
( iii ) random subtraction of the expected number of field stars from each cell from the cmd of the cluster region .
the algorithm is applied several times using different cell sizes ( @xmath102m@xmath100 , @xmath102(m@xmath101m@xmath100 ) ) so that many different decontamination results are derived . from these results
we calculate the final probability that a star is identified as a ` true ' cluster member , and the final decontaminated cmd of the cluster contains only the stars with the highest probability of being cluster members .
this process allows us to minimize any artificial effects intrinsic to the method itself .
a detailed description of the mathematical formulation of our method for field - subtraction is given in paper i. we performed the field - subtraction process , as described above , 15 times for each cluster considering combinations for the cmd cell sizes of @xmath102m@xmath103 0.5 , 1.0 , 1.5 , 2.5 and 3.0 mag and @xmath104m@xmath101m@xmath105 1.0 , 1.5 and 2.0 mag .
we obtained , thus , the probability for each star in each sample of being a member of cluster population . since our method for diagnosis of stellar stratification is completeness - sensitive in the subsequent analysis we limit the sample of stellar members of all three clusters to those detected with our photometry with completeness @xmath106 70% .
the derived stellar samples comprise 424 stars in ngc 1983 , 1 102 stars in ngc 2002 and 2 265 stars in ngc 2010 with probability larger than 70% of being true cluster members .
figure [ f : syscmds ] shows the corresponding field - star - decontaminated cmds of the cluster regions for all clusters .
stars found by the field - subtraction method with probability 100% of being cluster members are plotted with thick points . from these cmds
it can be seen that our statistical subtraction of the cmd contamination by the field stars was efficient in removing both young field populations of the immediate surroundings of the clusters , _ and _ old populations of the general lmc field . specifically , for ngc 1983 the ms stars remaining in the cmd of figure [ f : syscmds ] , after the application of the field - subtraction , form a sharp narrow ms down to @xmath107 mag .
this ms resembles that of the cmd within the @xmath46 of the cluster shown in figure [ f : totcmds ] ( bottom ) , with no indication of the broad lower ms stars of the field cmd of figure [ f : totcmds ] ( middle ) .
the cmd of ngc 2002 also demonstrates a very sharp ms similar to the @xmath46 cmd of the cluster .
the rsgs located within the @xmath46 of the cluster also remained in the cmd of figure [ f : syscmds ] with 100% of membership probability .
as far as ngc 2010 is concerned , the statistical field subtraction removed completely the bright ms young stars identified in [ ss : cmds ] as young field stars , while the older ms and evolved stars remained with high probabilities of being true cluster members .
it should be noted that the lower main sequence of the cmds of the clusters may be also sensitive to any potential pre - main sequence ( pms ) population that might exist in such young clusters .
however , due to observational constrains , we would be able to detect only the turn - on of these pms stars , which is known to coincide with the msto ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ) , very close to our detection limit .
as a consequence , any pms population in the clusters would be seriously affected by the incompleteness in our photometry and therefore does not have any significance in the measured stellar numbers .
another bias that may affect the final numbers of low main sequence cluster - members is differential reddening that could displace stars to the red - faint part of the cmd . however , differential reddening in young lmc clusters is known to account for no more than @xmath108 ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , which is well covered by the color uncertainties in our photometry at the low main sequence .
as a consequence , differential reddening is not considered in our treatment for field subtraction .
it is interesting to note that while the cluster cmds of figure [ f : syscmds ] for both ngc 2002 and ngc 2010 show the same features with the corresponding @xmath46 cmds of the clusters ( figure [ f : totcmds ] , bottom ) , that of ngc 1983 shows clear indications of multiple young populations with @xmath109 15 mag . indeed , as we discuss in
[ ss : cmds ] , multiplicity is a typical characteristic of young lmc clusters , and naturally it may lead to a merging process and mixing of different stellar populations within one cluster ( see e.g. , * ? ? ? * ) .
as we can not verify or reject this scenario for ngc 1983 , we treat all 424 stars that meet our membership criteria as true members of the cluster .
our diagnostic method for stellar stratification is applied in the following section on the final catalogs of stellar members of the clusters , with 70% membership , the cmds of which are shown in figure [ f : syscmds ] .
the _ effective radius _ method for the diagnosis of stellar stratification is based on the assumption that the observable counterpart of the _ spitzer radius _ , the _ effective radius _ , of stars in a specific magnitude ( or mass )
range will be a unique function of this range if the system is segregated . in paper
i the comparison of the results of this method with those from the ` classical ' methods of the radial dependence of the clusters lfs and mfs , showed that the _ effective radius _
method behaves more efficiently in the detection of the phenomenon , providing direct proof of stellar stratification in star clusters without considering any model or theoretical prediction , and with no application of any functional fit . in this section
we apply this method for the detection and quantification of stellar stratification to the young lmc clusters of our sample .
, scaledwidth=47.5% ] we calculated the effective radii , @xmath110 , of the stars in different magnitude ranges , according to eq .
( [ eq : reff ] ) , using the completeness corrected numbers of stars for all three clusters .
we performed this calculation only for the star - members of the clusters , as they are established in
[ ss : clusmem ] ( cmds of figure [ f : syscmds ] ) , and we used their brightness measured in the f814w filter , since this filter provided the most complete and accurate photometry ( see [ ss : phot ] ) . in order to identify any functional relation between @xmath110 and the corresponding brightness range , we binned the stars according to their magnitudes .
since the measurement of @xmath110 depends strongly to the number of stars per bin , this process naturally inherits systematic uncertainties to the derived relations , because of the fact that some stars could belong to neighboring bins due to photometric uncertainties .
therefore , we performed the same binning process per cluster several times by changing the bin sizes and/or the magnitude limits , within which the stars are binned .
we found that a reasonable bin size , which provides good statistics and allows for the detailed observation of the dependence of @xmath110 on magnitude , is that of 0.5 -2.truept @xmath111 -2.truept 1.0 mag .
for this range of bin sizes we did not find any significant change in the general trend of @xmath112 , which is also found to be independent of the selected magnitude limits for binning the stars . in figure
[ f : reffmag ] we indicatively show the relations between the computed effective radii and the corresponding magnitude bins derived from two different binning processes with bin sizes of 1 mag and magnitude ranges shifted by 0.5 mag from each other .
the combination of both distributions with these specific binning characteristics ( each plotted with different symbols ) provides measurements of @xmath110 for every 0.5 mag range .
the derived trends of @xmath112 , shown in these plots , are not different from the corresponding distributions derived for @xmath113 mag , except of being somewhat smoother . in general the relations of figure
[ f : reffmag ] are representative of those found from any applied binning , and therefore we consider them in the following section for the derivation of our results concerning stellar stratification in the clusters .
the @xmath112 graphs of figure [ f : reffmag ] are plotted on the same scale for reasons of comparison between the clusters . in these plots
it is shown that the effective radius does behave as a function of magnitude , but not necessarily providing proof of stellar stratification for all clusters .
a definite indication of stellar stratification would require the effective radii of the faintest stars to be systematically larger than those of the brightest ones for the whole observed brightness range .
this is not the behavior we observe at least for two of the three clusters of our sample . indeed , as we discuss in paper i , @xmath110 is not a monotonic function of brightness , and mass segregation , _ if any _
, is not expected to behave in the same manner in every cluster , but rather to depend on the intrinsic stellar characteristics of each of them .
for example , while in ngc 2002 there is a trend of fainter stars having systematically larger effective radii for the whole luminosity range , this is not the case for ngc 1983 that shows this trend for only a limited magnitude range , and certainly not for ngc 2010 , which one may argue is not even segregated .
specifically , for ngc 1983 , the brightest magnitude bins , down to @xmath114 16 mag , seem to include stars that are located away from the central part of the cluster .
indeed , the brightest stars in the cluster cmd ( figure [ f : syscmds ] ) with @xmath115 15.5 mag are found to be distributed within the largest @xmath110 of about 12 away from the center of the cluster and not closer than about 06 . on the other hand , taking into account the remaining magnitude bins of @xmath116 15.5 mag , the graph of @xmath112 for ngc 1983 provides clear indications that the cluster is segregated with an almost monotonic dependency of larger @xmath110 to fainter stars .
the graph for ngc 2002 is quite different , giving evidence that the cluster is segregated in the whole extend of observed stellar luminosities .
there is , however , a dependency of the appearance of a relation between @xmath110 and brightness on the considered magnitude range .
indeed , while the three brightest magnitude bins , comprising six stars ( the five rsg and the brightest ms star ) , correspond to the smallest effective radii , the next fainter bins , which include six more ms stars , show that these stars are located well away from the center of the cluster . moreover , from the plot of figure [ f : reffmag ] for the fainter stars , it is almost self - evident that they are segregated as the corresponding relation @xmath112 demonstrates an increase of @xmath110 with brightness .
the behavior of this relation for ngc 2010 is quite different from both the previous cases , as there is no general trend for larger effective radii with fainter magnitudes observable , and thus the cluster behaves as if there is no significant stratification occurring .
specifically , apart from the fact that the three brightest magnitude bins are inconclusive , for more faint stars there is a `` plateau '' in the function @xmath112 , which appears rather flat for magnitudes down to @xmath117 19 mag , while for even fainter stars , down to the faintest observed magnitude , the dependence of @xmath110 over @xmath118 appears again , giving evidence of mild mass segregation for the faintest stars at the outskirts of the cluster .
cccccc [ t : segdeg ] ngc 1983 & 12.0 - 15.5 & 3.5 & 1.20 - 0.58 & @xmath119 & @xmath120 + & 15.5 - 22.0 & 6.5 & 0.58 - 0.94 & @xmath121 & @xmath122 + & 12.0 - 22.0 & 10.0 & 1.20 - 0.94 & @xmath123 & @xmath124 + ngc 2002 & 11.0 - 14.5 & 3.5 & 0.11 - 0.99 & @xmath125 & @xmath126 + & 15.0 - 23.0 & 8.0 & 0.49 - 1.23 & @xmath127 & @xmath128 + & 11.0 - 23.0 & 12.0 & 0.11 - 1.23 & @xmath129 & @xmath128 + ngc 2010 & 13.0 - 14.0 & 1.0 & 1.24 - 0.10 & @xmath130 & @xmath131 + & 14.5 - 18.5 & 4.0 & 1.02 - 1.07 & @xmath132 & @xmath133 + & 19.0 - 23.5 & 4.5 & 0.89 - 1.51 & @xmath134 & @xmath135 + & 13.0 - 23.5 & 10.5 & 1.24 - 1.51 & @xmath136 & @xmath137 + in order to parametrize the phenomenon of stellar stratification as it exhibits ( or not ) itself in the considered clusters , we assume as null hypothesis that all clusters are segregated , but to a different _
degree_. this is evident , as discussed above , from the @xmath110 vs. magnitude plots of figure [ f : reffmag ] .
this _ degree of stratification _ in each cluster may be , thus , expressed in terms of the brightness range of its stars , and the effective radius within which they are confined .
as a consequence , in order to quantify stellar stratification as observed in every cluster , one may use the two primary output parameters of the effective radius method , i.e. , the _ magnitude range _ and the _ corresponding effective radii _ of segregation .
these parameters actually represent the ranges in magnitudes and radial distances that define _ the slope _ of the observed relation @xmath112 .
naturally , for a non - segregated cluster this slope , @xmath138 , should be @xmath139 , as it is derived from the graphs of figure [ f : reffmag ] . considering that a steeper positive slope of this relation represents _ a higher degree of stratification _ , we parametrize stellar stratification in the clusters by measuring @xmath140/@xmath141 and quantify the phenomenon through this _
degree_. as mentioned above , @xmath112 does not represent a systematic relation throughout the whole observed magnitude range in every cluster , but , as seen in figure [ f : reffmag ] , it shows differences depending on the selected magnitudes . for every cluster we select one indicative magnitude limit to separate the bright from the faint stars and measure the corresponding degree of stratification .
this selection is based on the magnitude bin , where we observe dramatic changes in the relation @xmath112 in each cluster . as a consequence , we select two magnitude ranges , one spanning from the brightest magnitude bin down to the selected limit , and the other starting at this limit until the faintest magnitude bin , and we estimate their corresponding @xmath138 for all three clusters .
these magnitude ranges , the corresponding @xmath141 , and @xmath142 , and the derived @xmath138 are given in table [ t : segdeg ] , along with the derived @xmath138 for the whole observed magnitude range for every cluster .
the values of @xmath138 , given in this table , correspond to the _ degree of stratification _ for each cluster in every given brightness range . from the @xmath138 measurements of table [ t : segdeg ]
one can see that while indeed the bright stars of ngc 1983 are _ not segregated _ , stars with @xmath118 -2.truept 15.5 mag show a definite trend in their relation @xmath112 , indicative of stratification with a degree of @xmath143 . on average , however , the cluster does not appear to demonstrate stellar stratification .
we characterize , thus , ngc 1983 as _ partially _ segregated cluster .
on the other hand , ngc 2002 shows a definite trend of larger @xmath110 for fainter stars within both selected magnitude ranges , as well as for the whole extend of observed magnitudes .
it is a definite case of stellar stratification with a degree of stratification of @xmath144 .
finally , ngc 2010 is the most peculiar case , as the results for the few brightest stars are quite inconclusive , and for the fainter stars with 14.5 -2.truept @xmath118 -2.truept 18.5 mag the cluster appears slightly segregated , with a relation @xmath112 that is rather flat . on the other hand ,
the most prominent trend in this relation appears for even fainter stars with @xmath118 -2.truept 18.5 mag and a degree of stratification @xmath145 .
however , if indeed the cluster was segregated for this magnitude range , the observed trend should appear as a continuation of the ( non stratified ) relation @xmath112 of the brighter stars .
more specifically the @xmath110 , where the trend starts for @xmath114 19 mag ( @xmath3 09 ) is _ smaller _ than that of the faintest non segregated stars of @xmath117 18.5 mag with @xmath146 .
this behavior of the relation @xmath112 corresponds to fluctuations in the spatial distribution of the faint stars around the center of the cluster rather than to true segregation . as a consequence ,
we characterize ngc 2010 as _ not segregated _ cluster . while the aforementioned results on three clusters can not be used for a statistically complete interpretation of the phenomenon of stellar stratification in young lmc clusters in connection to their individual characteristics , it is worthwhile to search for any dependence of the degree of stratification to the structural and evolutionary parameters we derived for each cluster in our sample . concerning the structural parameters shown in table [ t : structpar ] ,
all clusters are found with comparable concentration parameters , and tidal radii , @xmath47 , that are not far from each other .
the most important difference we find in these parameters is the core radius , @xmath46 , of ngc 2010 , the non ( or less ) segregated cluster , which is at least two times larger than those of the other two clusters .
in addition , from the characteristics of table [ t : param ] it is shown that this cluster is the most extended and least dense in the sample . moreover
, the cmd of the cluster ( figure [ f : syscmds ] ) shows indications of multi - age stellar populations , suggesting that ngc 2010 is a cluster with multiple stellar generations . in such clusters a large fraction of the first generation of stars
is lost early in the cluster evolution due to its expansion and stripping of its outer layers resulting from early mass loss associated ( e.g. , * ? ? ?
in general , the loose appearance of a low - density cluster with large @xmath46 , comes to the support of this speculation .
while this can not be verified with the present data , it would be reasonable to suggest that the lack of stellar stratification in this cluster will make the effect of early mass loss due to stellar evolution less destructive leading to longer lifetimes @xcite .
as far as ngc 2002 is concerned , it is interesting to note that this cluster , being the most segregated in the sample , is actually the youngest ( table [ t : param ] ) and located in a region , which appears to correspond to the emptiest general lmc field of the three ( table [ t : structpar ] ) .
naturally , these observations allow us to suggest that detection of stellar stratification may have been benefitted by the loose contaminating stellar field , and that the observed stellar stratification in ngc 2002 may be _ primordial in nature _ due to the star formation process , or due to very early dynamical evolution .
indeed recent studies show that cool , fractal clusters can dynamically mass segregate on timescales comparable to their crossing - times , far shorter than usually expected ( e.g. , * ? ? ?
* ) . the exact nature , though , of the observed segregation can not be fully understood without detailed kinematic information about the cluster members , or deeper photometry that will provide the complete imf of the cluster .
finally , the observed _ partial _ stellar stratification in ngc 1983 is driven by the fact that the brightest stars of the cluster are distributed in a wide range of distances from the cluster center .
in general , the behavior of @xmath110 as function of stellar magnitude can not be connected to any specific difference of its characteristics from those of the other clusters , apart from the fact that the cluster belongs to the richest region in field populations , and it is a rather elliptical cluster .
we present a coherent comparative investigation of stellar stratification in three young lmc clusters with the application of a robust method for the assessment of this phenomenon on deep hst imaging . in a cluster where stellar
stratification occurs , the segregated brighter stars are expected to be more centrally concentrated than the non - segregated fainter stars .
stellar stratification can , thus , be investigated from the dependence of the radial extent of stars in specific magnitude ( mass ) ranges on the corresponding mean magnitude ( mass ) . according to this notion , in paper i @xcite we developed and verified the efficiency of a robust method for the assessment of stellar stratification in star clusters .
this method is based on the calculation of the mean - square radius , the _ effective radius _
@xmath110 , of stars in different magnitude ranges , and the investigation of its dependence on magnitude as indication of stellar stratification in the cluster . in the present study
we apply the _ effective radius method _ for the detection and quantification of stellar stratification in young star clusters in the lmc .
we select three clusters observed with the high resolving efficiency of hst / acs , specifically ngc 1983 , ngc 2002 and ngc 2010 , on the basis of the differences in their appearance , structure , stellar content , and surrounding stellar field .
our photometry delivered complete stellar catalogs down to @xmath147 - 23 mag for all three observed fields ( [ ss : phot ] ) .
the stellar surface density maps of the observed regions , constructed from star counts on the photometric catalogs , demonstrate that all three clusters are morphologically quite different from each other .
ngc 1983 appears to be a centrally concentrated , rather elliptical compact cluster , ngc 2002 a compact , spherical stellar concentration , and ngc 2010 a large , loose and rather amorphous cluster ( [ ss : contmap ] ) .
the application of both eff @xcite and king s @xcite models on the stellar surface density profiles of the clusters allowed us the measurement of the background stellar field density of the regions , and the estimation of the core radii , @xmath46 , of the clusters .
while these clusters do not seem to be tidally truncated , a tidal radius , @xmath47 is estimated , indicative of the limiting radius of each cluster ( [ s : dynamics ] ) .
the latter is essential for the successful application of a random field subtraction technique for the decontamination of the stellar samples of the clusters from their surrounding young stellar ambient and the general lmc field ( [ ss : fldsub ] ) .
we derive the ages of the clusters from the stellar populations comprised within the @xmath46 of the clusters and we find ages of @xmath89 28 myr , 18 myr and 159 myr for ngc 1983 , ngc 2002 , and ngc 2010 respectively ( [ ss : clusage ] ) .
we finally apply the _ effective radius method _ for assessing stellar stratification in the clusters and for its quantitative study ( [ s : methappl ] ) .
we bin the stars according to their magnitudes in the f814w filter , and we estimate the corresponding effective radii , @xmath110 , for every cluster in our sample .
we then plot the derived radii versus the corresponding mean magnitude and we investigate the functional relation , @xmath112 , between these parameters . with our method
it is shown that stellar stratification behaves differently in every cluster .
ngc 1983 appears to be _ partially segregated _ , since its brightest stellar content does not appear to be centrally concentrated , while its fainter stars show a mild dependency of @xmath110 towards lower values for fainter magnitudes .
the results on the bright stars of ngc 2010 show evidence of lack of stratification , since the brightest stars are located quite far away from its center .
as far as the faint stars are concerned they show a rather flat @xmath112 relation , with no evidence of segregation in this relation for the faintest stars .
consequently , this cluster is characterized as _ not segregated_. finally , ngc 2002 is found to be well segregated for both its brighter and fainter stars .
it is a clear case of proof of stellar stratification with the effective radius method .
we propose the _ slope _
@xmath140/@xmath141 , measured for stars in selected magnitude ranges , as well as in the whole magnitude range , as the most appropriate parameter for the quantification of the _ degree of stratification _ , and we use it to characterize the phenomenon as it is observed in the clusters . from the derived values of @xmath138 ( [ ss : stratdeg ] ) we find that ngc 2002 is the most strongly segregated cluster in the sample with a degree of stratification @xmath148 , ngc 1983 is a partially segregated cluster with @xmath149 for the faintest stars with @xmath118 -2.truept 15.5 mag , and ngc 2010 is not segregated with indicative @xmath150 ( [ ss : results ] ) .
finally , it should be noted that the extension of this study to a larger sample of lmc clusters observed with hst will provide a more complete picture of the phenomenon of stellar stratification .
naturally , the development of such a comparative scheme , which will include star clusters in the whole extent of the evolutionary sequence , requires a comprehensive set of acs and , in the near future , wfc3 observations .
d. a. g. kindly acknowledges the support of the german research foundation ( deutsche forschungsgemeinschaft , dfg ) through grant go 1659/1 - 2 , and the german aerospace center ( deutsche zentrum fr luft- und raumfahrt , dlr ) through grant 50 or 0908 .
sincere acknowledgements go also to m. kontizas and e. kontizas the collaboration with whom originally planted the seed of the idea for the effective radius method .
based on observations made with the nasa / esa _ hubble space telescope _
, obtained from the data archive at the space telescope science institute .
stsci is operated by the association of universities for research in astronomy , inc . under nasa contract
nas 5 - 26555 .
markwardt , c. b. 2008 , in astronomical data analysis software and systems xviii , asp conference series , eds . d. bohlender , p. dowler & d. durand ( astronomical society of the pacific : san francisco ) , in press ( arxiv:0902.2850v1 ) | we present a comprehensive study of stellar stratification in young star clusters in the large magellanic cloud ( lmc ) .
we apply our recently developed _ effective radius method _ for the assessment of stellar stratification on imaging data obtained with the advanced camera for surveys of three young lmc clusters to characterize the phenomenon and develop a comparative scheme for its assessment in such clusters .
the clusters of our sample , ngc 1983 , ngc 2002 and ngc 2010 , are selected on the basis of their youthfulness , and their variety in appearance , structure , stellar content , and surrounding stellar ambient .
our photometry is complete for magnitudes down to @xmath0 mag , allowing the calculation of the structural parameters of the clusters , the estimation of their ages and the determination of their stellar content .
our study shows that each cluster in our sample demonstrates stellar stratification in a quite different manner and at different degree from the others .
specifically , ngc 1983 shows to be _ partially _ segregated with the effective radius increasing with fainter magnitudes only for the faintest stars of the cluster .
our method on ngc 2002 provides evidence of _ strong _ stellar stratification for both bright and faint stars ; the cluster demonstrates the phenomenon with the highest degree in the sample .
finally , ngc 2010 is _ not segregated _ , as its bright stellar content is not centrally concentrated , the relation of effective radius to magnitude for stars of intermediate brightness is rather flat , and we find no evidence of stratification for its faintest stars . for the parameterization of the phenomenon of stellar stratification and its quantitative comparison among these clusters ,
we propose the slope derived from the change in the effective radius over the corresponding magnitude range as indicative parameter of the _ degree of stratification _ in the clusters . a positive value of this slope indicates mass segregation in the cluster , while a negative or zero value signifies the lack of the phenomenon . |
deep - inelastic @xmath1 scattering ( dis ) at hera , in which a photon of virtuality , @xmath2 , is exchanged , can lead to the production of jets in the final state .
cross - section measurements of these types of processes provide stringent tests of the perturbative quantum chromodynamics ( pqcd ) formalism in next - to - leading order ( nlo ) calculations , the structure of the proton and allow measurements of the strong coupling constant , @xmath0 .
these features can be seen in the formula for the cross section , @xmath3 , which is factorised into a convolution of the hard partonic cross section , @xmath4 , and the proton s parton density , @xmath5 : @xmath6 the hard partonic cross section is a power series expansion in @xmath0 and is calculable in pqcd .
the proton s parton density is derived from fits to published data , such as inclusive dis measurements @xcite and high transverse energy jet production in @xmath7 production @xcite . in the region of momentum fraction , @xmath8 , and @xmath2 in which dis jet measurements are currently performed ,
the quark density in the photon is well constrained from particularly inclusive dis measurements . at high @xmath2 where the quark initiated process , qcd compton ( see figure [ fig : feyn](a ) ) , is dominant
, the jet cross section measurements test pqcd . at lower @xmath2 , where the boson - gluon fusion process ( see figure [ fig : feyn](b ) ) is dominant , measurements of the jet cross sections provide complementary information on the gluon density in the proton , which from dis is only constrained through the interpretation of scaling violations in nlo
fits .
( 0,0)(100,100 ) ( 137,110)*(a ) * ( 211,110)*(b ) * in this paper , the latest measurements from both the h1 and zeus collaborations on dis jet production will be discussed . initially the jet cross section comparisons with nlo
will be shown and then the extraction of the gluon density and @xmath0 will be presented . limitations of both the current data and theory and some of their possible solutions will be addressed . in particular , the measurements raise the following issues : * at which points are the theoretical or experimental errors dominant ? * which of the two `` natural '' scales , @xmath9 and transverse energy , @xmath10 , is the more appropriate ? * is there a region where the dglap @xcite formalism breaks down and bfkl @xcite is more appropriate ?
* is there an effect of the resolved photon ?
the two inclusive jet measurements presented here have different goals ; the first is concentrated in a `` safe '' region to test pqcd and extract @xmath0 , whilst the second extends the kinematic phase space to more extreme regions to look for suggestions of bfkl effects .
the safe region was defined by considering jets at a reasonably high@xmath11 of greater than 125 gev@xmath12 , high@xmath13 of greater than 8 gev in a central region of the detector , @xmath14 , where the jets are reconstructed in the breit frame of reference . to be more sensitive to bfkl effects
, a second region was probed ; @xmath15 gev@xmath12 , @xmath16 gev and @xmath17 .
the measured cross section as a function of @xmath2 , for @xmath18 gev@xmath12 , is shown in figure [ fig : incl](a ) compared to predictions from nlo qcd implemented in the disent @xcite program .
the data fall by five orders of magnitude and are reasonably well described by the predictions corrected for hadronisation effects and using two different values for the renormalisation scale , @xmath19 .
the description is quantified in figure [ fig : incl](b ) where the ratio of the data to the prediction ( using @xmath20 ) is shown .
the data lie above the theory by about @xmath21 at @xmath22 gev@xmath12 , although the size of the theoretical and experimental uncertainties rule out any firm conclusions . at higher @xmath2
, the prediction describes the data well .
similar results were observed when considering the cross section as a function of the transverse energy of the jet .
the data can therefore be used to extract a value of @xmath0 , which was done at high@xmath11 and high@xmath13 . as can be seen in figure [ fig : incl](b ) , extracting @xmath0 at @xmath23
gev@xmath12 leads to a smaller theoretical uncertainty , arising mainly from the variation of the renormalisation scale by a factor of two .
( 0,0)(100,100 ) ( 83,160)*(a ) * ( 83,115)*(b ) * ( 162,160)*(c ) * a further test of qcd is the structure of the cross section in the azimuthal angle , @xmath24 , which is defined as the angle between the lepton scattering plane and the jet production plane .
the cross section is predicted to have the form @xcite , @xmath25 and has not be seen before in neutral - current dis jet production .
the effect has been seen in charged hadron production in dis @xcite . without tagging the nature of the final state jets ,
the cross section is expected to reduce to : @xmath26 the measurement is shown in figure [ fig : incl](c ) compared to the nlo predictions .
the form of eq .
( [ eq : phi_red ] ) is clearly seen in the data for the first time in neutral current dis jet production and the calculation gives a good description of the shape .
comparisons of inclusive jet data at lower scales and for more forward - going jets with nlo are shown in figure [ fig : incl_forward ] .
the data are shown in different regions of pseudorapidity of the jet so as to maximise sensitivity to bfkl effects which are expected to be largest at forward values .
the nlo calculation has been evaluated using two different choices of scale ; @xmath10 and @xmath9 , shown in figures [ fig : incl_forward](a ) and [ fig : incl_forward](b ) , respectively .
( 0,0)(100,100 ) ( 166,122)*(a ) * ( 232,122)*(b ) * + overall , the predictions generally describe the data , although with very large uncertainties of sometimes up to @xmath27 .
the agreement is better at higher @xmath2 and at lower @xmath2 it is better when using @xmath9 rather than @xmath10 as the scale in the calculation .
it should be noted that at these low values of @xmath2 , @xmath28 . when using @xmath10 as the scale , the nlo does poorly in the forward region at low values of transverse energy .
this is exactly the region where bfkl effects are expected to show up .
however , the large uncertainties on the dglap - based calculation shown , preclude any statement of the need for bfkl effects in the measurement . these large uncertainties in the theory , and hence a significant difference between nlo and the next higher order ,
are probably not surprising given the lo and nlo cross sections differ by up to a factor of ten at low @xmath10 and forward @xmath29 .
clearly this kind of measurement needs more accurate calculations in order to search for a breakdown of the dglap formalism .
along with bfkl effects , at low@xmath11 the existence and nature of a resolved photon may also play a rle . the resolved photon is expected to be important when the transverse energy of the outgoing jets , @xmath10 is much greater than @xmath9 , whereas bfkl effects are expected to be most significant when the two quantities are roughly equal .
the rle of the resolved photon has been studied in dijet production with @xmath30 gev@xmath12 and at least two jets , such that again @xmath28 .
the ratio of the dijet data as a function of the momentum fraction , @xmath31 , is poorly described by the calculation when the renormalisation scale is set to the intuitive hard scale , @xmath32 .
when choosing the scale to be @xmath9 , the prediction is higher and describes the data well , but suffers from a lack of predictive power due to its large uncertainty .
therefore , using the natural scale , the impact of the resolved photon was studied as implemented in the calculation jetvip @xcite .
the inclusion of a resolved photon into the calculation compared to the data is shown in figure [ fig : dijets_jetvip ] .
it can be seen that the calculation lies below the data at low@xmath33 and that the description improves with the inclusion of a resolved photon .
the effect is larger at the lower range in @xmath2 , but is not enough to completely describe the data . due to large uncertainties in the calculations and unknowns in the underlying physics processes , fundamental parameters such as @xmath0 , are measured at higher values of @xmath2 or @xmath10 , where the uncertainties are reduced .
the dijet cross section at slightly higher @xmath10 is shown as a function of @xmath2 in figure [ fig : dijets_h1_q2 ] .
note that here the jets are reconstructed in the breit frame whereas in figure [ fig : dijets_jetvip ] they were in the hadronic centre - of - mass frame .
the description of the data by the nlo calculation in figure [ fig : dijets_h1_q2 ] is good over the whole range in @xmath2 .
both data and theory also scale similarly with different minimum transverse energy requirements .
the size of the lo to nlo corrections are reasonably large at low@xmath11 and the hadronisation corrections are also significant .
these factors again demonstrate the need to consider higher values of @xmath2 for the extraction of fundamental parameters .
the description of other variables ( not shown ) by the nlo is also generally good .
the dijet cross section at high-@xmath2 , greater than 470 gev@xmath12 , is shown in figure [ fig : dijets_zeus_q2 ] along with the inclusive cross section and dijet rate .
all measurements are well described by the nlo calculation .
the dijet rate is a particularly good variable to measure as it is sensitive to @xmath0 ( as demonstrated in figure [ fig : dijets_zeus_q2](b ) ) and experimental and theoretical uncertainties are expected to somewhat cancel .
cancellation of some of the theoretical uncertainty can be seen in figure [ fig : dijets_pdf ] where the size of the uncertainty related to the parton density function is shown .
this was estimated using the program epdflib @xcite .
this program propagates the statistical and systematic uncertainties of each data set used in the pdf , mbfit and provides additional pdf sets to quantify the theoretical uncertainties .
figure [ fig : dijets_pdf ] shows that the total uncertainty arising from the pdf for the rate is @xmath34 , smaller than for both the inclusive ( @xmath35 ) and dijet ( @xmath36 ) cross sections .
neutral current dis with three jets in the final state has recently been measured which , with a more complex final state , has the potential to provide a more stringent test of pqcd .
a new nlo three - jet calculation @xcite allows the comparison to be made for the first time .
the ratio of three- to two - jet cross sections can also be measured . as with the ratio of cross sections of two - jet and inclusive production
, it is expected that there will be cancellation of some experimental and theoretical uncertainties .
again , this measurement should allow an accurate extraction of @xmath0 , with much reduced uncertainties . in figures [ fig:3jets_xsec ] and [ fig:3jets_ratio](a ) ,
the cross - section measurements are shown as a function of bjorken - x , @xmath37 , three - jet mass , @xmath38 and @xmath2 .
predictions to lo and nlo , both corrected for hadronisation effects , are compared to the measurements .
the addition of the nlo corrections are significant , particularly at low-@xmath37 and @xmath2 and low mass , and their inclusion into the prediction provides a good description of the data . also shown , in figure [ fig:3jets_ratio](b ) , is the ratio of data to theory as a function of @xmath2 .
the nlo can be seen to describe the data to within about @xmath39 over the whole region in @xmath2 .
the uncertainty due to varying the value of the renormalisation scale is up to @xmath40 at low-@xmath2 and decreases with increasing @xmath2 .
the uncertainty arising from the value of @xmath0 and the gluon content of the proton are reasonably constant with @xmath2 and are about @xmath41 and @xmath39 , respectively .
( 0,0)(100,100 ) ( 167,140)*(a ) * ( 242,140)*(b ) * the ratio of three- to two - jet cross sections is also shown in figure [ fig:3jets_ratio](c ) as a function of @xmath2 .
the nlo again describes the data well and , as expected , the theoretical uncertainties are significantly reduced . in particular , the uncertainty due to the renormalisation scale and gluon pdf are both considerably smaller than the uncertainty arising from varying @xmath0 .
this demonstrates the potential of the distribution in providing an accurate measurement of @xmath0 .
this measurement awaits an increased event sample to realise its full potential .
( 0,0)(100,100 ) ( 157,235)*(a ) * ( 157,200)*(b ) * ( 157,150)*(c ) *
measurements of jet substructure in nc dis have been performed and used to test pqcd and perform extractions of @xmath0 .
measurements of both the integrated jet shape and mean subjet multiplicity in inclusive jet nc dis were made .
the lowest non - trivial - order contribution to these quantities is given by @xmath42 pqcd calculations .
thus measurements of the jet shape and subjet multiplicity provide a stringent test of pqcd calculations beyond lo and allow a determination of @xmath0 .
subjets were resolved within a jet by considering all particles that are associated with it and by repeating the application of the @xmath43 cluster algorithm @xcite until , for every pair of particles @xmath44 and @xmath45 , the quantity @xmath46 was above @xmath47 .
all remaining clusters were called subjets .
the subjet structure depends upon the value chosen for the resolution parameter , @xmath48 . in figure
[ fig : sub_data ] , the measurements of mean subjet multiplicity are shown as a function of the resolution scale , @xmath48 , and @xmath49 at a fixed value of @xmath48 equal to 0.01 .
also shown are the predictions from lo and nlo qcd , corrected for effects of hadronisation .
the nlo predictions give a good description of the data for all values of @xmath48 and @xmath49 and can therefore be used to perform a measurement of @xmath0 .
similar results are also seen for the measurements of the jet shape .
as stated in section [ sec : intro ] , jet production at hera is directly sensitive to the gluon content of the proton at lo .
therefore , the measurements shown here can in principle be used to constrain the gluon density , complementing the extractions from ( predominantly ) measurements of the proton structure function , @xmath50 .
this has been performed by simultaneously fitting the inclusive and dijet cross sections and the measurements from inclusive dis .
the result of this fit are shown in figure [ fig : h1_gluon ] , where the extraction is compared to other parametrisations of the gluon pdf .
the extraction from this simultaneous fit is consistent with all other parametrisations shown .
the effect of the jet data in further constraining the gluon density would be enhanced by using the data at lower @xmath2 . however , due to the large theoretical uncertainties , arising predominantly from varying the value of the renormalisation scale , an accurate extraction of the gluon density from jet data remains limited .
also shown in figure [ fig : h1_gluon ] is the result of simultaneously constraining both the gluon density and measuring @xmath0 .
the data are sensitive to the product @xmath51 , but do not permit a precise determination of both simultaneously .
( 0,0)(100,100 ) ( 157,140)*(a ) * ( 242,140)*(b ) *
as shown in the previous sections , measurements of jet cross sections and jet substructure provide an opportunity to extract a value for @xmath0 .
there now exist several of these measurements from hera as well as measurements of @xmath0 from inclusive dis data . at hera
there is also the opportunity to make a measurement of @xmath0 at different values of the scale , @xmath9 or @xmath10 and hence test the running of the value .
this is shown in figure [ fig : hera_alphas](a ) , where the running of @xmath0 is seen over a wide range in @xmath10 ( within one experiment ) , consistent with the renormalisation group equation .
the values of the different methods of measuring @xmath0 are shown in figure [ fig : hera_alphas](b ) and compared with the world averages .
all the values measured are consistent both with each other and the world averages .
in particular , the uncertainties on the measurements are also competitive with those on the world average
. it will be interesting to see the impact these hera measurements have on future determinations of the world average .
( 0,0)(100,100 ) ( 130,128)*(a ) * ( 242,120)*(b ) * the measurements of the inclusive jet cross sections at high-@xmath2 show significant potential for the future .
the uncertainty associated with that of the renormalisation scale is significantly smaller than in the nlo - qcd fits to @xmath50 data .
using the additional statistics available and those of the future , these measurements could provide very accurate measurements of @xmath0 .
this could also be true for the measurement of the three- to two - jet cross section ratio , with indications of small theoretical uncertainties already observed .
the theoretical uncertainties will also be significantly reduced with higher order calculations which are expected soon for fits to @xmath50 data @xcite and within a few years for jet calculations @xcite .
many measurements of jet production in dis have been made over a wide kinematic range at hera . in general , the data are well described by nlo qcd calculations and the current proton parton densities ( convoluted with a correction for hadronisation effects ) , particularly at high scales . from these data , several measurements of @xmath0
have been made , all of which agree with the world average and have uncertainties which are also competitive with the world average . although providing good tests of pqcd , further experimental and theoretical work is needed to make really precise comparisons .
most of the data discussed here , use luminosities of the order of 30 @xmath52 .
both experiments currently have about 100 @xmath52 of data on tape and expect to collect about ten times this amount by the end of hera ii .
this significant increase in the data sample will enable high - precision comparisons up to very high-@xmath2 values which are currently statistically limited .
experimentally , the main source of systematic uncertainty arises from uncertainty in the knowledge of the jet energy scale .
the current quoted uncertainty is generally @xmath53 , which can produce an uncertainty of @xmath54 in the cross section .
efforts are being made ( and succeeding ) to reduce the uncertainty to @xmath55 , which will lead to an uncertainty in the cross section of @xmath56 .
as most measurements already suffer from dominating theoretical errors , particularly for values of @xmath2 less than 500 @xmath57 , these further experimental improvements will require much more accurate theoretical predictions . at low-@xmath2 ,
the renormalisation scale uncertainty is often greater than @xmath58 , which is too large and hinders conclusions being drawn from the comparison with data , such as the existence of bfkl effects and the need for a resolved photon .
also in this region , the choice of scale is unclear with that of @xmath10 seeming to be the natural choice , as @xmath28 , but with @xmath9 appearing to better describe the data . at high-@xmath2 ,
the renormalisation scale uncertainty is still the dominant source of error and its reduction essential for future measurements .
there is a strong need for higher order calculations or resummed nlo programs to fully exploit the potential of the measurements being made . | jet cross sections in deep - inelastic scattering over a wide region of phase space have been measured at hera .
these cross section measurements provide a thorough test of the implementation of quantum chromodynamics in next - to - leading order ( nlo ) calculations .
they also provide the opportunity to test the consistency of the gluon distribution in the proton as extracted from ( mainly ) inclusive dis measurements .
comparison of the cross sections with nlo enables accurate extractions of the strong coupling constant , @xmath0 , to be made , several of which are reported here . |
fix a space dimension @xmath0 and denote by @xmath1 the standard gaussian measure on @xmath2 with density @xmath3 equal to @xmath4 given @xmath5 let @xmath6 be independent random points that are distributed on @xmath7 according to the probability measure @xmath1 . the random convex hull @xmath8\ ] ] of these points is a _
gaussian polytope_. random polytopes of this type are central objects considered in stochastic geometry and are also of importance in convex geometric analysis or coding theory .
for example , gluskin @xcite has used gaussian polytopes in his analysis of the diameter of the minkowski compactum and gaussian polytopes also arise as lower - dimensional shadows of randomly rotated high - dimensional regular simplices as shown by baryshnikov and vitale @xcite .
we refer to the survey article of reitzner @xcite for further background information and references .
we denote for @xmath9 by @xmath10 the @xmath11th intrinsic volume of @xmath12 , that is , @xmath13 here , @xmath14 is the grassmannian of @xmath11-dimensional linear subspaces of @xmath7 supplied with the unique haar probability measure @xmath15 and @xmath16 stands for the @xmath11-dimensional lebesgue measure of the orthogonal projection @xmath17 of @xmath12 onto @xmath18 .
moreover , for @xmath19 , @xmath20 denotes the volume of the @xmath21-dimensional unit ball .
the intrinsic volumes are of outstanding importance in convex geometry , since according to a classical theorem of hadwiger they form a basis of the vector space of all continuous and rigid - motion invariant real - valued valuations on convex sets , cf .
for example , @xmath22 is the volume , @xmath23 coincides with the surface area and @xmath24 corresponds to the mean width of @xmath12 .
it is well known from the work of affentranger @xcite that the expectation @xmath25 $ ] of @xmath10 satisfies @xmath26 = { d\choose \ell}{\k_d\over\k_{d-\ell}}\,(\log n)^{\ell\over 2}[1+o(1)]\,,\qquad\text{as}\qquad n\to\infty\,.\ ] ] more recently , the asymptotic behaviour of the variance @xmath27 $ ] of @xmath10 has moved into the focus of attention . using the classical efron - stein jackknife inequality hug and reitzner @xcite
have obtained a first upper bound of the form @xmath28
\leq c_d(\log n)^{{\ell-3\over 2}}$ ] with a constant @xmath29 only depending on the space dimension @xmath30 ( but not on @xmath11 ) . in a remarkable paper of calka and yukich @xcite
the precise variance asymptotic was derived , showing thereby that the upper bound from @xcite does not have the right order of magnitude .
in fact , ( * ? ? ?
* theorem 1.5 ) says that @xmath31 = c_{d,\ell}\,(\log n)^{\ell-{d+3\over 2}}\,[1+o(1)]\,,\qquad\text{as}\qquad n\to\infty\,,\ ] ] with constants @xmath32 only depending on @xmath30 and on @xmath11 .
however , using their methods the authors of @xcite were not able to exclude the possibility that @xmath33 .
the aim of the present paper is to fill this gap and to show that , in fact , @xmath34 .
this answers a question raised at several places in the literature , see ( * ? ? ?
* section 14 ) , the comment after ( * ? ? ?
* theorem 1.5 ) or ( * ? ? ?
* remark 3.6 ) .
our result reads as follows : [ thm : main ] let @xmath35 and let @xmath12 be a gaussian polytope .
then there exists a constant @xmath36 only depending on @xmath30 and on @xmath11 such that @xmath37 \geq v_{d,\ell}\,(\log n)^{\ell-{d+3\over 2}}\,,\ ] ] whenever @xmath38 is sufficiently large .
in particular , theorem [ thm : main ] in conjunction with shows that the limit @xmath39\,(\log n)^{{d+3\over 2}-\ell}\ ] ] exists and takes a stricly positive and finite value for any @xmath35 .
* let us first comment on the boundary case @xmath40 in theorem [ thm : main ] .
since @xmath41 for any convex set @xmath42 , we have that @xmath43 with probability one and hence @xmath44=0 $ ] . * since @xmath45 is the volume of the gaussian polytope @xmath12 , the case @xmath46 is already covered by theorem 6.1 in @xcite , which ensures that @xmath47 .
our proof comprises this situation as a special case .
a random polytope model closely related to @xmath12 can be described as follows . for each @xmath48 let @xmath49 be a poisson point process on @xmath7 with intensity measure @xmath50 .
the convex hull of the points of @xmath49 will be denoted by @xmath51 and is called the _ gaussian poisson polytope_. following the coupling construction in the proof of ( * ? ? ?
* lemma 7.1 ) one easily sees that the expectation and variance asymptotic for @xmath51 are literally the same as for @xmath12 .
moreover , the strict positivity of the constants @xmath52 in theorem [ thm : main ] implies that @xmath53(\log n)^{{d+3\over 2}-\ell}$ ] converges to a positive and finite limit .
we summarize the missing piece in the proof of this result in the following corollary : [ cor : poissonpolytopes ] let @xmath35 and let @xmath51 be the gaussian poisson polytope .
then there exists a constant @xmath36 only depending on @xmath30 and on @xmath11 such that @xmath54 \geq v_{d,\ell}\,(\log n)^{\ell-{d+3\over 2}}\,,\ ] ] whenever @xmath38 is sufficiently large .
the result of theorem [ thm : main ] and corollary [ cor : poissonpolytopes ] can be regarded as the missing piece of the jigsaw in the theory of gaussian polytopes .
let us mention some of the implications that are now immediate : * _ central limit theorems_. as explained in @xcite , the positivity of the limiting variance is the only missing piece in the proof of the central limit theorem for the normalized intrinsic volumes of @xmath51 .
the result follows by the methods developed in @xcite .
moreover , a de - poissonization argument similar to that in @xcite leads to the corresponding result for @xmath12 ; we omit the details . *
_ concentration inequalities_. as explained in the recent work @xcite , the positivity of the limiting variance is the only missing ingredient in the proof of a concentration inequality for @xmath55 .
the precise form of such an inequality can now be determined from ( * ? ? ?
* theorem 3.1 ) : for any @xmath35 one can find a constant @xmath56 only depending on @xmath30 and on @xmath11 such that @xmath57|\geq y\ , \sqrt{\bv[v_\ell(\pi_n)]}\,\big)\\ & \qquad \qquad \leq 2\exp\big(-{1\over 4}\min\big\{{y^2\over 2^{2d+\ell+5}},c\,(\log n)^{d-1\over 4(2d+\ell+5)}\ , y^{1\over 2d+\ell+5}\big\}\big)\end{aligned}\ ] ] for all @xmath58 and sufficiently large @xmath59 * _ marcinkiewicz - zygmund - type strong laws of large numbers .
_ the concentration inequality for @xmath55 mentioned in the previous paragraph can directly be used to derive marcinkiewicz - zygmund - type strong laws of large numbers along the lines of the proof of ( * ? ? ?
* theorem 1.3 ) : for any @xmath35 and @xmath60 one has that @xmath61}{(\log n)^{p \frac{\ell}{2 } } } \longrightarrow 0 \end{aligned}\ ] ] with probability one , as @xmath62 . using the monotonicity of intrinsic volumes and a simple coupling argument , one easily verifies that the same result also holds with @xmath51 replaced by @xmath12 . in that form , this refines the ordinary strong law of large numbers from ( * ? ? ?
* corollary 1.2 ) , which corresponds to the special case @xmath63 .
* _ moderate deviations_. moderate deviations for the volume and the face numbers of the gaussian poisson polytopes @xmath51 have also been investigated in @xcite .
again , the only missing piece for the extension of these results to the intrinsic volumes is the positivity of the limiting variances ; we omit the details .
let @xmath64 be an arbitrary real number , let @xmath65 be a poisson point process on @xmath7 with intensity measure @xmath66 and denote by @xmath67 the random convex hull induced by @xmath65 .
using the monotonicity of intrinsic volumes and a simple coupling argument , one easily verifies that the result of corollary [ cor : poissonpolytopes ] continues to hold with @xmath51 and @xmath68 replaced by @xmath67 and @xmath69 , respectively .
the same comment applies to the central limit theorem , the concentration inequalities , the marcinkiewicz - zygmund - type strong laws of large numbers and to the the moderate deviations mentioned above .
the rest of this paper is structured as follows . in section [ sec : preparations ] we recall the essential steps of a geometric construction from @xcite and prove some auxiliary results that are needed in the proof of theorem [ thm : main ] .
the latter is the content of the final section [ sec : proofmain ] .
the symbols @xmath70 and @xmath71 are used for the euclidean norm and scalar product in @xmath7 , respectively . moreover , for a set @xmath72 we write @xmath73 $ ] for the convex hull of @xmath74 .
we denote the @xmath30-dimensional unit ball by @xmath75 and write @xmath76 for the corresponding unit sphere .
our underlying probability space is @xmath77 and we implicitly assume that it is rich enough to carry all the random objects we consider in this paper . by @xmath78
$ ] we denote expectation ( integration ) with respect to @xmath79 and @xmath80 $ ] stands for the variance of the argument random variable .
the indicator function of an event @xmath81 is denoted by @xmath82 . for two sequences @xmath83 and @xmath84 we write @xmath85 ( or @xmath86 ) if we can find a constant @xmath56 not depending on @xmath38 and an index @xmath87 such that @xmath88 ( or @xmath89 ) for all @xmath90 .
finally , @xmath91 means that @xmath92 . in this paper constants
are denoted by @xmath93 it is implicitly assumed that these constants are finite and strictly positive , and only depend on the space dimension @xmath30 , unless otherwise stated . in this section
we recall a geometric construction as well as some of the results already obtained @xcite that we use below . we define @xmath94 and denote by @xmath95 the centred sphere of radius @xmath96 . by @xmath97
we denote a maximal system of points such that @xmath98 for some sufficiently large @xmath99 .
a simple volume comparison argument provides an estimate for the size of such a set , see ( * ? ? ?
* claim 5.1 ) : [ lem : estimatem ] one has that @xmath100 .
[ width=0.6]z1 ( 0,-5)@xmath101 ( 10,30)@xmath102 ( 45,10)@xmath103 ( 132,20)@xmath104 ( 220,10)@xmath105 ( 47,40)@xmath106 ( 220,40)@xmath107 ( 132,95)@xmath108 ( 132,50)@xmath109 for each @xmath110 define @xmath111 and notice that @xmath112 .
let further for @xmath110 , @xmath113 be the tangent hyperplane of @xmath101 at @xmath104 and fix a regular simplex in @xmath102 whose vertices @xmath114 are chosen from the @xmath115-dimensional sphere @xmath116 of radius @xmath117 in @xmath102 centred at @xmath104 .
the simplex @xmath118 $ ] is the convex hull of @xmath108 and the points @xmath119 , see figure [ fig1 ] .
it is not difficult to estimate the volume @xmath120 and the gaussian measure @xmath121 of these simplices , see ( * ? ?
* claim 5.2 ) : [ lem : contentdeltai ] for each @xmath110 one has that @xmath122 and @xmath123 . for each @xmath110 and @xmath124
we let @xmath125 be a homothetic copy of @xmath109 with @xmath126 being the centre of the homothety and the factor being a sufficiently small number @xmath127 , that is , @xmath128 .
since each @xmath125 is only a homothetic copy of @xmath109 with a scaling factor not depending on @xmath38 , the following holds by construction : [ lem : contentdeltaij ] for each @xmath110 and @xmath124 one has that @xmath129 and @xmath130 . for each @xmath110 and @xmath124 let @xmath131 be an arbitrary point in @xmath125 and define the cone @xmath132 , where we write @xmath133 for the positive hull of the argument set .
we recall the following fact about these cones from ( * ? ? ?
* lemma 5.4 ) , which ensures a certain independence property used below : [ lem : conesci ] one can choose the constant @xmath99 in the above construction sufficiently large and @xmath127 sufficiently small such that the cone @xmath134 contains all simplices @xmath135 with @xmath136 for each @xmath110 . next , for @xmath110 and @xmath137
we denote by @xmath138 the half - space containing @xmath139 for all @xmath140 , not containing @xmath141 and @xmath125 , and such that the hyperplane bounding @xmath138 touches all the simplices @xmath142 except for @xmath125 , see figure [ fig:2 ] ( left ) .
we are now in the position to define for each @xmath110 the event @xmath143 that precisely one point from the random sample @xmath6 is contained in each simplex of the form @xmath125 and no further point of @xmath6 is contained in @xmath144 , see figure [ fig:2 ] ( right ) . here , @xmath145 is the half - space bounded by @xmath102 not containing the origin .
the following probability estimate is taken from ( * ? ? ?
* lemma 6.2 ) : [ width=0.49]z2 ( 110,55)@xmath109 ( 109,103)@xmath141 ( 25,15)@xmath146 ( 180,15)@xmath147 ( 112,88)@xmath148 ( 27,30)@xmath149 ( 180,30)@xmath150 ( 150,75)@xmath151 ( 60,75)@xmath152 [ width=0.49]z3 ( 113,92)@xmath148 ( 28,35)@xmath149 ( 185,34)@xmath150 ( 110,55)@xmath109 ( 10,90)no points ( 20,80)here ( 175,90)no points ( 185,80)here [ lem : probabilityai ] there exists a constant @xmath153 such that @xmath154 for all @xmath110 . the facts summarized so far have been used in @xcite to prove a lower variance bound for the volume @xmath45 of @xmath12 .
since we are interested in all intrinsic volumes @xmath155 , a refinement is necessary to obtain such bounds .
in fact , we now follow and adapt the method applied already in @xcite to handle the more general situation .
let @xmath156 and @xmath157 for some @xmath9 .
the angle @xmath158 between @xmath159 and @xmath18 is defined as @xmath160 , where @xmath161 is the ordinary angle between @xmath159 and @xmath162 .
we observe the following geometric fact , see also ( * ? ? ?
* lemma 1 ) : [ lem : angle ] let @xmath156 and @xmath9 .
one can find a constant @xmath163 only depending on @xmath30 and on @xmath11 such that @xmath164 for all @xmath165 .
let @xmath166 be the relative grassmannian of @xmath167-dimensional linear subspaces of @xmath7 that are contained in the hyperplane @xmath168 orthogonal to the @xmath169-dimensional linear subspace spanned by @xmath159 .
for @xmath170 let @xmath171 be such that @xmath172 .
it is clear that the @xmath11-dimensional linear subspace spanned by @xmath18 and @xmath173 is contained in the set @xmath174 we are interested in . since @xmath175 has dimension @xmath176 , the set of points @xmath171 with @xmath172 forms a spherical cap in the @xmath177-dimensional subsphere @xmath178 of @xmath179 .
it has radius of order @xmath180 and volume of order @xmath181 , where by volume we mean here the @xmath177-dimensional hausdorff measure on @xmath178 .
hence , for sufficiently small @xmath180 , @xmath164 and the proof is complete .
let the assumptions and the notation from the previous section prevail and denote for each @xmath110 by @xmath182 the normal cone of the simplex @xmath183 $ ] at the vertex @xmath148
. by construction we can find constants @xmath184 only depending on @xmath30 such that @xmath185 compare also with @xcite .
the set dual to @xmath186 , @xmath187 , is denoted by @xmath188 for @xmath110 put @xmath189 $ ] and define @xmath190|l)\,\nu_\ell(\dint l)\,,\qquad z\in\d_i^0\,.\ ] ] the next lemma provides a lower bound for the variance of these local functionals . [ lem : variancei ] fix @xmath35 , let @xmath110 and let @xmath191 be a point chosen with respect to the normalized gaussian measure restricted to @xmath141 .
then @xmath192 \gg ( \log n)^{-(d-\ell+1)}\,,\ ] ] where the notation @xmath193 $ ] refers to the variance that is taken with respect to @xmath194 .
denote by @xmath195 the centre of the facet of @xmath141 opposite to the vertex @xmath108 , and define the points @xmath196 and @xmath197 .
furthermore , the regions @xmath198 are given by @xmath199 and @xmath200 .
it is crucial to observe that one can find a constant @xmath201 only depending on @xmath30 such that @xmath202 for @xmath203 and @xmath204 .
together with the first part of lemma [ lem : contentdeltaij ] and the fact that the gaussian density @xmath3 from satisfies @xmath205 we see that the gaussian measure of @xmath206 is @xmath207 we let @xmath208 and @xmath209 and notice that , clearly , @xmath210\subset[z_i^1,f_i]$ ] .
next , fix some @xmath157 with @xmath211 and let @xmath212 be an orthonormal basis in @xmath18 such that @xmath213 . define the hyperplane @xmath214 and let @xmath215 be the half - space bounded by @xmath216 not containing the origin . finally , let us define the set @xmath217 .
the whole construction is illustrated in figure [ fig4 ] .
[ width=0.99]z4 ( 0,8)@xmath149 ( 440,10)@xmath150 ( 210,0)@xmath218 $ ] ( 217,240)@xmath18 ( 228,72)@xmath195 ( 226,110)@xmath219 ( 243,157)@xmath220 ( 240,200)@xmath108 ( 236,178)@xmath221 ( 250,100)@xmath222 ( 200,90)@xmath223 ( 220,178)@xmath224 ( 210,130)@xmath225 ( 320,200)@xmath226 ( 270,108)@xmath227 ( 95,60)@xmath228 ( 380,142)@xmath216 ( 413,78)@xmath141 we have that @xmath229 $ ] , @xmath230=\{w_i^2\}$ ] and that @xmath231 we conclude that @xmath232|l)-\vol_\ell([z_i^2,f_i]|l ) \geq \vol_\ell(g_i|l)\ ] ] and hence , @xmath233|l)-\vol_\ell([z_i^2,f_i]|l)\big)\,\nu_\ell(\dint l)\\ & \geq { d\choose \ell}{\k_d\over\k_\ell\k_{d-\ell } } \int_{\gg(d,\ell)}{\bf 1}_{\{l\cap\sig_i^2\neq\emptyset\}}\,\vol_\ell(g_i|l)\,\nu_\ell(\dint l)\\ & \gg ( \log n)^{-{1\over 2}}\,\nu_\ell(\{l\in\gg(d,\ell):l\cap\sig_i^2\neq\emptyset\})\\ & \gg ( \log n)^{-{1\over
2}}\,(\log n)^{-{d-\ell\over 2 } } \\ & = ( \log n)^{-{d-\ell+1\over 2}}\,,\end{aligned}\ ] ] where we used lemma [ lem : angle ] , and the definition of @xmath234 .
selecting now @xmath235 , @xmath187 , independently at random according to the normalized gaussian measure restricted to @xmath141 ( i.e. , @xmath221 and @xmath222 are independent copies of @xmath191 ) , we conclude that @xmath236 & = { 1\over 2}\be\big[\big(\widetilde{v}_\ell(z_i^1;f_i)-\widetilde{v}_\ell(z_i^2;f_i)\big)^2\big]\\ & \geq { 1\over 2}\be\big[\big(\widetilde{v}_\ell(z_i^1;f_i)-\widetilde{v}_\ell(z_i^2;f_i)\big)^2\,{\bf 1}_{r_i^1}(z_1){\bf 1}_{r_i^2}(z_2)\big]\\ & \gg ( \log n)^{-(d-\ell+1)}\,\bp(z_i^1\in r_i^1,z_i^2\in r_i^2)\,.\end{aligned}\ ] ] to obtain a lower bound for @xmath237 we recall and combine this with the second assertion of lemma [ lem : contentdeltaij ] as well as with the independence of the random points @xmath221 and @xmath222 .
this implies that @xmath238 with a constant @xmath239 only depending on @xmath30 .
hence , @xmath240 \gg ( \log n)^{-(d-\ell+1)}\,,\ ] ] completing thereby the proof of the lemma .
recall the geometric construction and its properties from the previous section and denote by @xmath241 the @xmath242-field generated by the random points @xmath6 , except those in the simplices @xmath141 for which @xmath243 , @xmath110 .
the conditional variance formula implies that @xmath244 = \be\big[\bv[v_\ell(k_n)|\cf]\big]+\bv\big[\be[v_\ell(k_n)|\cf]\big]\geq\be\big[\bv[v_\ell(k_n)|\cf]\big]\ , .
\end{split}\ ] ] now , conditioned on @xmath245 , suppose that @xmath243 , write @xmath191 for the ( unique ) random point in @xmath141 and denote by @xmath246 the convex hull of the random points in @xmath125 with @xmath137 .
we notice that if @xmath243 for each @xmath247 in a subset @xmath248 , then @xmath249 is a family of independent random variables as a consequence of lemma [ lem : conesci ] .
this independence property implies that @xmath250 = \sum_{i=1\atop { \bf 1}_{a_i}=1}^m\bv_i[v_\ell(k_n)]=\sum_{i=1\atop { \bf 1}_{a_i}=1}^m\bv_i[\widetilde{v}_\ell(z_i;f_i)]\,,\ ] ] where , as in the previous section , the notation @xmath193 $ ] refers to the variance that is taken only with respect to the point @xmath194 and we only sum over those @xmath110 with the property that @xmath243 .
these variances can be controlled by means of lemma [ lem : variancei ] , which implies that @xmath250 \gg ( \log n)^{-(d-\ell+1)}\,\sum_{i=1}^m{\bf 1}_{a_i}\,.\ ] ] taking expectation and finally applying lemma [ lem : probabilityai ] as well as lemma [ lem : estimatem ] , we arrive at @xmath251 & \gg ( \log n)^{-(d-\ell+1)}\,\sum_{i=1}^m\bp(a_i ) \\ & \gg ( \log n)^{-(d-\ell+1 ) } \times ( \log n)^{d-1\over 2 } \\ & = ( \log n)^{\ell-{d+3\over 2}}\,.\end{aligned}\ ] ] this completes the argument and the proof of theorem [ thm : main ] .
@xmath252 we are greatful to julian grote ( bochum ) for useful comments on the concentration inequality for @xmath55 .
+ imre brny was partially supported by erc advanced research grant no 267165 ( disconv ) and by hungarian national research grant k 111827 . | the intrinsic volumes of gaussian polytopes are considered . a lower variance bound for these quantities
is proved , showing that , under suitable normalization , the variances converge to strictly positive limits .
the implications of this missing piece of the jigsaw in the theory of gaussian polytopes are discussed . + * keywords*. gaussian polytopes , random polytopes , stochastic geometry , variance lower bound + * msc*. primary 60d05 ; secondary 52a22 . |
in this paper , we study combinatorial optimization problems with a diseconomy of scale .
we consider problems in which we need to minimize the cost of resources used to accomplish a certain task .
often , the cost grows linearly with the amount of resources used . in some applications , the cost is sublinear e.g. , if we can get a discount when we buy resources in bulk .
such phenomenon is known as economy of scale " . however , in many applications the cost is superlinear . in such cases , we say that the cost function exhibits a `` diseconomy of scale '' .
a good example of a diseconomy of scale is the cost of energy used for computing .
modern hardware can run at different processing speeds . as we increase the speed
, the energy consumption grows superlinearly .
it can be modeled as a function @xmath19 of the processing speed @xmath20 , where @xmath21 and @xmath4 are parameters that depend on the specific hardware .
typically , @xmath22 $ ] ( see e.g. , @xcite ) .
as a running example , consider the minimum power routing problem studied by andrews , fernndez anta , zhang , and zhao @xcite .
we are given a graph @xmath23 and a set of demands @xmath24 .
our goal is to route @xmath25 ( @xmath26 ) units of demand @xmath27 from the source @xmath28 to the destination @xmath29 such that every demand @xmath27 is routed along a single path @xmath30 ( i.e. we need to find an unsplittable multi - commodity flow ) .
we want to minimize the energy cost .
every link ( edge ) @xmath31 uses @xmath32 units of power , where @xmath33 is a scaling parameter depending on the link @xmath34 , and @xmath35 is the load on @xmath34 . the straightforward approach to solving this problem is as follows .
we define a mathematical programming relaxation that routes demands fractionally .
it sends @xmath36 units of demand via the path @xmath37 connecting @xmath38 to @xmath39 .
we require that @xmath40 for every demand @xmath27 .
the objective function is to minimize @xmath41 where @xmath42 is the load on the link @xmath34 .
this relaxation can be solved in polynomial time , since the objective function is convex ( for @xmath43 ) .
but , unfortunately , the integrality gap of this relaxation is @xmath44 @xcite .
andrews et al .
@xcite gave the following integrality gap example .
consider two vertices @xmath20 and @xmath45 connected via @xmath46 disjoint paths .
our goal is to route 1 unit of flow integrally from @xmath20 to @xmath45 .
the optimal solution pays 1 .
the lp may cheat by routing @xmath47 units of flow via @xmath46 disjoint paths .
then , it pays only @xmath48 . for the case of uniform demands ,
i.e. , for the case when all @xmath49 , andrews et al .
@xcite suggested a different objective function : @xmath50 the objective function is valid , because in the integral case , @xmath35 must be a multiple of @xmath51 , and thus @xmath52 .
andrews et al .
@xcite proved that the integrality gap of this relaxation is a constant .
bampis et al .
@xcite improved the bound to the _ fractional bell number _
@xmath53 that is defined as follows : @xmath53 is the @xmath4-th moment of the poisson random variable @xmath54 with parameter 1 ( see figure [ fig : plot - aq ] in appendix [ sec : figures ] ) . i.e. , @xmath55}=\sum_{t=1}^{+\infty}t^q \frac{e^{-1}}{t!}.\ ] ] for the case of general demands no constant approximation was known .
the best known approximation due to andrews et al .
@xcite was @xmath56 where @xmath57 is the number of demands and @xmath58 is the size of the largest demand ( theorem 8 in @xcite ) . in this work ,
we give an @xmath53-approximation algorithm for the general case and thus close the gap between the case of uniform and non - uniform demands .
our approximation algorithm uses a general framework for solving problems with a diseconomy of scale which we present in this paper .
we use this framework to obtain approximation algorithms for several other combinatorial optimization problems .
we give @xmath59-approximation algorithm for load balancing on unrelated parallel machines ( see section [ section : loadbalancing ] ) , @xmath60-approximation algorithm for unrelated parallel machine scheduling with non - linear functions of completion times ( see section [ section : unrelatedparallel ] ) and @xmath53-approximation algorithm for the minimum degree balanced spanning tree problem ( see section [ section : qst ] ) .
the best previously known bound for the first problem with @xmath61 $ ] was @xmath62 ( see figure [ fig : plot - results ] for comparison ) .
the bound is due to kumar , marathe , parthasarathy and srinivasan @xcite .
there were no known approximation guarantees for the latter problems . in the analysis
, we use the de la pea decoupling inequality @xcite .
let @xmath63 be jointly distributed nonnegative ( non - independent ) random variables , and let @xmath64 be independent random variables such that each @xmath6 has the same distribution as @xmath7 .
then , for every @xmath43 , @xmath65 for some universal constant @xmath66 .
de la pea , ibragimov , and sharakhmetov ( @xcite , corollary 3.4 ) showed that @xmath67 for @xmath68 $ ] , and @xmath69 for @xmath11 .
we give an alternative proof of this inequality , and show that the inequality holds for @xmath70 for any @xmath13 , and moreover this bound is tight ( for any @xmath13 ) .
thus , we improve the known upper bound for @xmath66 for @xmath9 .
[ thm : decoupling ] inequality ( [ eq : de - la - pena ] ) holds for @xmath71 , where @xmath53 is the fractional bell number ( see equation ( [ def : frac - bell - number ] ) and figure [ fig : plot - aq ] ) .
moreover , @xmath59 is the optimal upper bound on @xmath72 .
in fact we prove a more general inequality for arbitrary convex functions and an analogous inequality for concave functions . in section [ sec : neg - depend ] ( see corollary [ cor : neg - assoc ] ) , we extend this theorem to negatively associated random variables @xmath6 .
we now describe the general framework for solving problems with a diseconomy of scale .
we consider optimization problems with @xmath46 decision variables @xmath73 .
we assume that the objective function equals the sum of @xmath74 terms , where the @xmath75-th term is of the form @xmath76 here @xmath77 s are nonnegative monotonically nondecreasing convex functions , @xmath78 are parameters
. the vector @xmath79 must satisfy the constraint @xmath80 for some polytope @xmath81^n$ ] .
therefore , the optimization problem can be written as the following boolean convex program ( ip ) : @xmath82 } f_j \big(\sum_{i\in [ n ] } d_{ij } y_{i}\big ) \label{ip : obj}\\ & & y\in { \cal p } \label{ip : poly}\\ & & y\in \{0,1\}^{n } \label{ip : int}\end{aligned}\ ] ] we assume that we can optimize any linear function over the polytope @xmath83 in polynomial time ( e.g. , @xmath84 is defined by polynomially many linear inequalities , or there exists a separation oracle for @xmath84 ) .
thus , if we replace the integrality constraint ( [ ip : int ] ) with the relaxed constraint @xmath85^n$ ] ( which is redundant , since @xmath81^n$ ] ) , we will get a convex programming problem that can be solved in polynomial time ( see @xcite ) .
however , as we have seen in the example of minimum power routing , the integrality gap of the relaxation can be as large as @xmath44 for @xmath86 . in this work ,
we introduce a linear programming relaxation of ( [ ip : obj])-([ip : int ] ) that has an integrality gap of @xmath3 for @xmath87 under certain assumptions on the polytope @xmath84 .
we define auxiliary variables @xmath88 for all @xmath89 $ ] and @xmath90 $ ] . in the integral solution , @xmath91 if and only if @xmath92 for @xmath93 and @xmath94 for @xmath95 .
@xmath96 } \sum_{s\subseteq [ n ] } f_j \left ( \sum_{i\in s } d_{ij}\right)z_{js } \label{r : obj}\ ] ] @xmath97}z_{js}&=1 , & \forall j\in [ k ] \label{const1}\\ \sum_{s : i\in s}z_{js}&=y_i,&\forall i\in [ n ] , j\in [ k ] \label{const2}&{}\\ z_{js}&\ge 0 , & \forall s\subseteq [ n ] , j\in [ k ] \label{r : rel3}\end{aligned}\ ] ] in the integral solution , @xmath98 for all @xmath99 and @xmath100 . the reason why we introduced many copies of the same integral variable @xmath101 to the lp is that the lp above is easier to solve than the lp with an extra constraint @xmath98 .
optimization problem ( [ r : obj])-([r : rel3 ] ) is a relaxation of the original problem ( [ ip : obj])-([ip : int ] ) .
the lp has exponentially many variables .
we show , however , that the optimal solution to this lp can be found in polynomial time up to an arbitrary accuracy @xmath102 .
we shall assume that all @xmath103 are integral and polynomially bounded , or , more generally , that all @xmath103 are multiples of some @xmath104 and that @xmath105 are polynomially bounded in @xmath46 and @xmath106 .
given an arbitrary instance of the problem , it is easy to round all @xmath103 s to multiples of a sufficiently small @xmath104 , so that the cost of any solution changes by at most @xmath102 assuming that functions @xmath77 satisfy some mild conditions . in section [ sec : discr ]
( theorem [ thm : discr ] ) , we show how to pick @xmath104 for functions @xmath77 satisfying the following conditions : 1 .
each @xmath77 is a convex increasing function ; 2 . for each @xmath75 , @xmath107
; 3 . for each @xmath75 and @xmath108 $ ] , @xmath109 , where @xmath37 is some polynomial .
note , that for @xmath110 , we have @xmath111
. so functions @xmath112 satisfy the conditions of theorem [ thm : discr ] if @xmath113 are polynomially bounded .
[ thm : efficient ] suppose that there exists a polynomial time separation oracle for the polytope @xmath83 , @xmath114 is computable in polynomial time as a function of @xmath75 and @xmath45 , and all @xmath103 are multiples of @xmath104 such that @xmath105 are polynomially bounded in @xmath46 and @xmath106 .
then , for every @xmath115 , there exists a polynomial time algorithm that finds a @xmath102-approximately optimal solution to lp ( [ r : obj])-([r : rel3 ] ) . for a convex non - decreasing function @xmath116
define @xmath117 as follows : @xmath118,\ ] ] where @xmath18 is a poisson random variable with parameter 1 .
note that @xmath119 = { \mathbb{e}}[p^q ] = a_q.\ ] ] we prove the following theorem .
[ thm : main ] let @xmath120 .
assume that there exists a randomized algorithm @xmath121 that given a @xmath80 , returns a random integral point @xmath122 in @xmath123 such that 1 .
@xmath124 for all @xmath27 ( where @xmath125 is the @xmath27-th coordinate of @xmath122 ) ; 2 .
random variables @xmath126 are independent or negatively associated ( see section [ sec : neg - depend ] for the definition ) for every @xmath75 .
then , for every feasible solution @xmath127 to lp ( [ r : obj])-([r : rel3 ] ) , we have @xmath128 } f_j \big(\sum_{i\in [ n ] } d_{ij } r_i(y^*)\big ) \big]\leq \sum_{j\in [ k ] } a(f_j ) \sum_{s\subseteq [ n ] } f_j\big ( \sum_{i\in s } d_{ij}\big)z^*_{js},\ ] ] where @xmath129 is defined as in ( [ eq : def - general - a ] ) .
particularly , since lp ( [ r : obj])-([r : rel3 ] ) is a relaxation for ip ( [ ip : obj])-([ip : int ] ) , if @xmath127 is a @xmath102-approximately optimal solution to lp ( [ r : obj])-([r : rel3 ] ) , then @xmath130 } f_j \big(\sum_{i\in [ n ] } d_{ij } r_i(y^*)\big ) \big]\leq ( 1+\varepsilon )
\max_j(a(f_j ) ) \,ip,\ ] ] where @xmath131 is the optimal cost of the boolean convex program ( [ ip : obj])-([ip : int ] ) .
this theorem guarantees that an algorithm @xmath121 satisfying conditions ( 1 ) and ( 2 ) has an approximation ratio of @xmath132 for @xmath87 . in the next section , section [ sec : applications ] ,
we show how to use the framework to obtain approximation algorithms for four different combinatorial optimization problems .
then , in section [ sec : thm : efficient ] , we give an efficient algorithm for solving lp ( [ r : obj])-([r : rel3 ] ) . in section [ sec : thm : main ] , we prove the main theorem
theorem [ thm : main ] .
the proof easily follows from the decoupling inequality , which we prove in section [ sec : thm : decoupling ] . finally , in section [ sec : generalizations ]
, we describe some generalizations of our framework .
in this section , we show applications of our general technique .
we start with the problem discussed in the introduction energy efficient routing .
recall , that andrews et al .
@xcite gave an @xmath56-approximation algorithm for this problem where @xmath133 and @xmath134 ( theorem 8 in @xcite ) .
we give an @xmath135-approximation algorithm for any fixed @xmath136 .
we write a standard integer program . each variable
@xmath137 indicates whether the edge @xmath34 is used to route the flow from @xmath38 to @xmath39 .
below , @xmath138 denotes the set of edges outgoing from @xmath139 ; @xmath140 denotes the set of edges incoming to @xmath139 .
@xmath141 @xmath142 using theorem [ thm : efficient ] , we obtain an almost optimal fractional solution @xmath143 of lp relaxation ( [ r : obj])-([r : rel3 ] ) of ip ( [ energy_obj])-([eqn : r5 ] ) .
we apply randomized rounding in order to select a path for each demand .
specifically , for each demand @xmath144 , we consider the standard flow decomposition into paths : in the decomposition , each path @xmath37 connecting @xmath38 to @xmath39 has a weight @xmath145 .
for every edge @xmath34 , @xmath146 ; and @xmath147 . for each @xmath27 ,
the approximation algorithm picks one path @xmath37 connecting @xmath38 to @xmath39 at random with probability @xmath148 , and routes all demands from @xmath38 to @xmath30 via @xmath37 .
thus , the algorithm always obtains a feasible solution .
we verify that the integral solution corresponding to this combinatorial solution satisfies the conditions of theorem [ thm : main ] .
let @xmath149 be the integral solution , i.e. , let @xmath150 if the edge @xmath34 is chosen in the path connecting @xmath38 and @xmath39 .
first , @xmath150 if the path connecting @xmath38 and @xmath39 contains @xmath34 , thus @xmath151 second , the paths for all demands are chosen independently .
each @xmath149 depends only on paths that connect @xmath38 to @xmath39 .
thus all random variables @xmath149 ( for a fixed @xmath34 ) are independent . therefore , by theorem [ thm : main ] , the cost of the solution obtained by the algorithm is bounded by @xmath152 , where @xmath153 is the cost of the optimal solution to the integer program which is exactly equivalent to the minimum energy efficient routing problem
. we are given @xmath46 jobs and @xmath154 machines .
the processing time of the job @xmath155 $ ] assigned to the machine @xmath156 $ ] is @xmath157 .
the goal is to assign jobs to machines to minimize the @xmath158-norm of machines loads .
formally , we partition the set of jobs into @xmath154 sets @xmath159 to minimize @xmath160 } ( \sum_{j\in s_i } p_{ij})^q\right)^{1/q}$ ] .
this is a classical scheduling problem which is used to model load balancing in practice } ( \sum_{j\in s_i } p_{ij})^q$ ] , can be used for energy efficient scheduling .
imaging that we need to assign @xmath46 jobs to @xmath154 processors / cores so that all jobs are completed by a certain deadline @xmath161 .
we can run processors at different speeds @xmath38 . to meet the deadlines we must set @xmath162 .
the total power consumption is proportional to @xmath163 . for this problem
, our algorithm gives @xmath164 approximation . ] .
it was previously studied by azar and epstein @xcite and by kumar , marathe , parthasarathy and srinivasan @xcite .
particular , for @xmath165 $ ] the best known approximation algorithm has performance guarantee @xmath62 @xcite ( theorem 4.4 ) .
we give @xmath166{(1+\varepsilon)a_q}$]-approximation algorithm for any @xmath136 substantially improving upon previous results ( see figure [ fig : plot - results ] ) .
we formulate the unrelated parallel machine scheduling problem as a boolean nonlinear program : @xmath167 } \big(\sum_{j\in [ n ] } & p_{ij}x_{ij}\big)^q & \label{sch_obj}\\ \sum_{i\in [ m]}x_{ij}&=1 , & \forall j\in [ n ] \label{sch_eqn : r2}\\
x_{ij}&\in \{0,1\ } , & \forall i\in [ m],\ , j\in [ n ] \label{sch_eqn : r5}\end{aligned}\ ] ] using theorem [ thm : efficient ] , we obtain an almost optimal fractional solution @xmath168 of the lp relaxation ( [ r : obj])([r : rel3 ] ) corresponding to the ip ( [ sch_obj])([sch_eqn : r5 ] ) .
we use the straightforward randomized rounding : we assign each job @xmath75 to machine @xmath27 with probability @xmath169 .
we claim that , by theorem [ thm : main ] , the expected cost of our integral solution is upper bounded by @xmath53 times the value of the fractional solution @xmath168 . indeed , the probability that we assign a job @xmath75 to machine @xmath27 is exactly equal to @xmath169 ; and we assign job @xmath75 to machine @xmath27 independently of other jobs .
that implies that our approximation algorithm has a performance guarantee of @xmath166{(1+\varepsilon)a_q}$ ] for the @xmath158-norm objective .
as in the previous problem , in unrelated parallel machine scheduling with nonlinear functions of completion times , we are given @xmath46 jobs and @xmath154 machines .
the processing time of the job @xmath155 $ ] assigned to the machine @xmath156 $ ] is @xmath157 . we need to assign jobs to machines and set their start times such that job processing intervals do not overlap .
the goal is to minimize @xmath170 where @xmath171 is the completion time of job @xmath75 in the schedule and @xmath172 .
using classical scheduling notation this problem can be denoted as @xmath173 . the problem @xmath173 is well studied for @xmath174 .
it is known to be apx - hard @xcite while the best known approximation algorithm has a performance guarantee of @xmath175 @xcite . for @xmath176
even the single machine scheduling problem is not understood : it is an open problem whether @xmath177 is @xmath178-hard for @xmath179 , @xmath180 .
bansal and pruhs @xcite and stiller and wiese @xcite gave constant factor approximation algorithms for more general functions of completion times for a single machine .
however , there were no known approximation algorithms for multiple machines .
we show how to use our framework for this problem in appendix [ section : unrelatedparallel - tech ] .
our algorithm gives @xmath181 approximation .
we are given an undirected graph @xmath23 with edge weights @xmath182 .
the goal is to find a spanning tree @xmath183 minimizing the objective function @xmath184 where @xmath185 is the set of edges in @xmath186 incident to the vertex @xmath187 . for @xmath188 ,
a more general problem was considered before in the operations research literature @xcite under the name of adjacent only quadratic spanning tree problem . a related problem , known as degree bounded spanning tree , received a lot of attention in theoretical computer science @xcite .
we are not aware of any previous work on degree balanced spanning tree problem .
let @xmath35 be a boolean decision variable such that @xmath189 if we choose edge @xmath31 to be in our solution ( tree ) @xmath183 .
we formulate our problem as the following convex boolean optimization problem @xmath190 where @xmath191 is the base polymatroid polytope of the graphic matroid in graph @xmath192 .
we refer the reader to schrijver s book @xcite for the definition of the matroid . using theorem [ thm : efficient ]
, we obtain an almost optimal fractional solution @xmath193 of lp relaxation ( [ r : obj])-([r : rel3 ] ) corresponding to the above integer problem . following calinescu et al .
@xcite , we define the continuous extension of the objective function ( [ tree : objective ] ) for any fractional solution @xmath194 @xmath195}f(s)\prod_{e\in s}x'_e\prod_{e\not\in s}(1-x'_e),\ ] ] i.e. @xmath196 is equal to the expected value of the objective function ( [ tree : objective ] ) for the set of edges sampled independently at random with probabilities @xmath197 . the function @xmath198 can be approximated with arbitrary polynomially small precision efficiently via sampling ( see @xcite ) . by theorem
[ thm : main ] , we get the bound @xmath199 , where @xmath200 is the value of the lp relaxation ( [ r : obj])([r : rel3 ] ) on the fractional solution @xmath193 .
the rounding phase of the algorithm implements the pipage rounding technique @xcite adopted to polymatroid polytopes by calinescu et al .
calinescu et al .
@xcite showed that given a matroid @xmath201 and a fractional solution @xmath202 , one can efficiently find two elements , or two edges in our case , @xmath203 and @xmath204 such that the new fractional solution @xmath205 defined as @xmath206 , @xmath207 and @xmath208 for @xmath209 is feasible in the base polymatroid polytope for small positive and for small negative values of @xmath210 .
they also showed that if the objective function @xmath211 is submodular then the function of one variable @xmath212 is convex . in our case ,
the objective function @xmath211 is supermodular which follows from a more general folklore statement .
the function @xmath213 is supermodular if @xmath214 for @xmath215 $ ] and @xmath216 is a convex function of one variable .
therefore , the function @xmath212 is concave .
hence , we can apply the pipage rounding directly : we start with the fractional solution @xmath193 . at every step ,
we pick @xmath203 and @xmath204 ( using the algorithm from @xcite ) and move to @xmath205 with @xmath217 or @xmath218 whichever minimizes the concave function @xmath212 on the interval @xmath219 $ ] .
we stop when the current solution @xmath220 is integral . at every step ,
we decrease the number of fractional variables @xmath35 by at least 1 .
thus , we terminate the algorithm in at most @xmath221 iterations .
the value of the function @xmath222 never increases .
so the cost of the final integral solution is at most the cost of the initial fractional solution @xmath193 , which , in turn , is at most @xmath223 .
note , that we have not used any special properties of graphic matroids .
the algorithm from @xcite works for general matroids accessible through oracle calls .
so we can apply our technique to more general problems where the objective is to minimize a function like ( [ tree : objective ] ) subject to base matroid constraints .
we now give an efficient algorithm for finding @xmath102 approximately optimal solution to lp ( [ r : obj])-([r : rel3 ] ) .
observe that for every @xmath80 , there exists a @xmath224 such that the pair @xmath143 is a feasible solution to lp ( [ r : obj])-([r : rel3 ] ) .
for example , one such @xmath224 is defined as @xmath225 . of course
, this particular @xmath224 may be suboptimal .
however , it turns out , as we show below , that for every @xmath226 , we can find the optimal @xmath224 efficiently .
let us denote the minimal cost of the @xmath75-th term in ( [ r : obj ] ) for a given @xmath80 by @xmath227 .
that is , @xmath227 is the cost of the following lp .
the variables of the lp are @xmath88 .
the parameters @xmath80 and @xmath90 $ ] are fixed .
@xmath228 } f_j\big ( \sum_{i\in s } d_{ij}\big)z_{js } \label{r - eff : obj}\ ] ] @xmath229}z_{js}=1 & & \label{const1-eff}\\ \sum_{s : i\in s}z_{js}=y_i , & & \forall i\in [ n]\label{const2-eff}\\ z_{js}\ge 0 , & & \forall s\subseteq [ n ] \label{r - eff : rel3}\end{aligned}\ ] ] now , lp ( [ r : obj])-([r : rel3 ] ) can be equivalently rewritten as ( below @xmath226 is the variable ) .
@xmath230 } h_j(y ) \label{lp : eff : obj}\\ y\in { \cal p } & & \label{lp : eff : poly}\end{aligned}\ ] ] the functions @xmath227 are convex and @xmath231 are the optimal solutions for vectors @xmath232 and @xmath233 , then @xmath234 is a feasible solution for @xmath235 .
hence , @xmath236 .
see section [ sec : convexfj ] in appendix for details . ] . in lemma
[ lem : find - z ] ( see below ) , we prove that lp ( [ r - eff : obj])-([r - eff : rel3 ] ) can be solved in polynomial time , and thus the functions @xmath227 can be computed efficiently .
the algorithm for finding @xmath227 also returns a subgradient of @xmath237 at @xmath226 .
hence , the minimum of convex problem ( [ lp : eff : obj])-([lp : eff : poly ] ) can be found using the ellipsoid method .
once the optimal @xmath232 is found , we find @xmath238 by solving lp ( [ r - eff : obj])-([r - eff : rel3 ] ) for @xmath232 and each @xmath90 $ ] . [
lem : find - z ] there exists a polynomial time algorithm for computing @xmath237 and finding a subgradient of @xmath237 . we need to solve lp ( [ r - eff : obj])-([r - eff : rel3 ] ) .
recall that in section [ sec : discr ] ( theorem [ thm : discr ] ) we show how to choose @xmath239 such that each @xmath105 is an integer polynomially bounded . for simplicity
we assume that @xmath240 ( the proof in general case is almost identical ) .
therefore @xmath103 are integral in this case and polynomially bounded .
we write the dual lp .
we introduce a variable @xmath241 for constraint ( [ const1-eff ] ) and variables @xmath242 for constraints ( [ const2-eff ] ) .
@xmath243\label{dual : constr}\end{aligned}\ ] ] the lp has exponentially many constraints .
however , finding a violated constraint is easy .
to do so , we guess @xmath244 for the set @xmath245 violating the constraint . that is possible , since all @xmath103 are polynomially bounded , and so is @xmath246
. then we solve the maximum knapsack problem @xmath247 } \sum_{i\in s}\eta_{i}\\ \sum_{i\in s } d_{ij}=b^*\end{aligned}\ ] ] using the standard dynamic programming algorithm and obtain the optimal set @xmath245 .
the knapsack problem is polynomially solvable , since @xmath246 is polynomially bounded . if @xmath248 , then constraint ( [ dual : constr ] ) is violated for the set @xmath245 ; otherwise all constraints ( [ dual : constr ] ) are satisfied .
let @xmath249 be the optimal solution of the dual lp .
the value of the function @xmath227 equals the objective value of the dual lp .
a subgradient of @xmath237 at @xmath226 is given by the equation @xmath250 this is a subgradient of @xmath237 , since @xmath249 is a feasible solution of the dual lp for every @xmath251 ( note that constraint ( [ dual : constr ] ) does not depend on @xmath226 ) , and , hence , ( [ eq : subgrad ] ) is a lower bound on @xmath252 .
in this section , we prove the main theorem theorem [ thm : main ] .
the theorem easily follows from the de la pea decoupling inequality ( theorem [ thm : decoupling ] and corollary [ cor : neg - assoc ] ) for @xmath87 , and from the more general inequality presented in theorem [ thm : cx - main ] ( see also corollary [ cor : a - f - inequality ] ) for arbitrary convex functions @xmath77 . consider a feasible solution @xmath127 to ip ( [ ip : obj])-([ip : int ] ) .
we prove inequality ( [ eq : thm : main ] ) term by term .
that is , for every @xmath75 we show that @xmath253\leq a(f_j ) \sum_{s\subseteq [ n ] } f_j\big ( \sum_{i\in s\cap d_j } d_{ij}\big)z^*_{js}.\ ] ] recall that @xmath254 . above , we dropped terms with @xmath255 , since if @xmath255 , then @xmath256 . fix a @xmath155 $ ] . define random variables @xmath7 for @xmath257 as follows : pick a random set @xmath89 $ ] with probability @xmath88 , and let @xmath258 if @xmath93 , and @xmath259 otherwise .
note that random variables @xmath7 are dependent .
we have @xmath260 it is easy to see that @xmath261 } y_i \big)\big ] = \sum_{s\subseteq [ n ] } f_j\big ( \sum_{i\in s } d_{ij}\big ) z_{js}.\ ] ] the right hand side is simply the definition of the expectation on the left hand side .
now , let @xmath262 for @xmath257 .
note that by conditions of the theorem , @xmath263 ( by condition ( 1 ) ) .
thus , each @xmath6 has the same distribution as @xmath7 .
furthermore , @xmath6 s are independent or negatively associated ( by condition ( 2 ) ) .
therefore , we can apply the decoupling inequality from theorem [ thm : cx - main ] ( see corollary [ cor : a - f - inequality ] ) @xmath264 \leq a(f_j ) { \mathbb{e}}\big[f_j\big(\sum_{i \in d_j } y_i\big)\big].\ ] ] the left hand side of the inequality equals the left hand side of ( [ eq : thm : main : need - to - prove ] ) , the right hand side of the inequality equals the right hand side of ( [ eq : thm : main : need - to - prove ] ) .
hence , inequality ( [ eq : thm : main : need - to - prove ] ) holds .
in this section , we prove the decoupling inequality ( theorem [ thm : decoupling ] ) with the optimal constant @xmath12 . in fact , we prove a more general inequality which works for arbitrary convex functions . to state the inequality we need the notion of _ convex stochastic order_. we say that a random variable @xmath265 is less than @xmath266 in the convex ( stochastic ) order , and write @xmath267 if for every convex function @xmath268 , @xmath269 whenever both expectations exist .
if @xmath267 and @xmath16 is a _ concave _ function , then @xmath270\geq { \mathbb{e}}[\psi(y)]$ ] , since the function @xmath271 is convex , and therefore @xmath272\leq -{\mathbb{e}}[\psi(y)]$ ] .
it is easy to see that the convex stochastic order defines a partial order on all random variables .
particularly , if @xmath267 and @xmath273 , then @xmath274 . note that the definition depends only on the distributions of @xmath265 and @xmath266 .
that is , if @xmath265 has the same distribution as @xmath275 and @xmath267 , then @xmath276 .
the random variables @xmath265 and @xmath266 may be defined on the same probability space or on different probability spaces .
we refer the reader to the book of shaked and shanthikumar @xcite for a detailed introduction to stochastic orders .
we now state the general inequality in terms of the convex order .
part ii of the theorem shows that the inequality is tight .
[ thm : cx - main ] i. let @xmath63 be jointly distributed nonnegative ( non - independent ) random variables , and let @xmath64 be independent random variables such that each @xmath6 has the same distribution as @xmath7 .
let @xmath18 be a poisson random variable with parameter 1 independent of @xmath7 s .
then , @xmath277 \ii . for every nonnegative function @xmath14 with a finite expectation
@xmath278 $ ] and every positive @xmath210 , there exists @xmath46 and random variables @xmath279 , @xmath63 as in part i such that @xmath280}\geq ( 1+\varepsilon ) { { \mathbb{e}}\left[\varphi\big(p \sum_{i=1}^n y_i \big)\right]}.\ ] ] theorem [ thm : cx - main ] implies theorem [ thm : decoupling ] , since the function @xmath281 is convex and thus @xmath282 } \leq { { \mathbb{e}}\left[\big(p\cdot ( y_1+\dots+y_n)\big)^q\right]}\\ & = { { \mathbb{e}}\left[p^q\right]}\cdot { { \mathbb{e}}\left[(y_1+\dots+y_n)^q\right ] } = a_q { { \mathbb{e}}\left[(y_1+\dots+y_n)^q\right]}\\ & = a_q\|y_1+\dots+y_n\|_q^q.\end{aligned}\ ] ] part ii of theorem [ thm : cx - main ] shows that we can not replace @xmath3 with a smaller constant . another immediate corollary of theorem [ thm : cx - main ] is as follows .
[ cor : a - f - inequality ] for an arbitrary nonnegative convex function @xmath116 ,
@xmath283 \leq a(f ) { \mathbb{e}}[f(y_1+\dots+y_n)],\ ] ] where @xmath117 is defined in ( [ eq : def - general - a ] ) .
write , @xmath283 \leq { \mathbb{e}}\big[f\big(p \cdot(y_1+\dots+y_n)\big)\big ] = { \mathbb{e}}{\mathbb{e}}\big[f\big(p \cdot(y_1+\dots+y_n)\big){\,\mid\,}y_1+\dots+y_n\big].\ ] ] for any @xmath284 , particularly for @xmath285 , we have @xmath286\leq a(f ) f(t)$ ] , hence @xmath283 \leq { \mathbb{e}}[a(f ) f(y_1+\dots+y_n ) ] = a(f ) { \mathbb{e}}[f(y_1+\dots+y_n)].\ ] ] we first prove part ii of theorem [ thm : cx - main ] .
consider the following example .
let @xmath287 , be random variables taking value @xmath288 with probability @xmath47 , and @xmath289 with probability @xmath290 .
we generate @xmath291 s as follows .
we pick a random @xmath292 $ ] and let @xmath293 and @xmath294 for @xmath295 .
random variables @xmath296 are i.i.d .
bernoulli random variables with @xmath297=1/n$ ] . then
, the sum @xmath298 always equals 1 , and @xmath299 } = { { \mathbb{e}}\left[\varphi(p)\right]}$ ] . as @xmath300
, the sum @xmath301 converges in distribution to @xmath18 ( by the poisson limit theorem ) .
thus ( see lemma [ lem : limsup ] for details ) , @xmath302 } \geq { { \mathbb{e}}\left[\varphi(p)\right]},\ ] ] and , hence , for some @xmath46 inequality ( [ eq : cx - counterexample ] ) holds . before proceeding to the proof of part i , we state some known properties of the convex order . [ lem : sumx - less - sumy ] suppose @xmath64 are independent random variables and @xmath63 are independent random variables . if @xmath303 for all @xmath27 , then @xmath304 .
[ lem : sum - ax - less - ay ] consider random variables @xmath279 and @xmath266 . if @xmath305 for all @xmath27 , then @xmath306 , for any sequence of nonnegative numbers @xmath307 . for completeness , we prove lemma [ lem : sumx - less - sumy ] and lemma [ lem : sum - ax - less - ay ] in appendix .
to simplify the proof , we will use the following easy lemma .
[ lem : simplify ] if for random variables @xmath265 and @xmath266 condition ( [ eq : def - cx ] ) is satisfied for all convex functions @xmath14 with @xmath308 , then @xmath267 .
consider an arbitrary convex function @xmath14 .
let @xmath309 .
since @xmath310 , we have @xmath311 = { \mathbb{e}}[\tilde{\varphi}(x ) ] + \varphi(0 ) \leq { \mathbb{e}}[\tilde{\varphi}(y ) ] + \varphi(0 ) = { \mathbb{e}}[\varphi(y)].\ ] ] first , we prove that a bernoulli random variable @xmath312 can be upper bounded in the convex order by a poisson random variable @xmath18 with @xmath313 = { \mathbb{e}}[b]$ ] .
[ lem : b - cx - p ] let @xmath18 be an integral random variable , and @xmath312 be a bernoulli random variable with parameter @xmath314 $ ] . then
, @xmath315 .
particularly , if @xmath18 is a poisson random variable with @xmath313={\mathbb{e}}[b]$ ] , then @xmath315 .
consider an arbitrary convex function @xmath316 with @xmath308 ( see lemma [ lem : simplify ] ) .
define linear function @xmath317 as @xmath318 .
the graph of @xmath319 intersects the graph of @xmath14 at points @xmath320 and @xmath321 .
since @xmath14 is convex , we have @xmath322 for @xmath323 . hence , for every integral @xmath74 , @xmath324 , and @xmath325 .
consequently , @xmath326 } \geq { { \mathbb{e}}\left[l(p)\right ] } = l({{\mathbb{e}}\left[p\right ] } ) = l({{\mathbb{e}}\left[b\right ] } ) = \varphi(1 ) { { \mathbb{e}}\left[b\right ] } = { \mathbb{e}}[\varphi(b)].\ ] ] now we consider a very special case of theorem [ thm : cx - main ] when all @xmath6 s and @xmath7 s are bernoulli random variables scaled by a factor @xmath327 , and all events @xmath328 are mutually exclusive .
as we see later , the general case can be easily reduced to this special case .
[ lem : main - cx - lemma ] consider bernoulli random variables @xmath329 such that @xmath330 , and all events @xmath331 are mutually exclusive .
( that is , with probability 1 one and only one @xmath332 equals 1 . )
let @xmath333 be independent bernoulli random variables such that @xmath334 .
then , for all nonnegative numbers @xmath335 , we have @xmath336 where @xmath18 is a poisson random variable with parameter 1 independent of @xmath332 s .
i. we prove the first inequality .
consider an arbitrary convex function @xmath268 with @xmath337 .
the function @xmath14 is superadditive i.e. , @xmath338 for all positive @xmath339 and @xmath340 ( since the derivative of @xmath14 is monotonically non - decreasing , and @xmath308 ) .
hence , we have @xmath341 and @xmath342}\geq { { \mathbb{e}}\left[\sum_{i=1}^n b_i \varphi(\alpha_i)\right ] } = \sum_{i=1}^n\pr(b_i = 1 ) \varphi(\alpha_i ) \\= \sum_{i=1}^n\pr(\chi_i = 1 ) \varphi(\alpha_i ) = { { \mathbb{e}}\left[\varphi\big(\sum_{i=1}^n \alpha_i \chi_i\big)\right]}.\end{gathered}\ ] ] \ii .
we prove the second inequality . using lemma [ lem : b - cx - p ] and lemma [ lem : sumx - less - sumy ]
, we replace bernoulli random variable @xmath333 with independent poisson random variables @xmath343 satisfying @xmath344 . the sum @xmath345 is distributed as a poisson random variable with parameter 1 . by coupling random variables @xmath346 and @xmath18 , we may assume that @xmath347 .
consider an arbitrary convex function @xmath14 with @xmath337 .
using convexity of the function @xmath14 , we derive @xmath348 } & = \sum_{k=1}^{\infty } { { \mathbb{e}}\left[\varphi\big(\sum_{i=1}^n \alpha_i p_i \big ) { \,\mid\,}p = k\right]}\pr(p = k)\\ & = \sum_{k=1}^{\infty } { { \mathbb{e}}\left[\varphi\big(\sum_{i=1}^n \frac{p_i}{k } \cdot \alpha_i k \big ) { \,\mid\,}p = k\right]}\pr(p = k)\\ & \leq \sum_{k=1}^{\infty } { { \mathbb{e}}\left [ \sum_{i=1}^n \frac{p_i}{k } \varphi(\alpha_i
k ) { \,\mid\,}p = k\right]}\pr(p = k)\\ & = \sum_{k=1}^{\infty } \sum_{i=1}^n { { \mathbb{e}}\left[\frac{p_i}{k}{\,\mid\,}p = k\right ] } \varphi(\alpha_i k)\pr(p = k).\end{aligned}\ ] ] we observe that @xmath349}= k\,{{\mathbb{e}}\left[p_i\right]}$ ] , which follows from the following well known fact ( see e.g. , feller @xcite , section ix.9 , problem 6(b ) , p. 237 ) .
[ fact : p12 ] suppose @xmath350 and @xmath351 are independent poisson random variables with parameters @xmath352 and @xmath353
. then , for every @xmath354 , @xmath355 = \frac{\lambda_a}{\lambda_a+\lambda_b}\;k$ ] . in our case , @xmath356 , @xmath357 , @xmath358 .
therefore , we have @xmath359 & \leq \sum_{k=1}^{\infty } \sum_{i=1}^n { \mathbb{e}}\big[\frac{p_i}{k}{\,\mid\,}p = k\big ] \varphi(\alpha_i k)\pr(p = k)\\ & = \sum_{k=1}^{\infty } \sum_{i=1}^n { { \mathbb{e}}\left[p_i\right ] } \varphi(\alpha_i k)\pr(p = k)\\ & = \sum_{i=1}^n { { \mathbb{e}}\left[\chi_i\right ] } \sum_{k=1}^{\infty } \varphi(\alpha_i k)\pr(p = k)\\ & = \sum_{i=1}^n { { \mathbb{e}}\left[\chi_i\right ] } { { \mathbb{e}}\left[\varphi(\alpha_i k)\right]}\\ & = { \mathbb{e}}\big[\varphi(p \sum_{i=1}^n \alpha_i \chi_i)\big].\end{aligned}\ ] ] denote by @xmath360 the support of the random vector @xmath361 . each @xmath7 can be represented as follows @xmath362 where @xmath363 is the indicator of the event @xmath364 .
here we assume that @xmath360 is finite .
we treat the general case in appendix [ sec : crv ] . applying lemma [ lem : main - cx - lemma ] to the random variables
@xmath363 , we get @xmath365 where @xmath366 are independent bernoulli random variables .
each @xmath366 is distributed as the random variable @xmath363 ; that is , @xmath367 .
since each @xmath6 has the same distribution as @xmath7 , we have @xmath368 by lemma [ lem : sumx - less - sumy ] , we can sum up this inequality over all @xmath27 from @xmath288 to @xmath46 : @xmath369 we apply lemma [ lem : sum - ax - less - ay ] to every sum in parentheses : @xmath370 again , using lemma [ lem : sumx - less - sumy ] , we get @xmath371 finally , by lemma [ lem : main - cx - lemma ] , @xmath372 this concludes the proof of theorem [ thm : cx - main ] .
the decoupling inequalities ( [ eq : de - la - pena ] ) and ( [ eq : cx - main ] ) can be extended to _ negatively associated _ random variables @xmath279 .
the notion of negative association is defined as follows .
random variables @xmath64 are negatively associated if for all disjoint sets @xmath373 $ ] and all non - decreasing functions @xmath374 and @xmath375 the following inequality holds : @xmath376 } \leq { { \mathbb{e}}\left[f(x_i , i\in i)\right ] } \cdot { { \mathbb{e}}\left[g(x_j , j\in j)\right]}.\ ] ] shao @xcite showed that if @xmath279 are negatively associated random variables , and @xmath377 are independent random variables such that each @xmath378 is distributed as @xmath6 , then for every convex function @xmath268 , @xmath379 } \leq { { \mathbb{e}}\left [ \varphi ( x^*_1+\dots+x^*_n)\right]}.\ ] ] in other words , @xmath380 ( see also theorem 3.a.39 in @xcite ) . [ cor : neg - assoc ]
let @xmath63 be jointly distributed nonnegative ( non - independent ) random variables , and let @xmath64 be negatively associated random variables such that each @xmath6 has the same distribution as @xmath7 .
let @xmath18 be a poisson random variable with parameter 1 independent of the random variables @xmath7 .
then , @xmath381 particularly , for every convex nonnegative @xmath116 , @xmath382 and for @xmath43 , @xmath383 where @xmath117 is defined in ( [ eq : def - general - a ] ) , and @xmath53 is the fractional bell number .
we can extend our results to maximization problems with the objective function @xmath384 } f_j\big(\sum_{i\in [ n ] } d_{ij } y_{i}\big),\label{eq : gener}\ ] ] if @xmath77 s are arbitrary non - decreasing nonnegative _ concave _ functions defined on @xmath385 .
the approximation ratio equals @xmath386 , where @xmath387.\ ] ] it is not hard to see that @xmath388 for all @xmath116 . indeed , if @xmath389 , then @xmath390 , thus @xmath391 \geq \pr(p\geq 1 ) = 1 - 1/e$ ] .
this bound is tight if @xmath392 for @xmath389 .
for example , @xmath393 for the function @xmath394 . note that the approximation ratio of @xmath395 for maximization problems of this form was previously known ( see calinescu et al .
however , for some concave functions @xmath116 we get a better approximation .
for example , for @xmath396 , we get an approximation ratio of @xmath397 .
we would like to thank the anonymous referees for valuable comments . 2 alexander a. ageev and maxim sviridenko .
pipage rounding : a new method of constructing algorithms with proven performance guarantee .
8(3 ) : 307 - 328 , ( 2004 ) .
s. albers .
energy - efficient algorithms .
acm 53(5 ) , 86 - 96 ( 2010 ) .
matthew andrews , antonio fernandez anta , lisa zhang , wenbo zhao .
routing for power minimization in the speed scaling model .
ieee / acm trans .
20(1 ) : 285 - 294 ( 2012 ) .
matthew andrews , spyridon antonakopoulos , lisa zhang .
minimum - cost network design with ( dis)economies of scale .
focs 2010 , pp . 585 - 592 .
a. assad and w. xu .
the quadratic minimum spanning tree problem . naval research logistics v39 .
( 1992 ) , pp .
399 - 417 .
yossi azar , amir epstein .
convex programming for scheduling unrelated parallel machines .
stoc 2005 , pp .
331 - 337 .
gruia calinescu , chandra chekuri , martin pal and jan vondrak . maximizing a submodular set function subject to a matroid constraint .
siam journal on computing 40:6 ( 2011 ) , pp .
1740 - 1766 . v. chvatal . on certain polytopes associated with graphs .
j. combinatorial theory ser .
b , 18 ( 1975 ) , pp .
138 - 154 .
evripidis bampis , alexander kononov , dimitrios letsios , giorgio lucarelli and maxim sviridenko .
energy efficient scheduling and routing via randomized rounding .
fsttcs 2013 , pp .
449 - 460 .
nikhil bansal and kirk pruhs . the geometry of scheduling .
focs 2010 , pp . 407 - 414 . stephen boyd and lieven vandenberghe .
convex optimization .
cambridge university press , 2004 .
c. durr and o. vasquez .
order constraints for single machine scheduling with non - linear cost .
the 16th workshop on algorithm engineering and experiments ( alenex ) , 2014 .
m. grotschel , l. lovasz and a. schrijver .
geometric algorithms and combinatorial optimization .
springer - verlag , berlin , 1988 .
w. feller .
an introduction to probability theory and its applications ( 3rd edition ) .
john wiley & sons , new york ( 1968 ) .
d. r. fulkerson . blocking and anti - blocking pairs of polyhedra .
programming , 1 ( 1971 ) , pp .
168 - 194 .
michel x. goemans .
minimum bounded degree spanning trees .
focs 2006 , 273 - 282 .
w. hohn and t. jacobs .
an experimental and analytical study of order constraints for single machine scheduling with quadratic cost .
14th workshop on algorithm engineering and experiments ( alenex ) , pp .
103 - 117 , siam , 2012 .
w. hohn and t. jacobs . on the performance of smith s rule in single - machine scheduling with nonlinear cost .
latin 2012 , pp . 482 - 493 .
han hoogeveen , petra schuurman , gerhard j. woeginger .
non - approximability results for scheduling problems with minsum criteria .
informs journal on computing 13(2 ) , pp .
157 - 168 ( 2001 ) .
s. irani and k. pruhs .
algorithmic problems in power management .
acm sigact news 36 ( 2 ) , 63 - 76 .
kumar joag - dev and frank proschan .
negative association of random variables with applications .
the annals of statistics 11 ( 1983 ) , no . 1 , 286 - 295 . v. s. a. kumar , m. v. marathe , s. parthasarathy and a. srinivasan .
a unified approach to scheduling on unrelated parallel machines .
journal of the acm , vol .
56 , 2009 .
s. maia , e. goldbarg , and m. goldbarg .
on the biobjective adjacent only quadratic spanning tree problem .
electronic notes in discrete mathematics , 41 , ( 2013 ) , 535 - 542 .
yurii nesterov and arkadi nemirovski .
interior - point polynomial algorithms in convex programming .
society for industrial and applied mathematics , 1987 .
t. oncan and a. punnen .
the quadratic minimum spanning tree problem : a lower bounding procedure and an efficient search algorithm .
computers and operations research v.37 , ( 2010 ) , 1762 - 1773 .
victor de la pea .
bounds on the expectation of functions of martingales and sums of positive rvs in terms of norms of sums of independent random variables .
proceedings of the american mathematical society 108.1 ( 1990 ) : 233 - 239 .
victor de la pea and e. gin .
decoupling : from dependence to independence .
springer , 1999 . victor de la pea , rustam ibragimov , shaturgan sharakhmetov .
on extremal distributions and sharp @xmath398-bounds for sums of multilinear forms .
the annals of probability , 2003 , vol .
31 ( 2),pp .
630675 .
d. pereira , m. gendreau , a. cunha .
stronger lower bounds for the quadratic minimum spanning tree problem with adjacency costs .
electronic notes in discrete mathematics , 41 , ( 2013 ) , 229 - 236 .
a. schrijver .
theory of linear and integer programming .
wiley , 1998 .
a. schrijver .
combinatorial optimization .
springer , 2002 .
moshe shaked and j. george shanthikumar .
stochastic orders .
springer , 2007 .
qi - man shao . a comparison theorem on moment inequalities between negatively associated and independent random variables .
journal of theoretical probability 13 , no . 2 ( 2000 )
: 343 - 356 .
andreas s. schulz and martin skutella .
scheduling unrelated machines by randomized rounding .
siam j. discrete math , 15(4 ) , pp .
450 - 469 ( 2002 ) .
mohit singh and lap chi lau .
approximating minimum bounded degree spanning trees to within one of optimal .
stoc 2007 : 661 - 670 .
martin skutella .
convex quadratic and semidefinite programming relaxations in scheduling .
j. acm 48(2 ) , 206 - 242 ( 2001 ) .
martin skutella .
approximation and randomization in scheduling .
technische universitat berlin , germany , 1998 .
s. stiller and a. wiese .
increasing speed scheduling and flow scheduling .
isaac ( 2 ) 2010 , pp.279 - 290 .
a. wierman , l. l. h. andrew , and a. tang . power - aware speed scaling in processor sharing systems .
infocom 2009 , pp . 2007 - 2015 .
in this section , we prove a corollary of theorem [ thm : cx - main ] , which we will need in the next section .
[ cor : decoupling ] let @xmath63 be jointly distributed ( non - independent ) nonnegative integral random variables , and let @xmath64 be independent bernoulli random variables taking values @xmath289 and @xmath288 such that for each @xmath27 , @xmath399 = { \mathbb{e}}[y_i]$ ] .
then , @xmath400 particularly , for every @xmath43 , @xmath383 where @xmath53 is the fractional bell number . consider independent random variables @xmath378 such that each @xmath378 is distributed as @xmath7 . by lemma [ lem : b - cx - p ] , @xmath401 for all @xmath27 .
hence , by lemma [ lem : sumx - less - sumy ] and theorem [ thm : cx - main ] , @xmath402
we consider the following linear programming relaxation of the scheduling problem @xmath173 .
the variable @xmath403 if job @xmath75 starts at time @xmath45 on machine @xmath27 . for convenience
we assume that @xmath404 for negative values of @xmath45 .
@xmath167 } \sum_{j\in [ n]}\sum_{t\ge 0 } w_j(t+p_{ij})^p x_{ijt } \label{obj_c}\\\ \sum_{i\in [ m ] , t\ge 0}x_{ijt}=1 & \hspace{2 cm } \forall j\in [ n ] \label{job_c}\\ \sum_{j\in [ n]}\sum_{\tau = t - p_{ij}+1}^tx_{ij\tau}\le 1 & \hspace{2 cm } \forall i\in [ m ] , t\ge 0 \label{time_c}\\ x_{ijt}\ge 0 & \hspace{2 cm } \forall i\in [ m ] , j\in [ n ] , t\ge 0 . \label{fr_c}\end{aligned}\ ] ] the constraints ( [ job_c ] ) say that each job must be assigned , the constraint ( [ time_c ] ) says that at most one job can be processed in a unit time interval on each machine .
such linear programming relaxation are known under the name of _ strong time indexed formulations_. the standard issue with such relaxations is that they have pseudo - polynomially many variables due to potentially large number of indices @xmath45 .
one way to handle this issue is to partition the time interval into intervals @xmath405 $ ] and round all completion times to the endpoints of such intervals .
this method leads to polynomially sized linear programming relaxations with @xmath406-loss in the performance guarantee ( see @xcite for detailed description of the method ) . from now on we ignore this issue and assume that the planning horizon upper bound @xmath407 is polynomially bounded in the input size .
* algorithm . *
our approximation algorithm solves linear programming relaxation ( [ obj_c])-([fr_c ] ) .
let @xmath193 be the optimal fractional solution of the lp .
each job is tentatively assigned to machine @xmath27 to start at time @xmath45 with probability @xmath408 , independently at random .
let @xmath409 be the tentative start time assigned to job @xmath75 by our randomized procedure .
we process jobs assigned to each machine in the order of the tentative completion times @xmath410
. * analysis .
* we estimate the expected cost of the approximate solution returned by the algorithm .
we denote the expected cost by @xmath411 . for each machine - job - tentative time
triple @xmath412 , let @xmath413 be the set of triples @xmath414 such that @xmath415 .
let @xmath416 be the random boolean variable such that @xmath417 if job @xmath75 is assigned to machine @xmath27 with tentative start time @xmath45 .
in addition , let @xmath418 be the random boolean variable such that @xmath419 if job @xmath99 is assigned to machine @xmath27 with tentative start time @xmath420 for some @xmath421 by our randomized rounding procedure .
then , @xmath422 suppose that job @xmath75 is tentative scheduled on machine @xmath27 at time @xmath45 i.e. , @xmath417 .
we start processing job @xmath75 after all jobs @xmath99 tentative scheduled on machine @xmath27 at time @xmath420 with @xmath423 are finished .
thus the weighted expected completion time to the power of @xmath37 for @xmath75 equals ( given @xmath417 ) @xmath424 } & = & { \mathbb{e}}\big [ \big(\sum_{j'\in [ n]\setminus \{j\ } } p_{i , j'}z^{ijt}_{j'}+p_{ij } \big)^p { \;\;\mid\;\;}x^{ijt}=1\big]\\ & = & { \mathbb{e}}\big [ \big(\sum_{j'\in [ n]\setminus \{j\ } } p_{i , j'}z^{ijt}_{j'}+p_{ij } \big)^p\big].\end{aligned}\ ] ] in the second equality , we used that random variables @xmath418 are independent from the random variable @xmath416 .
then , @xmath425 } \sum_{i\in [ m ] } \sum_{t\ge 0 } w_j{{\mathbb{e}}\left[w_j c_j^p{\;\;\mid\;\;}x^{ijt } = 1\right ] } \pr\big ( x^{ijt}=1\big ) \nonumber \\ & = & \sum_{j\in [ n ] } \sum_{i\in [ m ] } \sum_{t\ge 0 } w_j{\mathbb{e}}\big[\big(\sum _ { j'\in [ n]\setminus \{j\ } } p_{i , j'}z^{ijt}_{ij'}+p_{ij } \big)^p \big ] \;x^*_{ijt}. \label{unrelatedfirst}\end{aligned}\ ] ] note , that for fixed @xmath426 , j\in [ n],t\ge 0 $ ] random variables @xmath418 are independent from each other .
we claim that @xmath427\setminus \{j\ } } p_{i , j'}z^{ijt}_{j'}+p_{ij } \big)^p \big ] \le a_p ( t+2p_{ij})^p.\ ] ] combining ( [ unrelatedfirst ] ) and ( [ mainunrelated ] ) , we derive that the performance guarantee of our approximation algorithm is at most @xmath181 .
we now prove inequality ( [ mainunrelated ] ) .
let @xmath428 be the interval graph where the vertex set @xmath429 is the collection of intervals corresponding to triples in @xmath421 such that @xmath430 .
more precisely , every triple @xmath421 corresponds to the interval @xmath431 with corresponding weight @xmath430 .
let @xmath432 be the collection of all independent sets in @xmath428 .
the interval graph @xmath428 is perfect , and the weights @xmath433 satisfy the constraints ( [ time_c ] ) , so there is a collection of weights @xmath434 , @xmath435 ( for more formal argument see below ) such that @xmath436 formally , the claim above follows from the polyhedral characterization of perfect graphs proved by fulkerson @xcite and chvatal @xcite ( see also schrijver s book @xcite , section 9 , application 9.2 on p. 118 ) that a graph @xmath192 is perfect if and only if its stable set polytope is defined by the system below : @xmath437 in the interval graph @xmath428 all clique inequalities are included in the constraints ( [ time_c ] ) and therefore any set of weights @xmath433 can be decomposed into a convex combination of independent sets in @xmath428 .
we define a random variable @xmath438 as follows : sample an independent set @xmath435 with probability @xmath439 and let @xmath440 note that one job @xmath99 may have more than one interval @xmath441 in the set @xmath442 ( for different @xmath420 ) .
random variables @xmath438 may be dependent but @xmath443=\sum_{(i , j',t')\in j_{ijt } } x^*_{ij't ' } = { \mathbb{e}}[z_{j'}^{ijt}].\ ] ] therefore , by corollary [ cor : decoupling ] we have @xmath444\setminus \{j\ } } p_{i , j'}z^{ijt}_{j'}+p_{ij } \big)^p \big ] \le a_p{\mathbb{e}}\big[\big(\sum _ { j'\in [ n]\setminus \{j\ } } p_{i , j'}y^{ijt}_{j'}+p_{ij } \big)^p \big].\ ] ] now , observe , that @xmath445\setminus \{j\ } } p_{i , j'}y^{ijt}_{j'}$ ] is always bounded by @xmath446 , because all intervals in @xmath442 are disjoint ( @xmath442 is an independent set ) and all intervals are subsets of @xmath447 $ ] .
hence , @xmath444\setminus \{j\ } } p_{i , j'}z^{ijt}_{j'}+p_{ij } \big)^p \big ] \le a_p { \mathbb{e}}[((t + p_{ij } ) + p_{ij})^p ] \leq a_p ( t+2p_{ij})^p,\ ] ] which concludes the proof .
we show that functions @xmath237 defined in section [ sec : thm : efficient ] are convex . [
lem : f - convex ] fix real numbers @xmath13 and @xmath448 .
define a function @xmath449^n \to { \mathbb{r}}^+$ ] as follows : @xmath450 equals the optimal value of the following lp : @xmath451 } \big ( \sum_{i\in s } d_i\big)^{q}&z_{s}&\\ \sum_{s\subseteq [ n]}z_{s}&=1&\\ \sum_{s : i\in s}z_{s}&=y_i,&\forall i\in [ n]\\ z_{s}&\ge 0 , & \forall s\subseteq [ n]&\end{aligned}\ ] ] then , @xmath452 is a convex function .
consider two vectors @xmath453^n$ ] .
pick an arbitrary @xmath454 $ ] .
we need to show that @xmath455 consider the optimal lp solutions @xmath238 and @xmath231 for @xmath232 and @xmath233 .
then , by the definition of @xmath452 , @xmath456 } \big ( \sum_{i\in s } d_i\big)^{q } z^*_{s}\;\text { and } \;h(y^ { * * } ) = \sum_{s\subseteq [ n ] } \big ( \sum_{i\in s } d_i\big)^{q } z^{**}_{s}.\ ] ] observe , that @xmath234 is a feasible solution for @xmath235 ( since all lp constraints are linear ) .
hence , @xmath457 is at most the lp cost of @xmath234 , which equals @xmath458 } \big ( \sum_{i\in s } d_i\big)^{q } ( \lambda z^*_{s } + ( 1-\lambda ) z^{**}_{s } ) = \lambda h(y^ * ) + ( 1-\lambda ) h(y^{**}).\ ] ]
in this section , we show how to discretize values @xmath103 .
we assume that functions @xmath77 satisfy the following conditions : 1 .
all @xmath77 are convex increasing nonnegative functions computable in polynomial time
. 2 . for every @xmath75 , @xmath107 .
3 . for every @xmath75 and @xmath459 $ ] , @xmath460 .
we first find an approximate value of the optimal solution @xmath461 and then apply theorem [ thm : discr ] ( see below ) . note that if the gap @xmath462 is polynomially bounded , then we can pick @xmath210 such that the optimal value @xmath463 of the discretized problem is at most @xmath464 .
we pick such @xmath465 either by using the binary search or by enumerating all powers of 2 in the range @xmath466 $ ] .
[ thm : discr ] there exists a polynomial - time algorithm that given an instance of the integer program ( [ ip : obj])-([ip : int ] ) satisfying conditions ( 1 ) and ( 2 ) above , an upper bound @xmath465 on the cost of the optimal solution @xmath153 , and @xmath115 , returns a new set of coefficients @xmath467 , numbers @xmath468 , and an extra set of constraints @xmath469 for @xmath470 such that each @xmath467 is a multiple of @xmath104 ; @xmath471 is an integer polynomially bounded in @xmath46 , @xmath106 and @xmath18 ( see item 3 above ) such that the following two properties are satisfied . 1
. the cost of the optimal solution for the new problem is at most the cost of the original problem : @xmath472 2 . for every feasible solution of the new problem
@xmath473 satisfying @xmath469 for @xmath470 , we have @xmath384 } f_j \big(\sum_{i\in [ n ] } d_{ij } y_{i}\big)\leq ( 1+\varepsilon ) \sum_{j\in [ k ] } f_j \big(\sum_{i\in [ n ] } d'_{ij } y_{i}\big ) + \varepsilon \widetilde{opt}.\ ] ] particularly , @xmath474 the proof is fairly standard : we round all @xmath103 to be multiples of @xmath104 .
then we show that if @xmath104 s are sufficiently small , then the introduced rounding error is at most @xmath475 .
the details are below .
we algorithm finds the set @xmath476 .
if @xmath470 , then @xmath477 must be equal to @xmath289 in every optimal solution , because otherwise , @xmath478 .
thus , for all @xmath470 , we set @xmath477 and all @xmath467 s to be @xmath289 .
then , we let @xmath479 we round down all @xmath103 s to be multiples of @xmath104 . denote the rounded values by @xmath467 .
this is our new instance .
it is clear that @xmath467 are multiples of @xmath104 . since @xmath480 for all @xmath27 and @xmath75
, we have @xmath481 and therefore , all @xmath471 are polynomially bounded .
since @xmath482 and @xmath77 are monotone functions , we have for every @xmath226 , @xmath384 } f_j \big(\sum_{i\in [ n ] } d'_{ij } y_{i}\big ) \leq \sum_{j\in [ k ] } f_j \big(\sum_{i\in [ n ] } d_{ij } y_{i}\big).\ ] ] as we observed earlier if @xmath232 is the optimal solution to the original problem , then @xmath483 for @xmath470 , hence @xmath232 is a feasible solution to the new problem .
consequently , @xmath472 we now need to verify that @xmath384 } f_j \big(\sum_{i\in [ n ] } d_{ij } y_{i}\big ) - \sum_{j\in [ k ] } f_j \big(\sum_{i\in [ n ] } d'_{ij } y_{i}\big ) \leq \varepsilon \widetilde{opt } + \varepsilon\sum_{j\in [ k ] } f_j \big(\sum_{i\in [ n ] } d'_{ij } y_{i}\big).\ ] ] we prove that for every @xmath75 , @xmath484 } d_{ij } y_{i}\big ) - f_j \big(\sum_{i\in [ n ] } d'_{ij } y_{i}\big ) \leq \frac{\varepsilon}{k}\ , \widetilde{opt } + \varepsilon f_j \big(\sum_{i\in [ n ] } d'_{ij } y_{i}\big).\ ] ] consider two cases .
i. if @xmath485 } d_{ij } y_{i } < \varepsilon t_j / k$ ] , then @xmath486 } d_{ij } y_{i}\big)\leq \varepsilon f_j(t_j)/k \leq \varepsilon \widetilde{opt}/k,\ ] ] since @xmath116 is a convex function , @xmath487 , and @xmath488 .
hence , inequality ( [ eq : fj - diff ] ) holds .
now assume that @xmath485 } d_{ij } y_{i}\geq \varepsilon t_j / k$ ] .
observe , that @xmath489 } d_{ij } y_{i } - \sum_{i\in [ n ] } d'_{ij } y_{i}\leq n\delta_j = \varepsilon \eta t_j / k \leq \eta \sum_{i\in [ n ] } d_{ij } y_{i}.\ ] ] hence ( using that @xmath490 $ ] and thus @xmath491 ) , @xmath489 } d_{ij } y_{i}\leq \frac{1}{1-\eta } \sum_{i\in [ n ] } d'_{ij } y_{i } \leq ( 1 + 2\eta)\sum_{i\in [ n ] } d'_{ij } y_{i}.\ ] ] we now use claim [ cl : f - dir ] ( see below ) with @xmath492 } d'_{ij } y_{i};\;\;\;\text{and}\;\;\;\eta ' = \frac{\sum_{i\in [ n ] } d_{ij } y_{i}}{\sum_{i\in [ n ] } d'_{ij } y_{i } } - 1\leq 2\eta\leq \frac{\varepsilon}{2p}.\ ] ] we get @xmath486 } d_{ij } y_{i}\big ) - f_j \big(\sum_{i\in [ n ] } d'_{ij } y_{i}\big ) \leq \varepsilon f_j \big(\sum_{i\in
[ n ] } d'_{ij } y_{i}\big).\ ] ] this finishes the proof .
it only remains to prove claim [ cl : f - dir ] .
[ cl : f - dir ] suppose that @xmath116 is a nonnegative monotonically increasing function such that for every @xmath493 $ ] , @xmath494 .
let @xmath495 $ ] and @xmath496 .
then , for @xmath497 , @xmath498 write , @xmath499 thus , @xmath500 and ( since @xmath501 for @xmath502 $ ] ) @xmath503
[ lem : limsup ] suppose that a sequence of integer random variables @xmath504 converges in distribution to an integer random variable @xmath275 . then
, for every nonnegative function @xmath14 with a finite expectation @xmath505 $ ] , we have @xmath506\geq { \mathbb{e}}[\varphi(z)].\ ] ] let @xmath507 for @xmath508 , and @xmath509 , otherwise .
the function @xmath14 is nonzero on finitely many integral points .
hence , @xmath510 = { \mathbb{e}}[\varphi_t(z)]$ ] for every fixed @xmath45 .
observe that @xmath511\geq { \mathbb{e}}[\varphi_t(s^n)]$ ] , since @xmath512 for all @xmath2 . on the other hand , @xmath513 = { \mathbb{e}}[\varphi(z)]$ ] .
hence , @xmath514\geq \sup_{n } \sup_{t } { \mathbb{e}}[\varphi_t(s^n ) ] = \sup_{t } \sup_{n } { \mathbb{e}}[\varphi_t(s^n)]\geq\\ \sup_{t } \lim_{n\to\infty } { \mathbb{e}}[\varphi_t(s^n ) ] = \sup_t { \mathbb{e}}[\varphi_t(z)]={\mathbb{e}}[\varphi(z)].\end{gathered}\ ] ]
for completeness , we give proofs of lemma [ lem : sumx - less - sumy ] and lemma [ lem : sum - ax - less - ay ] in this section .
the reader may also find slightly different proofs of these lemmas in the book of shaked and shanthikumar @xcite .
[ lem : xyz ] consider three independent random variables @xmath265 , @xmath266 , and @xmath275 . if @xmath267 , then @xmath515 . for every convex function @xmath268 , we have @xmath516 } = { \mathbb{e}}{\mathbb{e}}[\varphi ( x+z){\,\mid\,}z ] \leq
{ \mathbb{e}}{\mathbb{e}}[\varphi ( y + z){\,\mid\,}z ] = { { \mathbb{e}}\left[\varphi ( y + z)\right]}.\ ] ] the inequality above holds , since for any fixed @xmath517 , the function @xmath518 is convex . using lemma [ lem : xyz ] , we replace @xmath6 s with @xmath7 s in the sum @xmath519 one by one : for every @xmath74 , let @xmath520 then , by lemma [ lem : xyz ] , @xmath521 we have @xmath522 this finishes the proof .
consider an arbitrary convex function @xmath14 .
let @xmath523 and @xmath524 . since @xmath525 is a convex function , we have @xmath526 and @xmath527}\leq { { \mathbb{e}}\left[\tilde{\varphi}(y)\right]}$ ] .
hence , @xmath528 } \leq { { \mathbb{e}}\left[\frac{1}{a}\sum_{i=1}^n a_i \tilde{\varphi}(x_i)\right ] } = \frac{1}{a}\sum_{i=1}^n a_i { { \mathbb{e}}\left[\tilde{\varphi}(x_i)\right ] } { \leq_{cx}}\frac{1}{a}\sum_{i=1}^n a_i { { \mathbb{e}}\left[\tilde{\varphi}(y)\right ] } = { { \mathbb{e}}\left[\tilde{\varphi}(y)\right]}.\ ] ] substituting @xmath529 , we get @xmath530 } \leq { { \mathbb{e}}\left[\tilde{\varphi}\big(\sum_{i=1}^n a_i y\big)\right]}.\ ] ]
in section [ sec : thm : decoupling ] , we proved theorem [ thm : cx - main ] for discrete random variables . in this section
, we extend this result to arbitrary random variables .
consider two sequences of nonnegative random variables @xmath64 and @xmath63 satisfying conditions of theorem [ thm : cx - main ] .
fix an arbitrary convex function @xmath316 .
we need to show that @xmath531 \leq { \mathbb{e}}[\varphi(p\cdot ( y_1+\dots + y_n))]\ ] ] assuming that both expectations exist .
observe that @xmath14 can be represented as the sum of two convex functions : a monotonically non - decreasing convex function @xmath532 and monotonically non - increasing convex function @xmath533 .
it suffices to show that @xmath534 & \leq { \mathbb{e}}[{\varphi^{\text{\tiny{$\mathbf{\uparrow}$}}}}(p\cdot ( y_1+\dots + y_n))];\label{eq : crv - phiup}\\ { \mathbb{e}}[{\varphi^{\text{\tiny{$\mathbf{\downarrow}$}}}}(x_1+\dots+x_n ) ] & \leq { \mathbb{e}}[{\varphi^{\text{\tiny{$\mathbf{\downarrow}$}}}}(p\cdot ( y_1+\dots + y_n))].\label{eq : crv - phidown}\end{aligned}\ ] ] note that if the expectations in ( [ eq : crv : phi ] ) exist than the expectations in ( [ eq : crv - phiup ] ) and ( [ eq : crv - phidown ] ) also exist .
we now prove inequality ( [ eq : crv - phiup ] ) .
the proof of ( [ eq : crv - phidown ] ) is almost the same . for natural @xmath535 ,
define a function @xmath536 as follows : @xmath537 the function @xmath536 truncates @xmath2 at the level @xmath535 and then rounds @xmath2 down to the nearest multiple of @xmath538 .
observe , that @xmath539 for every @xmath2 .
hence , @xmath540 converges a.s . to @xmath541 as @xmath542 ; and @xmath543 converges a.s . to @xmath544 as @xmath545 . since , @xmath532 is a continuous function @xmath546 and @xmath547 .
notice that @xmath548 for all @xmath2 .
hence , @xmath549 and @xmath550 . by lebesgue s dominated convergence theorem
, we get @xmath551&={\mathbb{e}}[{\varphi^{\text{\tiny{$\mathbf{\uparrow}$}}}}(x_1+\dots+x_n)];\label{eq : coverg1}\\ \lim_{m\to \infty } { \mathbb{e}}[{\varphi^{\text{\tiny{$\mathbf{\uparrow}$}}}}(p(g_{m}(y_1)+\dots + g_{m}(y_n ) ) ) ] & = { \mathbb{e}}[{\varphi^{\text{\tiny{$\mathbf{\uparrow}$}}}}(p(y_1+\dots+y_n))].\label{eq : coverg2}\end{aligned}\ ] ] for every fixed @xmath535 , the left hand side of ( [ eq : coverg1 ] ) is upper bounded by the left hand side of ( [ eq : coverg2 ] ) , because random variables @xmath552 and @xmath553 are discrete and satisfy the conditions of theorem [ thm : cx - main ] .
hence , the right hand side of ( [ eq : coverg1 ] ) is upper bounded by the right hand side of ( [ eq : coverg2 ] ) .
this proves inequality ( [ eq : crv - phiup ] ) and concludes the proof of theorem [ thm : cx - main ] for arbitrary random variables . | we present a new framework for solving optimization problems with a diseconomy of scale . in such problems ,
our goal is to minimize the cost of resources used to perform a certain task .
the cost of resources grows superlinearly , as @xmath0 , @xmath1 , with the amount @xmath2 of resources used .
we define a novel linear programming relaxation for such problems , and then show that the integrality gap of the relaxation is @xmath3 , where @xmath3 is the @xmath4-th moment of the poisson random variable with parameter 1 . using our framework , we obtain approximation algorithms for the minimum energy efficient routing , minimum degree balanced spanning tree , load balancing on unrelated parallel machines , and unrelated parallel machine scheduling with nonlinear functions of completion times problems .
our analysis relies on the decoupling inequality for nonnegative random variables .
the inequality states that @xmath5 where @xmath6 are independent nonnegative random variables , @xmath7 are possibly dependent nonnegative random variable , and each @xmath7 has the same distribution as @xmath6 .
the inequality was proved by de la pea in 1990 .
de la pea , ibragimov , and sharakhmetov showed that @xmath8 for @xmath9 and @xmath10 for @xmath11 .
we show that the optimal constant is @xmath12 for any @xmath13 .
we then prove a more general inequality : for every convex function @xmath14 , @xmath15 } \leq { { \mathbb{e}}\left[\varphi\big(p\sum_{i=1}^n y_i\big)\right]},\ ] ] and , for every _ concave _ function @xmath16 , @xmath17 } \geq { { \mathbb{e}}\left[\psi\big(p\sum_{i=1}^n y_i\big)\right]},\ ] ] where @xmath18 is a poisson random variable with parameter 1 independent of the random variables @xmath7 . |
* .. * the geometry on cayley s surface and the geometry in the ambient space of cayley s surface has been investigated by many authors from various points of view .
see , among others , @xcite , @xcite , @xcite , @xcite , and @xcite . in these papers
the reader will also find a lot of further references . as a by - product of a recent publication @xcite , it turned out that the cayley surface ( in the real projective @xmath0-space ) carries a one - parameter family of twisted cubics which have mutually contact of order four .
these curves belong to a well - known three - parameter family of twisted cubics @xmath1 on cayley s surface ; cf .
formula ( [ eq : abc_proj ] ) below .
all of them share a common point @xmath2 with a common tangent @xmath3 , and a common osculating plane @xmath4 , say .
however , according to @xcite such a one - parameter family of twisted cubics with contact of order four should not exist :
_ zwei kubiken dieser art , die einander in @xmath2 mindestens fnfpunktig berhren , sind identisch .
_ the aim of the present communication is to give a complete description of the order of contact ( at @xmath2 ) for the twisted cubics mentioned above .
in particular , it will be shown in theorem [ thm:1 ] that the twisted cubics with parameter @xmath5 play a distinguished role , a result that seems to be missing in the literature . furthermore , since the order of contact is not a self - dual notion , we also investigate the order of dual contact for twisted cubics @xmath1 .
somewhat surprisingly , in the dual setting the parameters @xmath6 and @xmath7 are exceptional ; see theorem [ thm:2 ] . in section [ para:2.4 ]
we show that certain results of theorem [ thm:1 ] have a natural interpretation in terms of the _ twofold isotropic geometry _ which is based on the absolute flag @xmath8 , and in terms of the _ isotropic geometry _ in the plane @xmath4 which is given by the flag @xmath9 .
section [ para:3.2 ] is devoted to the interplay between theorem [ thm:1 ] and theorem [ thm:2 ] . *
.. * the calculations which are presented in this paper are long but straightforward . hence a computer algebra system ( maple v )
was used in order to accomplish this otherwise tedious job .
nevertheless , we tried to write down all major steps of the calculations in such a form that the reader may verify them without using a computer .
* .. * throughout this paper we consider the three - dimensional real projective space @xmath10 .
hence a point is of the form @xmath11 with @xmath12 being a non - zero vector in @xmath13 .
we choose the plane @xmath4 with equation @xmath14 as _ plane at infinity _ , and we regard @xmath10 as a projectively closed affine space . for the basic concepts of projective differential geometry we refer to @xcite and @xcite . * .. * the following is taken from @xcite , although our notation will be slightly different .
_ cayley s _ ( _ ruled cubic _ ) _ surface _ is , to within collineations of @xmath10 , the surface @xmath15 with equation @xmath16 the line @xmath17 is on @xmath15 .
more precisely , it is a torsal generator of second order and a directrix for all other generators of @xmath15 .
the point @xmath18 is the cuspidal point on @xmath3 . in figure [ abb1 ] a part of the surface @xmath15
is displayed in an affine neighbourhood of the point @xmath2 .
in contrast to our general setting , @xmath19 plays the role of the plane at infinity in this illustration . on the surface @xmath15 there is a three - parameter family of cubic parabolas which can be described as follows : each triple @xmath20 with @xmath21 gives rise to a function @xmath22 if moreover @xmath23 then @xmath24 yields the mapping @xmath25 its image is a _
cubic parabola _ @xmath26 .
all these cubic parabolas have the common point @xmath2 , the common tangent @xmath3 and the common osculating plane @xmath4 .
we add in passing that for @xmath27 we have @xmath28 for all @xmath29 , whereas the points of the form @xmath30 comprise the affine part of a _ parabola _ , @xmath31 say , lying on @xmath15 .
each curve @xmath1 ( @xmath21 ) is on the _ parabolic cylinder _ with equation @xmath32 the mapping @xmath33 is injective , since different triples @xmath34 yield different parabolic cylinders ( [ eq : zylinder ] ) . figure [ abb2 ] shows some generators of @xmath15 , and five cubic parabolas @xmath35 together with their corresponding parbolic cylinders , where @xmath36 ranges in @xmath37 and @xmath5 .
1.0 cm ( 4.0,5.18 ) ( 0.0 , 0.0 ) ( 1.75,4.5)@xmath3 ( 2.1,2.1)@xmath2 ( 3.75,4.5)@xmath15 [ abb1 ] figure [ abb1 ] . ( 4.0,5.18 ) ( 0.0 , 0.0 ) [ abb2 ] figure [ abb2 ] . * .. * our first goal is to describe the order of contact at @xmath2 of cubic parabolas given by ( [ eq : abc_proj ] ) . since twisted cubics with contact of order five are identical @xcite , we may assume without loss of generality that the curves are distinct , and that the order of contact is less or equal four .
[ thm:1 ] distinct cubic parabolas @xmath1 and @xmath38 on cayley s ruled surface have 1 .
second order contact at @xmath2 if , and only if , @xmath39 or @xmath40 ; 2 . third order contact at @xmath2
if , and only if , @xmath39 and @xmath41 , or @xmath42 ; 3 .
fourth order contact at @xmath2 if , and only if , @xmath42 and @xmath41 .
we proceed in two steps : \(i ) first , we consider the quadratic forms @xmath43 which determine a hyperbolic paraboloid and a quadratic cone , respectively .
their intersection is the cubic parabola @xmath44 , given by @xmath45 and the line @xmath46 .
the tangent planes of the two surfaces at @xmath2 are different .
next , let @xmath47 be a lower triangular matrix , i.e. , @xmath48 for all @xmath49 .
the collineation which is induced by such a matrix @xmath50 fixes the point @xmath2 , the line @xmath3 , and the plane @xmath4 ; it takes @xmath44 to a cubic parabola , say @xmath51 . in order to determine the order of contact of @xmath44 and @xmath51 we follow @xcite . as @xmath52 , so we expand for @xmath53 the functions @xmath54 in terms of powers of @xmath55 and obtain @xmath56 the remaining coefficients @xmath57 will not be needed . note that the matrix entry @xmath58 does not appear in ( [ eq : koeffizienten ] ) .
\(ii ) we consider the collineation of @xmath10 which is induced by the regular matrix @xmath59 where @xmath20 and @xmath60 .
obviously , it fixes the point @xmath2 and takes @xmath44 to @xmath1 , since @xmath61 the ( irrelevant ) scalar factor in the definition of @xmath62 enables us to avoid fractions in the matrix @xmath63 the order of contact at @xmath2 of the cubic parabolas @xmath1 and @xmath38 coincides with the order of contact at @xmath2 of @xmath44 and that cubic parabola which arises from @xmath44 under the action of the matrix @xmath64 this matrix takes over the role of the matrix @xmath50 from the first part of the proof .
( its entry in the south - west corner has a rather complicated form and will not be needed ) .
therefore @xmath1 and @xmath38 have contact of order @xmath65 at @xmath2 if , and only if , in ( [ eq : expand ] ) the coefficients @xmath66 , @xmath67 , @xmath68
vanish for @xmath53 . by ( [ eq : koeffizienten ] ) , this leads for @xmath69 to the single condition @xmath70 which proves the assertion in ( a ) . by virtue of ( a ) , for @xmath71 there are two cases . if @xmath39 then @xmath72 vanishes and we obtain the condition @xmath73 whereas @xmath40 yields @xmath74 altogether this proves ( b ) .
finally , for @xmath75 there again are two possibilities : if @xmath39 and @xmath41 then @xmath76 vanishes , whence we get @xmath77 note that here @xmath78 , since @xmath79 . on the other hand , if @xmath42 then the conditions read @xmath80 this completes the proof .
alternatively , the preceding results could be derived from ( * ? ? ?
* theorem 1 ) which describes contact of higher order between curves in @xmath81-dimensional real projective space . * .. * in the following pictures we adopt once more the same alternative point of view like in figure [ abb1 ] , i.e. , the plane with equation @xmath19 is at infinity . in figure [ abb3 ] two curves @xmath1 and @xmath38
are displayed .
as @xmath82 and @xmath83 , they have contact of second order at @xmath2 .
a family of curves @xmath35 with @xmath84 and @xmath5 is shown in figure [ abb4 ] .
all of them have mutually contact of order four at @xmath2 .
these curves are , with respect to the chosen affine chart ( @xmath85 ) , cubic hyperbolas for @xmath86 , a cubic parabola for @xmath87 , and cubic ellipses for @xmath88 ; the corresponding values of @xmath36 are written next to the images of the curves .
see also figure [ abb2 ] for another picture of this family , although with different values for @xmath36 and @xmath14 as plane at infinity .
1.0 cm ( 4.0,5.59 ) ( 0.0,0.25 ) ( 1.5,4.9)@xmath3 ( 1.9,2.5)@xmath2 ( 3.0,5.0)@xmath15 ( 0.8,3.55)@xmath1 ( 3.05,2.45)@xmath38 [ abb3 ] figure [ abb3 ] .
( 4.0,5.59 ) ( 0.0 , 0.0 ) ( 2.15,5.35)@xmath3 ( 4.0,4.6)@xmath15 ( 0.2,3.05)@xmath89 ( 2.8,1.35)@xmath89 ( -0.2,4.2)@xmath90 ( 3.7,1.0)@xmath90 ( 3.5,-.15)@xmath91 ( 0.2,5.44)@xmath91 ( 0.8,5.40)@xmath92 ( 2.9,-.11)@xmath92 ( 4.05,3.9)@xmath93 ( 3.75,2.6)@xmath93 ( -0.05,1.68)@xmath93 ( 2.25,0.5)@xmath94 ( 2.4,0.7)(-3,4)0.3 ( 1.5,0.7)(3,4)0.4 ( 1.35,0.55)@xmath95 ( 3.5 , 5.2)@xmath94 ( 3.42,2.15)@xmath94 ( 2.15,1.1)@xmath93 ( 2.9 , 5.24)@xmath95 ( 3.2,1.8)@xmath95 ( 0.11,0.22)@xmath94 ( 0.71 , 0.18)@xmath95 ( 0.1,3.6)@xmath93 ( 0.365,4.28)(1,-1)0.4 ( 0.065,4.34)@xmath94 ( 0.5,4.55)(3,-4)0.4 ( 0.2,4.61)@xmath95 [ abb4 ] figure [ abb4 ] . * .. * [ para:2.4 ] it follows from theorem [ thm:1 ] that cubic parabolas @xmath1 with @xmath5 play a special role .
in order to explain this from a geometric point of view we consider the _ tangent surface _ of a cubic parabola @xmath1 and , in particular , its intersection with the plane at infinity . it is well known that this is a conic @xmath96 together with the line @xmath3 .
in fact , via the first derivative of the local parametrization @xmath97 of @xmath1 we see that @xmath98 is given by @xmath99 the plane at infinity carries in a natural way the structure of an _ isotropic _ ( or _ galileian _ ) _ plane _ with the absolute flag @xmath9 .
each point @xmath100 can be identified with the point @xmath101 . in this way
the standard basis of @xmath102 determines a unit length and a unit angle in the isotropic plane @xcite . from this point of view
each @xmath96 is an _
isotropic circle_. by ( [ eq : p_lokal ] ) , its _ isotropic curvature _
@xcite equals @xmath103 ; this bound is attained for @xmath5 .
it is well known that two isotropic circles @xmath96 and @xmath104 have second order contact at the point @xmath2 if , and only if , their isotropic curvatures are the same @xcite , i.e. for @xmath39 or for @xmath40 . from this observation
one could also derive the assertion in theorem [ thm:1 ] ( a ) as follows : we introduce an auxiliary euclidean metric in a neighbourhood of @xmath2 , and we take into account that the ratio of the euclidean curvatures at @xmath2 of the curves @xmath1 and @xmath96 ( the curves @xmath38 and @xmath104 ) equals @xmath105 ; see @xcite for this theorem of e. beltrami .
the flag @xmath8 turns @xmath10 into a _
twofold isotropic _
( or _ flag _ ) _ space_. the definition of metric notions in this space is based upon the identification of @xmath106 with @xmath107 , and the canonical basis of @xmath108 ; see @xcite . by @xcite
, each cubic parabola @xmath1 has the _ twofold isotropic conical curvature _ @xmath109 .
hence the following characterization follows . among all cubic parabolas @xmath1 on the cayley surface @xmath15 ,
the cubic parabolas with @xmath5 are precisely those with maximal twofold isotropic conical curvature .
yet another interpretation is as follows : the regular matrix @xmath110 yields a _
homothetic transformation _ of @xmath10 which maps the cubic parabola @xmath111 to @xmath112 , since @xmath113 as all points at infinity are invariant , the corresponding isotropic circles @xmath114 and @xmath115 coincide .
this homothetic transformation is identical if , and only if , @xmath5 .
the cayley surface @xmath15 admits a @xmath0-parameter collineation group ; see @xcite formula ( 9 ) .
the action of this group on the family of all cubic parabolas @xmath1 is described in @xcite , formula ( 12 ) .
( in the last part of that formula some signs have been misprinted .
the text there should read @xmath116 ) .
by virtue of this action , our previous result on homothetic transformations can be generalized to other cubic parabolas on @xmath15 .
* .. * the question remains how to distinguish between cubic parabolas @xmath1 and @xmath38 satisfying the first condition ( @xmath39 ) in theorem [ thm:1 ] ( a ) , and those which meet the second condition ( @xmath40 ) .
a similar question arises for the two conditions in theorem [ thm:1 ] ( b ) .
we shall see that such a distinction is possible if we consider the _ dual curves _ which are formed by the osculating planes ( i.e. cubic developables ) . recall that @xmath1 and @xmath38 have , by definition , _ dual contact of order @xmath65 _ at a common osculating plane @xmath117 , if their dual curves have contact of order @xmath65 at the `` point '' @xmath117 of the dual projective space .
we shall identify the dual of @xmath13 with the vector space @xmath118 in the usual way ; so planes ( i.e. points of the dual projective space ) are given by non - zero _ row vectors_. thus , for example , a plane @xmath119 is tangent to the cayley surface ( [ eq : cayley ] ) if , and only if , @xmath120 we note that all these tangent planes comprise a cayley surface in the dual space .
for each twisted cubic there exists a unique null polarity ( symplectic polarity ) which takes each point of the twisted cubic to its osculating plane . in particular , the null polarity of the cubic parabola @xmath44 is induced by the linear bijection @xmath121 1 . second order dual contact at @xmath4 if , and only if , @xmath39 ; 2 .
third order dual contact at @xmath4 if , and only if , @xmath39 and @xmath41 , or @xmath122 ; 3 .
fourth order dual contact at @xmath4 if , and only if , @xmath123 and @xmath41 .
the matrix @xmath124 determines a duality of @xmath10 which maps the set of points of @xmath44 onto the set of osculating planes of @xmath1 . since the product of a duality and the inverse of a duality is a collineation , we obtain the following : the order of dual contact at @xmath4 of the given curves @xmath1 and @xmath38 coincides with the order of contact at @xmath2 of the cubic parabola @xmath44 and that cubic parabola which arises from @xmath44 under the collineation given by the matrix @xmath125 here @xmath126 denotes an entry that will not be needed .
we now proceed as in the proof of theorem [ thm:1 ] . by substituting the entries of the matrix above into ( [ eq : koeffizienten ] ) , we read off necessary and sufficient conditions for dual contact of order @xmath65 at the plane @xmath4 of @xmath1 and @xmath38 . for @xmath69
we get the single condition @xmath127 which proves the assertion in ( a ) . by ( a )
, we let @xmath39 for the discussion of @xmath71 . then @xmath72 vanishes and we arrive at the condition @xmath128 from which ( b ) is immediate .
finally , for @xmath75 we distinguish two cases : if @xmath39 and @xmath41 then @xmath76 vanishes and we are lead to the condition @xmath129 note that here @xmath78 , since @xmath79 .
the proof of ( c ) will be finished by showing that the case @xmath122 does not occur . from the assumption
@xmath122 follows the first condition @xmath130 now , letting @xmath41 , the second condition @xmath131 is obtained .
however , both conditions can not be satisfied simultaneously , since the first condition and @xmath79 together imply that @xmath78 . * .. * by combining the results of theorem [ thm:1 ] and theorem [ thm:2 ] , it is an immediate task to decide whether or not two ( not necessarily distinct ) cubic parabolas @xmath1 and @xmath38 have contact at @xmath2 and at the same time dual contact at @xmath4 of prescribed orders .
in particular , we infer that two cubic parabolas of this kind , with fourth order contact at @xmath2 and fourth order dual contact at @xmath4 , are identical .
let us choose a _ fixed _ real number @xmath60 .
we consider the local parametrization @xmath132 of @xmath15 ; its image is @xmath133 , i.e. the affine part of @xmath15 . for our fixed @xmath134 and @xmath135 the affine parts of the parabolic cylinders ( [ eq : zylinder ] ) form a partition of @xmath136 ; see figure [ abb2 ] .
hence @xmath137 is injective so that through each point @xmath138 there passes a unique curve @xmath35 .
consequently , we can define a mapping @xmath139 of @xmath133 into the dual projective space by @xmath140 the image of the affine part of the cayley surface @xmath15 under the mapping @xmath139 described in _ ( [ eq : sigma ] ) _ consists of tangent planes of a cayley surface for @xmath141 , and of tangent planes of a hyperbolic paraboloid for @xmath142 .
as the null polarity of @xmath35 arises from the matrix @xmath143 so the @xmath139-image of a point @xmath144 is the plane which is described by the non - zero row vector @xmath145 in discussing @xmath146 there are two cases : \(i ) suppose that @xmath147 . then a duality of @xmath10 is determined by the regular matrix @xmath148 letting @xmath149 the transpose of @xmath150 is easily seen to equal the row vector in ( [ eq : polar ] ) . hence @xmath146 is part of a cayley surface in the dual space which in turn , by ( [ eq : dual_cayley ] ) , is the set of tangent planes of a cayley surface in @xmath10 .
\(ii ) if @xmath151 then the row vector ( [ eq : polar ] ) simplifies to @xmath152 thus the set @xmath146 is part of the non - degenerate ruled quadric in the dual space with equation @xmath153 ( in terms of dual coordinates ) . in other words
, @xmath154 consists of tangent planes of a hyperbolic paraboloid in @xmath10 .
let us add the following remark .
the linear fractional transformation @xmath155 is an involution such that our fixed @xmath141 goes over to @xmath156 , as defined in ( [ eq : betastrich ] ) , whereas @xmath157 . in particular , if @xmath7 then @xmath158 .
this explains the relation between theorem [ thm:1 ] ( c ) and theorem [ thm:2 ] ( c ) .
also the fixed values of @xmath159 are noteworthy : for @xmath160 the curves @xmath161 are _ asymptotic curves _ of @xmath15 , i.e. , the osculating plane of @xmath161 at each point @xmath162 is the tangent plane of @xmath15 at @xmath163 .
this means that the planes of the set @xmath146 are tangent planes of @xmath15 rather than tangent planes of another cayley surface . for @xmath164
it is immediate form ( [ eq : n_allgemein ] ) that the matrix @xmath165 does not depend on the parameter @xmath166 , whence in this particular case the mapping @xmath139 is merely the restriction of a null polarity of @xmath10 to the affine part of the cayley surface @xmath15 .
h. havlicek and k. list . a three - dimensional laguerre geometry and its visualization . in g.
wei , editor , _ proceedings dresden symposium geometry : constructive & kinematic ( dsg.ck)_ , pages 122129 , institut fr geometrie , technische universitt dresden , dresden , 2003 . | cayley s ( ruled cubic ) surface carries a three - parameter family of twisted cubics .
we describe the contact of higher order and the dual contact of higher order for these curves and show that there are three exceptional cases . _
2000 mathematics subject classification : _
53a20 , 53a25 , 53a40 . _
keywords : _ cayley surface , twisted cubic , contact of higher order , dual contact of higher order , twofold isotropic space .
_ dedicated to gunter wei on the occasion of his 60th birthday , in friendship _ |
from the cosmological observational data @xcite , it is now well established that the present observable universe is undergoing an accelerating expansion . while the source driving this cosmic acceleration
is known as ` dark energy ' its origin has not been well understood yet due to absence of a consistent theory of quantum gravity .
this acceleration is driven by the negative pressure of the dark energy .
the ` cosmological constant ' is the most simple and natural candidate for explaining cosmic acceleration but it faces serious problems of fine - tuning and large mismatch between theory and observations @xcite . hence there has been significant development in the construction of dark energy models by modifying the geometrical part of the einstein - hilbert action .
this phenomenological approach is called as the modified gravity which can successfully explain the rotation curves of galaxies , the motion of galaxy clusters , the bullet cluster , and cosmological observations without the use of dark matter or einstein s cosmological constant @xcite .
the @xmath5 theories can produce cosmic inflation , mimic behavior of dark matter and current cosmic acceleration , being compatible with the observational data @xcite ( also see a recent review @xcite on @xmath5 gravity and its cosmological implications ) .
in a recent paper @xcite , the authors considered a generalized gravity model @xmath0 , with @xmath2 being the trace of stress - energy tensor , manifesting a coupling between matter and geometry . by choosing different functional forms of @xmath6
, they solved the dynamical equations of astrophysical and cosmological interest . in the present work
, we study the same model by taking different kinds of the energy sources .
reconstruction of the cosmological models for this theory have been ivestigated by several authors @xcite .
the paper is organized as follows . in section
ii we present the general action and the equation of the motion for @xmath0 . in section iii
we reconstruct the cosmological models from the dust fluid . in section
iv we generalize the dust fluid models to the general fluid with eos @xmath7 .
section v is devoted to discussion and conclusions .
the action of @xmath0 gravity is of the form @xcite @xmath8 where @xmath0 is an arbitrary function of the scalar curvature @xmath9 and the trace @xmath10 of the energy - momentum tensor @xmath11 .
we define the lagrangian density for matter field @xmath12 by @xmath13 the equation of motion ( eom ) is obtained by varying the action ( [ 1d ] ) with respect to @xmath14 @xcite @xmath15 where @xmath16 and @xmath17 .
@xmath18 is the operator for covariant derivative and box operator ( or d alembert operator ) @xmath19 is defined via @xmath20 performing a contraction of indices in ( [ eom ] ) , we obtain @xmath21 here @xmath22
. we will use ( [ eq ] ) to reconstruct @xmath0 for different kinds of matter .
we consider the metric of a flat friedmann - robertson - walker ( frw ) spacetime @xmath23 following the conservation of the energy momentum tensor @xmath24 for metric ( [ metric ] ) we obtain @xmath25 the @xmath26component of ( [ eom ] ) reads @xmath27 in ( [ 8d ] ) , if we know the functional form of the hubble parameter @xmath28 as a function of the two independent variables @xmath1 and @xmath2 , then reconstruction of the model @xmath0 is straightforward .
but always , we have a physical intuition about the form of @xmath29 .
we clarify it here more . in a frw model
, the form of the ricci scalar is @xmath30 .
any cosmological era has a specific matter fields , radiation , phantom field or a mixed of them .
for example in @xmath3cdm model , we have @xmath31 .
it is easy to see that we can write @xmath1 in terms of @xmath28 as @xmath32 . thus usually , we can write @xmath33 . thus ( 11 )
can be integrated easily to give the functional form @xmath0 .
also , since @xmath34 , @xmath35 thus all terms of the ( [ tdot ] ) are functions of @xmath1 , @xmath2 . for term as @xmath36
we can write it as @xmath37 .
mathematically , ( [ 8d ] ) is a second order partial differential equation for @xmath0 .
we now discuss the solutions of ( [ 8d ] ) relevant in cosmological context . in the coming sections
we will show that , any cosmological epoch ( radiation , matter , dark energy dominated eras ) can be constructed in a model of @xmath0 only with dust fluid as the source .
we mention here that the form of the action @xmath0 is not unique .
for this reason we can continue our investigation of the cosmological reconstruction for another model with @xmath38 @xcite .
when we set @xmath39 , we recover the @xmath5 theory which has been discussed in @xcite . adopting the technique of the reconstruction of @xmath0 models , we have the following different models reconstructed from the dust fluid . in einstein gravity ,
the hubble parameter for a flat frw model with real matter field describes the @xmath3cdm model by @xmath41 , in units @xmath42 @xmath43 . since @xmath44 then the field equation ( [ 8d ] ) converts to the following form @xmath45f_{r}+ \frac{1}{6}f=\frac{2}{3}t(8\pi+f_t)+81\big(\frac{r}{3}-3 h_0 ^ 2\big)^{3}f_{rr}\\\nonumber+9\big(\frac{r}{3}-3 h_0 ^ 2\big)t^2 f_{tt}+54\big(\frac{r}{3}-3 h_0 ^ 2\big)^2tf_{rt}-3t\big(\frac{r}{6}-5h^2\big)f_t.\end{aligned}\ ] ] the solutions of ( [ eqa ] )
can be obtained under two special cases as : * @xmath46 and @xmath47 where @xmath48 , @xmath49 and @xmath2 are constants .
equation ( 15 ) gives @xmath50 and @xmath51 .
* @xmath52 and @xmath53 where @xmath54 , @xmath55 and @xmath1 are constants .
equation ( 15 ) gives @xmath56 and @xmath57 . in einstein gravity ,
a universe filled with a mixture of the phantom and non - phantom components obey from the following expression for h , @xmath58 . here
the set of the variables @xmath59 ( energy density of non - phantom matter ) @xmath60 ( energy density of phantom matter ) @xmath61 are the parameters of the model . in the early universe when the scale factor was small ,
the first term in @xmath62 dominates and _ it behaves as the universe described by the einstein gravity with a matter whose eos parameter is @xmath63 _ which means that the matter field is non - phantom like .
but for present era we have @xmath64 which means we live in a phantom epoch of the universe . as a special case , @xmath65 which is a special form discussed @xcite we have @xmath66.\end{aligned}\ ] ] thus ( [ 8d ] ) converts to the following form @xmath67)\\&&\nonumber+\frac{1}{6}f=\frac{2}{3}t(8\pi+f_t ) + ( 2\frac{\sqrt{\frac{a}{r}+br}(\frac{r}{6}-2 h^2)}{b-\frac{a}{r^2}})^2f_{rr}+9(\frac{r}{3}-3 h_0 ^ 2)t^2 f_{tt}\\&&\nonumber-6\sqrt{\frac{r}{3}-3 h_0 ^ 2}t\{2\frac{\sqrt{\frac{a}{r}+br}(\frac{r}{6}-2 h^2)}{b-\frac{a}{r^2}}\}f_{rt}-3t(\frac{r}{6}-5h^2)f_t.\end{aligned}\ ] ] we solve equation ( [ phantom ] ) using an ansatz @xmath68 .
the solution of ( [ phantom ] ) is @xmath69 where @xmath70 and @xmath71 are two constants of integration and @xmath72 @xmath73 if we choose @xmath74 a pure constant , then eq . ( [ phantom ] ) gives @xmath75 where @xmath2 is also a constant . furthermore if @xmath46 then eq . ( [ phantom ] ) can not be solved analytically or numerically .
if we live in a universe filled by dust and the scale factor increases exponentially with time as @xmath76 , then hubble parameter is constant .
such model which has been proposed firstly by de sitter , is called the de sitter universe . in this solution
, it is assumed that the only source of matter filling the universe is dust .
thus it is contained in our treatment on dust formation . for a de sitter universe in it
s static patch we ve @xmath77 eq . ( [ 8d ] ) reduces to @xmath78 solution of equation ( [ desitter ] ) is @xmath79 where @xmath80 and @xmath81 are two arbitrary constants and @xmath82 @xmath83 in this case since @xmath84 , thus we have @xmath85 .
thus we have @xmath86 the solution for this purely @xmath2 dependence equation is @xmath87
starting from ( [ eq ] ) and assuming eos @xmath88 , we have @xmath89 the trace of ( [ teta ] ) is given by @xmath90 since @xmath91 thus , we have @xmath92 now we rewrite ( [ eq ] ) in the following form @xmath93 which can be written in suitable form as @xmath94 where @xmath95 .
we will search for exact solutions of @xmath96 in the following cases . substituting this form @xmath97 in ( [ eq3 ] )
we obtain @xmath98 there is only one possibility to get an exact solution : when @xmath1 is constant.in this case , the solution for ( [ eq4 ] ) reads @xmath99 thus we have @xmath100 substituting this form of @xmath101 in ( [ eq3 ] ) we obtain @xmath102 note that the left and right hand sides of ( [ eq5 ] ) are functions of @xmath1 and @xmath2 respectively . thus solving the @xmath2 dependent part of ( [ eq5 ] )
, we obtain @xmath103 the solution for the @xmath1 dependent part of the ( [ eq5 ] ) is complicated .
indeed we must solve the following equation @xmath104 or the following equivalent form @xmath105 one simple but interesting solution is obtained by taking ricci curvature to be constant . from ( [ eq7 ] ) we have @xmath106 thus one of the interesting models is @xmath107 ) against redshift @xmath108 .
we choose @xmath109 , @xmath110 , @xmath111 . ]
( [ david2 ] ) against time @xmath112 .
we choose @xmath109 , @xmath113 , @xmath114 , @xmath115 , @xmath116 ] for this model , we can study the evolution of the hubble parameter and the scale factor . for this purpose , we assume @xmath117 and rewrite ( [ eq7 ] ) in terms of the hubble parameter @xmath118 ( and using @xmath119 ) and the scale factor @xmath120 : @xmath121 the two equations ( [ david1 ] ) and ( [ david2 ] ) are numerically solved and the result is shown in figures 1 and 2 .
the initial conditions are determined using the definitions of hubble parameter and jerk parameter .
specifically @xmath113 , @xmath122 , @xmath123 , @xmath124 , @xmath125 @xcite , @xmath126 , @xmath127 @xcite .
note that our model is compatible with the bao data @xcite for the hubble parameter .
the chaplygin gas ( cg ) has the equation of state @xmath128 .
there are some extensions of this model like the generalized chaplygin gas but we restrict ourselves only to the cg case .
thus we have quantities @xmath129 solving ( [ rho ] ) for @xmath130 and inserting the solution in ( [ tetarho ] ) and by defining a new parameter as @xmath131 we have @xmath132 using ( [ t32 ] ) in ( [ eq ] ) , we have @xmath133 we are searching for constant curvature solutions in which @xmath134 .
thus solution of ( [ t33 ] ) is @xmath135 .
\label{sol5}\end{aligned}\ ] ] we know that there is an important duality between @xmath5 models and scalar fields @xcite .
the lagrangian for a scalar field , which is minimally coupled to the background reads @xmath137 here @xmath138 is a free parameter .
the corresponding expression for stress - energy tensor is @xmath139 now , the expressions for @xmath140 and @xmath2 read @xmath141 eliminating the term @xmath142 from ( [ tetasf ] ) and ( [ tsf ] ) we have @xmath143 we can rewrite ( [ eq ] ) using ( [ tetat ] ) @xmath144 again , we limit ourselves to constant curvature solutions in which @xmath145 ( choosing @xmath146 ) .
then , the general solution for ( [ eqsf ] ) is @xmath147
in summary , we reconstruct cosmological models in the framework of a newly proposed model of @xmath0 gravity , where @xmath1 is the ricci scalar and @xmath2 is the trace of the stress - energy tensor .
we show that the dust fluid reproduces @xmath3cdm , einstein static universe , de sitter universe , phantom - non - phantom era and the phantom cosmology .
further , we reconstruct different cosmological models including , chaplygin gas , minimally coupled scalar field , with some specific forms of @xmath0 .
we demonstrated that for one case @xmath148 , we found that the behavior of hubble parameter @xmath118 and scale factor is compatible with the observational data of bao for small redshift @xmath4 .
m. jamil and d. momeni would like to thank the warm hospitality of eurasian national university , astana , kazakhstan where this work was completed .
all authors would like to thank the anonymous referee for enlightening comments on this paper .
harko t. , lobo f.s.n .
, nojiri s. , odintsov s.d . , ( 2011 ) phys . rev .
d 84 , 024020 .
myrzakulov ratbay . ,
eur.phys.j.c71:1752,2011 , arxiv:1006.1120v2 [ gr - qc ] .
nojiri s. , odintsov s.d .
, saez - gomez d. , ( 2009 ) phys .
b 681 , 74 . cognola g. , elizalde e. , nojiri s. , odintsov s.d . , sebastiani l. , zerbini s. , ( 2008 ) phys . rev .
d 77 , 046009 adham k.s , as@xmath149ss , doi:10.1007/s10509 - 011 - 0963 - 8 .
houndjo m. j. s. , piattella oliver f. , [ arxiv:1111.4275 ] boehmer christian g. , harko tiberiu , lobo francisco s. n. , [ arxiv:1110.5756 ] harko tiberiu , koivisto tomi s. , lobo francisco s.n .
, olmo gonzalo j. , [ arxiv:1110.1049 ] hamani daoud m. , rodrigues manuel e. , houndjo m. j. s. , isrn astronomy and astrophysics , 2011 , i d 341919 ( 2011)[arxiv:1109.2928 ] houndjo m. j. s. , [ arxiv:1107.3887 ] | * abstract * in this paper , we reconstruct cosmological models in the framework of @xmath0 gravity , where @xmath1 is the ricci scalar and @xmath2 is the trace of the stress - energy tensor .
we show that the dust fluid reproduces @xmath3cdm , phantom - non - phantom era and the phantom cosmology .
further , we reconstruct different cosmological models including , chaplygin gas , scalar field with some specific forms of @xmath0 .
our numerical simulation for hubble parameter shows good agreement with the bao observational data for low redshifts @xmath4 . |
the propagation of cosmic rays ( crs ) through turbulent magnetic fields is effectively described as a diffusion process .
this approach provides an excellent description of the observed spectra and chemical abundances of galactic crs up the cr knee at a few pev .
the arrival direction of crs is mostly isotropized in this process and the location of the sources is hidden .
the first order correction to the isotropic distribution is a small dipole anisotropy that scales with the local cr gradient according to fick s law .
the phase and strength of this dipole is expected to be a combined effect of the relative motion of the solar system with respect to the frame where crs are isotropic @xcite , the density gradient of crs , the strength of diffusion , and the strength and orientation of regular magnetic fields @xcite .
various cr observatories have revealed anisotropies in the arrival directions of tev pev galactic crs on large and small angular scales @xcite . in this letter
we will focus on small - scale anisotropies that have been studied via a power spectrum analysis by icecube @xcite and hawc @xcite ( see fig . [ fig2 ] ) . @xcite
have argued that small - scale anisotropies can naturally emerge as a next - to - leading order correction of the diffusive approximation if one takes into account the local scattering of crs within the diffusive scattering length .
@xcite showed that under certain simplifying assumptions these small - scale fluctuations necessarily arise from the application of liouville s theorem . assuming a simple hierarchical evolution equation of the multipoles predicts a power spectrum that matches well the observations .
fluctuations of the arrival direction of crs induced by local turbulent fields are not directly accessible in standard diffusion theory .
this becomes obvious if one considers the angular power spectrum of the cr phase - space distribution ( psd ) @xmath0 defined as @xmath1 where @xmath2 denotes a unit vector of @xmath3 and @xmath4 are legendre polynomials of degree @xmath5 . here and in the following we use the abbreviation @xmath6 for the local psd .
in general , the small - scale magnetic field that determines the psd is unknown .
therefore , the psd can only be predicted as an average over a statistical ensemble of random small - scale magnetic fields , usually characterized by its power spectrum . in ergodic systems , the average over turbulent magnetic field configurations , _
i.e. _ , the _ ensemble _ average denoted in the following by @xmath7 , is equivalent to a time - average .
however , the ensemble average of @xmath8 is proportional to @xmath9 , which is _ larger _ than the product of the ensemble averages , _
i.e. _ , the standard solution of diffusion theory .
on large timescales , the motion of crs in turbulent magnetic fields can be effectively approximated as a diffusion process .
the angular - integrated psd @xmath10 then satisfies a diffusion equation , @xmath11 , where the diffusion tensor @xmath12 can be expressed as @xmath13 here , @xmath14 denotes a unit vector along the regular magnetic field . in general , the ( scattering ) rates @xmath15 , @xmath16 and @xmath17 are independent of one another .
the presence of a small local gradient @xmath18 induces a dipole anisotropy in the quasi - stationary solution as a result of fick s law , @xmath19 .
therefore , standard diffusion theory predicts that the dipole strength of the relative intensity @xmath20 is given as @xmath21 we will see in the following that this relation becomes modified once we consider corrections of the psd product in the ensemble average .
this will also introduce multipoles at small angular scales , which can be related to properties of relative diffusion .
we now study the effect of small local fluctuations of the psd around the ensemble average , @xmath22 . according to liouville s theorem
we can relate the local ( _ i.e. _ , @xmath23 ) psd @xmath6 to the contribution backtracked along cr trajectories to an arbitrary time , @xmath24\nabla n\,,\end{gathered}\ ] ] where @xmath25 and @xmath26 denote the local cr density and gradient and @xmath27 and @xmath28 are the position and momentum of a cr ( that is at position @xmath29 and @xmath30 at time @xmath31 ) at backtracking time @xmath32 .
now , in the limit of large @xmath32 the last term in eq .
( [ eq : backtrack ] ) is dominated by the third term scaling with the position of the particle .
also , for two momenta @xmath33 we can assume that the ensemble average of fluctuations are uncorrelated , @xmath34 , for sufficiently large backtracking times when the cr trajectories eventually separate .
in the degenerate case @xmath35 the two backtracked cr trajectories stay correlated over arbitrarily long backtracking times .
it will be sufficient to assume that @xmath36 remains finite .
we can then express the multipole spectrum of the ensemble - averaged relative intensity as the limit @xmath37 @xmath38 being shorthand for @xmath39 .
note that the @xmath40 multipole spectrum is generated through _ relative _ diffusion : it can be easily seen that the sum over _ all _ ensemble - averaged multipoles of the relative intensity can be expressed via the symmetric part of the diffusion tensor @xmath41 in the limit of large backtracking times @xmath32 , @xmath42 on the other hand , the average monopole contribution in this limit can be expressed as @xmath43 where the symmetric part of the relative diffusion tensor is defined as @xmath44 with @xmath45 . therefore , the sum of multipoles @xmath40 is related to the relative diffusion tensor . for uncorrelated particle trajectories , this expression reduces to the normal diffusion tensor .
however , particle trajectories with a small relative opening angle will follow similar trajectories and the relative contribution ( [ eq : krel ] ) remains small over long timescales .
note that the multipoles in eq .
( [ eq : cell ] ) are expected to be finite in the limit of large backtracking times since particle trajectories with arbitrarily small opening angles will eventually become uncorrelated , @xmath46 . ) for a cr gradient parallel ( solid lines ) and perpendicular ( dotted lines ) to the regular magnetic field and the 3d turbulence model discussed in the main text .
we show results in terms of the dipole @xmath47 ( black ) , monopole @xmath48 ( blue ) , and medium-@xmath5 multipoles ( green ) .
we also show the asymptotic noise level ( [ eq : noise ] ) ( red ) and the dipole prediction ( [ eq : standardc1 ] ) of standard diffusion ( magenta ) evaluated by the replacement @xmath49 in eq .
( [ eq : cell ] ) . ]
in the following , we will study the development of small - scale anisotropies via numerical simulations ( see also @xcite ) .
we follow the approach of @xcite and define a three - dimensional ( 3d ) turbulent magnetic field as the sum @xmath50 with @xmath51 random phases @xmath52 and wave vectors @xmath53 with 3d random orientations , on top of a regular field @xmath54 .
the wave vector amplitudes @xmath55 range from @xmath56 to @xmath57 with equal logarithmic steps .
the vectors @xmath58 have 3d random orientations subject to the conditions @xmath59 and @xmath60 with a coherence scale @xmath61 .
we assume a kolmogorov - type phenomenology with @xmath62 and the strength of the turbulence relative to the regular magnetic field is fixed via a parameter @xmath63 and the normalization of the spatial average @xmath64 . for the galactic environment
we expect a regular magnetic fields of the order of @xmath65 g and a turbulent component with @xmath66 and a coherence scale @xmath67 pc .
the larmor radius of 1 - 10 tev crs is then of the order of @xmath68 corresponding to a few hundred au .
a numerical study of this configuration with the methods described above is very time - consuming since the asymptotic behavior relevant for the multipole configuration is reached at a very late stage . instead , here we will study a simple turbulent configuration that allows us to reach the necessary statistics to study the predicted effect qualitatively .
we choose @xmath69 with @xmath70 and @xmath71 with @xmath72 and fix the rigidity of the crs to @xmath73 .
we sample over @xmath74 different turbulent magnetic field configurations . for each field configuration
we uniformly sample @xmath75 cr orientations @xmath76 following the _ healpix _ parametrization ( @xmath77 ) @xcite , which we backtrack through the static magnetic field @xmath78 to find the initial position @xmath27 .
figure [ fig1 ] shows the power spectrum at different backtracking times determined via a multipole expansion of the relative intensity sky maps using the _ healpix _ utilities .
we show the cases of cr gradients parallel ( solid lines ) and perpendicular ( dotted lines ) to @xmath54 .
the black lines show the evolution of the ensemble - averaged dipole induced by the cr gradient .
the magenta lines correspond to the standard dipole anisotropy defined via eq .
( [ eq : standardc1 ] ) evaluated by the replacement @xmath49 in eq .
( [ eq : cell ] ) . as expected ,
this dipole estimate is smaller than the ensemble average .
note that for the case of perpendicular diffusion ( dotted lines ) the difference in the asymptotic values is about one order of magnitude .
the green lines show the multipole power for @xmath79 .
we also show the expected monopole component ( [ eq : c0 ] ) as blue lines . at large times , the high-@xmath5 components in the multipole expansion are dominated by noise , due to the finite number of trajectories .
the noise level can be estimated for large backtracking times @xmath32 as pixel shot noise @xmath80 this is indicated as red lines in fig .
[ fig1 ] and clearly influences the level of high-@xmath5 multipoles ( green lines ) in the map .
( normalized to @xmath81 ) for the parallel ( filled red diamonds ) and perpendicular ( filled green triangles ) cr gradient for the same simulations as shown in fig .
[ fig1 ] at a backtracking time @xmath82 .
we also show the icecube @xcite ( blue open circles ) and the hawc @xcite ( magenta open squares ) power spectra renormalized by a factor of @xmath83 for better comparison . the different noise level of the data is indicated as dotted lines .
the three gray lines correspond to the prediction of a relative scattering term @xmath84 in eq .
( [ eq : finalcl ] ) for three different values of @xmath85 . ]
the best estimator for the true power spectrum is then @xmath86 and the variance ( excluding cosmic variance ) is given by @xmath87 ; see,_e.g . _ ,
@xcite . in fig .
[ fig2 ] we show the estimators @xmath88 normalized to the dipole @xmath81 at a backtracking time @xmath89 , corresponding to the timescale of the transition into the diffusion regime ( cf
. fig .
[ fig1 ] ) . given the finite number of trajectories , it might be difficult to disentangle noise variance and cosmic variance in the simulations ; we hence limit ourselves to the former only . similar to the simulation , the sensitivity of experimental data to high-@xmath5 multipoles is limited to shot noise indicated as dotted lines in the plot .
overall , the simulation agrees well with the icecube and hawc data . for an isotropic gaussian field
, deviations at multipoles of @xmath90 can be due to cosmic variance with @xmath91 .
furthermore , the limited field of view of the observatories introduces additional systematic errors , in particular for low-@xmath5 multipoles , which are not included in the error bars of fig .
in the following , we derive the asymptotic form of the power spectrum via a microscopic description of relative diffusion . as before , we split the psd into the average and fluctuating parts , @xmath92 .
the ensemble - averaged boltzmann equation can be written as @xcite @xmath93 here we have introduced the angular momentum operator @xmath94 and the rotation vectors @xmath95 and @xmath96 . the rhs of eq .
( [ eq : genboltzmann ] ) encodes the particle scattering in the random fields .
the influence of the turbulence is typically approximated as a friction term introduced by @xcite ( bgk ) . in its original version
this term drives the ensemble - averaged distribution @xmath97 to its isotropic mean @xmath25 with an effective relaxation rate @xmath98 , _
i.e. _ , @xmath99 .
the parameters @xmath15 , @xmath16 , and @xmath17 are the effective parallel , perpendicular , and axial scattering rates , and can be related to the bgk rate @xmath98 via @xmath100 , @xmath101 , and @xmath102 . substituting into eq .
( [ eq : k ] ) we find that the standard @xmath103 and @xmath104 ( [ eq : standardc1 ] ) scale as @xmath105 and @xmath106 , respectively , similar to the sum ( [ eq : id1 ] ) in the relative diffusion approach .
we can generalize the bgk approximation of the friction term to higher multipoles if we consider the effect of turbulence on the particle momenta as a brownian motion on a sphere @xcite . for a distribution @xmath107 on a sphere the brownian diffusion of momenta
is described via @xmath108 . in particular ,
if initially the configuration is localized , @xmath109 , then the solution at later times is given as @xmath110 .
this reduces to a two - dimensional gaussian distribution of the opening angle @xmath111 defined via @xmath112 with a width @xmath113 .
therefore , for the rhs of eq .
( [ eq : genboltzmann ] ) we make the ansatz @xmath114 which is a natural generalization of the bgk approach . note that this simple treatment of the scattering term is only approximate .
depending on the strength and coherence length of the turbulent magnetic field , the effective relaxation can in general be anisotropic even for a fully isotropic turbulence .
however , in our case we are interested in the qualitative behavior and development of high-@xmath5 multipoles , and the first order bgk approximation seems appropriate .
now , for the asymptotic solution of @xmath8 at large backtracking times we have to look for stationary solutions of the equation @xmath115 as before , we make the ansatz @xmath116 with the stationary solution @xmath117 . assuming that the spatial dependence of @xmath118 is small compared to @xmath119 we can replace the gradient term in eq .
( [ eq : clevolution ] ) via @xmath120 . on the other hand , the second and third terms in eq .
( [ eq : clevolution ] ) are now non - trivial terms in the evolution that will lead to mixing and damping of multipoles .
analogous to the bgk - type approximation @xcite of eq .
( [ eq : bgk ] ) , we make the ansatz , @xmath121\langle f_1f_2\rangle\,,\end{gathered}\ ] ] with @xmath122 and @xmath123 .
the relative and correlated scattering rates , @xmath124 and @xmath125 , respectively , depend on the relative distance of the trajectories at early times , which in turn depends on the trajectories opening angle . note that eq .
( [ eq : bgk2 ] ) is necessarily symmetric under the interchange of @xmath126 as well as @xmath127 and is the most general linear approximation of all possible scalar combinations of @xmath128 and @xmath129 .
note that as in the original bgk ansatz , we do not account for the possibility of anisotropic scattering in turbulent fields .
we now look for the power spectrum of a stationary solution of eq .
( [ eq : clevolution ] ) using the approximation ( [ eq : bgk2 ] ) .
the product @xmath130 can be expanded into eigenstates of @xmath131 and @xmath132 and the power spectrum integral ( [ eq : celldef ] ) corresponds to projections onto the state @xmath133 with @xmath134 . the rotation term corresponding to the regular magnetic field in eq .
( [ eq : clevolution ] ) does not contribute .
since the relaxation rates @xmath135 and @xmath136 in the approximation ( [ eq : bgk2 ] ) are assumed to depend only on @xmath137 the relative rotation proportional to @xmath138 is the only non - vanishing term . for identical momenta and rigidities , the relative diffusion should vanish completely .
one way to see this is by considering a toy model where the initial configuration of crs is homogeneous @xcite . from this setup
one can infer that the psd at any time obeys @xmath139 .
therefore , in this case we must also have @xmath140 .
since this is independent of the initial configuration we must therefore conclude that @xmath141 .
a functional dependence can be obtained from a geometrical argument : two particles emitted locally under a small opening angle will separate with a reduced relative velocity @xmath142 as long as both particles are experiencing a strongly correlated magnetic field .
this leads to the ansatz @xmath143 with @xmath144 . to allow for a weaker or stronger @xmath145-dependence
, we have generalized this form to @xmath143 , with @xmath146 . under these assumptions and approximations ,
the stationary average @xmath147 spectrum then obeys the set of equations ( see appendix [ app1 ] ) @xmath148 with the effective dipole source term @xmath149 .
since the rhs of the previous equation is an expansion in terms of legendre polynomials , we can easily derive the stationary spectrum as @xmath150 with the assumed @xmath143 , this simplifies to @xmath151 in the limit @xmath152 , @xmath153 , and so @xmath154 .
for @xmath155 , this is close to the @xmath156 behavior exhibited by the data ( cf .
[ fig2 ] ) . figure [ fig2 ] also shows examples of the power spectrum ( [ eq : finalcl ] ) for three models @xmath157 with @xmath158 , @xmath159 , and @xmath160 .
while our ansatz @xmath143 is phenomenological in nature , a microscopic derivation of @xmath135 , in quasi - linear theory , will be presented in future work .
we have discussed the appearance of small - scale anisotropy from the propagation of crs in local turbulent magnetic fields . we have shown that this effect can be understood as a phenomenon of relative diffusion , where crs with similar arrival directions experience correlated motion in the local magnetic environment .
our numerical simulation showed that the ensemble - averaged dipoles can be significantly larger than those expected from standard diffusion theory .
in addition , the formalism predicts significant power in the ensemble - averaged small - scale multipoles .
we have also provided an analytic framework that can describe the multipole power spectrum via an effective relative relaxation rate of multipole components in the evolution of the ensemble - averaged psd products .
this formalism is a natural extension of the well - known bgk approach . assuming that the relative relaxation rate has a simple power - law dependence on the relative opening angle of cr arrival directions , we arrive at a power - law spectrum very similar to that obtained from simulations and the observed cr data . _
acknowledgements._the authors would like to thank paolo desiati , luke drury , gwenael giacinti , alex lazarian , vanessa lpez - barquero , martin pohl , and stefan westerhoff for comments .
we thank dan fiorino and marcos santander for discussions and providing the data points of fig .
is supported by the national science foundation under grants opp-0236449 and phy-0236449 .
p.m. is supported by doe contract de - ac02 - 76sf00515 and a kipac kavli fellowship .
the decomposition of the psd @xmath161 into spherical harmonics @xmath162 corresponds to a projection onto eigenfunctions @xmath163 of the angular momentum operator @xmath164 .
the products of psds that appear in the definition of the power spectrum in eq .
( [ eq : celldef ] ) are then eigenfunctions of @xmath165 that can be decomposed into eigenfunctions @xmath166 of the total angular momentum operator @xmath167 . as noted earlier , the integral over arrival directions in eq .
( [ eq : celldef ] ) corresponds to a projection of @xmath168 onto states with @xmath133 and @xmath134 and @xmath169 , which can be seen from the expansion into spherical harmonics @xmath170 following from the clebsch - gordan coefficients @xmath171 now , the bgk ansatz ( [ eq : bgk2 ] ) assumes that the scattering rates @xmath172 and @xmath173 are only functions of @xmath174 and can thus be expanded into a sum over states with @xmath133 and @xmath134 .
the projection of eq .
( [ eq : bgk2 ] ) onto @xmath175 can thus only receive contributions from the projection of the product of psds onto @xmath133 and @xmath134 states , @xmath176 finally , the stationary solution ( [ eq : stationary ] ) has the form @xmath177 and hence @xmath178 . from this
we immediately arrive at eq .
( [ eq : finalcl ] ) . | the arrival directions of multi - tev cosmic rays show significant anisotropies at small angular scales . it has been argued that this small - scale structure can naturally arise from cosmic ray scattering in _ local _ turbulent magnetic fields that distort a global dipole anisotropy set by diffusion .
we study this effect in terms of the power spectrum of cosmic ray arrival directions and show that the strength of small - scale anisotropies is related to properties of relative diffusion .
we provide a formalism for how these power spectra can be inferred from simulations and motivate a simple analytic extension of the ensemble - averaged diffusion equation that can account for the effect . |
vibronic sidebands are observed in the optical spectra of many solid - state materials @xcite , with ti : sapphire @xcite and nitrogen - vacancy ( nv ) centres in diamond @xcite being good examples .
the key point about a vibronic transition is that it involves the simultaneous absorption or emission of a photon and a phonon ( or phonons ) as the electron jumps between two electronic states .
the sideband spectra are continuous bands , although sub - structure can frequently be identified due to the involvement of specific phonon modes , especially at low temperatures .
for example , in the case of nv centres in diamond , clear structure can be identified that corresponds to the coupling of the electronic state to the a@xmath1 mode ( @xmath2 mev ) , giving rise to resolved side - peaks in both absorption and emission at integer multiples of the phonon energy @xcite .
the coupling to the vibrational modes is so strong that only a few percent of the emission occurs in the zero - phonon line , with most of the photons emitted from the sidebands @xcite .
the relatively weak intensity of the zero - phonon line has serious consequences for practical applications of nv centres in optical quantum information processing .
the reason why the phonon sidebands in materials like diamond nv centres and ti : sapphire are so strong is that both the electronic and vibrational modes are strongly localized on length scales similar to the unit cell size .
this means that the overlap between the electronic wave functions and the phonon modes is large , and hence the vibronic coupling strong .
by contrast , ingaas quantum dots ( qds ) have envelope wave functions localized on much larger length scales that are determined by the size of the dot , i.e. @xmath3 nm . this means that the coupling to phonons is relatively weak , with the dominant interaction being to longitudinal - acoustic ( la ) phonons via deformational potential scattering .
the weak vibronic coupling in these dots gives rise to very strong emission in the zero - phonon line , with only @xmath48% in the sideband at cryogenic temperature @xcite .
this makes ingaas qds excellent single - photon sources @xcite .
it also ensures that the light - matter coupling is strong , with optical dipole moments typically in the range of @xmath5 debye ( @xmath6 cm ) @xcite .
moreover , the relatively weak electron - phonon coupling leads to long coherence times that are ultimately limited only by the radiative lifetime @xcite which facilitates their application in coherent control experiments @xcite .
this contrasts with bulk and quantum well samples , where the coherence time is only a few picoseconds at best @xcite .
the vibronic sideband in ingaas quantum dots manifests itself in a number of important experimental situations .
one example is the emission of photons from a nano - cavity when the dot and cavity are out of resonance , with the emission involving either the absorption or emission of a phonon depending on the sign of the detuning @xcite .
another case is the observation of qd emission when pumping in the phonon sideband in a resonance fluorescence geometry @xcite .
the difference in the frequency of the pump laser and the qd exciton line facilitates spectral selection of the qd photons @xcite and can be exploited for stabilizing the frequency of the zero - phonon line @xcite .
the sideband also appears in four - wave mixing experiments @xcite .
a key point about the experiments described above is that the optical pumping is relatively weak , so that the qd - laser system is in the weak - coupling limit . in this paper
we explore the other limit where the qd - laser coupling is strong , such that a dressed - state picture is appropriate .
this limit was addressed in a theoretical paper in 2013 by glssl _
et al _ , where it was predicted that strong driving in the phonon sideband could lead to exciton populations approaching full inversion @xcite .
the predictions of the theory were confirmed independently by three experimental groups in 201415 , both for pumping of the neutral exciton and the biexciton @xcite .
most recently , it has also been demonstrated that the process works in reverse , so that an inverted system can be rapidly depopulated by pumping with a laser with red detuning relative to the exciton @xcite . in these experiments
the phonon coupling becomes strong when an intense laser field is present .
the pulse durations in the experiments are short , and so the vibronic coupling is turned from weak to strong and back to weak again on picosecond timescales .
we therefore call this process _ dynamic vibronic coupling _ ( dvc ) .
the dvc is fundamentally different from conventional excitation schemes , such as coherent rabi oscillation @xcite and adibatic rapid passage protocols @xcite where exciton - phonon coupling is usually an obstacle . in this paper
we first review the process underlying the dvc and summarise the results from our previous experiments @xcite . after discussing the experimental methods in section [ sec : experiment ] ,
we then present ramsey interference data to demonstrate that the mechanism of dvc is incoherent , i.e. that the exciton created by phonon - sideband pumping is incoherent with the pumping laser ( section [ sec : coherence ] ) .
we next consider the temperature dependence of the dvc in section [ sec : t - dependence ] , comparing the population generated at @xmath7k to that at base temperature ( 4.2k ) .
finally we show in section [ sec : push me pull you ] how we can combine pumping with a blue - detuned and a red - detuned pulse to create and destroy an exciton within @xmath8ps .
section [ sec : conclusions ] gives the conclusions and outlook .
= 0 with @xmath9 = + 1 mev and @xmath10 = 8.5 @xmath11 . ( a )
energies of the dressed states ( green lines ) @xmath12 and @xmath13 plotted against time . at @xmath14 ,
the states are split by the laser detuning @xmath9 . during the passage of the pulse the states
are admixed and the splitting becomes @xmath15 as shown by the shift centred at @xmath16 = 0 .
this enables relaxation from @xmath12 to @xmath13 ( blue arrows ) by emission of an la phonon with energy @xmath15 .
( b ) plot of the admixing angle @xmath17 ( black line ) and the population of the dressed state @xmath13 ( blue line ) against time .
@xmath17 also follows the envelope of the laser pulse and , as expected , relaxation into @xmath13 occurs only when the states are admixed . ]
the starting point for dvc is the coupling of excitons to the acoustic phonon bath by the deformation potential @xcite . in the absence of a laser field
this leads to nonexponential pure dephasing of the excitonic dipole @xcite .
the behaviour of the coupled system becomes somewhat more interesting however when a strong laser field is applied .
we consider a laser pulse with energy detuning @xmath18 and area @xmath10 : @xmath19 where @xmath20 is the time - dependent rabi frequency determined by the optical dipole moment and the time - varying electric field amplitude of the pulse .
using the rotating wave approximation , the hamiltonian in the rotating frame becomes time independent and one can thus define the dressed states of the laser - qd states as the corresponding eigenstates .
these are given by : @xmath21 @xmath22 where @xmath17 is an admixing angle defined by : @xmath23 the energies of the dressed states are given by : @xmath24 where @xmath25 is an effective rabi frequency defined as : @xmath26
the significance of these dressed states is that , in the presence of a driving laser such that @xmath27 , both eigenstates of the system contain an excitonic component and may couple to the acoustic phonon bath .
this enables phonon - mediated relaxation from @xmath12 to @xmath13 .
however , this relaxation is only possible when the states are admixed by the laser , giving rise to the dynamic nature of the vibronic coupling . figure .
[ fig : mixingj ] illustrates the dressed state energies ( green lines ) , admixture angle @xmath17 ( black line ) and occupancy of @xmath13 ( blue line ) against time during the passage of a @xmath28 pulse with @xmath29 .
the pulse has a temporal fwhm of 16.8 ps and is centred on @xmath30 .
the figure illustrates that the dressed state splitting rises and falls in line with the envelope of the laser pulse @xcite , and that the admixing angle follows it .
the transfer of population into @xmath13 only occurs whilst the states are admixed , turning the system from weakly to strongly vibronic and back again in approximately 20 ps .
and @xmath13 plotted against detuning @xmath31 at a fixed time .
the bare qd states in the frame of the laser are plotted for reference as dotted black lines . at @xmath32 the states anti -
cross with a splitting of @xmath33 .
it can be seen that @xmath12 is dominated by the ground state at @xmath34 , leading to phonon - assisted excitation of the exciton .
conversely , @xmath12 is primarily excitonic in character at @xmath35 , and thus phonon - assisted de - excitation of the exciton occurs .
the blue arrows illustrate the relaxation process corresponding to the emission of an la phonon whilst the red arrows show the competing phonon absorption process .
( b - e ) phenomenological energy level diagrams of the phonon emission and absorption processes for both positive and negative detunings . for simplicity
, the phonon relaxation is incorporated as a virtual state ( dashed lines ) . ]
the transfer between the dressed states shown in fig .
[ fig : mixingj ] can be exploited to achieve ultrafast incoherent excitation and de - excitation of the exciton by detuned laser pulses . beginning with excitation , in the case of @xmath34
the higher energy dressed state @xmath12 is dominated by the crystal ground state whilst the lower energy state @xmath13 is primarily excitonic in character as illustrated to the right of fig .
[ fig : relaxation](a ) . during the passage of the laser pulse , admixing of the states allows relaxation to occur from @xmath12 to @xmath13 by emission of an la phonon ( @xmath36 - blue arrows ) . for a sufficiently strong pulse ,
most of the population relaxes into @xmath13 and the exciton - dominated nature of @xmath13 means that the population of the exciton state after the passage of the pulse is high .
[ fig : relaxation](b ) shows a phenomenological level diagram of the phonon - assisted excitation process whilst fig .
[ fig : relaxation](c ) illustrates the competing process whereby a phonon is absorbed ( @xmath37 - red arrows ) and a photon emitted ( purple arrow ) .
the latter process is weak at low temperatures . in the case of @xmath35
the main difference is that the characteristics of the dressed states are now exchanged . during the passage of the laser pulse , relaxation by phonon emission
again occurs into the @xmath13 state .
however , as this state is now dominated by the crystal ground state , the effect of the relaxation is instead to de - excite the exciton with most of the population left in the ground state after the passage of the laser pulse .
[ fig : relaxation](d ) shows an illustrative schematic of this process with the emission of a photon ( purple arrow ) accounting for the rest of the energy difference .
this emission is stimulated by the laser and hence the process is termed la - phonon stimulated emission ( lapse ) .
[ fig : relaxation](e ) again illustrates the competing phonon absorption process . as determined by fitting data measured in ref .
the interaction strength peaks at a cut - off of around 2 mev and rolls off rapidly beyond this . ]
the physical factors that determine the efficiency of phonon - assisted excitation and de - excitation may be considered by analysing the parameters that enter eq .
[ eq : admixture ] .
the degree of admixing increases by increasing the driving strength ( @xmath38 ) or decreasing the detuning ( @xmath31 ) ( noting that @xmath31 should exceed the laser linewidth to exclude resonant coherent driving ) , leading to more efficient relaxation .
however , both of these dependencies are modified by the properties of the phonon bath .
the exciton - phonon interaction strength is characterised by the function @xmath39 which increases with @xmath40 at first due to the rising phonon density of states , and then rolls - off rapidly beyond a cut - off frequency which is typically around 1 - 2 mev @xcite .
the physical origin of this cut - off is the point at which the phonon wavelength is comparable to the spatial fwhm of the carrier wavefunction . as a result
, the cut - off frequency depends strongly on the height of the qd and phonon sideband measurements may be used to probe the confinement potential . a plot of @xmath39 derived from fitting data measured in ref .
@xcite is shown in fig .
[ fig : phononj ] .
the form of @xmath39 modifies the detuning dependence of the dvc and also weakens the exciton - phonon interaction for very strong driving where the effective rabi frequency @xmath41 exceeds the cut - off frequency ( although this regime has not been reached in present experiments ) .
the processes of interest for dvc both rely on phonon emission rather than phonon absorption , and are therefore highly sensitive to temperature .
increasing the bath temperature increases the phonon occupation leading to a higher probability of phonon emission and thus faster relaxation .
however , the probability of phonon absorption generally increases more than that of emission . as illustrated by the red arrows in fig .
[ fig : relaxation](a ) , phonon absorption ( @xmath37 ) causes the opposite population transfer between the dressed states to phonon emission ( @xmath36 - blue arrows ) and results in a lower final occupation of @xmath13 . as such , dvc processes are generally optimised by low bath temperatures in order to maximise the difference between phonon emission and absorption rates .
the influence of temperature on phonon - assisted excitation is studied both experimentally and theoretically in section [ sec : t - dependence ] .
it is only by combining all of these influences that the full spectral and power dependence of the dvc emerges .
analytical approximations do not accurately predict these dependencies ; hence numerical methods such as path integral calculations @xcite or the master equation formalism @xcite are employed .
the result of such calculations is a broad ( few mev ) spectral sideband feature that appears at high driving strengths and persists until the rabi splitting exceeds the cut - off energy for phonon coupling .
figure [ fig : set - up ] gives a schematic diagram of the apparatus used in the experiments .
a mode - locked ti : sapphire laser with 76 mhz pulse repetition rate is passed through two pulse shapers to obtain independently tunable pulses from within the @xmath42mev bandwidth of the @xmath43fs pulses @xcite .
the spectral fwhm is selected to be either either 0.2 or 0.42 mev corresponding to a fourier - transform - limited pulse duration of 16.8 or 8 ps .
both pulses pass through wave - plates to permit independent control of their polarisations , and one of them follows a variable path length controlled by a delay stage to enable pump - probe experiments with precise relative time control .
the pulses are incident on the sample in a variable temperature liquid he cryostat with piezo actuators for precise positioning of the sample relative to the focussed beams .
the sample consists of a layer of ingaas qds embedded in a schottky diode .
the exciton created by the laser pulses can be measured by detecting the photocurrent ( pc ) from the qd when a reverse bias is applied to the diode @xcite .
further details of the experimental methods may be found in ref .
in ref . @xcite , quilter et al . demonstrated the dvc by creating a qd exciton using a slightly blue - shifted laser pulse via emitting an la phonon .
since it involves emission of a phonon , in theory the dvc should be incoherent , namely the phase of the exciton is random relative to that of the blue - shifted laser pulse .
the exciton coherence time in this case is limited by the phonon relaxation time ( few ps @xcite ) . to demonstrate the incoherent nature of the dvc , we performed ramsey - like interference measurements @xcite using an unstabilized interferometer .
the qd is pumped with a blue - detuned pulse with @xmath44 mev at @xmath30 .
the pulse area ( @xmath45 ) is chosen to give an exciton population of @xmath46 .
( note that this is slightly higher than in ref .
@xcite due to different laser detunings . )
once the qd has been excited , the state is then probed by a @xmath47 pulse resonant with the exciton at @xmath48 .
the relative phase between the pump and probe is proportional to @xmath49 .
if the system is coherent , the probe pulse drives the qd to either @xmath50 or 1 depending on the phase of the exciton with respect to the probe .
the degree of coherence can then be determined by measuring the visibility of the ramsey fringes as a function of @xmath49 . in our experiment ,
the relative phase between the pump and probe is unstabilized , and so the final state fluctuates randomly within the visibility envelope .
k. in both cases the second @xmath47-pulse is resonant with the exciton .
the pulse area @xmath10 of the + 1 mev detuned pump pulse is set to be @xmath45 to generate @xmath51 .
the dashed line is a guide to the eye . ]
[ fig : ramsey fringes ] shows the time - integrated pc signal versus @xmath49 .
the red squares show the results for phonon - sideband pumping with @xmath52 mev .
a flat line is observed , with no ramsey - like fringes , indicating that the phase of the exciton is random relative to the probe pulse .
the slow increase of the pc signal in time is most probably related to the decay of the exciton population created by the pump via electron / hole tunnelling before the arrival of the probe . by contrast ,
the black dots show the results measured when the pump pulse is tuned to resonance ( @xmath53 ) and the pulse area is set to @xmath47 . in this fully resonant situation , ramsey - like fringes
are observed provided @xmath49 is shorter than or comparable to the exciton coherence time .
the exciton coherence time of @xmath54 ps can be estimated from the dashed line envelope and is limited by the electron tunnelling rate @xcite .
the incoherent nature of the dvc is clearly demonstrated by the absence of the fluctuations in the pc signal in the red data compared with that in the black data .
our result is consistent with that reported by weiler et al .
@xcite showing that the coherence properties of the emitted photon created via dvc is much worse than resonant cw excitation .
however , bounouar et al .
@xcite demonstrate that the photons generated by quasi - resonantly pumping the biexciton state using pulsed excitation show similar coherence properties as measured in the resonant two - photon scheme .
the relation of the coherence properties of the emitted photon and exciton prepared via dvc is still unclear and require further investigations .
it should be noticed that the fact that the dvc process is incoherent does not imply that the exciton spin is random .
the results published in ref .
@xcite clearly shows that the spin of the exciton is the same as that of the photon in the blue - shifted laser pulse
. in theory the spin of the emitted photon in the lapse process should also be the same as that of the laser photon , since the phonon - assisted excitation and deexcitation are two opposite processes in the dressed state picture [ see fig .
[ fig : relaxation](a ) ] . were obtained by preparing an exciton in the qd using a circularly - polarized @xmath55 pulse and then measuring the pc as a function of a ( a ) co- or ( b ) cross - polarized red - shifted laser pulse . to isolate the pc signal of the qd under study from other qds in the same sample , a reference spectrum measured with only a red - shifted laser pulse
is subtracted .
the pulse area @xmath56 of the red - shifted laser pulse is ( black ) @xmath55 or ( red ) @xmath57 .
the delay time between the @xmath55 pulse and the red - shifted pulse is 17 ps .
@xmath58 : biexciton . ] directly demonstrating the spin - preserving nature of the lapse process is beyond the scope of this paper , but it has been observed that the lapse process only occurs when the spin of the exciton and the lapse pulse are the same .
this spin selectivity of the lapse process can be demonstrated by preparing a spin - up exciton using a @xmath59 circularly - polarized @xmath55 pulse and then try to deexcite this exciton using a ( a ) co- or ( b ) cross - polarized red - shifted laser pulse [ see figure .
[ fig : spin ] ] . figure .
[ fig : spin ] shows the differential pc spectra @xmath60 measured by exciting the qd using a @xmath59 polarized @xmath55 pulse and then measuring the exciton population by detecting the pc signal as a function of the detuning of a subsequent red - shifted co / cross - polarized laser pulse . the first @xmath55 pulse creates a reference pc level corresponding to an exciton population of 1
[ see the dashed line in fig .
[ fig : spin](a ) ] .
this reference level is calibrated by the pc signal at @xmath61 in the differential spectrum measured with a weak ( @xmath62 ) co - polarized red - shifted pulse where the exciton - phonon coupling is negligible since the laser driving is quite weak [ see black line in fig .
[ fig : spin](a ) ] . in the case that a strong ( @xmath63 ) co - polarized red - shifted pulse is applied , the red - shifted pulse deexcites the exciton , hence reduces the pc single relative to the reference level and forms a negative sideband at @xmath61 [ see red line in fig .
[ fig : spin](a ) ] .
by contrast , no decrease of the pc single is observed at @xmath61 in the cross - polarized case [ see red line in fig .
[ fig : spin](b ) ] , indicating that no de - excitation occurs .
the peak and weak positive sideband at around @xmath64 mev can be attributed to the resonant and phonon - assisted excitation of the biexction .
the absence of the negative phonon sideband in the cross - polarized differential spectrum compared to the co - polarized spectrum [ see red lines in fig . [
fig : spin ] ] ambiguously proves the spin selectivity of the lapse process .
based on this observation and the fact that the spin is preserved while the emission of the la phonon as demonstrated in ref .
@xcite , we conclude that dvc is incoherent , but spin - preserving .
) at 15k ( red triangles ) compared to 4.2k ( black circles ) for excitation with pulse area @xmath10 = 4.4 @xmath11 .
the dashed line indicates a detuning of @xmath9 = + 0.83 mev .
( b ) calculated variation of the exciton population versus pulse area ( @xmath10 ) for @xmath65k ( black ) , @xmath66k ( red ) and @xmath67k ( blue ) for positive detuning of @xmath9 = + 0.83 mev .
the dashed line shows @xmath10 = 4.4 @xmath11 . ] in previous studies dvc processes have mainly been studied at @xmath65k @xcite , the temperature of liquid - he .
it is interesting however to consider the influence of temperature on the mechanism .
increasing the temperature increases the strength of the exciton - phonon interaction , observed for instance by stronger damping and frequency renormalization of excitonic rabi rotations @xcite . as discussed in section [ sec : mechanism ] , this increases the rate of phonon relaxation between the dressed states , and , for relatively weak excitation , increases the resulting exciton population as more population relaxes during the pump pulse .
this is shown by the higher population observed at positive detuning for @xmath66k ( red triangles ) compared to @xmath65k ( black circles ) in fig .
[ fig : detuning](a ) and also by the steeper initial gradient for higher temperatures observed in the calculations of fig .
[ fig : detuning](b ) .
however , this does not illustrate the complete picture . at higher bath temperatures the increased number of phonons increases the probability of the competing process at positive detuning , namely the annihilation of an exciton by absorption of a phonon ( see fig . [
fig : relaxation](c ) ) .
this is shown in our measurements by the reduced detuning asymmetry of the phonon sidebands at @xmath66k in fig .
[ fig : detuning](a ) .
the origin of the negative detuning signal is the creation of exciton population by phonon absorption , illustrating the increased probability of the competing absorption processes discussed in section [ sec : mechanism ] . as such , the reduced asymmetry in the phonon sidebands
can be considered equivalent to the reduced difference between the probability of phonon emission and absorption .
the consequence of this competition is that for driving that is sufficiently strong to allow complete relaxation ( i.e. in the plateau region of fig .
[ fig : detuning](b ) ) , the final occupancy of the exciton state is in fact lower at higher temperatures owing to the thermalisation of the dressed states .
this is illustrated by the reduced maximum exciton population attained at higher bath temperatures in fig .
[ fig : detuning](b ) . in summary ,
a higher bath temperature will lead to a faster rise with pulse area , but a lower maximum population transfer by dvc .
as such , increasing the bath temperature may provide a means to enhance dvc processes when the driving is relatively weak as in fig .
[ fig : detuning](a ) .
the three first experimental observations of dvc in quantum dots @xcite focused on the la - phonon - assisted excitation process [ see fig . [
fig : relaxation](a ) , ( b ) ] .
recently liu .
et al . have demonstrated the reverse process - la - phonon - assisted de - excitation [ see fig . [
fig : relaxation ] ] - by observing the reduction of the exciton population of an excited qd by a red - shifted laser pulse @xcite .
a combination of the la - phonon - assisted excitation and de - excitation processes allows us to create and destroy an exciton within picosecond time scales , enabling ultrafast incoherent optical switching with a single qd much faster than the exciton radiative lifetime ( @xmath4ns @xcite ) . to demonstrate this ,
a two - color pump - probe experiment is performed with a blue - shifted ( excitation ) and red - shifted ( de - excitation ) pulse with the same pulse area .
the pulses are equally detuned ( @xmath68 mev ) from the exciton . at negative delay times
, the excitation pulse creates a certain exciton population corresponding to the background pc level .
the de - excitation pulse arrives before excitons are created , and therefore does nothing .
when the two pulses overlap , the pc signal decreases to a minimum within @xmath69 ps , as shown in fig .
[ fig : push - pull ] .
this occurs due to erasure of the exciton created by the excitation pulse by the de - excitation pulse .
the pc signal then slowly recovers due to the electron tunnelling out from the qd before the de - excitation pulse arrives @xcite .
the increase of the pc signal at @xmath70 ps is most probably related to the phonon - assisted excitation of biexciton by a very small portion of imperfectly circularly polarized light in the excitation / deexcitation pulses .
[ fig : push - pull](b ) shows the calculated exciton population with similar parameters as fig .
[ fig : push - pull](a ) .
the phase relation between the two pulses has been assumed to be random and we show the average of multiple repetitions of the simulation .
the numerically complete path - integral method takes into account the electron tunnelling in the pc experiment by incorporating it as an lindblad - type relaxation term @xcite . before the arrival of the pulses the qd
is assumed to be in the ground - state and the phonon modes follow a thermal distribution .
this model well reproduces the deexcitation of the qd and the subsequent recovery of the photo current signal .
the oscillation of the exciton population at @xmath45 ps is not clearly visible in the measured data in fig .
[ fig : push - pull](a ) due to the limited time resolution of the experiment . , where @xmath71 and @xmath72 are the arrival time of the de - excitation and excitation pulses , respectively
. @xmath18 = @xmath730.7 mev .
@xmath75 mev .
@xmath76 ps .
inset : pulse sequence .
( b ) exciton population detected in the two - color pump - probe measurement calculated using similar parameters as fig .
[ fig : push - pull](a ) .
@xmath18 = @xmath730.7 mev .
@xmath74 . @xmath77 ps . ]
the @xmath8 ps excitation / de - excitation time in our experiment is determined by the laser pulse duration .
the ultimate limit , , is set by the phonon thermalisation time ( few ps @xcite ) .
we note that in contrast to the optical switching scheme employing resonant coherent excitation @xcite , our scheme is robust against the fluctuation of laser power and detuning @xcite .
furthermore , the incoherent nature of the phonon - assisted relaxation process determines that no phase locking is needed , in contrast to fully coherent excitation schemes .
in summary , we have investigated the dynamic vibronic coupling in a single ingaas qd .
the dvc enables the population inversion and ultrfast depopulation of a qd by incoherent quasi - resonant excitation with the assistant of phonons @xcite .
we prove that the dvc is an incoherent process by performing ramsey - like interference measurements .
the temperature dependence study shows that at high temperature the phonon - assisted absorption induced by both the blue- and red - shifted laser pulses are enhanced and the strong asymmetry of the phonon sidebands observed at low temperatures disappears .
furthermore we demonstrate that it is possible to create and destroy an exciton within a pulse - width limited ps time scale by combining a blue - detuned and a red - detuned pulses , opening the door to ultrafast incoherent optical switching with a single qd @xcite .
the dvc may also be used to reduce the timing jitter of single / entangled photon sources @xcite or make tunable single qd lasers @xcite . compared with resonant coherent excitation , one factor which limits the potential applications of dvc is the high laser power required to achieve an efficient phonon - assisted excitation / deexcitation .
this problem may be solved by incorporating the qd into a nanocavity with a small mode volume where the light matter interaction is strongly enhanced @xcite .
when cavities are involved , the knowledge of phonon interactions in the strong qd - phonon coupling regime becomes an essential requirement for a full understanding of the behaviour of optically driven qds .
various recent studies have shown new physical phenomena that arise from this coupling @xcite .
this work was funded by the epsrc ( uk ) ep / j007544/1 . m. g. , a. m. b. , and v. m. a. gratefully acknowledge the financial support from deutsche forschungsgemeinschaft via the project no .
ax 17/7 - 1 .
the authors thank h.y .
liu and m. hopkinson for sample growth .
romana schirhagl , kevin chang , michael loretz , and christian l. degen , `` nitrogen - vacancy centers in diamond : nanoscale sensors for physics and biology '' , annual review of physical chemistry * 65 * , 83 - 105 ( 2014 )
. i. favero , g. cassabois , r. ferreira , d. darson , c. voisin , j. tignon , c. delalande , g. bastard , and ph .
roussignol , `` acoustic phonon sidebands in the emission line of single inasgaas quantum dots '' , physical review b * 68 * , 233301 ( 2003 ) s.l .
portalupi , g. hornecker , v. giesz , t .
grange , a. lematre , j. demory , i. sagnes , n.d .
lanzillotti - kimura , l. lanco , , a. auffves , and p. senellart , `` bright phonon - tuned single - photon source '' , nano letters , * 15 * , 629062945 ( 2015 ) .
p. g. eliseev , h. li , a. stintz , g. t. liu , t. c. newell , k. j. malloy and l. f. lester , `` transition dipole moment of inas / ingaas quantum dots from experiments on ultralow - threshold laser diodes '' , applied physics letters * 77 * , 262 ( 2000 ) .
a. j. ramsay , t.m .
godden , s. j. boyle , e.m .
gauger , a. nazir , b.w .
lovett , a.m. fox , and m.s .
skolnick , `` phonon - induced rabi - frequency renormalization of optically driven single ingaas / gaas quantum dots '' , phys .
lett . * 105 * , 177402 ( 2010 ) t. m. godden , j. h. quilter , a. j. ramsay , y. wu , p. brereton , s. j. boyle , i. j. luxmoore , j. puebla - nunez , a. m. fox and m. s. skolnick , `` coherent optical control of the spin of a single hole in an inas / gaas quantum dot '' , phys . rev . lett .
* 108 * , 017402 ( 2012 ) p .-
ardelt , l. hanschke , k. a. fischer , k. mller , a. kleinkauf , m. koller , a. bechtold , t. simmet , j. wierzbowski , h. riedl , g. abstreiter , and j. j. finley , `` dissipative preparation of the exciton and biexciton in self - assembled quantum dots on picosecond time scales '' , phys .
b 90 , 241404(r ) ( 2014 ) .
s. ates , s. m. ulrich , a. ulhaq , s. reitzenstein , a. lffler , s. hfling , a. forchel and p. michler , `` non - resonant dot - cavity coupling and its potential for resonant single - quantum - dot - spectroscopy '' , nature photonics * 3 * , 724 - 728 ( 2009 ) s. weiler , a. ulhaq , s. m. ulrich , d. richter , m. jetter , p. michler , c. roy and s. hughes , `` phonon - assisted incoherent excitation of a quantum dot and its emission properties '' , phys .
b * 86 * , 241304(r ) ( 2012 )
. k. h. madsen , s. ates , j. liu , a. javadi , s. m. albrecht , i. yeo , s. stobbe , and p. lodahl , `` efficient out - coupling of high - purity single photons from a coherent quantum dot in a photonic - crystal cavity '' , phys .
b * 90 * , 155303 ( 2014 ) j. hansom , c. h. h. schulte , c. matthiesen , m. j. stanley and m. atatre , `` frequency stabilization of the zero - phonon line of a quantum dot via phonon - assisted active feedback '' , appl .
. lett . * 105 * , 172107 ( 2014 ) .
p. borri , w. langbein , u. woggon , v. stavarache , d. reuter , and a. wieck , `` exciton dephasing via phonon interactions in inas quantum dots : dependence on quantum confinement '' , phys .
b * 71 * , 115328 ( 2005 ) .
m. glssl , a. m. barth , and v. m. axt , `` proposed robust and high - fidelity preparation of excitons and biexcitons in semiconductor quantum dots making active use of phonons '' , phys .
* 110*,147401 ( 2013 ) .
quilter , a.j .
brash , f. liu , m. glssl , a.m. barth , v.m .
axt , a.j .
ramsay , m.s . skolnick and a.m. fox , `` phonon - assisted population inversion of a single ingaas / gaas quantum dot by pulsed laser excitation '' , phys .
* 114 * , 137401 ( 2015 ) .
s. bounouar , m. mller , a. m. barth , m. glssl , v. m. axt and p. michler , `` phonon - assisted robust and deterministic two - photon biexciton preparation in a quantum dot '' , phys .
b * 91 * , 161302 ( 2015 ) .
f. liu , l.m.p .
martins , a.j .
brash , a.m. barth , j.h .
quilter , v. m. axt , m.s .
skolnick , and a.m. fox `` ultrafast depopulation of a quantum dot by la - phonon - assisted stimulated emission '' , phys .
b * 93 * , 161407(r ) ( 2016 )
. a. j. ramsay , a. v. gopal , e. m. gauger , a. nazir , b.w .
lovett , a. m. fox , and m. s. skolnick , `` damping of exciton rabi rotations by acoustic phonons in optically excited ingaas / gaas quantum dots '' , phys .
* 104 * , 017402 ( 2010 ) .
l. monniello , c. tonin , r. hostein , a. lemaitre , a. martinez , v. voliotis , and r. grousson , `` excitation - induced dephasing in a resonantly driven inas / gaas quantum dot '' , phys .
* 111 * , 026403 ( 2013 ) .
ramsay , a. j. , godden , t. m. , boyle , s. j. , gauger , a. nazir , b.w .
lovett , a.v .
gopal , a.m. fox , and m. s. skolnick , `` effect of detuning on the phonon induced dephasing of optically driven ingaas / gaas quantum dots '' , journal of applied physics , * 109 * , 102415 ( 2011 ) y.j .
wei , y.m .
chen , y.n .
hu , y. he , d. wu , c. schneider , m. kamp , s. hfling , c.y .
lu and j .- w .
pan,``deterministic and robust generation of single photons from a single quantum dot with 99.5@xmath78 indistinguishably using adiabatic rapid passage '' , nano lett .
* 14 * , 6515 ( 2014 ) .
r. mathew , e. dilcher , a. gamouras , a. ramachandran , h. y. shi yang , s. freisem , d. deppe , and k. c. hall , `` subpicosecond adiabatic rapid passage on a single semiconductor quantum dot : phonon - mediated dephasing in the strong - driving regime '' , phys .
b * 90 * , 035316 ( 2014 ) .
s. j. boyle , a. j. ramsay , a. m. fox , m. s. skolnick , a. p. heberle and m. hopkinson , `` beating of exciton dressed states in a single semiconductor ingaas / gaas quantum dot '' , phys .
. lett . * 102 * , 207401 ( 2009 ) .
s. j. boyle , a. j. ramsay , f. bello , h. y. liu , m. hopkinson , a. m. fox and m. s. skolnick , `` two - qubit conditional quantum - logic operation in a single self - assembled quantum dot '' , phys .
b * 78 * , 075301 ( 2008 ) . a. j. brash , l. m. p. p.
martins , f. liu , j. h. quilter , a. j. ramsay , m. s. skolnick , and a. m. fox , `` high - fidelity initialization of long - lived quantum dot hole spin qubits by reduced fine - structure splitting '' , physical review b * 92 * , 121301 ( 2015 ) .
kolodka , r. s. , ramsay , a. j. , skiba - szymanska , j. , fry , p. w. , liu , h. y. , fox , a. m. , and skolnick , m. s. , `` inversion recovery of single quantum - dot exciton based qubit '' , phys .
b , * 75 * , 193306(r ) ( 2007 ) p.a .
dalgarno , j.m .
smith , j. mcfarlane , b.d .
gerardot , k. karrai , a. badolato , p.m. petroff , and r.j .
warburton , `` coulomb interactions in single charged self - assembled quantum dots : radiative lifetime and recombination energy '' , phys
b , * 77 * , 245311 ( 2008 ) d. heinze , d. breddermann , a. zrenner , and s. schumacher , `` a quantum dot single - photon source with on - the - fly all - optical polarization control and timed emission '' , nature communications * 6 * , 8473 ( 2015 ) .
dara p. s. mccutcheon and ahsan nazir , `` model of the optical emission of a driven semiconductor quantum dot : phonon - enhanced coherent scattering and off - resonant sideband narrowing '' , phys .
lett . * 110 * , 217401 , ( 2013 ) .
yu - jia wei , yu he , yu - ming he , chao - yang lu , jian - wei pan , christian schneider , martin kamp , sven hfling , dara p. s.mccutcheon , and ahsan nazir , `` temperature - dependent mollow triplet spectra from a single quantum dot : rabi frequency renormalization and sideband linewidth insensitivity '' , phys .
lett . * 113 * , 097401 , ( 2014 ) .
kai mller , kevin a. fischer , armand rundquist , constantin dory , konstantinos g. lagoudakis , tomas sarmiento , yousif a. kelaita , victoria borish , and jelena vukovi . , `` ultrafast polariton - phonon dynamics of strongly coupled quantum dot - nanocavity systems '' , phys .
x * 5 * 031006 , ( 2015 ) .
f. hargart , m. mller , k. roy - choudhury , s. l. portalupi , c.schneider , s. hfling , m. kamp , s. hughes , and p. michler , `` cavity - enhanced simultaneous dressing of quantum dot exciton and biexciton states '' phys .
b * 93 * , 115308 , ( 2016 ) . | the electron - phonon coupling in self - assembled ingaas quantum dots is relatively weak at low light intensities , which means that the zero - phonon line in emission is strong compared to the phonon sideband .
however , the coupling to acoustic phonons can be dynamically enhanced in the presence of an intense optical pulse tuned within the phonon sideband .
recent experiments have shown that this dynamic vibronic coupling can enable population inversion to be achieved when pumping with a blue - shifted laser and for rapid de - excitation of an inverted state with red detuning . in this paper
we confirm the incoherent nature of the phonon - assisted pumping process and explore the temperature dependence of the mechanism .
we also show that a combination of blue- and red - shifted pulses can create and destroy an exciton within a timescale @xmath0ps determined by the pulse duration and ultimately limited by the phonon thermalisation time . |
after bourbaki [ 1989 11 ] we use _ cogebra , bigebra _ and hopf _ gebra _ instead of coalgebra , bialgebra and hopf algebra .
let @xmath5 be an @xmath6-space and @xmath7 be an @xmath6-dual @xmath6-space .
if @xmath5 is an @xmath6-cogebra and @xmath8 is an @xmath6-algebra then the @xmath6-space @xmath9 inherits the structure of an @xmath6-algebra with a convolution product : this is a convolution @xmath6-algebra , an @xmath6-convolution for short .
a dual @xmath6-space @xmath10 inherits a structure of an @xmath6-cogebra with coconvolution coproduct : this is a coconvolutional @xmath6-cogebra , an @xmath6-coconvolution for short .
in particular if an @xmath6-space @xmath11 carries an @xmath6-biconvolution algebra & cogebra structure , then do also the @xmath6-spaces @xmath12 as well as all iterated @xmath6-spaces @xmath13 inherit also @xmath6-biconvolution algebra & cogebra structures .
if the @xmath6-space @xmath5 is an @xmath6-cogebra with a coproduct @xmath14 then @xmath7 is an @xmath6-algebra with product @xmath15 if an @xmath6-space @xmath8 is a finite dimensional @xmath6-algebra having a binary product @xmath16 then an @xmath6-dual @xmath6-space @xmath17 ( or @xmath18-graded dual in the case @xmath8 is not a finite dimensional @xmath6-space ) is an @xmath6-cogebra with a binary coproduct @xmath19 however there are several important situations ( free tensor algebra , exterior algebra , clifford algebra , weyl algebra , ... ) where the dual space of an algebra is also an algebra in a natural way by construction . if this is the case , then by the above ( @xmath18-graded ) duality , both mutually dual @xmath6-spaces carry both structures , algebra & cogebra , and therefore we have a dual pair of @xmath6-biconvolutions . an unital and associative convolution possessing an ( unique ) antipode
is said to be a hopf gebra or an _ antipodal convolution _
( definition [ dfn2.2 ] ) .
this terminology has been introduced in [ oziewicz 1997 , 2001 ; cruz & oziewicz 2000 ] and is different from sweedler s [ 1969 p. 71
] . a general theory of the finite - dimensional antipodal and antipode - less biconvolutions ( convolutions and coconvolutions ) has been initiated in [ cruz & oziewicz 2000 ] .
nill in 1994 and bhm & szlachnyi in 1996 introduced weak bigebras and weak hopf gebras with antipode @xmath20 defined as the galois connection with respect to the binary convolution @xmath21 which does not necessarily needs to be unital [ nill 1998 , nill et all .
1998 ] , @xmath22 if @xmath23 is invertible then @xmath24 if @xmath25 is a clifford @xmath6-algebra , then a dual @xmath6-space @xmath26 is a clifford @xmath6-cogebra .
it was shown in [ oziewicz 1997 , and ff . ]
that a clifford convolution @xmath27 is antipode - less .
the aim of this paper is to show that a clifford convolution @xmath28 for @xmath29 _
i.e._for @xmath30 does posses an antipode iff @xmath31 ( main theorem 4.1 ) .
applications to physics have been proposed in [ fauser 2000b ] .
] ] let the convolution be unital , fig .
[ conv_unit ] .
this is the case if an @xmath6-algebra @xmath8 is unital with unit @xmath32 and an @xmath6-cogebra @xmath5 is counital with counit @xmath33 [ dfn2.1 ] the convolutive inverse , w.r.t .
the convolutive unit @xmath34 , of the identity map on @xmath11 , fig .
[ antipode ] , is said to be an _ antipode _ , @xmath35 .
if an antipode exists w.r.t .
an unital associative convolution it must be unique .
an antipodal biconvolution defines a unique crossing as given in fig .
[ crossing ] in the last section , see [ oziewicz 1997 , 2001 ; cruz & oziewicz 2000 ] . using the axioms of the antipode , fig .
[ antipode ] , and biassociativity , this crossing is equivalent to the algebra homomorphism between algebra and crossed algebra , as well as , to the cogebra homomorphism from crossed cogebra to cogebra . a proof is given in [ cruz & oziewicz 2000 ] .
the gramann wedge product , the gramann coproduct and the unique antipode , as given by sweedler [ 1969 , ch .
xii ] and extended by woronowicz [ 1989 ,
3 ] , see section 3.1 below for details , gives the gramann hopf gebra . this motivates the following definition : [ dfn2.2 ] an unital and associative convolution possessing a ( unique ) antipode is said to be a hopf gebra or an antipodal convolution .
a closed structure is given by the evaluation and coevaluation [ kelly & laplaza 1980 ; lyubashenko 1995 ] : @xmath36 following [ lyubashenko 1995 , fig .
3 , p. 250 ]
the evaluation is represented by cup and coevaluation by cap .
the evaluation intertwines the transposed endomorphisms @xmath37 with @xmath38 , fig .
[ dual_op ] . up and down arrows indicate the spaces @xmath39 and @xmath11 . . ]
a coconvolution is counital if e.g. there exits a dual counit @xmath40 and a dual unit @xmath41 , fig .
[ conv_counit ] . ] ] we have to use the dual coconvolution counit @xmath42 in fig .
[ conv_counit ] .
the gramann hopf gebra was constructed by sweedler [ 1969 , ch .
xii ] factoring the couniversal shuffle tensor hopf gebra sh@xmath11 by the _ switch _ , @xmath43 .
this construction was generalized to any braid by woronowicz [ 1989 , 3 ,
an exterior hopf gebra can be defined in terms of an unique braid dependent homomorphism of universal tensor hopf gebra into couniversal tensor hopf gebra and this implies that an exterior hopf gebra is couniversal and braided [ oziewicz , paal & raski 1995 8 ; raski 1996 ; oziewicz 1997 p. 1272 - 1273 ] . ]
the tensors @xmath44 are said to be scalar products on @xmath11 or coscalar products on @xmath39 ( @xmath45 is the transpose of @xmath46 ) .
the tensors @xmath47 are said to be scalar products on @xmath39 and coscalar on @xmath11 . in particular @xmath48 is the symmetric part of @xmath49 the scalar products are displayed by decorated ( or labelled ) cups and coscalar products by decorated caps , see fig .
[ cliffordization ] .
rota and stein [ 1994 ] introduced the clifford product as a deformation of exterior biconvolution .
this deformation process was called _
a cliffordization introduces an internal loop in a binary product having two inputs and one output , employing the @xmath50-cup ( or @xmath51-cap for the coproduct ) , fig .
[ cliffordization ] .
clifford biconvolution was defined in [ oziewicz 2001 ] as the @xmath52-bicliffordization of an exterior biconvolution . in sweedler
s notation , @xmath53 if the product coproduct duality of fig .
[ duality ] is used with cup as an evaluation , then every product on @xmath54 induces a coproduct on @xmath55 and vice versa . if @xmath50-cup s and @xmath56-cap s are used , one gets a correlation between products on @xmath54 and coproducts on @xmath54 .
a clifford @xmath6-algebra together with a clifford @xmath6-cogebra on @xmath57 @xmath58 is said to be a clifford @xmath6-convolution .
it was shown in [ oziewicz 1997 ] that a clifford @xmath6-convolution for @xmath59 and for @xmath60 is antipode - less .
an antipode - less clifford convolution @xmath61 for an _ invertible _
tensor @xmath46 _ can not _ be a deformation of an exterior gramann hopf gebra .
the tensors @xmath46 and @xmath56 are said to be dependent if @xmath62 and @xmath63 exist such that one of the following relations hold , @xmath64 if the tensors @xmath65 are independent then the clifford product and coproduct are defined independently by rota & stein s _
deformation_. cliffordization and cocliffordization of the gramann convolution does not change the convolutive unit @xmath66 however since the @xmath18-grading is changed due to the skewsymmetric parts of @xmath46 and @xmath67 the counit is no longer the projection onto @xmath68 [ fauser 1998 - 2000a ] .
[ mnthm ] a clifford biconvolution @xmath69 is a clifford hopf gebra iff @xmath70 then @xmath71.\end{aligned}\ ] ] theorem 4.1 was proved for @xmath72 in [ oziewicz 1997 ] . if @xmath73 then for @xmath74 @xmath75 therefore @xmath76 the clifford antipode @xmath77 is computed from fig .
[ antipode ] .
we give the proof for @xmath78 only .
the general case will be treated elsewhere .
let @xmath79 and @xmath80 then we find together with ( i ) of the main theorem : @xmath81 an action of @xmath82 on tensors @xmath83 and @xmath84 is given as follows @xmath85 it would be desirable to study @xmath86-orbits on @xmath87 and present full classification of all orbits in terms of invariants .
however , this topic exceeds the scope of this paper and will be presented elsewhere .
theorem [ mnthm ] solves and improves conjecture 2.2 posed in [ oziewicz 1997 , p. 1270
the above clifford hopf gebra includes as a particular case for @xmath88 the construction made by urevich [ 1994 ] . in the case
@xmath89 one can take @xmath90 . also if @xmath91 one can choose @xmath92 . in order to prevent such possibilities we need to supplement the definition of the clifford hopf gebra with the extra conditions : @xmath93
the antisymmetric tensor @xmath94 can be adjusted in such a way that @xmath95 with invertible symmetric tensor @xmath96 we found a clifford antipode for @xmath97 @xmath98 the following matrices for @xmath99 represent the same tensor from @xmath100 or from @xmath101 with respect to the different bases , these matrices are on the same @xmath102 orbit , @xmath103 an antipode for regular scalar and coscalar tensors can be found also .
a clifford biconvolution @xmath69 is antipode - less if @xmath104 . in particular
this is the case if @xmath105 it was shown [ oziewicz 1997 ] that @xmath106 .
what axioms for clifford biconvolution implies the condition @xmath104 ?
in particular , does a braid exists for which such a clifford biconvolution is a braided hopf gebra in the usual sense ?
if such a braid exits how much freedom remains for choices ?
compare with [ oziewicz 1997 , p. 1272
] where it was shown that for @xmath72 , @xmath107 exists a 12-parameter family of crossings .
let @xmath108 and @xmath109 then the following equations are equivalent : @xmath110 @xmath111 according to lemma 5.2 we have to ask that @xmath112 we present three examples of antipode - less clifford biconvolutions @xmath69 , @xmath104 for @xmath78 , @xmath113 and for @xmath65 given by with signature @xmath114 @xmath115 @xmath116
let @xmath117 be a commutative ring . an exact sequence of homomorphisms of @xmath117-modules , @xmath118 @xmath119 splits if @xmath120 is a direct summand of @xmath121 the following statements are equivalent @xmath122 let @xmath123 split the exact sequence of @xmath6-algebra homomorphisms @xmath124 thus @xmath125 in this case the element of the crossed @xmath6-algebra , viz .
@xmath126 , is said to be a _ splitting idempotent _ ( a cleft of @xmath127 ) , @xmath128 . if @xmath129 then @xmath130 a crossing defined in fig .
[ crossing ] gives a cogebra map @xmath131 and an algebra homomorphism @xmath132 [ oziewicz 1997 , 2001 , cruz & oziewicz 2000 ] .
however an algebra homomorphism @xmath133 in general does not need to split .
a clifford convolution @xmath134 is said to be a gramann convolution .
if @xmath69 is a gramann hopf gebra , @xmath135 then @xmath133 splits , @xmath136 iff @xmath137
a crossing for an antipodal convolution is defined on fig .
[ crossing ] . for antipodal convolution ] the crossing @xmath138 fig .
[ crossing ] is equivalent that there is a cogebra map @xmath139 and an algebra homomorphism @xmath140 [ oziewicz 1997 , 2001 , cruz & oziewicz 2000 ] .
a crossing for @xmath141 and thus for @xmath142 is the involutive graded switch [ sweedler 1969 , ch.xii ] , @xmath143 in the sequel @xmath70 the degree of the minimal polynomial of the crossing @xmath144 we denote by : degree@xmath145 let @xmath146 then @xmath147 let @xmath18 be the ring of integers , @xmath148 $ ] be the ring of polynomials in @xmath149 with coefficients in @xmath150 and @xmath151 $ ] be the following polynomial , @xmath152 let @xmath69 be a gramann hopf gebra , _ i.e._an antipodal gramann convolution ( definitions 2.3 & 6.4 ) with @xmath153 and @xmath78 .
the minimal polynomial of the crossing is of order 30 in @xmath154 if @xmath155
then the minimal polynomial of @xmath138 vanishes , as also no antipode exists in this case . moreover @xmath156,\\ & p(\sqrt{\mu}-1)\neq 0,\ ; \text{then the minimal polynomial of the crossing}\\ & \text{is of degree 3.}\\ ( ii)&\quad\text{if}\;p(\sqrt{\mu}-1)=0,\ ; \text{then the the minimal polynomial of the crossing is}\\ & \text{of degree}\;\leq 2.\end{aligned}\ ] ]
clifford , a maple v package by rafa abamowicz [ abamowicz , 1995 - 2000 ] and extensions like bigebra , by rafa abamowicz and bertfried fauser , have been used to check and get some of the results .
99 rafa abamowicz , clifford , an maple v package and bigebra , available at http://math.tntech.edu/rafal/ nicolas bourbaki , _ elements of mathematics , algebra i _ , chapter iii , springer verlag , berlin 1989 henry cartan & samuel eilenberg , _ homological algebra _ , princeton university press 1956 jos de jess de cruz guzman and zbigniew oziewicz , _ unital and antipodal biconvolution and hopf gebra _ , ( 2000 ) this volume micho urevich , _ braided clifford algebras as braided quantum groups _ , advances in applied clifford algebras * 4 * ( 2 ) ( 1994 ) 145156 micho urevich , _ braided clifford algebras as quantum deformations _ , international journal of theoretical physics * 40 * ( 1 ) ( 2001 ) samuel eilenberg , _ extensions of general algebra _ , annales soc .
polon . math .
( rocznik polskiego towarzystwa matematycznego ) * 21 * ( 1 ) ( 1948 ) 125134 bertfried fauser , _ on an easy transition from operator dynamics to generating functionals by clifford algebras _ ,
journal of mathematical physics * 39 * ( 1998 ) 49284947 bertfried fauser and rafa abamowicz , _ on the decomposition of clifford algebras of arbitrary bilinear form _ , in : `` clifford algebras and their applications in mathematical physics '' , rafa abamowicz and bertfried fauser , eds . , birkhuser , boston 2000 p. 341366
bertfried fauser , _ on the hopf algebraic origin of wick normal - ordering _ , 2000a , j. phys .
a. : math . and gen .
accepted ( hep - th/0007032 ) bertfried fauser , _ quantum clifford hopf gebras for quantum field theory _ , talk at `` quantum group symmposium '' at group23 , dubna , 2000b , ( hep - th/0011026 ) alexander j. hahn , _
quadratic algebras , clifford algebras , and arithmetic witt groups _ , springer verlag , berlin 1994 , chap .
2 gregory maxwell kelly & miguel l. laplaza , _ coherence for compact closed categories _ , journal of pure and applied algebra * 19 * ( 1980 ) 193213 volodimir lyubashenko , _ tangles and hopf algebras in braided categories _ , journal of pure and applied algebra * 98 * ( 1995 ) 245278 florian nill , _ axioms for weak bialgebras _ , q - alg/9805104 florian nill , k. szlachnyi & h .- w .
wiesbrock , _ weak hopf algebras and reducible jones inclusion of depth 2 .
i : from crossed products to jones towers _ , q - alg/9806130 zbigniew oziewicz , eugen paal and jerzy raski , _ derivations in braided geometry _ , acta physica polonica * 26 * ( 7 ) ( 1995 ) 12531273 zbigniew oziewicz , _
clifford hopf gebra and bi - universal hopf gebra _ , czechoslovak journal of physics * 47 * ( 12 ) ( 1997 ) 12671274 zbigniew oziewicz , _ guest editor s note : clifford algebras and their applications _
, international journal of theoretical physics * 40 * ( 1 ) ( 2001 ) 111 richard s. pierce , _ associative algebras _
, graduate texts in mathematics # 88 , springer - verlag , new york 1982 gian - carlo rota & joel a. stein , _ plethystic hopf algebras _ , proceedings of natl .
academy of sciences usa * 91 * ( 1994 ) 1305713061 jerzy raski , _ braided antisymmetrizer as bialgebra homomorphism _ , reports in mathematical physics * 38 * ( 2 ) ( 1996 ) 273277 gnter scheja , uwe storch , _ lehrbuch der algebra _ , teil 1 , b.g
. teubner , stuttgart , 1980 moss e. sweedler , _ hopf algebras _ , benjamin , new york 1969 stanisaw lech woronowicz , _ differential calculus on compact matrix pseudogroups ( quantum groups ) _ , communications in mathematical physics * 122 * ( 1989 ) 125170 stanisaw zakrzewski , _ quantum and classical pseudogroups .
part i. union pseudogroups and their quantization _ , communications in mathematical physics * 134 * ( 1990 ) 347370 | a clifford algebra @xmath0 jointly with a clifford cogebra @xmath1 is said to be a clifford biconvolution @xmath2 we show that a clifford biconvolution for @xmath3 does possess an antipode iff @xmath4 an antipodal clifford biconvolution is said to be a clifford hopf gebra . + * 2000 mathematics subject classification : * 15a66 clifford algebra , 16w30 coalgebra , bialgebra , hopf algebra + * 2000 pacs : * 02.10.tq associative rings and algebras + * keywords : * cliffordization , clifford algebra , clifford cogebra , antipode , hopf gebra , clifford bigebra , gramann algebra |
the ordering of anisotropic particles is a universal phenomenon appearing widely in nature , ranging from thermally driven molecules or colloids @xcite to active particles such as bacteria colonies @xcite , actin filaments @xcite , animal groups @xcite , and living liquid crystals @xcite . in equilibrium lyotropic systems , such as hard rods interacting only through excluded volume interactions ,
the transition of sufficiently anisotropic particles into various ordered states is entropy driven .
the loss in rotational degrees of freedom in the ordered state is compensated by the gain in the translational ones @xcite .
taking a two - dimensional system of hard rectangles as an example , a tetratic state with four - fold rotational symmetry was discovered in monte carlo ( mc ) simulations @xcite , and studied theoretically with density functional theory ( dft ) @xcite .
the number density and the length - to - width ratio ( aspect ratio ) of the particles were found to be the key parameters determining the ordered states of hard rectangles with only excluded volume interactions @xcite . given the ubiquity of ordering transitions in nature , it is important to ask how well the existing knowledge about such transitions in equilibrium ( thermal ) systems can be extended to nonequilibrium ( athermal ) systems . due to the dissipative interactions between particles ,
agitated granular matter has been frequently used as a nonequilibrium model system for phase transitions @xcite .
rich and often counterintuitive dynamical behavior @xcite has been discovered for granular rods , including vortex patterns @xcite , collective swirling motions @xcite , giant number fluctuations @xcite , violation of the equipartition theorem @xcite , and an enhanced ordering transition in an effective ` thermal ' bath of spherical particles @xcite .
reminiscent to equilibrium systems , ordering transitions of vertically agitated granular rods were investigated in three - dimensional ( 3d ) and quasi - two - dimensional systems . in 3d , the aspect ratio of the rods was found to influence the ordered states of cylindrical rods @xcite . in quasi - two - dimensional systems , a bulk isotropic - uniaxial nematic ( i - u ) transition
was observed for cylindrical rods with large aspect ratios @xcite and an effective elastic constant was characterized quantitatively @xcite . particularly in strict monolayer systems , the shape of the rods
was found to play an important role in determining the ordered states : tetratic , nematic or smectic order was found for cylindrical rods , tapered rods or rice particles , respectively @xcite .
moreover , tetratic order was also found for tubular shaped particles and the influence of the container shape was discussed in @xcite . despite all of these progresses
, it is still unclear to which extent one can draw quantitative connections between systems in and out of thermodynamic equilibrium .
more specifically , a quantitative comparison between the state diagram of dissipative granular rods and that of the corresponding equilibrium system is still lacking . this quantitative comparison is the purpose of the present work .
here we investigate experimentally the orientational order in monolayers of cylindrical granular rods driven far from thermodynamic equilibrium , and compare the results to mc simulations as well as dft of the analogous equilibrium system .
focusing on the bulk region of the system , we detect both tetratic and uniaxial nematic states by varying the aspect ratio of the rods .
we demonstrate that the aspect ratio and the number density of rods are the key parameters determining the state diagram in both systems .
we find a common aspect ratio that separates tetratic and uniaxial nematic states in both experiments and mc simulations .
such an agreement illustrates the universal aspects of the ordering of rod - shaped particles .
( color online ) sketch of the experimental set - up .
the closed cylindrical container of height @xmath1 and radius @xmath2 is driven sinusoidally against gravity with an electromagnetic shaker .
the rods have a length @xmath3 and a diameter @xmath4 .
the embedded image shows a close view of the detected particles . ] a sketch of the experimental set - up is shown in fig .
[ fig : setup ] .
monodisperse polyvinyl chloride ( pvc ) rods of diameter @xmath4 and length @xmath3 , cut from welding wires of @xmath5 mm ( aspect ratio @xmath6 ) or @xmath7 mm ( @xmath8 ) , are confined in a cylindrical container of height @xmath1 and radius @xmath9 cm .
the ratio @xmath10 is chosen for both diameters to ensure a monolayer of particles ; that is , no rods can cross or jump over each other .
the inner surface of the container is covered with antistatic spray ( kontakt chemie , antistatik 100 ) to minimize electrostatic forces .
an electromagnetic shaker ( tira tv50350 ) is employed to drive the sample sinusoidally against gravity with frequency @xmath11hz and peak acceleration @xmath12 , where @xmath13 is the peak vibrational amplitude and @xmath14 is the gravitational acceleration .
the acceleration is monitored with an accelerometer ( dytran 3035b2 ) .
we capture high contrast images of the rods using backlight led illumination and a camera ( idt motionscope m3 ) mounted above the container .
the camera is synchronized with the shaker so as to capture images at a fixed phase of each vibration cycle .
the images are subjected to an analysis algorithm that determines the center of mass @xmath15 and the orientation @xmath16 of the @xmath17-rod with @xmath18}$ ] .
@xmath19 is the angle of the main rod axis with respect to a fixed laboratory axis , and @xmath20 is the total number of rods in the container .
the detection rate is @xmath21 % for @xmath5 mm and @xmath22 % for @xmath23 mm . to systematically study the collective behavior of the rods , we vary the global area fraction @xmath24 between @xmath25 and @xmath26 , and the aspect ratio @xmath27 between @xmath28 and @xmath29 . for each @xmath30 and @xmath27 , we vary the peak acceleration @xmath31 with a step of @xmath32 from @xmath33 to @xmath34 and back . the waiting time between each step
is fixed at @xmath35 minutes .
we repeat the whole cycle at least @xmath36 times . correspondingly , we model the particles as two - dimensional hard rectangles of length @xmath3 and width @xmath4 that interact through excluded volume interactions .
@xmath20 of such particles are placed in a box with dimensions @xmath37 and @xmath38 along the @xmath39- and @xmath40-axes , respectively .
we use periodic boundary conditions along both axes and study the equilibrium bulk configurations by means of standard mc simulations @xcite in the canonical ensemble .
that is , we fix the number of particles @xmath20 and the system area @xmath41 ( the temperature is irrelevant in hard models ) .
the number of particles is similar to that in the experiments , @xmath42 .
we use simulation boxes with rectangular and square shapes .
no difference has been found between the two geometries .
the simulation method is as follows . in order to equilibrate the system we start at very high area fractions , @xmath43 .
we place the particles , with their main axes pointing in the same direction , in a rectangular lattice .
next we run @xmath44 monte carlo steps ( mcss ) .
each mcs is an attempt to move and rotate all the particles in the system .
the maximum displacement @xmath45 and maximum rotation @xmath46 that each particle is allowed to perform in a mcs is determined such that the acceptance probability is @xmath47 .
then we remove a few randomly chosen particles , recalculate @xmath45 and @xmath46 , and start a new simulation .
the number of removed particles is such that the change in area fraction is @xmath48 . in order to rule out metastable configurations related to the preparation of the initial state
, we discard simulations with @xmath49 .
when the area fraction is below that limit we start the proper simulation . for each simulation
we first run @xmath50 mcss to equilibrate the system and then accumulate data over @xmath44 mcss . for selected @xmath27
we have also simulated the system by increasing the number of particles , i.e. , by adding particles instead of removing them . we have found no differences between both methods .
we use an onsager - like dft with parsons - lee rescaling .
a similar dft was previously used to analyze the state diagram of two - dimensional rods confined in a circular cavity @xcite .
we are interested in the behavior of fluid states in which the density is spatially homogeneous .
hence we can write , without loss of generality , the one body density distribution as @xmath51 where @xmath52 is the number density and @xmath53 is the orientational distribution function . here
@xmath54 is the angle with respect to the director .
@xmath53 is normalized such that @xmath55 we split the free energy into two parts @xmath56=f_{\text{id}}[\rho]+f_{\text{ex}}[\rho],\ ] ] where @xmath57 is the ideal gas part and @xmath58 is the excess part accounting for the excluded volume interactions .
the ideal free energy per unit of area @xmath59 is given exactly by @xmath60}{a}=\int_0^\pi { \rm d } \gamma\rho h(\gamma)\ln(\lambda\rho h(\gamma)-1),\ ] ] where @xmath61 with @xmath62 the boltzmann s constant and @xmath63 the absolute temperature .
@xmath64 is the ( irrelevant ) thermal volume that we set to one .
the excess part is approximated by @xmath65}{a}=\frac{\psi_{\text{ex}}(\phi)}{4ld}\rho\int_0^\pi { \rm d } \gamma_1\int_0^\pi { \rm d } \gamma_2 h(\gamma_1)h(\gamma_2)v_{\text{exc}}(\gamma_{12}).\label{eqex}\ ] ] @xmath66 is the excluded area between two rectangles with relative orientation @xmath67 : @xmath68 and @xmath69 is the excess free energy per particle of a reference system of hard disks at the same area fraction as our system of hard rectangles .
the diameter of the disks is selected such that both disks and rectangles have the same area .
following baus and colot @xcite we approximate @xmath70 by : @xmath71 with @xmath72 .
( [ eqex ] ) recovers the onsager approximation in the low density limit .
finally , the grand potential is given by @xmath73/a = f[\rho]/a-\mu\rho,\ ] ] with @xmath74 the chemical potential .
we minimize @xmath75 with respect to @xmath52 and @xmath53 in order to find the equilibrium states .
we use a standard conjugated gradient method to minimize the functional .
we use a truncated fourier expansion to describe @xmath53 .
we truncate the expansion such that the absolute value of the last coefficient in the expansion is smaller than @xmath76 .
this section is organized as follows : we first introduce the ordered states observed in experiments and mc simulations in section [ sec : os ] . in section [ sec : exp ] , we analyze the influence of the container walls and the driving conditions in the experiments . finally in section [ sec :
cmp ] , we quantify the ordering transition threshold for various aspect ratios and compare the state diagrams obtained experimentally , via mc simulations and with dft .
( color online ) raw experimental images ( topview ) showing typical configurations of rods with aspect ratio @xmath77 ( a ) and @xmath78 ( b ) at high global area fractions .
the yellow ( light gray ) dashed circle indicates the region of interest . ]
figure [ fig : snapshot ] shows typical snapshots of the ordered states obtained experimentally .
short rods ( a ) tend to develop tetratic order with two alignment directions perpendicular to each other . long rods ( b ) form uniaxial nematic order with only one preferred alignment direction . in both cases , the container promotes either homeotropic ( perpendicular ) or planar ( parallel ) anchoring of the rods close to the boundary . to minimize the boundary effects ,
we consider only those particles located in the central region of the container , as marked in fig .
[ fig : snapshot ] . a quantitative justification of this region of interest ( roi ) will be given in section [ sec : exp ] . sometimes during the experiments , especially at low global area fractions , we observe regions with very low number density of rods ( almost empty regions ) .
as we are interested in the bulk behavior , we discard those configurations in which the empty regions " and the roi overlap .
( color online ) typical snapshots of tetratic ( left column , @xmath77 ) and uniaxial nematic ( right column , @xmath78 ) states .
( a ) and ( d ) are reconstructed from the positions and orientations of the particles detected in the center region of the container .
( b ) and ( e ) are from mc simulations with periodic boundary conditions .
the particles are color coded according to their orientations .
( c ) and ( f ) show the orientational distribution functions @xmath53 of the particles in experiments ( gray bars ) and simulations ( solid line ) .
@xmath54 is the angle with respect to the director . ]
figure [ fig : sa ] shows a direct comparison of the ordered states obtained in both experiments and mc simulations .
the color coded particle configurations are reconstructed from granular rods in the roi ( upper panels ) and from mc simulations ( middle panels ) with periodic boundary conditions . in the tetratic state with fourfold rotational symmetry ( left column ) ,
the orientational distribution function @xmath53 , where @xmath54 is the angle with respect to the director @xmath79 , has two peaks at @xmath80 and @xmath81 ( c ) .
in contrast , in the uniaxial nematic state ( right column ) , the elongated particles are oriented on average along the director , yielding only one peak at @xmath80 ( f ) .
the director @xmath79 is calculated as the normalized eigenvector of the largest eigenvalue of the tensorial order parameter @xmath82 . here
@xmath83 is the @xmath84th cartesian coordinate of the unit vector @xmath85 , @xmath86 is the kronecker delta , and @xmath87 denotes an average over the rods @xcite . to quantify the orientational order we measure @xmath88 where @xmath89 and @xmath90 are the uniaxial and tetratic order parameters , respectively . in an isotropic state ( no orientational order )
@xmath89 and @xmath90 vanish . in a uniaxial nematic state @xmath91 and @xmath92 . finally in a tetratic state @xmath93 and @xmath92 . the states in fig .
[ fig : sa ] are selected such that @xmath89 and @xmath90 are comparable in both experiments and mc simulations .
( color online ) wall - rod angular correlation function @xmath94 as a function of the rescaled distance @xmath95 to the wall for various @xmath27 .
a sketch with various definitions is shown in the inset .
the data are obtained through an average over all @xmath31 , global area fractions @xmath30 and experimental runs .
the typical error for @xmath94 ( @xmath96 ) is comparable to the size of the symbols for @xmath97 .
note the logarithmic scale on the y - axis . ]
experiments @xcite and mc simulations @xcite have shown that the container induces a preferential alignment of the particles close to the wall . in order to facilitate the investigation in the bulk , we first need to characterize such an influence quantitatively . following the ideas in @xcite ,
we calculate the wall - rod angular correlation function @xmath98\rangle$ ] , where @xmath99 is the shortest distance from the rod center to the container wall , the angle @xmath100 quantifies the tangential direction of the corresponding point on the wall ( see inset in fig .
[ fig : wallrod_correlation ] ) , and @xmath101 denotes an average over all the particles at a distance @xmath99 .
either homeotropic or planar alignment of the particles with respect to the wall results in @xmath102 . in fig .
[ fig : wallrod_correlation ] , @xmath103 is presented as a function of the rescaled distance to the wall @xmath95 with a binning width of @xmath104 .
for all aspect ratios investigated , @xmath103 decays exponentially with @xmath95 . to minimize the influence of the wall ,
we consider only those particles with @xmath105 to be in the roi . in this region
, @xmath103 is always smaller than @xmath106 and remains in a range comparable to the experimental uncertainties .
we characterize the state of the system by measuring the area fraction @xmath107 and @xmath53 in circular regions with radius @xmath108 inscribed in the roi .
subsequently , we calculate @xmath109 from @xmath53 accumulated over all the regions that share the same @xmath107 .
figure [ fig : varacc ] shows the order parameters as a function of @xmath107 for short rods with @xmath77 .
it indicates an area fraction @xmath110 above which the tetratic order parameter @xmath90 grows from its initial low value , while the uniaxial order parameter @xmath89 remains low .
such a combination of @xmath89 and @xmath90 suggests a gradual isotropic - tetratic ( i - t ) transition .
as shown in ( a ) , the behavior of @xmath89 and @xmath90 does not depend on the peak vibration acceleration .
this is further confirmed through a comparison among data obtained for all @xmath31 in the range of @xmath111 and also for all aspect ratios investigated . as shown in ( b ) , a variation of the vibration frequency @xmath112 from @xmath113 hz to @xmath114 hz for @xmath77 also yields the same behavior of @xmath109 .
such agreements indicate that the details of how the rods are effectively ` thermalized ' in our nonequilibrium system are not essential in determining the ordering transitions , providing us the opportunity to draw connections to the corresponding equilibrium systems .
accordingly , we accumulate the data over all @xmath31 at @xmath11 hz for a more accurate characterization of the transition threshold @xmath110 . by fitting @xmath90 with a constant value in the isotropic region and with a straight line in the ordered state ,
we obtain @xmath110 as the intersection point which minimizes the standard error . only data with sufficient statistics ( i.e.
, error bar @xmath115 ) and @xmath116 are chosen for the fits .
moreover , the height of the container is found to play a minor role in determining the ordering transition : a variation of @xmath117 from @xmath118 to @xmath33 leads to the same behavior of @xmath119 .
experiments with @xmath120 for @xmath77 and @xmath78 give rise to slightly lower transition thresholds @xmath110 .
more specifically , we find a decrease of @xmath121 for short rods and of @xmath122 for long rods , which is in both cases within the uncertainty of the fit .
in addition , for a specific aspect ratio of @xmath123 , the same experiments have been performed for two different rod diameters .
the results agree with each other within the error bar , suggesting that the mass of the rods does not play a dominating role in the ordering transition .
( color online ) tetratic and uniaxial order parameters @xmath124 as a function of the area fraction @xmath107 for ( a ) @xmath77 and ( b ) @xmath78 .
the insets show the corresponding results from mc simulations .
the experimental data is an accumulation over all @xmath31 at @xmath11 hz .
straight lines are linear fits to determine the threshold @xmath110 . ] based on the above characterizations of the boundary influence , we compare the ordering transitions of granular rods in the roi to the corresponding thermal system .
figure [ fig : order_bulk ] shows the averaged order parameters obtained in both experiments and mc simulations ( insets ) for rods with @xmath77 ( a ) and @xmath78 ( b ) . as discussed above , tetratic
ordering occurs in the system of short rods . for long rods ,
both order parameters start to grow above @xmath110 , suggesting a gradual i - u transition .
qualitatively , the agreement between experiments and mc simulations on the behavior of both tetratic @xmath90 and uniaxial @xmath89 order parameters is remarkable for both aspect ratios .
such similarities indicate that the ordering of granular rods is governed by the geometric constrain of non - overlapping rods , which is the only interaction considered in the simulations .
quantitatively , the threshold @xmath125 obtained experimentally for rods with @xmath77 agrees with the one @xmath126 obtained from mc simulations within the error .
however , the experimentally obtained threshold @xmath127 for rods with @xmath78 is larger than the one obtained for the corresponding thermal system , @xmath128 . as @xmath27 and @xmath107 are the key parameters determining the state of the system , we compare the experimental ( nonequilibrium ) results with the mc ( equilibrium ) simulations in a state diagram shown in fig .
[ fig : phasediag ] . in both systems
short rods form a tetratic state and long rods a uniaxial nematic state at sufficiently high area fractions .
the aspect ratio at which the ordered state changes from tetratic to uniaxial nematic agrees quantitatively .
it is found to be @xmath129 in both experiments and simulations ( uniaxial ) and @xmath130 ( tetratic ) . ] .
this result agrees with previous simulations in which a tetratic phase was found for @xmath131 and some evidence of uniaxial ordering for @xmath132 @xcite .
the quantitative agreement of @xmath133 across systems in and out of thermodynamic equilibrium illustrates the universal aspects of the ordering transitions . on the other hand ,
the threshold @xmath110 for agitated rods differs from that in mc simulations , indicating the non - universal aspects of the ordering transitions .
first , the experimentally determined @xmath110 exhibit a peak around @xmath133 .
in contrast , mc simulations show a monotonic decay with @xmath27 .
second , @xmath110 measured in experiments deviates systematically from that obtained via mc simulations as @xmath27 grows . for the largest aspect ratio investigated experimentally , @xmath134
, much higher area fraction is required for the uniaxial state to develop .
this difference might be attributed to the following mechanisms .
( i ) the strong fluctuations in the nonequilibrium steady states of granular rods may lead to temporal disorder in a system that could in principle relax into an ordered state .
( ii ) due to the dissipative rod - rod interactions , the tendency of clustering for granular rods is larger than in mc simulations , especially for large @xmath27 [ compare panels ( d ) and ( e ) of fig . [
fig : sa ] ] .
( iii ) finally , the container wall may frustrate the orientational order of the agitated rods in the entire cavity .
further experiments using containers with different sizes and shapes might shed light on such a discrepancy .
concerning the fluctuations , it is known that the velocity distributions of agitated granular spheres are non - gaussian and exhibit exponential tails , no matter whether the particles form clusters @xcite or not @xcite . as the dissipative nature
does not depend on the shape of the particles , we expect a similar behavior in our system .
this feature sets agitated granular rods apart from thermally driven liquid crystals , and triggers the question of how to define an effective ` thermal ' energy scale for a nonequilibrium system .
monitoring the mobility of individual granular rods with high speed photography could help to shed light on the difference between equilibrium and nonequilibrium systems found here .
( color online ) state diagram in the plane of aspect ratio @xmath27 , and area fraction @xmath107 obtained via experiments of agitated granular rods ( circles ) and mc simulations of thermally driven hard rectangles ( triangles ) .
the labels denote the states : isotropic ( i ) , uniaxial nematic ( u ) , and tetratic ( t ) .
they are colored according to the experimental data .
the diameter of the rods used in the experiments is @xmath135 mm for @xmath136 ( gray symbols ) and @xmath23 mm for @xmath137 ( black symbols ) .
closed and open symbols indicate the i - t and i - u transitions , respectively .
the inset shows the state diagram of equilibrium hard rods according to dft in comparison to mc simulations in an extended region of @xmath27 .
dashed lines are continuous transitions and solid lines denote first order transitions . ] in the inset of fig .
[ fig : phasediag ] we show the state diagram according to dft together with the thresholds obtained from mc simulations in an extended region of @xmath27 .
it is similar to the one predicted by the scaled particle theory @xcite .
dft also predicts i - t transitions for small @xmath27 and i - u transitions for large @xmath27 .
however , the tetratic state is stable only for @xmath138 , most likely because only two - body correlations are considered in the theory @xcite . concerning the ordering transition threshold @xmath110 , there is a good agreement between dft and mc simulations for @xmath139 . for low aspect ratios ,
the deviations between both approaches are due to the mean field character of the theory . for @xmath140 , dft predicts a t - u transition at very high area fractions . due to the limitations in both experiments and mc simulations , the region of very high area fractions , where the t - u transition may arise , has not been explored .
to conclude , the ordering in agitated granular rod monolayers is found to be determined predominately by the aspect ratio of the rods and the area fraction , while the frequency and the strength of the agitation are not essential .
it suggests that the detailed nature of energy injection into such a nonequilibrium system is not important , analogous to the role that temperature plays in equilibrium hard rod models . in comparison to previous experimental investigations on monolayer systems ,
we have focused on the bulk region of the container and found both tetratic and uniaxial nematic order for cylindrical rods .
this enables a direct comparison to the state diagram of the corresponding equilibrium system .
we find that , depending on whether the aspect ratio is smaller or larger than @xmath141 , a gradual isotropic - tetratic or an isotropic - uniaxial nematic transition arises as the area fraction grows , in both experiments and simulations .
this agreement suggests some degree of universality for the ordering of rod shaped particles across systems in and out of thermodynamic equilibrium .
nevertheless , we have also found a qualitative difference between both systems , namely the trend of the area fraction threshold at the ordering transitions .
further investigations will focus on characterizing the area fraction and velocity fluctuations of the system , in order to find an effective ` thermal ' energy scale for such a nonequilibrium system .
moreover , a comparison to molecular dynamics simulations @xcite with tunable rod - rod dissipation energy could help to elucidate how fluctuations influence the ordering transition threshold .
the authors would like to thank wilhelm august for the preliminary work on the experimental set - up .
inspiring discussions with m. schmidt , d. van der meer , and c. krlle are greatly acknowledged .
tm and kh acknowledge the support from the dfg through grant no .
hu1939/2 - 1 . 48ifxundefined [ 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.46.617 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1111/j.1749 - 6632.1949.tb27296.x [ * * , ( ) ] link:\doibase 10.1088/0034 - 4885/55/8/003 [ * * , ( ) ] link:\doibase 10.1073/pnas.1001651107 [ * * , ( ) ] link:\doibase 10.1038/nature09312 [ * * , ( ) ] http://www.nature.com/nature/journal/v491/n7424/full/nature11591.html [ * * , ( ) ] link:\doibase 10.1126/science.1125142 [ * * , ( ) ] link:\doibase 10.1038/nature03236 [ * * , ( ) ] link:\doibase 10.1073/pnas.1222339110 [ * * , ( ) ] http://www.pnas.org/content/111/4/1265 [ * * , ( ) ] link:\doibase 10.1016/s0378 - 4371(98)00501 - 9 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) \doibase http://dx.doi.org/10.1063/1.1849159 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physreve.79.011711 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physreve.80.011707 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.68.1259 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.79.833 [ * * , ( ) ] link:\doibase 10.1103/physreve.74.011304 [ * * , ( ) ] link:\doibase 10.1063/1.2815745 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/10/5/053020 [ * * , ( ) ] link:\doibase 10.1002/zamm.201000110 [ * * , ( ) ] link:\doibase 10.1103/physreve.88.062201 [ * * , ( ) ] link:\doibase 10.1039/c3sm50298h [ * * , ( ) ] http://link.aps.org/doi/10.1103/physreve.67.031303 [ * * , ( ) ] link:\doibase 10.1103/physreve.75.051301 [ * * , ( ) ] link:\doibase 10.1126/science.1140414 [ * * , ( ) ] link:\doibase 10.1126/science.1153456 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.144102 [ * * , ( ) ] http://www.nature.com/doifinder/10.1038/ncomms5688 [ * * , ( ) ] link:\doibase 10.1103/physreve.88.052203 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.96.028002 [ * * , ( ) ] link:\doibase 10.1039/b918034f [ * * , ( ) ] link:\doibase 10.1088/1742 - 5468/2006/01/p01005 [ * * , ( ) ] @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) link:\doibase 10.1103/physreva.36.3912 [ * * , ( ) ] link:\doibase 10.1103/physreva.42.2126 [ * * , ( ) ] link:\doibase 10.1039/c3sm52650j [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.105.168001 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.058001 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.2209000 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.4369 [ * * , ( ) ] link:\doibase 10.1103/physreve.62.r1489 [ * * , ( ) ] link:\doibase 10.1103/physreve.70.051312 [ * * , ( ) ] | the orientational order in vertically agitated granular rod monolayers is investigated experimentally and compared quantitatively with equilibrium monte carlo simulations and density functional theory . at sufficiently high number density ,
short rods form a tetratic state and long rods form a uniaxial nematic state
. the length - to - width ratio at which the order changes from tetratic to uniaxial is around @xmath0 in both experiments and simulations .
this agreement illustrates the universal aspects of the ordering of rod - shaped particles across equilibrium and nonequilibrium systems .
moreover , the assembly of granular rods into ordered states is found to be independent of the agitation frequency and strength , suggesting that the detailed nature of energy injection into such a nonequilibrium system does not play a crucial role . |
veritas ( figure [ veritas ] ) is an imaging atmospheric cherenkov telescope array , now entering its tenth year of operations .
the array consists of four identical telescopes , located at the fred lawrence whipple observatory ( flwo ) in arizona , each with a @xmath0 diameter tessellated reflector .
cherenkov light from gamma - ray and cosmic - ray initiated particle cascades is focused by the telescope reflectors onto 499-pixel photomultiplier tube ( pmt ) cameras , which cover a @xmath1 field of view .
the array has undergone two major upgrades over the past decade .
the first , in 2009 , involved the relocation of the original prototype telescope to a more favorable location , resulting in an approximate diamond - shaped layout with sides of @xmath2 .
this improved the angular reconstruction capabilities , and enhanced the sensitivity of the array . in 2012 ,
a major overhaul of the telescope electronics saw the installation of new trigger systems , and the replacement of all of the photosensors with higher quantum efficiency ( _ super bialkali _
photocathode ) pmts .
this again improved sensitivity , and led to a dramatic enhancement of the low energy response .
thanks to these upgrades , and to improvements in analysis and calibration tools , veritas now detects a source with 1% of the steady crab nebula flux in under 25@xmath3 ; less than half of the exposure required in the original array configuration . of similar importance is the increase in duty cycle provided by moonlight observations .
veritas now commonly conducts observations with the lunar disk up to 50% illuminated , including with reduced pmt voltage for higher illuminations .
moonlight observations account for as much as @xmath4 of the total annual observing yield @xcite , which averages approximately 1300 hours per year .
figure [ sensitivity ] shows the differential sensitivity of veritas for three different epochs .
further details are available in @xcite .
the veritas source catalog ( figure [ catalog ] ) now stands at 56 sources , and is composed of eight different source classes . to make the most efficient use of limited observing time ,
the balance of the observing plan has shifted from attempts to detect new sources , to precision measurements requiring deep exposures , or observations of variable sources during exceptional high flux states .
we summarize here some of the most interesting results from the past two years of veritas operations .
figure modified from tevcat ( http://tevcat.uchicago.edu ) .
, scaledwidth=100.0% ]
supernova remnants ( snrs ) are key targets for precision measurements with veritas , since the details of their gamma - ray morphology and spectra can be used to inform and constrain models of cosmic - ray acceleration and diffusion near the source .
exposures on tycho s snr , ic 443 and cassiopeia a have all more than doubled since the original veritas publications , and much of the more recent data has been taken with the most sensitive configuration of the observatory .
tycho s snr is the remnant of a historical type ia supernova , well - studied over all wavelengths .
the relative simplicity of the environment around this remnant makes it a favored system for theoretical models ( e.g. ) .
gamma - ray emission from the source was discovered by veritas in 2011 , at the level of 0.9% of the crab nebula flux @xcite .
the broadband spectral energy distribution ( sed ) has been interpreted within both hadronic and leptonic scenarios .
the updated veritas spectrum @xcite , shown in figure [ tychocasa ] , extends the published measurement down in energy from @xmath5 to @xmath6 , and , while consistent , can be fit with a softer power - law than the previous result ( @xmath7 ) .
this may imply a steeper spectrum of accelerated particles than predicted by standard diffusive shock acceleration models .
like tycho , cas a is a young remnant , unresolved at tev energies , but the relatively high gamma - ray flux allows for much more precise spectral measurements .
figure [ tychocasa ] shows the most recent veritas spectrum @xcite , calculated using over 60 hours of data collected over 6 years ( three times the originally published exposure ) .
when combined with fermi - lat results at lower energies , a broken power - law fit is favored .
the centroid of the emission is consistent with an unresolved source at the the center of the remnant , and with other measurements at tev and gev energies .
ic 443 is an older remnant ( 3 - 30@xmath8 ) expanding into an extremely inhomogeneous environment .
it has been established as a hadronic accelerator , through the measurement of a pion bump signature in the mev - gev range @xcite .
the initial tev gamma - ray detections @xcite identified a single site of emission co - located with dense molecular clouds , with which the snr is interacting .
updated results from veritas now resolve emission from the entire shell of the remnant , as illustrated in figure [ ic443 ] @xcite .
also shown is the @xmath9 map from fermi - lat , which correlates well with the tev emission @xcite .
the spectrum steepens from the gev to the tev regime ( from @xmath10 to @xmath11 ) , and this behaviour is remarkably uniform across the entire remnant , despite the widely varying environmental conditions and integrated flux values .
the interpretation of these results is ongoing , but ic 443 clearly provides a valuable laboratory for the exploration of particle acceleration and diffusion over a wide range of environmental conditions in a single object .
precision measurements of gamma - ray binary systems with veritas have also been conducted over the almost decade - long baseline .
these attempt to provide dense and complete sampling of the gamma - ray lightcurve over all orbital phases , as well as to monitor the systems for unusual flux states .
new results on hess j0632 + 057 are presented at this conference @xcite .
the full 315-day orbit is now sampled by veritas observations , providing clear detections and spectral measurements over all phases in which emission is detected ( figure [ binaries ] ) .
also shown are observations of ls i@xmath12 which , in 2014 , revealed exceptionally bright tev flares , reaching @xmath13 of the crab nebula with variability on daily timescales @xcite .
plans for the coming seasons include monitoring of psr j2032 + 4127 as it approaches periastron in 2017 - 2018 .
this system is a recently identified 20 - 30 year - period pulsar - be star binary co - located , and possibly associated , with tev 2032 + 4130 @xcite . in 2014 , reaching 30% of the crab nebula flux ( right).,title="fig:",scaledwidth=45.0% ] in 2014 , reaching 30% of the crab nebula flux ( right).,title="fig:",scaledwidth=55.0% ] veritas results on a deep exposure ( @xmath14 ) of the galactic center region were also recently published @xcite . with an average elevation angle of @xmath15 ,
the energy threshold for this analysis is @xmath16 , complementing studies by h.e.s.s .
point - like emission from the direction of both sgra * and the composite snr g0.9 + 0.1 has been measured . when these sources are subtracted from the excess maps ,
residual emission is observed along the galactic ridge , including a relatively bright component close to sgra * , which we label ver j1746 - 289 ( figure [ gc ] ) .
with its location in the northern hemisphere , much of veritas science naturally focuses on the extragalactic sky , and active galactic nuclei ( agn ) are the most numerous source class in the catalog .
veritas has detected 34 agn , the majority of which are high- or intermediate - frequency peaked bl lacertae objects ( hbls / ibls ) .
the remainder are the flat spectrum radio quasars ( fsrqs ) , pks 1222 + 216 and pks 1441 + 25 , the nearby radio galaxies m87 and ngc 1275 , and the starburst galaxy m82 .
multiwavelength monitoring programs provide contemporaneous measurements of the broadband seds of many of these sources , in particular through close collaboration with the swift x - ray telescope team .
veritas blazar results are covered in more detail elsewhere in these proceedings @xcite , as is their use as probes of the extragalactic background light ( ebl ) and intergalactic magnetic fields @xcite . here
we highlight just a couple of the most exciting recent extragalactic source results .
pks 1441 + 25 is one of only two fsrqs detected by veritas @xcite and , at a redshift of @xmath17 , is the second - most distant source in the very high energy sky .
observations in april 2015 , triggered by alerts from fermi - lat @xcite and magic @xcite , allowed a precise measurement of the emission during a high state , in which the flux reached @xmath18 of the crab nebula flux above @xmath19 .
the spectral index , @xmath20 , is the softest yet measured by veritas for any source , corresponding to an intrinsic index of @xmath21 after correction for absorption by the ebl ( figure [ blazars ] ) .
no emission is detected above @xmath22 , illustrating the importance of the energy threshold reduction provided by the 2012 camera and trigger upgrades for extragalactic science .
gamma - ray opacity arguments , as well as multiwavelength correlations , indicate that the origin of the emission must lie far from the center of the agn , outside of the broad line region .
the extreme distance of the source , coupled with the highly significant measurement of its spectrum , allow to place constraints on the intensity of the near - ultraviolet to near - infrared ebl which are comparable to those derived from stacking multiple sources with lower redshifts , and agree with ebl estimates from galaxy counts .
1es 1959 + 650 was among the first tev blazars to be detected @xcite and holds particular interest as the clearest example of a tev source which has exhibited `` orphan '' flaring - a prominent gamma - ray flare with no x - ray counterpart @xcite .
the source has been in an extended active state since summer 2015 , and veritas observed flares at the level of the crab nebula flux in fall 2015 @xcite .
further flaring activity in 2016 triggered additional veritas monitoring , revealing the source to be in an unprecedented high state , with flares exceeding 2.5 crab ( figure [ blazars ] ) .
cooperation and communication between the northern hemisphere gamma - ray instruments , fermi - lat and swift places this among the best - sampled extreme blazar flaring events , with observations often conducted multiple times within a 24-hour period .
gamma - ray astronomy is an important and growing field , but imaging atmospheric cherenkov telescopes ( iacts ) provide the means to study more than gamma - ray emission from specific astrophysical objects .
we summarize some recent updates from veritas below .
the most recent update on the search for gamma rays from wimp annihilation in dwarf spheroidal galaxies with veritas was presented at the 2015 icrc @xcite . as the exposure for targets with the
highest estimated luminosity due to dark matter annihilation approaches @xmath23 , further deep observations of individual targets provide diminishing returns , and are prone to systematic biases . an alternative approach , which offers considerable improvements , is to combine the data from multiple sources , weighting individual events according to their energies , locations , and the properties of the instrumental and astrophysical backgrounds @xcite . using this method ,
the combination of over 200 hours of observations , spread over 4 dwarf spheroidal galaxies , has been used to provide improved limits to the annihilation cross - section .
the cosmic - ray electron / positron spectrum provides a direct measurement of cosmic - ray acceleration and diffusion in our local galactic neighbourhood .
interest in this area has grown with the observation , by multiple instruments , of a positron component which increases up to an energy of @xmath22 .
while they are unable to easily discriminate electrons and positrons , the large effective collection area of iacts allows to extend measurements of the spectrum of cosmic - ray electrons beyond those energies probed by ams and fermi - lat .
veritas has now measured the combined electron - positron spectrum up to a few tev @xcite .
the results are consistent with those from h.e.s.s . ,
and the spectrum is best fit with two power - laws with a break energy of @xmath24 ( figure [ physics ] ) . .
the gray band shows the systematic uncertainty on the veritas measurement @xcite.,scaledwidth=60.0% ] with their large mirror area , iacts are powerful optical instruments , and probe a unique temporal parameter space .
veritas has recently developed an efficient method to search for ultra - fast optical transients from astronomical objects , sensitive to nanosecond pulses with fluxes as low as @xmath25 .
we have applied this technique to search for optical pulses in serendipitous archival observations of kic 8462852 , a star with a peculiar variability pattern which has been suggested as an exceptional target for seti searches @xcite . unsurprisingly ,
no emission was detected , but the technique is now established , and can be applied to searches of the extensive veritas data archives for brief optical transient events .
veritas has developed ongoing partnerships with many collaborations around the world , working both in gamma - rays and at other wavelengths , as well as with multi - messenger astroparticle physics observatories .
collaborative efforts with two relatively new facilities have recently enhanced the scientific program of veritas .
the icecube observatory has detected an isotropic flux of tev - pev astrophysical neutrinos , whose origin remains unclear @xcite .
iact arrays , including veritas , can attempt to identify a gamma - ray counterpart , or to constrain the properties of the neutrino source population .
veritas has made follow - up observations of 18 well - located icecube events ; those containing muon tracks , with @xmath26 angular uncertainty ( figure [ icecube ] ) .
no significant emission has been detected , and upper limits are typically a few % of the crab nebula flux @xcite . angular resolution ) and a high probability ( @xmath27 ) of astrophysical origin @xcite.,scaledwidth=100.0% ] the sources of astrophysical neutrinos may be variable , or have a variable component . in this case ,
prompt follow - up observations of any neutrino event could be the key to the discovery of a gamma - ray counterpart .
veritas and magic have both taken part in a long - running program , responding to potential neutrino flares originating from the direction of a list of candidate sources ( mostly blazars ) @xcite . with the identification of a high energy astrophysical neutrino flux
, more stringent conditions can now be applied to ensure that follow - up observations are targeting the locations of astrophysical neutrinos ( as opposed to fluctuations in the background of atmospheric neutrinos ) . using infrastructure developed for gamma - ray burst studies , icecube is now providing prompt gamma - ray coordinates network ( gcn ) alerts for high - energy starting events ( hese ) through the astrophysical multimessenger observatory network ( amon ) .
veritas received its first alert in april 2016 , with observations starting after 193 seconds .
almost all of the data were taken under bright moonlight conditions .
unfortunately , a revised location , issued @xmath28 after the initial alert , placed the neutrino point of origin outside of the field of view of the veritas observations .
the revised location was targeted by veritas the following night , but showed no excess .
veritas flux upper limits were posted as a gcn circular .
the prompt response provided a useful demonstration of the capabilities of the alert program , and observations will continue over the coming observing seasons .
the most recent partner facility to be fully commissioned is the high altitude water cerenkov ( hawc ) observatory , which observes the northern hemisphere very high energy gamma - ray sky continuously from its location in mexico .
hawc is able to provide alerts to the tev observatories , which allows prompt follow - up of extreme flaring events ( e.g. @xcite ) .
the much higher instantaneous sensitivity of iacts can then be used to study such events in detail and veritas , located at almost the same longitude as hawc , is particularly well - suited to this task .
the wide - field survey capabilities of hawc provide another promising avenue to new science .
hawc has recently released a preliminary skymap based on 341 day s live - time with almost the full detector in operation .
further results were shown in detail at this meeting @xcite .
while the veritas field - of - view ( @xmath29 ) is much smaller than hawc , veritas has been in operation since 2007 , and has archival observations of a significant fraction of the celestial sphere ( figure [ exposure ] ) .
this includes exposures , totalling @xmath30 , which cover at least 10 of the sources and source candidates in the preliminary hawc catalog which do not have an existing tev counterpart .
a consistent re - analysis of these data using the most recent analysis tools is in progress , and the results will be presented at future meetings .
we note that there exists some overlap with the veritas survey of the cygnus region , an update to which was presented here @xcite .
dedicated follow - up observations have been made of two particularly interesting regions identified by hawc .
the first is the region close to snr g54.1 + 0.3/psr 1930 + 1852 .
veritas detected a gamma - ray counterpart to this object , ver j1930 + 188 , in 2010 @xcite .
the emission is point - like , with a spectral index of @xmath31 , and is likely associated with the pulsar wind nebula .
the hawc view of this region displays emission from the location of the veritas source ( 2hwc j1930 + 188 ) , along with two further sources located @xmath32 ( 2hwc j1927 + 187 ) and @xmath33 ( 2hwc j1928 + 178 ) from psr 1930 + 1852 .
both of these excesses are close to relatively high spin - down luminosity pulsars ( psr j1928 + 1746 and psr b1925 + 18.8 ) , which may provide plausible candidates for the hawc emission .
we note that the original veritas publication already placed limits to point - like emission from psr j1928 + 1746 at the level of @xmath34 .
veritas has 27 hours of archival data on this region , taken in standard `` wobble '' mode with @xmath35 offset around g54.1 + 0.3 .
all of these data were recorded prior to 2009 ( before any of the upgrades to veritas were implemented ) , and half are with only 3 of the 4 telescopes in operation .
we have supplemented these data with an additional 12 hours of recent observations ( june 2016 ) , targeting a point roughly equidistant from the 3 hawc sources .
the resulting skymap , for the complete dataset , is shown on the left in figure [ hawc ] .
ver j1930 + 188 is re - detected , as we would expect , but there is no evidence for emission from either of the new hawc sources .
studies are ongoing , but the non - detection may be explained by a combination of effects : for example , the limited sensitivity of the archival data for sources with large offsets , and/or the influence of source characteristics such as spatial extension and hard spectra .
dedicated veritas observations have also been performed this summer of another of the preliminary hawc source candidates , 2hwc j1953 + 294 .
the veritas dataset in this case totals 37 hours , and reveals a statistically significant gamma - ray source located within the hawc source contours ( figure [ hawc ] , right ) .
the emission is coincident with the pwn da 495 ( g65.7 + 1.2 ) and its central object , wga j1952.2 + 2925 .
karpova et al .
@xcite have argued that , despite the lack of observed pulsations , the fermi - lat source 3fgl j1951.6 + 2926 is likely the counterpart of the pulsar in this system , and not of the pwn .
the addition of vhe measurements from hawc and veritas to the broadband sed will soon help to determine the nature of the high energy emission from this interesting object .
veritas continues to operate smoothly , and the instrument is currently in its most sensitive configuration to - date .
ongoing analysis developments promise further incremental improvements in the coming years , while new scientific opportunities are presented by the introduction of hawc , icecube , and other facilities .
operational funding for veritas is secure , and the collaboration plans to continue observing with the array until at least 2019 , with a 10-year anniversary workshop planned for 2017 .
the construction of the prototype mid - size schwarzchild - couder telescope @xcite at the flwo provides a firm link to cta , and to the future of ground - based gamma - ray astronomy .
this research is supported by grants from the u.s .
department of energy office of science , the u.s . national science foundation and the smithsonian institution , and by nserc in canada .
we acknowledge the excellent work of the technical support staff at the fred lawrence whipple observatory and at the collaborating institutions in the construction and operation of the instrument .
the veritas collaboration is grateful to trevor weekes for his seminal contributions and leadership in the field of vhe gamma - ray astrophysics , which made this study possible .
archambault , s. , et al .
, * apj * , 808 , 110 , 2015 park , n. , for the veritas collaboration , in _ proc .
34@xmath36 icrc _ , ( the hague , the netherlands , 2015 ) , arxiv:1508.07070 morlino , g. , & caprioli , d. , * a&a * , 538 , a81 , 2012 atoyan , a. , & dermer , c. d. , * apjl * , 749 , l26 , 2012 berezhko , e. g. , ksenofontov , l. t. , voelk , h. j. , * apj * , 763 , 14 , 2013 slane , p. , et al . , * apj * , 783 , 33 , 2014 .
acciari , v. a. , et al .
, * apjl * , 730 , l20 , 2011 .
park , n. , for the veritas collaboration , in _ proc .
34@xmath36 icrc _ , ( the hague , the netherlands , 2015 ) , arxiv:1508.07068 .
morlino , g. , & blasi , p. , * a&a * , 589 , a7 , 2016 .
kumar , s. , for the veritas collaboration , in _ proc .
34@xmath36 icrc _ , ( the hague , the netherlands , 2015 ) , arxiv:1508.07453 ackermann , m. , et al . , * science * , 339 , 807 , 2013 .
albert , j. , et al .
, * apjl * , 664 , l87 , 2007 .
acciari , v. a. , et al . , * apjl * , 698 , l133 , 2009 .
humensky , b. , for the veritas collaboration , in _ proc .
34@xmath36 icrc _ , ( the hague , the netherlands , 2015 ) , arxiv:1512.01911 hewitt , j. w. , hays , e. a. , tajima , h. , schmid , j. , & lat collaboration , v. c. 2016 , american astronomical society meeting abstracts , 227 , 238.10 schlensted , s. for the veritas collaboration .
this conference .
archambault , s. , et al .
, * apjl * , 817 , l7 , 2016 .
lyne , a. g. , et al .
, * mnras * , 451 , 581 , 2015 .
archer , a. , et al .
, * apj * , 821 , 129 , 2016 .
pacciani , l. , * the astronomer s telegram * , 7402 , 2015 .
mirzoyan , r. , * the astronomer s telegram * , 7416 , 2015 .
holder , j. , et al .
, * apjl * , 583 , l9 , 2003 .
krawczynski , h. , et al . , * apj * , 601 , 151 , 2004 .
mukherjee , r. , * the astronomer s telegram * , 8148 , 2015 .
zitzer , b. , for the veritas collaboration , in _ proc .
34@xmath36 icrc _ , ( the hague , the netherlands , 2015 ) , arxiv:1509.01105 .
geringer - sameth , a. , koushiappas , s. m. , & walker , m. g. , * phys .
d. * , 91 , 083535 , 2015 .
staszak , d. , for the veritas collaboration , in _ proc .
34@xmath36 icrc _ , ( the hague , the netherlands , 2015 ) , arxiv:1508.06597 .
abeysekara , a. u. , et al .
, * apjl * , 818 , l33 , 2016 .
aartsen , m. g. , et al .
letters * , 113 , 101101 , 2014 .
santander , m. , for the veritas & icecube collaborations , in _ proc .
34@xmath36 icrc _ , ( the hague , the netherlands , 2015 ) , arxiv::1509.00517 . the icecube collaboration , in _ proc .
34@xmath36 icrc _ , ( the hague , the netherlands , 2015 ) , * pos(icrc2015 ) * 1052 .
biland , a. , et al . , * the astronomer s telegram * , 9137 , 2016 .
sandoval , a. for the hawc collaboration .
this conference .
riviere , c. for the hawc collaboration .
this conference .
krause , m. for the veritas collaboration .
this conference .
acciari , v. a. , et al .
, * apjl * , 719 , l69 , 2010 .
karpova , a. , zyuzin , d. , danilenko , a. , & shibanov , y. , * mnras * , 453 , 2241 , 2015 .
benbow , w. for the cta collaboration . this conference . | the veritas imaging atmospheric cherenkov telescope array has been observing the northern tev sky with four telescopes since summer 2007 .
over 50 gamma - ray sources have been studied , including active and starburst galaxies , pulsars and their nebulae , supernova remnants and galactic binary systems .
we review here some of the most recent veritas results , and discuss the status and prospects for collaborative work with other gamma - ray instruments , and with multimessenger observatories . |
a few binaries have been identified as variable very - high - energy ( vhe ) @xmath1-ray sources ( aharonian et al .
2005a , b ; 2006 , albert et .
al . 2006 , 2007 ) . for ls 5039 ,
a periodicity in the @xmath1-ray flux , consistent with the orbital timescale as determined by casares et al .
( 2005 ) , was found with amazing precision ( aharonian et al .
short timescale variability displayed on top of this periodic behavior , both in flux and spectrum , was also reported .
indeed , the h.e.s.s .
observations of ls 5039 ( @xmath2 hours distributed over many orbital cycles , aharonian et al .
2006 ) constitute one of the most detailed datasets of high energy astrophysics .
accordingly , it has been subject of intense theoretical studies ( e.g. , bednarek 2006 , 2007 ; bosch - ramon et al . 2005 ; bttcher 2007 ; bttcher & demer 2005 ; dermer & bttcher 2006 ; dubus 2006a , b ; paredes et al . 2006 ; khangulyan et al . 2007 )
. however , the rich observed phenomenology of the system has precluded an unifying explanation . the discovery of a jet - like radio structure in ls 5039 and the fact of it being the only radio / x - ray source co - localized with an egret detection , prompted a microquasar interpretation ( paredes et al .
2001 ) . however , the current findings at radio and vhe @xmath1-rays in the cases of ls i + 61 303 ( dhawan et al . 2006 , albert et al . 2006 ) or psr b1259 - 63 ( aharonian et al .
2005 ) , gave the perspective that all three systems are different realizations of the same scenario : a pulsar - massive star binary .
dubus ( 2006a , b ) has studied these similarities .
he provided simulations of the extended radio emission in the case of ls 5039 , showing that the features found in high resolution radio observations could also be interpreted as the result of a pulsar wind .
this result , the rest of similarities found with systems for which a pulsar companion is known , and the assessment of the low x - ray state ( martocchia et al .
2005 ) , makes the pulsar hypothesis tenable . high energy emission from pulsar binaries
have been subject of study for a long time ( e.g. , moskalenko et al .
1993 , moskalenko 1995 ; arons & tavani 1993 ; maraschi & treves 1981 ; romero et al .
but the dichotomy between pulsars or black hole binaries is not settled ( see , e.g. , mirabel 2006 ) . here , for testing whether the h.e.s.s .
observations could be explained in a pulsar scenario , we have assumed that the compact object in ls 5039 is in fact a pulsar . under this assumption , this letter presents the results of a theoretical modeling of ls 5039 which is able to describe reasonably well the phenomenology found in the system .
following sierpowska & bednarek ( 2005 ) ( see also bednarek 1997 ) , we consider that the volume of the binary system is separated by a shock wave into two regions with different properties , with the shock appearing as a result of the collision between the pulsar and the massive star winds .
the position of the shock is defined by the parameter @xmath3 , where the wind velocity , @xmath4 , depends on the radial distance from the massive star , @xmath5 is the star mass - loss rate , and @xmath6 is the spin - down luminosity of the pulsar ( ball & dodd , 2001 ) . note that for @xmath7 the star wind dominates over the pulsar s : the termination shock wraps around it .
it is assumed that the pulsar and stellar winds are symmetric . in the case of ls 5039
( see table 1 ) the value of @xmath8 is between @xmath9 ( periastron ) and @xmath10 ( apastron ) . between the pulsar and the shock , inside the pulsar wind zone ( pwz )
, we assume that relativistic leptons are frozen in the magnetized pulsar wind which propagate radially from the pulsar .
therefore , their synchrotron losses are neglected in our cascade calculations .
we consider then ( klein nishina ) inverse compton ( ic ) cascades in this pwz , assuming that the thermal radiation from the massive star dominates in this region ( see e.g. , ball & kirk 2000 ) .
synchrotron emission are typically considered in models where the @xmath1-emission mainly comes from the shock region , especially applicable in the case of binaries with more eccentric orbits ( e.g. , kirk , ball & skjaeraasen 1999 ) . in such very eccentric massive binaries with large separations
the radiation field of the massive star does not dominate over the whole orbit and ic scattering is not effective in the pwz .
we assume that the cascade initiated by a primary lepton in the pwz develops in one dimension , i.e. in the direction of propagation of the primary particle .
the charged products of this cascade arrive finally to the shock region in the pulsar wind and follow the flow along the shock surface . due to the magnitude of the opacity to ic and @xmath1-@xmath1 , most electrons effectively interact in the pwz . indeed ,
for instance for inclination @xmath11 , for 1 tev ( 100 gev ) electrons , the opacity to ic is @xmath12 @xmath13 in the pwz .
the energy carried by electrons that actually reach the shock to be trapped by the magnetic field is less than @xmath14 of the primary injected energy .
the secondary @xmath1-rays move into the massive star wind region ( mswr ) .
some of them escape the binary system but a significant part can be absorbed due to @xmath1-@xmath1 process .
these cascades are followed by means of a monte carlo procedure ( explained in general terms in sierpowska & bednarek 2005 ) .
the details of the ls 5039 implementation worked out for this paper ( including the detailed account of the system geometry , the computation of opacities to the different processes as a function of orbital phase , and many other details specific to our study ) will be presented elsewhere .
we assume that the pulsar - provided injected power in relativistic electrons is a fraction of the spin - down power , and that they are described by a power - law in energy .
table 1 presents the few free parameters of the model . , which also relates the stellar mass - loss rate @xmath15 and @xmath16 .
] additional model parameters are needed , but these are bound to fixed values ( given also in table 1 ) under multiwavelength observations ( see , e.g. , the papers quoted at the introduction ) .
figure [ lc ] shows the phase folded h.e.s.s .
data for the integral flux above 1 tev ( aharonian et al .
2006 ) and the results of our theoretical model .
casares et al.s ( 2005 ) ephemeris was used to fold the observational data .
each experimental point represents a typical time span of observation of 28 min .
the corresponding phases for apastron , periastron , inferior ( infc ) , and superior ( supc ) conjunction are marked .
two periods and two different inclinations angles are shown .
the trends in the h.e.s.s .
datapoints are reproduced : orbital periodicity , a non - zero emission close to periastron , and the maximum of the fluxes is around infc .
vhe gamma - rays produced close to the star suffer severe absorption via pair production .
the absorption process strongly depends on the direction of propagation with respect to the massive star .
indeed , the cross - section for pair production depends on the angle @xmath17 between the vhe gamma - ray and optical photons , and has an energy threshold @xmath18 .
the level of absorption therefore depends on geometry and leads to orbital modulation of the vhe gamma - ray flux . for ls 5039 , apastron and periastron
are near infc and supc .
absorption provides flux maxima at infc , when @xmath1-photons propagate towards the observer outward the system , and opacities are the lowest ( cos @xmath19 .
flux minima happen at supc , when photons propagate toward the massive star ( absorption more effective , opacities the highest ) , as earlier discussed , e.g. , by dubus ( 2006a ) .
absorption alone would however produce strict modulation ( zero flux ) in the energy range 0.2 to 2 tev whereas the observations show that the flux at @xmath20 tev is stable .
additional processes , i.e. cascading , must be considered to explain the spectral modulation .
our model have these processes consistently included and the lightcurve details arise then as the interplay of the absorption of @xmath1-rays with the cascading process , in the framework of a varying geometry along the orbit of the system .
it can be seen that the differences between the two lightcurves corresponding to different inclinations are not large , but it is the detail at infc what is worth noticing .
the spectra for ls 5039 was presented in two broad orbital phase intervals around infc and supc ( aharonian et al .
2006 ) . a clear tendency for a harder spectrum at higher flux level was found .
this is reproduced by the theoretical model herein presented . in order to compare with the broad phase - intervals spectra presented by h.e.s.s .
, we have also computed the phase - average of our predictions , see figure 1 .
we found an overall agreement .
observations have also provided the evolution of the normalization and slope of a power - law fit to the 0.25 tev data in 0.1 phase - binning ( aharonian et al .
the variability of the spectral index and normalization of these fits along the orbit is impressive .
the use of a power - law fit has to do with low statistics in such shorter sub - orbital intervals : higher - order functional fittings ( such as a power - law with exponential cutoff ) provide a no better fit and were not justified .
we can directly compare with these h.e.s.s .
results fitting our spectral predictions in the same way .
however , we note that particularly for some phases close to periastron , the theoretical spectrum is badly mimicked by a power - law ( see the examples of figure [ single - spec ] ) . using these fits to each of
the theoretically predicted spectra as a function of phase along the orbit , we compare with h.e.s.s .
observations in figure [ phase - bin ] .
the overall similarity of the theoretical and observational fittings is notable .
depending on inclination there are a few points around infc for which the observational data seem to be described by a harder spectrum than what we obtain as a result of fitting the theoretical predictions .
this can be modeled with a slightly harder injection spectrum for electrons .
the normalization data point close to periastron is missed by our predicted fitting ranges .
this is the result of our spectrum being badly fitted by a power - law at these phases .
a detailed modeling of ls 5039 , under a pulsar scenario , accounting for the system geometry , klein - nishina ic , @xmath1-@xmath1 absorption , and a monte carlo computation of cascading , describes well the phenomenology found in the system at all timescales , both flux and spectrum - wise .
the predictive power of the model is impressive : it is based on just a handful of free parameters .
the results shown here were not obtained varying these parameters searching for a best fit , but rather exploring the predictions under minimal assumptions , what enhances their value .
the overall agreement found does not however constitute a proof that the system is indeed formed with a pulsar companion ( although contributes to circumstantial evidence in this direction ) .
this agreement does provide a framework in which the pulsar assumption can be precisely tested with future observations .
we acknowledge w. bednarek , j. cortina , i. ribas , j. rico , and n. sidro for comments , extended use of ieec - csic parallel computers cluster , and support by grants aya 2006 - 00530 and csic - pie 200750i029 .
lll + spin - down power of assumed pulsar & @xmath16 & @xmath21 erg s@xmath22 + fraction of spin - down power injected in relativistic electrons & @xmath23 & @xmath24 + slope of the injected power - law description of the electron spectrum & @xmath25 & @xmath26 + vhe cutoff of the injection spectra ( consistent with upper end of @xmath1-spectrum in h.e.s.s .
results ) & & @xmath27 tev + inclination of the orbit of the system with respect to the line of sight ( consistent with casares et al .
2005 ) & @xmath28 & 30@xmath29 , 60@xmath29 + + radius of star & @xmath30 & @xmath31 + mass of star & @xmath32 & @xmath33 + temperature of star & @xmath34 & @xmath35 k + mass loss rate of star & @xmath5 & @xmath36 + stellar wind ( sw ) termination velocity & @xmath37 & @xmath38 + sw initial velocity ( sw velocity at radius @xmath39 from the star : @xmath40 ) & @xmath41 & @xmath42 + distance to the system & @xmath43 & @xmath44 + eccentricity of the orbit & @xmath45 & @xmath46 + semimajor axis & @xmath47 & @xmath48 @xmath49 + longitude of periastron & @xmath50 & @xmath51 + | under the assumption that ls 5039 is a system composed by a pulsar rotating around an o6.5v star in a @xmath0 day orbit , we present the results of a theoretical modeling of the high energy phenomenology observed by the high energy stereoscopy array ( h.e.s.s . ) .
this model ( including detailed account of the system geometry , klein - nishina inverse compton , @xmath1-@xmath1 absorption , and cascading ) is able to describe well the rich observed phenomenology found in the system at all timescales , both flux and spectrum - wise . |
consider the cauchy problem for a strictly hyperbolic system of conservation laws @xmath7 where @xmath8 and @xmath9 is sufficiently smooth , @xmath10 open . if the initial data @xmath5 is of small total variation , the global existence was proved first in @xcite .
moreover a series of paper @xcite establishes the uniqueness and well posedness of the cauchy problem .
however , when @xmath5 has large total variation or even more generally @xmath5 belongs to @xmath11 , the solution @xmath12 may not exist globally in @xmath11 @xcite : only for special system it is possible to consider initial data with large total variation .
we recall some of the results available in this direction . * for scalar conservation laws , the entropic solution to generates a contractive semigroup w.r.t the @xmath13 distance , on a domain of @xmath11 data @xcite .
* for general temple class system , in @xcite it is proved the existence and stability of the entropic solution for initial data with arbitrarily large but bounded total variation . *
if all characteristic families are genuinely nonlinear and the system is temple class , the existence and stability for initial data in @xmath11 is proved in @xcite .
* for special @xmath14 systems , in which one of the equation is autonomous , various results have been proved in @xcite , with initial data with unbounded total variation .
an open question is if the semigroup of solutions to the systems of case 2 ) , defined on all the initial data @xmath5 with total variation arbitrary large but bounded , can be extended to data in @xmath11 . in many model systems , in fact , some of the characteristic fields are linearly degenerate , so that the result in @xcite does not apply : for example , the traffic model considered in @xcite or a @xmath14 model for chromatography .
aim of this paper is to prove that , at least in the case where the flux function @xmath15 is convex and shocks and rarefactions coincide , the solution to can be defined for @xmath3 . to illustrate the heart of the matter , we assume that the system admits a system of riemann coordinates @xmath16 , and that shock and rarefaction curves coincide in @xmath10 .
moreover we assume that each characteristic field is linearly degenerate or genuinely nonlinear .
we do not assume that rarefaction curves are straight lines .
we consider a set @xmath17 of the form @xmath18 , \
i=1,\dots , n \bigr\}.\ ] ] with @xmath19 we denote the space of @xmath11 functions with values in @xmath17 .
the main result of this paper is the following : [ t : final0 ] there exists a unique semigroup @xmath20 such that the following properties are satisfied : * for all @xmath21 , @xmath22 , with @xmath23 in @xmath6 , @xmath24 as @xmath25 , @xmath26 * each trajectory @xmath27 is a weak entropic solution to the cauchy problem with @xmath28 ; * if @xmath5 is piecewise constant , then , for @xmath29 sufficiently small , @xmath27 coincides with the function obtained by piecing together the solutions of the corresponding riemann problems .
the first two properties imply the well posedness of the cauchy problem with initial data in @xmath11 .
the uniqueness follows because @xmath4 satisfies iii ) and it is limit of wave front approximations .
as it is shown in the last example of @xcite , the semigroup @xmath4 can not be uniformly continuous : thus we can not apply any compactness argument to construct the solution @xmath30 .
the fundamental problem is that , differently from @xcite , the total variation of the riemann invariants corresponding to linearly degenerate families does not decrease in time .
the main idea of this paper is to study how the solution to the characteristic equation @xmath31 depends on the solution @xmath12 of .
it will be shown that , for a fixed time @xmath32 , the map @xmath33 depends lipschitz continuously on the initial data @xmath5 , and moreover the lipschitz constant is independent of the total variation of @xmath5 . since the riemann invariant @xmath34 is the broad solution to @xmath35 a simple argument gives the convergence of the wave front tracking approximations .
note that this also implies the well posedness of the ode when @xmath36 is an @xmath11 solution of the system . as noted in @xcite , for general system the solution to does not exist .
the paper is organized as follows .
section [ s : setting ] contains the basic assumptions on the system .
moreover we construct the wave front approximation of the solution @xmath37 . in section [ s : shiftdif ]
we analyze carefully the shift differential map , i.e. the evolution of a perturbation in @xmath5 in which only the position of the initial jumps has changed .
the method we use is essentially the one in @xcite , with slight modifications due to the fact that in our system the rarefaction curves do not need to be straight lines .
the main result is here the explicit computation of the shift differential map .
section [ s : charact ] is concerned with the equation for characteristics .
we prove the lipschitz dependence of the map @xmath38 w.r.t . both the initial data @xmath5 and @xmath39 .
moreover we will show that the lipschitz constant is independent from the total variation of @xmath5 .
finally , in section [ s : semicont ] , we prove theorem [ t : final0 ] .
we consider a strictly hyperbolic system of conservation laws @xmath40 where @xmath41 is a smooth vector field defined on some open set @xmath42 .
let @xmath43 be the jacobian matrix of @xmath15 and denote with @xmath44 its eigenvalues and with @xmath45 , @xmath46 its right and left eigenvectors , respectively .
we assume that the eigenvalues @xmath47 can be either genuinely nonlinear or linearly degenerate . in the following the @xmath48-th rarefaction curve through @xmath49
will be written as @xmath50 , with @xmath51 , while the @xmath48-th shock curve will be denoted by @xmath52 , and its speed by @xmath53 .
the directional derivative of a function @xmath54 in the direction of @xmath45 will be denoted as @xmath55 while the left and right limit of a bv function @xmath15 in a point @xmath56 will be written as @xmath57 we assume that the rarefaction curves @xmath58 generate a system of riemann coordinates @xmath59 .
we recall that a necessary and sufficient condition for the local existence of riemann coordinates is the frobenius involutive condition : if @xmath60 $ ] denotes the lie bracket of the vector fields @xmath61,@xmath62 , the condition is @xmath63 \in { \text{\rm span}}\{r_i , r_j\ } \quad \text{for all}\ i , j = 1,\dots , n.\ ] ] in the following we will use indifferently the conserved coordinates @xmath12 or the riemann coordinates @xmath1 . fix a domain @xmath64 , \
i=1,\dots , n \bigr\}.\ ] ] since @xmath17 is compact , there is a constant @xmath65 such that @xmath66 we suppose that the system is uniformly strictly hyperbolic in @xmath10 : this means that there exists a constant @xmath67 such that @xmath68 we also assume that in the system shock and rarefaction curves coincide : this implies @xcite that either the rarefaction curve @xmath50 is a straight line or the eigenvalue is linearly degenerate . in fact , one can prove that @xmath69 and for the shock curve @xmath52 we have @xmath70 if @xmath47 is genuinely nonlinear , the left hand side of is zero if and only if the rarefaction curve is a straight line , because @xmath71 is orthogonal to @xmath45 .
the flux function @xmath15 thus satisfies the following assumptions : * the eigenvalues @xmath47 of @xmath72 are linearly degenerate or genuinely nonlinear ; * the rarefaction curves form a system of coordinates ; * shock and rarefaction curves coincide .
the system has thus @xmath73 linearly degenerate fields @xmath74 , corresponding to the riemann invariants @xmath75 , and @xmath76 genuinely nonlinear fields @xmath77 , corresponding to the riemann invariants @xmath78 .
in the latter case we have @xmath79 for all @xmath80 . [ r:2x2case ]
if @xmath81 , then the rarefaction curves @xmath50 always generate a system of riemann coordinates .
thus our assumptions are satisfied by the following classes of systems : * both eigenvalues are linearly degenerate ; * one eigenvalue is linearly degenerate , the other genuinely nonlinear and the rarefaction curves of the latter are straight lines ; * both eigenvalues are genuinely nonlinear and the system is of temple class .
the various situations are shown in fig .
[ fi:2x2exmpl ] .
case ii ) corresponds to the traffic model considered in @xcite , while case i ) corresponds to @xmath14 chromatography .
given the two points @xmath82 , with coordinates @xmath83 and @xmath84 , with @xmath85 , consider the intermediate states @xmath86 , where @xmath87 for all @xmath88 , we denote with @xmath89 the vectors defined as @xmath90 and we define @xmath91 as @xmath92 where @xmath45 is the @xmath48-th eigenvector of @xmath93 .
we assume that the vectors @xmath91 are linearly independent for all @xmath94 .
this condition is satisfied for data in a sufficiently small neighborhood of a given point @xmath95 .
we denote also with @xmath96 the dual base .
we now define an approximated semigroup of solutions @xmath97 on a set @xmath98 .
the construction is similar to the one in @xcite . for any integer @xmath99 , set @xmath100 and let @xmath101 be the domain defined as @xmath102 given @xmath103
, we construct a solution @xmath37 by wave front tracking .
we first define how to solve the riemann problem @xmath104 $ ] , with @xmath105 .
the solution to the riemann problem @xmath106 is constructed by piecing together the solutions to the simple riemann problems @xmath107 $ ] , where @xmath108 is defined in .
if the @xmath48-th field is linearly degenerate , then @xmath107 $ ] is solved by a contact discontinuity travelling with speed @xmath109 .
if the @xmath48-th field is genuinely nonlinear and @xmath110 , then @xmath111 $ ] is solved by a shock travelling with the rankine - hugoniot speed @xmath112 . finally ,
if the @xmath48-th field is genuinely nonlinear and @xmath113 , then @xmath111 $ ] is solved by a rarefaction fan : if @xmath114 , @xmath115 , consider the states @xmath116 the solution will consist of @xmath117 shock waves @xmath118 $ ] , travelling with the corresponding shock speed @xmath119 . at time
@xmath120 we solve the initial riemann problems of @xmath121 .
note that the number of wave fronts is bounded by @xmath122 .
when two or more fronts interact , we solve again the riemann problem they generate , and so on .
it is easy to show that at each interaction at least one of the following alternatives holds : * the number of waves decreases at least by @xmath123 ; * the total variation of the solution @xmath37 decreases by @xmath124 , * the interaction potential @xmath125 , defined as @xmath126 , @xmath127 of the families @xmath128 , @xmath129 , located at points @xmath130 , @xmath131 , are considered as _ approaching _ if @xmath132 and @xmath133 .
this implies that there are at most a finite number of interactions , so that we can construct our approximate solution for all @xmath134 .
note that @xmath135 is a semigroup of solutions , but not entropic due to the presence of rarefaction fronts .
if the @xmath48-th family is linearly degenerate , the @xmath48-th riemann coordinate @xmath136 of the solution can be constructed by solving the semilinear system @xmath137 since @xmath12 is a piecewise constant solution , with a finite number of jumps , the broad solution to is well defined @xcite : if we denote with @xmath138 the solution to the ode @xmath139 then the solution to is given by @xmath140 in the following sections we will consider the dependence on the initial data @xmath5 of the genuinely nonlinear riemann coordinates @xmath141 and the map @xmath142 defined as @xmath143 where @xmath144 is the solution to .
in this section we prove some properties of the shift differential map .
these properties are closely related to the structure of , i.e. the conservation form , the coinciding shock and rarefaction assumption , which prevents the creation of shock when two jumps of the same family collide , and the existence of riemann invariants , which prevents the creation of shock when two jumps of different families interact . consider a wave front solution @xmath145 of , and assume that the initial datum @xmath146 has a finite number @xmath147 of jumps @xmath148 , located in @xmath149 : @xmath150
if @xmath151 is the shift rate of the jump @xmath148 , define @xmath152 as the front tracking solution with initial datum @xmath153 in the following , we will use the integral shift function , defined by @xmath154 if @xmath145 has a shock @xmath127 , located in @xmath155 , and if @xmath156 is its shift rate , it is clear that the following relation holds : @xmath157 we first recall the following result in @xcite , obtained using the conservation equation : [ l : cons ] consider a bounded , open region @xmath158 in the @xmath29-@xmath56 plane .
call @xmath159 , @xmath160 , the fronts entering @xmath158 and let @xmath161 be their shifts .
assume that the fronts leaving @xmath158 , say @xmath162 , @xmath163 , are linearly independent . then their shifts @xmath164 are uniquely determined by the linear relation @xmath165 as observed in @xcite , formula implies that the shift rates of the outgoing fronts depend only on the shift rates of the incoming ones , and not on the order in which these wave - fronts interact inside @xmath158 . in particular we can perform the following operations , without changing the shift rates of the outgoing fronts : * switch the order of which three or more fronts interact ; * invert the order of two fronts at time @xmath166 , if they have zero shift rate . the second lemma is concerned with a configuration where a sequence of contact discontinuities interacts with a wave of another family .
[ l : lindeg ] consider a family of parallel contact discontinuities @xmath148 , @xmath160 of the @xmath48-th linearly degenerate family and a single wave - front @xmath167 of the @xmath168-th family , @xmath169 .
let @xmath151 and @xmath170 be their initial shifts , respectively , and let @xmath171 , @xmath172 be their shifts after interaction .
assume that @xmath173 for all @xmath174 . then after the interactions all the shift rates @xmath171 of the @xmath48-th family have the same value @xmath175 and @xmath176 where @xmath177 , @xmath178 and @xmath179 , @xmath180 are the speeds of the shocks @xmath148 and @xmath167 , before and after interaction , respectively .
define the vector @xmath181 in the @xmath29-@xmath56 plane as the shift of the first collision point . by a direct computation
one finds @xmath182 since all the incoming shock of the linearly degenerate family have the same speed @xmath177 , by simple geometrical considerations it follows that the vector @xmath181 is constant during all interactions ( fig .
[ fi : intlind ] ) .
formula follows easily .
note that this lemma allows us to perform the following new operation , without changing the shift rates : * replace a family of contact discontinuities @xmath148 of a linearly degenerate , all with the same shift rate @xmath183 , by a single wave @xmath184 with shift rate @xmath183 . in the next lemma
we will show that the existence of riemann coordinates @xmath1 implies a strong relation among shocks of different families .
[ l : temple ] consider two adjacent jumps belonging to different families , @xmath185 and @xmath186 , @xmath187 , located at @xmath188 .
let @xmath189 , @xmath190 be their strength after interaction .
then the following holds : @xmath191 if @xmath192 , @xmath193 are the shift rates before interaction , and @xmath194 , @xmath195 after interaction , then follows easily from the conservation relation @xmath196 because by assumptions no waves of other families are generated . by condition
the conclusion follows .
[ r : o3 ] note that the previous lemma implies that the conservation relation is bidimensional , i.e. the shocks @xmath185 , @xmath186 and @xmath189 , @xmath190 lie on a two dimensional plane ( fig .
[ fi : bidime ] ) .
we can obtain then an identity which relates the the strengths @xmath167 with the speeds @xmath178 : substituting in , since @xmath183 , @xmath170 are arbitrary , we get @xmath197 one can show that if a riemann solver verifies for all couple of waves @xmath198 , then there exists a flux function @xmath15 such that the wave front approximation is a weak solution to . an important property of the shift differential map for temple class systems is the fact that a perturbation to the initial data , initially localized in @xmath199 $ ] , remains in the neighborhood of the set @xmath200 $ ] , where @xmath201 is the solution of the @xmath48-th characteristic equation starting at @xmath39 .
we now extend this property to hyperbolic systems satisfying the hypotheses * h1 ) * , * h2 ) * , * h3 ) * of section [ s : setting ] .
consider @xmath147 jumps @xmath148 , @xmath160 , of some linearly degenerate family @xmath48 , located at @xmath130 and corresponding to the jumps @xmath202 in the riemann coordinates @xmath1 : @xmath203 for some constants @xmath204 , @xmath160 .
we say that the jumps @xmath148 defined in are in involution if @xmath205 i.e. the initial and final riemann coordinate @xmath75 is the same : @xmath206
. note that , by the existence of riemann coordinates , this relation does not depend on the positions and strength of the shocks of the other families .
we can now extend lemma 2 in @xcite to our systems : [ l : excg ] consider a wave front tracking solution @xmath12 .
assume that there are @xmath147 shocks @xmath148 * either of the @xmath48-th linearly degenerate family in involution , * or of the @xmath168-th genuinely nonlinear family , and let @xmath207 , @xmath208 , be the position of the shock @xmath148 , @xmath160 .
then it is possible to assign shift rates to all shocks such that @xmath209 and the shift of all fronts outside the strip @xmath210 , x_1(t ) \leq x \leq x_n(t ) \}$ ] is zero .
we consider only the case of linearly degenerate family @xmath48 , since in the other case the proof is exactly the one given in @xcite .
let @xmath207 , @xmath211 , be the position of the shock @xmath159 of the @xmath48-th family in involution , and let @xmath212 be the value of the riemann coordinate at @xmath213 . for @xmath214 , define @xmath215 as the projection of @xmath1 on the hyperplane @xmath216 , and @xmath217 .
we choose the shift rates such that @xmath218 where @xmath219 is a scalar function different from @xmath166 only in @xmath220 $ ] , and we recall that @xmath221 . by imposing the value @xmath222 , i.e. @xmath223 , @xmath224 , we need to prove that can be satisfied at time @xmath120 .
we have two cases . *
if the jump @xmath185 belongs to the @xmath48-th family and is inside @xmath225 $ ] , then set @xmath226 . *
if the jump @xmath185 belong to the @xmath168-th family with @xmath227 , then by assumption and by there exists a unique shift @xmath192 and a unique constant @xmath228 such that @xmath229 since we assume that the shocks are in involution , setting @xmath230 we have that holds at time @xmath231 : in fact the last jump has size @xmath232 .
we now show that this property is conserved for all @xmath134 .
this follows easily from conservation and lemma [ l : temple ] .
the proof is exactly the same as in @xcite : we repeat it for completeness .
consider the interaction between two shocks @xmath185 and @xmath186 in the point @xmath233 , see fig .
[ fi : collision1 ] . by inductive assumption , we have for the states @xmath234 , @xmath235 and @xmath234 that @xmath236 using conservation we have @xmath237 so that for the new middle state @xmath238 we have @xmath239 and using lemma [ l : temple ] we conclude @xmath240 the same relation proves that they vanish outside @xmath158 : in fact , assume for example that @xmath241 and @xmath242 .
then from we get @xmath243 which implies that @xmath244 .
this concludes the proof .
[ r : lidegsh ] note that for discontinuities of a linearly degenerate family all shift rates has the same sign .
note moreover that if no waves of other families are present , then we shift all jumps @xmath148 by unit rate @xmath123 .
this corresponds to the case considered in lemma [
l : lindeg ] , i.e. to the substitution of a family of contact discontinuities with a single jump , whose strength in this case is @xmath166 by the involution assumption . using conservation and the previous lemmas , we obtain explicitly the shift differential map at a given time @xmath32 .
we recall that , given the states @xmath94 , we denote with @xmath91 the vectors defined in , and with @xmath245 its dual base .
let @xmath246 be the projection operator on @xmath247 : @xmath248 where @xmath249 denotes the the scalar product in @xmath250 .
given a point @xmath251 , with @xmath36 continuous in @xmath56 , define @xmath252 the intersection of the backward @xmath48-th characteristics starting at @xmath251 with the real axis @xmath253 , and for all @xmath254 let @xmath255 the index such that @xmath256 , @xmath257 , with @xmath258 and @xmath259 . without any loss of generality , we can assume that in @xmath254 there is a jump @xmath167 of the @xmath168-th family .
define the points @xmath260 by @xmath261 moreover define the point @xmath262 by @xmath263 if @xmath264 , and in a similar way , if @xmath265 , @xmath266 define @xmath267 as the vector @xmath268 where @xmath269 , @xmath270 , @xmath271 and @xmath167 is the initial jump in @xmath254 .
consider now a front tracking solution @xmath272 , obtained by shifting the initial jumps @xmath148 in @xmath149 with rates @xmath151 .
[ t : shiftmap ] if @xmath273 is the integral shift function of @xmath152 , defined in , then @xmath274 the theorem will be proved outside the times of interaction , because the lipschitz dependence in @xmath13 of the approximate semigroup implies the validity of for all @xmath275 .
if is sufficient to show that @xmath276 is piecewise constant , with jumps only at the points @xmath277 where @xmath145 has a shock @xmath127 , and the following relation holds : @xmath278 where @xmath156 is the shift rate of @xmath127 , located in @xmath56 . note that by the second equality of is trivially satisfied . by linearity in the shift rates
@xmath151 , we can consider the case in which a single shock is shifted , let us say @xmath167 at @xmath39 : becomes @xmath279 formula follows from the following considerations : consider a wave front pattern , fig .
[ fi : riemnprb ] , where for simplicity we assume that @xmath264 .
the states @xmath280 , @xmath281 are computed considering the riemann problem generated by adding to the @xmath168-jump @xmath167 in @xmath254 all the @xmath48-waves starting from the left of @xmath254 and ending in the right of @xmath251 and all the @xmath48-waves , with @xmath282 , starting from the right of @xmath254 and ending in the left of @xmath251 .
the jump @xmath283 is a single wave of the @xmath168-th family formed by adding all the @xmath168-waves between @xmath254 and @xmath251 . using the definition of @xmath273 given ,
one obtains easily the second case of : in fact the shift rates of the shocks in the left of @xmath251 is given by the shift rates of the jumps of the riemann problem @xmath284 ending in the left of @xmath251 , @xmath285 , plus the shift rate of the shock @xmath283 , @xmath286 . since only the @xmath48-waves with @xmath287
are present in @xmath288 , then @xmath289 .
the other cases can be computed in a similar way : in this case one solves the riemann problem @xmath283 in @xmath254 , and consider the @xmath168-wave @xmath284 starting in the left of @xmath254 and ending in the right of @xmath251 . from the above considerations it is clear that @xmath267 is piecewise constant , with jumps only when in @xmath251
there is a @xmath48-shock @xmath290 : in fact otherwise the wave front pattern used to compute @xmath267 remains the same .
let @xmath291 be the set of the starting points of all shocks arriving in @xmath251 , and define @xmath292 we consider two cases : * the shocks arriving in @xmath251 start on both sides of @xmath254 : @xmath293 . in this case , @xmath294 is the shift rate of the @xmath48-shock starting in the riemann problem @xmath295 if @xmath296 ( @xmath297 if @xmath298 ) which collides with a @xmath168-shock @xmath299 ( @xmath284 if @xmath298 ) :
in fact the only difference is that in @xmath300 there is a shock of the @xmath48-th family starting in @xmath254 , and @xmath48 is genuinely nonlinear . finally , using @xmath301 and lemma [ l : temple ]
, one can change position to the @xmath48-wave and the remaining @xmath168-wave @xmath283 , whose strength does not change .
if @xmath302 , there are no @xmath168-shocks starting on the right ( left ) of @xmath254 and ending on the right ( left ) of @xmath251 , so that @xmath294 is the shift rate of the @xmath48-shock of the riemann problem @xmath303 .
* the shocks of the @xmath48-th family arriving in @xmath251 start either in @xmath304 or @xmath305 : assume for definiteness that @xmath306 . in this case
the difference @xmath294 is the shift rate of the shock @xmath290 colliding with the shifted shocks of the riemann problem @xmath307 in @xmath254 , crossing the jump @xmath308 , and finally overtaking @xmath290 .
in fact one can use lemma [ l : temple ] ( and @xmath301 if @xmath48 is genuinely nonlinear ) to obtain the wave pattern of fig .
[ fi : riemnprb2 ] .
the various cases will be proved in the following lemmas .
[ l : simplify1 ] assume that @xmath293 , i.e. case 1 ) .
if the shock @xmath290 is of the @xmath48-th family , then its shift @xmath172 is @xmath309 we follow closely the method of @xcite .
assume for definiteness @xmath264 , the other cases being similar .
the basic idea is to reduce the computation to the single riemann problem @xmath295 , with eventually a single @xmath168-wave @xmath283 .
consider fig .
[ fi : riemnprb1 ] . by lemma [ l : cons ] , we can simplify the wave configuration considering only the fronts crossing starting in the right of @xmath254 and ending in the left of @xmath251 : in fact we can move the other fronts to @xmath310 without changing the shift rate of @xmath290 .
we can now shift the initial position of the waves of the @xmath48-th family merging in @xmath56 such that their initial position coincide with @xmath39 , without changing the shift rate @xmath172 .
this operation can be repeated for all shocks of genuinely nonlinear families .
finally , we can move the shocks of the linearly degenerate families such that they have the same sequence of interaction with the other shocks .
this means that , if @xmath311 is the position of the @xmath312-th shock of the @xmath48-th linearly degenerate family , the only interactions among shocks occurring in the sector @xmath313 $ ] are the one involving one @xmath48-th wave and one @xmath168-th wave , with @xmath169 . using lemma [ l : lindeg ] , we can at this point substitute them with a single shock , whose strength is the sum of the strengths of the @xmath48-waves .
finally we move their position at @xmath314 such that it coincides with @xmath39 : we obtain the wave patterns of fig .
[ fi : riemnprb1 ] . to conclude
, we just need to prove that the riemann problem obtained in this way is exactly @xmath295 and that the remaining @xmath168-wave is @xmath299 . by the previous argument ,
the strength of the shock of the @xmath312-th family @xmath242 , @xmath315 , is given by the @xmath312-waves starting in the right of @xmath39 and ending in the left of @xmath56 : since they are the only @xmath312-wave crossing the segment @xmath316 $ ] , it follows @xmath317 the other relations for @xmath318 and @xmath319 follows in the same way . finally , for @xmath320 the jump is @xmath321 .
note that the wave pattern is the same obtained in 1 ) .
we consider only the case @xmath306 , since the other is entirely similar .
[ l : simplify2 ] assume that @xmath306 .
then the shift @xmath172 of @xmath290 is given by @xmath322 the hypothesis implies that the @xmath48-th shocks ending at @xmath56 starts in the right of @xmath39 . with the same simplification considered in lemma [ l : simplify1 ] ,
we reduce to the riemann problem @xmath323 in @xmath324 , such that the waves of the @xmath312-th families , @xmath319 , generated at @xmath324 collide with the @xmath48-wave in @xmath130 ( see fig .
[ fi : riemnprb2 ] ) , after overtaking the @xmath168-wave @xmath283 .
the conclusion follows easily , since the wave pattern is the same considered in 2 ) .
this concludes the proof of theorem [ t : shiftmap ] .
finally we extend to our case the following result proved in @xcite : [ p : finitetv ] let @xmath12 be a wave - front tracking solution , and consider two wave - fronts , @xmath325 and @xmath326 , @xmath327 $ ] .
then there exists a second front tracking solution @xmath328 such that the initial and final positions of the two shocks is the same , and @xmath329 is uniformly bounded . for genuinely nonlinear fields ,
the proof is the same as in @xcite .
we then restrict the proof to the case of a linearly degenerate fields @xmath48 .
assume that there exists two jumps @xmath330 , @xmath331 of the @xmath48-th family , with positions @xmath332 , such that @xmath333 \
\text{and } \
y(0 ) \notin [ z_1(0),z_2(0 ) ] , \quad x(t ) \notin [ z_1(t),z_2(t ) ] \ \text{and } \
y(t ) \notin [ z_1(t),z_2(t)].\ ] ] for definiteness , assume @xmath334 , and the following conditions is satisfied : @xmath335 . \ ] ] let @xmath148 , @xmath160 be the jumps of linearly degenerate family @xmath48 in the strip @xmath336 : if we define @xmath337 it is easy to verify that the shocks @xmath148 , @xmath338 , are in involution . by lemma [ l : excg ]
, we can then moving the jumps to the left until either @xmath339 meets the wave fronts @xmath325 , or @xmath339 coincides with another shock of the @xmath48-th family ( fig .
[ fi : lidecanc ] ) .
it is clear that we can repeat the same procedure also in the following cases : * @xmath340 and @xmath341 $ ] ; * @xmath342 and @xmath343 $ ] ; * @xmath344 and @xmath345 $ ] .
it is now easy to prove that the total variation of the jumps of the @xmath48-th family satisfying can at most be @xmath346 . since @xmath325 , @xmath326 divide the lines @xmath231 and @xmath347 in three regions , the total variation of @xmath75 is bounded by @xmath348 .
in this section we prove some estimates on the solution @xmath201 of the characteristic equation : @xmath349 we assume for simplicity that the @xmath48-th family is linearly degenerate , however the same results are valid for characteristics of a genuinely nonlinear family if the following condition holds : for all @xmath32 there exists an @xmath350 such that in the strip @xmath351 there are no shock waves of the @xmath48-th family . given front tracking approximation @xmath12 , @xmath201 is unique , since it crosses only a finite number of transversal jumps , and it depends lipschitz continuously on the initial data @xmath39 ( see @xcite ) .
we want to give uniform estimates on this dependence .
the idea is to suppose that in @xmath39 there is a shock @xmath352 of the @xmath48-family of size @xmath350 : @xmath353 .
since by assumption no shocks of the @xmath48-family collide with @xmath352 , it is easy to construct a wave front solution : for @xmath354 , the solution @xmath355 takes values in @xmath356 \cap 2^{-\nu } { \mathbb{z } } , j=1,\dots , n \bigr\},\ ] ] while for @xmath357 , enlarging @xmath17 and assuming @xmath350 sufficiently small , @xmath358 \cap 2^{-\nu } { \mathbb{z } } , j \not= i , w_i(u ) \in [ a_i , b_i ] \cap \bigl\ { 2^{-\nu } { \mathbb{z}}+ \epsilon \bigr\ } \bigr\}.\ ] ] the following lemma proves the continuous dependence of the solution @xmath359 and the position @xmath360 of the shock @xmath352 w.r.t . @xmath350 .
[ l : littleshock ] consider a front tracking solution @xmath12 , with initial data @xmath5 and the characteristic lines @xmath361 , defined in for a linearly degenerate family @xmath48 .
let @xmath362 the wave front solution with initial data @xmath363 , where @xmath364 is defined as @xmath365 then there exists constants @xmath366 , @xmath367 , depending only on the total variation of the initial data @xmath5 , such that for all @xmath275 @xmath368 where @xmath369 is the position of the shock @xmath370 starting in @xmath371 .
the first inequality is an easy consequence of the @xmath13 continuous dependence for front tracking solutions , see @xcite . for the second one , note that all the shocks different from @xmath352 have size uniformly bigger than @xmath166 , so that their position is shifted of the order @xmath350 .
thus the second inequality follows by standard ode perturbation estimates , see @xcite .
an easy application of the previous lemma together with proposition [ p : finitetv ] implies that to compute @xmath372 and @xmath373 , we can actually consider in a solution @xmath328 with uniformly bounded total variation , so that the constant @xmath367 in is independent on the total variation of @xmath5 .
we now estimate the dependence of @xmath201 w.r.t .
@xmath12 .
[ p : shiftode ] let @xmath151 be the shift rate of the jump @xmath148 in @xmath146 , and denote with @xmath374 the solution to @xmath375 where @xmath376 is the shifted front tracking solution . then there exists a constant @xmath377 independent of the total variation of @xmath12 such that @xmath378 we give a sketch of the proof , for details one can see @xcite
. if @xmath350 is the size of the shock @xmath352 located in @xmath254 , then we can apply theorem [ t : shiftmap ] to compute its shift @xmath379 : by formula we obtain @xmath380 if @xmath381 is sufficiently small , then we have @xmath382 where @xmath383 is the position of the shifted shock and @xmath384 is its original position .
note that @xmath385 is the shift rate of the shock @xmath352 , after colliding with the shocks of the riemann problems @xmath284 and @xmath283 .
their total shift is proportional to @xmath386 , and after the interaction with @xmath352 , the shift of the latter is proportional to @xmath387 . thus taking the limit
as @xmath350 tends to @xmath166 of , we obtain for @xmath350 sufficiently small @xmath388 which implies .
we prove now the uniform lipschitz continuity of the map @xmath389 for all @xmath275 .
[ p : uniflip1 ] consider two characteristic lines @xmath390 @xmath391 , solution to .
there exists @xmath392 , depending only on the system and the set @xmath17 , such that @xmath393 as in the previous proposition , let @xmath350 be the size of the shock @xmath394 located in @xmath254 in riemann coordinates .
if @xmath395 is its shift rate , then for @xmath381 sufficiently small by theorem [ t : shiftmap ] we obtain @xmath396 in fact , by assumption , in the simplified wave patterns to compute the shift rate of @xmath352 , there are no waves of the @xmath48-th family different from @xmath352 . dividing by @xmath350 and taking the limit
as @xmath350 tends to @xmath166 , we obtain @xmath397 which implies @xmath398 we use the fact that @xmath399 tends to @xmath400 as @xmath401 .
since @xmath17 is compact , the conclusion follows easily .
[ r : uniflip1 ] the above proposition implies that the map @xmath402 defined in is uniformly lipschitz , independent on the total variation of @xmath5 , together with its inverse map @xmath403 . to end this section
, we give a different proof of the following result given in @xcite : [ p : decay ] if @xmath325 , @xmath326 are the positions of two adjacent @xmath168-rarefaction waves , then for some constant @xmath404 one has @xmath405 thus for all @xmath406 the total variation of the riemann invariant @xmath78 of the @xmath168-th genuinely nonlinear family with @xmath147 shocks at @xmath231 is bounded by @xmath407 \ } \leq { 2 ( b - a ) \over \kappa \tau } + \bigl\|w_{k}\bigr\|_{l^\infty } + ( n+1 ) 2^{1-\nu}.\ ] ] consider two adjacent @xmath168-rarefaction fronts @xmath325 and @xmath326 , and let @xmath408 , @xmath160 , be the interaction times of @xmath325 , @xmath326 with other waves in the interval @xmath409 $ ] .
fixed @xmath410 for some @xmath411 , let @xmath412 be the characteristic line of the @xmath168-th genuinely nonlinear family starting in @xmath413 ( see fig .
[ fi : decay ] ) .
assume @xmath414 sufficiently close to @xmath415 such that @xmath416 and @xmath417 does not collide with shocks of other families for @xmath418 $ ] .
let @xmath419 be the characteristic curve starting in @xmath420 . by the assumption of genuinely nonlinearity , at time @xmath421 we have @xmath422 for some constant @xmath423 , depending only on @xmath17 .
using proposition [ p : uniflip1 ] , at time @xmath32 we have @xmath424 repeating the process , it is possible to find a countable number of times @xmath415 such that @xmath425 and using we get @xmath426 repeating the process for @xmath326 and for all intervals @xmath427 , we obtain where @xmath428 .
the second equation follows noticing that the total amount of positive jumps in the interval @xmath199 $ ] is bounded by @xmath429 .
in this section we construct the semigroup @xmath4 on @xmath19 . in @xcite
it is shown that for all @xmath430 , there exists a semigroup @xmath431 defined on the domain @xmath432 which is the only limit of the wave front tracking approximations constructed in section [ s : setting ] .
we study now the dependence of the solution on the initial data @xmath433 .
we consider separately the case for genuinely nonlinear and linearly degenerate families .
[ p : gennlin1 ] consider a front tracking solution @xmath12 , such that @xmath146 has @xmath147 jumps @xmath148 , @xmath160 , and let @xmath151 be their shift rates .
given @xmath434 , denote with @xmath127 the jumps in the riemann invariant @xmath435 of the @xmath168-th genuinely nonlinear family .
then there exists a constant @xmath436 , depending only on the system and the domain @xmath17 such that @xmath437 the proof follows by theorem [ t : shiftmap ] and proposition [ p : decay ] .
in fact , fixed a shock @xmath438 , using theorem [ t : shiftmap ] , we have that at time @xmath32 for a shock @xmath127 of the @xmath48-th family there exist @xmath439 @xmath440 if the shock @xmath127 starts on both sides of @xmath441 , or , using the same estimate of proposition [ p : shiftode ] , @xmath442 if @xmath127 start on one side of @xmath438 .
since there is at most @xmath123 shocks such that holds , and the interval of influence is @xmath443 $ ] , using proposition [ p : decay ] together with and we obtain @xmath444 \bigr\ } \leq f ( 1 + 2^{-\nu } ) \bigl| \xi_{\bar \alpha } \sigma_{\bar \alpha } \bigr|.\ ] ] the conclusion follows the linearity of the shift differential map .
[ p : lindeg1 ] consider a wave front solution @xmath12 , such that @xmath146 has @xmath147 jumps @xmath148 , @xmath160 , and let @xmath151 be their shifts .
consider the equation , whit the eigenvalue @xmath47 linearly degenerate .
fixed @xmath434 , then the shift @xmath445 of @xmath446 is bounded by @xmath447 [ t : uniflipdep1 ] consider two initial data @xmath448 and @xmath449 , and denote with @xmath450 , the @xmath168-th riemann coordinate of @xmath451 , @xmath452 , corresponding to the @xmath168-th genuinely non linear family .
moreover , let @xmath453 , @xmath452 , the map defined in for the @xmath48-th linearly degenerate family .
then there exists a constant @xmath454 , independent of @xmath430 , such that the following estimates hold : @xmath455 @xmath456 consider two piecewise constant initial data @xmath457 , @xmath458 in @xmath459 , and construct a pseudo polygonal path @xmath460 , connecting @xmath448 and @xmath449 , such that @xmath461 we can assume that @xmath462 has a finite number @xmath147 of jumps .
if we denote with @xmath463 the path @xmath464 , we have by proposition [ p : gennlin1 ] @xmath465 if now @xmath466 , since @xmath467 converges to @xmath468 , we obtain .
since this estimate does not depend on the number of initial jumps @xmath147 , we can extend it uniformly on @xmath469 . using the same pseudo polygonal path , in a similar way we can prove that @xmath470 this shows that @xmath471 converges uniformly to the solution @xmath472 as @xmath466 and @xmath473 .
it also implies that @xmath474 this concludes the proof .
it is easy to prove that the right hand side of is a cauchy sequence in every compact set @xmath199 $ ] : in fact , using the finite speed of propagation , we can consider @xmath12 with compact support @xmath480 $ ] . for the components @xmath78 of the @xmath168-th genuinely nonlinear family , it follows directly from , while for a linearly degenerate component @xmath75 , let @xmath215 be a lipschitz continuous function such that @xmath481 by theorem [ t : uniflipdep1 ] we have for @xmath482 such that @xmath483 , @xmath484 , @xmath485 and it follows by easy computations that @xmath486 where @xmath366 is the lipschitz constant of @xmath215 .
this shows that @xmath487 is a cauchy sequence for all @xmath134 , because the right hand side of can be made arbitrarily small .
we can now prove the main theorem : * for all @xmath489 , @xmath22 , with @xmath490 in @xmath6 , @xmath24 as @xmath25 , @xmath491 * each trajectory @xmath492 is a weak entropic solution to the cauchy problem @xmath493 with @xmath28 ; * if @xmath5 is piecewise constant , then , for @xmath29 sufficiently small , @xmath27 coincides with the function obtained by piecing together the solutions of the corresponding riemann problems .
[ r : finalrem ] note that what we also proved that the characteristic equation is well posed for @xmath11 data : the solution @xmath144 is lipschitz continuous w.r.t . both variables .
this is not trivial , since even for @xmath14 systems not in conservation form the dependence is hlder continuous , while for general @xmath494 the solution does not exists .
note moreover that semigroup @xmath4 is continuous , but not uniformly continuous .
however if the initial data takes values is a compact set of @xmath495 , then the semigroup becomes uniformly continuous .
this extend the lipschitz continuity when the initial data have bounded total variation .
26 a. aw , m. rascle , _ resurrection of `` second order '' models of car traffic _ , siam appl . math . * 60 - 3 * ( 2000 ) , pag .
916 - 938 .
p. baiti , a. bressan , _ the semigroup generated by a temple class system with large data _ , diff .
* 10 * ( 1997 ) , pag .
401 - 418 .
baiti , h.k .
jenssen , _ well - posedness for a class of @xmath496 conservation laws with @xmath11 data _ , j. diff . eq .
* 140 - 1 * ( 1997 ) , pag .
161 - 185 .
s. bianchini , _ the semigroup generated by a temple class system with non - convex flux function _ , diff .
( to appear ) .
a. bressan , hyperbolic systems of conservation laws .
the one dimensional cauchy problem , oxford univ .
press , 2000 .
, _ the unique limit of the glimm scheme _ , arch .
130 * ( 1995 ) , pag .
205 - 230 . , _
unique solutions for a class of discontinuous differential equations _
soc . * 104 * ( 1988 ) , pag .
772 - 778 .
a. bressan , g. crasta , b. piccoli , well posedness of the cauchy problem for @xmath497 systems of conservation laws , _
memoir a.m.s . _
* 694 * ( 2000 ) .
a. bressan , r.m .
colombo , _ the semigroup generated by @xmath498 conservation laws _ , arch .
rational mech .
* 133 * ( 1995 ) , pag . 1 - 75 .
a. bressan , p. goatin , _ oleinik type estimates and uniqueness for @xmath497 conservation laws _ ,
156 * ( 1999 ) , pag .
26 - 49 . ,
_ stability of @xmath11 solutions of temple class systems _ , diff .
, to appear .
a. bressan , p. lefloch , _ uniqueness of weak solutions to hyperbolic systems of conservation laws _ , arch .
rational mech .
* 140 * ( 1997 ) , pag .
301 - 317 .
a. bressan , m. lewicka , _ a uniqueness condition for hyperbolic systems of conservation laws _ , discr .
systems * 6 - 3 * ( 2000 ) , pag .
673 - 682 .
a. bressan , t.p .
liu , t. yang , _
@xmath13 stability estimates for @xmath497 conservation laws _ , arch . rational mech . anal .
* 149 * ( 1999 ) , pag . 1 - 22 .
a. bressan , w. shen , _ uniqueness for discontinuous o.d.e . and conservation laws _ , nonlinear analysis * 34 * ( 1998 ) , pag
. 637 - 652 .
diperna , _ convergence of approximate solutions to conservation laws _ , arch .
rational mech .
* 82 * ( 1983 ) , pag .
j. glimm , _ solutions in the large for nonlinear hyperbolic systems of equations _ , comm .
pure appl . math . * 18 * ( 1965 ) , pag .
697 - 715 .
j. glimm , p. lax , decay of solutions of systems of nonlinear hyperbolic conservation laws , _ amer .
soc . memoir _
* 101 * ( 1970 ) .
jenssen , _
blowup for systems of conservation laws _ , siam math .
31 - 4 * ( 2000 ) , pag .
894 - 908 .
s. kruzhkov , _ first - order quasilinear equations with several space variables _ , mat .
* 123 * ( 1970 ) , 228255 .
english transl . in math .
* 10 * ( 1970 ) , pag .
lax , _ hyperbolic systems of conservation laws ii _ , comm .
pure appl . math . *
10 * ( 1957 ) , pag .
537 - 566 .
m. lewicka , _ on the well posedness of a system of balance laws with @xmath11 data _ , rend .
sem . mat .
padova * 102 * ( 1999 ) , pag 319 - 340 .
d. serre , _ solutions variation borne pour certains systmes hyperboliques de lois de conservation _ , j. diff . eq .
* 67 * ( 1983 ) , pag .
137 - 168 . , systmes de lois de conservation , diderot editor , paris 1996 . j. smoller , shock waves and reaction - diffusion equations , springer - verlag , new york , 1983 . b. temple , _ systems of conservation laws with invariant submanifolds _ , trans .
* 280 * ( 1983 ) , pag . | we consider a hyperbolic system of conservation laws @xmath0 where each characteristic field is either linearly degenerate or genuinely nonlinear . under the assumption of coinciding shock and rarefaction curves and the existence of a set of riemann coordinates @xmath1 , we prove that there exists a semigroup of solutions @xmath2 , defined on initial data @xmath3 .
the semigroup @xmath4 is continuous w.r.t .
time and the initial data @xmath5 in the @xmath6 topology .
moreover @xmath4 is unique and its trajectories are obtained as limits of wave front tracking approximations .
s.i.s.s.a . ref . 65/2000/m |
since the very beginning of the feasibility studies , the gaia mission design was driven by the need of determining the position and the brightness of huge numbers of stars with the uttermost accuracy . during the concept and technologt study completed in july 2000 ( see @xcite and @xcite ) , however , it clearly emerged that such a star - driven mission could also provide , several socalled by - products which would substantially enrich its already impressive scientific yield .
in particular , it was realized that the scientific case for gaia imaging of high - surface brightness sky regions such as the central regions of nearby galaxies was dramatic .
the issues connected with this opportunity have been addressed in a certain detail in a number of studies ( see @xcite , @xcite , @xcite and @xcite ) , and such observations were included in the mission baseline design as described in @xcite under the name of gaia galaxy survey . in this paper
the current ideas on the implementation of such a survey and on the expected performance are presented . in section 2 the overall strategy for galaxy detection and observation with gaia
is outlined . in sections 3 and 4 the detection and observation ,
respectively , are discussed in greater detail and results of dedicated studies are presented . in section 5 the substantial scientific case for performing such a survey is briefly sketched , and in section 6 the additional efforts required in terms of mission preparation , operations and telemetry are discussed
. only the measurement of relatively large enhancements of the surface brightness with respect to the sky background is discussed here , although it is believed that the rather similar issue of the observation of galactic nebulae could be addressed along the same lines .
the measurement of the surface brightness of the sky background , instead , calls for a substantially different approach , as described in @xcite .
during its scientific operations , the gaia satellite will continuously spin about its symmetry axis , and the charges contained in the ccd pixels will correspondingly be shifted along - scan to integrate the image for a longer exposure time , a technique known as time - delay integration ( tdi ) . in order to limit the ccds reading frequency and the corresponding readnoise and telemetry rate , a dedicated ccd readout process was devised , consisting in detecting objects as they enter the field of view , determining their position , magnitude and signal - to - noise ratio , and , if the latter exceeds a certain limit , collecting data from regions around such stars only . while this approach was the subject of detailed studies as far as point - like objects
were concerned , leading to the definition of the astro telescopes focal plane and the ccd binning strategy described e.g. in @xcite , the process of galaxy detection and observation called for a rather different approach in order to identify and measure faint surface brightness variations . as first suggested in @xcite ,
galaxies could be detected in the astro sky mapper ( asm ) as an average surface brightness significantly in excess over the local sky background and observed in different colours in the broad band photometer ( bbp ) .
this observing strategy preventing from optimally observing stars whenever a galaxy is being observed ( which is however less than 1% percent of the time ) , it could be implemented in one of the two astros only . in order to follow this general idea , one had to optimize :
* the size of the areas over which the average surface brightness and the local sky background values are computed . it is presently envisaged to determine the former through trimmed median filtering of asm1 samples of @xmath4 pixels and the latter over a few degrees of scan with the same algorithm .
however , the exact method used to determine the local sky background is not of interest for our purposes , since the readnoise is by far the dominant noise source .
* the value for the detection area and @xmath5 limit so that useful data are transmitted to the ground without being swamped in less interesting data from the milky way or the zodiacal light .
the larger the detection area , the fainter the detection limit can be for objects of constant surface brightness , if the error on the sky background is negligible . on the other hand , the detection area should be small enough that a large number of small objects would not be missed . *
the sampling scheme for galaxy observations , so as to establish a trade - off between angular resolution , readnoise and telemetry . as with stars ,
a larger sample size yields a smaller error on the ( average ) surface brightness , lower readnoise and telemetry rate but also a lower angular resolution .
this aspect is critical for the observations of galaxies , since their potentially very large angular extension will require in some cases the full readout of ccds . in the context of the studies carried out to demonstrate the feasibility of the gaia galaxy survey ( summarized in @xcite )
, these questions have been given satisfactory ( if subject to further improvements ) answers .
as for the detection , a statistical model of galaxy number density , size and surface brightness distribution was developed in order to characterize the `` typical '' galaxy ( see @xcite ) .
while this model obviously can not do justice to the well - known strong individuality displayed by many galaxies , it is believed to yield sufficiently reliable results when , as in our case , only statistical properties , i.e.properties averaged over large samples , are of interest .
on the basis of such a model , adopting current estimates of gaia sensitivity and noise , and adapting the statistical formulae for the estimation of the @xmath5 obtained in the observation of point - like objects presented in @xcite to the case of extended objects , one can estimate whether a given galaxy could be significantly detected above the sky background .
it is thus concluded that a typical galaxy of @xmath6 would be detected about 60% of the times ( i.e. 50 times on average during a 5-year mission ) with a @xmath7 using an area of @xmath4 for the detection , as shown in figure [ detection.fig ] . according to our afore - mentioned statistical model ,
there are about 3 million galaxies brighter than this limit away from the galactic plane ( i.e. with @xmath8 , where galaxy detection should nt be hampered by the high density of stars ) .
it can thus be conservatively concluded that gaia would be able to reliably detect at least 3 million galaxies , in agreement within a factor of two with the more optimistic estimation obtained in @xcite following a different approach .
bright galaxies such as those appearing in @xcite would thus typically be detected down to @xmath9 or @xmath10 .
in order to assess the angular resolution and the photometric accuracy obtainable from gaia galaxy observations , complete simulation software based on hst wfpc2 images and generating representative all - mission sets of gaia observations of the same field was developed . among other things , this has also allowed to determine the optimal sample size and to test different stacking techniques .
the astro psf resulting from different smearing effects was modelled using the tools provided in @xcite , whereas the expected ccd readnoise was calculated using the formulae given in @xcite and conservatively assuming full ccd readout in all cases .
the results of the simulations are shown in figure [ fluxmap.fig ] , where the original hst wfpc2 900 s image is compared with gaia bbp flux map obtained from stacking of 50 observations of 0.9 s each , i.e. with an effective exposure time of 45 s , and a sample size of @xmath11 pixels , chosen as the best trade - off between the needs of angular resolution and photometric accuracy . notwithstanding
the great difference in exposure time , most details are still clearly visible in gaia flux map , and the median photometric accuracy is of @xmath1 0.2 at a median surface brightness of @xmath12 20 , interestingly very similar to the predictions based on statistical formulae taking into account photon noise and readnoise . the angular resolution achievable in gaia bbp flux maps can instead be evaluated in figure [ fluxmap_detail.fig ] , where two bright hii regions with a separation of less than 0.5 arcsec are clearly resolved in gaia flux map as well .
more accurate estimates of the angular resolution based on model psfs give an angular resolution as low as 0.35 arcsec , comparable to superb ground - based observing sites in the rare moments of excellent seeing .
galaxy observations could therefore be profitably carried out in astro 2 , where a sample size of @xmath11 is not in conflict with the baseline sample size of @xmath13 adopted for the observation of stars .
the scientific case for a wide - angle high - resolution photometric ( and possibly astrometric , even though the latter issue has not been studied in detail as yet ) survey of bright galaxies is enormous .
the high reliability catalogues extending down to low galactic latitudes and high spatial resolution imaging of all sufficiently high surface brightness galaxies provided by gaia will make up a remarkably vast , homogeneous and well - defined database that will have a tremendous impact at two levels : for statistical analysis of the photometric structure of the central regions of tens of thousands of well - resolved galaxies , and for the study of the large scale structure of the local universe from the spatial distribution of all detected galaxies .
detailed analysis of the inner luminosity profiles of a large sample of galaxies will define the true incidence of core structures and complex morphologies .
inner color gradients will map recent star formation and dust lanes , and central luminosity cusps may indicate massive black holes .
multi - colour and multi - epoch information will allow the identification of variable sources for dedicated follow - up by other telescopes .
more in general , in conjunction with the proposed high - resolution survey for the detection of stars fainter than the nominal detection limit and the low - resolution survey of the sky background surface brightness , gaia promises to yield the first uniform measurements of brightness gradients of the `` real sky '' at all spatial scales .
several aspects of the picture which was outlined require careful definition .
the galaxy detection process must be validated through the development of a dedicated algorithm , suitable for the reliable detection of faint extended objects on a bright complex background .
one must then make sure that the astro ccds can be operated so that they can switch to the galaxy observation mode , possibly reading the full ccd , whenever a galaxy is detected .
the opportunity of galaxy observations in the af17 and in the ssm , where the much higher sensitivity and spectral resolution , respectively , would result in a much higher accuracy in surface photometry and astrophysical characterization of the observed galaxies , should be discussed .
finally , but perhaps most importantly , it must be evaluated how the required telemetry ( on average 120 before compression , where most readout samples will carry little signal and thus allow efficient compression ) can be accommodated within the overall telemetry budget .
the proposed gaia galaxy survey will provide a nearly all - sky , multi - color and multi - epoch astrometric and photometric galaxy survey . in the framework of the present mission design ,
the feasibility , scientific case and optimization of such a survey were discussed . from both statistical considerations and numerical simulations
it appears that galaxies could be reliably detected in the asm1 within square areas of @xmath4 and observed in the astro 2 bbp with a sample size of @xmath11 .
the first choice should yield the highest number of detected galaxies without too may false detections , whereas the second one would provide the best trade - off between angular resolution , readnoise and telemetry . under the present assumptions about the instrumental performance of the satellite payload , and provided some effort is put into its planning in the near future , the following measurement capabilities are expected from a 5-year mission : * at least galaxies brighter than @xmath14 will be detected . *
all detected galaxies will be observed with a 0.35 angular resolution and an all - mission accuracy in surface photometry of 0.2 at 20.0 in the @xmath15 band .
* multi - color ( in the 45 bbp broad bands ) and multi - epoch ( @xmath16 epochs ) information will be available for all observed objects .
white light and medium - band photometry could as well be obtained .
these outstanding measurement capabilities will result in a unique dataset providing state - of - the - art information on galaxy spatial distribution and surface photometry over a well - defined sample extending down to low galactic latitudes , which is in turn expected to yield significant scientific results concerning the large - scale structure of the local universe and the multi - color photometric structure of galaxy innermost regions .
99 de vaucouleurs g. , de vaucouleurs a. , corwin h.g .
, buta r.j . , paturel g. and fouqu p. 1991
, third reference catalogue of bright galaxies , springer esa 2000 , gaia : composition , formation and evolution of the galaxy , concept and technology study report , esa
sci(2000)4 hg e. , fabricius c. , knude j. and makarov v.v .
1998 , gaia surveys of nebulae and sky background , technical report sag_cuo_32 hg e. , fabricius c. , knude j. and makarov v.v .
1999 , sky survey and photometry by the gaia satellite , technical report sag_cuo_53 and baltic astronomy , 8 , 2556 hg e. 2001a , photometric and imaging performance , technical report gaia_cuo_91 and this volume hg e. 2001b , multi - colour photometry with gaia of the diffuse sky background , technical report gaia_cuo_92 and this volume lindegren l. 1998 , point spread functions for gaia including aberrations , technical report gaia_ll_25 lindegren l. 2000 , detection of faint galaxies with gaia , technical report gaia_ll_29 perryman m.a.c . et al .
2001 , gaia : composition , formation and evolution of the galaxy , a&a , 369 , 339363 vaccari m. and hg e. 1999a , statistical model of galaxies , technical report gaia_cuo_61 , html version available at ` http://hal.pd.astro.it/\simmattia/research/smog/ ` vaccari m. and hg e. 1999b , simulated gaia observations of galaxies , technical report gaia_cuo_69 vaccari , m. 2000 , gaia galaxy survey , master thesis , university of padova .
available at ` http://hal.pd.astro.it/\simmattia/research/ ` or upon request writing to ` [email protected] ` vannier m. 1998 , noise of gaia astro instrument , technical report gaia_mv_04 | the performance expected from a galaxy survey to be carried out with gaia , the gaia galaxy survey , is outlined . from a statistical model of galaxy number density , size and surface brightness distribution , and from detailed numerical simulations based on real images
, it is conservatively estimated that gaia would be able to detect and observe about 3 million galaxies brighter than @xmath0 and to provide multi - colour and multi - epoch broad - band photometry of these with an end - of - mission angular resolution of @xmath1 0.35 and a photometric accuracy of @xmath2 0.2 at @xmath3 . the substantial scientific case for performing such a survey and the additional efforts required in terms of mission preparation ,
operations and telemetry are also discussed . |
hypercomlex numbers are high - dimensional extensions of real numbers @xcite , and they have been introduced to solve multivariate signal processing problems , such as colour image processing @xcite , vector - sensor array signal processing @xcite , human gesture spotting @xcite , wind profile prediction @xcite , and wireless communications @xcite . in the last application ,
anemometer readings are modeled with hypercomplex numbers and the wind profile is predicted by adaptive filtering algorithms . in particular , pure quaternions have been widely used to model three - dimensional ( 3d ) wind speed .
when external atmospheric parameters are available , a full quaternion - valued model can be considered @xcite .
the quaternion - valued model leads to improved performance over real - valued models , since it accounts for the coupling of the wind measurements and can be extended to exploiting the augmented quaternion statistics @xcite . however
, pure quaternions do not form a mathematical ring @xcite , since the product of two pure quaternions is not a pure quaternion in general . as a result
, the related adaptive algorithms for 3-d wind profile prediction initialised in the pure quaternion will have to work in the full quaternion domain .
the prediction results have to be truncated from full quaternions to pure quaternions , implying redundant computations in the update process .
for example , it takes 16 real - valued multiplications and 12 real - valued additions to implement a full - quaternion - valued multiplication . by comparison
, it only takes 9 real - valued multiplications and 6 real - valued additions to calculate the multiplication of any two 3-d numbers . in this work
, we aim to develop a more compact 3-d wind speed model based on a 3-d mathematical ring called trinions , termed by an anonymous author who provided all possible definitions of 3-d numbers @xcite .
we first define the gradient operation in the trinion domain and then derive the least mean squares ( lms ) adaptive algorithm , which is further extended to augmented statistics .
the rest of this paper is organised as follows .
a brief introduction to trinions is provided in section ii .
the trinion - valued lms ( tlms ) algorithm is developed in section iii , and the augmented tlms ( atlms ) algorithm in section iv .
simulation results are presented in section v and conclusions are drawn in section vi .
as shown in fig .
1 , the 3-d wind speed is a tri - variate signal composed of three perpendicular components , and can be modeled with 3-d hypercomplex numbers ( see fig .
1 ) . there are various definitions of a 3-d number @xmath0 composed of one real part ( @xmath1 ) and two imaginary parts ( @xmath2 and @xmath3 ) , @xmath4 according to definitions about the relationships among the three base elements 1 , @xmath5 , and @xmath6 .
+ to form an abelian group of these three , the following rules apply @xcite : @xmath7 trinions subject to the above rules form a commutative ring , namely , for two trinions @xmath0 and @xmath8 , we have @xmath9 .
the modulus of @xmath0 is given by @xcite @xmath10 we define the following conjugate of @xmath0 , @xmath11 so that @xmath12 is equal to the real part of @xmath13 , i.e. @xmath14 . to develop an lms - like algorithm in the trinion domain , the trinion - valued gradient needs to be defined . in the complex domain , a variable @xmath15 and its conjugate can be viewed as two independent variables , based on which the complex - valued gradient can be defined @xcite .
similarly , in the quaternion domain , a variable @xmath16 and its three involutions can be viewed as four independent variables , based on which the quaternion - valued gradient can be derived @xcite . however , to our best knowledge , the trinion involution is not available in general . in this paper , we define the following gradients of a function @xmath17 with respect to the trinion - valued variable @xmath18 and its conjugate , @xmath19 respectively , where @xmath20 .
unlike the quaternion - valued gradient , the trinion - valued gradient remains the same , no matter which side of the sub - gradients the imaginary units @xmath5 and @xmath6 are on , since trinions are commutative .
we can then calculate the derivatives of some simple functions , for instance , @xmath21
similar to the lms algorithm in the complex - valued domain , the trinion - valued error @xmath22 is given by @xmath23 where @xmath24 is the reference signal , @xmath25 is the weight vector , @xmath26^t$ ] is the filter input , @xmath27 is the filter length , and @xmath28 is the prediction step . the cost function is expressed as @xmath29 using the steepest descent method , the following gradient need to be calculated @xmath30.\label{gra_def}\ ] ] for details of the calculation , please refer to appendix at the end . using results there , the update of the weight vector can be obtained as @xmath31 where @xmath32 is the step size . the algorithm formulated above
is termed as the trinion - valued lms ( tlms ) algorithm .
a zero - mean trinion - valued vector @xmath33 is composed of three zero - mean real - valued random vectors @xmath34 , @xmath35 , and @xmath36 , and their complete second - order statistics can be found in the following six real - valued covariance matrices : @xmath37 and they can be more efficiently represented by three trinion - valued covariance matrices : @xmath38 where superscript @xmath39 denotes the hermitian transpose and the two additional mappings of @xmath33 are defined as @xmath40 neither of these two mappings is an involution , and they are defined as shorthand notations only
. then we can obtain @xmath41 to take the complete second - order statistics into consideration , we need to make use of the augmented input vector composed of the original input vector @xmath42 and its two mappings @xmath43 , namely , @xmath44 and the predicted estimate @xmath45 will be given by @xmath46 where @xmath47 $ ] .
analogous to the derivation of the tlms algorithm , the update equation of the augmented tlms ( atlms ) algorithm is given by @xmath48 where @xmath49 is the step size .
the computational complexity for each update of the weight vector of the trinion - based and quaternion - based filtering algorithms are summarised in table i. it can be seen that the trinion model can effectively reduce the computational complexity compared to the quaternion model .
.computional complexity per update of the weight vector [ cols="^,^,^",options="header " , ] + + +
in this section , both proposed algorithms ( tlms and atlms ) are applied to anemometer readings provided by google s re@xmath50c initiative @xcite .
the wind speed measured on may 25 , 2011 is used for demonstration .
the step size is set to be @xmath51 .
the filter length is 8 , and the prediction step is 1 .
all algorithms are initialised with zeros .
the predicted results provided by tlms and atlms algorithms are shown in figs . 2 and 3 , respectively .
from the results , we can see that both algorithms can track the wind data effectively .
the learning curves averaged over 200 trials of the proposed algorithms are shown in fig .
4 , compared with the quaternion - based qlms and aqlms algorithms .
it can be observed that both augmented algorithms ( aqlms and atlms ) have a faster convergence rate than the original ones ( qlms and tlms ) , due to their full exploitation of the second - order statistics .
meanwhile , the proposed tlms algorithm has a similar performance with the qlms algorithm , while the atlms algorithm is comparable with the aqlms algorithm .
however , as shown in table i , the proposed trinion - based algorithms have a much lower computational complexity .
a compact model for three - dimensional wind profile prediction based on trinion algebra has been proposed , with two lms - type adaptive filtering algorithms derived using partial and full trinion - valued second - order statistics , respectively .
numerical simulations using recorded wind data have shown that the two algorithms have a similar performance to their quaternion - valued counterparts , but with a much lower computational complexity .
the cost function @xmath52 as a function of real - valued variables is given by @xmath53 where the time index ` @xmath54 ' has been dropped for the sake of compact notation .
then the three component - wise gradients can be computed as @xmath55 , \end{split } \label{sub_gra1}\ ] ] @xmath56 , \end{split } \label{sub_gra2}\ ] ] @xmath57 .
\end{split } \label{sub_gra3}\ ] ] finally , we can obtain the expression for the gradient of @xmath52 by substituting ( [ sub_gra1])([sub_gra3 ] ) into ( [ gra_def ] ) , @xmath58 which yields the update equation in ( [ update ] ) , as the term @xmath59 can be absorbed into the step size .
this work was supported in part by the national natural science foundation of china ( 61331019 , 61490690 ) , the china scholarship council postgraduate scholarship program ( 2014 ) , and the uk national grid .
i. l. kantor and a. s. solodovnikov , _ hypercomplex numbers : an elementary introduction to algebras_. springer - verlag , 1989 .
j. p. ward , _ quaternions and cayley numbers : algebra and applications_. dordrecht : kluwer academic publisher , 1997 . n. le bihan and s. j. sangwine , `` quaternion principle component analysis of color images , '' in _ proceedings of the ieee international conference on image processing _ , barcelona , spain , pp . 809812 , sep .
s. c. pei , j. h. chang , and j. j. ding , `` commutative reduced biquaternions and their fourier transform for signal and image processing applications , '' _ ieee transactions on signal processing _
52 , no . 7 , pp .
20122031 , jul .
t. a. ell , `` multi - vector color - image filters , '' in _ proceedings of the ieee international conference on image processing _ , san antonio , tx , usa , pp . 245248 , sep . 2007 .
l. q. guo , m. zhu , and x. h. ge , `` reduced biquaternion canonical transform , convolution and correlation , '' _ signal processing _ ,
91 , no . 8 , pp .
21472153 , aug .
s. miron , n. le bihan , and j. i. mars , `` quaternion - music for vector - sensor array processing , '' _ ieee transactions signal processing _ , vol .
. 12181229 , apr .
n. le bihan , s. miron , and j. i. mars , `` music algorithm for vector - sensors array processing using biquaternions , '' _ ieee transactions on signal processing _ , vol .
9 , pp . 45234533 , sep . 2007 .
j. f. jiang and j. q. zhang , `` geometric algebra of euclidean 3-space for electromagnetic vector - sensor array processing , part i : modeling , '' _ ieee transactions on antennas and propagation _ , vol .
39613973 , dec . 2010 .
x. f. gong , z. w. liu , and y. g. xu , `` coherent source localization : bicomplex polarimetric smoothing with electromagnetic vector - sensors , '' _ ieee transactions on aerospace and electronic systems _ ,
3 , pp . 22682285 , jul . 2011 . x. r. zhang , w. liu , y. g. xu , and z. w. liu , `` quaternion - valued robust adaptive beamformer for electromagnetic vector - sensor arrays with worst - case constraint , '' _ signal processing _ , vol .
104 , pp . 274283 , nov .
m. b. hawes and w. liu , `` design of fixed beamformers based on vector - sensor arrays , '' , vol .
2015 , 2015 .
b. che ujang , c. jahanchahi , c. c. took , and d. p. mandic , `` adaptive convex combination approach for the identification of improper quaternion processes , '' _ ieee transactions on neural networks and learning systems _ , vol .
1 , pp . 172182 , jan .
f. a. tobar and d. p. mandic , `` quaternion reproducing kernel hilbert spaces : existence and uniqueness conditions , '' _ ieee transactions on information theory _
9 , pp . 57365749 , sep . 2014 . c. c. took , g. strbac , k. aihara , and d. p. mandic , `` quaternion - valued short term joint forecasting of three - dimensional wind and atmospheric parameters , '' _ renewable energy _ , vol .
6 , pp . 17541760 , jun . 2011 .
j. navarro - moreno , r. m. fernandez - alcal , c. c. took , and d. p. mandic , `` prediction of wide - sense stationary quaternion random signals , '' _ signal processing _
9 , pp . 25732580 , sep . 2013
. q. barthlemy , a. larue , and j. i. mars , `` about qlms derivations , '' _ ieee signal processing letters _ , vol .
2 , pp . 240243 , feb . 2014 .
r. b. allenby .
_ rings , fields and groups : an introduction to abstract algebra_. butterworth - heinemann , 1991 .
omic s n - nion s site _ [ online ] .
available : http://asyncbrain.baf.cz/m/nt/index.htm .
d. assefa , l. mansinha , k. f. tiampo , h. rasmussen , and k. abdella , `` the trinion fourier transform of color images , '' _ signal processing _ , vol .
91 , no . 8 , pp .
18871900 , aug .
a. van den bos , `` complex gradient and hessian , '' _ iee proceedings - vision image signal process _ ,
141 , no . 6 , pp .
380383 , dec . 1994 .
m. d. jiang , w. liu , and y. li , `` a general quaternion - valued gradient operator and its applications to computational fluid dynamics and adaptive beamforming , '' in _ proceedings of the 19th international conference on digital signal processing _ , hong kong , china , pp .
821826 , aug .
w. liu , `` antenna array signal processing for a quaternion - valued wireless communication system , '' in _ proc .
the benjamin franklin symposium on microwave and antenna sub - systems ( benmas ) _ , philadelphia , us , september 2014 .
m. d. jiang , y. li , and w. liu , `` properties and applications of a restricted gradient operator , '' , july 2014 .
google , `` re@xmath50c : surface level wind data collection , '' _ google code 2011 _ [ online ] .
available : http://code.google.com/p/google-rec-csp/ | the problem of three - dimensional ( 3-d ) wind profile prediction is addressed based a trinion wind model , which inherently reckons the coupling of the three perpendicular components of a wind field . the augmented trinion statistics are developed and employed to enhance the prediction performance due to its full exploitation of the second - order statistics .
the proposed trinion domain processing can be regarded as a more compact version of the existing quaternion - valued approach , with a lower computational complexity .
simulations based on recorded wind data are provided to demonstrate the effectiveness of the proposed methods .
wind profile prediction , trinion - valued representation , least mean squares , adaptive filtering , widely linear processing |
quantumness is very fragile and its surviving over long distances and times is subtly interconnected to the amount of noise coupling into the quantum system through its unavoidable interaction with the classical environment . among the different quantum signatures , entanglement
, a concept firstly introduced by schrdinger @xcite in response to the famous einstein , podolsky and rosen paper in 1935 @xcite , probably represents the most fascinating one .
the conceptual consequences of entanglement are somehow subtle and deeply affect many aspects of modern physics as the interaction between reality and observer @xcite , the actual limit for defining metrology standards @xcite , the limit to communication security @xcite .
the optical realization of such an intriguing quantum feature has gained , in the last decades , a central role for realising complex quantum communication protocols @xcite . among the different approaches the use of bright beams , _
i.e. _ optical modes carrying many photons , appears as the most robust carrier of entangled information @xcite no matter if the latter will be coded into single photon like states ( discrete quantum coding ) or over the quadratures of the carrier beams ( continuous valued quantum coding ) @xcite . by interacting with the external world , a pure quantum state decoheres into a mixture @xcite and its quantumness , if not completely death @xcite , need to be restored or re enhanced by suitable protocols shor95 . as a matter of fact ,
decoherence limits both the attainable protocol fidelity and the type of accessible protocols .
the aim of this paper is to discuss and experimentally analyse the effects of the transmission over a lossy channel on the quantumness of bi partite gaussian continuous variable ( cv ) optical entangled states . while these effects on cv entangled states obtained by above threshold opos have been already investigated @xcite , we focus our analysis on the states generated by a type - ii sub threshold opo @xcite .
the birth of an entangled state happens whenever a genuine quantum correlation , _
i.e. _ without classical analogue , is set between two distinguishable quantum systems .
this implies that the system wavefunction can not factorize into the product of wavefunctions of the single sub
systems . while this definition is almost unique entanglement marks itself into different properties of the quantum state of the whole system and can be seen under different perspectives @xcite .
for this reason , since the first experimental demonstration of entanglement aspect81 several criteria and quantities have been introduced for analyzing _ entanglementness _ @xcite .
furthermore , the definition of an entangled state is conceptually connected to pure states @xcite .
the properties of these states are described by wave functions and _ gedanken _ experiments can be though for testing different theoretical aspects but we all know that the states we can prepare and manipulate in the laboratory are not pure but mixed . so that , experimental entanglement tests , especially in cv regime , are described in terms of density matrices rather than wavefunctions .
far from the ideal concept of pure states and systems that do not interact with the environment and/or observers , any quantum state undergoes to decoherence phenomena . in optics
the most common process leading to decoherence is the phase insensitive loss of photons through diffusion and absorption mechanisms .
this process is described by a lindblad equation @xcite for the evolution of the field operators that translates into a master equation for the state density matrix .
eventually , if the state is gaussian , _ i.e. _ its wigner function is gaussian , it admits a full description in term of its covariance matrix @xcite .
the investigated cv entangled state represents one of the possible quantum resources in cv teleportation protocols @xcite .
the experimental data we present extend the analysis of ref .
@xcite discussing the behaviour of the cv entangled system to the strong decoherence regime ( up 99% of loss ) .
moreover , we discuss in details the relationship between the three different entanglement criteria used in the cv field linking them to the teleportation fidelity and quantum discord , two possible quantum signatures for evaluating the possibility of using this class of states in quantum communication protocols @xcite .
the discussion on the experimental results is preceded by a detailed theoretical foreword aimed at giving a physical insight the different criteria and quantities actually used for classifying cv entanglement . with this paper
we prove that the state we are able to generate never disentangles .
this rather counter intuitive result has also been shown theoretically for the case of qubits states , showing a purity at the generation stage above a threshold value @xcite .
moreover , the state we analyse can be used as the quantum resource for an effective quantum teleportation , of a coherent state , for any value of the transmission loss although the teleportation fidelity slowly reduce to its quantum limit .
optical entangled states can be obtained in non
linear processes such as parametric amplifiers that , depending on their operating regime , can prepare either single photon @xcite or continuous variable gaussian entangled states @xcite . in the latter case
the non linear medium is allocated in a optical cavity and the opos , below the oscillation threshold and in a semiclassical approach , are described by bilinear hamiltonians so realising the paradigm for gaussian state generation @xcite .
in particular below threshold a single continuous wave opo , generating squeezing in a fully degenerate operation @xcite , can give raise to a pair of bright cv entangled beams in the non degenerate case @xcite .
both the cases lead to states that represent robust resources for implementing different quantum communication tasks @xcite . in this paper
we experimental discuss the behaviour , under strong loss , of a bright bi
partite cv gaussian entangled state outing a sub threshold type
in particular , we analyse , for this state , the behaviour of different quantumness and entanglement markers in order to discuss the limit at which this state can be transmitted before loosing its quantum ability of being employed in quantum communication protocols .
more in details , we discuss : state purity , mutual information , quantum discord and three different entanglement criteria : peres horodechi
simon peres96,simon00 , duan @xcite , and epr
reid @xcite .
in addition we relate these criteria to the teleportation fidelity @xmath0 , _
i.e. _ the state ability of overcoming the quantum limit in a teleportation protocol for a coherent state .
these states , already successfully generated in the laboratories ou92,laurat05,prl09 , are gaussian , _
i.e. _ they are completely characterized by their covariance matrix , namely , by the first and second moments of their quadratures so that they can be easily characterized by a single homodyne detector @xcite .
the paper is organized as follows : first , in sect .
[ sect : state ] a general overview over the state generated by a sub threshold opo is given and the model for its evolution in a lossy gaussian channel is discussed . in sect .
[ sect : markers ] the different entanglement markers are introduced in terms of the state covariance matrix .
the section includes a detailed discussion on the relationship between the different entanglement criteria and quantum markers for cv gaussian states .
next , in sect .
[ sect : evolved ] the quantum markers evolution in term of channel transmission is given . a brief description of the experimental apparatus , given in sect .
sect : experiment , precedes the discussion , in sect .
[ sect : results ] , of our experimental findings . eventually , in sect .
[ sect : conclusions ] , conclusions are drawn .
a sub threshold frequency degenerate type ii opo generates a bi
partite state made of the two downconverted cross
polarized beams @xcite .
these states are gaussian _
i.e. _ the wigner function ( @xmath1 ) of the state @xmath2 , expressed in terms of field quadratures is gaussian . in our case
, we can assume that the average values of quadratures to be zero prl09 so that we can write @xmath1 as@xmath3 } { \pi ^{2}\sqrt{\text{det}\left [ \mathbf{\sigma } \right ] } } ~ , \label{wigner0}\]]where @xmath4 is the vector of the field quadratures ( hereafter , whenever not necessary , we will omit the `` @xmath5 '' on operators so that : @xmath6 and @xmath7 ) for mode @xmath8 and @xmath9 respectively .
a gaussian state can be fully characterized by the matrix of the second order statistical moments of the field quadratures , _
i.e. _ the covariance matrix ( * cm * ) @xmath10 , whose elements are defined by @xmath11 where @xmath12 . moreover , for a bi partite field state , @xmath10 can be always given in the standard form by local symplectic operations@xmath13where @xmath14 @xmath15 and @xmath16 are @xmath17 real matrices , representing respectively the self and mutual correlations matrices .
the quantities , @xmath18 , @xmath19 , @xmath20 and @xmath21 are determined by the four local symplectic invariants @xmath22 , @xmath23 , @xmath24 , @xmath25 .
we note that the experimental matrices we have measured and that will be discussed in sect . [
sect : results ] are all in the above standard form . as a matter of fact , a sub threshold type ii opo , due to the symmetry of its hamiltonian , can only produce states of this kind .
so that , hereafter , whenever we refer to * cm * we will intend a covariance matrix in the standard form of eq .
( [ cm initial ] ) .
moreover , at the time of their birth , states produced in an opo show @xmath26 and @xmath27 , and the matrix is called symmetric and represents a bi
partite state where the energy is equally distributed between the two modes .
@xmath28 is a _ bona fide _ * cm * _ i.e. _ it describes a physical state iff @xmath29where @xmath30 . the above condition can be written in terms of the four symplectic invariants@xmath31the above inequalities are equivalent to the heisenberg uncertainty principle for a two mode state and to ask covariance matrix positivity .
the * cm * is also characterized by its symplectic eigenvalues@xmath32inequality ( [ heis rel .
sympl invariant ] ) assumes a simple form in term of @xmath33:@xmath34 we also note that a gaussian pure state is a minimum uncertainty state and that the * cm * relative to a pure state necessary has @xmath35 ( see eq .
( [ purity ] ) below ) so that the case @xmath36 implies @xmath26 , to ensure a _ bona fide _ * cm , * and for a pure fully symmetric state@xmath37moreover , for mixed fully symmetric states , @xmath38 with the inequality saturated only by pure states .
hereafter we will indicate these fully symmetric states as _
diagonal_. the meaning of this name will be clarified in sect .
[ sect : markers ] .
decoherence indicates the detrimental effect on a quantum system that stochastically interacts with the external world .
the state of the system , initially pure , becomes mixed . in the density matrix language it translates into the fact that while a pure state is represented by an idempotent density matrix @xmath2 ( @xmath39 ) this is not true for a decohered mixed state .
the lossy transmission of an arbitrary optical quantum state between two sites is an irreversible decoherence process that can be described by using the open systems approach @xcite . in this approach the environment
is seen as a reservoir ( thermal bath ) made up of infinite modes at thermal equilibrium .
the kossakowski - lindblad equation describes the time evolution in a noisy quantum channel of a multi
mode quantum state @xmath40 .
this model is valid if the system @xmath41 and the reservoir @xmath42 satisfy the following general conditions : * _ _ weak coupling ( born approximation ) _ _ * * * * the coupling between system and environment is so that the density matrix @xmath43 of the environment is negligibly influenced by the interaction ( the thermal bath state is stationary ) .
this approximation allows to write the state @xmath44 of the global system as @xmath45 @xmath46 ; * _ markovianity _ there are not memory effects neither on the system and the reservoir .
this approximation implies that the time scale @xmath47 over which @xmath48 changes appreciably under the influence of the bath is large compared to the time scale @xmath49 over which the bath forgets about its past , @xmath50 . * _ secularity ( rotating wave approximation ) _ the typical time scale @xmath51 of the intrinsic evolution of the system @xmath41 is small compared to the relaxation time @xmath47 . for an optical field it implies that the reservoir reacts to the average field and not to its instantaneous value .
the evolution of an arbitrary two mode gaussian state in a noisy channel , at thermal equilibrium , can be translated into the formalism of the wigner function obtaining the following fokker - planck equation for @xmath52 ( see eq.([wigner0 ] ) ) @xcite,@xmath53where @xmath54 is the damping rate of the channel for both modes , while @xmath55 and @xmath56 .
the evolution of eq.([fokker - planck ] ) preserves the gaussian character of the initial state and in terms of the * cm * @xmath10 reads@xmath57where @xmath58 is the covariance matrix at @xmath59 and @xmath60 is the @xmath61 identity matrix ( @xmath62 is the vacuum state covariance matrix setting the standard quantum limit sql ) .
this form is in all equal to the effects of a fictitious beam
splitter ( bs ) that mimics the channel losses and couples into the system the vacuum quantum noise through its unfilled port .
being @xmath63 the su(2 ) transformation induced by the bs on the @xmath64mode ( @xmath65 , with @xmath66 the modal operator for the vacuum ) and @xmath67 the power transmission of the beam splitter ( @xmath68 ) , the above equation becomes:@xmath69 in this form we can drop the temporal dependence and label the * cm * of the initial state as @xmath70 . for states in the form of eq .
( [ cm initial ] ) the different elements evolve as@xmath71 for an infinite transmission channel @xmath72 implies @xmath73 _ i.e. _ the fully decohered * cm * represents a two mode vacuum state .
the quantumness of a bi partite * cv * state can be tested by two classes of markers .
the first is intimately related to the state itself and includes inequalities that fix bounds for distinguishing among entangled and un entangled states .
the second has been translated into the quantum context from the classical information theory and is related to the amount of quantum information carried by the state .
the latter includes quantitative measures such as mutual information , von neumann entropy , quantum discord and teleportation fidelity .
the former is made of entanglement criteria usually named by the authors that have theoretically found them .
they are the phs ( peres - horodecki - simon ) @xcite , duan@xcite , epr
reid @xcite criteria . before going into details it is useful to start by introducing the state purity as an extra marker that tells how far the analysed state is from a pure one .
every quantum information protocol and every measurement scheme must take into account the fact that pure states are inevitably corrupted by the interaction with the environment and , therefore , in real experiments only mixed states are available .
the purity @xmath74 $ ] ( @xmath75 for mixed state , @xmath76 for pure ) may be expressed as a function of the * cm * ( [ cm initial ] ) being@xmath77 } } ~. \label{purity}\]]intimately related to the purity , it is possible to introduce the von neumann entropy @xmath78 .
it is zero for pure states while strictly positive for mixed ones . in the case of the two mode gaussian state
the entropy @xmath78 can be written in terms of its covariance matrix symplectic eigenvalues @xmath79 ( see eq .
( symplectic ) ) as @xmath80 @xcite where @xmath81we note that both @xmath82 and @xmath78 are invariant under symplectic transformations .
although it has not been found a general solution to the problem of a quantitative measure of entanglement for mixed states , there exist necessary and sufficient conditions to asses whether a given state is entangled or not .
these criteria provide a test for entanglement , and can be employed for studying the behaviour of entanglement in a quantum transmission channel . a bi
partite quantum state is separable iff its density operator can be written as a convex combination of the tensor product of density operators relative to the two different sub
systems @xcite@xmath83where @xmath84 while @xmath85 @xmath86 are the density matrices of subsystems @xmath8 and @xmath9 . by performing a partial transposition ( _ i.e. _ transposition of the density matrix with respect to only one of the two hilbert subspaces ) @xmath2 transforms into @xmath87 . in the case
@xmath2 is a separable state , @xmath87 should still represent a physical state so that @xmath87 must be a non negative density operator . on the contrary
if @xmath87 no more represents a physical state the system does not admit the form ( [ sepst ] ) .
thus , all separable states have a non - negative partially transposed density operator . from this consideration
it is possible to deduce a necessary condition for separability or viceversa a sufficient condition for entanglement . in view of this
the criterion is sometime referred to as the _ ppt _ criterion ( _ positivity _ under _ partial transposition _ ) @xcite . in the following
we prefer to indicate it as the phs criterion .
it can be proven that it becomes a necessary and sufficient condition for gaussian states @xcite . from the * cm * point of view
, partial transposition implies a sign flip for @xmath88 so that phs criterion has a simple expression in terms of * cm * elements : a bi
partite gaussian state is separable iff @xmath89and it is entangled otherwise . * * * * we also note that the phs criterion is invariant under symplectic transformations .
phs criterion relies on the possibility of describing independently the two subsystems . in phase space
, transposition is defined as time reversal ( or mirror reflection ) , which is given by a change of sign of out of phase quadrature ( usually indicated as @xmath90 ) .
partial transposition is , therefore , a local time reversal which inverts the @xmath90 quadrature of only one subsystem . if any _ true quantum _ correlation is set between @xmath91 and @xmath92 a sign flip on @xmath91 ( or @xmath92 )
will affect the sign of @xmath20 ( or @xmath21 ) in eq .
( [ cm initial ] ) making @xmath87 no more physical ( see eq .
( [ heis rel .
sympl invariant ] ) ) . for every bi
partite state there exists a pair of epr like conjugate operators defined by@xmath93with @xmath94 an arbitrary non
zero real number and @xmath95 = \frac{i}{2}\delta _ { jj^{\prime } } $ ] ( @xmath96 ) and where subscript refers to the @xmath8 ( @xmath9 ) subsystem . then , the birth of non classical correlation between sub
systems @xmath8 and @xmath9 will lead to states for which the variance of epr like operators will reduce below the standard quantum limit ( sql ) . by calculating the total variance of such a pair of operators on @xmath2 , a separable state of the form of eq .
( sepst ) , it can be proven @xcite that @xmath97setting a lower bound for separable states .
contrarily to the phs criterion ( see sect . [
sect : phs ] ) inequality ( [ duan ] ) has been formulated only as a necessary condition for separability so that it is a sufficient condition for entanglement of a generic cv gaussian state .
as shown in @xcite it becomes a necessary and sufficient condition for entangled cv gaussian states .
the sufficient and necessary condition can be expressed in terms of the covariance matrix elements iff the matrix itself is expressed in the form of eq .
( 10 ) of ref
. @xcite@xmath98with the matrix elements satisfying the constrains ( 11a ) and ( 11b ) of ref .
@xcite that , for the sql equal to @xmath99 , read@xmath100 in this case the epr operators pair of eq .
( [ eprop ] ) are rewritten as@xmath101where @xmath102 with the sufficient and necessary duan criterion given by@xmath103we note that , as proved in ref .
@xcite , any gaussian state can be transformed into the form ( [ cm_duan ] ) by local linear unitary bogoliubov operations , _
i.e. _ by acting independently on one or both the subsystems by applying local squeezing and/or rotations . in a more general fashion
it is possible to write the sufficient but not necessary condition of eq .
( [ duan ] ) for a * cm * in the standard form of eq .
( [ cm initial ] ) as@xmath104where @xmath94 can be set to @xmath105 to minimize the left hand side of the inequality:@xmath106we note that , while for symmetric states ( @xmath107 ) @xmath108 and the epr pair consists of two orthogonal field quadratures , this is not true , in general . the duan criterion is strictly related to the heisenberg principle for the single sub system .
if the state is separable the indeterminacy on a single operator is disjoint from the indeterminacy of the twin operator on the second sub
system ; so that the total indeterminacy can not violate the heisenberg limit .
this has nothing to do with conditional measurement and with the possibility of gaining information on one sub
system measuring the other .
as we will see this approach leads to a stricter criterion : the so called epr `` _ _ reid _ _ '' criterion .
a stronger bound can be found by analysing a bi partite state under the shadow of conditional measurements .
this concept descends directly from the original epr _ gedanken _ experiment @xcite . for this reason it is usually indicated as the epr criterion and
was firstly introduced by reid in 1989 @xcite , in the very early days of quantum information .
it describes the ability to deduce the expectation value of an observable on a sub
system by measuring the epr companion observable on the second sub system .
the epr `` reid '' criterion sets , as the duan one in its generic form ( [ duancr ] ) , only a sufficient condition for assessing entanglement .
mathematically this criterion can be deduced by calculating the conditional variance for an observable on sub
system @xmath8 given the result of a measurement on sub
system @xmath9 and comparing it with the standard quantum limit . for gaussian states
it can be written in terms of * cm
* elements : @xmath109we note that , being based on conditional variances ( and thus on conditional states ) this last criterion is not symmetric under the exchange of the two sub - systems .
this asymmetry reveals itself in a asymmetric sensitivity to loss @xcite .
then , the criterion itself can be recast if sub system @xmath8 is measured and the conditional variance on @xmath9 is given@xmath110the two definitions of the epr criterion can make it ambiguous if one of the relations are not satisfied .
this is not the case of balanced systems ( @xmath107 ) .
moreover , it can be proved that no pure state can asymmetrically violate the epr criterion . at the same time
this asymmetry can be useful for revealing eavesdropping in one sided device
independent qkd schemes branciard12 .
it is easy to see that the above two expressions for the epr criterion are invariant for symplectic transformations like the phs one ( see eq .
( phscr ) ) .
the epr criterion is necessary and sufficient for assessing the steering form of nonlocality as recently discussed in refs .
this concept is very recently gaining much attention as a peculiar consequence of entaglement @xcite .
it is interesting to note that the concept of conditional variance , used for quantum steering , was used for assessing state preparation ability in quantum demolition experiments @xcite . in that case steering
happens between the probe and the signal systems .
all the above criteria ( [ phscr ] ) , ( [ duancr ] ) , and ( [ eprcr ] ) can not be used other than as bounds .
they are not suitable for measuring entanglement in a quantitative way . quantifying entanglement for mixed states
is a complicated issue still under discussion .
there is nt , to date , a single , universal measure that quantifies the entanglement for a mixed state . in discussing the experimental results in sect .
[ sect : results ] , we will make use of three different witnesses , one each for the above criteria , in order to evaluate their behaviours under lossy transmission:@xmath111 the three @xmath112 s do nt satisfy the requirements for being a measure of entanglement . for example they do nt verify the basic axiom stating that a good measure should be equal to @xmath113 for any separable state @xcite . in summary , for a gaussian bi
partite state the three criteria ( see eqs .
( [ phscr ] ) , ( [ duancr ] ) , and ( [ eprcr ] ) ) reduce to@xmath114 however , once the state @xmath2 is entangled @xmath115 , @xmath116 and @xmath117 provide suitable markers for evaluating how far the state is from being separable , somehow measuring the robustness of the entanglement .
eventually we note that for _ diagonal _ fully symmetric states ( @xmath26 and @xmath36 in eq .
( [ cm initial ] ) ) @xmath115 , @xmath116 and @xmath117 read@xmath118and the two bounds ( @xmath119 ) for @xmath115 and @xmath120coincide while the bound for @xmath117 is @xmath121 so that the epr criterion is stricter than the phs and duan ones for any allowed value of @xmath18 .
the quality of an entangled state can be , indeed , evaluated , operatively , by looking at the effectiveness such a state could , in principle , give in quantum communication protocols . to this end
we use @xmath0 , the fidelity of cv teleportation of a coherent state , as a benchmark .
in fact , @xmath0 for a gaussian resource depends only on the entanglement quality of the resource itself . using as resource a gaussian two - mode state described by the * cm * in eq . * * ( [ cm initial ] ) , @xmath0 is given by @xcite@xmath122it is possible to show that a teleportation protocol of a coherent state , fully based on classical strategies , provides @xmath123 .
then , @xmath124 implies that the resource is entangled .
so that , the value of @xmath0 becomes an indicator of the quality of the entanglement .
we note that for _ diagonal _ states ( @xmath26 and @xmath36 ) @xmath125 . in this particular case
, perfect fidelity would be obtained for @xmath126 that would imply an unphysical * cm * so that gaussian resources , as the state produced by opos , can not guarantee perfect teleportation dellanno07 .
moreover , @xmath127 gives @xmath119 and coincides with both the duan and phs bounds ( see the end of the previous subsection ) for such _ diagonal _ states .
we note that , similarly to the duan ( eq .
( duancr ) ) but contrarily to the phs and epr criteria ( eqs .
( [ phscr ] ) and ( [ eprcr ] ) ) @xmath0 is not invariant under symplectic transformations . ) and teleportation fidelity ( eq .
( [ fidelity_th ] ) ) as an entanglement marker .
the _ light gray _ ( labelled with ( vi ) ) areas indicate un - physical * * cm**s ( _ i.e. _ violating inequality ( [ heis rel .
sympl invariant ] ) ) .
the different criteria show different regions of entanglement ( see text for details).,scaledwidth=48.0% ] as already mentioned , only the phs criterion can be written for a generic * cm * as a sufficient and necessary condition while the duan and the epr ones set only sufficient bounds . in order to discuss the relations between the different bounds and the teleportation fidelity , as an entanglement marker ,
it is possible to draw a region plot , similar to the one drawn in ref .
@xcite for a different case . in fig .
fig : criteria , we have visualized the different bounds set by the three entanglement witnesses of eq .
( [ witnesses ] ) ) and the region for which @xmath127 ( see eq .
( [ fidelity_th ] ) ) .
the plot has been computed considering a * cm * in the form of eq .
* * ( [ cm initial ] ) with @xmath107 ( balanced system ) .
the axes report the value of @xmath128 ( below ) and @xmath129 ( left ) , the two correlation terms of the covariance matrix are normalized to @xmath130 so that @xmath131 represents a pure maximally entangled state ( _ i.e. _ the state showing the maximum quantum correlation for a given total energy of the system ) .
we note that fully symmetric states , that we indicated as _ diagonal _ states , lay on the plot diagonal ( top left to bottom
these states , besides their symmetry , can be obtained at the end of a lossy propagation , described by eq .
( [ cm evolution ] ) , of an initially pure state . on the contrary ,
states outside the diagonal , having @xmath132 , can not be obtained by propagating pure states . as a matter of fact , cv entangled states produced by type
ii opo , show , in view of the symmetry in the interaction hamiltonian , @xmath133 .
this is not true for cv entangled states obtained by mixing the outputs of two independent type
i opos . in such a case the two fields have disjoint hamiltonians and the symmetry
is broken .
the _ light gray _ ( labelled with ( _ vi _ ) ) areas indicate un physical states _ i.e. _ * * cm**s violating inequality ( heis rel
. sympl invariant ) .
_ diagonal _ states satisfy the conditions ( see eq .
( duan_conditions ) ) for which the duan criterion becomes also necessary so that the coincidence between the duan and the phs bounds , along the diagonal , is not a surprise .
being both necessary and sufficient they coincide . for these _ diagonal _ states entanglement ( seen as non separability property ) implies @xmath127 .
so that entanglement is a pre
requisite for using the state as a resource for the teleportation of a coherent state .
there are two interesting regions in the plot that deserve some comments . the first one , encompassing areas labelled as ( _ iii _ ) and ( _ iv _ ) in the plot ( _ lights green _ and _ blue _ ) , represents states that violate the phs bound ( @xmath134 ) while they do not the duan one ( @xmath135 ) .
this apparent ambiguity can be solved noting that the * * cm**s relative to these regions do not respect the condition ( duan_conditions ) so that the non negativity of @xmath116 does not imply a separability of the state . on the other hand for such states
@xmath134 implies that they are effectively entangled and that their density matrix @xmath2 can not factorize into a convex combination of the tensor product of density operators relative to the two different sub
systems .
it is possible to see that if these * * cm**s are transformed by local squeezing operations , as outlined in ref .
@xcite , into a form that fulfills conditions ( [ duan_conditions ] ) , the transformed states show @xmath136 .
we have numerically done a few tests on such _ odd _ matrices , verifying that once taken into that form , the states violate the duan bound ( [ duan_necessary ] ) .
we note that , being the latter written in a more general form , it is more useful from the practical point of view .
the second interesting region , labelled with ( _ iv _ ) ( _ light green _ ) in fig .
[ fig : criteria ] , represents states that , although entangled , can not be used for teleporting coherent states . *
* cm**s lying inside this area will not give @xmath127 .
it is interesting to note that such states fall also inside the region for which the duan criterion ( [ duancr ] ) is not fulfilled . as above
mentioned , once the relative * * cm**s are transformed into the form ( [ cm_duan ] ) by local squeezing the transformed state will fulfill the duan criterion in the form ( [ duan_necessary ] ) so that , in this new scenario , the system will be entangled . at the same time , if this novel state is used as a quantum resource for teleportation of a coherent state it will give @xmath127 @xcite so that local squeezings unveil entanglement . the initial state lying in this area
is entangled for phs , being @xmath134 , but , from the point of view of teleportation , entanglement manifests itself in an useless way .
this entanglement can be made useful by locally transforming the two subsystems .
we can see that the epr criterion ( region ( _ i _ ) , _ light brown _ ) offers a more restrictive condition with respect to the other two criteria even for diagonal states .
region ( _ ii _ ) ( _ salmon _ ) represents the bound fixed by the duan criterion as a sufficient but not necessary condition . while region ( _ v _ )
( _ white _ ) represent separable states .
recent studies have shown that some particular classes of separable correlated states , traditionally considered classical , show quantum features useful for application in quantum technology @xcite .
in statistical theory any correlation between two random variables @xmath137 and @xmath138 can be measured by their _ mutual information _ defined by two equivalent expressions:@xmath139where @xmath140 is the shannon entropy ( the classical analogue of the @xmath78 , von neumann entropy defined in sect .
sect : purity ) associated to the statistical distribution of the variable @xmath141 ( @xmath142 indicating the joint probability distribution and @xmath143 the conditional distribution of @xmath90 given the value of @xmath144 ) .
these two definitions , translated into the quantum language by substituting @xmath145 with @xmath78 , the von neumann entropy , do not coincide anymore @xcite .
so , while the translation of the first one into the quantum language is straightforward and univocal , this is not true for the second one .
the first of the two definitions ( [ mutual_classical ] ) , is unambiguously referred to as the quantum mutual information @xcite between state @xmath8 and @xmath9 ( @xmath2 representing the state of the bi partite system as a whole)@xmath146where
@xmath147 $ ] are the partial traces .
it relies only on the quantum state of the whole system compared with the single sub systems states . in a bi
partite system , described by a density matrix @xmath2 , @xmath148 quantifies the total correlation between the subsystems @xmath149 and @xmath150 .
it is @xmath151 for entangled states , while it is strictly @xmath152 for separable systems .
it can be written in terms of the * cm * invariants ( see sect . [
sect : state ] ) and symplectic eigenvalues ( see eq .
( symplectic))@xmath153with @xmath154 given in eq .
( [ funct_f ] ) . the second definition of eq .
( [ mutual_classical ] ) , translated into the quantum world , necessarily involves the conditional state of a subsystem after a measurement performed on the other one .
so that , the symmetry between the two subsystems is broken .
since the conditional entropy @xmath155 requires to specify the state of @xmath156 given the state of @xmath137 , its definition , in quantum theory is ambiguous until the to be measured observables on @xmath137 are selected so that the conditional state of @xmath156 can be defined .
this discrepancy has lead to the concept of quantum discord @xmath157 @xcite
. a non zero @xmath158 signals the presence of quantum features in the correlation between the two sub
systems notwithstanding their separability or entangled nature . so that @xmath157 is a measure of genuine quantum correlation@xmath159where @xmath160 is the amount of genuinely classical correlation . for a gaussian state described by the * cm * ( [ cm initial ] ) ,
@xmath161 becomes @xcite @xmath162 we note that , as for the epr criterion , in quantum discord there is an asymmetry in the exchange of the two sub systems .
again this is due to the use of the concept of conditional states .
as shown the * cm * of a bi
partite state undergoing to a lossy transmission evolves as eq .
( [ cm post bs ] ) .
consequently the different quantum markers evolve as:@xmath163 } } ~ , \notag \\ \mathcal{f}_{t } & = & \frac{1}{\sqrt{1+\left ( m_{t}+n_{t}\right ) ^{2}+2\left ( c_{2,t}-c_{1,t}\right ) \left ( 1+m_{t}+n_{t}\right ) + 2\left ( m_{t}+n_{t}-2c_{1,t}c_{2,t}\right ) } } \notag \\ w_{phs , t } & = & n_{t}^{2}+m_{t}^{2}+2\left\vert c_{1,t}c_{2,t}\right\vert -4\left ( n_{t}m_{t}-c_{1,t}^{2}\right ) \left ( n_{t}m_{t}-c_{2,t}^{2}\right ) \notag \\ w_{duan , t } & = & \sqrt{\left ( 2n_{t}-1\right ) \left ( 2m_{t}-1\right ) } -\left ( c_{1,t}-c_{2,t}\right ) \notag \\ w_{epr , t } & = & n_{t}^{2}\left ( 1-\frac{c_{1,t}^{2}}{n_{t}m_{t}}\right ) \left ( 1-\frac{c_{2,t}^{2}}{n_{t}m_{t}}\right ) -\frac{1}{4 } \notag \\ \mathcal{i}_{t } & = & f\left ( n_{t}\right ) + f\left ( m_{t}\right ) -f\left ( d_{+,t}\right ) -f\left ( d_{-,t}\right ) \notag \\
\mathcal{d}_{t } & \mathcal{=}&f(m_{t})\mathcal{-}f\left ( d_{+,t}\right ) -f\left ( d_{-,t}\right ) + f\left ( \frac{n_{t}+2n_{t}m_{t}+2c_{1}c_{2}}{1 + 2m_{t}}\right ) \label{ev_markers}\end{aligned}\]]where the subscript @xmath164 indicates the quantity undergone to a lossy transmission described by eq.([cm post bs ] ) .
we note that the vacuum state obtained for @xmath165 is a pure one _ i.e. _ @xmath166 .
moreover , in the ideal case ( no loss ) , the opo would generate a pure state as well .
being @xmath167 , for @xmath168 , the evolution of @xmath169 is not monotonic .
the purity of the composite system can not be considered a general entanglement marker @xcite . as a matter of fact
, any pair of physical systems in a pure state have @xmath170 even if they are disentangled . on the other hand , in our specific case , having a precise hypothesis on the ideal state ( a pure twin
beam diagonal state ) outing the opo crystal allows us to consider @xmath82 as a measure of the total decoherence that has affected the state .
it is easy to see that @xmath171 , @xmath172 , and @xmath173 describe properties very robust under decoherence .
once @xmath127 , @xmath134 and @xmath136 for @xmath174 they will keep breaking the respective bounds for every level of loss .
both mutual information and quantum discord show , with respect to loss , the same feature even if decoherence affects their amount . on the contrary a state that show epr
like correlation ( @xmath175 ) for @xmath174 will not keep this property along the propagation so that some particular protocol based on this property can not be reproduced . under a total loss greater than @xmath176 any state looses this correlation property
the transmission over a gaussian channel is described by eq .
( fokker - planck ) .
as already mentioned this evolution is in all equivalent to a fixed amount of loss introduced by a fictitious beam
the actual experimental apparatus is made of the cv entangled state source , a variable attenuator ( mimicking the bs ) , and a state characterization stage .
a block diagram of the experimental setup is presented in fig .
fig : setup the state source relies on a cw internally frequency doubled nd : yag laser ( innolight diabolo ) whose outputs @532 nm and @1064 nm are respectively used as the pump for a non degenerate opo and the local oscillator ( lo ) for the homodyne detector .
the opo is set to work below the oscillation threshold and it provides at its output two entangled thermal states ( the signal @xmath94 and the idler @xmath177 ) .
the opo is based on an @xmath178-cut periodically poled ktp non linear crystal ( ppktp , _ raicol crystals ltd_. on custom design ) which allows implementing a type ii phase matching with frequency degenerate and cross polarized signal and idler beams , for a crystal temperature of @xmath179 53@xmath180c .
the transmissivity of the cavity output mirror , @xmath181 , is chosen in order to guarantee , together with crystal losses ( @xmath182 ) and other losses mechanisms ( @xmath183 ) , an output coupling parameter @xmath184 @1064 nm of @xmath179 @xmath185 , corresponding to an experimental linewidth of @xmath186 mhz @1064 nm . in order to obtain a low oscillation threshold ,
opo cavity geometry is set to warrant simultaneous resonance on the pump , the signal and the idler @xcite .
measured oscillation threshold is around @xmath187 mw ; during measurement runs the system has been operated at @xmath17960% of the threshold power to avoid unwanted non gaussian effects @xcite .
the two beams outing the opo are transmitted through a filter of variable optical density that mimics the bs .
the loss level introduced by the filter is polarization independent and can be tuned from a few percent up to more than 99% .
the signal and idler modes are then sent to the covariance matrix measurement set - up : this consists in a preliminary polarization system , that allows choosing the beam to be detected and a standard homodyne detector .
the polarization system is made of an half - wave plate ( @xmath188 ) followed by a polarizing beam splitter ( pbs ) ; the different wave - plate orientations allow choosing the beam to be transmitted by the pbs : the signal ( @xmath94 ) , the idler ( @xmath177 ) or their combinations @xmath189 or @xmath190 .
two other combinations , @xmath191 and @xmath192 , can be obtained by inserting before the pbs an additional quarter wave plate ( @xmath193 ) @xcite .
acquisition times are considerably short thank to pc - driven mechanical actuators that allow setting the @xmath188 and @xmath193 positions in a fast and well calibrated manner .
once a beam is selected , it goes to an homodyne detector put downstream the pbs .
this exploits , as local oscillator , the laser output @1064 nm , previously filtered and adjusted to match the geometrical properties of the opo output : a typical interferometer visibility is @xmath194 .
the lo phase @xmath195 is spanned to obtain a @xmath196 variation in @xmath197 ms . in order to avoid low frequency noise
the homodyne current is demodulated at @xmath198=3 mhz and low - pass filtered ( @xmath156=@xmath199 khz ) .
then it is sampled by a pci acquisition board obtaining @xmath200pts / run , with 14-bit resolution .
the electronic noise floor is 16 dbm below the shot noise level , corresponding to snr @xmath17940 .
data are analysed by a mathematica routine that extract from the six homodyne traces the relevant quadrature variances necessary for reconstructing the whole covariance matrices @xcite .
we have performed different sets of measurement under lossy transmission in order to investigate different loss regimes .
each experimental * cm * comes from seven homodyne traces : the shot noise calibration trace , obtained by obscuring the opo output , six traces each for one of the six modes required for a full state characterization .
contrarily to other previous experiments , where quantum tomographic routine were used in order to retrieve experimental * * cm**s @xcite , we have evaluated * * cm**s by a simpler mathematica routine that calculates relevant second order moments of homodyne distributions in a faster way without enhancing the experimental indeterminacy on the * cm * elements .
we have tested on a few * * cm**s that this procedure gives results in all compatible with quantum tomography .
we have also checked , with the standard procedure outlined in @xcite , that the states under scrutiny were effectively gaussian . once a * cm * is obtained the different entanglement witnesses ( @xmath201 , @xmath116 , and @xmath117 eqs .
( [ witnesses ] ) ) , state purity ( @xmath82 eq .
( [ purity ] ) ) , teleportation fidelity ( @xmath0 eq .
( fidelity_th ) ) , quantum discord ( @xmath161 eq .
( [ discord ] ) ) , and mutual information ( @xmath202 eq . ( [ quantum_mutual ] ) ) are calculated .
then , the overall decoherence , _
i.e. _ the total level of loss that includes opo cavity escape efficiency , propagation loss , filter absorption , homodyne efficiency , is assigned as a label to the measurement @xcite .
this expected level of decoherence is then compared to the theoretical one obtained by inverting eq .
( [ cm post bs ] ) and solving for @xmath203 under the condition @xmath204 ; thus requiring that @xmath205 represents a pure state ( see the discussion at the end of sect .
[ sect : state ] ) .
we have verified that , even if experimental * * cm**s do not reproduce exactly _ diagonal _ states , all of them respect the duan conditions ( [ duan_conditions ] ) within experimental indeterminacies .
so that for the analysed matrices the duan witness @xmath116 represents a sufficient and necessary condition for entanglement . in all the reported plots we have considered the less decohered datum ( obtained for @xmath206 ) as a reference so that all the reported theoretical curves are obtained imposing that eqs .
( [ cm_elements_evolution ] ) and ( [ ev_markers ] ) evaluted for @xmath206 give the measured values .
the total losses we have measured span the interval @xmath207 ( @xmath208 ) . we note that @xmath206 , in absence of extra loss and having a cavity escape efficiency of @xmath209 , implies an overall state detection efficiency of @xmath210 in agreement with an homodyne visibility of @xmath211 , a photodiode ( nominal ) efficiency of @xmath212 and residual transmission loss between the opo output mirror and the detector surface of @xmath213 . in eq .
( [ cm post bs ] ) .
as expected the correlation reduces linearly with @xmath203 . the full ( _ dark orange _ )
line represents the linear behaviour calculated starting from the first experimental point we have measured ( @xmath206 ) .
error bars are smaller than data points and amount to @xmath214.,scaledwidth=48.0% ] in fig .
[ fig : c ] we report the behaviour vs. @xmath203 of the averaged correlation term ( @xmath215 see eq .
( [ cm_elements_evolution ] ) ) .
as expected the correlation between the two sub systems degrades linearly with the total loss ( @xmath72 ) .
the expected behaviour ( full _ dark orange _ line ) , obtained by considering the less absorbed * cm * ( @xmath216 ) as a reference , follows quite well the reported data .
actually , data refer to acquisition taken on different days so that , the scattering of the point around that line is more due to source long term dynamics then to actual deviation from the lindblad model . at the same time
the fact that the points are reasonably close to that line proves that the long term stability of the source can be considered quite good .
vs. @xmath203 . the full ( _ dark orange _ )
line represents the expected behaviour calculated by the third of eqs .
( [ ev_markers ] ) setting the first experimental point at @xmath206 as the intial datum .
error bars , obtained by propagating the experimental indeterminacies in eq .
( [ witnesses]a ) , range between @xmath217 and @xmath218 . in the inset
we report the high loss regime ( @xmath219 ) for better enlightning the un
separability , as witnessed by @xmath115 , even in presence of strong decoherence.,scaledwidth=48.0% ] vs. @xmath203 . the full ( _ dark orange _ ) line represents the expected behaviour calculated by the fourth of eqs .
( [ ev_markers ] ) setting the first experimental point at @xmath206 as the intial datum .
error bars , obtained by propagating the experimental indeterminacies in eq .
( [ witnesses]b ) , range between @xmath220 and @xmath221 . in the inset
we report the plot for @xmath219 in order to better visualize the persistence of entanglement , as witnessed by @xmath116 , even in presence of strong decoherence.,scaledwidth=48.0% ] as already mentioned , @xmath115 and @xmath116 describe a physical property of the state that is strong under decoherence as proved for lower loss ( below 90% ) in ref .
they are symptoms of un separability , in the sense that the system state can not be described by a density matrix in the form of eq .
( [ sepst ] ) @xcite .
this can be seen in figs .
[ fig : phs ] and [ fig : duan ] where @xmath115 and @xmath116 are plotted vs. @xmath203 .
we have also enlarged , in the insets the region for strong loss ( @xmath219 ) to prove that , even if the analysed state is very close to a two mode vacuum ( the total average number of photon ( @xmath222 ) reduces to @xmath223 ) it is still experimentally possible to prove that the state is non separable .
it has to be noted that , while , for @xmath72 , @xmath115 approaches its classical limit non linearly ( see the third of eqs .
( [ ev_markers ] ) ) , @xmath116 ( see the fourth of eqs .
( [ ev_markers ] ) ) is linear .
thus , in the very high loss regime it becomes more reliable to assess entanglement using the latter than the former .
vs. @xmath203 . the full ( _ dark orange _ )
line represents the expected behaviour calculated by the fifth of eqs .
( [ ev_markers ] ) setting the first experimental point at @xmath206 as the intial datum .
error bars , obtained by propagating the experimental indeterminacies in eq .
( [ witnesses]c ) , range between @xmath224 and @xmath225 .
they are considerably larger for point at low losses . as expected for total losses larger than @xmath226 @xmath227 and the state does not show _ epr _
correlation.,scaledwidth=48.0% ] @xmath175 indicates that the state exhibits epr like correlation so that it is possible to gain information on the expectation value of one observable on one sub system with a precision better than the standard quantum limit once the epr companion is measured on the other sub system .
this feature is by far the most fragile under decoherence : for @xmath228 no state can keep this quantum feature .
this can be understood from the fact that loss , a stochastic process , affects directly the degree of correlation between the two sub systems while the system representation ( _ i.e. _ its un separability ) is only smoothed by this process .
it is relevant to note that @xmath229 also corresponds to the minimum state purity @xmath82 . so that , loosing the epr character coincides with the maximum mixedness for the state during its propagation . in fig .
[ fig : epr ] we report the experimental behaviour found for @xmath117 for our system . measured * * cm**s for @xmath228 all show @xmath227 .
a positive @xmath117 indicates that , for these states , any attempt to gain information on one sub system by measuring the other would result less precise than the standard quantum limit . vs. @xmath203 . the full ( _ dark orange _ )
line represents the expected behaviour calculated by the second of eqs .
( [ ev_markers ] ) setting the first experimental point at @xmath206 as the intial datum .
error bars , obtained by propagating the experimental indeterminacies in eq .
( [ fidelity_th ] ) , range between @xmath217 and @xmath220 . in the inset we report the plot for @xmath219 in order to underline the persistence of a quantum teleportation regime even in presence of strong decoherence ( high loss ) thus proving that even for , in principle , infinite distance this class of states would allow to perform the teleportation of a coherent state with a fidelity above @xmath230.,scaledwidth=48.0% ] an important signature for an entangled cv state is its ability of acting as a quantum resource in the cv teleportation protocol for coherent state . in eq .
( [ fidelity_th ] ) we have expressed the fidelity @xmath0 as a function of the * cm * elements .
@xmath0 , as @xmath115 and @xmath116 , represents a robust signature of quantum properties for the state undergoing to a lossy transmission .
in particular , in fig .
[ fig : fid ] , we see that even in the high loss regime , @xmath0 remains above the classical limit of @xmath226 ( see the inset for greater details ) . thus proving that cv entangled state , as the one produced by our source , could be used as resource for realising teleportation protocol of coherent state , in principle , at an infinite distance .
eventually we have retrieved , from our * * cm**s , the value for the quantum mutual information @xmath231 ( eq . (
[ mutual ] ) ) and quantum discord @xmath232 ( eq . ( [ discord ] ) ) . and @xmath161 vs.@xmath203 . the full ( _ dark orange _ ) and
dashed ( _ blue _ ) lines represent the expected behaviours calculated by the sixth and seventh of eqs .
( ev_markers ) setting the first experimental point at @xmath206 as the intial datum .
error bars , obtained by propagating the experimental indeterminacies in eqs .
( [ mutual ] ) and ( [ discord ] ) , respectively , range between @xmath233 and @xmath225 for @xmath202 and @xmath217 and @xmath234 for @xmath161 . in the inset
we report the @xmath161 data for @xmath219 in order to underline the persistence of true quantum correlation even in presence of strong decoherence ( high loss ) .
note that the data for @xmath202 scatter more from the expected behaviour may be signalling extra classical correlations.,scaledwidth=48.0% ] in fig .
[ fig : discord ] we report the experimental data together with the expected behaviours , as usually calculated considering the less decohered matrix as a reference , for @xmath202 and @xmath161 vs. @xmath203 .
as it can be seen the quantum discord follows very well its _ theoretical _ line while quantum mutual information is a little more scattered around it .
moreover , our data prove that even in presence of strong decoherence , it is possible to evaluate that @xmath161 keeps @xmath151 , within the experimental indeterminacies , all the way down to an highly absorbed state .
we note thatgaussian quantum discord is attracting , very recently , a lot of experimental interest @xcite . in particular , in ref .
@xcite , the authors give an operational significance to quantum discord as the possibility of encoding quantum information in separable states . in ref .
@xcite the optimal strategy for evaluating @xmath235 in homodyne measurement is presented .
it is interesting to compare our experimental plot with the one reported in ref .
@xcite where the authors analyse the quantum discord under the lossy transmission of one of the two sub systems .
we note that in their case the scattering of the experimental points around the theoretical curve is almost equivalent for @xmath202 and @xmath161 while in our case there is a clear difference .
gaussian bi partite states are one of the most renown resources for implementing cv quantum communication protocols such as cv teleportation of coherent states . in this paper
we have experimentally analysed how decoherence affects different entanglement criteria and quantum markers for a cv bi
partite state outing a sub threshold type
the decoherence is experimentally introduced by transmistting the quantum state through a variable attenuator . before illustrating our experimental results we have discussed in details the relationship between the three different entanglement criteria used in the cv framework and linked them to the teleportation fidelity and quantum discord .
the latters represent two possible quantum signatures for evaluating the ability of this class of states in quantum communication protocols .
on one hand , our findings prove that the lindblad approach for describing lossy transmission is valid all the way down to strongly decohered states . on the other hand , with this paper ,
we prove that the particular class of states we have analysed keeps , within the experimental indeterminacies , its main quantum signatures , _ i.e. _ the possibility of realizing quantum teleportation of coherent states with a fidelity above @xmath226 and a quantum discord above @xmath113 for a total loss of @xmath236 .
this proves that the class of cv entangled states , we analysed , would allow , in principle , to realize quantum teleportation over an infinitely long gaussian channel . in analysing how quantum discord ( see fig .
[ fig : discord ] ) and quantum mutual information behave under decoherence we interestingly found that the scattering of the points around the theoretical curve is significantly more evident for the quantum mutual information may be signalling that a key role , in our case , is played by unexpected classical correlations .
this point will be subject of further theoretical and experimental investigation .
wheeler , j. a. , and w. h. zurek , `` quantum theory and measurement '' , princeton university press , princeton ( usa ) , 1984 ; a. a. clerk , m. h. devoret , s. m. girvin , florian marquardt , and r. j. schoelkopf , rev .
phys . * 82 * , 1155 ( 2010 ) ; v. dauria , s. fornaro , a. porzio , s. solimeno , s. olivares , and m.g.a .
paris , phys .
* 102 * , 020502 ( 2009 ) ; d. buono , g. nocerino , v. dauria , a. porzio , s. olivares , and m. g. a. paris , j. opt .
b , * 27 * , a110 ( 2010 ) ; borivoje daki , yannick ole lipp , xiaosong ma , martin ringbauer , sebastian kropatschek , stefanie barz , tomasz paterek , vlatko vedral , anton zeilinger , aslav brukner , and philip walther , nat .
physics * 8 * , 666 ( 2012 ) ; virginia dauria , alberto porzio , salvatore solimeno , stefano olivares , and matteo g a paris , j. opt .
b : quantum semiclass . opt .
* 7 * s750 ( 2005 ) ; alberto porzio , virginia dauria , salvatore solimeno , stefano olivares , and matteo g.a .
paris , int .
j. .quantum information , * 5 * , 63 ( 2007 ) ; s. l. w. midgley , a. j. ferris , and m. k. olsen , phys .
a * 81 * , 022101 ( 2010 ) ; d. j. saunders , s. j. jones , h. m.wiseman , and g. j. pryde , nat . physics
* 6 * , 845 ( 2010 ) ; _ corrigendum _ in nat .
physics * 7 * , 918 ( 2011 ) ; devin h. smith , geoff gillett , marcelo p. de almeida , cyril branciard , alessandro fedrizzi , till j. weinhold , adriana lita , brice calkins , thomas gerrits , howard m. wiseman , sae woo nam , and andrew g. white , nat
. communications * 3 * , 625 ( 2012 ) ; a. j. bennet , d. a. evans , d. j. saunders , c. branciard , e. g. cavalcanti , h. m. wiseman , and g. j. pryde , phys .
x * 2 * , 031003 ( 2012 ) ; vitus hndchen , tobias eberle , sebastian steinlechner , aiko samblowski , torsten franz1 , reinhard f. werner , and roman schnabel , nat .
photonics * 6 * , 596 ( 2012 ) ; m. j. holland , m. j. collett , d. f. walls , and m. d. levenson , phys
a * 42 * , 2995 ( 1990 ) , j .- ph .
poizat , j .- f .
roch and p. grangier , ann .
fr . , * 19 * , 265 ( 1994 ) ; f. e. harrison , d. f. walls , opt . comm . ,
* 123 * , 331 ( 1996 ) ; hai wang , yun zhang , qing pan , hong su , a. porzio , changde xie , and kunchi peng , phys . rev .
82 * , 1414 ( 1999 ) ; virginia dauria , antonino chiummo , martina de laurentis , alberto porzio , and salvatore solimeno , opt .
express * 13 * , 948 ( 2005 ) ; v. dauria , c. de lisio , a. porzio , s. solimeno , javaid anwar , and m. g. a. paris , phys .
a * 81 * , 033846 ( 2010 ) ; | quantum properties are soon subject to decoherence once the quantum system interacts with the classical environment . in this paper
we experimentally test how propagation losses , in a gaussian channel , affect the bi
partite gaussian entangled state generated by a sub threshold type ii opticla parametric oscillator ( opo ) .
experimental results are discussed in terms of different quantum markers , as teleportation fidelity , quantum discord and mutual information , and continuous variable ( cv ) entanglement criteria .
to analyse state properties we have retrieved the composite system covariance matrix by a single homodyne detector .
we experimentally found that , even in presence of a strong decoherence , the generated state never disentangles and keeps breaking the quantum limit for the discord .
this result proves that the class of cv entangled states discussed in this paper would allow , in principle , to realize quantum teleportation over an infinitely long gaussian channel . |
dynamics in the solar corona takes on a wide range of forms . on one hand ,
the corona is the setting for the most violent eruptions in the solar system : solar flares and coronal mass ejections @xcite . on the other , coronal
heating makes the corona almost a thousand times hotter than the photosphere , even in the quiet sun @xcite .
@xcite unified these two phenomena by proposing that micro- and nano - flares , less energetic cousins of eruptive flares , heat the corona .
this model gained credence from studies showing that solar flares exhibit power law statistics @xcite over a wide range of scales for many quantities .
[ see @xcite for a review . ] in addition , stellar flares have similar light curves to solar flares @xcite and also exhibit power law statistics @xcite , suggesting that the physics of the solar corona is generic to sun - like stars .
coronal dynamics remains an active research area @xcite .
details of the eruption process including how magnetic energy is stored , how eruptions onset , and how the stored energy is converted to other forms are still open questions .
in addition , while micro- and nano - flares are believed to be a major contributor to coronal heating , the authors know of no theory which explains why the coronal temperature and density have the values they have , as opposed to larger or smaller values . in this paper
, we propose that the condition of the corona is regulated by magnetic reconnection @xcite , a dynamical process which converts magnetic energy into kinetic energy and heat and energizes particles . magnetic energy is stored during collisional ( slow ) reconnection , which has been shown to drive the coronal plasma toward lower collisionality @xcite .
if the plasma becomes marginally collisionless , a bifurcation in the underlying dynamics of reconnection occurs @xcite .
this bifurcation , which occurs when two length scales @xmath0 and @xmath1 ( to be defined below ) are comparable , catastrophically initiates fast ( hall ) reconnection , releasing the stored energy in the form of an eruption .
the condition of marginal collisionality , therefore , sets an upper bound on how hot the coronal plasma can be for a given density . the continual driving toward lower collisionality of the pre - flare corona by slow reconnection enforces the self - organization of the corona to a state of marginal collisionality where @xmath2 .
we present finer details of this process below .
then , we perform the first observational test of this model using a large sample of data from stellar flares on sun - like stars .
we find that @xmath0 and @xmath1 are comparable for every event in the sample , indicating that stellar coronae do self - organize into a marginally collisional state .
magnetic reconnection depends strongly on the collisionality of the plasma .
collisional ( sweet - parker ) reconnection @xcite is exceedingly slow . the thickness @xmath0 of the sweet - parker diffusion region is given by @xcite @xmath3 where @xmath4 is the alfvn speed , @xmath5 is the strength of the reconnecting magnetic field , @xmath6 is the ion mass , @xmath7 is the density , @xmath8 is the resistivity , and @xmath9 is the length of the sweet - parker diffusion region in the outflow direction .
the normalized reconnection rate @xmath10 is @xmath11 for coronal parameters , where @xmath12 is the inflow speed .
collisionless ( hall ) reconnection has a reconnection rate of the order of 0.1 @xcite , six orders of magnitude faster than sweet - parker reconnection for coronal parameters .
recent studies @xcite showed that the transition from collisional to collisionless reconnection is catastrophic , occurring when @xmath0 becomes smaller than the ion gyroradius @xmath1 . at this scale , magnetohydrodynamics ( mhd ) breaks down and the hall effect ( absent in mhd ) allows reconnection to be fast @xcite .
the transition can be described as a bifurcation @xcite which takes stable equilibria out of existence as a control parameter ( @xmath13 ) varies .
the relevant gyroradius @xmath1 for anti - parallel reconnection is the ion inertial scale @xmath14 @xcite , @xmath15 where @xmath16 is the ion cyclotron frequency , @xmath17 is the ion plasma frequency , and @xmath18 is the ion charge . for reconnection with a ( guide ) magnetic field along the current sheet , the relevant gyroradius becomes @xmath19 , where @xmath20 is the sound speed @xcite . to see how magnetic reconnection self - organizes the corona , consider an active region . before an eruption
, the plasma can not be collisionless : if it were , the stored magnetic energy would be rapidly released by hall reconnection .
therefore , the pre - flare active region must be collisional .
since ( collisional ) sweet - parker reconnection is exceedingly slow , magnetic energy can be stored .
[ by `` collisional '' , we mean @xmath21 , which , using eqs .
( [ deltaspdef ] ) and ( [ didef ] ) , is equivalent to @xmath22 where @xmath23 is the ion - electron collision frequency .
therefore , reconnection is collisional when the ion transit time along the sweet - parker diffusion region is longer than the ion - electron collision time .
see also @xcite . ]
when sweet - parker reconnection begins in the corona , _
e.g. , _ as a result of two coronal flux tubes coming together , the reconnecting magnetic field @xmath5 is initially much weaker than the strong asymptotic magnetic field in the core of the flux tube , _
i.e. , _ the reconnection is embedded within a wider current sheet . from equation ( [ deltaspdef ] )
, the thickness of the diffusion region will be relatively wide , so that @xmath24 .
it was shown @xcite that embedded sweet - parker reconnection spontaneously self - drives the current sheet to thinner scales , even without external forcing .
this is because the reconnection inflow convects stronger magnetic fields into the diffusion region , which causes @xmath0 to decrease [ see eq .
( [ deltaspdef ] ) ] .
thus , the reconnection process itself self - drives the system toward lower collisionality .
if the asymptotic field is strong enough so that @xmath25 , _ i.e. , _ the system becomes marginally collisionless , then a bifurcation causes hall reconnection to begin , eruptively releasing the stored energy .
[ we note in passing that if @xmath5 in an active region is not strong enough to ever satisfy @xmath26 for a given density and temperature then no eruption occurs , potentially providing an observational constraint on which active regions erupt and which do not . ] after the eruption , the corona returns to a collisional state , and the process begins again .
the continual self - driving of the corona toward lower collisionality keeps coronal parameters near the critical condition where the bifurcation occurs ( @xmath2 ) .
we propose that this process regulates the temperature of the corona .
if the temperature @xmath27 of the corona is larger than the critical value , the spitzer resistivity @xmath8 is smaller ( since @xmath28 ) . from equation ( [ deltaspdef ] ) ,
a smaller @xmath8 allows a smaller @xmath5 to initiate an eruption . as such , less magnetic energy is stored and released , and the corona cools . alternately ,
if the corona is cooler than the critical value , a larger magnetic field is required to set off an eruption .
more magnetic energy is stored and released , increasing the temperature .
either way , the temperature is driven back to the critical value .
@xcite independently proposed a similar model based on the density , which we discuss in sec .
[ sec - disc ] .
observational verification of this model entails confirming that @xmath0 and @xmath1 are comparable at fast reconnection onset .
laboratory experiments @xcite are consistent with this condition , but direct observations are impossible because the length scales are not resolvable .
indirect verification is possible by estimating @xmath0 and @xmath1 for coronal parameters . for a solar active region ,
one finds both length scales to be a few meters , as has been noted previously @xcite .
an important question is whether this agreement is indicative of a general mechanism or is just a coincidence for solar parameters .
we use data from a recent study @xcite which analyzed 134 eruptive flares from 44 stars of spectral type f , g , k , and m , using the deep survey / spectrometer instrument on the _ extreme ultraviolet explorer _ ( _ euve _ ) satellite .
the data come from a single instrument on a single satellite , so there are no spurious variations due to different instrumental characteristics .
see @xcite for a thorough discussion of the data .
the @xmath18-folding time @xmath29 for the flare signal to decay and the emission measure @xmath30 were extracted from flare light curves .
because of interruptions in _ euve _ data during some flares , @xcite presented the @xmath29 values for each flare as a unique value , a range of values or an upper bound .
we retain events having a particular value or a range ( using the average ) and omit events given as bounds , leaving 107 events from 37 stars .
post - flare parameters ( the temperature @xmath27 , density @xmath7 , minimum magnetic field @xmath31 , and length @xmath32 and cross sectional area @xmath33 of coronal loops ) were derived @xcite from @xmath29 and @xmath30 using an approach due to @xcite , which assumes that @xmath29 is comparable to the radiative and conductive cooling times .
taking @xmath34 , one finds @xcite @xmath35 where @xmath36 , and @xmath37 are constants ( in cgs units ) . a lower bound for the magnetic field @xmath31
is estimated by requiring the magnetic pressure @xmath38 to be at least as large as the gas pressure @xmath39 , where @xmath40 is boltzmann s constant , to maintain a coronal loop . as a test of the model , @xcite surveyed the literature for independent measurements of @xmath41 , and @xmath31 for the stars in their study , finding that 178 of 212 measurements were consistent with the haisch model .
this justifies treating the derived parameters as valid independent of the haisch model .
we first verify that the haisch model gives reasonable results for solar parameters .
eruptive solar flares have @xmath42 sec and @xmath43 @xcite .
using these values , the haisch model predicts post - flare parameters of @xmath44 mk , @xmath45 @xmath46 , @xmath47 cm , and @xmath48 g. compact solar flares have @xmath49 sec and @xmath50 @xmath51 @xcite . using these values
, the haisch model gives @xmath52 mk , @xmath53 @xmath51 , @xmath54 cm , and @xmath55 g. these ranges of @xmath56 and @xmath31 are consistent with independent empirical values obtained from images and x - ray data for flaring loops in the sun @xcite . to calculate @xmath0 and @xmath14 from the haisch model , we use @xmath31 for the upstream magnetic field and @xmath7 for the density .
sweet - parker current sheets extend to system scales @xcite , so @xmath9 is on the order of the coronal loop radius @xmath57 , consistent with the haisch model .
finally , we use @xmath27 to calculate the spitzer resistivity @xcite @xmath58 where @xmath59 is the electron mass and @xmath60 $ ] is the coulomb logarithm .
use of this formula is justified because the electron mean free path ( @xmath61 km for solar conditions , where @xmath62 is the electron thermal speed and @xmath63 is the electron - ion collision frequency ) is small compared to length scales in the outflow direction ( @xmath64 km ) and along the current sheet ( @xmath65 km ) .
the result of comparing @xmath0 to @xmath14 using the stellar flare data is plotted in fig .
[ spvsdi ] .
representative solar values based on @xmath66 sec and @xmath67 for eruptive flares ( @xmath68 cm and @xmath69 cm ) and @xmath70 sec and @xmath71 for compact flares ( @xmath72 cm and @xmath73 cm ) are plotted as the asterisk and plus , respectively .
a dashed line with slope of unity is plotted .
the agreement is extremely good .
a least squares analysis gives a best fit slope of @xmath74 with a correlation coefficient of 0.981 .
it is encouraging that the slope of the line in fig .
[ spvsdi ] is consistent with unity .
however , there are ambiguities in the data analysis .
for example , we used @xmath14 as the critical length scale , whereas @xmath75 is more applicable to the corona ( but more difficult to estimate ) .
these scales differ by a factor of @xmath76 , where @xmath77 is the ratio of gas pressure to total magnetic pressure . if @xmath78 in the corona , this introduces a factor of a few .
the present analysis does not intend to distinguish between the two gyroradii ; rather , the results demonstrate that @xmath0 is within a factor of a few of the critical length scale @xmath1 in active stellar coronae . a caveat of the result in fig .
[ spvsdi ] pertains to how the parameters are derived in the haisch model . using equations ( [ deltaspdef ] ) , ( [ didef ] ) , and ( [ spitzer ] ) , using @xmath79 , and eliminating @xmath5 by defining the ratio of gas pressure to magnetic pressure in the reconnecting magnetic field as @xmath80
, we find @xmath81 .
treating @xmath82 as a fixed parameter and eliminating @xmath83 and @xmath32 using equation ( [ haisch ] ) gives @xmath84 where @xmath85 is a constant .
the slow dependence on @xmath86 significantly suppresses scatter in the observational data when evaluating @xmath87 .
however , the magnitude of @xmath87 is unconstrained by the haisch model , so the slope of the line in fig .
[ spvsdi ] being of order unity is significant .
furthermore , since the data obtained using the haisch model agrees with independent determinations of the same quantities from other studies @xcite , it is reasonable to assert that data obtained independently from the haisch model would fall close to the same line .
we can avoid suppression of the scatter in the data by solving equation ( [ dspodiandtaudem ] ) for @xmath29 and taking a logarithm of both sides .
this yields @xmath88 , where @xmath89 using a value of @xmath90 , which is representative of the stellar data in our study . if @xmath91 , this predicts a linear relationship between @xmath92 and @xmath93 , with @xmath94 being the @xmath95-intercept .
the stellar data are plotted in fig .
[ fgkstars ] .
the gray boxes show the range of values for eruptive and compact flares on the sun @xcite . assuming @xmath91 and taking @xmath82 to be of order unity
, the predicted line is plotted . while the data do not fall on a line , the line predicted by the hypothesis that @xmath26 does pass through the data . to see why this is significant , note that if @xmath0 was , say , @xmath96 ( at @xmath97 m , still a very small length scale compared to coronal loop radii )
, then @xmath94 would be @xmath98 instead of @xmath99 and the line in fig .
[ fgkstars ] would lie 32 units higher , orders of magnitude removed from the data .
the hypothesis that @xmath91 brings significant ordering to the data .
we note that the theory predicts @xmath91 at flare onset , while the haisch model refers to post - flare conditions .
following an eruption on the sun , temperatures typically increase by a factor of a few @xcite , while the density increases due to chromospheric evaporation by at least a factor of 10 [ compare pre - flare data @xcite with post - flare data @xcite ] . from the relation above equation ( [ dspodiandtaudem ] ) , @xmath100 . for an increase in @xmath7 by a factor of 10 and @xmath27 by a few ,
@xmath87 does not change appreciably . assuming this is true for other stars ,
if @xmath91 after a flare , it is also true before a flare .
the data analyzed in this paper pertain to flares in sun - like stars , but the underlying dynamics of reconnection is general .
our model applies equally well to micro- and nano - flares in the quiet corona . using values for the quiet sun of @xmath101 1 mk , @xmath102 @xmath51 ,
@xmath103 5 g , and @xmath104 cm , we find @xmath105 cm and @xmath106 cm , in agreement with the model .
the present result may have important implications for self - organized criticality ( soc ) models of the solar corona .
soc occurs in driven , dissipative systems when the system is driven to a critical state where it undergoes a major reconfiguration @xcite .
soc leads to power law statistics , which encouraged @xcite to propose the corona undergoes soc .
subsequent studies of soc in the corona exist @xcite , but a firm physical foundation of the mechanism for self - driving and the physical condition setting the critical state is often traded for the ease of performing cellular automaton simulations [ see @xcite for a review ] .
the present result provides a physical mechanism for self - driving ( embedded sweet - parker reconnection ) and the critical state ( marginal collisionality ) , which may provide an avenue for developing quantitative predictions of soc to compare with coronal observations .
an alternate mechanism @xcite for heating the solar corona uses a change in density to achieve self - regulation . after an eruption , chromospheric evaporation increases the coronal density , decreasing the ion gyroradius [ eq . ( [ didef ] ) ] and making subsequent eruptions more difficult .
the extent to which uzdensky s and our mechanisms regulate coronal heating is an open question .
the present model assumes that sweet - parker scaling is appropriate for thin current sheets of large extent .
long current sheets are known to fragment due to secondary instabilities , but the effect of this on the reconnection rate is unknown .
verification of the present model would entail testing whether sweet - parker reconnection in extended current sheets remains much slower than hall reconnection .
[ see @xcite for further discussion of this point as well as other future research directions . ] | there is currently no explanation of why the corona has the temperature and density it has .
we present a model which explains how the dynamics of magnetic reconnection regulates the conditions in the corona .
a bifurcation in magnetic reconnection at a critical state enforces an upper bound on the coronal temperature for a given density .
we present observational evidence from 107 flares in 37 sun - like stars that stellar coronae are near this critical state
. the model may be important to self - organized criticality models of the solar corona . |
investigations on the spin - dependent transport in a magnetic tunnel junction ( mtj ) have attracted much interest in the last decade , as the mtjs have essential applications in spintronic devicesprinz , wolf , zutic , maekawa , su .
one of phenomena in mtjs is the so - called tunnel magnetoresistance ( tmr ) effect , which states that the tunneling current through the junction depends sensitively on the relative orientation of the magnetizations of both ferromagnetic electrodes , that is caused by the spin - dependent scattering of conduction electrons .
the difference between the currents through spin up and down channels in mtjs is usually referred to as the spin current@xcite . on the other hand , a reverse effect to tmr
was predicted independently by slonczewski@xcite and berger@xcite , i.e. , the spin - polarized electrons passing from the left ferromagnetic layer into the right layer in which the magnetization deviates the left by an angle may exert a torque to the right ferromagnet and can change the orientation of its magnetization .
this effect was coined as the spin transfer effect ( for review , see e.g. ref.@xcite ) , which can lead to the current - induced magnetization reversal , and might offer a promise for the current - controlled spintronic devices .
it has recently attracted intensive investigations both experimentallytsoi , myers , katine , wegrowe , sun , asamitsu and theoreticallywaintal , brataas , zhu , tserkovnyak , x.waintal , brouwer , mu , parcollet . among other magnetic tunnel structures ( e.g. refs.@xcite ) , the ferromagnet - quantum dot - ferromagnet ( fm - qd - fm )
coupled systems have also received much attention recently .
previous theoretical works on spin - dependent transport through qds are mainly focused on the tunnel electrical current and the tmr effect for collinear configurationbarnas , fransson , zhang , bulka , lopez , martinek , pedersen and noncollinear configurationsergueev , konig , zhu1,braun , rudzinski , wetzels , fran1,fran2,weymann . however , the investigations on the spin current and the current - induced spin transfer torque ( cistt ) in such systems are sparsely reported .
it is the purpose of this paper to study the spin current and spin transfer torque in fm - qd - fm coupled mtjs . in terms of keldysh s nonequilibrium green function method , the tunnel electrical current , the spin current and the spin transfer torque in the fm - qd - fm coupled system will be investigated .
it is found the spin current exhibits different behaviors from its electrical counterpart .
it is also shown that the resonant positions for the tunneling electrical current and spin current become far separated with the increase of the coulomb interaction @xmath0 in the qd .
the magnitudes of the cistt at two resonant positions are found to be greatly enhanced at resonant positions . at the bias larger than @xmath1 ,
the cistt can reach a saturation plateau which is independent of the coulomb interaction .
the cistt shows a kink - like behavior with the increase of the spin current .
the relationship between the cistt , the electrical current and the spin current is also addressed .
the rest of this paper is outlined as follows . in sec .
ii , the model and formalism for the tunnel electrical current , the spin current and the current - induced spin transfer torque will be established . in sec .
iii , the corresponding numerical results will be given .
finally , a summary will be presented .
let us consider a single - level qd coupled to two ferromagnetic electrodes .
the left ( l ) and right ( r ) electrodes are connected with the bias voltage @xmath2 and @xmath3 , respectively , as shown in fig .
1 . the magnetic moment @xmath4 of the left fm is assumed to be parallel to the @xmath5 axis , while the moment @xmath6 of the right fm is aligned along the @xmath7 axis which deviates the @xmath5 axis by a relative angle @xmath8 . the tunnel current flows along the @xmath9 axis and perpendicular to the junction plane .
the system consists of the left , right , qd and coupling parts , and can be described by the following hamiltonian : @xmath10 with @xmath11 @xmath12a_{kr\sigma } ^{\dag } a_{kr\sigma } -\mathbf{m}_{r}\sin \theta a_{kr\sigma } ^{\dag } a_{kr\bar{\sigma}},\ ] ] @xmath13 @xmath14 where @xmath15 is the single - electron energy for the wavevector @xmath16 and spin @xmath17 in the @xmath18 electrode , @xmath19 and @xmath20 are annihilation operators of electrons with spin @xmath17 in the @xmath18 electrode and the qd , respectively , @xmath21 , @xmath0 represents the on - site coulomb interaction between electrons in the qd , and @xmath22 is the coupling matrix elements between the @xmath18 electrode and the qd .
( @xmath23 ) stands for the coupling matrix between the @xmath24 electrode and the qd , and both magnetizations are aligned by a relative angle @xmath25 . ]
the tunnel electrical current is composed of the sum of the currents carried by spin up and down electrons:@xmath26while the spin current is defined by the difference between the electrical currents through the spin up and down channels@xcite : @xmath27where@xmath28with @xmath29 the lesser green function . by applying the langrenth theorem and fourier transform ,
we may obtain the following equation@xmath30 , \label{sel - l}\]]where @xmath31 @xmath32 @xmath33 , @xmath34 , @xmath35 and @xmath36 are the retarded and lesser green functions of the qd , respectively , and @xmath37 is the fermi distribution function in the @xmath18 electrode . by defining @xmath38\gamma _ { l}(\varepsilon )
$ ] , we can simplify @xmath39 as @xmath40where @xmath41 is a @xmath42 matrix , @xmath43 with @xmath44 .
the lesser green function @xmath45 can be calculated by the keldysh equation @xmath46 . to get @xmath47 , we invoke ng s ansatz@xcite : @xmath48 , where @xmath49 $ ] , @xmath50 , @xmath51 $ ] , @xmath52 , and @xmath53 . under these considerations
, we finally get @xmath54where @xmath55 @xmath56 is also a @xmath42 matrix .
consequently , the tunnel current and spin current have the forms of@xmath57with @xmath58 the pauli matrix . the spin torque exerting on the right ferromagnet is determined by the time evolution rate of the total spin in the right ferromagnetslonczewski , berger , which is composed of two parts : one is caused by the spin - dependent potential that is known as the equilibrium torque , the other is the current - induced spin transfer torque caused by the tunnel hamiltonian @xmath59 . by means of the nonequilibrium green functions ,
the cistt exerting on the right ferromagnet can be obtained by@xcite @xmath60+\sin \theta \re e\underset{kr}{\sum } \int \frac{d\varepsilon } { 2\pi } % tr[g_{kr}^{<}(\varepsilon ) \hat{\sigma}_{3}t_{kr}^{\ast } ] , \label{torque}\]]where @xmath61 . by a treatment similar to eq .
( [ sel - l ] ) , we can obtain @xmath62.\]]therefore , the cistt can be rewritten as @xmath63.\]]the remaining task is to calculate the retarded green function @xmath64 . by the equation of motion
, we can derive @xmath65where @xmath66 is a unit matrix , @xmath67 @xmath68 @xmath69 and @xmath70substituting these equations into eq .
( [ eqn - motion ] ) , up to the third - order of green functions , we arrive at @xmath71 , \label{self - con}\]]where @xmath72 @xmath73 @xmath74,@xmath75@xmath76 , \\
\tilde{g}_{kl}^{r-1}g_{kl}^{2(2)r}(\varepsilon ) & = & t_{kl}^{\ast } g^{(2)r}(\varepsilon ) + \underset{k^{\prime } \alpha } { \sum } t_{k^{\prime } \alpha } ^{\ast } [ g_{klk^{\prime } \alpha } ^{3(3)r}(\varepsilon ) + g_{klk^{\prime } \alpha } ^{4(3)r}(\varepsilon ) ] , \\
g_{\bar{k}l}^{r-1}g_{kl}^{3(2)r}(\varepsilon ) & = & t_{kl}[g^{r}(\varepsilon ) -g^{(2)r}(\varepsilon ) ] + \underset{k^{\prime } \alpha } { \sum } [ -t_{k^{\prime } \alpha } g_{klk^{\prime } \alpha } ^{5(3)r}(\varepsilon ) + t_{k^{\prime } \alpha } ^{\ast } g_{klk^{\prime } \alpha } ^{6(3)r}(\varepsilon ) ] , \end{aligned}\]]with @xmath77 @xmath78 and the third - order green functions defined by : @xmath79it is worthily mentioning that one usually takes the hatree - fock decoupling approximation up to the second - order for green functions in most systems .
such a decoupling scheme is not adequate for the present single - level qd coupled system , as it could smear some characteristic features that come from electronic correlations . in order to extract more information from the many - body interactions
, we should consider the equation of motion up to the third - order of green functions @xmath80 in eq .
( [ self - con ] ) . in above derivations , we have invoked the following decoupling approximations for green functions@xcite : @xmath81 @xmath82 @xmath83g^{\sigma ^{\prime } \sigma , r}(\varepsilon ) .$]as we ignore the spin - flip scatterings in the present system , the off - diagonal elements associated with different spins of the third - order green function are sent to zero , namely , the higher - order spin correlations in the ferromagnetic leads are neglected . with these results ,
we may obtain @xmath84 , \\
g_{k , q}^{3(2)r}(\varepsilon ) & = & -t_{kl}g_{\bar{k}\alpha } ^{r}[g^{(2)r}(\varepsilon ) -\bar{f}_{l}g^{r}(\varepsilon ) ] , \end{aligned}\]]with @xmath85 and @xmath86 .
similarly , a more straightforward but somewhat complicated calculation gives rise to the following equations @xmath87% \hat{\sigma}_{1}g^{r}(\varepsilon ) \notag \\ & & + t_{kr}\left [ \cos ^{2}\frac{\theta } { 2}g_{kr}^{r}(\varepsilon ) + \sin ^{2}% \frac{\theta } { 2}g_{\bar{k}r}^{r}(\varepsilon ) \right ] g_{q , q}^{(2)r}(% \varepsilon ) \end{aligned}\]]@xmath88\hat{% \sigma}_{1}g^{r}(\varepsilon ) \notag \\ & & -t_{kr}\left [ \cos ^{2}\frac{\theta } { 2}\tilde{g}_{kr}^{r}(\varepsilon ) % \bar{f}_{r}+\sin ^{2}\frac{\theta } { 2}\tilde{g}_{\bar{k}r}^{r}(\varepsilon ) f_{r}\right ] g^{r}(\varepsilon ) \notag \\ & & + \frac{1}{2}t_{kr}\sin \theta \lbrack \tilde{g}_{kr}^{\uparrow , r}(\varepsilon ) -\tilde{g}_{kr}^{\downarrow , r}(\varepsilon ) ] \hat{\sigma}% _ { 1}g^{(2)r}(\varepsilon ) \notag \\ & & + t_{kr}\left [ \cos ^{2}\frac{\theta } { 2}\tilde{g}_{kr}^{r}(\varepsilon ) + \sin ^{2}\frac{\theta } { 2}\tilde{g}_{\bar{k}r}^{r}(\varepsilon ) \right ] g^{(2)r}(\varepsilon ) , \end{aligned}\]]@xmath89 g^{(2)r}(\varepsilon ) \notag \\ & & + t_{kr}^{\ast } \left [ \sin ^{2}\frac{\theta } { 2}g_{kr}^{r}(\varepsilon ) f_{r}+\cos ^{2}\frac{\theta } { 2}g_{\bar{k}r}^{r}(\varepsilon ) \bar{f}_{r}% \right ] g^{r}(\varepsilon ) , \end{aligned}\]]where @xmath90 and @xmath91 . on the other hand , the second - order retarded green s function can be expressed as of the following form@xmath92@xmath93where @xmath94 \\ & & -\underset{kr}{\sum } t_{kr}t_{kr}^{\ast } \frac{1}{2}\sin \theta \left [ \tilde{g}_{kr}^{\uparrow , r}(\varepsilon ) -\tilde{g}_{kr}^{\downarrow , r}(\varepsilon ) \right ] \hat{\sigma}_{1 } \\ & & -\underset{kr}{\sum } t_{kr}t_{kr}^{\ast } \left [ \cos ^{2}\frac{\theta } { 2}% \tilde{g}_{kr}^{r}(\varepsilon ) + \sin ^{2}\frac{\theta } { 2}\tilde{g}_{\bar{k}% r}^{r}(\varepsilon ) \right ] \\ & & -\underset{kr}{\sum } t_{kr}t_{kr}^{\ast } \left [ \cos ^{2}\frac{\theta } { 2}% g_{\bar{k}r}^{r}(\varepsilon ) + \sin ^{2}\frac{\theta } { 2}g_{\bar{k}% r}^{r}(\varepsilon ) \right ] , \end{aligned}\ ] ] @xmath95% \hat{\sigma}_{1 } \\ & & + \underset{kr}{\sum } t_{kr}t_{kr}^{\ast } \frac{1}{2}\sin \theta \lbrack \tilde{g}_{kr}^{\downarrow , r}(\varepsilon ) f_{r}(\varepsilon _ { kr\uparrow } ) -\tilde{g}_{kr}^{\uparrow , r}(\varepsilon ) f_{r}(\varepsilon _ { kr\downarrow } ) ] \hat{\sigma}_{1 } \\ & & -\underset{kr}{\sum } t_{kr}t_{kr}^{\ast } \left [ \cos ^{2}\frac{\theta } { 2}% \tilde{g}_{kr}^{r}(\varepsilon ) \bar{f}_{r}+\sin ^{2}\frac{\theta } { 2}\tilde{% g}_{\bar{k}r}^{r}(\varepsilon ) f_{r}\right ] \\ & & -\underset{kr}{\sum } t_{kr}t_{kr}^{\ast } \left [ \cos ^{2}\frac{\theta } { 2}% g_{\bar{k}r}^{r}(\varepsilon ) \bar{f}_{r}+\sin ^{2}\frac{\theta } { 2}% g_{kr}^{r}(\varepsilon ) f_{r}\right ] .\end{aligned}\ ] ] by combining eq .
( [ 3rd - order ] ) with eq .
( [ self - con ] ) , we get @xmath96 from this equation , the retarded green function @xmath64 can be obtained .
we would like to point out that the averaged values involved in eq .
( self - con ) should be obtained self - consistently by@xmath97these above - mentioned equations establish the fundamental basis for numerically investigating the spin - dependent transport properties of the fm - qd - fm coupled mtjs .
without losing the generality , in the following numerical calculations we may further suppose that the two ferromagnets are made of the same materials , i.e. , @xmath98 , where @xmath99 is the polarization of the left ( right ) ferromagnet .
then , we can introduce @xmath100 , where @xmath101 @xmath102 will be taken as an energy scale .
the tunnel matrix elements @xmath103 is presumed to take values at the fermi level .
we will take @xmath104 and @xmath105 as scales for the currents and corresponding differential conductance , respectively . for different polarization @xmath106 in parallel ( a ) and antiparallel ( b ) configurations of magnetizations .
the parameters are taken as @xmath107 , @xmath108 , @xmath109.,width=340,height=340 ] the bias dependences of the tunnel electrical current and the differential conductance @xmath110 in the fm - qd - fm coupled system have been reexamined within the present scheme .
it is found that the tunnel electrical current exhibits two - step features , which corresponds to the resonant tunneling of electrons through the qd at energy levels @xmath111and @xmath112 .
the corresponding differential conductance as a function of the bias thus exhibits resonant peaks , where the large peak at a lower voltage corresponds to the discrete level @xmath111 , and a small resonant peak corresponds to the level @xmath112 .
it is uncovered that with the increase of the interaction @xmath0 , the peaks of @xmath113 corresponding to the discrete level @xmath112 shift towards the higher voltage side without changing the peak amplitude for both parallel and antiparallel configurations .
this is because an increase of @xmath0 makes two discrete levels @xmath111 and @xmath112become far apart , thereby leading to that the resonant positions of main and charging peaks of @xmath113 are far _ _ _ _ separated .
these observations are quite consistent with the previous studies ( see e.g. ref.@xcite ) .
the bias dependence of the differential conductance for different polarizations in the parallel and antiparallel configurations of magnetization is shown in fig .
it can be found that in the parallel configuration there is only slight changes for different polarizations except for @xmath114 , as displayed in fig .
. we can understand this property from the point of view of the resonant tunneling through two barrierssergueev , namely , the two barriers are of the same height , and the resonant probability is unity in this case , that is independent with @xmath106 . the small changes with @xmath106 observed in fig .
2(a ) is caused by the interaction @xmath0 .
however , it is quite different for the case with full polarization @xmath114 , which comes from the fact that the conductance for @xmath114 has only one channel .
it is interesting to note that with increasing @xmath106 in the case of parallel alignment , although the main peaks at the bias corresponding to the energy level @xmath111 are suppressed , the charging peaks at the bias corresponding to the level @xmath112are enhanced . in this case ,
more electrons from the left enter into the qd through the channel @xmath112 to tunnel into the right lead.for the antiparallel configuration , it can be seen that the main peaks of the conductance decrease as the polarization @xmath106 increases , while the charging peaks remain almost intact , as shown in fig .
2(b ) , that is different from the case of parallel configuration , and is due to the conventional spin - valve effect in mtjs . for @xmath114 ,
the conductance becomes zero , which is nothing but a perfect spin - valve effect .
( a ) and spin differential conductance @xmath115 ( b ) for different angle @xmath25 at @xmath116 .
the other parameters are taken the same as in fig .
2.,width=340,height=340 ] the bias dependence of the spin current @xmath117 and its corresponding spin differential conductance defined by @xmath118 for different angles is shown in fig .
it can be seen that , by changing the relative angle @xmath8 , the spin current can change its magnitude even its sign . at @xmath119 , when @xmath120 , the spin current @xmath117 shows step - like behaviors , and @xmath121 ; while for @xmath122 , @xmath117 exhibits behaviors similar to a basin - like shape , and @xmath123 at the bottom of the `` basin '' .
this shows that the effect of the relative orientation of magnetizations on the spin current is more obvious . at two resonant positions corresponding to levels @xmath111 and @xmath112
, @xmath117 begins to sharply change .
it is clearly seen in fig .
3(b ) that at @xmath124 , the spin conductance shows sharp resonant peaks and dips at the levels @xmath111 and @xmath112 , although for @xmath125 and @xmath126 the resonance corresponding to @xmath111 shows a peak , while for @xmath127 , @xmath128 and @xmath129 the first resonance shows a sharp dip .
the resonances corresponding to the level @xmath112 exhibit peaks for all @xmath8 . at @xmath125 ,
the two resonant peaks of @xmath130 have comparable weights ; at @xmath131 , the first peak has a small weight , while the second peak has a large weight ; at @xmath127 , the first dip of @xmath117 has a small amplitude , while the other peak has a large amplitude ; at @xmath132 , the resonant dip and peak have almost the same amplitude ; and at @xmath133 , the first dip has a larger amplitude than the peak .
evidently , the spin current and spin conductance have no spin - valve effect , showing that the spin current and spin conductance have quite different characteristics from their electrical counterparts . for different polarization @xmath106 in parallel ( a ) and antiparallel ( b ) configurations of magnetizations .
the other parameters are taken the same as in fig .
2.,width=340,height=340 ] the step - like and basin - like behaviors of the bias dependence of the spin current for different polarizations in parallel and antiparallel configurations are presented in figs .
4(a ) and ( b ) , respectively . for @xmath134 , as there is no polarization in electrodes , the contribution of electrons with spin up is the same as that of spin down , leading to the spin current vanishes . in the case of parallel alignment ( @xmath125 ) , with increasing the polarization @xmath106 , the contribution of spin up
is increased while that of spin down is decreased , thus resulting in the magnitude of the spin current increased , as shown in fig .
. the step - like behavior of @xmath117 against @xmath135 comes from the resonant tunneling .
the situation becomes a bit complicated in the case of antiparallel alignment ( @xmath133 ) .
when @xmath136 and @xmath137 , the spin current is zero , because for the former @xmath138 while for the latter @xmath139 , as shown in fig .
4(b ) . for @xmath140 and @xmath137 , with increasing the bias voltage the spin current decreases steeply at the resonant position corresponding to the qd energy level @xmath111 , and then keeps a constant until the other resonant position corresponding to the level @xmath141 where @xmath117 increases sharply , leading to the spin current shows a basin - like behavior . the larger the polarization @xmath106 @xmath142 , the deeper the bottom of the basin .
these results demonstrate that the spin current @xmath117 has quite different behaviors in parallel and antiparallel configurations . in parallel ( a ) and antiparallel ( b ) configurations for different coulomb interaction @xmath0 , where @xmath116 , and the other parameters
are taken the same as in fig .
2.,width=340,height=340 ] fig .
5 shows the bias dependence of spin current in collinear configurations for different coulomb interaction @xmath0 .
it can be found that the overall qualitative behavior of @xmath117 against @xmath143 looks similar to that shown in fig .
4 . with increasing @xmath0 , in the parallel alignment the second plateau of
the spin current @xmath117 becomes wider , as shown in fig .
5(a ) ; while in the antiparallel alignment apart from that the bottom of the `` basin '' becomes deeper , the width of the `` basin '' also becomes wider , as shown in fig .
this is again from the fact that an increase of @xmath0 makes the resonant positions corresponding qd discrete levels @xmath111 and @xmath141 become far separated . for different polarization @xmath106 ( a ) and different interaction @xmath0 ( b ) at the bias voltage @xmath144 .
the other parameters are taken the same as in fig .
2.,width=340,height=340 ] the angular dependence of the cistt for different polarization and coulomb interaction is shown in fig .
it is seen that the cistt versus the angle @xmath145 shows a sine - like behavior , consistent with the finding in ref.slonc . at @xmath125 and @xmath129 ,
i.e. , the magnetizations of the two ferromagnetic electrodes are collinearly aligned , the cistt vanishes , which is obvious as the spin torque @xmath146 , where @xmath147 and @xmath148 are the spin moments of the left and right ferromagnets . with increasing the polarization @xmath106 , the magnitude of the cistt
is enhanced , as displayed in fig .
this is in agreement with the statement that the torque is proportional to the polarization of the other ferromagnet@xcite . with the increase of coulomb interaction @xmath0 , the magnitude of the cistt
is decreased , as seen in fig .
this can be understood in such a way that the contribution of the tunneling through the discrete level @xmath149 is decreased with increasing the interaction @xmath0 , thus leading to a decrease of the spin transfer torque .
( a ) and @xmath0 ( b ) at @xmath150 @xmath151 .
the other parameters are taken the same as in fig .
2.,width=340,height=340 ] the bias dependences of the cistt for different polarizations and coulomb interactions at a given angle @xmath131 are shown in figs .
7(a ) and ( b ) .
it is seen that with increasing the bias voltage , the cistt first increases slowly to a sharp peak , then decreases dramatically to almost zero , and after undergoing a wider flat , it increases suddenly to another plateau .
the position of the first sharp peak in the curve of cistt versus @xmath152 is observed to be independent of the polarization as well as the coulomb interaction , showing that it is a resonant peak at the resonant position @xmath153 .
the second sharp increase of the cistt against @xmath143 is also from the resonance at the qd discrete energy level @xmath112 . from fig .
7(a ) , one may find that with increasing the polarization the magnitudes of the cistt at the two resonant positions increase .
this is because the transmission coefficient of electrons is proportional to the polarization@xcite , while the cistt depends on the tunneling electrical current that is determined by the transmission coefficient .
_ _ _ _ thus , an increase of polarization would enhance the cistt at the resonant positions .
the effect of the coulomb interaction @xmath0 on the cistt is shown in fig .
at the first peak , it is independent of @xmath0 , suggesting that it is a resonance at the level @xmath111 ; while with increasing @xmath0 , the second resonant positions move to higher voltages , showing that the second resonance takes place at the level @xmath112 .
it is interesting to note that at a given polarization the cistt goes to a saturation plateau at the bias larger than @xmath1 , which is independent of the coulomb interaction . from fig .
7 , we could see that the cistt can be remarkably enhanced at the qd discrete energy levels which may be adjusted by changing the gate voltage . at @xmath154 @xmath155 and @xmath116 .
the other parameters are taken the same as in fig .
2.,width=340,height=340 ] what is the relationship between the cistt , the electrical current and the spin current in this fm - qd - fm coupled system ?
the answer is presented in fig . 8 for different @xmath0 at @xmath156 @xmath157 . with increasing the electrical current
, the cistt goes up linearly to a round peak , then decreases slowly to a broad minimum , and then increases sharply , as shown in fig .
the first round peak is owing to the resonant tunneling at the level @xmath111 ; while the second sharp increase is caused by the resonant tunneling at the level @xmath112 .
the spin current @xmath158 is found to increase non - monotonically with increasing the electrical current @xmath159 , as illustrated in fig .
this is quite different from the spin current in fm - sc - fm tunnel junctions where the spin current is proportional to the injection current@xcite . in the curves of @xmath130 against @xmath159 ,
the two resonances can be clearly seen .
the cistt as a function of @xmath117 shows a kink - like behavior , as displayed in fig .
8(c ) . with increasing @xmath117
, the cistt first shows a linear behavior , then suddenly drops to nearly zero , and then increases almost linearly again . this kink - like behavior
is caused by the resonant tunneling at the levels @xmath153 and @xmath112 , which manifests the same thing as in fig .
_ _ _ away from the kink , the cistt appears to be proportional to the spin current .
the reason is that the electrons with opposite spins from the left ferromagnetic electrode entering into the right electrode exert the torques with opposite directions on the spins of the right electrode , giving rise to that the magnitude of the cistt is determined by the difference between the magnitudes of torques exerted by electrons with spin up and spin down .
considering that the spin current is defined by the difference between the electrical currents of spin up and spin down , while the spin transfer torque is proportional to the electrical current@xcite , the proportionality between the cistt and the spin current is conceivable . _ _ _ _ it is seen that for different @xmath0 the curves in fig .
8 are almost unchanged , showing that the coulomb interaction has less effect on the cistt versus @xmath159 and @xmath117 , as well as @xmath130 versus @xmath159 .
this is because the coulomb interaction @xmath0 only changes the resonant positions .
we have investigated the spin - dependent transport properties in a fm - qd - fm coupled system by means of the nonequilibrium green functions .
it has been found that by changing the relative orientation of both magnetizations , the spin current can change its magnitude even its sign . for positive bias voltages ,
when the relative orientation angle is less than @xmath160 , the spin current shows a step - like behavior , and the spin current is positive ; when the relative orientation angle is greater than @xmath160 , the spin current behaviors similar to a basin - like shape , and is negative at the bottom of the `` basin '' . with increasing the coulomb interaction @xmath0 in the qd , it has been uncovered that the resonant positions for the tunneling electrical current and spin current become far separated .
the cistt is observed to first increase slowly to a sharp peak with increasing the bias voltage , then decrease dramatically to almost zero ; and after undergoing an unchanged stage , it increases suddenly to another plateau .
such a behavior is obviously resulted from the resonant tunneling through the central qd .
the magnitudes of the cistt at two resonant positions are observed to increase with increasing the polarization of the ferromagnet . at the bias larger than @xmath1
, the cistt is seen to reach a saturation plateau which is independent of the coulomb interaction .
in addition , it has been demonstrated that the cistt as a function of the spin current shows a kink - like behavior , and away from the kink , the cistt increases almost linearly with increasing the spin current .
the cistt as a function of the electrical current shows a non - monotonic behavior , while the spin current is found to be nonlinearly proportional to the electrical current . besides
, the coulomb interaction is shown to have less effect on the behaviors of the cistt as functions of the tunneling electrical current and spin current , as well as the spin current as a function of the electrical current .
finally , we would like to mention that the cistt can be greatly enhanced when the bias voltage meets with the discrete levels of the qd at two resonant positions . while the energy level of the qd can be adjusted by tuning the gate voltage , the resonant property of the cistt offers an alternative premise to develop a spintronic device through the current - controlled magnetization reversal effect . we have benefitted from discussions with b. jin , z. c. wang , and z. g. zhu .
this work is supported in part by the national science foundation of china ( grants nos .
90403036 , 20490210 , 10247002 ) and by the chinese academy of sciences . | based on keldysh s nonequilibrium green function method , the spin - dependent transport properties in a ferromagnet - quantum dot ( qd)-ferromagnet coupled system are investigated .
it is shown the spin current shows quite different characteristics from its electrical counterpart , and by changing the relative orientation of both magnetizations , it can change its magnitude even sign .
the current - induced spin transfer torque ( cistt ) is uncovered to be greatly enhanced when the bias voltage meets with the discrete levels of the qd at resonant positions .
the relationship between the cistt , the electrical current and the spin current is also addressed . |
consider a billiard on the plane with @xmath1-periodic rectangular obstacles as in figure [ fig : windtree ] . in @xcite , it is shown that for almost all parameters @xmath2 of the obstacle ( i.e. , lenghts @xmath3 of the sides of the rectangular obstacles ) , for almost all initial directions @xmath4 , and for any starting point @xmath5 the billiard trajectory @xmath6 escapes to infinity with a rate @xmath7 ( unless it hits the corner ) : @xmath8 note that changing the height and the width of the obstacle ( see figure [ fig : different : billiards ] ) we get quite different billiards , but this does not change the diffusion rate .
the number `` @xmath9 '' here is the lyapunov exponent of a certain renormalizing dynamical system associated to the initial one .
more precisely , it is the lyapunov exponent of certain subbundle of the hodge bundle under the kontsevich zorich cocycle .
the lyapunov exponents of the hodge bundle also govern the deviation spectrum for the asymptotic cycle of an orientable measured foliation , as well as the rate of convergence of ergodic averages for interval exchange transformations and for certain area preserving flows on surfaces see @xcite , @xcite , @xcite .
the range of phenomena where the lyapunov exponents of the hodge bundle are extremely helpful keeps growing : nowadays , it includes , in particular , the evaluation of volumes of the moduli spaces of quadratic differentials on @xmath10 , see @xcite , and the classification of commensurability classes of all presently known non - arithmetic ball quotients @xcite . in this paper
we develop the study of the lyapunov spectrum of invariant subbundles of the hodge bundle under the kontsevich zorich cocycle ( with respect to general @xmath0-invariant measures ) .
we revisit variational formulas of g. forni from @xcite for the hodge norm interpreting them in more geometric terms and we generalize them to invariant subbundles .
we generalize forni s formulas for partial sums of the lyapunov exponents of the kontsevich
zorich cocycle in terms of geometric characteristics of the hodge bundle ( theorem [ thm : partialsum ] in [ ssec : formulas ] ) .
we establish the reduciblity of the second fundamental form with respect to any decomposition into hodge star - invariant , hodge orthogonal subbundles and generalize the kontsevich formula for the sum of all non - negative exponents ( corollary [ cor : kzformgen ] in [ ssec : reduc ] ) .
we investigate the occurrence of zero exponents with a particular emphasis on the relation between the central oseledets subbundle and the kernel of the second fundamental form .
our main theorem in this direction ( theorem [ th : ann ] in [ ssec : central ] ) establishes sufficient conditions for the inclusion of one into the other , hence for their equality .
we illustrate our conclusions with two examples .
the first model case , inspired by recent work of a. eskin , m. kontsevich and a. zorich @xcite , is given by arithmetic teichmller curves of _ square - tiled cyclic covers _ ( introduced in @xcite ) ; the second model case is a certain @xmath0-invariant locus @xmath11 ( inspired by a paper of c. mcmullen @xcite ) supporting an @xmath0-invariant ergodic probability measure with some zero exponents .
the study of square - tiled cyclic covers was motivated by the discovery of two arithmetic teichmller curves of square - tiled cyclic covers with maximally degenerate kontsevich zorich spectrum ( see @xcite , @xcite and @xcite ) .
conjecturally there are no other @xmath0-invariant probability measures with maximally degenerate spectrum .
progress on this conjecture has been made by m. mller @xcite in the case of teichmller curves and , recently , by d. aulicino @xcite in the general case .
a conditional proof of the conjecture in sufficiently high genera can also be derived from a quite explicit formula for the sum of all non - negative exponents ( see @xcite ) .
the case of square - tiled cyclic covers is especially rigid . in this case , the central oseledets subbundle is always @xmath0-invariant , smooth and in fact coincides with the kernel of the second fundamental form .
this pictures does not hold in general .
in fact , in our second model case the central oseledets subbundle does not coincide with the kernel of the second fundamental form even though it has the same rank .
we emphasize that the rank of the kernel of the second fundamental form at generic points can be explained in all examples by symmetries ( automorphisms ) of the underlying surfaces . indeed , the discovery of the two above - mentioned maximally degenerate examples was based on a symmetry criterion for the vanishing of the second fundamental form on the complement of the tautological subbundle ( see @xcite , @xcite and [ symmetry ] of this paper ) .
this article is organized as follows . in 2 , we introduce the kontsevich
zorich cocycle on the hodge bundle over the moduli space of abelian differentials and we compute the relevant geometric tensors of the bundle , endowed with the hodge hermitian product , namely , the second fundamental form and the curvature of the hermitian connection with respect to the gauss - manin connection .
we then prove first and second variational formulas for the hodge norm in terms of the second fundamental form and of the curvature . in 3 ,
we derive formulas for partial sums of the exponents of the restriction of the kontsevich
zorich cocycle to any @xmath0-invariant subbundle of the hodge bundle . in
4 we investigate the presence of zero exponents and we prove results on the relation between the central oseledets subbundle of the cocycle and the kernel of the second fundamental form . in appendix [ s :
rk : b : cyclic : covers ] we describe the case of ( arithmetic ) curves of square - tiled cyclic covers .
finally , in appendix [ s : rk : b : z ] , we present our second model case .
conjecturally , this second example is representative of the general features related to the presence of zero exponents on invariant subbundles of the hodge bundle .
the moduli space of abelian differentials @xmath12 has the structure of a complex vector bundle over the moduli space @xmath13 of riemann surfaces of genus @xmath14 .
the fiber over a point of @xmath13 represented by a riemann surface @xmath15 corresponds to the complex @xmath14-dimensional vector space of all holomorphic 1-forms @xmath16 on @xmath15 .
the space @xmath12 admits a natural action of the group @xmath17 ( see , for instance , @xcite or @xcite for an elementary description of this action ) .
it is well - known that the orbits of the diagonal subgroup @xmath18 of @xmath17 project on @xmath13 to the geodesics with respect to the teichmller metric .
for this reason the flow on @xmath12 defined by the action of the diagonal subgroup is called the _ teichmller geodesic flow_. the real ( respectively complex ) hodge bundle @xmath19 ( respectively @xmath20 ) over the moduli space @xmath13 is the vector bundle having the first cohomology @xmath21 ( respectively @xmath22 ) as its fiber over a point represented by the riemann surface @xmath15 . by identifying the lattices @xmath23 ( respectively @xmath24 ) in the fibers of these vector bundles it is possible to canonically identify fibers over nearby riemann surfaces
this identification is called the _ gauss
manin _ connection .
let us consider now the pullback of the hodge bundle to @xmath12 with respect to the natural projection @xmath25 .
we can lift the teichmller geodesic flow to the hodge bundle by parallel transport of cohomology classes with respect to the _ gauss manin _ connection , thus getting a cocycle @xmath26 called the _ kontsevich - zorich cocycle_. the lyapunov exponents of this cocycle were proved to be responsible for some fine dynamical properties of flows on individual riemann surfaces , see @xcite , @xcite , which motivated the study of these exponents for all known teichmller flow - invariant ( and especially @xmath0-invariant ) ergodic measures on @xmath12 . in order to analyze the lyapunov spectrum ( i.e. , collection of lyapunov exponents ) of @xmath26
, we need to understand the evolution of cohomology classes @xmath27\in h^1(s,\mathbb{r})$ ] under @xmath26 .
a particularly useful tool for this task is the _
hodge norm _
, the main object of the next subsection . the natural pseudo - hermitian intersection form on the complex cohomology @xmath28 of a riemann surface @xmath15 can be defined on any pair @xmath29 of complex - valued closed @xmath30-forms on @xmath15 representing cohomology classes in @xmath28 as @xmath31 restricted to the subspace @xmath32 of holomorphic @xmath30-forms
, the intersection form induces a positive - definite hermitian form ; restricted to the subspace @xmath33 of anti - holomorphic @xmath30-forms , it induces a negative - definite hermitian form , so on the entire complex cohomology space the pseudo - hermitian form has signature @xmath34 . for later use ,
we define ( with the aid of the hodge norm ) @xmath35 that is , @xmath36 is the moduli space of _ unit area _
abelian differentials on riemann surfaces of genus @xmath14 .
the _ hodge representation theorem _ affirms that for any cohomology class @xmath37 in the _ real _ cohomology space @xmath21 of a riemann surface @xmath15 there exists a unique holomorphic form @xmath38 such that @xmath37 is the cohomology class of the closed 1-form @xmath39 in @xmath40 . by the hodge representation theorem
, the positive - definite hermitian form on @xmath41 induces a positive - definite bilinear form on the cohomology @xmath21 : for any @xmath42 , @xmath43 , @xmath44 the hodge bundle @xmath45 is thus endowed with an inner product , called the _ hodge inner product _ and a norm , called the _ hodge norm_. given a cohomology class @xmath46 , let @xmath38 be the unique holomorphic @xmath30-form such that @xmath47 $ ] .
define @xmath48 to be the real cohomology class @xmath49 $ ] .
the hodge norm @xmath50 satisfies @xmath51 or , in other words , @xmath52 is the value of @xmath53 on the fundamental cycle .
the operator @xmath54 on the real cohomology @xmath21 of a riemann surface @xmath15 is called the _ hodge star operator_. we will denote the hodge inner product of cycles @xmath55 by round brackets : @xmath56 , and their symplectic product by angular brackets : @xmath57 . by definition ,
the spaces @xmath32 and @xmath33 are hodge - orthogonal , hence the following relations hold : @xmath58 consider the complex hodge bundle @xmath59 over the moduli space @xmath13 of complex structures having the complex cohomology space @xmath22 as a fiber over the point of @xmath13 represented by a riemann surface @xmath15 . this complex @xmath60-dimensional vector bundle is endowed with the flat gauss
manin connection @xmath61 which preserves the hermitian form of signature @xmath34 .
the bundle @xmath62 admits a decomposition into a direct sum of two orthogonal subbundles @xmath63 with respect to the hermitian form .
this decomposition is _ not _ invariant with respect to either the flat connection on @xmath62 or with respect to the teichmller flow .
the decomposition defines an orthogonal projection map @xmath64 of the vector bundles @xmath65 .
the subbundle @xmath41 is a hermitian vector bundle with respect to the hermitian form restricted to @xmath41 .
consider the unique connection @xmath66 on @xmath41 compatible with the hermitian metric in the fiber and with the complex structure on the base of the bundle .
this ( nonflat ) connection coincides with the connection defined as a composition of the restriction of @xmath61 to the subbundle @xmath41 composed with the projection @xmath64 : @xmath67 ( see , for example , @xcite , page 73 ) . the _ second fundamental form _ @xmath68 defined as @xmath69 is a differential form of type @xmath70 with values in the bundle of complex - linear maps from @xmath41 to @xmath71 , hence @xmath68 can be written as a matrix - valued differential form of type @xmath70 ( see , for example , @xcite , page 78 ) . in the literature
, @xmath68 is also known as the _ kodaira - spencer map_. note that we work with the pullbacks of the vector bundles @xmath62 , @xmath41 and @xmath71 to the moduli spaces @xmath12 or @xmath72 of abelian ( correspondingly quadratic ) differentials with respect to the natural projections @xmath25 ( correspondingly @xmath73 ) .
we recall that there is a canonical identification between the tangent bundle of the moduli space of riemann surfaces , which can be naturally described as the bundle @xmath74 of equivalence classes of beltrami differentials , and its cotangent bundle , which is naturally identified to the bundle @xmath72 of holomorphic quadratic differentials .
this identification follows from the existence of a canonical pairing between the bundle @xmath72 of quadratic differentials and the bundle of beltrami differentials given by integration . for any quadratic differential @xmath75 and any beltrami differential @xmath76 on a riemann surface @xmath15
, the pairing is given by the formula : @xmath77 in fact , quadratic differentials are tensors of type @xmath78 while beltrami differentials are tensors of type @xmath79 , hence the product of a beltrami and a quadratic differential is a tensor of type @xmath80 which can be integrated .
beltrami differentials corresponding to trivial deformations of the complex structure are exactly those which are orthogonal to all quadratic differentials @xcite , hence the pairing between beltrami and quadratic differentials induces a non - degenerate pairing between the tangent bundle to the moduli space @xmath13 of riemann surfaces and the bundle of holomorphic quadratic differentials .
the bundle of quadratic differentials is thus identified to the cotangent bundle of the moduli space of riemann surfaces .
there exists a natural map @xmath81 defined as follows , @xmath82 \in { { \mathcal b}}_g \ , , \quad \text { for all } \ , q\in { { \mathcal q}}_g\,,\ ] ] which yields a canonical identification between the bundles of quadratic and beltrami differentials , that is , between the cotangent and the tangent bundles to the moduli space @xmath13 of riemann surfaces . taking into account this canonical identification , the differential form @xmath68 with values in the bundle of complex - linear maps from @xmath41 to @xmath71 defines a vector bundle map @xmath83 over the moduli space of quadratic differentials . in other terms , for any @xmath84 by evaluating the form @xmath68 at the tangent vector @xmath85 under the identification between the tangent bundle and the bundle of quadratic differentials , we get a complex - linear map @xmath86 for any abelian differential @xmath87 , let @xmath88 be the complex - linear map corresponding to the quadratic differential @xmath89 .
the curvature tensor of the metric connections of the holomorphic hermitian bundles @xmath41 , @xmath41 are differential forms @xmath90 , @xmath91 of type @xmath80 with values in the bundle of complex - linear endomorphisms of @xmath41 , @xmath71 respectively , hence they can be written , with respect to pseudo - unitary frames , as skew - hermitian matrices of differentials forms of type @xmath80 on the moduli space @xmath92 ( see @xcite , page 73 ) .
let @xmath93 be a pseudo - unitary frame with respect to the pseudo - hermitian intersection form , that is , a frame which verifies the pseudo - orthonormality conditions @xmath94 let us consider the connection matrix @xmath95 associated to the gauss - manin connection @xmath96 . by the identities @xmath97 it follows that @xmath4 has a block structure @xmath98 with @xmath99 blocks @xmath100 , @xmath101 , @xmath102 and @xmath103 that verify the relations @xmath104 observe that for any unitary frame @xmath105 , there is an associated pseudo - unitary frame @xmath106 , with respect to the intersection form , defined as follows , @xmath107 and , with respect to the above pseudo - unitary frame , the blocks @xmath100 , @xmath101 are equal to the connection matrices @xmath108 , @xmath109 of the connections @xmath66 and @xmath110 , respectively , that is , @xmath111 and @xmath112 , and @xmath113 is the matrix of the second fundamental form ( see @xcite , page 76 ) .
let us now consider the curvatures @xmath114 , @xmath90 and @xmath91 of the connections @xmath61 , @xmath66 and @xmath110 on the vector bundles @xmath62 , @xmath41 and @xmath71 respectively .
it follows from the above relations that @xmath115 by cartan s structure equation ( see for instance @xcite , page 75 ) we have the identity @xmath116 , hence we conclude that @xmath117 it follows that ( compare with @xcite , page 78 ) @xmath118 note that @xmath114 is the curvature of the gauss manin connection , which is flat .
so @xmath114 is null , and the curvature @xmath90 can be written as @xmath119 similarly to the case of the second fundamental form , we pull back the bundle @xmath41 to a point @xmath120 of @xmath72 and take the value of the curvature form on the tangent vectors @xmath121 , where the tangent vector @xmath122 under the identification discussed above between the tangent bundle of the moduli space and the bundle of quadratic differentials .
thus , instead of a differential form of type @xmath80 with values in the bundle of complex endomorphisms of the bundle @xmath41 , we get a section of that bundle over @xmath72 , that is , we get a complex endomorphism @xmath123 of the space @xmath32 for any pair @xmath84 . for any abelian differential @xmath87 , let @xmath124 denote the complex endomorphism corresponding to the quadratic differential @xmath89 . for any riemann surface @xmath15 and any orthonormal basis @xmath125 of the space @xmath32 , the system @xmath126 is a pseudo - orthonormal basis of the space @xmath33 .
let @xmath127 be the matrix of the complex endomorphism @xmath128 with respect to the basis @xmath129 and @xmath102 be the matrix of the second fundamental form operator @xmath130 with respect to the bases @xmath129 and @xmath131 .
formula can be written in matrix form as follows : @xmath132 it is immediate from the above formulas that the matrix @xmath127 of the curvature of the hodge bundle is a negative - semidefinite hermitian matrix .
a formula for the second fundamental form @xmath68 was implicitly computed in @xcite , 2 .
we state such a formula below .
we remark that all formulas in @xcite are written with different notational conventions , which we now explain for the convenience of the reader .
any abelian differential @xmath16 on a riemann surface @xmath15 induces an isomorphism between the space @xmath32 of all abelian differentials on @xmath15 , endowed with the hodge norm , and the subspace of all square integrable meromorphic functions ( with respect to the area form of the abelian differential @xmath16 on @xmath15 ) . in @xcite , @xcite variational formulas are written in the language of meromorphic functions . in this paper
we will adopt the language of abelian differentials .
let @xmath133 be a pair ( riemann surface @xmath15 , holomorphic @xmath30-form @xmath16 on @xmath15 ) .
following @xcite , 2 , for any @xmath134 we define : @xmath135 the complex bilinear form @xmath136 depends continuously , actually ( real ) analytically , on the abelian differential @xmath137 .
[ lemma : secfundform ] for any @xmath87 , the second fundamental form @xmath130 can be written in terms of the complex bilinear form @xmath136 , namely @xmath138 in particular , for any orthonormal basis @xmath139 of holomorphic forms in @xmath32 and any @xmath140 , @xmath141 the argument is a simplified version of the proof of lemma 2.1 in @xcite rewritten in the language of holomorphic differentials . as everywhere in this paper we use the same notation for holomorphic ( antiholomorphic ) forms and their cohomology classes , while for other closed 1-forms we use square brackets to denote the cohomology class .
let @xmath142 denote a teichmller deformation , that is , a trajectory of the teichmller flow .
let @xmath143 be any given holomorphic differential on the riemann surface @xmath144 .
there exists @xmath145 such that there is a natural identification @xmath146 by parallel transport for all @xmath147 , so that locally constant sections are parallel for the gauss - manin connection .
let @xmath148 be a smooth one - parameter family of closed @xmath30-forms such that @xmath149 and @xmath150 is holomorphic on @xmath151 for all @xmath147 , that is , @xmath148 is a smooth local section of the bundle @xmath41 .
let @xmath152 denote the natural projection ( see formula ) .
by definition , @xmath153\right)\,.\ ] ] thus , for any @xmath30-form @xmath154 , in order to compute the pseudo - hermitian intersection @xmath155 , it is sufficient to compute the derivative @xmath156 _ up to exact @xmath30-forms and up to @xmath30-forms of type @xmath70_. in fact , it follows from formula that the cohomology class @xmath157 \in h^1(s , \mathbb c)$ ] is holomophic .
hence , by the definition of the pseudo - hermitian intersection form , it is orthogonal to @xmath158 for any holomorphic form @xmath159 , @xmath160\,,\ , \bar\beta\right)=0\,,\ ] ] which implies @xmath161\,,\,\bar\beta\right ) = \frac{i}{2 } \int_s \frac { d\alpha}{dt } ( 0 ) \wedge \beta\,.\ ] ] it is immediate from the definition of the teichmller deformation that @xmath162 by writing @xmath163 and differentiating , taking into account , we derive the following formula : @xmath164 by formula , the @xmath30-form @xmath165 is of type @xmath70 , hence ( taking into account that @xmath159 is holomorphic ) we have @xmath166 the first formula in the statement follows from formulas and . finally , let @xmath167 be any orthonormal basis ( in the sense that @xmath168 ) .
the system @xmath169 is pseudo - orthonormal ( in the sense that @xmath170 ) , hence @xmath171 thus the second formula in the statement is proved .
we warn the reader that in general ( unless @xmath172 ) @xmath173 in fact , the @xmath30-form @xmath174 is closed by definition , while the @xmath30-form @xmath175 is in general not closed . in order to compute @xmath174 directly ,
it is necessary to consider the appropriate projection of the @xmath30-form @xmath175 onto the subspace of closed @xmath30-forms .
we carry out such a direct calculation below , following lemma 2.1 in @xcite . we stress that this calculation , although not needed for first variation formulas , is important for the correct derivation of second variation formulas along the teichmller flow .
let @xmath176 and @xmath177 denote respectively the type @xmath70 and the type @xmath178 exterior differentials on the riemann surface @xmath15 , defined as the projections of the ( total ) exterior differential @xmath179 on the subspaces of @xmath30-forms of type @xmath70 and @xmath178 respectively . by definition , for all @xmath180 ,
the @xmath30-form @xmath181 is of type @xmath70 , the @xmath30-form @xmath182 is of type @xmath178 and the following formula holds @xmath183 note that the @xmath30-form @xmath175 is @xmath177-closed ( but not @xmath179-closed , unless @xmath184 ) , hence its @xmath177-cohomology class has a unique anti - holomorphic representative @xmath185 . in other words ,
there exist a unique anti - holomorphic form @xmath185 and a complex - valued function @xmath180 ( unique up to additive constants ) such that @xmath186 ( the linear operator @xmath187 is equivalent to the restriction to the subspace of meromorphic functions of the orthogonal projection from the space of square - integrable functions on @xmath15 onto the subspace of anti - meromorphic functions , which appears in the formulas of @xcite and @xcite ) .
since @xmath188 and @xmath189 are closed forms and any closed form of type @xmath70 is holomorphic , by formulas and it follows that @xmath190 hence , by formulas and , we conclude that the second fundamental form has the following expression : @xmath191 in conclusion , the form @xmath192 is equal to @xmath193 only up to a @xmath194-exact correction term . if such a correction term were identically zero , the theory of the kontsevich
zorich cocycle would be much simpler ! the second fundamental form of the hodge bundle
is related to the derivative of the period matrix along the teichmller flow .
we recall the definition of the period matrix .
let @xmath195 be a canonical basis of the first homology group @xmath196 of a riemann surface @xmath15 and let @xmath197 be the dual basis of holomorphic @xmath30-forms , that is , the unique basis with the property that @xmath198 the period matrix @xmath199 is the @xmath99 complex symmetric matrix with positive - definite imaginary part defined as follows : @xmath200 let @xmath201 denote the lie derivative along the teichmller flow on the space of abelian differentials .
the following formula holds : @xmath202 by rauch s formula ( see for instance @xcite , prop .
a.3 ) , for any beltrami differential @xmath76 on @xmath15 , we have @xmath203 by definition , at any holomorphic quadratic differential @xmath204 on @xmath15 the teichmller flow is tangent to the equivalence class of beltrami differentials represented by the beltrami differential @xmath205 the statement then follows immediately from rauch s formula .
for any @xmath206 ( that is , for any abelian differential @xmath137 of unit total area , see formula ) , the second fundamental form @xmath136 satisfies a uniform upper bound and a spectral gap bound , proved below .
[ lemma : apriori_b_bounds ] for any abelian differential @xmath206 on a riemann surface @xmath15 , the following uniform bound holds : @xmath207 and the maximum is achieved at @xmath208 , in fact we have @xmath209 let @xmath210 be hodge orthogonal complement of the complex line @xmath211 .
the following spectral gap bound holds : @xmath212 by the cauchy - schwarz inequality in the space @xmath213 , it follows that , for all @xmath143 , @xmath214 , @xmath215 the uniform upper bound in formula is therefore proved .
the bound is achieved at @xmath216 since it is immediate by the definition that @xmath217 in fact , @xmath218 is by definition equal to the area of surface @xmath15 with respect to the flat metric associated to the abelian differential @xmath219 , which , by definition of @xmath36 , is normalized ( equal to @xmath30 ) .
the spectral gap bound in formula is proved as follows . by a fundamental property of the cauchy - schwarz inequality , _ equality _ holds in formula if and only if there exists a constant @xmath220 such that @xmath221 the functions @xmath222 and @xmath223 are meromorphic on the riemann surface @xmath15 , hence @xmath224 is anti - meromorphic . since the only meromorphic functions which are also anti - meromorphic are the constant functions , if follows that equality holds in the cauchy - schwarz inequality if and only if @xmath143 and @xmath159 belong to the complex line @xmath225 . thus , if @xmath226 , the cauchy - schwarz inequality is strict and the spectral gap bound stated above on the second fundamental form @xmath136 is proved . the curvature form of the hodge bundle appears in the second variation formulas for the hodge norm computed in @xcite , 2 - 5 , which we will recall in the next section . for consistence with the notations of that paper
, we adopt below a sign convention for the curvature matrix which is opposite to that of formula ( [ eq : curvatureformula ] ) . for any abelian differential @xmath87 ,
let @xmath227 be the hermitian curvature form defined as follows : for all @xmath228 , @xmath229 it follows from lemma [ lemma : apriori_b_bounds ] that , for any abelian differential @xmath230 on a riemann surface @xmath15 , the second fundamental form operator ( the kodaira - spencer map ) @xmath231 is a contraction with respect to the hodge norm and , as a consequence , the hermitian positive semi - definite curvature form @xmath227 is uniformly bounded .
in fact , the following result holds : [ lemma : apriori_curv_bounds ] for any abelian differential @xmath206 on a riemann surface @xmath15 , the following bounds hold : @xmath232 and the maximum is achieved at @xmath208 , in fact we have @xmath233 the following spectral gap result holds : @xmath234 by lemma [ lemma : secfundform ] , it follows that , for all @xmath235 , we have @xmath236 and by the definition of the curvature form @xmath237 , we also have @xmath238 the upper bounds in formulas and therefore follow from the corresponding results for the form @xmath136 established in lemma [ lemma : apriori_b_bounds ] .
the identities follow from formula in lemma [ lemma : apriori_b_bounds ] , which states that @xmath239 .
in fact , by lemma [ lemma : secfundform ] we have @xmath240 finally , by the definition of the curvature form it follows that @xmath241 the argument is complete . for any abelian differential @xmath87 on a riemann surface @xmath15 the matrix @xmath242 of the hermitian curvature form @xmath227 with respect to any hodge - orthonormal basis @xmath243
can be written as follows .
let @xmath103 be the matrix of the bilinear form @xmath136 on @xmath32 with respect to the basis @xmath129 , that is : @xmath244 by formula and lemma [ lemma : secfundform ] , the matrix @xmath242 of the hermitian curvature form @xmath227 of the vector bundle @xmath41 over @xmath245 in the orthonormal basis @xmath129 can be written as follows : @xmath246 ( the above formula is the corrected version of the formula @xmath247 which appears as formula @xmath248 in @xcite and as formula @xmath249 in @xcite .
this mistake there is of no consequence ) . in particular , since the form @xmath136 is symmetric , the forms @xmath227 and @xmath136 have the same rank and their eigenvalues are related .
let @xmath250 and @xmath251 denote the set of eigenvalues of the forms @xmath227 and @xmath136 respectively .
the following identity holds : @xmath252 for every abelian differential @xmath253 , the eigenvalues of the positive - semidefinite form @xmath227 on @xmath32 will be denoted as follows : @xmath254 we remark that the top eigenvalue @xmath255 is equal to @xmath30 and the second eigenvalue @xmath256 for any abelian differentials @xmath206 as a consequence of lemma [ lemma : apriori_curv_bounds ] , in particular all of the above eigenvalues give well - defined , continuous , non - negative , bounded functions on the moduli space @xmath36 of all ( normalized ) abelian differentials . by the hodge representation theorem for riemann surfaces ,
the forms @xmath227 and @xmath136 induce _ complex - valued _ bilinear forms @xmath257 and
@xmath258 on the real cohomology @xmath21 : for all @xmath42 , @xmath43 , @xmath259 the forms @xmath257 and @xmath258 on @xmath21 have the same rank , which is equal to twice the common rank of the forms @xmath227 and @xmath136 on @xmath32 .
the bilinear form @xmath257 induces a real - valued , positive semi - definite quadratic form , while the quadratic form induced by the bilinear form @xmath258 is complex - valued .
we recall below some basic variational formulas from @xcite , 2 , 3 and 5 , reformulated in geometric terms .
such formulas generalize the fundamental kontsevich formula for the sum of all non - negative lyapunov exponents of the hodge bundle @xcite .
the second fundamental form of the hodge bundle measures the first variation of the hodge norm along a parallel ( locally constant ) section for the gauss - manin connection .
in fact , the formula given below ( implicit in the computation of @xcite , 2 ) holds .
let @xmath133 be a pair ( riemann surface @xmath15 , holomorphic @xmath30-form @xmath16 on @xmath15 ) .
for any cohomology class @xmath46 , let @xmath260 be the unique holomorphic @xmath30-form such that @xmath37 is the cohomology class of the closed @xmath30-form @xmath261 in the de rham cohomology @xmath262 .
we remark that for any given @xmath46 the holomorphic @xmath30-form @xmath260 only depends on the riemann surface @xmath15 . however , the riemann surface @xmath15 underlying a given holomorphic @xmath30-form @xmath16 is unique and it will be convenient to write below the harmonic representative as a function of the holomorphic @xmath30-form @xmath16 on @xmath15 .
[ lemma : varfor1 ] the lie derivative @xmath263 of the hodge inner product @xmath264 of parallel ( locally constant ) sections @xmath265 in the direction of the teichmller flow can be written as follows : @xmath266 the argument is just a rephrasing of the proof of lemma 2.1 of @xcite in the language of differential geometry .
let us recall that by definition , for any pair @xmath267 , the hodge inner product is defined as @xmath268 since the gauss - manin connection is compatible with the hermitian intersection form , we can compute @xmath269 and , since @xmath270 , we also have @xmath271 any cohomology class @xmath46 can be interpreted as a parallel ( constant ) local section of the bundle @xmath20 . since by definition of the differential @xmath272 we have
@xmath273/2 $ ] and since the gauss - manin connection is real ( on real tangent vectors of the moduli space it commutes with the complex conjugation ) , it follows that @xmath274 from formulas and we can derive the identities @xmath275 in conclusion , from formula , by the above identities and by the definition ( [ eq : sec : fund : form ] ) of the second fundamental form , it follows that @xmath276 the stated first variation formula is therefore proved .
the fundamental variational formula , computed in @xcite , lemma 2.1 , for the lie derivative of the hodge norm of a parallel ( locally constant ) section @xmath46 in the direction of the teichmller flow can now be derived from lemma [ lemma : varfor1 ] and lemma [ lemma : secfundform ] .
[ lemma : varfor2 ] the following variational formula holds : @xmath277 lemma [ lemma : secfundform ] can also be derived from the variational formulas of lemma [ lemma : varfor1 ] ( proved above ) and lemma [ lemma : varfor2 ] ( proved as part of lemma 2.1 in @xcite ) .
in fact , by comparison of the variational formulas of lemmas [ lemma : varfor1 ] and [ lemma : varfor2 ] , for any cohomology classes @xmath278 , @xmath279 which implies the main identity of lemma [ lemma : secfundform ] since the operator @xmath130 is complex linear , the intersection form is hermitian by definition and the form @xmath136 is complex bilinear .
the variational formula of lemma [ lemma : varfor2 ] implies a uniform bound and a spectral gap result on the exponential growth of hodge norms based on the uniform bound and on the spectral gap estimate of lemma [ lemma : apriori_b_bounds ] ( see @xcite , lemma 2.1 and corollary 2.2 ) .
let @xmath280 be the function defined as follows : for all @xmath206 , @xmath281 by definition @xmath282 is a continuous function on the moduli space of normalized ( unit area ) abelian differentials and by lemma [ lemma : apriori_b_bounds ] it is everywhere strictly less than @xmath30 , hence it achieves its maximum ( strictly less than @xmath30 ) on every compact subset .
it is proved in @xcite that @xmath282 has supremum equal to @xmath30 on every connected component of every stratum of the moduli space . for any abelian differential @xmath137 on a riemann surface @xmath15 , let @xmath283 denote the hodge norm of a real cohomology class @xmath46 , that is , the hodge norm of the holomorphic @xmath30-form @xmath284 .
[ cor : universal_1 ] the lie derivative of the hodge norm along the teichmller flow admits the following bounds : for any abelian differential @xmath219 on a riemann surface @xmath15 and for any cohomology class @xmath46 , @xmath285 for any cohomology class @xmath286 , [ { \operatorname{im}}(\omega)]\rangle^\perp$ ] , @xmath287 by lemma [ lemma : varfor2 ] , for any abelian differential @xmath137 on a riemann surface @xmath15 and for any cohomology class @xmath288 we have @xmath289 hence the statement follows from lemma [ lemma : apriori_b_bounds ] .
in fact , for any cohomology class @xmath46 , the abelian differential @xmath260 belongs to the hodge orthogonal complemement @xmath290 if and only if @xmath37 belongs to the hodge orthogonal complemement @xmath291 , [ { \operatorname{im}}(\omega)]\rangle^\perp \subset h^1(s,{{\mathbb{r}}})$ ] .
the above universal bound and spectral gap estimate immediately extends to all exterior powers of the hodge bundle . for every abelian differential @xmath87 on a riemann surface @xmath15 and for every @xmath292 , the hodge norm @xmath293 on @xmath294 induces a natural norm ( also called the hodge norm ) @xmath295 on polyvectors @xmath296 . [ cor : universal_2 ] the lie derivative of the hodge norm along the teichmller flow admits the following bounds : for any abelian differential @xmath219 on a riemann surface @xmath15 , for any @xmath297 and for any polyvector @xmath298 such that the span @xmath299 is an isotropic subspace , we have : @xmath300 for any @xmath301 the following stronger bound holds : @xmath302 the @xmath0-orbit of any holomorphic abelian differential @xmath303 is isomorphic to the unit tangent bundle of a hyperbolic surface ( generically a copy of the poincar disk ) .
thus the left quotient @xmath304 is a hyperbolic surface , called a _
teichmller disk_. there is a natural action of @xmath305 on the space @xmath12 by multiplication of abelian differentials by nonzero complex numbers .
the corresponding projectivization @xmath306 is foliated by teichmller disks endowed with the hyperbolic metric .
we remark that for consistency with a standard normalization for the teichmller geodesic flow adopted in the literature the hyperbolic metric is normalized to have curvature equal to @xmath307 .
we have the following basic variational formula for the _ leafwise _ hyperbolic laplacian @xmath308 of the hodge norm @xmath309 at a `` point '' @xmath16 of the projectivized moduli space @xmath310 ( see @xcite , lemmas 2.1 and 3.2 , @xcite , lemma 4.3 ) : [ lemma : varlapl ] the following variational formula for the hodge norm holds : @xmath311 [ rm : so : invariant ] in fact , given a cohomology class @xmath37 in @xmath21 , the hodge norm @xmath312 at a point @xmath133 of @xmath12 is completely determined by the complex structure of the underlying riemann surface @xmath15 .
thus , for any holomorphic form @xmath313 one has @xmath314 .
whenever , in addition @xmath315 with real @xmath5 , then @xmath316 and @xmath317 .
thus , all the quantities in the above formula are @xmath318-invariant , which makes it legitimate to consider them defined on a teichmller disk in the projectivized moduli space @xmath310
. it will often be convenient to pull back to @xmath12 the functions defined on the projectivized moduli space @xmath310 .
when operating with the leafwise hyperbolic laplacian @xmath308 , we will always tacitly verify the @xmath318-invariance of the functions involved .
the above variational formulas can be generalized to all the exterior powers of the kontsevich
zorich cocycle . for any @xmath319 ,
let us denote by @xmath320 the total space of the grassmannian bundle of isotropic @xmath321-dimensional subspaces of the hodge bundle . by definition ,
the fiber @xmath322 of the bundle @xmath320 at any abelian differential @xmath137 on a riemann surface @xmath15 is the grassmannian manifold of all @xmath321-dimensional isotropic subspaces of the symplectic vector space @xmath294 .
let @xmath323 denote the function on the grassmannian bundle @xmath320 defined as follows ( see @xcite , 5 ) .
let @xmath87 be an abelian differential on a riemann surface @xmath15 and let @xmath324 be any isotropic subspace of dimension @xmath319 .
let @xmath325 be any hodge - orthonormal basis and let @xmath326 be any hodge - orthonormal lagrangian completion .
let @xmath327 [ lm : fi : k ] the function @xmath328 depends only on @xmath87 and on the isotropic subspace @xmath329 and is independent of the choice of the orthonormal basis @xmath325 and of its hodge - orthonormal lagrangian completion @xmath330 .
it can be also expressed as @xmath331 ( for @xmath332 the second sum on the right hand side is defined to be null ) . for any normalized ( unit area ) abelian differential @xmath253 on a surface @xmath15 and for any @xmath321-dimensional isotropic subspace @xmath329 , the following bound holds : @xmath333 be an orthonormal basis of abelian differentials representing the orthonormal basis @xmath334 . by the definition of @xmath335 and @xmath336 we have that @xmath337 and @xmath338 . by formula
we also have the relation @xmath339 , which implies that @xmath340 by definition one has @xmath341 by formulas and and by taking into account that @xmath103 is symmetric , we transform formula as follows : @xmath342 formula is therefore established . note that formula does not depend on the choice of the system @xmath343 in the orthonormal lagrangian completion , while formula does not depend on the choice of the orthonormal basis @xmath344 .
hence , the equality between formulas and proves that neither formula depends on the choice of the orthonormal basis of @xmath345 and of its orthonormal lagrangian completion .
the bound in formula follows from formulas and , by taking into account that @xmath346 and @xmath347 , for all @xmath348 , according to the upper bounds in formulas and respectively . finally , by the spectral gap bound in lemma [ lemma : apriori_curv_bounds ] and by the consequent strict bound for the second curvature eigenvalue in formula , for @xmath349 the inequality in formula is strict .
the formulas below , computed in @xcite , lemma 5.2 and 5.2 , extend the formula in lemma [ lemma : varlapl ] to all exterior powers of the hodge bundle .
let @xmath325 be any hodge - orthonormal basis of an isotropic subspace @xmath329 on a riemann surface @xmath15 .
recall that the euclidean structure defined by the hodge scalar product on @xmath21 defines the natural norm @xmath350 of any polyvector which we also call the hodge norm . similarly to the case of the hodge norm , it is defined only by the complex structure of the underlying riemann surface .
it follows from the definition of @xmath351 that this function is @xmath318-invariant ( compare to remark [ rm : so : invariant ] ) .
thus , for any @xmath352 in @xmath12 the hodge norm @xmath353 defines a smooth function on the hyperbolic surface obtained as a left quotient @xmath354 of the orbit of @xmath355 .
[ lemma : var2 ] for all @xmath297 the following formula holds : @xmath356 see lemma 5.2 and 5.2 in @xcite .
recall that our hyperbolic laplacian @xmath308 is written in the hyperbolic metric of constant curvature @xmath307 .
any different choice of the constant negative curvature would change the rhs in the above formula by a constant factor .
for @xmath357 the formula of lemma [ lemma : var2 ] reduces to that of lemma [ lemma : varlapl ] .
[ rm : phi : g ] note that for @xmath332 the lagrangian subspace @xmath358 is not present in the right - hand side of definition of @xmath359 : @xmath360 thus , the function @xmath361 is the pull - back to the grassmannian bundle of lagrangian subspaces of a function on the moduli space @xmath12 .
moreover , by definition the above sum is the trace of the hermitian form @xmath227 , hence it is by definition the curvature of the hermitian bundle @xmath41 .
this fundamental fact discovered in @xcite is crucial for the validity of the kontsevich formula for the sum of exponents .
a version of this formula is stated in corollary [ cor : kzformula ] below .
in this section we derive formulas for the lyapunov exponents of the kontsevich zorich cocycle on the hodge bundle in terms of the second fundamental form and curvature of the hodge bundle .
let @xmath362 be a flow preserving a borel ergodic probability measure @xmath76 on a locally compact topological space @xmath363 .
let @xmath364 be a real or complex @xmath179-dimensional vector bundle .
in other words , the fiber @xmath365 of the vector bundle above any @xmath366 is a real or complex vector space isomorphic to @xmath367 or @xmath368 respectively .
a _ real or complex linear cocycle _ @xmath369 over the flow @xmath370 is a flow on the total space @xmath371 of the vector bundle such that , for all @xmath372 , the map @xmath373 is well - defined and linear over @xmath374 or @xmath375 respectively .
suppose that there exists a family of norms @xmath376 on the fibers of the vector bundle and , for all @xmath377 , let @xmath378 denote the operator norm of the linear map @xmath379 with respect to the norm @xmath380 on @xmath381 and @xmath382 on @xmath383 . under the condition of @xmath384-_integrability _ of the cocycle @xmath385 , that is , under the condition that @xmath386 the so - called _ oseledets theorem _ states that there exists a collection of real numbers @xmath387 , @xmath388 , such that , for @xmath76-almost every @xmath366 , one has a splitting @xmath389 with @xmath390 for every @xmath391 . moreover , the subspaces @xmath392 depend measurably on @xmath366 . in the literature ,
the numbers @xmath393 are called _ lyapunov exponents _ and the subspaces @xmath392 are called _
oseledets subspaces_. for the sake of convenience , one writes the list of lyapunov exponents as @xmath394 ( where @xmath179 is the dimension of the fibers of the vector bundle ) by _ repeating _ each exponent @xmath393 a number of times equal to the real or complex dimension of the corresponding oseledets space @xmath395 ( which by ergodicity is constant almost everywhere )
. we will loosely refer to both the list @xmath387 and @xmath394 as the _ lyapunov spectrum _ of the linear cocycle @xmath396 , although the second list also contains the information about the real or complex multiplicities of lyapunov exponents ( that is , about the real or complex dimensions of the oseledets subspaces ) .
the following general facts will be relevant in this paper : * the natural _ complexification _ of any real linear cocycle has the _ same lyapunov spectrum _ of the original real linear cocycle ; in particular , the complexified cocycle has complex multiplicities equal to the real multiplicities of the original real cocycle ; * the lyapunov spectrum of a real _ symplectic _ cocycle , that is , a real cocycle preserving a family of symplectic forms on the fibers @xmath397 , @xmath398 , of the vector bundle @xmath364 , has the form @xmath399 in other words , the lyapunov spectrum of a symplectic cocycle is symmetric with respect to the origin @xmath400 . for more details on oseledets theorem and the general theory of lyapunov exponents ,
see the books @xcite and @xcite ( and references therein ) .
coming back to the kontsevich - zorich cocycle , let us recall that , by corollary [ cor : universal_1 ] , for any ( teichmller ) flow - invariant borel probability ergodic measure @xmath76 on @xmath36 , the cocycle is @xmath384-integrable with respect to the hodge norm , hence it has well - defined lyapunov spectrum , with top exponent @xmath401 ; since the cocycle on @xmath19 preserves the symplectic intersection form on the ( @xmath60-dimensional ) fibers @xmath402 of @xmath19 , its lyapunov spectrum is symmetric .
thus the lyapunov spectrum of the kontsevich - zorich cocycle with respect to any ( teichmller ) flow - invariant borel probability ergodic measure @xmath76 on @xmath36 has the following form : @xmath403 in particular , the kontsevich zorich spectrum always has @xmath14 non - negative and @xmath14 non - positive exponents .
let @xmath404 be the kontsevich zorich spectrum of all _ distinct non - zero _ lyapunov exponents of the hodge bundle @xmath19 . applying the teichmller flow both in forward and backward directions we get the corresponding oseledets decomposition @xmath405 at @xmath76-almost every point @xmath133 of @xmath36 , where @xmath406 is omitted if the lyapunov spectrum of @xmath76 does not contain zero . by definition
all nonzero vectors of each subspace @xmath407 or @xmath406 share the same lyapunov exponent @xmath408 ( correspondingly @xmath409 ) which changes sign under the time reversing . [
rm : invariant : versus : sl : invariant ] by convention , when saying that a measure ( function , line subbundle , etc ) is _ `` invariant '' _ we mean that it is _ `` invariant with respect to the teichmller flow''_. if the corresponding object is _
`` invariant with respect to the @xmath0-action '' _ , we explicitly indicate that it is _
`` @xmath0-invariant''_. in particular , decomposition is defined by any probability measure invariant and ergodic with respect to the teichmller flow .
[ l.symp-orth ] every subspace @xmath410 of the oseledets direct sum decomposition except @xmath411 is isotropic .
any pair of subspaces @xmath410 , @xmath412 such that @xmath413 is symplectic orthogonal .
the restriction of the symplectic form to each subspace @xmath414 , where @xmath415 , and on @xmath411 is nondegenerate .
the absolute value of the symplectic product of any two cocycles @xmath416 in @xmath21 is uniformly bounded on any compact part @xmath417 of @xmath36 by the product of their hodge norms , @xmath418 by ergodicity of the flow , it returns infinitely often to the compact part @xmath417 . consider a pair of cocycles @xmath419 such that @xmath420 , @xmath421 and such that @xmath422 . by definition of @xmath423
, we have @xmath424 when @xmath425 the latter expression tends to zero when @xmath426 ; when @xmath427 the latter expression tends to zero when @xmath428 . in both cases
, we conclude that for a subsequence of positive or negative times @xmath429 ( chosen when the trajectory visits the compact set @xmath417 ) the symplectic product @xmath430 tends to zero .
since the symplectic product is preserved by the flow this implies that it is equal to zero , in particular @xmath431 .
thus , we have proved that every subspace @xmath410 except @xmath411 is isotropic , and that any pair of subspaces @xmath410 , @xmath412 such that @xmath413 is symplectic orthogonal .
hence , the cohomology space decomposes into a direct sum of pairwise symplectic - orthogonal subspaces @xmath411 , @xmath432 , etc , where we couple all pairs @xmath410 and @xmath433 . since the symplectic form is nondegenerate and the summands are symplectic - orthogonal , it is nondegenerate on any summand . for any @xmath297 such that @xmath434 , let @xmath435 be the index of the exponent @xmath436 in the ordering without multiplicities , @xmath437 .
let us define the @xmath321-th unstable oseledets subbundle as @xmath438 let @xmath439 be any @xmath440-dimensional flow - invariant measurable subbundle of the hodge bundle , almost everywhere defined with respect to a flow - invariant ergodic probability measure @xmath76 on the moduli space @xmath36 .
let us denote @xmath441 the lyapunov spectrum of the restriction of the kontsevich
zorich cocycle to the subbundle @xmath442 with respect to the teichmller geodesic flow and the invariant measure @xmath76 on @xmath36 .
let us also denote @xmath443 the lyapunov spectrum of all distinct lyapunov exponents and let @xmath444 be the corresponding oseledets decomposition .
it follows from oseledets theorem that the lyapunov exponents on @xmath445 form a subset of the lyapunov spectrum of the cocycle on the hodge bundle , that is , @xmath446 and the oseledets subspaces @xmath447 are the non - trivial intersections of oseledets spaces for the cocycle on the full hodge bundle @xmath19 , as in , and the subbundle @xmath448 , that is , @xmath449 where @xmath406 is omitted if the lyapunov spectrum of @xmath76 does not contain zero . for any @xmath319 denote by @xmath320 the total space of the grassmannian bundle of isotropic @xmath321-dimensional subspaces of the real hodge bundle @xmath19
let us denote by @xmath450 the space of all borel probability measures on @xmath320 which project onto any probability measure , _ absolutely continuous _ with respect to the flow - invariant ergodic probability measure @xmath76 on @xmath36 under the canonical projection .
the kontsevich
zorich cocycle on the hodge bundle @xmath19 preserves the symplectic form in the fibers .
hence , it defines a natural action on the grassmannian bundle @xmath320 . since the subbundle @xmath445 is flow - invariant , the measurable grassmannian subbundle @xmath451 is also flow - invariant .
let @xmath452 be the subset of those measures in @xmath450 that are invariant with respect to the kontsevich
zorich cocycle @xmath453 on @xmath320 .
note that all measures @xmath454 project onto the flow - invariant ergodic probability measure @xmath76 on @xmath36 under the canonical projection .
it follows that the set @xmath455 is a compact subset of the set of all borel probability measures on the locally compact space @xmath320 endowed with the weak - star topology .
consider a subbundle @xmath445 of the hodge bundle satisfying the properties stated at the beginning of section [ ssec : formulas ] .
ergodicity of the measure and invariance of the intersection form with respect to the flow implies that the restrictions of the symplectic form to @xmath76-almost any fiber of @xmath445 have the same rank .
we do not exclude the situation when the resulting form is degenerate .
denote by @xmath456 the rank of the restriction of the symplectic form to @xmath445 ; denote by @xmath457 the difference @xmath458 between the dimension of the fiber of @xmath445 and @xmath459 .
the restriction of the symplectic form to @xmath76-almost any fiber of @xmath445 is nondegenerate if and only if @xmath460 , otherwise the form is degenerate and @xmath461 . note also that @xmath457 is the maximal dimension of an isotropic subspace in the fiber of @xmath445 .
let @xmath451 denote the total space of the grassmannian bundle of all @xmath321-dimensional isotropic subspaces contained in the fibers of @xmath445 . by definition , for @xmath76-almost all @xmath219 , the fiber @xmath462 of the bundle
@xmath451 is equal to the space of all @xmath321-dimensional isotropic subspaces of @xmath463 .
clearly , for each individual fiber , one has @xmath464 , so the grassmanian bundle @xmath451 is a measurable subbundle of @xmath320 .
let @xmath465 be the subset of all borel probability measures on @xmath320 essentially supported on a subset of the measurable grassmannian bundle @xmath466 .
note that there is no flow - invariance assumption in the definition of the sets @xmath450 and @xmath467 .
in fact , we will prove existence of flow - invariant probability measures essentially supported on the grassmannian @xmath468 , which project to any given flow - invariant ergodic probability measure @xmath76 on @xmath36 , in lemma [ lemma : inv_meas_grass ] below .
we start with an elementary preparatory lemma .
[ lemma : bundle_meas ] let @xmath469 be any borel probability measure absolutely continuous with respect to @xmath76 on @xmath36 . let @xmath448 be a measurable subbundle defined @xmath469-almost everywhere . for any @xmath321 such that @xmath451 is non - empty ( i.e. , for any @xmath470 ) , there exist measures on @xmath451 which project onto @xmath469 on @xmath36 , under the canonical projection .
in particular , the space @xmath467 is also non - empty .
any borel measurable bundle can be trivialized on the complement of a subset of measure zero with respect to any given borel measure .
so , there exists a set @xmath471 of full @xmath469-measure such that the restriction @xmath472 is measurably isomorphic to the product bundle @xmath473 .
it follows that the restriction @xmath474 of the grassmannian bundle of isotropic subspaces is isomorphic to the product bundle @xmath475 . by definition , for any borel probability measure @xmath476 on @xmath477 , the product measure @xmath478 on the space @xmath475 induces ( by push - forward under the bundle isomorphism ) a borel probability measure @xmath479 on @xmath451 , which projects onto @xmath469 on @xmath36 under the canonical projection .
note that we do not assume that the measure @xmath480 constructed above is flow - invariant .
let @xmath481 be the subset of those measures in @xmath467 that are invariant with respect to the restriction of the kontsevich
zorich cocycle @xmath453 to @xmath445 .
we are going to show that the set @xmath482 is non - empty whenever the set @xmath467 is non - empty . for any @xmath483 ,
let @xmath484 be the set of all _ weak limits _ of the family of probability measures @xmath485 in the space of all borel probability measures on the locally compact space @xmath320 .
here by a `` weak limit '' we mean any limit in the weak - star topology along some diverging sequence of positive times @xmath486 .
of course , in general distinct sequences may lead to different weak limits .
[ lemma : inv_meas_grass ] for any @xmath487 , the set @xmath488 is a non - empty compact subset of the set @xmath482 of probability measures in @xmath467 which are invariant under the kontsevich
zorich cocycle .
the fiber of the grassmannian bundle @xmath320 is the space @xmath489 of all isotropic subspaces of dimension @xmath321 in a symplectic space of dimension @xmath60 , so that the fiber is a compact manifold . also , the full grassmannian bundle @xmath320 is a _ continuous _ bundle .
thus , the space of all borel measures on @xmath320 of finite total mass is a montel space ( in the sense that all closed bounded sets are compact ) with respect to the weak - star topology .
hence , for any diverging sequence of ( positive ) times @xmath490 we can extract from the sequence @xmath491 of measures in @xmath450 a converging subsequence .
the limit measure is a probability measure since , by the birkhoff ergodic theorem , the projection of the sequence @xmath491 under the canonical projection converges weakly to the flow - invariant ergodic probability measure @xmath76 on @xmath36 .
it follows that the subset @xmath488 of @xmath455 is nonempty . since
the set of all accumulation points of any given set in a topological space is closed and the set @xmath455 is compact , we get that @xmath488 is a nonempty compact subset of @xmath455 .
as the flow @xmath492 is continuous on @xmath320 , by the usual ( relative ) bogolyubov krylov argument ( see , e.g. , page 135 of the book @xcite ) , one has that any measure @xmath493 is @xmath494-invariant .
we reproduce the argument below for the convenience of the reader .
for any ( fixed ) @xmath495 and for all @xmath496 , we have @xmath497 it follows that ( for fixed @xmath495 ) the total mass @xmath498 of the signed measure @xmath499 converges to zero as @xmath500 .
in fact , @xmath501 let then @xmath491 be a sequence converging weakly to a measure @xmath502 .
since the map @xmath503 is continuous on @xmath320 , the sequence @xmath504 converges weakly to @xmath505 , hence , for all @xmath495 , we have @xmath506 we conclude that any measure @xmath493 is @xmath453-invariant as stated .
it remains only to show that @xmath507 , that is , any measure @xmath508 is essentially supported on @xmath451 , in the sense that @xmath509 .
let @xmath510 denote the projection of @xmath511 on @xmath512 . by definition
, @xmath510 is absolutely continuous with respect to @xmath76 : in particular , given @xmath513 , we can choose @xmath514 such that @xmath515 implies @xmath516 for all measurable @xmath517 . on the other hand , by luzin s theorem ( see , e.g. , page 2 of the book @xcite ) , given @xmath145
, we can fix @xmath518 a compact subset such that @xmath519 and @xmath520 is a _ continuous _ subbundle of the measurable bundle @xmath445 .
in particular , @xmath521 is a compact subset of @xmath320 .
let @xmath522 be any real - valued continuous function on @xmath320 such that @xmath523 and @xmath522 is identically equal to @xmath30 on @xmath521 .
since , by definition , @xmath480 means that @xmath479 is supported on @xmath451 and it projects to @xmath524 on @xmath36 , it follows from our choice of @xmath514 above that , for any @xmath496 , @xmath525 hence , for any weak limit @xmath493 , one has @xmath526 because this holds for every @xmath522 as above , we conclude that @xmath527 and hence @xmath509 , as claimed . for any measure @xmath528
we define below the average lyapunov exponent @xmath529 over the grassmannian bundle @xmath451 .
let us consider an isotropic subspace @xmath530 in the fiber of the hodge bundle over some point @xmath219 .
let @xmath531 and @xmath532 be a pair of bases in it .
let us consider the parallel transport of these vectors to a neighborhood of @xmath16 . clearly , at any point of the neighborhood of @xmath16 the polyvectors @xmath533 and @xmath534 remain proportional with the same constant coefficient .
hence , the hodge norms of these polyvectors are proportional with the same constant coefficient .
this implies , in particular , that the logarithmic derivative of the hodge norm of a polyvector along the teichmller geodesic flow , depends only on the isotropic subspace @xmath530 , @xmath535 thus , slightly abusing notations , in the following we shall sometimes denote the logarithmic derivative as above by @xmath536 the theorem establishes that , for @xmath76-almost every abelian differential @xmath219 on a riemann surface @xmath15 and for every polyvector @xmath537 in @xmath538 , the lyapunov exponent @xmath539 is well defined .
let us assume that the vectors @xmath540 form a basis of an isotropic subspace @xmath541 , and that the norm is the hodge norm . then , according to the above discussion about the logarithmic derivative of the hodge norm of an isotropic subspace , for @xmath76-almost all @xmath206 , the lyapunov exponent @xmath542 depends only on the isotropic subspace @xmath543 and can be written as follows : @xmath544 by corollary [ cor : universal_2 ] the function @xmath545 is bounded above by @xmath546 for any point @xmath547 of the grassmannian @xmath320 .
hence , for any @xmath548 and any @xmath496 , we get the following uniform estimate : @xmath549 by averaging the lyapunov exponent @xmath542 over @xmath451 with respect to measure @xmath550 , we define the _ average lyapunov exponent _ : @xmath551 note that the interchange of the limit with the integral and the change in the order of integration ( fubini theorem ) in formula are justified by the uniform upper bound established above in formula .
also , note that the definition of @xmath529 does nt assume the flow - invariance of @xmath552 .
suppose now that @xmath76 is an ergodic @xmath0-invariant probability measure on @xmath36 ( and not just flow - invariant as above ) .
let @xmath442 be any @xmath318-invariant measurable subbundle defined @xmath76-almost everywhere .
a measure @xmath528 will be called _
@xmath318-invariant _ if it is invariant under the natural lift of the action of the group @xmath318 to the grassmannian bundle @xmath451 .
the subset @xmath553 consisting of @xmath318-invariant probability measures is non - empty whenever @xmath467 is .
in fact , since by assumption the measure @xmath76 is @xmath0-invariant and the bundle @xmath445 is @xmath318-invariant , and since @xmath318 is an amenable ( compact ) group , the @xmath318-average of any probability measure in @xmath467 is a well - defined probability measure in @xmath554 . [
thm : partialsum ] let @xmath76 be any @xmath0-invariant borel probability ergodic measure on the moduli space @xmath36 of normalized abelian differentials .
let @xmath448 be any @xmath0-invariant measurable subbundle defined @xmath76-almost everywhere . for any @xmath318-invariant probability measure @xmath555 , the following formula holds : @xmath556 let @xmath557 denote the poincar disk and let @xmath558 denote the geodesic polar coordinates on the teichmller disk @xmath559 centered at an abelian differential @xmath206 on a riemann surface @xmath15 , defined as follows .
let @xmath560 and let @xmath561 for @xmath76-almost all abelian differential @xmath206 on a riemann surface @xmath15 and for any @xmath321-dimensional isotropic subspace @xmath562 , let @xmath563 be any hodge orthonormal basis at @xmath564 and let @xmath565 denote the hodge norm of the polyvector @xmath566 at the point @xmath567 of coordinates @xmath568 . from the variational formulas of lemma [ lemma : var2 ] for the hyperbolic laplacian of the hodge norm of a polyvector on a teichmller disk , by the green formula ( or , equivalently , by explicit integration of the poisson equation for the hyperbolic laplacian on the poincar disk @xmath557 )
, we derive the formula stated below ( see formula @xmath569 in @xcite ) .
let @xmath570 denote the disk of hyperbolic radius @xmath571 centered at the origin of the poincar disk , let @xmath572 denote its hyperbolic area and let @xmath573 denote the poincar area form .
let us also adopt the convention , introduced above , on the logarithmic derivative of the hodge norm of isotropic subspaces .
we have @xmath574 let us now integrate formula over the grassmannian @xmath451 with respect to the @xmath318-invariant probability measure @xmath575 .
note that by corollary [ cor : universal_2 ] all the integrands are uniformly bounded , hence it is possible to exchange the order of integrations . on the left - hand side ( lhs for short ) of formula , by the @xmath318-invariance of the measure @xmath479 on @xmath451 ,
we compute as follows : @xmath576 on the right - hand side ( rhs for short ) of formula we compute as follows .
let us recall that the poincar area form can be written as @xmath577 in geodesic polar coordinates @xmath578 and , as a consequence , the poincar area of the disk @xmath570 of geodesic radius @xmath571 is @xmath579 . by taking into account the uniform bound for @xmath580 , and by the above elementary formulas of hyperbolic geometry , and the @xmath318-invariance of the measure @xmath479 on @xmath451 , we proceed as follows : @xmath581 to sum up our computations so far , by integration of formula over the grassmannian @xmath451 with respect to the @xmath318-invariant probability measure @xmath582 we have :
@xmath583 let us then average the above formula over the interval @xmath584\subset { { \mathbb{r}}}$ ] and take the limit as @xmath585 .
the average of the lhs of formula converges , by the definition in formula , to the average lyapunov exponent of the @xmath318-invariant measure @xmath586 , that is , @xmath587 we claim that for any probability measure @xmath588 there exists a diverging sequence @xmath589 such that the average over @xmath590 $ ] of the rhs of formula converges to the integral @xmath591 so that , taking into account the limit in formula , the theorem follows from formula .
the above claim is proved as follows . for any continuous function @xmath522 with compact support on @xmath320 ,
the function @xmath592 converges to zero uniformly as @xmath593 .
in fact , the hyperbolic tangent converges to @xmath30 , the function @xmath522 is uniformly continuous , and for any @xmath145 , the mass assigned by the probability measure @xmath594 over @xmath595 $ ] to the interval @xmath596 $ ] converges to zero as @xmath597 .
it follows that the measure @xmath598 converges to zero weakly as @xmath599 .
thus for any @xmath588 there exists a diverging sequence @xmath589 such that the sequence of measures @xmath600 converges weakly to the measure @xmath601 on @xmath320 , essentially supported on @xmath468 . since the function @xmath323 is continuous and bounded on @xmath320 it follows that the average over @xmath590 $ ] of the rhs of formula converges to the integral in formula , as claimed , and the proof of the theorem is completed . for any @xmath602 such that @xmath603 , let @xmath604 be the index such that @xmath605 .
let us define @xmath606 in general , the subbundle @xmath607 does not need to be a bundle of isotropic subspaces , that is , a measurable section of the grassmannian @xmath451 .
however , note that if @xmath608 , then the bundle @xmath607 is a subbundle of the unstable oseledets bundle which is isotropic ( since the kontsevich
zorich cocycle is symplectic ) , hence it is itself isotropic .
[ cor : partialsum1 ] let @xmath76 be any @xmath0-invariant borel probability ergodic measure on the moduli space @xmath36 of normalized abelian differentials .
let @xmath448 be any @xmath0-invariant measurable subbundle defined @xmath76-almost everywhere .
assume that there exists @xmath609 such that @xmath603 and that the subbundle @xmath607 is a bundle of @xmath321-dimensional isotropic subspaces ( that is , it defines a measurable section of the grassmanian @xmath451 ) .
then the following formula holds : @xmath610 let @xmath611 be an @xmath318-invariant probability measure on @xmath451 such that all of its conditional measures on the fibers @xmath462 of the grassmannian bundle are equivalent to the lebesgue measure , for @xmath76-almost all @xmath206 .
theorem [ thm : partialsum ] in this special case implies formula .
in fact , by the assumption that @xmath603 the family of measures given in formula converges to the unique probability measure @xmath612 on @xmath451 given by the condition that for @xmath76-almost all @xmath206 the conditional measure @xmath613 is the dirac measure at the point @xmath614 ( in other terms , the measure @xmath615 is defined as the push - forward of the measure @xmath76 on @xmath36 under the section @xmath616 ) . in other words ,
the set @xmath617 of all weak limits of the family of measures given in formula is equal to @xmath618 . by the oseledets theorem , the average lyapunov exponent @xmath619 of the kontsevich
zorich cocycle with respect to the measure @xmath615 on the bundle @xmath451 is given by the formula @xmath620 thus formula is indeed a particular case of formula . in the particular case of the full hodge bundle we derive below a result first proved in @xcite , corollary 5.5 , for the canonical absolutely continuous invariant measures on connected components of strata of the moduli space .
[ cor : partialsum2 ] let @xmath76 be any @xmath0-invariant borel probability ergodic measure on the moduli space @xmath36 of normalized abelian differentials .
assume that there exists @xmath621 such that @xmath622 .
then the following formula holds : @xmath623 by remark [ rm : phi : g ] , corollary [ cor : partialsum2 ] for @xmath332 holds without any assumptions on the lyapunov exponents and provides a version of the kontsevich formula for their sum ( see @xcite and @xcite , corollary 5.3 ) : [ cor : kzformula ] let @xmath76 be any @xmath0-invariant borel probability ergodic measure on the moduli space @xmath36 of normalized abelian differentials .
the following formula holds : @xmath624 let @xmath625 be a subspace invariant under the hodge operator .
since for any nonzero @xmath626 one has @xmath627 , this implies that @xmath37 can not be symplectic - orthogonal to @xmath445 .
thus , invariance of @xmath445 under the hodge star - operator implies , in particular , that restriction of the symplectic form to @xmath445 is nondegenerate , in particular , @xmath445 is even - dimensional . for any hodge star - invariant subspace @xmath628 , let us define @xmath629 and @xmath630 to be the subspaces of cohomology classes of all holomorphic , respectively ,
anti - holomorphic forms @xmath16 such that @xmath631 \in v$ ] .
invariance of @xmath445 under the hodge operator implies that the sets @xmath632 and @xmath633 are indeed complex vector spaces , that @xmath634 and that @xmath635 . in particular , @xmath636
let us denote by @xmath637 and @xmath638 the restrictions of the forms @xmath227 and @xmath136 to @xmath639 and by @xmath640 and @xmath641 the restrictions of the forms @xmath257 and @xmath258 to @xmath642 respectively .
[ lm : equivalence : hodge : and : symp : orthogonal : under : star : invariance ] a subspace @xmath625 is invariant under the hodge star - operator if and only if the subspace @xmath643 , hodge - orthogonal to @xmath445 , coincides with the subspace @xmath644 , symplectic - orthogonal to @xmath445 . in that case
the subspace @xmath645 is hodge star - invariant and @xmath646 .
assume that a subspace @xmath625 is hodge star - invariant .
let @xmath644 be the subspace symplectic - orthogonal to @xmath445
. by for any @xmath647 and @xmath648 one has @xmath649 . since @xmath445 is hodge star - invariant , we have @xmath650 and hence the right expression is equal to zero .
it follows that @xmath651 .
the converse inclusion is proved similarly .
in fact , for any @xmath652 and @xmath648 one has @xmath653 .
again since @xmath445 is hodge star - invariant , we have @xmath654 , hence the right expression is equal to zero .
thus @xmath655 and equality holds .
conversely , assume that @xmath656 .
let @xmath657 and let @xmath658 . by , one has @xmath659 , hence @xmath660 . thus @xmath445 is hodge star - invariant .
let us show that if @xmath445 is hodge star - invariant , then @xmath643 is also hodge star - invariant .
let @xmath661 and take any @xmath662 .
by the same equation one has @xmath663 .
since @xmath42 and @xmath664 belong to hodge - orthogonal subspaces the right expression is equal to zero .
hence @xmath665 is symplectic - orthogonal to any @xmath666 in @xmath445 , which implies that @xmath667 .
finally , for any @xmath668 and any @xmath669 let @xmath670\in v$ ] and let @xmath671\in v^\perp$ ] . by formula @xmath672 which is equal to zero since @xmath445 and @xmath643 are both symplectic - orthogonal and hodge - orthogonal .
this implies that @xmath646 .
note that the above property is not related to either the @xmath0-action or the teichmller flow .
[ pr : b : reducible ] let @xmath673 be a hodge star - invariant subspace in the fiber of the hodge bundle over @xmath674 and let @xmath675 denote its symplectic orthogonal .
let @xmath676-{\varepsilon},{\varepsilon}\right[$ ] be any open interval along the trajectory of the teichmller flow passing through @xmath355 .
let us identify the fibers of the hodge bundle over @xmath677 by parallel transport with respect to the gauss manin connection .
the following properties are equivalent : \(i ) for any @xmath678 the subspace @xmath445 stays hodge star - invariant at @xmath679 .
\(ii ) for any @xmath678 the subspaces @xmath445 and @xmath644 are @xmath680-orthogonal .
an analogous equivalence holds when @xmath677 is replaced by a small open ball in @xmath0 containing the identity element , or by a small open neighborhood of the initial point @xmath352 in its teichmller disc .
suppose that property ( i ) is satisfied .
by lemma [ lm : equivalence : hodge : and : symp : orthogonal : under : star : invariance ] we have a direct sum decomposition @xmath681 where @xmath445 and @xmath644 are simultaneously symplectic - orthogonal and hodge - orthogonal with respect to the hodge - inner product @xmath682 on @xmath683 . since the symplectic structure is preserved by the gauss manin connection
, @xmath644 is constant over @xmath677 under our identification of the real cohomology spaces @xmath684 given by the connection .
hence for any pair @xmath685 , the hodge inner products @xmath686 for all @xmath678 , so that by lemma [ lemma : varfor2 ] @xmath687 it follows that @xmath445 and @xmath643 are @xmath680-orthogonal for all @xmath678 .
conversely , suppose that property ( ii ) is satisfied .
let @xmath644 be the symplectic orthogonal of @xmath445 .
since the gauss - manin connection preserves the symplectic structure , the space @xmath644 is constant over @xmath677 .
in addition , since @xmath445 is hodge star - invariant , the symplectic - orthogonal @xmath644 and the hodge - orthogonal @xmath643 coincide at @xmath688 .
it follows from formulas that since @xmath445 and @xmath644 are @xmath680-orthogonal for all @xmath678 and they are hodge - orthogonal for @xmath688 , then they are hodge - orthogonal for all @xmath678 .
thus the symplectic orthogonal and the hodge - orthogonal subspaces of @xmath445 coincide , hence by lemma [ lm : equivalence : hodge : and : symp : orthogonal : under : star : invariance ] the space @xmath445 is hodge star - invariant , for all @xmath678 . in the setting of proposition [ pr : b : reducible ] let @xmath689 be the eigenvalues of the positive - semidefinite hermitian form @xmath690 restricted to @xmath691 , where @xmath692 . in the setting of proposition [ pr : b : reducible ]
the following sets with multiplicities coincide : @xmath693 let @xmath139 be an orthonormal basis such that @xmath694 spans @xmath445 , @xmath695 spans @xmath696 and @xmath136 has block - diagonal matrix in the basis @xmath139 . by formula
the matrix @xmath697 is also block - diagonal in the basis @xmath698 .
hence , @xmath699 which is exactly the statement of the corollary .
suppose that at some point @xmath133 of the moduli space of normalized abelian differentials @xmath36 all the eigenvalues @xmath700 are distinct .
the corollary above implies that there is only a finite number of subspaces ( namely @xmath701 ) which might a priori serve as fibers of hodge star - invariant subbundles , namely , those spanned by @xmath702 for some subcollection @xmath703 of eigenvectors @xmath139 of @xmath227 . the condition that @xmath136 is block - diagonal in the corresponding basis in a small neighborhood @xmath677 of the initial point is a necessary and sufficient condition for extension of the corresponding subspace to a local hodge star - invariant subbundle over @xmath677 .
let @xmath448 be an @xmath0-invariant and hodge star - invariant subbundle of dimension @xmath704 over a full measure set for an @xmath0-invariant ergodic probability measure @xmath76 on the moduli space @xmath36 of normalized abelian differentials .
the hodge star - invariance is a very strong property of an @xmath0-invariant subbundle . in particular , the restriction of the kontsevich
zorich cocycle to @xmath445 mimics most of the properties of the cocycle on hodge bundle @xmath19 , where @xmath705 plays a role of a `` virtual genus '' .
let us give several illustrations of this general philosophy .
we have seen that the symplectic structure on @xmath19 restricts to a nondegenerate symplectic structure on @xmath445 , hence the kontsevich
zorich cocycle on @xmath445 is symplectic .
it follows that for any @xmath0-invariant borel probability ergodic measure @xmath76 on @xmath36 the kontsevich zorich spectrum on @xmath445 is symmetric : @xmath706 we get the following generalization of the kontsevich formula for the sum of positive lyapunov exponents ( compare analogous formulas in @xcite , @xcite and , for the specific case of teichmller curves , in @xcite ) .
[ cor : kzformgen ] let @xmath76 be any @xmath0-invariant ergodic borel probability measure on the moduli space @xmath36 .
the following formula holds for the kontsevich
zorich exponents of any subbundle @xmath707 ( of dimension @xmath704 ) , @xmath76-almost everywhere @xmath0-invariant and hodge star - invariant : @xmath708 let us consider a maximal isotropic subspace @xmath709 of @xmath463 , and some hodge - orthonormal basis @xmath710 of @xmath709 .
let @xmath694 be abelian differentials in @xmath32 such that @xmath711 $ ] for @xmath712 .
since @xmath445 is hodge star - invariant , we get @xmath713 for @xmath712 .
since @xmath710 are symplectic - orthogonal and hodge - orthonormal , the collection @xmath694 is orthonormal , see . complete the latter collection of abelian differentials to an orthonormal basis @xmath139 in @xmath32 . by construction @xmath695 is an orthonormal basis in @xmath714 .
finally , let @xmath711 $ ] for @xmath715 .
we have constructed a hodge - orthonormal basis @xmath716 of a lagrangian subspace in @xmath21 which completes the initial hodge - orthonormal basis in the isotropic subspace @xmath717 . by formula
we have @xmath718 by proposition [ pr : b : reducible ] the matrix @xmath136 is block - diagonal in the chosen basis , so we obtain the following relations : @xmath719 plugging the latter formula in definition we finally obtain the following expression for @xmath720 : @xmath721 since the function @xmath722 has no dependence on the maximal isotropic subspace @xmath723 and the subbundle @xmath445 is @xmath0-invariant , the statement follows from the variational formula given in lemma [ lemma : var2 ] and mimics the proof of the kontsevich formula ( see corollary [ cor : kzformula ] above and @xcite , corollary 5.3 ) . alternatively , the statement can be now immediately obtained from the more general formula on partial sums of exponents in corollary [ cor : partialsum2 ] above ( see @xcite , corollary for the proof ) .
in this section we collect several results which address the occurrence of zero kontsevich zorich exponents .
in particular , we prove that all the exponents are zero on a given @xmath0-invariant subbundle if and only if the cocycle is isometric and that happens whenever the second fundamental form vanishes on that subbundle . in
all known examples the vanishing of the second fundamental form can be derived from symmetries ( automorphisms ) of ( almost ) all abelian differentials in the support of an @xmath0-invariant measure .
we conclude with a partial converse which gives a lower bound on the number of strictly positive exponents in terms of the rank of the second fundamental form . by the variational formulas , whenever the second fundamental form vanishes identically on any flow - invariant subbundle then the kontsevich
zorich cocycle acts isometrically , hence all of its exponents are zero .
we prove below partial converse results in the special case of @xmath0-invariant , hodge star - invariant subbundles .
[ lemma : degeneratespectrum ] let @xmath724 be a flow - invariant subbundle over a full measure set for a flow - invariant ergodic borel probability measure @xmath76 on the moduli space @xmath36 of normalized abelian differentials . consider the following two properties : 1 .
the bilinear form @xmath641 vanishes for @xmath76-almost all @xmath219 ; 2 .
the restriction of the kontsevich
zorich cocycle to @xmath445 is isometric with respect to the hodge norm .
then , one has that @xmath725 implies @xmath726 . moreover , if one also assumes that @xmath445 is hodge star - invariant , then @xmath726 implies @xmath725 . by lemma [ lemma : varfor2 ] ,
the real part of the bilinear form @xmath258 gives the derivative of the hodge inner product under the action of the kontsevich
zorich cocycle at any @xmath206 .
whenever @xmath641 vanishes for @xmath76-almost all @xmath206 , by continuity it vanishes identically on the support of the measure , and , hence , it follows from the variational formula that the kontsevich
zorich cocycle acts isometrically on @xmath445 with respect to the hodge inner product .
let us show now that @xmath726 implies @xmath725 ( assuming also that @xmath445 is hodge star - invariant ) . by the variational formula the real part of the bilinear form @xmath103 vanishes on @xmath632 , i.e. @xmath727 for all @xmath728 .
note that , since @xmath136 is complex bilinear , @xmath729 since the hodge star - invariance of @xmath445 implies that @xmath632 is a complex space , we see that if @xmath730 for all @xmath728 , then @xmath731 . under either condition @xmath725 or condition @xmath726 of lemma [ lemma : degeneratespectrum ] above , all lyapunov exponents of the restriction of the kontsevich
zorich cocycle to @xmath445 are equal to zero .
in fact , any isometric cocycle has a lyapunov spectrum reduced to the single exponent zero . under the extra assumption that the invariant subbundle @xmath445 and the ergodic measure @xmath76 are invariant not only with respect to the teichmller flow , but with respect to the action of @xmath0
, one can prove a converse statement and prove that vanishing of all lyapunov exponents of an invariant subbundle implies vanishing of the second fundamental form on this subbundle , see theorem [ th : ann ] below . in the particular case
that @xmath445 is hodge star - invariant , the converse result becomes a straightforward corollary of the generalized kontsevich formula ( see formula in corollary [ cor : kzformgen ] ) .
[ cr : rank:0 ] let @xmath724 be an @xmath0-invariant , hodge star - invariant subbundle over a full measure set for an @xmath0-invariant ergodic borel probability measure @xmath76 on the moduli space @xmath36 of abelian differentials .
the following properties are equivalent : 1 .
the bilinear form @xmath641 vanishes for @xmath76-almost all @xmath219 ; 2 .
the restriction of the kontsevich
zorich cocycle to @xmath445 is isometric with respect to the hodge norm ; 3 .
the non - negative lyapunov spectrum of @xmath445 has the form @xmath732 the first two statements are equivalent by lemma [ lemma : degeneratespectrum ] .
the second statement implies the third by the definition of lyapunov exponents .
all the above statements hold for any flow - invariant subbundle .
if @xmath445 is @xmath0-invariant and hodge star - invariant , the third statement implies the first statement by the generalized kontsevich formula .
in fact , by that formula ( see corollary [ cor : kzformgen ] ) the vanishing of all lyapunov exponents of @xmath445 implies that @xmath733 it follows that @xmath637 vanishes on the support of the measure @xmath76 in @xmath36 . by formula
this implies that the bilinear forms @xmath638 and , thus , @xmath641 also vanish for @xmath76-almost all @xmath219 .
an important particular case of the above corollary [ cr : rank:0 ] is given below . in appendix
[ s : rk : b : cyclic : covers ] we shall see other examples .
the hodge bundle @xmath19 over @xmath36 splits into a direct sum of two subbundles .
the first one ( the _ tautological subbundle _ ) has dimension two ; its fiber are spanned by cohomology classes @xmath631 $ ] and @xmath734\in h^1(s , { { \mathbb{r}}})$ ] .
the second subbundle , @xmath735 , is the orthogonal complement to the first one with respect to the symplectic intersection form ( and with respect to the hodge inner product ) on the hodge bundle @xmath19 . clearly , both the tautological subbundle and its orthogonal complement @xmath735 are @xmath0-invariant and hodge star - invariant . in particular , for all @xmath219 the space @xmath736 is the orthogonal complement to @xmath16 in @xmath32 with respect to the hermitian form .
by corollary [ cr : rank:0 ] we have the following result ( see corollary 7.1 in @xcite ) : [ cr : rank:1 ] let @xmath76 be an @xmath0-invariant ergodic borel probability measure on the moduli space @xmath36 of normalized abelian differentials .
the second fundamental form @xmath136 has rank equal to @xmath30 for @xmath76-almost all @xmath219 if and only if all the non - trivial lyapunov exponents of the kontsevich zorich cocycle with respect to @xmath76 vanish , that is , if and only if @xmath737 it can be verified explicitly that the lyapunov spectrum of the tautological bundle is @xmath738 .
it follows that the non - negative lyapunov spectrum of its symplectic orthogonal complement @xmath739 is @xmath740 . since by definition @xmath239 , the rank of @xmath136 is @xmath30 if and only if the rank of @xmath741 is zero , for all @xmath206 .
the statement then follows from corollary [ cr : rank:0 ] for the @xmath0-invariant , hodge star - invariant bundle @xmath739 .
we recall below a simple symmetry criterion for the vanishing of the second fundamental form found in @xcite , 7 .
let @xmath206 be an abelian differential on a riemann surface @xmath15 .
suppose that @xmath15 has a holomorphic automorphism @xmath742 and that the holomorphic 1-form @xmath16 is an eigenvector of the induced action @xmath743 .
denote by @xmath744 the corresponding eigenvalue , @xmath745 .
note that the action @xmath743 preserves the restriction of the hermitian intersection form ( [ eq : intform ] ) which is positive - definite on @xmath32 , which implies that @xmath746 and @xmath747 is diagonalizable . consider a basis @xmath139 of eigenvectors of @xmath748 in @xmath32 and
denote the corresponding eigenvalues by @xmath749 .
the following statement is a simplified version of lemma 7.2 in @xcite .
[ thm : degenerateb ] let @xmath750 be an @xmath0-invariant suborbifold in some stratum of abelian differentials in genus @xmath14 .
let @xmath750 be endowed with an ergodic probability measure .
suppose that almost every flat surface @xmath133 in @xmath750 is endowed with a holomorphic automorphism @xmath751 , and that @xmath16 is an eigenvector of @xmath748 with an eigenvalue @xmath744 .
denote by @xmath752 all eigenvalues of @xmath753 . if for all but one couple of indices @xmath754 , where @xmath755 , one has @xmath756 , then the rank of the bilinear form @xmath136 on @xmath41 is equal to @xmath30 for all @xmath757 , and , hence , all the non - trivial lyapunov exponents of the hodge bundle with respect to the teichmller geodesic flow on @xmath750 vanish : @xmath758 consider a holomorphic automorphism @xmath751 .
for any two holomorphic differentials @xmath759 of our basis of eigenvectors of the linear map @xmath748 in @xmath32 , by definition of the form @xmath136 and by change of coordinates , we get @xmath760 hence , for every pair of indices @xmath761 such that @xmath756 , the element @xmath762 of the matrix of the form @xmath136 is equal to zero .
the result now follows from corollary [ cr : rank:1 ] .
in appendix [ s : rk : b : cyclic : covers ] ( more precisely , subsection [ ss : fmt ] below ) , we will give an application of this symmetry criterion ( for the vanishing of exponents ) provided by theorem [ thm : degenerateb ] in a particular interesting case ( of an arithmetic teichmller disk of a square - tiled cyclic cover in genus @xmath763 ) .
an important example of a hodge star - invariant subspace is given by the kernel of the second fundamental form .
in fact , the following elementary result holds . for any @xmath764 ,
let @xmath765 denote the kernel of the form @xmath336 on @xmath294 , that is : @xmath766 [ lemma : ann : hodgeinvariance ] for any @xmath767 the subspace @xmath765 in @xmath21 is hodge star - invariant .
moreover , @xmath768 .
let @xmath769 be any two cohomology classes , and let @xmath770 be holomorphic @xmath30-forms such that @xmath670 $ ] and @xmath671 $ ] .
by definition @xmath771 $ ] , so we have @xmath772 $ ] .
thus , by the definition of @xmath336 given in section [ ss : eval : sec : fund : form ] and by bilinearity of the form @xmath136 defined in one gets @xmath773 hence , if for some @xmath774 one has the identity @xmath775 for all @xmath43 , one also has @xmath776 for all @xmath43 .
the last statement is a direct corollary of the definition of @xmath336 .
another remarkable , however simple , property of the bundle @xmath765 is described below : [ lemma : invcompl ] for any flow - invariant ( resp . ,
@xmath0-invariant ) measurable subbundle @xmath777 the hodge orthogonal splitting @xmath778 is flow - invariant ( resp .
, @xmath0-invariant ) . by the variational formula ( [ eq : varfor2 ] ) of lemma [ lemma : varfor2 ] for the hodge innner product
, the condition @xmath777 implies that the hodge product @xmath779 is constant for any parallel ( locally constant ) sections @xmath780 and @xmath781 .
in particular , the equation @xmath782 is invariant under parallel transport .
we prove below our strongest result on the central oseledets bundle .
[ th : ann ] let @xmath76 be a flow - invariant ergodic probability measure on the moduli space @xmath36 of normalized abelian differentials .
if @xmath783 is a flow - invariant subbundle of the hodge bundle such that @xmath784 is @xmath76-almost everywhere flow - invariant , then the dimension @xmath785 is the same for @xmath76-almost all @xmath206 ( so @xmath784 defines a vector bundle over the support of @xmath76 ) and @xmath786 if @xmath76 is @xmath0-invariant and @xmath735 is any @xmath0-invariant measurable subbundle of the oseledets ( measurable ) bundle @xmath411 , then @xmath787 in particular , if @xmath76 is @xmath0-invariant and @xmath788 is @xmath76-almost everywhere @xmath0-invariant , then @xmath789 consequently , if @xmath76 is @xmath0-invariant and @xmath448 is a hodge star - invariant and @xmath0-invariant subbundle of the hodge bundle such that @xmath784 is @xmath76-almost everywhere flow - invariant and @xmath788 is @xmath76-almost everywhere @xmath0-invariant , then @xmath790 since @xmath791 is , by hypothesis , a flow - invariant integer - valued function on @xmath36 , the ergodicity of @xmath76 implies that it is constant @xmath76-almost everywhere .
( the dimension can jump and become larger , say , on suborbifolds of nontrivial codimension ) .
thus , @xmath784 defines a flow - invariant vector bundle over the support of @xmath76 and we can apply lemma [ lemma : degeneratespectrum ] to this bundle to prove the first statement . the last statement is a trivial combination of the first two , so the essential part of the theorem is the second statement , which is proved below . let @xmath792 be the space of all probability measures on the grassmannian bundle @xmath793 of @xmath321-dimensional isotropic subspaces projecting on @xmath36 to some @xmath510 absolutely continuous with respect to @xmath76 .
let us denote by @xmath456 the rank of the restriction of the symplectic form to the subbundle @xmath735 and by @xmath457 the difference @xmath794 .
by lemma [ lemma : bundle_meas ] , for any @xmath546 satisfying the relations @xmath795 , the space @xmath792 is non - empty .
let @xmath796 be the subset of all @xmath318-invariant measures . as we have already seen in section [ ssec : formulas ] , since @xmath318 is a compact amenable group , the set @xmath797 is non - empty whenever @xmath798 is non - empty : the @xmath318-average of any measure in @xmath799 is a measure in @xmath800 . since by assumption @xmath801 , all the lyapunov exponents of the restriction of kontsevich zorich cocycle to @xmath735 are zero , hence by the oseledets theorem the average lyapunov exponent @xmath529 of any probability measure @xmath802 is equal to zero . by theorem [ thm : partialsum ]
it follows that , for any measure @xmath803 and for any weak limit @xmath588 , @xmath804 since @xmath323 is by definition non - negative , the above formula implies that @xmath323 vanishes @xmath805-almost everywhere . hence , for @xmath805-almost all @xmath806 we get the following conclusion . by applying the identity in the middle of formula to any lagrangian hodge - orthonormal completion @xmath807 of a hodge - orthonormal basis @xmath531 of the @xmath321-dimensional isotropic subspace @xmath808 we get @xmath809 by the definition of the form @xmath810
, it follows that @xmath811 for all @xmath812 , or , equivalently , that @xmath813 .
thus we have proved that , for any borel probability measure @xmath814 and for any weak limit @xmath588 , @xmath815 for every borel probability measure @xmath814 , we then define , over the support of the measure @xmath76 on @xmath36 , a flow - invariant borel measurable subbundle @xmath816 as follows . for @xmath76-almost all abelian differentials @xmath219 we let the fiber @xmath817 be the linear span of all isotropic subspaces @xmath808 such that @xmath818 belongs to the essential support of at least one measure @xmath588 on the grassmannian @xmath819 . since @xmath488 is a compact set of probability measures , and , by rokhlin s disintegration theorem @xcite , for each @xmath820 , the conditional measures @xmath821 of @xmath822 on the fibers @xmath823 depend measurably on @xmath253 , one can check that @xmath824 depends measurably on @xmath253 . by construction , since all measures @xmath588 are flow - invariant , the family of subspaces @xmath824 is defined @xmath76-almost everywhere and flow - invariant . since the measure @xmath76 is ergodic , this implies that the dimension @xmath825 is @xmath76-almost everywhere constant , hence @xmath816 is a @xmath76-measurable flow - invariant subbundle . by construction ,
every such @xmath826 is a subbundle of @xmath827 , since at @xmath76-almost all @xmath206 the fiber @xmath817 is spanned by subspaces @xmath345 of @xmath828 .
we then define , for every @xmath829 one more flow - invariant measurable subbundle @xmath830 over the support of @xmath76 on @xmath36 . for @xmath76-almost any @xmath206 ,
the fiber @xmath831 is defined as a linear span of the family of vector spaces @xmath832 . finally , we let @xmath833 be the flow - invariant measurable subbundle , defined @xmath76-almost everywhere , such that the fiber @xmath834 over @xmath76-almost any @xmath206 is the linear span of vector subspaces @xmath835 . as above , since @xmath76 is ergodic , the bundles @xmath836 and @xmath837 are indeed measurable subbundles . by construction ,
the measurable bundle @xmath837 is a subbundle of @xmath838 , since its fiber at @xmath76-almost any abelian differential @xmath206 is spanned by subspaces of @xmath839 .
let us argue by contradiction .
let us assume that the subset @xmath840 of all @xmath219 such that @xmath841 has positive measure ( with respect to the @xmath0-invariant ergodic probability measure @xmath76 on @xmath36 ) .
since the subbundles @xmath842 and @xmath735 are @xmath318-invariant , it follows that the set @xmath843 is @xmath318-invariant . for @xmath76-almost
all @xmath206 , let @xmath844 denote the smallest @xmath318-invariant linear subspace which contains the subspace @xmath845 .
the vector space @xmath846 can be defined as the intersection of the ( non - empty ) family of all @xmath318-invariant subspaces of the vector space @xmath847 which contain @xmath834 or , equivalently , as the span of the union of the family @xmath848 . by construction @xmath849 , hence @xmath850 , which implies that @xmath851 , for all @xmath852 .
note that it is not restrictive to assume that the dimension of the vector space @xmath846 is constant for all @xmath853 .
in fact , the positive measure set @xmath843 has a finite partition @xmath854 hence at least one of these sets has positive measure .
we can therefore assume that the collection of vector spaces @xmath855 forms a proper measurable @xmath318-invariant subbundle @xmath856 of the restriction @xmath857 of the subbundle @xmath735 to @xmath843 . by the above construction
, it follows that the bundle @xmath858 is a non - trivial @xmath318-invariant subbundle of the restriction of the subbundle @xmath859 to @xmath843 .
we claim that there exists @xmath860 such that for all @xmath861 there exists an @xmath318-invariant probability measure @xmath862 essentially supported on the grassmannian @xmath863 .
in fact , let @xmath864 be the restriction of the measure @xmath76 on @xmath36 to the positive measure set @xmath843 , normalized to have unit total mass .
there exists @xmath865 such that the grassmannian of @xmath866-dimensional isotropic subspaces of the ( non - trivial ) subbundle @xmath867 is non - empty . by lemma [ lemma : bundle_meas ] , for any @xmath861 there exists a probability measure @xmath868 on @xmath869 essentially supported on @xmath870 , which projects onto the probability measure @xmath864 on @xmath36 under the canonical projection .
let @xmath871 be the @xmath318-average of the measure @xmath868
. by construction @xmath871 belongs to the set @xmath800 : in fact , it is @xmath318-invariant , it projects onto the probability measure @xmath872 , which is absolutely continuous with respect to @xmath76 on @xmath36 , and it is essentially supported on @xmath863 .
the above claim is therefore proved . since @xmath873 and by construction it is flow - invariant , by lemma [ lemma : invcompl ] there is a flow - invariant hodge - orthogonal splitting @xmath874 it follows that , on the one hand , @xmath875 by construction , since @xmath871 is essentially supported on @xmath863 and the bundle @xmath859 is flow invariant ; on the other hand , @xmath876 by the definition of the bundle @xmath837 given above .
of course , this is a contradiction since @xmath877 .
the argument is therefore complete .
concerning the invariance assumptions ( under teichmller flow and/or @xmath0 ) in the previous theorem , we would like to stress out that they are really necessary .
more precisely , while in appendix [ s : rk : b : cyclic : covers ] below we introduce a class of ergodic @xmath0-invariant probability measures supported on the @xmath0-orbit of _ square - tiled cyclic covers _ such that we can show @xmath878 is @xmath0-invariant ( see theorem [ thm : cyclic : covers ] below ) , we construct in appendix [ s : rk : b : z ] below an example of a closed @xmath0-invariant locus @xmath11 supporting an ergodic @xmath0-invariant probability measure such that @xmath879 is not @xmath0-invariant , @xmath880 is not teichmller invariant and hence @xmath881 _ even though _ they have the same dimension ! in other words , although @xmath880 does nt coincide with @xmath879 in the case of the locus @xmath11 , the number of zero exponents is still predicted by the corank of @xmath882 . partly motivated by the features of these examples , we pose below the problem to establish whether the corank of @xmath882 always give the correct number of central exponents ( see problem [ pb : rank ] below ) . note that invariance of a hodge star - invariant vector subbundle @xmath448 of the hodge bundle under @xmath0 does not imply invariance of the corresponding subbundle @xmath883 under @xmath0 or under the flow . for example , the tautological bundle , spanned by @xmath884 $ ] and @xmath885 $ ] is @xmath0-invariant , while the line bundle @xmath886 is not .
the lemma below shows that @xmath887 is special in this sense .
assume that the kernel @xmath887 of the bilinear form @xmath810 on @xmath19 is invariant over an open interval @xmath676-{\varepsilon},{\varepsilon}\right[$ ] along the trajectory of the teichmller flow passing through @xmath888 .
then the kernel @xmath889 of the form @xmath103 on @xmath41 is also invariant over @xmath677 .
an analogous statement holds when @xmath677 is replaced by a small open ball in @xmath0 containing the identity element , or by a small open neighborhood of the initial point @xmath352 in the teichmller disc of @xmath352 . since by lemma [ lemma : ann : hodgeinvariance ] the subspace @xmath890 is hodge star - invariant , the subspace @xmath891 is invariant over the interval @xmath677 and admits a ( a priori non - invariant ) decomposition @xmath892 , for all @xmath678 . by lemma [ lemma :
ann : hodgeinvariance ] @xmath893 , hence by the definition ( [ eq : sec : fund : form ] ) of the second fundamental form the projection of the covariant derivative @xmath894 onto @xmath895 vanishes for every @xmath896 and for all @xmath678 .
it follows that the subspace @xmath897 is constant over @xmath678 , as stated . in this subsection
we prove a general lower bound on the number of strictly positive lyapunov exponents of the kontsevich
zorich cocycle in terms of the rank of @xmath103 .
such a bound is considerably weaker than a sharp estimate ( roughly by a factor @xmath898 ) and no upper bound other than lemma [ lemma : degeneratespectrum ] is known .
[ thm : lowerbound ] let @xmath76 be any @xmath0-invariant ergodic probability measure on the moduli space of abelian differentials .
let @xmath448 be any hodge star - invariant , @xmath0-invariant subbundle ( of dimension @xmath899 ) defined @xmath76-almost everywhere .
if for some @xmath900 the bottom @xmath901 exponents of the restriction of the kontsevich
zorich cocycle to @xmath445 with respect to the measure @xmath76 vanish , that is , @xmath902 then the rank of the bilinear form @xmath903 satisfies the inequality : @xmath904 the proof follows closely the argument given in @xcite , corollary 5.4 , which was stated for the hodge bundle with respect to the canonical absolutely continuous invariant measures only . in that case , the bilinear form @xmath103 has maximal rank ( equal to @xmath14 ) on a subset of positive ( lebesgue ) measure on each connected component of every stratum , as proved in @xcite , 4 , which implies that @xmath905 } > 0\,.\ ] ] the generalized argument proceeds as follows . if all lyapunov exponents vanish , then the @xmath906 also vanishes by corollary [ cr : rank:0 ] , hence the statement holds in this case .
it follows that without loss of generality we may assume that @xmath907 : if this assumption only holds for a smaller value of @xmath908 , we would prove a statement even stronger than the claim .
let us assume then that @xmath909 .
we are in the setting of corollary [ cor : partialsum1 ] and we can apply formula . for @xmath76-almost all @xmath206 , let @xmath910 be a hodge orthonormal basis .
let @xmath911 be a completion of the system @xmath531 to a orthonormal basis of any lagrangian subspace of @xmath294 . by formula and by the definition of the function @xmath323 in formula we have @xmath912 since @xmath445 is hodge star - invariant , by the reducibility of the second fundamental form
( see proposition [ pr : b : reducible ] ) it follows that @xmath913 by the generalized kontsevich formula ( see corollary [ cor : kzformgen ] ) we get @xmath914 since @xmath915 the above two expressions coincide , which implies that @xmath916 this means that @xmath76-almost everywhere we have @xmath917 for any pair @xmath754 such that @xmath918 .
thus , for @xmath76-almost all @xmath206 there exists an orthonormal basis @xmath919 of the subspace @xmath920 of holomorphic differentials on the riemann surface @xmath15 such that @xmath921 hence , the matrix of the form @xmath903 with respect to the basis @xmath919 has a @xmath922 zero diagonal block .
it follows that @xmath903 has rank at most @xmath923 .
[ see @xcite , cor .
5.4 ] [ cor : lowerbound ] let @xmath76 be any @xmath0-invariant ergodic probability measure on the moduli space of abelian differentials . if for some @xmath924 the bottom @xmath925 exponents of the kontsevich
zorich cocycle with respect to the measure @xmath76 vanish , that is , @xmath926 then the rank of the bilinear form @xmath136 on @xmath41 satisfies the inequality : @xmath927 in @xcite the first author introduced analytic subvarieties @xmath928 of the moduli space @xmath36 , defined as follows : @xmath929 in terms of such subvarieties theorem [ thm : lowerbound ] can be formulated as follows : if the support of the @xmath0-invariant measure @xmath76 on @xmath36 is _ not _ contained in the subvariety @xmath930 , then formula does not hold , that is , @xmath931 a lower bound for the number of strictly positive exponents which holds in general for @xmath0-invariant probability measures supported on _ regular _ orbifolds is proved in @xcite .
such a bound can not be derived from the above corollary [ cor : lowerbound ] without assumptions on the minimal rank of the fundamental form @xmath103 on the moduli space @xmath36 . however , to the authors best knowledge , lower bounds on the rank of the second fundamental forms for @xmath932 are not available .
it could be conjectured that it grows linearly with respect to the genus of the surface .
the argument given in the proof of theorem [ thm : lowerbound ] is _ not _ sufficient to prove the non - vanishing of _ all _ the kontsevich zorich exponents .
in fact , any improvement on theorem [ thm : lowerbound ] based on the formulas for sums of lyapunov exponents , given in corollary [ cor : partialsum2 ] , seems to require some control _ a priori _ on the position of the unstable bundle @xmath933 of the cocycle on a set of positive measure of abelian differentials . in the case of a canonical absolutely continuous invariant measure
such a set can be found near the boundary of the moduli space and the full non - vanishing of the lyapunov spectrum can thus be proved ( see @xcite , 4 and 8.2 ) .
later a. avila and m. viana @xcite proved the simplicity of the lyapunov spectrum , that is , that all the exponents are non - zero and distinct .
the simplicity of the top exponent , that is , the strict inequality @xmath934 is much simpler .
w. veech @xcite proved it for the canonical absolutely continuous invariant measures on connected components of strata .
this result was generalized by the first author in @xcite to an arbitrary teichmller invariant ergodic probability measure .
recently , forni has developed his method from @xcite to give a general criterion for the non - uniform hyperbolicity of the kontsevich zorich spectrum for a wide class of @xmath0-invariant measures @xcite .
the criterion is based on a topological condition on completely periodic directional foliations contained in the support of the measure . closing the considerations of this section , we observe that the following fundamental question is wide open : [ pb : rank ]
does there exist any finite @xmath0-invariant ergodic measure such that the number of strictly positive exponents for the kontsevich
zorich cocycle differs from the maximal rank of the bilinear form @xmath136 at a positive measure set in the space of abelian differentials ?
arithmetic teichmller curves of square - tiled cyclic covers ( see e.g. the subsection below and/or @xcite for definitions ) provide a basic model for the discussion of relations between the lyapunov spectrum and the geometry of the hodge bundle .
in particular , we will see that , _ in the case of square - tiled cyclic covers _ , the annihilator @xmath935 of the second fundamental form @xmath936 _ coincides _ with the neutral oseledets bundle @xmath406 .
therefore , in the context of square - tiled cyclic covers , the rank of the second fundamental form ( at the generic point with respect to an @xmath0-invariant measure ) equals the number of strictly positive lyapunov exponents , and the corank of the second fundamental form is always completely explained by the symmetries of the underlying surfaces .
however , we advance that these relations _ are not true in general _ , as an example in appendix [ s : rk : b : z ] below will show : actually , the best _ general _ results available ( on the relationship between @xmath935 and @xmath406 ) were already given above in theorem [ th : ann ] , theorem [ thm : lowerbound ] and theorem [ thm : degenerateb ] . in the sequel , we recall the definition of square - tiled cyclic covers and some of its basic properties . for more details and proofs of the statements below , see e.g. @xcite .
let @xmath937 be an integer , and consider @xmath938 be a @xmath763-tuple of integers satisfying the following conditions : @xmath939 let @xmath940 be four distinct points .
conditions imply that , possibly after a desigularization , a riemann surface defined by equation @xmath941 is closed , connected and nonsingular .
we denote this riemann surface by @xmath942 . puncturing the ramification points we obtain a regular @xmath943-fold cover over @xmath944 .
the group of deck transformations of this cover is the cyclic group @xmath945 with a generator @xmath946 given by @xmath947 where @xmath948 is a primitive @xmath943th root of unity , @xmath949 .
next , we recall that any meromorphic quadratic differential @xmath950 with at most simple poles on a riemann surface defines a flat metric @xmath951 with conical singularities at zeroes and poles of @xmath75 .
consider a meromorphic quadratic differential @xmath952 on @xmath953 .
for a convenient choice of parameters @xmath954 the rectangles become unit squares . therefore , given @xmath955 as above and denoting by @xmath956 the canonical projection @xmath957 , we have that the quadratic differential @xmath958 on @xmath955 induces a flat structure naturally tiled by unit squares ( by construction ) . in other words , we get in this way a _
square - tiled surface _ ( or _ origami _ or _ arithmetic translation surface _ ) @xmath959 . during the present discussion
, we will focus _ exclusively _ on the orientable case @xmath204 for an abelian differential @xmath16 .
by lemma 2 of @xcite , this amounts to assume that @xmath943 is _ even _ , and @xmath960 is _ odd _ ,
@xmath961 in what follows .
finally , closing this preliminary subsection on square - tiled cyclic covers , we recall ( for later use ) the following property : [ lm : minus : omega ] in the case @xmath962 for an abelian differential @xmath16 , one has that the form @xmath16 is anti - invariant with respect to the action of a generator of the group of deck transformations , @xmath963 before entering into the discussion of the relation between the annihilator @xmath935 of the second fundamental form and the neutral oseledets bundle @xmath964 in the context of square - tiled cyclic covers , we recall below one of the _ only _ two examples of teichmller curves of square - tiled cyclic covers with maximally degenerate kontsevich zorich spectrum .
the first example to be discovered , in genus three , is the teichmller curve of the well - known eierlegende wollmilchsau ( see @xcite , 7 , and @xcite ) .
the second example , in genus four , was announced in @xcite and will be presented below . these examples were
the motivation for a full investigation of the spectrum of square - tiled cyclic covers carried out in @xcite and , from a slightly different perspective , in the next subsections of this appendix . in @xcite
we showed that there are no other ( maximally degenerate ) examples among square - tiled cyclic covers ( see @xcite , for a stronger result in the class of all veech surfaces and @xcite for progress on the general case ) .
both examples were discovered as an application of the symmetry criterion given by theorem [ thm : degenerateb ] to the arithmetic teichmller curves of the square - tiled cyclic covers @xmath965 and @xmath966 respectively .
below , we will follow the presentation in @xcite and apply the symmetry criterion to the arithmetic teichmller curve of the square - tiled cyclic cover @xmath966 to derive that it is maximally degenerate .
the lyapunov spectrum of the hodge bundle over the geodesic flow on the teichmller curve of cyclic covers @xmath966 is maximally degenerate , @xmath967 . by lemma [ lm :
minus : omega ] , the generator @xmath742 of the group of deck transformations of @xmath966 acts on @xmath16 as @xmath968 , see ; in particular @xmath969 , and @xmath970 .
we have an explicit basis @xmath971 in the space of holomorphic @xmath30-forms on @xmath966 ; the corresponding eigenvalues are @xmath972{-1}\ .\ ] ] it is easy to see that for all couples of indices @xmath973 except @xmath974 one has @xmath975 . by theorem [ thm : degenerateb ] , the proof is complete . in both cases of @xmath976 and @xmath966 , the symmetries given by the deck transformations force the rank of the second fundamental form to be equal to @xmath30 ( its lowest possible value ) and there is exactly @xmath30 strictly positive exponent , the top exponent of the tautological bundle . in the sequel , we will extend this picture to the _ whole _ class of square - tiled cyclic covers . the identity between the rank of the second fundamental form and the number of strictly positive kontsevich
zorich exponents always holds for square tiled - cyclic covers .
in fact , we have the following result .
[ th : n : of : exp : equals : rk ] let @xmath955 be a square - tiled cyclic cover with @xmath943 even and all @xmath960 odd .
the rank of @xmath136 is constant for all @xmath16 in the corresponding @xmath0-orbit and it is equal to the number of strictly positive kontsevich zorich exponents on the hodge bundle @xmath19 .
the result was inspired by @xcite , and it can be obtained as a corollary of the results of that paper . for the sake of completeness , we present in the remaining part of this appendix a proof of theorem [ th : n : of : exp : equals : rk ] . such a proof is based on theorem [ th : ann ] and a remarkable property of square - tiled cyclic covers , namely , the existence of an explicit @xmath0-invariant , @xmath810-orthogonal splitting of the hodge bundle @xmath19 over a teichmller curve of a square - tiled cyclic cover into subbundles of small dimension ( @xmath898 or @xmath763 ) .
we start with the description of the splitting , see @xcite and @xcite for more details .
consider a generator @xmath742 of the group of deck transformations of the cyclic cover , see .
the induced linear map @xmath977 verifies @xmath978 , hence its eigenvalues are @xmath943-th roots of unity , that is , they form a subset of the set @xmath979 , where @xmath948 is an @xmath943-th primitive root of unity . for all @xmath980 ,
let @xmath981 since the deck transformation @xmath742 commutes with the @xmath0-action , each @xmath982 is @xmath0-invariant subbundle of @xmath20 . on the other hand , @xmath748 has a well - defined restriction to @xmath983 , and @xmath984 , so that we can also define @xmath985 in _ general _ , the subspaces @xmath986 do _ not _ form @xmath0-invariant subbundles of the complex cohomology bundle .
let us consider the decomposition @xmath987 since the operator @xmath748 preserves the subspace @xmath32 and commutes with complex conjugation , it follows that @xmath988 in particular , by defining @xmath989 and @xmath990 we see that , for any @xmath991 , one has @xmath992 and , for @xmath993 , one gets @xmath994 thus , each @xmath995 is @xmath0-invariant and hodge star - invariant . concerning the dimensions of these subbundles , we have the following lemma .
denote @xmath996 where @xmath997 denotes a fractional part of @xmath5 .
conditions imply that @xmath998 is integer , hence , clearly , @xmath999 .
for any @xmath1000 , one has @xmath1001 define the following two complementary subsets @xmath1002 and @xmath1003 of the set @xmath1004 : @xmath1005 i_0:=\{k | 1kn/2 , v^1,0_k v^1,0_n - k } , @xmath1006 i_1:=\{1,
,n/2}i_0 , @xmath1007 h_i_0 ^ 1(s , ) : = _
ki_0 h^1_k(s , ) , h_i_1 ^ 1(s,):=_ki_1 h^1_k(s , ) .
@xmath1008h^1(s,)=h_i_0 ^ 1(s , ) h_i_1 ^ 1(s,).@xmath1005 in this language , a consequence of the previous lemma is : [ cor : q ] an integer @xmath321 , such that @xmath1009 , belongs to the subset @xmath1003 if and only if @xmath1010 the definition of the integers @xmath998 and conditions on the sum of @xmath960 imply that @xmath1011 , or , equivalently , @xmath1012 it follows that if @xmath1013 then the integers @xmath1014 and @xmath1014 are both different from zero , hence they have to be both equal to @xmath30 , as claimed . in any case , for a square - tiled cyclic cover , we can use these subbundles to compute the complex bilinear form @xmath103 on @xmath41 as follows .
[ lm : b : i : j ] let @xmath1015 , @xmath1016 be eigenvectors of the linear map @xmath1017 with eigenvalues @xmath1018 and @xmath1019 respectively .
the following formula holds : @xmath1020 symmetry arguments analogous to the ones in the proof of theorem [ thm : degenerateb ] show that , in our case , if the eigenvalues @xmath1021 and @xmath1022 are not complex - conjugate , the value @xmath1023 on the corresponding eigenvectors is equal to zero . since @xmath1024 ,
one has that @xmath1025 implies @xmath1026 .
now , consider the action of @xmath748 on the abelian differential @xmath1027 . by lemma [ lm :
minus : omega ] , we get @xmath1028 that is , @xmath1027 is @xmath748 anti - invariant . in other words , since @xmath1029 , we have that @xmath1030 . by corollary [ cor : q ] , this implies that @xmath1031 is proportional to @xmath16 with a nonzero constant coefficient @xmath1032 .
thus , @xmath1033 this completes the proof of the lemma .
our discussion so far can be summarized by the following lemma : [ lemma : realsplitting ] the real hodge bundle @xmath19 over an arithmetic teichmller curve of a square - tiled cyclic cover splits into a direct sum @xmath1034 of @xmath0-invariant , @xmath810-orthogonal , hodge star - invariant subbundles . by lemma [ lm :
equivalence : hodge : and : symp : orthogonal : under : star : invariance ] and lemma [ lemma : realsplitting ] the subspaces @xmath995 in the splitting are symplectic - orthogonal and hodge - orthogonal .
of course , it can be also immediately seen directly .
given a teichmller curve @xmath1035 associated to a square - tiled cyclic cover , we denote @xmath1036 the bundle over @xmath1035 formed by the subspaces @xmath1037 , and @xmath1038 be the oseledets bundle corresponding to the zero lyapunov exponents ( with respect to the unique @xmath0-invariant probability supported on @xmath1035 ) .
the lemma below is a consequence of theorem [ th : ann ] and lemma [ lm : b : i : j ] .
[ lemma : inclusion ] the following inclusions hold : @xmath1039 for each @xmath1040 , one has @xmath1041
. hence @xmath1042 by lemma [ lm : b : i : j ] .
it follows that @xmath1043 , that is , @xmath1044 . on the other hand
, we have that , by construction , @xmath1045 is @xmath0-invariant and hodge star - invariant .
thus , the inclusion @xmath1046 follows from theorem [ th : ann ] ( and the inclusion @xmath1044 ) . in what follows ,
we want to show that the inclusions in the previous lemma are actually equalities .
the first step is to prove the following fact : [ cor : hp1 ] the following identity holds : @xmath1047 for any @xmath1013 , take @xmath1048 and @xmath1049 .
it follows from corollary [ cor : q ] that such @xmath1016 and @xmath1050 exist , and @xmath1051 , @xmath1052 . by and
, we obtain that @xmath1053 , @xmath1054 , @xmath1055 , @xmath1056 is a basis of @xmath995 when @xmath1057 , and @xmath1058 , @xmath1059 is a basis of @xmath1060 . by lemma [ lm :
b : i : j ] , we deduce that the @xmath1061 is non - degenerate for any @xmath1013 .
since @xmath1062 is a @xmath810-orthogonal sum ( see lemma [ lemma : realsplitting ] ) , it follows that the restriction of @xmath810 to @xmath1062 is non - degenerate .
finally , since @xmath21 splits as a @xmath810-orthogonal sum of @xmath1062 and @xmath1063 ( see again lemma [ lemma : realsplitting ] ) , it follows that @xmath1064 . because the converse inclusion was proved as part of lemma [ lemma : inclusion ] , the proof of this lemma is complete .
next , we invoke the following key result : [ lemma : qexp ] for every @xmath1013 the kontsevich
zorich cocycle has no zero exponents on the @xmath0-invariant subbundle @xmath1065 .
moreover , for each @xmath1066 , the lyapunov spectrum of kontsevich - zorich cocycle restricted to @xmath1067 has the form @xmath1068 with @xmath1069 ( i.e. , there is only one double positive lyapunov exponent ) . this theorem could be deduced _
directly _ from the properties of @xmath1070 discussed in this paper .
however , we prefer to skip the presentation of this proof because , contrary to the arguments in @xcite , it does nt yield the _ precise _ value @xmath1071 . at this stage , we are ready to get the following result stating in particular that for square - tiled cyclic covers the central oseledets subbundle indeed coincides with the kernel of the second fundamental form : [ thm : cyclic : covers ] suppose that @xmath943 is even and all @xmath960 are odd .
consider the @xmath0-orbit of the square - tiled cyclic cover @xmath955 .
the following identities hold : @xmath1072 in particular the number of strictly positive lyapunov exponents of the kontsevich
zorich cocycle is given by the following formula : @xmath1073 the decomposition of the hodge bundle into a direct sum of @xmath0-invariant subbundles implies that the lyapunov spectrum of the hodge bundle is the union of the spectra of the subbundles .
lemmas [ lemma : inclusion ] and [ cor : hp1 ] show that @xmath1074 , while theorem [ lemma : qexp ] , shows that the inclusion @xmath1075 is , actually an equality .
finally , the formula counting the number of nonnegative lyapunov exponents is obtained by a combination of theorem [ lemma : qexp ] with corollary [ cor : q ] .
theorem [ th : n : of : exp : equals : rk ] is now an immediate corollary of the more precise theorem [ thm : cyclic : covers ] .
this appendix is devoted to the description of an @xmath0-invariant locus @xmath1076 supporting an ergodic @xmath0-invariant probability @xmath76 with the following properties : * there are precisely @xmath1077 vanishing exponents out of @xmath1078 non - negative kontsevich zorich exponents ( with respect to @xmath76 ) , * the corank of @xmath1079 is @xmath1077 for every @xmath1080 , but * @xmath1081 , @xmath1082 is not @xmath0-invariant and @xmath880 is not flow - invariant .
denote by @xmath11 the family of riemann surfaces @xmath1083 equipped with the abelian differentials @xmath1084 here , @xmath1085 are @xmath1077 pairwise distinct points of the riemann sphere .
a quick inspection reveals that these riemann surfaces have genus 10 , and any such @xmath16 has a zero of order @xmath1086 at the branch point over @xmath1087 and a double zero at the branch point over @xmath1088 , @xmath1089 . in other words , @xmath1076 .
observe that @xmath1090 where @xmath1091 , @xmath1092 , and @xmath75 is the following quadratic differential with a simple zero and 5 simple poles of the riemann sphere : @xmath1093 alternatively , @xmath1094 where @xmath1095 , @xmath1096 , @xmath1097 is a genus two riemann surface and @xmath1098 is the following abelian differential on @xmath1099 : @xmath1100 since @xmath11 , @xmath1101 and @xmath1102 are @xmath763-dimensional loci , it follows that @xmath1103 and @xmath11 is the closure of the @xmath1104-orbit of @xmath1105 for a generic choice of @xmath1106 .
similarly to the discussion of the case of square - tiled cyclic covers in appendix [ s : rk : b : cyclic : covers ] above , we notice that the group of deck transformations of the cover @xmath1091 ( ramified over @xmath1107 ) is generated by @xmath1108 where @xmath1109 is a primitive @xmath1077th root of unity . for sake of concreteness
, we take @xmath1110 . again
, we denote by @xmath1111 the induced linear map . of course , since @xmath1112 , its eigenvalues are a subset of @xmath1113 . for every @xmath1114
, we put @xmath1115 again , because @xmath0 acts by monodromy of the flat gauss - manin connection , these eigenspaces @xmath982 form @xmath0-invariant subbundles of the hodge bundle over @xmath11 .
next , we consider the restriction of @xmath1116 to @xmath1117 . because there is no abelian differential on @xmath953 ,
we see that the eigenvalues of @xmath1118 form a subset of @xmath1119 and we denote @xmath1120 so that @xmath1121 a quick computation shows that @xmath1122 is a basis of holomorphic differentials on the genus 10 riemann surface @xmath1123 .
in particular , @xmath1124 for each @xmath1114 .
since @xmath1127 and @xmath1128 are invariant under complex conjugation , each @xmath982 is @xmath0-invariant and @xmath1129 , we have the following hodge-@xmath1130 and @xmath0-invariant splitting @xmath1131 moreover , a direct application of the arguments of the proof of theorem [ thm : degenerateb ] shows that the @xmath1132 are pairwise @xmath882-orthogonal .
[ [ neutral - oseledets - bundle - versus - textrmannbmathbbr - over - mathcalz ] ] neutral oseledets bundle versus @xmath1133 over @xmath11 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in the present case , we just saw that @xmath1134 , so that @xmath1135 , @xmath1136 and , _ a fortiori _ ,
@xmath1137 and @xmath1138 ( compare with theorem [ thm : degenerateb ] and lemma [ lm : b : i : j ] ) .
furthermore , from theorem [ th : ann ] , it follows also that @xmath1139 for any ergodic @xmath0-invariant @xmath76 supported on @xmath11 , that is , any such measure automatically possesses at least @xmath763 vanishing exponents among the @xmath1078 non - negative exponents of kontsevich - zorich cocycle restricted to the hodge bundle over @xmath11 .
the fact @xmath1140 says that the jacobian of @xmath1123 has a _ fixed part _
( of dimension @xmath1141 ) , i.e. , the complex torus @xmath1142 obtained from the quotient of @xmath1143 by the lattice @xmath1144 is a fixed part ( rigid factor ) of the jacobian @xmath1145 of @xmath1123 in the sense that we have a isogeny @xmath1146 .
the fact that @xmath1143 is a fixed part of the jacobian was already known by c. mcmullen ( * ? ? ? * theorem 8.3 ) and it was our starting point to study the locus @xmath11 .
next , we pass to the analysis of @xmath1147 . because @xmath1148 equals @xmath1149 ( where @xmath1095 , @xmath1096 is the covering map used above to construct the isomorphism @xmath1150 ) , by the results of bainbridge @xcite and eskin , kontsevich and zorich @xcite , we conclude that the non - negative exponents of the kontsevich - zorich cocycle restricted to @xmath1147 are @xmath30 and @xmath1151 for any ergodic @xmath0-invariant probability supported on @xmath11 and the rank of the second fundamental form @xmath103 restricted to @xmath1149 has full rank ( equal to @xmath898 ) . for sake of concreteness
, we observe that @xmath1153 is a copy of the first homology group @xmath1154 of the genus @xmath763 riemann surface @xmath1155 moreover , the square of the abelian differential @xmath16 on @xmath1123 projects into the quadratic differential @xmath1156 . in other words , the study of the restriction of the kontsevich
zorich cocycle over @xmath11 to @xmath1153 is equivalent to the study of the kontsevich
zorich cocycle over the locus determined by the family of quadratic differentials @xmath1157 of genus @xmath763 .
in fact , in the forthcoming paper @xcite , we will adopt the latter point of view ( i.e. , we will study directly this family of quadratic differentials of genus @xmath763 ) .
[ lemma : rk : b : z ] the restriction of the form @xmath103 to @xmath1158 has rank @xmath898 . in particular , @xmath1159 is a @xmath763-dimensional ( real ) subspace of the @xmath1086-dimensional space @xmath1153 .
let @xmath1160 and @xmath1161 be a basis of @xmath1158 . by the arguments in the proof of theorem [ thm : degenerateb
] , we have that @xmath731 and @xmath1162 for all @xmath1163 . hence , the matrix of @xmath1164 in the basis @xmath1165 is @xmath1166 where @xmath1167 , @xmath1168 .
hence , it suffices to prove that one of the entries @xmath1169 is non - zero to conclude that the rank of @xmath1164 is @xmath898 . to do so
, we make the following choice of basis @xmath1170 , @xmath1171 , and we compute @xmath1172 that is @xmath1173 .
[ lemma : rk : ec : z]the non - negative part of the lyapunov spectrum of the restriction of the kontsevich - zorich cocycle to @xmath1153 with respect to any @xmath0-invariant ergodic probability @xmath76 has the form @xmath1174 in particular , the neutral oseledets bundle @xmath1175 intersects @xmath1153 in a subspace of real dimension @xmath763 .
we begin the proof of this lemma by following @xcite to see that the lyapunov exponents of @xmath1176 have multiplicity @xmath898 ( at least ) . indeed , given a vector @xmath1177 corresponding to a lyapunov exponent @xmath1178
, we have that @xmath1179 corresponds to the same lyapunov exponent ( as @xmath1116 commutes with the monodromy ) . on the hand ,
since the eigenvalues of @xmath1180 are not real ( i.e. , @xmath1181 ) , it follows that @xmath1179 is not collinear to @xmath666 , that is , @xmath1178 has multiplicity @xmath898 at least . on the other hand , we know that @xmath810 restricted to @xmath1153 is not degenerate ( see lemma [ lemma : rk : b : z ] ) .
hence , from theorem [ th : ann ] , we conclude that @xmath1183 in other words , the proof of the previous lemma is reduced to show that @xmath1184 , that is , in the forthcoming paper @xcite , we will deduce this lemma along the following lines . since @xmath1185 , @xmath1186 , and @xmath1187 ,
the intersection form has signature @xmath1188 and hence the action of @xmath0 through monodromy of the gauss - manin connection on the subspace @xmath1189 of the complex hodge bundle @xmath1190 is represented by @xmath1191 matrices . in @xcite , we will see that , _ in general _ , a cocycle preserving a pseudo - hermitian form of signature @xmath1192 ( i.e. , with values in the matrix group @xmath1193 ) has @xmath1194 zero lyapunov exponents _ at least_. by applying this general principle in the context of the previous lemma , we have that the kontsevich zorich cocycle has ( at least ) @xmath1195 zero lyapunov exponents , so that the lemma follows .
in any event , by lemmas [ lemma : rk : b : z ] and [ lemma : zero : exp ] , we can discuss the main result of this appendix . in the following theorem ,
we denote by @xmath76 the natural @xmath0-invariant ergodic probability fully supported on @xmath11 ( obtained from the so - called _ masur - veech _ probability on @xmath1102 via the isomorphism @xmath1150 previously constructed ) . assuming momentarily this theorem
, one can conclude theorem [ thm : annb : ec ] as follows .
we have that @xmath1199 is a continuous ( actually , real - analytic ) and @xmath318-invariant @xmath898-dimensional subbundle of @xmath1158 . from theorem [ thm : matrix :
comp ] it follows that @xmath1199 is not flow - invariant , and hence @xmath1159 is not flow - invariant .
since @xmath1200 is flow - invariant , it follows that @xmath1200 can not coincide with @xmath1159 . by theorem
[ th : ann ] , if @xmath879 were @xmath0-invariant , it would be a subbundle of the bundle @xmath1133 , hence it would coincide with it by dimensional reasons , thereby contradicting the first part of the statement . concerning the proof of theorem [ thm : matrix : comp ] ,
let us just say a few words ( for more details see @xcite ) : the basic idea is that @xmath0-invariance of a continuous subbundle @xmath445 can be tested along _ pseudo - anosovs _
( i.e. , periodic orbits of teichmller flow ) ; indeed , the existence of @xmath0-invariant continuous subbundles @xmath445 implies that the monodromy matrices along pseudo - anosovs passing by the same riemann surface should share a common subspace , and ( the non - validity of ) this last property can be tested by direct calculation . closing this appendix ,
let us mention that the _ precise _ value of the positive lyapunov exponent coming from @xmath1153 can be computed from the main formula of @xcite and a computation with siegel - veech constants related to @xmath1101 ( see @xcite ) : [ pr : exp : value : z]let @xmath479 be any @xmath0-invariant ergodic probability supported on @xmath11 .
then , the non - negative part of the lyapunov spectrum of the kontsevich zorich cocycle with respect to @xmath479 has the form @xmath1201 we would like to heartily thank a. avila , a. eskin , p. hubert , m. kontsevich , m. mller and j .- c .
yoccoz for several discussions related to the material of this paper . also , we are thankful to the anonymous referees for their comments and suggestions leading to the current version of this article .
the authors thank collge de france , institut des hautes tudes scientifiques ( ihes ) , hausdorff research institute for mathematics ( him ) and max planck institute for mathematics ( mpim ) for their hospitality during the preparation of this paper .
a. zorich , _ flat surfaces _ , in collection `` frontiers in number theory , physics and geometry .
vol . 1 : on random matrices , zeta functions and dynamical systems ''
; ecole de physique des houches , france , march 921 2003 , p. cartier ; b. julia ; p. moussa ; p. vanhove ( editors ) , springer - verlag , berlin , ( 2006 ) , 439586 . | we study the lyapunov spectrum of the kontsevich zorich cocycle on @xmath0-invariant subbundles of the hodge bundle over the support of @xmath0-invariant probability measures on the moduli space of abelian differentials .
in particular , we prove formulas for partial sums of lyapunov exponents in terms of the second fundamental form ( the kodaira spencer map ) of the hodge bundle with respect to gauss
manin connection and investigate the relations between the central subbundle and the kernel of the second fundamental form .
we illustrate our conclusions in two special cases . |
exploration of comets with space missions is critical to probe the coma and to access the detailed features of a cometary nucleus .
however , remote spectroscopic observations can provide ample information about the global composition of comets . the coma composition associated with main species
can be constrained from the analysis of airglow emissions using ground- and space - based telescopes .
this nevertheless requires the quantitative assessment of physical processes that govern these emissions in the coma . among many observed cometary ultraviolet and visible airglow spectra
, metastable emission lines have gained special interest .
solar resonance fluorescence is not an effective excitation mechanism to populate excited metastable states in the coma due to optically forbidden transitions .
the dissociative excitation by photons , suprathermal electrons such as photoelectrons , and thermal recombination of ions are the main channels for producing various species in metastable excited states . by observing the emissions from the daughter products , which are particularly from metastable state ,
the estimation of mixing ratios of their respective parent species has been done in several comets @xcite .
the lifetime of a metastable excited species is shorter ( @xmath10.7 s for o(@xmath2s ) , @xmath1110 s for o(@xmath2d ) , and @xmath13 ms for co(a@xmath3 ) ) compared to their respective parent species ( e.g. , for h@xmath0o it is @xmath18 @xmath4 10@xmath5 s@xmath6 at 1 au ) , and can not travel large radial distances in the coma from the place of formation without being lost through the emission of photons or quenched in collision .
hence , these forbidden emissions are good tracers to quantify the gas production rates of respective parent species in the coma .
though water is the dominant species in comets , it is very difficult to access cometary h@xmath0o infrared emissions from ground - based observatories because of strong absorption by terrestrial water molecules .
however , the spatial profiles of cometary h@xmath0o have been observed by ground - based telescopes by observing non - resonance fluorescence emissions @xcite .
since h@xmath0o does not have any transitions in ultraviolet and visible regions , the forbidden emissions of its dissociative metastable products ( [ oi ] 6300 , 6364 , 5577 , 2972 ) have been used as tracers .
oxygen atoms that are produced in the @xmath2s state , decay 95% of the time to the @xmath2d state emitting photons at 5577 ( green line ) wavelength , while 5% of them decay directly to the ground @xmath7p state which yields 2972 and 2958 emission lines .
the radiative decay of @xmath2d @xmath8p leads to 6300 and 6364 emission lines ( red - doublet emission ) .
thus , if the green line is present in a cometary spectrum , the red - doublet must also be present , although the red - doublet can be formed without the green line . the quantification of h@xmath0o in the cometary coma has been done by observing [ oi ] 6300 emission in many comets @xcite .
the direct de - excitation of o(@xmath2s ) yields 2972 emission line which has been observed in cometary spectrum only once @xcite .
since there are other oxygen bearing species , such as co@xmath0 and co , which can also produce these metastable states via dissociative excitation reactions , the observed green to red - doublet emission intensity ratio ( hereafter referred as g / r ratio ) has been used to confirm the parent source of these emission lines @xcite .
based on o(@xmath2s ) and o(@xmath2d ) photorates calculated by @xcite , the observed high g / r ratio values ( @xmath9 0.1 ) were ascribed to large co@xmath0 and co volume mixing ratios in the coma @xcite .
the spin forbidden atomic oxygen emission line [ oi ] 1356 has also been detected using rocket- and space - borne uv spectrometers in comets @xcite .
since fluorescence efficiency ( g - factor ) for this transition is small to explain the observed intensity , production sources for this emission are attributed to electron impact excitation mechanisms @xcite .
@xcite accounted for various electron impact excitation sources to explain the observed emission intensity in comet halley , which had gas production rate of 1.3 @xmath4 10@xmath10 s@xmath6 at 1 au .
co@xmath0 is also an important oxygen bearing species in the cometary coma but is difficult to detect directly in the visible and ultraviolet cometary spectra because of the lack of electronic transitions . in order to quantify co@xmath0 in comets , co cameron band emissions (
a@xmath3 @xmath11 a@xmath12 ) have been used as a tracer of co@xmath0 by assuming that the excited metastable state a@xmath3 ( lifetime is @xmath13 ms , * ? ? ?
* ) originate primarily from photodissociation of co@xmath0 @xcite .
this spin - forbidden electron transition ( a@xmath3 @xmath11 a@xmath12 ) yields a band emission in the ultraviolet spectral range 18002600 . using hubble space telescope ( hst ) , @xcite detected this band emission on comet 103p / hartley 2 .
the observation of co cameron band emission on comet 103p led to the re - examination of international ultraviolet explorer ( iue ) observed spectra and @xcite identified this emission on four comets ( viz .
, c/1979 y1 ( bradfield ) , 1p / halley , c/1989 x1 ( austin ) , c/1990 k1 ( levy ) ) .
the volume mixing ratios of co relative to water have been derived in these comets by assuming that photodissociation of co@xmath0 is the main source of co(a@xmath3 ) . however , besides photodissociation of co@xmath0 , co(a@xmath3 ) can also be produced via electron impact and dissociative recombination of co - bearing species @xcite .
atomic carbon [ ci ] 1931 emission line has been observed in several comets @xcite .
the excited state of this emission is a metastable state c(@xmath2d ) which has a lifetime of about 4080 s @xcite . the radiative decay of carbon from @xmath2d to the ground @xmath7p state results in photons at 9823 and 9850 , which are analogous to atomic oxygen red - doublet emissions at 6300 and 6464 .
the [ ci ] 9850 emission line has been detected in comet hale - bopp by @xcite .
the carbon atom in @xmath2d metastable state fluoresces the solar photons at 1931 before radiative decay to the ground @xmath7p state can occur .
most of the emission line intensity is attributed to photodissociative excitation of co in comets @xcite .
hence this emission line is a good tracer for co production rate in comets @xcite .
the model developed by @xcite calculated this emission line intensity in comet halley , which is smaller by a factor 5 than that observed by iue , and suggested an involvement of additional carbon - bearing species in the coma . in order to derive the parent species production rates in the coma based on the observed forbidden emission intensities
a quantitative study of various processes that govern these emissions is necessary .
we have developed a coupled chemistry - emission model which accounts for major production and loss reactions of o(@xmath7p ) , o(@xmath2s ) , o(@xmath2d ) , c(@xmath7p ) , c(@xmath2d ) , and co(a@xmath3 ) in cometary comae @xcite .
this model has been applied on several comets and results have been compared with the earth - based observations @xcite .
the model calculations for comets 103p / hartley 2 and 1p / halley have shown that suprathermal electron impact is an important excitation process in the formation of co(a@xmath3 ) , which is more important than photodissociation of co@xmath0@xcite . the model applied to study atomic oxygen emission lines in comets has shown that the collisional quenching in the inner coma can significantly change the observed g / r ratio@xcite .
the model calculations in active comets , such as ( c/1996 b2 ) hyakutake and ( c/1995 o1 ) hale - bopp , have shown that the g / r ratio varies as a function of projected distance and depends on the collisions in the cometary comae @xcite .
recently , we have applied our model for the analysis of high - resolution spectroscopic observations made from eso very large telescope ( vlt ) on four comets ( viz .
, c/2002 t7 ( linear ) , 73p - c / schwassmann - wachmann 3 , 8p / tuttle , and 103p / hartley 2 ) .
this study has allowed to constrain the co@xmath0 volume mixing ratios in these comets @xcite . after a successful rendez - vous in august 2014
, esa s rosetta spacecraft is exploring comet 67p / churyumov - gerasimenko(here - after , referred as 67p ) by escorting it from @xmath14 au towards perihelion at 1.29 au reached in the summer 2015 . assessing the chemical evolution of the cometary coma , as the comet approaches the sun ,
is one of the main aims of rosetta mission . in support to rosetta , different space- and ground - based observation campaigns are taking place to understand the spatial distribution of different volatile species in the coma .
we apply our coupled chemistry - emission model to comet 67p to identify and quantify the processes driving formation and loss of co cameron band , atomic oxygen forbidden emissions and [ ci ] 1931 emission line at perihelion and at 3 au heliocentric distance .
model calculations are necessary to understand the physical processes governing these metastable emissions and to derive co@xmath0 and co mixing ratios .
the motivation for this work is to provide a theoretical support for the interpretation of earth - based and rosetta - based uv observations of comet 67p .
we present the model input parameters , which may represent the gaseous environment of comet 67p at its perihelion and at 3 au , in section [ model - inputs ] .
the modelled various production and destruction profiles for different electronic states of atomic oxygen , and atomic carbon , and co(a@xmath3 ) , and emission intensities as a function of projected distance are presented in section [ results ] .
the implications of modelled emission profiles for comet 67p are discussed in section [ discussion ] .
conclusions are given in section [ conclusions ] .
the coupled chemistry - emission model has been used in the present study which accounts for main production and loss processes for co(a@xmath3 ) , o(@xmath7p ) , o(@xmath2d ) , o(@xmath2s ) , c(@xmath7p ) , and c(@xmath2d ) species in the inner cometary coma , as described in the earlier works @xcite .
calculations are done at perihelion distance ( 1.29 au ) .
the h@xmath0o out - gassing rate at perihelion is assumed to be 1 @xmath4 10@xmath13 s@xmath6 @xcite .
the number density relative to water is taken to be 5% for both co@xmath0 ( hearafter @xmath14(co@xmath0 ) ) and co ( hearafter @xmath14(co ) ) as a standard neutral composition .
the on - board osiris instrument on rosetta mission observed 67p s nucleus has bi - lobed structure with dimension 2.5 @xmath4 2.5 @xmath4 2.0 km for the small lobe and 4.1 @xmath4 3.2 @xmath4 1.3 km for the large lobe @xcite . in our model
we have assumed a spherical nucleus of 2 km radius for simplification .
the neutral atmosphere is calculated using haser s formula @xcite which assumes spherical expansion of coma with a constant velocity of 1 km s@xmath6 . the electron temperature profile , which is required to calculate electron - ion recombination rate , is assumed to be same as on halley @xcite .
we vary co@xmath0 and co mixing ratios to quantify the change in contributions of different productions and loss processes yielding co(a@xmath3 ) , o(@xmath2s ) , o(@xmath2d ) , c(@xmath2d ) , and atomic carbon and atomic oxygen in ground states .
the incident solar flux is based on the measurements from the thermosphere ionosphere mesosphere energetics and dynamics ( timed)/solar euv experiment ( see ) @xcite on 2 january 2005 ( for solar activity phase with f10.7 = 100 @xmath4 10@xmath15 wm@xmath16 hz@xmath6 ) at earth and scaled to heliocentric distance of 1.29 au .
it is expected to be representative of conditions encountered in summer 2015 near perihelion from solar decreasing active period @xcite .
the theoretical water collisional zone of comet 67p , with gas production rate 10@xmath13 s@xmath6 , is around 2000 km @xcite .
the calculations presented in this work are relevant for the inner coma and inside the diamagnetic cavity .
we also made calculations at heliocentric distance of 3 au assuming the total gas production rate of 5 @xmath4 10@xmath17 s@xmath6 @xcite .
the neutral coma composition is assumed to be 80% h@xmath0o , 15% co and 5% co@xmath0 .
the solar flux on 1 november 2014 is used in the model and scaled to 3 au using inverse square of heliocentric distance .
the electron temperature - dependent reactions play a minor role in governing the intensities of these emission lines .
recently rosina / dfms has made several important discoveries such as the d / h ratio @xcite , presence of n@xmath0 @xcite and o@xmath0 @xcite in 67p s coma .
the observation of molecular oxygen has an important implication in the production of o(@xmath2s ) and o(@xmath2d ) .
@xcite found that the local abundance of molecular oxygen is varying between 1 and 10% around the 67p nucleus relative to h@xmath0o production rate .
the mean value of molecular oxygen abundance in 67p coma is 3.8 @xmath18 0.85% relative to h@xmath0o production rate @xcite . in order to quantify the contribution of molecular oxygen on the forbidden emission lines we have taken 4% molecular oxygen relative to h@xmath0o production rate in the model .
hence we also have done a case study by assuming the water production rate 5 @xmath4 10@xmath17 s@xmath6 and 25% co , 8.3% co@xmath0 , and 4% o@xmath0 as relative abundances with respect to h@xmath0o for the month of august 2014 .
these abundances are mostly in agreement with the rosina measurements between august and october 2014 @xcite when the comet was between 3 and 3.5 au from the sun .
the modelled co(a@xmath3 ) rate profiles for different formation processes in the coma of 67p are shown in figure [ prd - cam ] .
the number of excited co(a@xmath3 ) molecules produced per unit volume per second is referred to as volumetric production rate . for equal ( 5% relative to water ) co@xmath0 and co volume mixing ratios relative to water in the coma the major production sources of co(a@xmath3 ) is electron impact on co. the electron impact on co@xmath0 and photodissociation of co@xmath0 are next most dominant sources of co(a@xmath3 ) .
the thermal electron recombination of hco@xmath19 and co@xmath20 ions and fluorescence of co are minor co(a@xmath3 ) production sources . above 500 km
the contributions from photodissociative excitation of co@xmath0 and electron impact on co and co@xmath0 are nearly equal .
since the lifetime of co(a@xmath3 ) is short ( @xmath13 ms , * ? ? ?
* ) , most of the excited molecules decay to ground state by spontaneous emission .
hence the radiative decay is the major loss source of this excited state .
other loss processes , such as collisional quenching and ionization by photons and photoelectrons , are smaller compared to radiative decay by several orders of magnitude .
the number of species de - excite to ground state per second by various loss mechanisms is referred to as loss rate .
the cross section for electron impact excitation of various excited states and the calculated suprathermal electrons intensity in 67p coma at 10 km radial distance is presented in figure [ csc - eflux ] .
the suprathermal electrons intensity in the energy range between 1015 ev mainly determines the excitation rate of co(a@xmath3 ) through electron impact on co bearing species .
the modelled major production rate profiles of atomic oxygen and atomic carbon in ground states are presented in figure [ prod - ocg ] . below 50 km radial distances
, various sources are contributing to the formation of o(@xmath7p ) .
the production of atomic oxygen in ground state is mainly due to strong collisional quenching of o(@xmath2d ) with water .
the next important sources of atomic oxygen are charge exchange of oh@xmath19 and o@xmath19 ions with water .
photodissociation of co@xmath0 and co are next important o(@xmath7p ) production sources . above 50 km radiative decay of o(@xmath2d ) is the major source of atomic oxygen in ground state . the calculated major chemical loss rate profile for o(@xmath7p ) via different destruction mechanisms
are presented in figure [ los - ocg ] .
the major loss process for the atomic oxygen is due to collisions with oh molecules which yields atomic hydrogen and molecular oxygen .
the atomic oxygen can travel to large distances before getting lost in chemical reactions .
hence , we accounted transport loss by taking 1 km s@xmath6 as advection velocity .
we have accounted for many o(@xmath2s ) and o(@xmath2d ) formation and destruction processes in the coma as described in @xcite .
the modelled major production rate profiles for o(@xmath2s ) in comet 67p are shown in figure [ prd - o1s ] .
the photodissociation of h@xmath0o and co@xmath0 are equally important sources in producing o(@xmath2s ) in the inner coma of comet 67p .
below 100 km , suprathermal electron impact on co@xmath0 is next important o(@xmath2s ) source .
the photodissociation of co and electron impact on h@xmath0o and co@xmath0 are minor sources of o(@xmath2s ) .
electron recombination of h@xmath0o@xmath19 ion is a minor source of o(@xmath2s ) in the inner coma whereas its contribution is significant at large ( @xmath910@xmath7 km ) radial distances .
the calculated o(@xmath2d ) production rates profiles for different formation processes are presented in figure [ prd - o1d ] . the photodissociation of h@xmath0o is a dominant source of o(@xmath2d ) throughout the inner coma .
contribution from other o(@xmath2d ) formation processes is minor ( @xmath225% to the total ) . at large radial distances ( @xmath9 10@xmath7 km ) the contributions from dissociative recombination of h@xmath0o@xmath19 , radiative decay of o(@xmath2s ) and photodissociation of oh become significant in the formation of o(@xmath2d ) .
the modelled destruction rate profiles of o(@xmath2s ) and o(@xmath2d ) are shown in figure [ los - o1ds ] .
the o(@xmath2s ) and o(@xmath2d ) atoms are strongly quenched by h@xmath0o up to radial distances of @xmath110 and @xmath1200 km , respectively . above these radial distances the radiative decay , which leads to forbidden visible emission lines , is the major loss process for the o(@xmath2s ) and o(@xmath2d ) .
collisional quenching of o(@xmath2s ) and o(@xmath2d ) by other neutrals is smaller compared to h@xmath0o quenching by several orders of magnitude , hence these processes are not shown in the figure .
the production of atomic oxygen in @xmath21s state yields [ oi ] 1356 emission line via immediate decay to ground state ( lifetime is 185 @xmath23s , * ? ? ? * ) .
the calculated [ oi ] 1356 emission rates are presented in figure [ prod - oemis ] .
electron impact on atomic oxygen is the major production source for [ oi ] 1356 emission followed by electron impact on co@xmath0 and h@xmath0o . in case of atomic carbon formation
the photodissociation of co is the major source of c(@xmath7p ) as shown in figure [ prod - ocg ] .
collisional quenching of c(@xmath2d ) is next important source of atomic carbon in ground state .
all other production processes described in @xcite contribute little ( @xmath225% ) to the total .
the loss of atomic carbon is mainly due to collisions with oh which yields atomic hydrogen and co. the next main loss source is due to collisions with h@xmath24o@xmath19 which leads to the formation of hco@xmath19 and h@xmath0 .
the model accounts for transport of atomic carbon in @xmath7p and @xmath2d sates with an advection velocity of 1 km s@xmath6 .
transport is the major loss processes for atomic oxygen and atomic carbon compared to the total loss due to chemical reactions .
the calculated formation rates of metastable c(@xmath2d ) atom via different production processes are presented in figure [ prod - c1d ] . the major formation mechanism for c(@xmath2d )
is photodissociation of co. at large radial distance ( 10@xmath7 km ) dissociative recombination of co@xmath19 ion is also an important source of c(@xmath2d ) .
other formation reactions are smaller compared to photodissociation of co by more than an order of magnitude .
the modelled c(@xmath2d ) loss rates presented in figure [ los - c1d ] show that the collisional quenching with water is the dominant loss process up to 300 km radial distance , above which radiative decay takes over .
collisional quenching of co and co@xmath0 are relatively less significant c(@xmath2d ) loss processes .
the radial emission rate profiles are integrated for each emission line along the line - of - sight perpendicular to the sun - comet direction at different radial distances to obtain limb brightness profiles .
the model calculated intensity profiles , as a function of projected distance , for co cameron band , atomic oxygen ( [ oi ] 6300 + 6364 , 5577 , 2972 , and 1356 ) and atomic carbon emissions ( [ ci ] 1931 , 9823 , 9850 ) are shown in figure [ inten - emis ] . among the calculated emission intensities
co cameron band emission peaks close to the nucleus ( @xmath2220 km ) .
the calculated intensity profiles of 5577 and red - doublet ( 6300 + 6364 ) are flat up to radial distances 20 km and 200 km , respectively , due to strong collisional quenching of o(@xmath2s ) and o(@xmath2d ) with h@xmath0o in the inner coma .
the calculated g / r ratio as a function of projected distance is also presented in the same figure on the right y - axis . since lifetime of metastable c(@xmath2d )
is large ( 4080 s ) , the collisional quenching with water makes [ ci ] 1931 , 9850 and 9823 emission profiles flat up to 1000 km .
the [ oi ] 1356 line is the weakest emission among the calculated emissions ( presented in figure [ inten - emis ] after multiplying a factor 10 ) .
model calculated intensity profiles , when comet was at 3 au heliocentric distance , are presented in figure [ inten - emis3au ] as a function of projected distance . very close to the nucleus surface the oxygen red - doublet , green line , co cameron , and [ oi ] 1356 emissions are intense .
since neutral gas production rate is low ( 5 @xmath4 10@xmath17 s@xmath6 ) at 3 au the calculated intensity profiles are decreased by two order of magnitude compared to those at perihelion .
due to high radiative lifetime ( @xmath1110 s ) , the collisional quenching of o(@xmath2d ) is significant for radial distance up to 20 km , which alters the g / r ratio from 0.9 to 0.1 .
inspite of having high mixing ratio ( 15% ) , the role of co in determining the oxygen visible emission intensities as well as in determining the g / r ratio is insignificant . in the case of co cameron band ,
most of emission intensity ( @xmath990% ) close to nucleus is mainly via electron impact excitation of co. for radial distance higher than 50 km , the major ( @xmath150% ) source for co(a@xmath3 ) is co@xmath0 via electron impact and photodissociation excitation processes .
the long radiative lifetime ( @xmath14080 s ) of c(@xmath2d ) makes the atomic carbon emission intensity profile flat up to 1000 km . by varying the @xmath14(co@xmath0 ) and @xmath14(co ) the contribution of different processes producing co(a@xmath3 )
is calculated at three different projected distances .
the calculations are presented in table [ tab : per - cam ] by varying @xmath14(co@xmath0 ) and @xmath14(co ) from zero to two , and then to five percent . in all these cases the contribution from photodissociation of co@xmath0 and electron impact on co@xmath0 processes are nearly equal . keeping 1% @xmath14(co@xmath0 ) in the coma we varied @xmath14(co ) between 1 and 5% . in this case
the contribution of photodissociation of co@xmath0 and electron impact on co@xmath0 in the inner coma is @xmath2215% , whereas electron impact on co is about 6590% .
the contribution from other chemical reactions , such as electron recombination of hco@xmath19 and co@xmath20 ions contribute less than 10% to the formation of co(a@xmath3 ) . when @xmath14(co@xmath0 ) is increased to 2% , the contribution from electron impact on co
is reduced to 5080% .
about 1020% of co(a@xmath3 ) is produced from photodissociation of co@xmath0 . by increasing @xmath14(co@xmath0 ) to 5%
the contribution of co@xmath0 , from both photodissociation and electron impact reactions , in producing co(a@xmath3 ) increases to a total of 30 60% . in the case of
equal ( 5% ) @xmath14(co ) and @xmath14(co@xmath0 ) , the contribution from co is around 65% and the rest comes from co@xmath0-associated reactions in the inner coma . modelling results in table [ tab : per - cam ]
show that in the absence of co@xmath0 the main source for production of co(a@xmath3 ) is electron impact of co and other processes have negligible contribution to the total .
when co is absent in the coma electron impact on co@xmath0 and photodissociation of co@xmath0 processes are producing co(a@xmath3 ) with nearly equal contributions . in this case at 1000 km projected distances , the contribution of the thermal recombination of hco@xmath19 and co@xmath20 producing co(a@xmath3 ) is 15% . the calculated percentage contribution of different processes in the formation of o(@xmath2s ) and o(@xmath2d ) are presented in table [ tab : per - o1sd ] .
the main processes controlling the formation of o(@xmath2s ) is photodissociation of h@xmath0o and co@xmath0 .
the photodissociation of h@xmath0o is the dominant source of o(@xmath2d ) production in the inner coma . beyond 1000 km radial distances , the photodissociation of oh , electron recombination of h@xmath0o@xmath19 , and radiative decay of o(@xmath2s )
are also important o(@xmath2d ) sources .
the calculations presented in table [ tab : per - o1sd ] show that below 100 km radial distance the formation of o(@xmath2d ) is mainly ( 8090% ) through photodissociation of h@xmath0o .
above these distances this contribution changes to around 50% , while the rest is via photodissociation of oh and dissociative recombination of h@xmath0o@xmath19 and radiative decay of o(@xmath2s ) . in the case of o(@xmath2s ) production , both photodissociation of co@xmath0 and h@xmath0o are important formation processes in the inner coma .
it is found that the role of co photodissociation is very small ( @xmath225% ) in the o(@xmath2s ) production .
calculations presented in table [ tab : per - o1sd ] show that for 1% of @xmath14(co@xmath0 ) , below 100 km projected distance , the contributions in the formation of o(@xmath2s ) are 6575% from photodissociation of h@xmath0o , 15% from co@xmath0 photodissociation , and 20% from other reactions . at 1000 km projected distance the photodissociation of h@xmath0o is contributing around 45% and @xmath145% contribution mainly from dissociative recombination of h@xmath0o@xmath19 . in this case
the calculated g / r ratio varies between 0.05 and 0.4 for less than 1000 km projected distance .
when we increased @xmath14(co@xmath0 ) to 5% the contribution from both photodissociation of h@xmath0o and co@xmath0 is similar ( 30 to 45% ) for the projected distances less than 1000 km . in this case g / r ratio
is found to vary between 0.07 to 0.7 . in the absence of co@xmath0
, the photodissociation of water and dissociative recombination of h@xmath0o@xmath19 mainly controls the formation of o(@xmath2s ) in the cometary coma . in this case by changing the co alone between 1 and 5% , it is found that the change in the calculated g / r ratio profile is insignificant . assuming the absence of co in the coma , the calculated contributions are not changed from the previous cases whereas the calculated g
/ r ratio profile is increasing linearly by increasing @xmath14(co@xmath0 ) .
we have calculated the [ oi ] 1356 emission intensity by varying @xmath14(co@xmath0 ) and @xmath14(co ) between 1% and 5% .
the calculations show that electron impact on atomic oxygen is important ( 50% ) excitation process for [ oi ] 1356 emission ( see figure [ prod - oemis ] ) .
electron impact on co@xmath0 is next important emission source of [ oi ] 1356 line ( 30% to the total ) .
below 50 km , the formation of atomic oxygen is through collisional quenching of o(@xmath2d ) with water ( 35% ) , charge exchange between o@xmath19 and oh@xmath19 with water ( 45% ) . above this radial distance 75% atomic oxygen
is produced due to radiative decay of o(@xmath2d ) .
the role of co@xmath0 and co in producing atomic oxygen is less than 5% . hence , by increasing the @xmath14(co@xmath0 ) in the coma it is found that 30% of this emission line intensity is increased only below 50 km radial distance .
the role of co in producing this emission line is insignificant .
the @xmath14(co ) can significantly influence the [ ci ] 1931 emission intensity compared to that of co@xmath0 . by increasing the @xmath14(co ) from 1 to 5%
it is found that the intensity of this emission line also increases .
the [ ci ] 1931 is mainly through photodissociation of co ( 75% ) and co@xmath0 ( 20% ) .
the role of resonant scattering of c(@xmath2d ) is significant ( 50% ) for radial distances larger than 500 km .
similar effects also have been observed on 9850 and 9823 emission lines .
the maximum gas production in this comet at perihelion for high activity case is about 1 @xmath4 10@xmath13 s@xmath6 .
we have also done calculations for this comet by considering low activity case with water production rate of 5 @xmath4 10@xmath25 s@xmath6 and keeping the @xmath14(co@xmath0 ) and @xmath14(co ) equal ( 5% ) . by decreasing the gas production rate by a factor of 2
it is found that the calculated emission intensities are decreased by 30% . similarly the collisional quenching radius for o(@xmath2s ) ,
o(@xmath2d ) , and c(@xmath2d ) also decreased by 30% and is moved towards the nucleus .
alice ultraviolet spectrometer on - board rosetta mission is designed to observe many emission lines from 67p in the wavelength region 750 2050 @xcite .
this range overlaps with part of co cameron bands covering 18002600 , [ oi ] 1356 , and [ ci ] 1931 emission lines @xcite . by making limb scan observations alice spectrograph
can be used to derive the spatial distribution of co and/or co@xmath0 around 67p nucleus .
alice can observe the shortward part of co cameron band emission in its longward limit where unfortunately the sensitivity is small @xcite .
similarly [ ci ] 1931 emission line also falls into the longer end of the alice spectral range .
it may however possible to detect it because of strong resonant fluorescence efficiency of c(@xmath2d ) atom .
the rosetta onboard optical , spectroscopic , and infrared remote imaging system ( osiris ) is a scientific camera system with 12 discrete filters which is designed to observe 67p cometary coma over the wavelength rage 2501000 nm @xcite .
osiris can also map the release of certain daughter species such as oh and [ oi ] based on observed emission intensities at 3090 and at 6300 , respectively .
outcome of this study can be compared with on - board rosina mass spectrometers in - situ measurements of the neutral composition in the coma @xcite .
space - based observations from earth , such as from the hubble space telescope ( hst ) , can also observe these ultraviolet emissions during rosetta mission observation period .
several ground - based observatories have been observing comet 67p ( http://www.rosetta-campaign.net/planned-observations ) in visible and infrared regions to study the spatial distribution of various species . in this context
the present modelling work can provide better understanding of different processes governing co cameron band , atomic oxygen and atomic carbon emission lines in comet 67p to derive parent neutral composition in the coma . by making several observations on different comets @xcite
demonstrated that there is a strong correlation between 1931 emission line intensity and co column density .
the radiative decay of c(@xmath2d ) to ground state yields 9823 and 9850 emission lines . by observing these emission lines on hale - bopp ,
@xcite concluded that they can be used as direct tracers of co photodissociation in the cometary coma .
the model calculations also show that the major production source of c(@xmath2d ) in the inner coma is mainly due to photodissociation of co and the contribution from other production processes is smaller by an order of magnitude compared to the former ( see fig [ prod - c1d ] ) .
the model calculated 1931 , 9850 , and 9823 emission line profiles are flat up to 1000 km projected distances due to strong collisional quenching of c(@xmath2d ) with h@xmath0o . since these atomic carbon emission lines are mainly controlled by photodissociation of co , observed intensity profiles can be used to derive the co gas production rate in the coma .
the calculated emission intensity profile can be useful as a baseline prediction to constrain the co mixing ratios in the coma for the alice observation of carbon emission lines which can be then compared with the rosina observations for the same regions under similar solar illumination .
measuring atomic oxygen visible emission line intensities is an important diagnostic tool in estimating the water production rate as well as to understand the spatial distribution of h@xmath0o in the cometary coma @xcite .
@xcite analysed several eso s vlt observed high resolution green and red - doublet emission line spectra on various comets .
co@xmath0 mixing ratios are derived in these comets by comparing the eso vlt observations with modelled g / r ratio profiles @xcite .
the model calculated g / r ratio profile is presented on 67p in figure [ inten - emis ] . by modelling green and red - double emission intensity profiles on various comets at different heliocentric distances , @xcite have shown that g / r ratio value increases linearly by increasing @xmath14(co@xmath0 ) in the coma whereas the affect of co is minor in determining either green or red - doublet emission intensities . hence , the observed g / r ratio profile on 67p can be used to constrain co@xmath0 mixing ratio in the coma .
when a comet is far away from the sun ( 3 au ) , it is expected to have higher co and co@xmath0 volume mixing ratios which are species associated with low sublimation temperatures ( e.g. , * ? ? ?
the calculations made at 3 au heliocentric distance ( see fig .
[ inten - emis3au ] ) , with mixing ratios 5% co@xmath0 and 15% co , show that atomic oxygen red - doublet emission is the most intense emission in the inner coma .
this emission can be observed by rosetta onboard osiris instrument which can be subsequently used to derive water production rate .
unfortunately there are no filters on osiris to measure [ oi ] 5577 , [ ci ] 9823 , and 9850 emission lines @xcite .
the predicted oxygen red - doublet intensity along the projected distance ( figures [ inten - emis ] and [ inten - emis3au ] ) could be useful in analysing osiris visible spectra of 67p and subsequently deriving h@xmath0o distribution around the nucleus .
the photon cross section for the formation of o(@xmath2s ) from h@xmath0o has never been reported in the literature @xcite . in this model
the formation of o(@xmath2s ) from photodissociation of h@xmath0o has been accounted by assuming 0.5% yield for h@xmath0o at ly-@xmath26 wavelength @xcite .
the onboard rosina spectrometers can measure the co@xmath0 number densities during this mission period at different radial distances in the coma . by combining the observed g / r ratio profile with onboard rosina co@xmath0 measurements it would be possible to constraint the o(@xmath2s ) average yield value at solar lyman-@xmath26 .
the high - resolution spectroscopic observations , such as analysis of @xcite , can provide information about collisional quenching of o(@xmath2s ) and o(@xmath2d ) metastable states in the coma of 67p .
the observation of both green and red - doublet emission line widths and g / r ratio profiles along with rosina measurements can solve the puzzle that green line is wider than either of red - doublet emission lines in comets .
the modelling of production rates of co(a@xmath3 ) has shown that suprathermal electron impact reactions mainly govern the co cameron band emission with contribution of around 75% whereas co@xmath0 photodissociation contributes about 25% ( see section [ results_cam ] ) . in the absence of co ,
the electron impact on co@xmath0 is an equally important production source of co(a@xmath3 ) as photodissociation of co@xmath0 ( see table [ tab : per - cam ] ) .
this suggest that electron impact excitation mechanism should be considered for the estimation of parent species production rates in the coma . with sufficient co ( @xmath273% ) in the coma , the contribution from electron impact reactions in producing this band emission close to the nucleus ( @xmath22100 km ) is about 80% .
the excited state co(a@xmath3 ) is mainly populated in the coma by suprathermal electrons in the energy range between 10 and 15 ev ( see figure [ csc - eflux ] ) . since the major source for the production of co(a@xmath3 ) is electron impact , the observed co cameron band emission close to the nucleus would be suitable to track the suprathermal electron intensity @xcite rather than co@xmath0 neutral density .
the [ oi ] 1356 emission line is an excellent tracer for electron impact processes in the coma .
modelling of electron impact excitation processes shows that this emission is mainly due to electron impact excitation of atomic oxygen followed by electron impact dissociative emission of co@xmath0 and h@xmath0o ( see figure [ prod - oemis ] ) .
however , the intensity of this emission is weaker by three orders of magnitude compared to co cameron band emission .
the electron impact excitation cross section for atomic oxygen producing [ oi ] 1356 emission line peaks at 15 ev whereas for co@xmath0 and h@xmath0o it is between 30 and 60 ev .
the contribution from atomic oxygen is about 50% and the rest is through co@xmath0 ( 30% ) and h@xmath0o ( 20% ) .
hence , half of the observed emission intensity profile is linked to the suprathermal electron intensity at 15 ev .
recently alice observed several hi , oi , and ci emission near the cometary nucleus when comet was at around 3 au @xcite .
the observation of oi 1356 emission line intensity , which is varying between @xmath11.5 and @xmath13 rayleigh at 10 km projected distance , @xcite , is close to our predicted calculation ( @xmath11.5 rayleigh , see fig .
[ inten - emis3au ] ) . detailed analysis of this emission line will be presented in the future work using the rosetta measured neutral density distribution around the nucleus . the onboard rosetta plasma consortium ( rpc)/ ion and electron sensor ( ies ) is capable of measuring the electron energy spectra in the energy rage 1 ev / e to 22 kev / e @xcite . since
both co cameron band and [ oi ] 1356 emission lines are governed mainly by electron impact excitation reactions , the observed emission intensities may be supportive for the ies measured suprathermal electron intensity at around 15 ev .
molecular oxygen is the major source for the production of o(@xmath2s ) and o(@xmath2d ) in the terrestrial atmosphere .
the recent discovery of o@xmath0 in 67p s coma by rosina / dfms @xcite demands the inclusion of o@xmath0 in the model in order to calculate green and red - doublet emission intensities . by including 4% molecular oxygen with respect to water production rate , and for the input conditions described in section [ model - inputs ] ,
the g / r ratio is found to be increase by around 20% close to the nucleus ( @xmath2220 km projected distance ) .
@xcite observed that the relative abundance of molecular oxygen ranges 1 to 10% with respect to h@xmath0o production rate .
hence in order to determine co@xmath0 abundance based on the g / r ratio the contribution from molecular oxygen should also be considered . in case of higher o@xmath0 abundance in comets ,
the observed g / r ratio can be significantly controlled by photodissociation of o@xmath0 and may lead to underestimation of co@xmath0 mixing ratio .
the estimated diamagnetic cavity on the sunlit side of this comet at perihelion is around 3040 km @xcite .
the extent of diamagnetic cavity depends on the gas production rate and solar wind conditions during comet perihelion visit . beyond this cavity ,
most of the ions are transported towards the tail side due to solar wind interaction .
the assessment of solar wind interaction on the emission intensities is beyond the scope of this work but we would like to discuss the possible sources that can alter the emission intensities . outside the diamagnetic cavity ,
the chemical lifetime of neutrals can be significantly altered by charge exchange between solar wind ions and cometary species .
hence , it is expected that the calculated intensities outside the diamagnetic cavity can be changed based on the solar wind conditions during that time .
the electrons outside the diamagnetic cavity are primarily solar wind electrons or shocked solar wind electrons @xcite .
the population of suprathermal electrons outside the diamagnetic cavity is a complex problem due to admixture of solar wind electrons .
however , the radius of collisional zone and diamagnetic cavity are subjected to the gas production rate and solar wind conditions during the comet perihelion passage . for electron impact
driven emissions , such as co cameron band and [ oi ] 1356 , due to strong solar wind interaction both neutral density and electron population may change outside the diamagnetic cavity region , thus the observed emission intensities vary significantly . in this region the solar wind electrons may also contribute to the total emission intensity @xcite .
however , the dissociative recombination co - bearing ions to the total emission intensity contribute little ( @xmath225% ) , whereas formation of atomic oxygen significantly ( 50% ) , because of charge exchange between o@xmath19 and oh@xmath19 with h@xmath0o .
we do not expect the radiative decay and collisional quenching of o(@xmath2d ) can be altered significantly due to solar wind interaction .
the evolution of cometary ionosphere around 67p nucleus has been monitored by rosetta plasma consortium and rosetta orbiter spectrometer for ion and neutral analysis ( rosina ) instruments .
the recent observations of rosina / double focussing mass spectrometer ( dfms ) , rpc / ion and electron sensor ( ies ) , and rpc / ion composition analyzer ( ica ) , when the comet was beyond 2 au , have shown that due to low outgassing rate no contact surface is formed and most of the solar wind has directly accessed the 67p s nucleus , though the plasma close to the comet is dominated by cometary water ions @xcite .
@xcite found that the suprathermal electrons are accelerated to several hundreds of ev .
the high energetic solar wind charged particles may also be involved in producing the excited atomic and molecular states discussed here .
the formation o(@xmath2s ) , o(@xmath2d ) , and c(@xmath2d ) are mainly due to photochemical reactions .
the contribution from ions and thermal electron recombination reactions for the inner coma is very small ( @xmath225% ) .
hence , we do not expect the predicted oxygen visible and [ ci ] 1931 line emission intensities to change due to solar wind interaction for the inner coma unless the radial distribution of h@xmath0o is changed . in this model
we have accounted for main parent oxygen- and carbon - bearing species to compute the emission intensities .
however , the contribution from photodissociation and electron impact of other minor species is also possible . in case of atomic oxygen visible emissions ,
the dissociation of other oxygen - bearing species , such as hcooh or h@xmath0co , are unlikely to be the parent because they can not decay fast enough to produce o(@xmath2d ) and o(@xmath2s ) @xcite . however , in the case of carbon emissions there could be an involvement from other carbon - bearing species , such as hydrocarbons . since the major processes governing these emissions are via photochemical reactions , the role of electron temperature for the inner coma is not significant ( @xmath225% ) .
the model calculations are done for gas phase so scattering of solar photons by dust grains could be a significant factor in governing these emission intensities . the recent rosina / dfms observations on 67p show that the cometary coma contains a variety of species with heterogeneous distribution which varies with time and latitude ( e.g. , * ? ? ?
* ; * ? ? ?
the microwave instrument on the rosetta orbiter ( miro ) mapped around 67p s nucleus when it was at 3.4 au @xcite .
the water column density in the inner coma ( within 3 km from the nucleus ) is found to vary even by two orders of magnitude .
@xcite further investigated the heterogeneity of 67p s coma by measuring various major ( h@xmath0o , co@xmath0 , and co ) and minor ( hcn , ch@xmath24oh , ch@xmath28 , and c@xmath0h@xmath29 ) volatile species using rosina / dfms . our calculated emission intensities may change significantly due to variable neutral densities around comet 67p .
future work will include the use of in - situ measured neutral densities from rosina sensors to drive model calculated emission intensities .
results could then be compared with alice and osiris observations as well as ground - based observations .
rosetta - remote and earth - based spectroscopic observations , combined with modelling of comet 67p , offer a unique opportunity to assess the main production and destruction processes governing various forbidden visible and ultraviolet emissions as the comet gets closer to the sun .
the combined analysis applied to rosetta remote and in - situ observations could be used as a ground truth for the interpretation of earth - based observations on 67p cometary coma .
the model calculations suggest that the electron impact reactions are the dominant sources in producing co cameron band emission .
hence , the observed co cameron emission intensity close to the cometary nucleus can be used to track the suprathermal electron intensity in the energy range 10 to 15 ev close to the nucleus .
the observed g / r ratio away from the collisional zone can be used to confirm the parent oxygen species producing these emissions .
measurement of the g / r ratio close to the comet as a function of projected distance can be used to constrain the @xmath14(co@xmath0 ) .
presence of high mixing ratio of molecular oxygen can affect the g / r ratio significantly which may lead to underestimation of @xmath14(co@xmath0 ) .
the observation of [ oi ] 1356 can give a clear indication of the role of electron impact processes in the coma , while [ ci ] 1931 emission is a good tracer to probe co distribution near the nucleus . both cameron band and atomic oxygen emission observations are useful to assess h@xmath0o , co@xmath0 , and co volume mixing ratios in the coma and to understand the spatial distribution and their time evolution in comet 67p .
the quantitative assessment of different excitation processes is essential to study the evolution of the chemistry in the inner cometary coma with the increasing neutral gas production rate .
s.r . and m.g .
acknowledge the support of the science and technology facilities council ( stfc ) of uk through the consolidated grant st / k001051/1 to imperial college london .
the work of a.b . is supported by indian space research organization .
feldman , p. d. , cochran , a. l. , & combi , m. r. 2004 , spectroscopic investigations of fragment species in the coma : comets ii ( m. c. festou , h. a. weaver , & h. u. keller ( ed.)(tucson : univ . of arizona ) ) , 425447 ) rate profiles in comet 67p / churyumov - gerasimenko for an h@xmath0o out gassing rate of 10@xmath13 s@xmath6 and for 5% co@xmath0 and 5% co volumetric mixing ratios relative to water at 1.29 au .
the photodissociation of co@xmath0 ( gray curve ) and suprathermal electron impact on co@xmath0 ( solid dashed curve ) are producing co(a@xmath3 ) with nearly equal rates and both curves overlap .
h@xmath30 , e@xmath31 , and e@xmath32 stand for photon , thermal electron , and suprathermal electron , respectively .
[ prd - cam ] ] ) from co and co@xmath0 , and for [ oi ] 1356 from atomic oxygen , co@xmath0 , and h@xmath0o .
calculated suprathermal electron intensity at cometocentric distance of 10 km is also shown with magnitude on right side y - axis .
the cross section for the formation of [ oi ] 1356 emission line from electron impact dissociation of h@xmath0o is estimated by taking the ratio of cross sections for oi 1304 to [ oi ] 1356 at 100 ev from @xcite .
e@xmath32 stand for suprathermal electron .
[ csc - eflux ] ] s@xmath6 for 5% co@xmath0 and 5% co volume mixing ratios relative to water at 1.29 au .
the calculated atomic carbon production rate profiles , which are represented with symbols and curves , are divided by a factor 5 .
the production rate profile of o(@xmath7p ) through charge exchange between o@xmath19 and h@xmath0o is multiplied by a factor 1.5 .
stand for photon .
[ prod - ocg ] ] s ) rate profiles in comet 67p / churyumov - gerasimenko for an h@xmath0o out gassing rate of 10@xmath13 s@xmath6 and for 5% co@xmath0 and 5% co volumetric mixing ratios relative to water at 1.29 au .
h@xmath30 , e@xmath32 , and e@xmath32 stand for photon , thermal electron , and suprathermal electron , respectively .
[ prd - o1s ] ] d ) rate profiles in comet 67p / churyumov - gerasimenko with the water production rate of 10@xmath13 s@xmath6 for 5% co@xmath0 and 5% co volume mixing ratios relative to water at 1.29 au .
h@xmath30 and e@xmath32 stand for photon and thermal electron , respectively .
[ prd - o1d ] ] s ) and o(@xmath2d ) loss rate profiles in comet 67p / churyumov - gerasimenko with the water production rate of 10@xmath13 s@xmath6 for 5% co@xmath0 and 5% co volume mixing ratios relative to water at 1.29 au .
stand for photon.[los - o1ds ] ] 1356 emission rate profiles in comet 67p / churyumov - gerasimenko with the water production rate of 10@xmath13 s@xmath6 for 5% co@xmath0 and 5% co volume mixing ratios relative to water at 1.29 au .
e@xmath32 stand for suprathermal electron .
[ prod - oemis ] ] d ) production rate profiles in comet 67p / churyumov - gerasimenko with the water outgassing rate of 10@xmath13 s@xmath6 for 5% co@xmath0 and 5% co volume mixing ratios relative to water at 1.29 au .
h@xmath30 , e@xmath32 , and e@xmath32 stand for photon , thermal electron , and suprathermal electron , respectively .
[ prod - c1d ] ] d ) loss rate profiles in comet 67p / churyumov - gerasimenko with the water production rate of 10@xmath13 s@xmath6 for 5% co@xmath0 and 5% co volume mixing ratios relative to water at 1.29 au .
stand for photon;[los - c1d ] ] and 5% co volume mixing ratios relative to water in comet 67p / churyumov - gerasimenko with water production rate 1 @xmath4 10@xmath13 s@xmath6 as a function of projected distance at 1.29 au .
the calculated g / r ratio values are shown on the right y - axis .
the [ oi ] 1356 emission line profile is multiplied by a factor 10 .
one rayleigh = @xmath33 photons cm@xmath16 s@xmath6 sr@xmath6 .
[ inten - emis ] ] and 15% co volume mixing ratios relative to water in comet 67p / churyumov - gerasimenko with water production rate 5 @xmath4 10@xmath17 s@xmath6 as a function of projected distance at 3 au .
the calculated g / r ratio values are shown on the right y - axis .
one rayleigh = @xmath33 photons cm@xmath16 s@xmath6 sr@xmath6 .
[ inten - emis3au ] ] | remote observation of spectroscopic emissions is a potential tool for the identification and quantification of various species in comets .
co cameron band ( to trace co@xmath0 ) and atomic oxygen emissions ( to trace h@xmath0o and/or co@xmath0 , co ) have been used to probe neutral composition in the cometary coma . using a coupled - chemistry emission model , various excitation processes controlling co cameron band and different atomic oxygen and atomic carbon emissions
have been modelled in comet 67p / churyumov - gerasimenko at 1.29 au ( perihelion ) and at 3 au heliocentric distances , which is being explored by esa s rosetta mission .
the intensities of co cameron band , atomic oxygen and atomic carbon emission lines as a function of projected distance are calculated for different co and co@xmath0 volume mixing ratios relative to water .
contributions of different excitation processes controlling these emissions are quantified .
we assess how co@xmath0 and/or co volume mixing ratios with respect to h@xmath0o can be derived based on the observed intensities of co cameron band , atomic oxygen , and atomic carbon emission lines .
the results presented in this work serve as base line calculations to understand the behaviour of low out - gassing cometary coma and compare them with the higher gas production rate cases ( e.g. comet halley ) .
quantitative analysis of different excitation processes governing the spectroscopic emissions is essential to study the chemistry of inner coma and to derive neutral gas composition . |
the longstanding problem of the generation of turbulence in various hydrodynamically stable situations has found a solution in recent years with the mhd shear flow instability , also called magnetorotational instability ( mri ) , in which the presence of a magnetic field has a destabilizing effect on a differentially rotating flow with the angular velocity decreasing outwards .
the mri has been formulated decades ago @xcite for ideal taylor - couette flow , but its importance as the source of turbulence in accretion discs with differential ( keplerian ) rotation was first recognized by balbus and hawley , @xcite .
however , the mri has never been observed in the laboratory @xcite .
after goodman and ji @xcite was the absence of mri due to the small magnetic prandtl number approximation used in @xcite .
the magnetic prandtl number pm is really very small under laboratory conditions ( @xmath10 and smaller , see table [ tab0 ] ) .
let us now concentrate to the small magnetic prandtl number for liquid sodium , i.e. pm @xmath11 .
we start with the results for containers with insulating walls and outer cylinders at rest , ( fig .
there are then linear instabilities even without magnetic fields
. for ha@xmath12 solutions for @xmath13 ( re@xmath14 ) and @xmath15 ( re@xmath16 ) are known , see @xcite .
the axisymmetric mode possesses the lowest eigenvalue .
this is also true for nonvanishing magnetic field ; we do not find any crossover of the instability lines for axisymmetric and nonaxisymmetric modes .
the same is true for containers with rotating outer cylinder ( fig . [
for growing @xmath17 the reynolds number for the hydrodynamic solution moves upwards , reaching infinity for @xmath18 ( here ) .
the mri is represented by characteristic minima , in our case for @xmath19 at hartmann numbers of order @xmath20 and reynolds numbers of order @xmath9 .
the exact coordinates of the minima are given in table [ tab1 ] . in order to characterize the hartmann numbers note that for liquid sodium @xmath21 hence , for @xmath22 the magnetic field and the hartmann number have the same numerical values . with @xmath23 @xmath24/s and @xmath25=0.5
it follows from ( [ re ] ) and ( [ ha ] ) f_in= 64re/10 ^ 6(r_out/10cm)^2 hz [ fff ] for the frequency of the inner cylinder .
hence , a container with an outer radius of 22 cm ( see above ) and an inner radius of 11 cm filled with liquid sodium and embedded in vacuum requires a _ rotation of about 19 hz _ in order to find the mri . following ( [ b2.22 ] ) the required magnetic field is about 1400 gauss . the results for containers with _
conducting walls _ are given in fig .
note that the minimal reynolds numbers given in fig .
[ f2]b are higher than for insulating cylinder walls .
the influence of the boundary conditions is not as small as expected .
the main difference between the two sorts of boundary conditions , however , is the existence of crossovers of the instability lines for @xmath13 and @xmath15 in case of conducting walls . for both resting and rotating outer cylinder hartmann numbers exist above which the nonaxisymmetric mode possesses a lower reynolds number than the axisymmetric mode .
we have already shown the existence of such crossovers for conducting walls for @xmath26 in @xcite .
it is now clear that the occurrence of nonaxisymmetric solutions as the preferred modes is a rather general phenomenon for containers with conducting walls which can become important for the design of future dynamo experiments ( cowling theorem ) .
there are not only stationary patterns of flow and field possible but the instability can also onset in form of oscillating solutions .
this effect is called overstability . in case of rotating convection between two layers heated from below the onset of the instability in form of oscillating solutions
even possesses the lowest eigenvalues for certain prandtl numbers , @xcite .
we find a very similar behavior for the mhd taylor - couette flow between conducting cylinders for resting outer cylinder ( see fig .
[ over ] ) .
it is a pair of waves traveling in positive and negative @xmath27-direction .
note that the cylinder considered here has no bound in vertical direction .
if the cylinder is finite , however , the possibility exists that traveling waves might be combined to standing waves .
e. p. velikhov , sov .
jetp * 9 * , 995 ( 1959 ) .
s. chandrasekhar , _ _ h__ydrodynamic and hydromagnetic stability ( clarendon , oxford , 1961 ) .
s. a. balbus and j. f. hawley , astrophys .
j. * * 3**76 , 214 ( 1991 ) .
r. j. donnelly and m. ozima , phys . rev .
lett . * 4 * , 497 ( 1960 ) .
r. j. donnelly and m. ozima , proc .
a * * 2**66 , 272 ( 1962 ) .
r. j. donnelly and d. r. caldwell , j. fluid .
mech . * * 1**9 , 257 ( 1964 ) .
a. brahme , physica scripta * 2 * , 108 ( 1970 ) .
j. goodman and h. ji , j. fluid mech .
* * 4**62 , 365 ( 2002 ) . k. noguchi , v. i. pariev , s. a. colgate , h. f. beckley and j. nordhaus , astrophys .
j. * * 5**75 , 1151 ( 2002 ) .
d. richard and j .- p .
zahn , astron .
astrophys . * * 3**47 , 734 ( 1999 ) . f. wendt , ing . arch .
* 4 * , 577 ( 1933 ) .
f. schultz - grunow , zamm * * 3**9 , 101 ( 1959 ) .
d. a. shalybkov , g. rdiger and m. schultz , astron .
astrophys . * * 3**95 , 339 ( 2002 ) . which is unknown in the hydrodynamic regime , counter rotation excluded , see e. r. krger , a. gross and r. c. diprima , j. fluid mech . *
* 2**4 , 521 ( 1966 ) .
d. elstner , r. meinel and g. rdiger , geophys .
fluid dyn . *
* 5**0 , 85 ( 1990 ) .
g. rdiger , differential rotation and stellar convection : sun and solar - type stars ( gordon & breach science publishers , new york , 1989 ) .
a. p. willis and c. f. barenghi , astron . astrophys . * * 3**88 , 688 ( 2002 ) .
g. rdiger and d. a. shalybkov , phys .
e * * 6**6 , 016307 ( 2002 ) .
.- k . chen and m. chang , j. fluid mech . * * 3**66 , 135 ( 1998 ) . | the linear stability of mhd taylor - couette flow of infinite vertical extension is considered for liquid sodium with its small magnetic prandtl number pm of order of @xmath0 .
the calculations are performed for a container with @xmath1 with an axial uniform magnetic field excluding counter rotating cylinders .
the sign of the constant @xmath2 in the basic rotation law @xmath3 strongly influences the presented results .
it is negative for resting outer cylinder .
the main point here is that the subcritical excitation which occurs for large pm disappears for small pm ( cf .
fig .
[ minima ] ) .
this is the reason that the existence of the magnetorotational instability remained unknown over decades . for rotating outer cylinder
the limiting case @xmath4 ( i.e. @xmath5 ) plays an exceptional role .
the hydrodynamic instability starts to disappear while the hydromagnetic instability exists with minimal reynolds numbers at certain hartmann numbers of the magnetic field .
these reynolds numbers exactly scale with pm@xmath6 resulting in moderate values of order @xmath7 for pm=10@xmath8 .
however , already for the smallest positive value of @xmath2 the reynolds numbers start to scale as 1/pm leading to much higher values of order @xmath9 for pm=10@xmath8 .
hence , for outer cylinders rotating faster than the limit @xmath4 it is exclusively the _ magnetic _ reynolds number rm which directs the excitation of the instability .
they are resulting as lower for insulating walls ( ` vacuum ' ) than for conducting walls .
generally , the magnetic reynolds numbers for liquid sodium have to exceed values of order 10 leading to frequencies of about 20 hz for the rotation of the inner cylinder if containers with ( say ) 10 cm radius are considered .
the required magnetic fields are about 1000 gauss . also nonaxisymmetric modes have been considered . with vacuum boundary conditions
their excitation is always more difficult than the excitation of axisymmetric modes ; we never observed a crossover of the lines of marginal stability . for conducting walls , however ,
such crossovers exist for both resting and rotating outer cylinders , and this might be essential for future dynamo experiments . in this case , however , the instability also can onset in form of _ oscillating _ axisymmetric patterns of flow and field and the reynolds numbers of these solutions are lower than the reynolds numbers for the stationary solutions . |
quantum effects are expected to drastically affect the structure of space and time at the smallest of scales . however , our current theories of gravity and quantum mechanics are ( fortunately ) very stingy with the options they leave us as far as the small scale of structure of spacetime is concerned .
for example , attempts to model such a structure by violating or deforming lorentz invariance ( li ) are either very strongly constrained by experiments , or run into deeper conceptual issues when one goes beyond the simple one - particle models .
it is much more plausible that instead of li , it is the assumption of locality that might have to be given up at small scales @xcite .
however , abandoning locality then also necessitates that we give up the classical description of spacetime in terms of local tensorial objects , in particular the metric tensor @xmath1 .
finding the right geometric variables that can describe spacetime geometry down to smallest scales is of utmost significant not only for quantum gravity , but also for the proper physical interpretation of the field equations of gravity at the classical level .
in particular , the deep connection between einstein equations , thermodynamics , and information theory that has been studied in depth for over a decade very strongly suggests that we question the conventional description of gravitational dynamics based on einstein - hilbert ( eh ) action . in any case , to properly understand the implications of results that have been accumulated from the study of quantum fields in curved spacetime , it is extremely important that one must first identify the correct geometric variables to describe spacetime geometry at the classical level itself .
fortunately , the hint for doing so also comes from these very same results .
in particular , one of the most significant results to have come out of semiclassical studies is the existence of a minimal spacetime length , say @xmath9 , below which spacetime intervals loose any operational significance @xcite .
such a _ zero - point length _ appears in various forms in several candidate models of quantum gravity , and is often considered as the universal regulator for divergences in quantum field theory ( qft ) and general relativity ( gr ) . in a recent work ,
one of us @xcite proposed that a more appropriate description of spacetime geometry in presence of a minimal length scale must be based on non - local bi - tensors instead of the metric tensor .
the geodesic distance between two spacetime events , in particular , was proposed as a more fundamental object than the metric tensor .
the relevant bi - tensor in this context is the so called synge world function @xmath2 defined by @xcite @xmath23(x(\lambda ) ) \dm \lambda \nn \\ &
= & \frac{1}{2 } \sigma(p , p_0)^2\end{aligned}\ ] ] where @xmath24 is the square of geodesic interval , with the corresponding geodesic distance given by @xmath25 here , @xmath26 is tangent to the geodesics , and @xmath27 .
( in this paper , we shall use @xmath28 and @xmath29 interchangeably to keep the expressions and notation convenient .
we will also often use @xmath30 in covariant taylor series to keep track of terms of various orders without messing up the notation . ) assuming that the small scale structure of spacetime is characterised ( at least at a semiclassical level ) by the existence of a minimal length , it was shown that one can construct a second rank bi - tensor , @xmath31 , such that it yields geodesic distances with a lower bound @xmath9 .
the construction of @xmath15 in @xcite was based on two inputs : * p1 : * the requirement that geodesic distances have a lorentz invariant lower bound _ and _ this arises from modification of geodesic distances as @xmath32 .
* p2 : * the requirement that the modified dalembartian @xmath14 yields the following modification for the two point functions @xmath33 of fields in _ flat spacetime _ : @xmath34 \rightarrow \widetilde g\l[\sigma^2\r ] = g\l [ \sigma^2 + \lp^2 \r]$ ] .
( this essentially regulates the uv divergences in qft , and is based on several earlier works on the subject . ) these two requirements then completely fix the form of @xmath15 .
being manifestly covariant , the extension to arbitrary curved spacetimes suggests itself naturally . in subsequent work @xcite
, it was shown that the ricci bi - scalar @xmath16 corresponding to this so called _ qmetric _
@xmath15 has a very specific non - analytic structure which results in a non - trivial result for the coincidence limit @xmath35 $ ] when @xmath36 .
the specific result proved there was @xmath37 with @xmath26 being arbitrary normalised vectors at each spacetime event .
a similar analysis for the surface term @xmath38 of the eh action was also presented subsequently , and the very same term as above was shown to appear there as well ( see second reference in @xcite ) .
the above results have many deep implications , in particular for understanding better the notion of entropy associated with each spacetime event @xmath6 and it s causal boundaries , and for the _ emergent gravity paradigm _ ; these were discussed in detail in @xcite .
they not only provide a very strong hint towards the importance of the quantity @xmath39 in description of gravitational dynamics , but also give a precise quantitative manner in which the transmutation of gravitational lagrangian @xmath40 can arise as a relic of a minimal length . on the other hand
, the analysis presented in @xcite does not give much insight on the robustness of the conclusions drawn from the final result , and much less insight on some of the miraculous cancellations responsible for it .
in particular , the following issues concerning the main inputs * p1 * and * p2 * were left unclear : \1 . the analysis _ assumed _ ( see * p1 * ) that a lower bound on geodesic distances is realized via the modification @xmath32 .
while the motivation for such a modification comes from several older results , it remained unclear as to how much the final result depends on it .
this question is of fundamental significance , since the precise manner in which a minimal length is introduced in spacetime can come only from a complete framework of quantum gravity . in absence of such a framework ,
it is important not to make any assumptions on how distances can get modified . in particular
, the modifications introduced by quantum gravity can be non - perturbative , and hence need not possess a series expansion in @xmath9 near @xmath41 . in this paper
, we shall establish our result without making any such assumption .
technically , we shall keep the function @xmath18 , which represents modification of distances , completely arbitrary and satisfying only @xmath21(0 ) < \infty$ ] in addition to it s defining properties ( which will be given below ) .
in particular , the function @xmath42 need not admit a perturbative expansion in @xmath9 , unlike the form @xmath43 which was used in @xcite .
the construction of @xmath15 for arbitrary @xmath22 was already sketched in appendix a of @xcite , except for a crucial difference , which brings us to our second point concerning the input * p2*. \2 .
the requirement * p2 * that two point functions get modified as @xmath34 \rightarrow \widetilde g\l[\sigma^2\r ]
= g\l [ \sigma^2 + \lp^2 \r]$ ] makes sense only when @xmath33 depends on @xmath44 and @xmath6 only through @xmath45 @xmath46 .
this can not happen in arbitrary curved spacetimes , which very much reduces the possibilities available to fix the qmetric .
this is , of course , good , since it reduces the room available for _ adhoc _ choices .
the most general space(time)s in which @xmath33 is only a function of @xmath45 @xmath46 are the maximally symmetric spaces , of which flat space(time ) is but the simplest possibility with zero curvature .
fixing the qmetric based on flat spacetime , although it captures the correct leading singularity of the two point functions in the coincidence limit , wipes away all information about curvature .
more precisely , the leading singular structure of two point functions associated with the dalembartian @xmath47 in arbitrary spacetime is given by the hadamard form @xmath48 as is evident , the information about curvature , at the leading order , therefore appears in the two point functions through the so called van vleck determinant ( vvd ) @xmath3 . in flat spacetime , @xmath49 exactly , whereas in arbitrary curved spacetimes , @xmath50 one might therefore think that the dependence of qmetric on @xmath3 can not possibly affect the coincidence limit of the ricci biscalar @xmath16 ( or any other curvature invariant ) associated with @xmath15 .
this expectation is , however , wrong .
curvature involves second derivatives of the metric , and the coincidence limit of the second ( or higher ) derivatives of @xmath3 is not zero in general .
( the exact form of the coefficients in covariant taylor expansion of vvd are well known , and are quoted later in this paper . )
this makes it crucial to identify the dependence of qmetric on vvd .
as we shall show , doing so leads to some remarkable results , all following from certain identities satisfied by the vvd .
the modified @xmath16 derived in @xcite is in fact singular in the coincidence limit @xmath51 .
this divergence is cubic in @xmath26 s , and can be regularised using known methods in point splitting regularisation , as was suggested in @xcite .
however , this still leaves a certain amount of discomfort at the mathematical level .
it does not make much sense to appeal to point splitting regularisation since our starting point , based on existence of a minimal length , does not invoke point splitting at any level , but is instead based on use of a non local second rank bi - tensor .
it is therefore important to have a deeper look at this divergence and it s origin , particularly so because it depends on @xmath52 and hence vanishes for all maximally symmetric spaces !
the argument given in 2 above advocating the use of maximally symmetric spaces therefore is not expected to help here , since the divergence is anyway zero for these spaces .
one requires a much more mathematically rigorous analysis to probe the structure of this divergent term .
as we shall show , identifying the correct dependence of @xmath15 on the vvd in fact cancels this divergence in a rather surprising manner .
in fact , the reason for this cancellation is buried deep within the expansion of extrinsic curvature of _ equi - geodesic surfaces _ @xmath7 ( see below ) in an arbitrary spacetime to @xmath53th order in covariant taylor series , and a close relationship between vvd and extrinsic curvature of @xmath7 .
we address all the above issues in this paper , and while doing so , reveal the mathematical robustness of the final result , and hence it s inevitability ( given the two basic inputs ) in any theory which admits a li short distance cut - off .
our key inputs would be much less restrictive and/or specialized than the ones used in @xcite , which makes the results presented here significantly stronger .
these can be stated as : * q1 : * the requirement that geodesic distances have a lorentz invariant lower bound . *
q2 : * the requirement that the modified dalembartian @xmath14 yields the following modification for the two point functions @xmath33 of fields in _ all maximally symmetric spacetimes _ : @xmath34 \rightarrow \widetilde g\l[\sigma^2\r ] = g\l [ { \mathcal s}_{\lp}\l [ \sigma^2 \r ] \r]$ ] . as is evident , these inputs are much more minimalistic compared with * p1 , p2 * , and hence can be expected to provide much more general insights into small scale structure of spacetime .
the paper is structured as follows : sec .
[ sec : vvd ] : in this section , we discuss the geodesic structure of arbitrary curved space(time)s , with focus on intrinsic and extrinsic geometry of equi - geodesic surfaces @xmath7 @xcite , which comprise of points @xmath44 which are at some constant geodesic distance @xmath45 from @xmath6 , and connected to @xmath6 by non - null geodesics .
we also discuss the geometric significance of vvd in studying the small scale structure of spacetime , and highlight some elementary identities relating derivatives of vvd to the extrinsic curvature of @xmath7 , which are used later in sec .
[ sec : rs ] .
[ sec : qmetric ] : we present the derivation of the 2nd rank bi - tensor , the qmetric @xmath31 , based on the two inputs * q1 , q2 * stated above .
in particular , we identify the dependence of the qmetric on the vvd using our condition * q2*. sec . [ sec : rs ] : the ricci bi - scalar @xmath16 for the qmetric is obtained in a closed form based on certain tools developed in @xcite , and it s coincidence limit @xmath51 is evaluated to obtain a local scalar @xmath35(p_0)$ ] at @xmath6 .
it is then shown @xmath35(p_0 ) \neq { \rm \bf ric}(p_0)$ ] , which is one of the key results of this paper
[ sec : ksqrth ] : in this section , we complete our analysis of the eh action by evaluating the gibbons - hawking - york ( ghy ) surface term in the action on equi - geodesic surfaces , for the qmetric . in sec .
[ sec : discussion ] , we finally conclude with a general discussion and implications of the results obtained in this work . the key results of this paper are contained in the boxed eqs .
( [ eq : qmfinal ] ) , ( [ eq : finalrsq ] ) and ( [ eq : grin ] ) .
we work in @xmath54 dimensions , and use the sign convention @xmath55 for lorentzian spaces .
latin alphabets denote spacetime indices .
also , for notational convenience , we use @xmath56 throughout to denote short distance cutoff on geodesic distances ; for timelike / spacelike cases , the replacement @xmath57 must be made in the final results after which @xmath58 .
for convenience , we give below a quick list of some of the most recurring symbols / notation used in the text : + + @xmath59 @xmath60 + @xmath59 @xmath61 + @xmath59 @xmath62 + @xmath63 @xmath64 + @xmath59 @xmath35(p_0)$ ] is the coincidence limit of @xmath16
mathematically , a key role in our analysis would be played by the congruence of geodesics emanating from a fixed spacetime event @xmath6 and the surface comprised of events @xmath44 lying at constant geodesic interval from @xmath6 , which we call as the _ equi - geodesic _ surface of event @xmath6 and denote it by @xmath7 .
the relevant geometrical properties of such surfaces in arbitrary curved spacetimes were discussed in @xcite , and we simply quote the results which we will need here .
we start with the affinely parametrized tangent vector @xmath26 to the geodesic connecting @xmath6 to @xmath44 @xmath65 and note that it is also the normal to @xmath7 .
the extrinsic curvature tensor of @xmath7 , is therefore given by @xmath66 this particular foliation , which characterizes the local geodesic structure of any spacetime , has many interesting properties , and all of these derive from the well known covariant taylor series expansion of the bi - tensor @xmath67 at @xmath44 near @xmath6 @xcite : @xmath68 where @xmath69 .
therefore , we see that the extrinsic geometry of such a equi - geodesic foliation " is very special , and completely characterized by the _ tidal tensor _ @xmath70 . in fact , the intrinsic and extrinsic curvatures can be characterized by systematic taylor expansions around @xmath6 , given by @xcite @xmath71 where @xmath72 , and @xmath73 . for later use
, we also quote here the combination ( easily derived from above ) : @xmath74 for any arbitrary @xmath75 .
as we shall see , the structure of the above expression , which requires keeping upto @xmath53th order terms in @xmath76 and @xmath77 ( that is , terms of @xmath78 ) , hold the key to elimination of coincidence limit divergences in @xmath16 , in conjunction with a couple of differential identities ( involving @xmath76 and @xmath77 ) satisfied by the vvd , which we discuss next .
the van vleck determinant , @xmath3 , is an extremely important object in semi - classical physics .
geometrically , this bi - scalar governs the properties of geodesic congruences emanating from a point , say @xmath6 , as a function of an arbitrary point @xmath44 .
the immense physical importance of this object certainly warrants a longer discussion than presented here , and we refer the reader to @xcite for the same .
in fact , as we shall see , the geometrical significance of vvd holds the key to its relevance for the small scale structure of spacetime , a theme that will resonate constantly throughout this paper .
the vvd is defined as follows : @xmath79 two of the most important differential identities that we shall use , connecting the vvd with the extrinsic curvature of @xmath7 , are the following : @xmath80 where @xmath69 and @xmath81 .
_ proofs of @xmath82 and @xmath83 _ : the above elementary identities follow trivially from the expression @xcite @xmath84 = 2 d \delta\ ] ] noting that the acceleration @xmath85 of @xmath86 is zero since @xmath86 represents tangents to geodesics , we can write the above identity as @xmath87 which is @xmath82 . operating once more with @xmath88 , and using the ( easily proved ) differential geometric identity @xmath89 with @xmath90 , we get @xmath83 . as an aside , let us point out the possible relevance of the vvd in the gravitational action when one focusses on an observer dependent description of gravitational dynamics based on causal structure associated with an arbitrary event ( observer " ) @xmath6 .
the relevance of such a description has gained increased attention since the proposal by jacobson @xcite of using local rindler frames as probes of gravitational dynamics .
we shall focus on the equi - geodesic surfaces @xmath7 straddling the causal boundaries of an arbitrary event @xmath6 , and briefly comment on the null limit in the end .
the complete eh action is given by @xcite @xmath91 where @xmath92 are covariant volume elements in bulk and boundary respectively .
the subtraction term , @xmath93 , is usually taken to be the trace of extrinsic curvature of the boundary surface embedded in flat spacetime .
usually , the boundary @xmath94 is taken at infinity . however , given the fact that the causal structure of spacetime limits the amount of information accessible at an event @xmath6 , it is interesting to ask for the contribution of the boundaries @xmath7 ( which , in the null limit , would make the null cone of @xmath6 ) .
we therefore write @xmath95 ( note the subscript @xmath6 on @xmath96 ; we put it as a reminder that we are now focussing on quantities from the point of view of a specific event ( observer ) @xmath6 . ) the trace of extrinsic curvature of @xmath7 , as embedded in _
spacetime , is @xmath97 .
recalling @xmath82 , this immediately implies @xmath98 using divergence theorem , @xmath99 .
( we must include a similar contribution from @xmath100 ; we do not write this explicitly since we are only interested in the contribution from @xmath7 ) . putting all this together , we get @xmath101 where we have dumped all contributions from @xmath100 in @xmath102 .
let us now comment briefly on the null limit .
since the term @xmath8 is purely geometrical , we expect the null limit of the bulk term above to be straightforward . however , we must point out that the issue of boundary term for null boundaries is not completely unambiguous @xcite , and it might therefore require more care to repeat the above steps for a strictly null surface .
the above analysis strongly suggests that an observer dependent study of gravitational dynamics might require us to change the conventional description based on eh lagrangian . in the rest of this paper
, we will actually present a much stronger result suggesting a very natural transmutation of gravitational lagrangian from @xmath103 to @xmath39 in presence of a lorentz invariant short distance cut - off .
we now have the basic geometric tools using which we can implement * q1 * and * q2 * to arrive at a geometrical description of spacetime at small scales .
our aim in this section would be to construct the so called _ qmetric _ @xmath31 , as described in @xcite , which would reduce to the background spacetime metric @xmath1 for @xmath104 , but which yields geodesic distances bounded from below by @xmath9 , while maintaining lorentz invariance . as was shown in @xcite , the general form of @xmath15 turns out to be ( throughout this paper , @xmath105 ) @xmath106 with corresponding covariant components @xmath15 @xmath107 where @xmath108 , where @xmath109 is the induced metric on @xmath7 , and @xmath110 are functions of events @xmath111 to be fixed by * q1 * and * q2*. the requirement of minimal length , * q1 * , can be imposed @xcite using the hamilton - jacobi equation satisfied by @xmath112 @xcite @xmath113 and requiring @xmath114 we make no assumptions about precisely how quantum gravity would actually affect geodesic intervals , that is , we construct @xmath15 for _ arbitrary _ modification of distances @xmath115
. we will only require : \(i ) @xmath19 ( the condition of minimal length ) .
+ ( ii ) @xmath116 is identity : @xmath117 . + ( iii ) @xmath21(0 ) < \infty$ ] .
the hamilton - jacobi equation , eq .
( [ eq : hjq ] ) , then partially fixes the following combination in the qmetric , as was sketched in appendix a of @xcite : @xmath118 we now use * q2 * to fix the qmetric completely ( which is where we differ significantly with the presentation in @xcite ) . recalling the condition * q2 * explained in detail in the introduction
, we will require that the two point functions of the modified dalembartian @xmath14 satisfy @xmath119 = g\l[\mathcal s_{\lp}(\sigma^2)\r]$ ] in all maximally symmetric spacetimes ( rather than just flat spacetime ) .
we start with the dalembartian operator corresponding to @xmath15 for arbitrary backgrounds @xmath12 ( not necessarily maximally symmetric ) .
after some algebra , we obtain : @xmath120\slashed{\partial}+\slashed{\partial}^2 \biggr\ } + \sqrt{\epsilon \sigma^2}\alpha'\slashed{\partial}\end{aligned}\ ] ] where @xmath121 , @xmath122 .
to impose * q2 * , we will analyse this operator for maximally symmetric spacetimes , in which @xmath123 and @xmath124 are functions of only @xmath45 , and eq .
( [ boxq ] ) becomes @xmath125'\frac{\partial}{\partial\sigma^2}\ ] ] on the other hand , the dalembertian @xmath126 for maximally - symmetric spacetimes is given by @xmath127 where @xmath128 is the exact expression of the van vleck determinant in maximally symmetric spacetimes of positive , zero , and negative curvature respectively ( with radius of curvature @xmath129 ) .
the quantity @xmath130 below is defined as above with @xmath131 .
we are now ready to impose * q2*. as shown in the appendix , the condition that @xmath132 = g\l[\mathcal s_{\lp}(\sigma^2)\r]$ ] is the two point function corresponding to @xmath133 , which translates into @xmath134=0 $ ] when @xmath135 = 0 $ ] ( for @xmath136 ) , gives a differential equation @xmath137 whose solution is @xmath138 where the constant of integration is fixed by the condition @xmath139 when @xmath140 .
we have therefore accomplished our aim of identifying the dependence of @xmath123 , and hence the qmetric , on the vvd by appealing to maximally symmetric spaces and * q2*. equations ( [ eq : fix1 ] ) and ( [ eq : fix2 ] ) fix the final form of the qmetric as @xmath141 with the inverse metric given by @xmath142 for maximally - symmetric spacetimes , it can be shown that the metrics @xmath12 and @xmath15 are related by a non - local , singular diffeomorphism .
this is most easily seen from the line element corresponding to @xmath15 when @xmath12 is maximally symmetric ( assuming @xmath143 and constant positive curvature below for purpose of demonstration ) : @xmath144 the above relationship between @xmath12 and @xmath15 will , of course , not hold in arbitrary curved spacetimes , and therefore the two metrics would have different curvatures .
the form derived in @xcite and @xcite turn out to be special cases of the one derived above if one chooses @xmath145 and @xmath146 .
as we will see , while the choice of @xmath42 is just that , a choice , putting @xmath146 can be potentially dangerous , since one then risks missing important contributions to @xmath16 arising from derivatives of @xmath147 .
in fact , this is just what happens .
having found the qmetric , eq .
( [ eq : qmfinal ] ) , in terms of modification of geodesic distances @xmath42 and the vvd , we can now proceed to evaluate the ricci _
bi_-scalar @xmath16 corresponding to it .
this is the simplest curvature invariant associated with any spacetime , and more importantly for us , the ricci scalar is the simplest lagrangian describing gravitational dynamics in general relativity .
we can then construct a scalar from @xmath16 by taking the coincidence limit @xmath148 , and compare it with @xmath149 , the ricci scalar of the background spacetime @xmath12 .
naively , one might expect that @xmath150(p_0 ) \overset{?}{= } \rsg + { \rm terms~of~order~ } \lp\ ] ] we will explicitly calculate @xmath151(p_0)$ ] to verify this , and show that the leading term is @xmath152 .
we will find an exact expression for the leading term as well as sub - leading terms in terms of the geometry of @xmath7 and the first two derivatives of the vvd . to proceed with the calculation , we use the following expression , derived in @xcite , relating ricci scalars of metrics related in a manner similar to @xmath15 and @xmath12 . @xmath153 ( where , borrowing notation of @xcite , @xmath154 ) , and @xmath155 \nn \\ \vspace{.2 cm } & & \hspace{.55 cm } \ ; + \l ( k + d_1 \nabla_{\bm q } \ln \omega \r ) \times \nabla_{\bm q } \ln \alpha \omega^2 \nn \\ \nn \\
\mathcal{j}_d & = & 2 r_{ab } q^a q^b + k_{ab}^2 -k^2 \nn \\ & = & \epsilon \l ( r - \mathcal r_{\sigg } \r)\label{eq : jcjd}\end{aligned}\ ] ] one can now plug in the form of @xmath123 and @xmath156 from eqs .
( [ eq : fix1 ] ) and ( [ eq : fix2 ] ) to find the form of rhs .
this is the most important , and also the most lengthy , part of the calculation .
the computation is largely aided by identities @xmath82 and @xmath83 ( eqs .
( [ eq : vvd - ids ] ) ) satisfied by the vvd .
some of the key steps are sketched in the appendix [ app : rsdetail ] .
the final result turns out to be @xmath157 } _ { \bm{q_0 } } \;-\ ; \underbrace{\frac{\mathcal s_{\lp}}{\lambda^2 \mathcal s'^2_{\lp } } \biggl\ { k_{ab } k^{ab } - \frac{1}{d_1 } k^2 \biggl\}}_{\bm{q_{\rm k } } } \\ & \phantom{\biggl [ \frac{\sigma^2}{\mathcal s_{\lp } } \zeta^{-{2}/{d_1 } } \ ; \mathcal r_{\sigg } \;-\ ; \frac{d_1 d_2}{\mathcal s_{\lp } } + 4 ( d+1 ) \overset{\bullet}{\delta}_{\mathcal s } \biggl ] \;-\ ; \;\;\;\;\ ; } + \underbrace{4 \mathcal s_{\lp } \biggl\ { - \frac{d}{d_1 } \l [ ( \ln \delta_{\mathcal s})^{\bullet } \r]^2 + 2 ( \ln \delta_{\mathcal s})^{\bullet\bullet } \biggl\}}_{\bm{q_\delta } } \end{aligned } } \ ] ] where we have defined @xmath158 , @xmath159 , and @xmath160 .
it is not too difficult to see that for @xmath161 , the rhs above reduces to @xmath149 .
( to verify this , one has to use @xmath162 from eqs .
( [ eq : vvd - ids ] ) above along with @xmath163 . )
it is crucial to note here that we have not used any of the covariant taylor expansions yet ; the above expression , therefore , does not assume the region of spacetime under consideration to be smooth ( i.e. , having finite curvature ) .
this will be important for discussing the implications of our framework for cosmological and black hole singularities , which we wish to address in future work . for the purpose of this paper ,
however , we shall assume that we are working in smooth regions of spacetime , so that the various taylor expansions given in sec .
[ sec : vvd ] can be used .
the significance of separating out the rhs into @xmath164 and @xmath165 will become evident shortly .
the above expression holds the key to understand non - perturbative effects of a covariant short distance cut - off on spacetime curvature .
let us therefore first highlight some of it s most important mathematical aspects , before taking it s @xmath36 limit .
the expression contains no derivatives of the function @xmath22 higher than one .
this is an extremely delicate mathematical point ; as can be seen from the details provided in the appendix , terms of the form @xmath166 do in fact appear in the intermediate steps , but they cancel out in the final expression .
since @xmath22 represents ( in general non - perturbative ) effects of quantum gravity on invariant distance between spacetime events , the non - existence of higher derivatives of @xmath42 in @xmath16 is of deep conceptual importance it tells us that semi - classical effects of quantum gravity can be captured only via limited information about the precise details of quantum gravity .
the ricci bi - scalar @xmath16 is completely described by _
geodesic structure of spacetime _
, characterized by : + ( a ) @xmath167 ( intrinsic curvature ) + ( b ) @xmath76 ( extrinsic curvature ) , and + ( c ) the van vleck determinant @xmath3 .
the extrinsic curvature of @xmath7 appears only in a very special combination , which ( as we will see in a moment ) , is responsible for _ no coincidence limit divergences _ !
this is essentially a consequence of eq .
( [ eq : kabcomb ] ) for @xmath168 .
the coincidence limit of @xmath16 gives us a local scalar @xmath35(p_0)$ ] at each spacetime event @xmath6 which will depend on @xmath9 .
we wish to ask whether this scalar gives back @xmath149 , the ricci scalar of the background spacetime , when @xmath9 is set to zero .
the limit @xmath36 must be taken with care .
first of all , note that any @xmath9 independent contribution must come from @xmath170 .
the contribution from @xmath171 and @xmath165 can only be @xmath172 , since @xmath19 .
let us therefore first focus on @xmath170 . to do this
, we invoke the coincidence limit expansions of various quantities given in eqs .
( [ eq : taylor - exp ] ) , in addition to the following well known covariant taylor expansion of the vvd @xmath173 from which it is easy to see that @xmath174(p_0 ) \label{eq : q0step1}\end{aligned}\ ] ] also , using the last of eqs .
( [ eq : taylor - exp ] ) for @xmath167 , and the fact that @xmath175 , we get @xmath176 where @xmath177(p_0 ) + \ldots$ ] the limit @xmath36 limit of the rhs above is most easily evaluated using the _ lhospital _ s rule ( note that both the numerator and denominator vanish in this limit ) : @xmath178(p_0 ) \label{eq : q0step2}\end{aligned}\ ] ] from eqs .
( [ eq : q0step1 ] , [ eq : q0step2 ] ) , we immediately get @xmath179 \l [ r_{ab } q^a q^b \r](p_0 ) \nn \\ & = & \epsilon d \l [ r_{ab } q^a q^b \r](p_0 ) \nn \\ & = & \epsilon d \mathcal e(p_0)\end{aligned}\ ] ] + which is one of the most important results in this paper .
the above limit , being independent of @xmath22 , is precisely the relic left by the presence of a zero point length .
( qualitatively , this is similar to the various quantum anomalies one encounters in qft in curved spacetimes @xcite ; see @xcite ( first reference ) for a much detailed conceptual discussion of this and related points . )
now consider @xmath171 , with the condition @xmath21(0 ) < \infty$ ] .
this term would be divergent in the coincidence limit were it not for the fact that the combination of extrinsic curvature appearing here is @xmath180 , as is readily seen from eq .
( [ eq : kabcomb ] ) with @xmath168 .
this is a very strong result .
any other combination of extrinsic curvature tensor would lead to coincidence limit divergences in @xmath35(p_0)$ ] , and the reason the right combination appears here is completely due to the presence of vvd . in the absence of it , we would indeed get divergences , as can be seen from the result in @xcite . the final term , @xmath165 , is a smooth function which would yield further @xmath172 dependent terms coupled to the background curvature .
these terms can be read off from the known covariant taylor expansions of the vvd .
to summarize , then , we have proved the following : @xmath181(p_0 ) & = & \epsilon d \l [ r_{ab } q^a q^b \r](p_0 ) \end{aligned } } \ ] ] @xmath182 . for the sake of completeness
, we quote below the higher order terms in @xmath9 which can be obtained if one further assumes that all the quantities involved allow a legitimate series expansion in @xmath9 in the coincidence limit .
we must , however , caution that such an assumption could be highly objectionable ( even wrong ) in full quantum gravitational context .
nevertheless , the expansion turns out to be : @xmath183(p_0 ) & = & \epsilon d { \mathcal e}(p_0 ) + \frac{2 \epsilon
( d+1)}{3}(\nabla_{\bm q } { \mathcal e } ) \lp \nn \\ & + & \biggl\{\frac{1}{18}\biggl(d+2-\frac{2}{\mathcal s'(0)^2}\biggr)\l({\mathcal e}_{ab}^2 - \frac{{\mathcal e}^2}{d_1}\r ) \nn \\ & + & \frac{1}{4}(d+2)\nabla_{\bm q}^2 { \mathcal e } \biggl \ } \lp^2+o(\lp^3)\end{aligned}\ ] ]
in the previous section , we analyzed the ricci bi - scalar @xmath16 for the qmetric and showed that it s coincidence limit becomes proportional to @xmath184 in the limit @xmath185 .
we now calculate the contribution of the surface term @xmath186 for the equi - geodesic surfaces using the qmetric .
this is a much simpler calculation than that of @xmath16 , and is along the same lines as the one given in @xcite ( second reference ) , so we will be brief .
we start with the following relation between induced metrics and the extrinsic curvatures @xcite @xmath187 since by definition @xmath188 , the series for @xmath77 in eq .
( [ eq : taylor - exp ] ) readily yields @xmath189 putting everything together , we get @xmath190 the above limit of the surface term corresponds to the surfaces @xmath7 straddling very close to the causal horizon of @xmath6 . if one takes the limit @xmath36 above , it gives zero , and hence there is no non - trivial effect on the surface term as was found in @xmath16 .
@xmath182 . again , _ if one can justify a series expansion in @xmath9 _ , one gets @xmath191 \nn \\ & \overset{d=4}{= } & 3 \lp^2 - \frac{5}{6}{\mathcal e}(p_0 ) \lp^4 + o(\lp^5)\end{aligned}\ ] ] in which the first two terms have the same form as in @xcite ( second reference ) , where no expansion in @xmath9 was needed .
as mentioned in the introduction , one needs only a very few conceptual inputs from semiclassical gravity to deduce some basic facts about spacetime at small scales .
the existence of a lower bound on spacetime intervals is one such input , and if it turns out to be a fundamental feature of quantum gravity , it makes sense to look for a geometric description of spacetime in terms of objects which are likely to be more useful in incorporating such a lower bound .
bi - tensors , then , provide a natural choice , and indeed , if one looks deeper into our theories of classical gr and qft , some of the most important features of these theories , such as geodesic deviation , focussing of geodesics , causal structure , singularity structure of two point functions etc , are characterized in terms of bi - tensors . characterizing the small scale topology of spacetime
is another aspect which would necessitate a description directly in terms of bi - tensors such as the distance function @xmath29 .
this is the point of view which was stressed in @xcite , developed and described in much greater conceptual detail in @xcite , and has been put on a much more rigorous basis in the present paper .
our general expression ( [ eq : finalrsq ] ) for the ricci bi - scalar @xmath16 of the qmetric presents a natural basis for the description of gravitational dynamics by a non - local action
. this can be particularly relevant for study of spacetime singularities , where one can not use covariant taylor expansions .
although mathematical complexity forced us to look at only the ricci scalar ( instead of the full riemann tensor ) for geometries with a covariant short distance cut - off , the resulting expression very elegantly and concretely expresses the key idea : curvature of spacetime might be solely expressible in terms of behaviour of it s geodesics and related bi - tensorial quantites . in absence of a lower bound on geodesic distances ,
such a description would coincide with the standard one in terms of local tensors such as @xmath192 etc .
however , if there exists a minimal length , then the non - local character of bi - tensors , combined with the non - analytic deformation of geometry necessitated by such a minimal length , might lead to a very different description of spacetime curvature at smallest of scales ; in particular , it may leave a relic independent of the details or value of the short distance cut - off , thereby acting as a crucial guidepost towards our understanding of classical gravity itself .
the mathematical results derived here , for example , seem to strongly support the so called emergent gravity paradigm , in which gravitational dynamics is described in terms of thermodynamics of future causal horizon of an event @xmath6 .
two of the key ideas in this context the use of _ local rindler frames _ as probes of spacetime curvature ( due to jacobson @xcite ) , and a variational principle based on _ entropy functional _ ( due to padmanabhan et .
@xcite ) find a unified and purely geometric description in our framework in terms of equi - geodesic _ surfaces _ ( which replace the rindler _ trajectories _ ) and the @xmath193 term of the coincidence limit of ricci bi - scalar of the qmetric , which happens to have the same form as the entropy functional .
the possibility of description of geometry in terms of two point functions of quantum fields has also been emphasized , in a series of paper , by kempf and collaborators @xcite .
the connection with the work presented here is obvious : the uv behavior of two point functions is in one to one correspondence with vanishing of geodesic distances in the coincidence limit , which , of course , was the basis for our input * q2*. in fact , there could also be a more fundamental connection at the level of geometry itself .
for example , it was pointed out in @xcite that the commutation relation between position and momenta , @xmath194 $ ] , would in general acquire a correction on a curved manifold , thereby affecting the resultant uncertainties .
we hope to present more details on this particular connection in a future work ( which is in progress ) .
we hope to apply the results derived here to analyze implications of a lorentz invariant minimal length for issues such as cosmological and black hole singularities , trans - planckian problem in black hole physics and cosmology , and possible relevance for the cosmological constant problem .
_ acknowledgements _ dk thanks t. padmanabhan for comments and discussion on various aspects of the ideas presented here .
djs is supported by fellowship from the ministry of human resource development ( mhrd ) , india .
the complete derivation of eq .
( [ eq : finalrsq ] ) is lengthy , but there are several key structural aspects of the result which are worth highlighting , since these are responsible not only for the final simple form of @xmath16 , but also for the second fact is more non - trivial , and there is no simple reason to have expected it ! in fact , since @xmath15 depends on @xmath156 , and @xmath156 depends on @xmath197 , one expects @xmath16 to involve terms such as @xmath198 , which it does not .
although we do not have a completely geometric understanding of why this _ must _ happen , we highlight below _ how _ the cancellations happen which point to the following two reasons for these subtle cancellations : * the fact that the coupling between @xmath15 and @xmath12 is _ disformal _ rather than _
conformal_. * the fact that the @xmath166 contribution of the conformal part is cancelled by a contribution coming from the disformal one . to demonstrate these , we start from eq .
( [ eq : mrs - final ] ) , and note that the only possibility of occurrence of second derivatives of @xmath42 is from @xmath199 , which we write as @xmath200}_{\mathcal{j}_{c1 } } \nn \\ \vspace{.2 cm } & & \hspace{.55 cm } \ ; + \underbrace{\l ( k + d_1 \nabla_{\bm q } \ln \omega \r ) \times \nabla_{\bm q } \ln \alpha
\omega^2 } _ { \mathcal{j}_{c2}}\end{aligned}\ ] ] the cancellation of higher order derivatives happen between the terms @xmath201 , which is the contribution of purely conformal transformation , and @xmath202 , contributed by the disformal part . simplifying the above expression
, we get @xmath203 and @xmath204 if the metrics @xmath15 and @xmath12 are conformally coupled , @xmath205 and hence @xmath206 . in such a case
, the higher derivatives of @xmath42 would not cancel ; the cancellation is solely a consequence of the disformal coupling between the metrics .
+ 99 for a small sample of relevant work which argue for the breakdown of locality , see : l. garay , phys .
lett . * 80 * , 2508 ( 1998 ) [ gr - qc/9801024 ] ; s. doplicher , k. fredenhagen , j. roberts , comm .
phy . * 172 * , 187 ( 1995 ) [ hep - th/0303037 ] ; r. sorkin , _ does locality fail at intermediate length - scales _ ( 2007 ) [ gr - qc/0703099 ] ; s. giddings , phys . rev
. d * 74 * , 106006 ( 2006 ) [ hep - th/0604072 ] .
bryce dewitt , _ the global approach to quantum field theory _
( clarendon press , oxford , 2003 ) , 2 vols ; matt visser , phys . rev . d * 47 * , 2395 ( 1993 ) [ arxiv : hep - th/9303020 ] ; m. c. gutzwiller , _ chaos in classical and quantum mechanics _
, interdisciplinary applied mathematics 1 , springer verlag ( 1990 ) .
ivan g. avramidi , _ covariant methods for the calculation of the effective action in quantum field theory and investigation of higher - derivative quantum gravity _
( phd thesis , moscow state university ( 1986 ) ; udk 530.12:531.51 ) [ hep - th/9510140 ] .
b. s. dewitt , phys .
lett . * 47 * , 1647 ( 1981 ) ; e. alvarez , phys .
lett . b * 210 * , 73 ( 1988 ) ; e. alvarez , j. cespedes , e. verdaguer : phys .
b * 289 * , 51 ( 1992 ) , phys .
d * 45 * , 2033 ( 1992 ) , phys .
b * 304 * , 225 ( 1993 ) . | it has recently been argued that if spacetime @xmath0 possesses non - trivial structure at small scales , an appropriate semi - classical description of it should be based on non - local bi - tensors instead of local tensors such as the metric @xmath1 .
two most relevant bi - tensors in this context are synge s world function @xmath2 and the van vleck determinant ( vvd ) @xmath3 , as they encode the metric properties of spacetime and ( de)focussing behaviour of geodesics .
they also characterize the leading short distance behavior of two point functions of the dalembartian @xmath4 .
+ + we begin by discussing the intrinsic and extrinsic geometry of _ equi - geodesic _ surfaces @xmath5 in a geodesically convex neighbourhood of an event @xmath6 , and highlight some elementary identities relating the vvd with geometry of @xmath7 . as an _ aside _
, we also comment on the contribution of @xmath7 to the surface term in the einstein - hilbert ( eh ) action and show that it can be written as a _ volume _ integral of @xmath8 . + + we then proceed to study the small scale structure of spacetime in presence of a lorentz invariant short distance cut - off @xmath9 using @xmath2 and @xmath3 , based on some recently developed ideas . we derive a @xmath10nd rank bi - tensor @xmath11 $ ] which naturally yields geodesic intervals bounded from below and reduces to @xmath12 for @xmath13 .
we present a general and mathematically rigorous analysis of short distance structure of spacetime based on ( a ) geometry of equi - geodesic surfaces @xmath7 of @xmath12 , ( b ) structure of the non local dalembartian @xmath14 associated with @xmath15 , and ( c ) properties of vvd .
in particular , we prove the following : ( i ) the ricci _ bi - scalar _ @xmath16 of @xmath15 is completely determined by @xmath7 , the tidal tensor and first two derivatives of @xmath3 , and has a non - trivial _ classical _ limit ( see text for details ) : @xmath17 ( ii ) the ghy term in eh action evaluated on equi - geodesic surfaces straddling the causal boundaries of an event @xmath6 acquires a non - trivial structure .
+ + these results strongly suggest that the mere existence of a lorentz invariant minimal length @xmath9 can leave _ unsuppressed residues _ independent of @xmath9 and ( surprisingly ) independent of many precise details of quantum gravity . for e.g. ,
the coincidence limit of @xmath16 is finite as long as the modification of distances @xmath18 satisfies ( i ) @xmath19 ( the condition of minimal length ) , ( ii ) @xmath20 , and ( iii ) @xmath21(0 ) < \infty$ ] .
in particular , the function @xmath22 , which should eventually come from a complete framework of quantum gravity , need not admit a perturbative expansion in @xmath9 .
+ + finally , we elaborate on certain technical and conceptual aspects of our results in the context of entropy of spacetime and classical description of gravitational dynamics based on noether charge of diff invariance instead of the eh lagrangian . 0.5 in |
wise [email protected] ( hereafter wise j0720@xmath00846 ) is an m9.5 dwarf originally identified by as a possible new member of the local 8 pc sample .
it was previously missed in searches for nearby low - mass dwarfs due to its low galactic latitude ( @xmath7 ) and small proper motion ( [email protected] mas yr@xmath8 ) .
subsequent follow - up observations by @xcite ; ( * ? ? ?
* hereafter b15 ) ; and confirmed the late - type nature and proximity of the source ( [email protected] pc ) , and have identified it as having an unusually high recessional velocity ( + 83.8 km s@xmath8 ) , indicative of old disk / thick disk kinematics . @xcite
have deduced that wise j0720@xmath00846 had one of the closest stellar approaches to the sun inferred to date , passing within 0.25@xmath9 pc over a period of 60,00085,000 years ago , possibly penetrating the outer oort cloud .
evidence for a t - type brown dwarf companion to wise j0720@xmath00846 was reported by b15 , based on both near - infrared spectral analysis and the presence of a candidate resolved source in high - resolution laser guide star adaptive optics ( lgsao ) imaging . in support of this , i15 reported infrared excess in 11 and 22 photometry from the wide - field infrared survey explorer ( wise ; @xcite ) that could be attributed to a low - temperature secondary .
however , the detection of the secondary was marginal due to its close separation ( 140 mas ) , large magnitude difference with the primary ( @xmath10 ) , and poor lgsao correction , and neither b15 nor i15 detected significant radial velocity ( rv ) variability over overlapping @xmath113 month timescales . confirming the presence of this putative companion and assessing the degree of its gravitational perturbation on the primary is important for determining an accurate parallax and space motion for wise j0720@xmath00846 .
this system is only the second ( candidate ) binary to be identified among the 14 known late - m dwarfs within 10 pc of the sun .
another remarkable trait of wise j0720@xmath00846 reported in b15 and i15 is its weak yet highly variable magnetic emission .
quiescent h@xmath3 emission was observed to be at or below typical values for m9l0 dwarfs , with a relative power of @xmath12 @xmath13 @xmath05 .
however , b15 reported the detection of multiple flaring events , both in white - light photometry and h@xmath3 line emission .
these flares were infrequent ( @xmath141% effective duty cycle ) but nevertheless produced order - of - magnitude variations in nonthermal emission .
unfortunately , the x - ray limit of @xmath15 @xmath16 @xmath03.2 reported by i15 does not provide a stringent constraint on the high - energy nonthermal emission from this source .
overall , wise j0720@xmath00846 appears to be similar in magnetic behavior to the rapidly rotating bri 0021 - 0040 @xcite , being weakly active but with occasional strong bursts .
deducing how these bursts relate to age or rotation , or in the case of wise j0720@xmath00846 interaction with a ( putative ) close companion , may provide critical clues for understanding the overall decline in optical and x - ray magnetic activity across the m dwarf / l dwarf transition and contrary trends in nonthermal radio emission . in this article , we report new observations of wise j0720@xmath00846 at optical , near - infrared and radio wavelengths which confirm both the bursting and binary nature of this source . in section 2
we report the detection of low - level quiescent and bursting radio emission based on data obtained with the karl g. jansky very large array ( hereafter vla ) , and limits on h@xmath3 variability from asynchronous narrow - band photometric monitoring . in section 3
we report new imaging and high resolution spectroscopic observations that confirm the presence of a t5.5 companion , and provide the first indications of orbital motion .
we use these data to make constraints on the orbital configuration of the system in section 4 .
our results are discussed in section 5 .
wise j0720@xmath00846 was observed with the vla ( project code 14b@xmath0313 , pi burgasser ) in the compact c - configuration ( baselines of [email protected] km ) on 2014 november 11 from ut 08:48:50 to 14:47:48 .
the widar correlator was set up for c - band continuum observations with two basebands , each having eight 128mhz sub - bands centered at 5.0ghz and 6.0ghz , for a frequency range of 4.488ghz to 6.512ghz and a total bandwidth of @xmath132ghz .
each sub - band had four polarization products ( rr , ll , rl , lr ) and sixty - four 2mhz channels ; a 5sec dump rate was used . after observation of our primary calibration source
3c48 to set the absolute flux scale and measure the complex bandpass , we conducted a sequence of 4min cycles , with 3min on wise j0720@xmath00846 and 1min on the gain calibrator qsoj0730@xmath01141 .
this observational strategy optimizes image quality by frequent monitoring ( and hence correction ) of phase fluctuations .
all data were reduced with the astronomical image processing software package ( aips ; @xcite ) following best practices for wide - band radio data reduction .
radio - frequency interference ( rfi ) was present throughout the observation , being especially persistent and strong in the sub - bands covering 5.782ghz to 6.512ghz , and also present at varying levels in several other sub - bands .
all data affected by rfi were removed .
in addition , sub - band 0 of each baseband failed to produce robust data , and these measurements were also removed .
after data flagging , our average frequency was 5.27ghz and our total bandwidth was reduced to @xmath131.2ghz .
the final cleaned and calibrated data were analyzed by imaging and performing time - series analysis on the @xmath17-data .
figure [ fig : vla - image ] shows imaging of the entire data set ( all unflagged times and frequencies ) at the coordinates of wise j0720@xmath00846 , which reveals a weak , elongated source .
this is interpretted to be a blend between weak quiescent emission from wise j0720@xmath00846 and an unrelated faint background source , each having similar flux densities of @[email protected]@xmath2jy ; the map rms noise level is 2.8@xmath2jybm@xmath8 ] using the aips task dftpl and the procedures described in @xcite , we analyzed the time - series data at the peak flux position , shown in figure [ fig : vla - series ] .
this series reveals weak but nonzero emission over the entire course of the observation , with a short burst of emission at ut 13:21:37 . masking out the burst
, we measure a mean quiescent flux density of 15@xmath13 @xmath2jy with no evidence of statistically significant variability .
= 296.0 for 295 degrees of freedom , yielding a @xmath19-value of 0.46 . note that with the bursting emission included @xmath20 = 382.5 and @xmath19 @xmath140.1% . ]
this is the weakest quiescent emission detected for a very low mass dwarf to date , comparable to the quasi - quiescent emission reported for the t6.5 dwarf 2mass j10475385 + 2124234 ( * ? ? ?
* @xmath2jy ) .
the apparent flux density translates into a specific radio luminosity of @xmath21 = ( 7@xmath12)@xmath2210@xmath23 erg s@xmath8 hz@xmath8 at the [email protected] pc distance of wise j0720@xmath00846 , and @xmath24 = @[email protected] assuming @xmath25 = @[email protected] ( b15 ; i15 ) and @xmath26 = 5.28ghz .
the specific luminosity is 12 orders of magnitude below measurements of previously detected late m and l dwarfs and the relative luminosity falls below the activity - rotation relation of @xcite given this source s @xmath27 = [email protected] km s@xmath8 ( section 3.4 ) .
indeed , in terms of rotation and radio power , wise j0720@xmath00846 is similar to the m7 dwarf vb 8 @xcite and the m8 dwarf denis j1048 - 3956 @xcite , both nearby late m dwarfs with weak but variable magnetic emission and modest rotation rates .
the weakness of the quiescent emission prevents us from determining either its spectral behavior across the 4.46.5 ghz band or its polarization .
a close - up view of the radio burst around ut 13:21 is shown in figure [ fig : vla - series ] . the burst appears to have begun during observations of the secondary calibrator , so our data sample it only at or after its peak emission .
reimaging of the radio data in the period around this burst shows it to be associated with a bright point source at the coordinates expected for the brown dwarf ( figure [ fig : vla - flare - image ] ) .
the burst is highly polarized , with emission in stokes @xmath28 of 175@xmath128 @xmath2jy and stokes @xmath29 of @xmath0150@xmath128 @xmath2jy , implying left - handed circular polarization of 84@xmath30% ( 1@xmath31 equivalent uncertainties ) ; i.e. , consistent with full polarization .
there is no evidence of spectral variation in the burst to within our measurement uncertainties .
we modeled the time series emission as both a rotating spot ( `` pulse '' ) feature with a gaussian profile , and as an exponentially declining flare , @xmath32 after peak emission .
the former yields a peak flux density of 310@xmath140 @xmath2jy , @xmath1320 times greater than the quiescent emission , and consistent with a total emitted energy of ( [email protected])@xmath2210@xmath33 erg over the 4.56.5 ghz band ( assuming a flat spectrum ) .
the flare model yields a decay time constant of @xmath34 = [email protected] min and , depending on whether the burst initiated at the start or end of the calibration period , a total emitted energy of ( 13)@xmath2210@xmath33 erg .
the burst emission occurs for at most 3 min during our 5 hr of observation of wise j0720@xmath00846 , or 1% of the total time on - source .
this is consistent with the flaring duty cycle inferred from the aperiodic white light bursts reported in b15 .
thus , we favor an infrequent flare mechanism for this emission as well .
we nevertheless note that the pulse / spot model implies an emission region @xmath1312@xmath35 in longitudinal extent ( assuming a rotation period of @xmath1314 hr ; see section 4 ) , which is consistent with previous periodic pulse detections @xcite .
longer - term monitoring would be required to distinguish between the flare and pulse hypotheses .
lcccccc 2014 feb 18 & 56706.26 & 900 & 11 & 4:42:32 & 3.03 & 4 + 2014 feb 20 & 56708.24 & 900 & 10 & 4:38:51 & 2.63 & 3 + 2014 feb 22 & 56710.24 & 300 & 38 & 3:55:13 & 3.97 & 4 + 2014 feb 24 & 56712.28 & 300 & 13 & 4:55:02 & 3.52 & 3 wise j0720@xmath00846 was monitored over four nights on 2014 february 18 , 20 , 22 and 24 ( ut ) using the facility ccd camera on the 1-meter nickel telescope at lick observatory ( table [ tab : halpha ] ) .
the ccd was configured for 2@xmath222 binning for a pixel scale of 0@xmath3637 pixel@xmath8 .
after acquisition and centering at @xmath28-band , wise j0720@xmath00846 was monitored without dithering through the narrow - band h@xmath3 filter ( @xmath37 = 6557 , @xmath38 = 15 ) , with integration times of 900 s on february 18 and 20 and 300 s on february 22 and 24 .
total monitoring periods per night spanned 2.633.97 hr , but due to overheads and pauses during occasional clouds the on - source time totalled 7.25 hr over the entire observing run .
bias frames and quartz flat field lamps were also acquired each night for detector calibration .
data were reduced using standard image reduction techniques .
aperture photometry of wise j0720@xmath00846 and nearby non - saturated stars was measured using a variable aperture scaled to encapsulate 80% of each source s peak brightness , and an annulus of 63100 pixels ( 23@xmath3936@xmath39 ) was used to subtract foreground emission .
we did not observe a photometric calibrator during these observations , so no attempt was made to measure absolute h@xmath3 fluxes .
instead , we used the mean flux of non - variable stars in the field of view to compute a reference light curve , and used this to normalize the photometry for wise j0720@xmath00846 over the course of each night .
uncertainties were dominated by source photometry , of order 2050% .
we found no significant flux variations over the course of the four nights of observation . at the measured noise level
, we can rule out bursts 23 times above quiescent emission .
b15 reported a single h@xmath3 flare with a line flux 5 times greater than quiescent emission , and an order of magnitude brighter than the local continuum .
such a flare would have been easily detected in our observations had it occured .
assuming a flare period of @xmath135 min , we infer a flare indicidence rate of @xmath141% , similar to the white light flare rates reported in b15 and the radio flare reported here .
all of these observations reinforce the conclusion that wise j0720@xmath00846 is a weakly active and infrequently bursting source .
wise j0720@xmath00846 was re - observed with the near - infrared camera 2 ( nirc2 ) and lgsao system @xcite on the keck ii 10 m telescope on 2015 january 11 ( ut ) in mostly clear and windy conditions with 0@xmath368 seeing .
the narrow field - of - view ( fov ) camera was used to obtain dithered observations in broad - band mko @xmath40 , @xmath41 and @xmath42 filters , and the medium - band @xmath43 filter sampling 1.541.65 @xmath44 .
the @xmath45 mag field star usno 0812 - 0137390 was used to correct for tip - tilt aberrations .
we achieved better strehl ratios than observations reported in b15 , 1020% depending on wavelength .
this allowed us to easily resolve the candidate companion reported in that study in all four filters , as shown in figure [ fig : image ] . to extract relative photometry and astrometry , we performed point source function ( psf ) fitting of each individual image using a two - dimensional asymmetric moffat profile optimized to the psf of the primary component ( i.e. , with the secondary masked ) .
measurements are reported in table [ tab : nirc2 ] .
we find a separation of 197@xmath13 mas at position angle 256@xmath57@xmath10@xmath56 , wider than , and at a marginally distinct position angle as , the candidate source previously reported ( 139@xmath114 mas at 262@xmath462@xmath35 ) .
as the source detected in these data would have been easily resolved in prior observations , we conclude that statistically significant relative motion has been observed between the two epochs .
ll @xmath47 & [email protected] + @xmath48 & [email protected] + @xmath49 & [email protected] + @xmath50 & [email protected] + @xmath51 ( mas ) & 197@xmath13 + @xmath51 ( au ) & [email protected] + @xmath52 ( @xmath35 ) & [email protected] + primary spt & [email protected] + secondary spt & [email protected] relative motion can be due to differential motion between two physically unrelated sources or orbital motion in a gravitationally bound binary .
figure [ fig : cpm ] displays the estimated center of mass motion and component positions between our 2014 and 2015 imaging data , assuming systemic astrometry from b15 and @xmath53 = 0.4 ( see below ) . it is clear that the predominant motion of both sources is co - aligned with the proper motion of the system , particularly in declination ; the change in position angle is inconsistent with the secondary being an unmoving background source at the 16@xmath31 confidence level .
we therefore determine that wise j0720@xmath00846 is a co - moving binary system , and identify significant astrometric orbital motion over a one year - period .
our new observations provide greatly improved relative photometry of the two components in the four bands measured , and refine the coarse @xmath49 estimate from b15 ( table [ tab : nirc2 ] ) .
there is a significant difference in relative brightnesses in the @xmath43 and @xmath41-band filters consistent with strong ch@xmath54 absorption in the secondary .
we used the relative photometry to better constrain the spectral types of the components through spectral template fitting . following the procedures described in @xcite
, we combined 512 m7-l1 and 125 t0-t7 spectral templates from the spex prism library ( spl ; @xcite ) , scaled so that that relative spectrophotometry agreed to within 1@xmath31 of the nirc2 @xmath55 measurements .
photometry was not used for this analysis as the filter profile was unavailable in a digital format ( h. tran , 2015 , priv .
comm . ) . ]
we compared the 16394 binary templates that satisfied these constraints to the combined light spex spectra of wise j0720@xmath00846 from both @xcite and b15 , following the methods described in the latter paper .
figure [ fig : binfit ] shows the best fitting template to the b15 spectrum .
both analyses yield identical results , with component types [email protected] and [email protected] .
the secondary is a half subtype later than , but formally consistent with , the classification reported in b15 .
new high resolution optical and near - infrared spectroscopy of wise j0720@xmath00846 were obtained with the near infrared spectrometer ( nirspec ; @xcite ) on the keck ii telescope on 2014 december 8 ( ut ) , and with the hamilton echelle spectrograph @xcite on the lick observatory shane 3 m telescope on 2015 march 10 ( ut ) .
data were acquired and reduced as described in b15 .
the nirspec spectrum ( figure [ fig : nirspec ] ) is of high quality ( median s / n = 82 ) and similar to data reported in b15 .
the hamilton data had a somewhat lower signal - to - noise ( s / n ) than prior observations ( s / n @xmath13 5 at 7500 ) due to a shorter total integration of 3000 s. both spectra were analyzed as described in b15 for radial motion and , for the hamilton data , h@xmath3 emission equivalent width ( ew ) . despite the low s / n , cross - correlation of the hamilton spectrum with contemporaneous observations of the rv standard gj 251 yielded an rv = + [email protected] km s@xmath8 , significantly different than the mean motion reported in b15 ( + [email protected] km s@xmath8 ) .
marginal h@xmath3 emission was observed in these data , with ew = @xmath02@xmath11 , consistent with the lowest emission states observed in b15 . for the nirspec data , we re - analyzed the observations reported here and the 2014 observations using an updated markov chain monte carlo ( mcmc ) adaptation of the forward - modeling method described in @xcite and b15 .
we used the bt - settl atmosphere models @xcite with updated solar abundance values from @xcite over an effective temperature ( t@xmath56 ) range of 16002900 k and surface gravity ( @xmath57 ) range of 4.55.5 ( cgs ) .
the solar atlas of @xcite was used to model telluric absorption features .
table [ tab : nirspec ] summarizes the rv and rotational velocities ( @xmath27 ) inferred from these analyses , while figure [ fig : nirspec ] shows the best - fit model , with t@xmath56 = 2700 k and @xmath57 = 5.5 , to the most recent nirspec observations .
the rotational velocity measurements are mutually consistent , with a mean value of @xmath27 = [email protected] km s@xmath8 . however , like the hamilton observations , the most recent nirspec rv measurement of + [email protected] km s@xmath8 is significantly distinct from the mean rv reported in b15 ( + [email protected] km s@xmath8 ) and follows the trend of increasing recessional motion . with these independent measurements ,
we conclude that wise j0720@xmath00846a is an rv variable gravitationally perturbed by its brown dwarf secondary .
llcc 2014 jan 19 & 56676.00968 & + [email protected] & [email protected] + 2014 jan 20 & 56677.00048 & + [email protected] & [email protected] + 2014 mar 10 & 56725.71832 & + [email protected] & [email protected] + 2014 apr 12 & 56758.74056 & + [email protected] & [email protected] + 2014 dec 8 & 56999.00802 & + [email protected] & [email protected] + mean & & + [email protected] & [email protected] we note that the best - fitting models for the forward - modeling analyses were consistently in the ranges t@xmath56 = 25002800 k and @xmath57 = 5.05.5 .
the temperatures are somewhat higher than those expected for an m9.5 dwarf ( t@xmath56 @xmath13 2300 k ; @xcite ) , but the surface gravity is consistent with the lack of li i absorption and low - surface gravity spectral features , indicating an age @xmath581 gyr .
we verified that lower temperature and lower surface gravity models yielded identical rvs , so this measurement appears to be insensitive to the specific model over the parameter range examined here .
a robust investigation of the atmospheric parameters of this source is deferred to a later study .
with wise j0720@xmath00846ab verified as a gravitationally bound binary with orbital motion detected in all three spatial dimensions , we can begin to constrain the orbital properties of the system and the physical properties of its components .
we adapted the mcmc analysis described in @xcite to include both rv and relative astrometry measurements .
we employed an orbit model with nine parameters , @xmath59 where @xmath60 is the period of the orbit in years , @xmath61 is the semi - major axis in au , @xmath62 is the eccentricity , @xmath63 is the inclination , @xmath64 is the argument of periastron , @xmath65 is the longitude of nodes , @xmath66 is the mean anomaly at epoch @xmath67 = 2014.0896 ( mjd = 56675.982 ) , @xmath68 is the system mass ratio , @xmath69 is the center of mass ( systemic ) radial velocity in km s@xmath8 , and @xmath70 is the distance in pc . the primary radial velocity as a function of time @xmath71 , @xmath72 , is @xmath73 + v_{com}\ ] ] where @xmath74 and the true anomaly @xmath75 is related to the eccentric anomaly @xmath76 through @xmath77 which is solved by kepler s equation : @xmath78 the angular separation vector from primary component to secondary component , @xmath79 = ( @xmath80(t ) , @xmath81(t ) ) is determined from @xmath82 \\ \delta\delta(t ) = \frac{a}{d } \left[b(\cos{e(t ) } - e ) + g\sqrt{1-e^2}\sin{e(t)}\right]\end{aligned}\ ] ] where @xmath80 and @xmath81 are the angular separations on sky measured in arcseconds , and @xmath83 , @xmath84 , @xmath85 and @xmath86 are the thiele - innes constants @xcite : @xmath87 note that the total system mass ( m@xmath88 in solar masses ) and component masses ( m@xmath89 = m@xmath90/[1@xmath91 , m@xmath92 = @xmath53m@xmath93 ) could in principle be uniquely inferred from these parameters if sufficiently constrained .
we selected an initial parameter set that visually coincided with the observations through manual experimentation .
we then computed a chain of 10@xmath94 parameter sets using the metropolis - hastings algorithm @xcite , at each step varying parameter @xmath95 by drawing a random offset from a normal distribution @xmath96 where @xmath97 is the set of jump steps .
= ( 3 yr , 0.5 au , 0.2,10@xmath35,10@xmath35,10@xmath35,10@xmath35 , 0.2 , 1.0 km s@xmath8 , 1.0 pc ) . ]
we applied additional parameter constraints of 0.5 yr @xmath98 30 yr , @xmath99 and 4 pc @xmath100 8 pc to eliminate improbable regions of parameter space ; note that the eccentricity cutoff is beyond the @xmath101 limit suggested in empirical data by @xcite .
we also limited the component masses to 0.055 m@xmath102 @xmath14 m@xmath89 @xmath14 0.15 m@xmath102 given the spectral classification of the primary and lack of li i absorption in its optical spectrum ( b15 ; i15 ) , and m@xmath103 m@xmath102 given the substellar nature of the secondary .
orbit models were compared to the data using a @xmath20 statistic that combined both rv and relative astrometric measurements : @xmath104 where @xmath105 and @xmath106 are the number of rv and astrometric measurements , respectively , and @xmath31 the measurement errors .
model values were calculated at the same epochs as the observations .
the criterion to adopt successive parameter sets was @xmath107 , where @xmath108 is a random number drawn from a uniform distribution between 0 and 1 .
we compared models to data after each individual parameter change rather than changing all parameters , a procedure we found greatly improved the acceptance rate , which declined from 10% to 1% through the chain .
separate fits were made to the nirspec + nirc2 and hamilton + nirc2 measurements , and to all of the data .
note that in the last case we did not take into account a possible velocity offset between the rv datasets @xcite . to test for convergence , we monitored parameter autocorrelation and the evolution of subchain variance through the chain .
we also generated @xmath109 = 10 chains of length @xmath110 = 10@xmath111 steps on the hamilton + nirc2 data , varying the initial parameter set @xmath112 as @xmath113 where @xmath114 and @xmath115 are the median and half quantile ranges of the posterior parameter distributions ( see below ) and @xmath116 is a normal distribution centered on zero with unit variance .
we quantified the within chain and between chain variance of all parameters using the @xcite scale reduction factor @xmath117 where @xmath118 is the within chain variance for parameter @xmath119 , equal to the average of parameter variances @xmath120 for each of the ten chains using the second half of the chains as the sample ; and @xmath121 is the between chain variance , equal to the variance in parameter averages @xmath122 .
scale reduction factors were within 5% of unity for all modeled parameters with the exception of @xmath60 ( 1.35 ) , @xmath61 ( 1.37 ) and @xmath64 ( 1.51 ) .
as described below , these three parameters were weakly constrained , and this analysis suggests that the primary mcmc chains may not converge for our limited datasets , re - emphasizing that the orbital results presented here should be considered preliminary . figure [ fig : orbit ] shows the best - fit orbits from our separate analyses of the rv and imaging datasets .
table [ tab : orbit_parameters ] lists the best - fit orbital and component parameters , as well as the median and 16% and 84% quantiles of the parameter distributions , after eliminating the first 10% of each chain .
figures [ fig : orbit_parameters ] through [ fig : orbit_parameters_all ] display the distributions and correlations of parameters @xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath53 and m@xmath90 for all three datasets .
+ + + lcccccc + best @xmath20 & 5.83 & & 2.84 & & 27.6 & + @xmath60 ( yr ) & 10.4 & 6.1@xmath123 & 2.9 & 4.1@xmath4 & 3.1 & 4.5@xmath124 + @xmath61 ( au ) & 2.4 & 1.7@xmath125 & 1.1 & 1.3@xmath126 & 1.1 & 1.4@xmath126 + @xmath62 & 0.79 & 0.72@xmath127 & 0.79 & 0.77@xmath128 & 0.80 & 0.76@xmath129 + @xmath63 ( @xmath35 ) & 92.1 & 94.1@xmath130 & 92.8 & 93.6@xmath131 & 94.9 & 94.4@xmath132 + @xmath64 ( @xmath35 ) & 109 & 83@xmath133 & 70 & 76@xmath134 & 54 & 74@xmath135 + @xmath65 ( @xmath35 ) & 81.7 & 82.2@xmath136 & 82.4 & 82.7@xmath137 & 82.2 & 83.3@xmath138 + @xmath66 ( @xmath35 ) & 5 & 20@xmath139 & 10 & 13@xmath140 & 24 & 17@xmath141 + @xmath53 & 0.36 & 0.61@xmath142 & 0.61 & 0.77@xmath143 & 0.91 & 0.75@xmath144 + @xmath69 ( km s@xmath8 ) & 87.2 & 87.3@xmath145 & 87.3 & 87.4@xmath146 & 87.4 & 87.5@xmath147 + @xmath70 ( pc ) & 4.5 & 6.7@xmath148 & 4.5 & 5.4@xmath149 & 6.8 & 6.2@xmath147 + + @xmath150 ( m@xmath102 ) & 0.13 & 0.13@xmath151 & 0.17 & 0.15@xmath152 & 0.16 & 0.15@xmath152 + @xmath153 ( m@xmath102 ) & 0.098 & 0.080@xmath154 & 0.105 & 0.082@xmath155 & 0.081 & 0.081@xmath156 + @xmath157 ( m@xmath102 ) & 0.035 & 0.051@xmath158 & 0.064 & 0.064@xmath159 & 0.074 & 0.062@xmath160 + @xmath161 ( km s@xmath8 ) & 3.0 & 4.4@xmath162 & 7.2 & 6.3@xmath163 & 8.7 & 6.0@xmath164 + @xmath165 ( km s@xmath8 ) & 8.4 & 7.3@xmath166 & 11.9 & 8.5@xmath167 & 9.5 & 8.2@xmath167 the best - fit orbits are exceptionally good fits for the individual rv datasets , with @xmath20 = 5.83 for nine data points ( zero degrees of freedom ) for nirspec + nirc2 and @xmath20 = 2.84 for twelve data points ( three degrees of freedom ) for hamilton + nirc2 .
note that these low @xmath20 values largely reflect the underconstained nature of the solution in orbital phase ; i.e. , a relatively large family of solutions is formally consistent with these data .
the combined dataset yields a somewhat poorer best fit , with modest disagreement between contemporary nirspec and hamilton rvs driving the @xmath20 values .
nevertheless , this solution is nearly identical to that of the hamilton + nirc2 dataset . while the best - fit solutions between the nirspec + nirc2 and the hamilton + nirc2 datasets are distinct , the median parameters are consistent within the uncertainties for all datasets . since parameters inferred from the hamilton + nirc2 data are best constrained , we refer to these in the following discussion .
several of the parameters are very well constrained , most notably the orbital inclination ( @xmath63 = 93@xmath56@xmath6 ) , the longitude of ascending node ( @xmath65 = 82@xmath57@xmath12@xmath51 ) and the center of mass radial velocity ( @xmath69 = + [email protected] km s@xmath8 ) .
the tight constraints on @xmath63 and @xmath65 stem from the near - radial motion of the secondary between the two nirc2 epochs .
the position angle of the secondary changed by only 5@xmath35@xmath12@xmath35 between these two observations but the secondary has moved 40% further away ; this is possible only when the orbit is viewed very close to edge - on or @xmath62 @xmath13 1 . as discussed below
, even with an eccentricity at the proscribed limits , the inclination remains close to 90@xmath35 .
similarly , @xmath65 is constrained by the orientation of the orbit on the sky .
the tight constraint on @xmath69 arises from the small variance in observed rvs and consistency in the values measured in the earlier epochs .
the nirspec + nirc2 and hamilton + nirc2 datasets yield nearly identical values for all three of these parameters .
parameter distributions for @xmath60 , @xmath62 and @xmath53 are more weakly constrained , and the latter two abut the imposed limits , emphasizing that we do not yet have sufficient coverage of the orbit of wise j0720@xmath00846ab to robustly determine them . to gain some insight on how the quality of the fits vary as @xmath62 changes
, we performed additional mcmc chains on hamilton + nirc2 data using models with fixed @xmath62 = 0.2 , 0.4 , 0.6 and 0.8 .
figure [ fig : orbit_eccentricity ] and table [ tab : orbit_parameters_eccentricity ] display the results of this experiment .
we found that all parameters remain consistent with the unconstrained mcmc analysis , although @xmath53 and m@xmath92 values are modestly smaller for the @xmath62 = 0.2 .
parameter uncertainties are considerably larger for the @xmath62 = 0.8 case , in part reflecting the lower @xmath20 values of viable solutions .
it is clear that the @xmath62 = 0.2 and @xmath62 = 0.4 models are poor representations of the rv data , with best - fit @xmath20 values that are significantly worse and can be eliminated with @xmath16895% confidence as compared to the unconstrained model .
= 2.84 , 3 degrees of freedom ) to the constrained models ( @xmath20 = 39.3 and 25.1 , 4 degrees of freedom ) . ]
this analysis supports an eccentic orbit for wise j0720@xmath00846ab , but a firm constraint on its value should emerge with further monitoring . .
[ fig : orbit_eccentricity],title="fig:",width=302 ] .
[ fig : orbit_eccentricity],title="fig:",width=302 ] + .
[ fig : orbit_eccentricity],title="fig:",width=302 ] .
[ fig : orbit_eccentricity],title="fig:",width=302 ] lcccc best @xmath20 & 39.3 & 25.1 & 10.4 & 2.8 + @xmath60 ( yr ) & 4.4@xmath149 & 4.3@xmath169 & 4.2@xmath170 & 3.8@xmath171 + @xmath61 ( au ) & 1.40@xmath172 & 1.38@xmath173 & 1.4@xmath174 & 1.3@xmath175 + @xmath63 ( @xmath35 ) & 94.1@xmath176 & 94.4@xmath177 & 94.2@xmath176 & 93.5@xmath178 + @xmath53 & 0.69@xmath179 & 0.80@xmath180 & 0.82@xmath181 & 0.73@xmath182 + @xmath69 ( km s@xmath8 ) & 85.9@xmath183 & 86.5@xmath184 & 87.0@xmath184 & 87.3@xmath145 + @xmath150 ( m@xmath102 ) & 0.14@xmath152 & 0.14@xmath185 & 0.15@xmath186 & 0.15@xmath152 + @xmath153 ( m@xmath102 ) & 0.083@xmath154 & 0.079@xmath187 & 0.079@xmath188 & 0.083@xmath189 + @xmath157 ( m@xmath102 ) & 0.059@xmath190 & 0.064@xmath191 & 0.066@xmath192 & 0.062@xmath160
our follow - up observations of wise j0720@xmath00846ab verify both the magnetic and binary nature of this nearby system , and provide first constraints on the physical properties of the components .
while these first measurements of orbital motion do not robustly constrain system or component masses , they do constrain the orbital geometry , most notably inclination , which is within a few degrees of edge - on .
the scale of this system is such that it is unlikely to eclipse ; nevertheless , the orbit orientation allows us to make some constraints on the rotational properties of the primary .
the orbital and primary rotational axes of main sequence binaries with close separations ( @xmath1620au ) are generally aligned @xcite .
the only very low mass binary for which spin - orbit alignment has been tested , 2massw j0746425 + 200032ab , shows alignment between orbital and rotational axes of both components to within 5@xmath35 . assuming similar alignment of the rotational and orbital axes of wise j0720@xmath00846a and a radius of 0.1r@xmath193
, our @xmath27 measurement implies a rotation period of [email protected] hr .
this value is remarkably close to the marginally - indicated variability period from white light monitoring reported in b15 , [email protected] hr for a [email protected]% variability amplitude .
the rotation period is somewhat slower than the mean for comparably classified sources @xcite and considerably slower than most highly variable
l and t dwarfs @xcite .
its slow rotation rate and orientation may explain the weaker than average magnetic emission of wise j0720@xmath00846a , in both optical and radio bands , and its somewhat low flaring rate compared to other late m dwarfs . in addition , with a 12 au semimajor axis , magnetospheric interaction between primary and secondary is unlikely to play a role in driving magnetic emission in either source @xcite .
a more promising role for wise j0720@xmath00846ab is as a testbed for brown dwarf evolutionary models .
as demonstrated by several studies ( e.g. , @xcite ) , the inferred masses and atmospheric parameters of several very low mass binaries diverge to varying degrees from evolutionary model predictions . limiting factors for such analyses include lack of knowlege of detailed spectra for individual components , inability to measure component masses , low quality distance determinations , and/or lack of independent age determinations .
while the last factor may be challenging to overcome for wise j0720@xmath00846ab , the proximity and detection of reflex motion in both astrometric and radial coordinates allows the first three to be addressed , and we anticipate that this system will provide a high - quality test of models in the next 5 - 10 years .
finally , we note that despite being able to detect reflex motion in the primary , the inferred close passage of wise j0720@xmath00846ab to the sun deduced by @xcite on the basis of the system kinematics is not ruled out .
the systemic motion estimated here is about 4 km s@xmath8 larger than that assumed by these authors , so the impact parameter is roughly 5% smaller .
the astrometric perturbation of the primary induced by the secondary , of order 100 mas , is sufficiently slow to play little role in modulating the parallax or proper motion of the system significantly given current measurement uncertainties .
improved monitoring of the systemic orbital motion will be needed to make a more precise estimate of the geometry and timescale of wise j0720@xmath00846ab s closest approach , and the corresponding perturbation it has made on our oort cloud comet population .
the authors thank randy campbell , heather hershley , and marc kassis at keck observatory ; and pavl zachary and shawn stone at lick observatory for their assistance with the observations .
we thank our anonymous referee for her / his / their helpful comments , particularly on the mcmc analysis .
c.m . acknowledges funding support from the national science foundation under award no .
the material is based upon work supported by the national aeronautics and space administration under grant no .
nnx15ai75 g .
this research has made use of the simbad database , operated at cds , strasbourg , france ; nasa s astrophysics data system bibliographic services ; the m , l , t , and y dwarf compendium housed at dwarfarchives.org ; and the spex prism libraries at http://www.browndwarfs.org/spexprism . research at lick observatory is partially supported by a generous gift from google .
the authors recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community .
we are most fortunate and grateful to have the opportunity to conduct observations from this mountain .
, i. s. , graham , j. r. , becklin , e. e. , et al . 2000 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
4008 , society of photo - optical instrumentation engineers ( spie ) conference series , ed .
m. iye & a. f. moorwood , 10481055 | we report the detection of radio emission and orbital motion from the nearby star - brown dwarf binary wise [email protected] .
radio observations across the 4.56.5ghz band with the very large array identify at the position of the system quiescent emission with a flux density of 15@xmath13 @xmath2jy , and a highly - polarized radio source that underwent a 23 min burst with peak flux density 300@xmath190 @xmath2jy .
the latter emission is likely a low - level magnetic flare similar to optical flares previously observed for this source .
no outbursts were detected in separate narrow - band h@xmath3 monitoring observations .
we report new high - resolution imaging and spectroscopic observations that confirm the presence of a co - moving t5.5 secondary and provide the first indications of three - dimensional orbital motion .
we used these data to revise our estimates for the orbital period ( 4.1@xmath4 yr ) and tightly constrain the orbital inclination to be nearly edge - on ( 93@xmath56@xmath6 ) , although robust measures of the component and system masses will require further monitoring .
the inferred orbital motion does not change the high likelihood that this radio - emitting very low - mass binary made a close pass to the sun in the past 100 kyr . |
instabilities of the magnetized shear flows play an important role in enhancing transport processes in various astrophysical bodies and laboratory experiments .
shear flows are rather common in astrophysics , and the study of their stability properties is important for the understanding of many phenomena in stars , jets , accretion flows , galaxies , etc .
likely , the best studied shear flow is differential rotation .
it is well known since the classical papers by velikhov @xcite and chandrasekhar @xcite that a differentially rotating flow with a negative angular velocity gradient and a weak magnetic field can be subject to the magnetorotational instability .
this instability has been analyzed in detail in several astrophysical contexts ( @xcite ) because it can be responsible for transport of the angular momentum in various objects . in accretion disks ,
this instability is also well studied by numerical simulations in both linear and non - linear regimes ( see , e.g. , @xcite ) .
astrophysical applications rise great interest in trying to study this instability in laboratory @xcite . the experiments , however , are complicated because very large rotation rates should be achieved .
the plane couette flow is another example of shear flows well studied in laboratory conditions .
the pioneering work on the stability of a magnetized plane couette flow has been done by velikhov @xcite who obtained that a longitudinal magnetic field with the strength @xmath2 has to stabilize the flow ; @xmath3 is the velocity in the center of the channel and @xmath4 is the fluid density .
a sufficient condition of the ideal instability in a parallel magnetic field has been considered by chen & morrison @xcite .
they argued that the magnetic field can provide a destabilizing effect such as the flow , which is stable in the absence of a magnetic field , can be driven unstable by a relatively weak magnetic field .
also , they found that although strong magnetic shear can stabilize shear flow , there exist a range of magnetic shear that causes destabilization .
the linear stability properties of dissipative shear flow in a parallel magnetic field have been considered by lerner & knobloch @xcite .
the authors argued that misaligned linear perturbations can exhibit enhanced decay in such dissipative flows .
stability of incompressible flow in a transverse magnetic field has been studied by takashima @xcite who found that there exist both the stationary and traveling modes of instability .
note that many previous stability analyses have adopted the boussinesq approximation , and have therefore neglected the effect of compressibility .
this is allowed if the magnetic field strength is essentially subthermal , and the sound speed is much greater than the alfvn velocity , @xmath5 but often this can not be realized in real astrophysical conditions and in many numerical simulations . as it was shown by bonanno & urpin @xcite , the compressibility profoundly alters the stability properties of shear flows .
the number of new instabilities may occur in a compressible flow if the magnetic field has a component perpendicular to the flow .
bonanno & urpin @xcite have considered the particular case of differentially rotating flows but , likely , the shear - driven instabilities are typical for other shear flows as well . in this paper , we show that the same sort of mhd instabilities can occur also in a plane couette flow if the magnetic field has a transverse component .
the instability can arise even in a sufficiently strong magnetic field that suppresses other mhd instabilities .
stability analysis done in this paper will hopefully prove to be a useful guide in understanding various numerical simulations that explore the nonlinear development of instabilities and their effects on the resulting turbulent state of shear flows .
consider a plane couette flow with the velocity @xmath6 where @xmath7 , @xmath8 , and @xmath9 are the cartesian coordinates ; @xmath10 , @xmath11 , and @xmath12 are the unit vectors . for the sake of simplicity ,
we assume shear to be linear , @xmath13 , where @xmath3 and @xmath14 are constant . we restrict ourselves to an inviscid fluid .
the equations of compressible mhd read in this case @xmath15
@xmath16 @xmath17 @xmath18 @xmath19 our notation is as follows : @xmath4 and @xmath20 are the density and fluid velocity , respectively ; @xmath21 is the gas pressure ; @xmath22 is the magnetic field , @xmath23 is the magnetic diffusivity , and @xmath24 is the adiabatic index ; @xmath25 is a scalar force introduced in order to provide hydrostatic equilibrium in the basic state . for the sake of simplicity
, the flow is assumed to be isothermal .
the basic state on which the stability analysis is performed is assumed to be quasi - stationary with the magnetic field that has non - vanishing components in all directions , @xmath26 .
generally , a quasi - stationary basic state in such a magnetic shear flow can be achieved only if dissipative effects are taken into account . in the basic state
, the magnetic field @xmath22 should satisfy the stationary induction equation @xmath27 since @xmath22 depends only on the @xmath9-coordinate , we have @xmath28 integrating this equation , we obtain @xmath29 where @xmath30 and @xmath31 are constant .
we can choose the boundary conditions in such a way that @xmath32 that corresponds to the absence of electric currents at the low boundary @xmath33 .
we will assume that the longitudinal magnetic field at the low boundary is much stronger than @xmath34 where @xmath35 is the thickness of the couette flow .
then , @xmath36 , and one can neglect the change of @xmath37 across the basic flow when considering the behaviour of small perturbations .
we assume also that the basic state satisfies the condition of hydrostatic equilibrium in the @xmath7- and @xmath9-directions . in the @xmath7-direction ,
this condition is satisfied always . for the chosen magnetic field , hydrostatic equilibrium in the @xmath9-direction yields @xmath38 eq .
( 9 ) can be satisfied if @xmath39const and the vertical component of the lorentz force is balanced by a scalar force in the basic state , @xmath40 note that , generally , there is no hydrostatic equilibrium in the @xmath8-direction in our model , but departures from equilibrium are small and can lead only to a very slow change of the basic state .
for example , the lorenz force changes the basic velocity profile @xmath41 in accordance with the @xmath8-component of the momentum equation , @xmath42 where @xmath43 . integrating this expression ,
we obtain @xmath44 the lorentz force changes essentially the initial velocity profile on the timescale @xmath45 ( we assume @xmath46 ) .
therefore , one can neglect this departure from hydrostatic equilibrium if the growth rate of instability , @xmath47 , is greater than @xmath48 , or @xmath49 under this condition , the chosen basic state can be considered as quasi - stationary .
we will show that this condition is satisfied in many cases of interest .
we consider the stability of perturbations with the spacetime dependence @xmath50 .
small perturbations will be indicated by subscript 1 , while unperturbed quantities will have no subscript .
then , the linearized mhd - equations read @xmath51 \nonumber \\ - \frac{\nabla p_{1}}{\rho } \ ! + \frac{1}{4 \pi \rho } [ ( \nabla \times \vec{b}_{1 } ) \times \vec{b } + ( \nabla \times \vec{b } ) \times \vec{b}_{1 } ] , \end{aligned}\ ] ] @xmath52 @xmath53 @xmath54 @xmath55 this set of equations determines the behaviour of small perturbations .
a general set of eqs .
( 15)-(19 ) can be substantially simplified under our assumptions regarding the basic state .
as it was mentioned , small departures from hydrostatic equilibrium in the basic state can not influence the behavior of perturbations if inequality ( 14 ) is satisfied .
therefore , the term proportional to @xmath56 on the r.h.s . of eq .
( 15 ) can be neglected since it is proportional to small departures from hydrostatic equilibrium . from eq .
( 19 ) , we have @xmath57 and , hence , @xmath58 .
then , the @xmath7- , @xmath8- , and @xmath9-components of the momentum equation are @xmath59 eq . ( 23 ) yields for the @xmath7- and @xmath8-components of the magnetic field @xmath60 combining now eqs .
( 21 ) , ( 22 ) , and ( 24 ) and taking into account that @xmath61 , we can obtain the equation that contains only perturbation @xmath62 , @xmath63 \left ( \ !
\sigma^2 - c_s^2 \frac{\partial^2}{\partial z^2 } + \frac{b_y}{b_z } \sigma v ' \ !
\right ) \right .
\left . - \frac{b_y}{b_z } \sigma^2 \left [ v ' \left ( \sigma - \eta \frac{\partial^2}{\partial z^2 }
\right ) + \frac{b_y}{b_z } c_{az}^2 \frac{\partial^2}{\partial z^2 } \right ] \right\ } v_{1z } = 0,\end{aligned}\ ] ] where @xmath64 .
this equation can be solved easily since all coefficients are approximately constant in our model .
to solve eq .
( 25 ) , one needs the boundary conditions .
note that the eigenvalues are not very sensitive to the boundary conditions .
therefore , we choose the simplest model conditions and assume that @xmath65 at @xmath33 and @xmath66 .
then , @xmath67 where @xmath68 and @xmath69 is integer . from eq .
( 25 ) , we have the following dispersion equation for the fundamental mode ( @xmath70 ) @xmath71 where @xmath72 , @xmath73 , @xmath74 , @xmath75 , and @xmath76 ; @xmath77 .
this equation describes fast and slow magnetosonic waves modified by shear .
the conditions under which eq .
( 26 ) has unstable solutions can be obtained by making use of the routh - hurwitz theorem ( see @xcite , @xcite ) . in the case of the dispersion equation of a fourth order ,
the routh - hurwitz criteria are written , for example , in @xcite . according to these criteria , eq .
( 26 ) has unstable solutions if one of the following inequalities is fulfilled @xmath78 where @xmath79 .
two conditions ( 27 ) never apply because @xmath80 , @xmath81 , and @xmath82 are positive . in the limit of small magnetic diffusivity , eqs .
( 28)-(29 ) are equivalent to @xmath83 that is the generalization of the condition derived by bonanno and urpin @xcite for differentially rotating flows .
apart from shear , condition ( 30 ) requires non - vanishing @xmath8- and @xmath9-components of the magnetic field .
the direction of @xmath22 and the sign of @xmath14 are insignificant , and the instability may occur for both positive and negative @xmath14 . note that the instability given by eq .
( 30 ) can arise even in a very strong field .
consider criteria ( 28)-(29 ) in the case when dissipation can not be neglected .
condition ( 28 ) can be rewritten as @xmath84 as it was mentioned , our consideration is valid only if the condition @xmath85 is satisfied ( see eq .
( 8) ) that is equivalent to @xmath86 since inequalities ( 31 ) and ( 32 ) are incompatible in the chosen longitudinal field , criterion ( 28 ) can not be fulfilled and , hence , @xmath87 . since @xmath88 in the considered flow
, we can transform criterion ( 29 ) into @xmath89 taking into account eq .
( 32 ) , we can estimate @xmath90 .
then , the last term on the l.h.s .
is of the order of @xmath91 and can be neglected compared to the second term .
hence , criterion ( 33 ) is approximately equivalent to @xmath92 this condition can be fulfilled only if @xmath93 that is the necessary condition of instability . in accordance with this condition
, the imposed longitudinal field should have the same direction as the field stretched from @xmath94 .
if inequality ( 35 ) is satisfied , then the instability arises if @xmath95 , or @xmath96 this inequality can be fulfilled in a wide variety of strongly magnetized flows where the magnetic pressure is greater than the thermal pressure .
( 35 ) and ( 36 ) determine the necessary and sufficient conditions of shear - driven instability in a couette flow .
since the necessary condition of instability is given by eq .
( 35 ) , we consider the roots of eq .
( 26 ) only in the case of negative @xmath97 when @xmath98 . to calculate the growth rate it is convenient to introduce dimensionless quantities
@xmath99 then , eq . ( 26 ) becomes @xmath100 this equation was solved numerically for different values of the parameters by computing the eigenvalues of the matrix whose characteristic polynomial is given by eq .
( 26 ) ( see @xcite , for details ) .
moreover it is not difficult to see that in order to satisfy the constrain eq.(8 ) we must choose @xmath101 .
on @xmath102 for @xmath103 , @xmath104 , and @xmath105 .
solids lines show the growth rate and frequency of the real roots , and the dashed line corresponds to the complex root.,width=340 ] in fig . 1
, we plot the dependence of real roots and real part of complex roots on @xmath102 for @xmath106 , @xmath107 and @xmath108 .
the solid lines show roots when they are real , and the dashed line show the real part of complex roots .
our calculations clearly indicate that two real roots are positive for the considered parameters and , hence , there should exist a new shear - driven instability .
the pair of complex roots split into a pair of real ones at @xmath109 , but these roots always correspond to stable modes .
one unstable root is rather large with the growth rate @xmath110 , and another one is typically about 10 times smaller . for these roots ,
the growth rate varies very slowly with the parameter @xmath102
. only if the alfven frequency is smaller than the characteristic shear frequency and @xmath111 , the growth rate of the most unstable mode decreases .
note that the considered instability occurs at a very large magnetic pressure that exceeds the gas pressure by two orders of magnitude .
.,width=340 ] in fig . 2
, we plot the same dependence as in fig . 1 but for @xmath112 .
the higher value of @xmath113 corresponds to a larger magnetic viscosity and , hence , to a stronger dissipation of perturbations . due to this
, the instability turns out to be suppressed .
indeed , all roots are either negative or have a negative real part . in this case ,
complex conjugate roots have a very small negative part , but real roots dissipate much more rapidly .
note that the critical value @xmath113 that discriminate between stable and unstable flows is @xmath0 , and the instability occurs if @xmath114 .
for example , the growth rate can reach @xmath115 in a flow with @xmath116 .
.,width=340 ] fig .
3 shows the dependence of a real part of @xmath117 on @xmath102 for @xmath108 , @xmath118 , and @xmath105 .
the right panel shows the behavior of roots at the top left region of the left panel where roots are small . comparing with fig .
1 , it is seen that a decrease of the ratio @xmath119 results naturally in a smaller growth rate .
this dependence is qualitatively clear since the considered instability is due to the presence of a transverse field component in a flow and , therefore , a decrease of this component leads to a weaker instability . in the considered range of @xmath102 , both oscillatory and non - oscillatory
modes can arise .
two non - oscillatory modes are unstable if @xmath120 . after merging ,
this couple forms a pair of complex conjugate modes that are unstable if @xmath121 . the growth rate of one non - oscillatory is larger than that of oscillatory modes and can reach @xmath122 .
the growth rate of oscillatory modes is a factor @xmath123 smaller .
note that another couple of modes is always stable .
on @xmath102 for @xmath124 , @xmath125 , and @xmath108.,width=340 ] in fig . 4 , we plot the dependence of a real part of @xmath117 on @xmath102 for @xmath126 , @xmath127 , and @xmath108 .
we show only three roots in this figure since the fourth root has a large negative value , @xmath128 .
the increase of @xmath129 leads to a corresponding increase in the growth rate . like the previous case ,
the instability can arise either in oscillatory or non - oscillatory regimes .
two non - oscillatory modes are unstable if @xmath130 .
after merging at @xmath131 , these real roots form a couple of complex conjugate roots that are unstable at @xmath132 .
another pair of modes is always stable .
the growth rate of unstable modes is @xmath133 and increases slightly with @xmath102 .
note that oscillatory modes can grow faster than non - oscillatory ones in this case .
on @xmath102 for @xmath134 , @xmath135 , and @xmath136.,width=340 ] fig .
5 shows the growth rate of instability for a very small value of @xmath136 .
this @xmath137 corresponds to the magnetic pressure approximately three orders of magnitude greater than the gas pressure . despite a very high magnetic pressure
, the instability can still occur .
this is in an agreement with our analytic result that the instability should not be suppressed by a strong magnetic field .
the dependences in fig
. 5 are qualitatively very similar to those shown in fig .
two non - oscillatory modes are unstable in this case as well .
one unstable mode has a very small growth rate @xmath122 , but another one grows much faster , @xmath138 .
the growth rate of the fastest growing mode is even higher than in the case @xmath108 despite a strong magnetic field .
to summarize then , we have considered the instability caused by shear in a compressible magnetized gas . to illustrate the main qualitative features of the instability associated to compressibility and shear
, we analyzed a particular case of perturbations that depend on the vertical coordinate alone .
the plane couette flow with a non - vanishing transverse magnetic field turns out to be unstable even in this simplest case .
the necessary condition of instability is @xmath139 , and it can be easily satisfied in laboratory flows . since the shear flow in the presence of a transverse magnetic field always stretches the longitudinal field satisfying the necessary condition ( 35 ) , one can expect that the instability likely operates if @xmath140 is entirely generated by shear and @xmath141 in eq .
we consider this case elsewhere .
the newly found instability is relatively slow : its growth rate reaches @xmath142 and is small compared to the shear timescale , @xmath143 .
however , even this growth rate can be sufficient to generate hydrodynamic motions in many real flows , for example , in astrophysics . basically , the growth rate is larger for non - oscillatory modes which are unstable at relatively not very large @xmath102 .
the growth rate depends on the ratio of the magnetic and gas pressure , being smaller for a low ratio .
the considered instability is related basically to shear and compressible properties of a magnetized gas . in the incompressible limit that corresponds to @xmath144
, we have from eq . ( 26 ) @xmath145 and the instability does not occur for the chosen perturbations .
it can be not the case , however , for perturbations of a more general form which depend also on the @xmath7- or @xmath8-coordinates
. this new instability can be either oscillatory or non - oscillatory , depending on the value of the ratio @xmath146 .
typically , the considered instability is non - oscillatory if @xmath102 is not large and oscillatory in the opposite case .
the critical @xmath102 that determines the transition between oscillatory and non - oscillatory regimes depends strongly on the parameters @xmath129 , @xmath113 , and @xmath137 and can vary within a wide range .
one more important feature of the instability is associated with the dependence on the magnetic field strength .
generally , a sufficiently strong magnetic field can suppress instabilities of a shear flow . on the contrary
, the instability discovered in our study can not be suppressed even in very strong magnetic fields as it is seen from the criterion ( 40 ) .
all this comparison allows us to claim that our analysis demonstrates the presence of the new instability in compressible shear flows .
_ acknowledgments .
_ this research project has been supported by a marie curie transfer of knowledge fellowship of the european community s sixth framework program under contract number mtkd - ct-002995 .
vu thanks also inaf - ossevatorio astrofisico di catania for hospitality . | we study the stability of a compressible magnetic plane couette flow and show that compressibility profoundly alters the stability properties if the magnetic field has a component perpendicular to the direction of flow .
the necessary condition of a newly found instability can be satisfied in a wide variety of flows in laboratory and astrophysical conditions .
the instability can operate even in a very strong magnetic field which entirely suppresses other mhd instabilities .
the growth time of this instability can be rather short and reach @xmath0 shear timescales .
# 1@xmath1#1 |
let @xmath7 be a finite connected graph without loops , but with multiple edges allowed .
the _ adjacency matrix _ of @xmath0 , denoted by @xmath8 , is given by @xmath9 , where @xmath10 is the number of the edges between @xmath11 and @xmath12 in @xmath13 .
the _ laplacian matrix _ of @xmath0 is the matrix @xmath14 where @xmath15 and @xmath16 is the degree of the vertex @xmath11 in @xmath0 . if @xmath1 is a vertex of @xmath0 , the _ reduced laplacian matrix _ ,
denoted by @xmath17 , is the matrix obtained from @xmath18 by removing the row and column @xmath1 .
critical group _ of @xmath0 , denoted by @xmath19 , is the cokernel of @xmath17 , @xmath20 where @xmath21 .
the critical group is an abelian group that is isomorphic to the sandpile group introduced by dhar in @xcite , which generalizes the case of a grid from @xcite .
the critical group has been studied by several authors , see for instance @xcite .
the concept of critical group can be generalized easily to an arbitrary commutative ring with identity @xmath22 .
more precisely , if @xmath23 , then the _ critical module _ of @xmath24 , denoted by @xmath25 , is defined as : @xmath26 given @xmath27 and @xmath28 , we say that @xmath29 are _
@xmath30-equivalent _ , denoted by @xmath31 , if there exist @xmath32 and @xmath33 such that @xmath34 .
when @xmath35 and @xmath36 , then we simply say that @xmath24 and @xmath37 are _ unitary equivalent _ and will be denoted by @xmath38 .
also , if @xmath39 and @xmath40 are the subgroups generated by the elementary matrices , we simply say that @xmath24 and @xmath37 are _ elementary equivalent _ and will be denoted by @xmath41 .
finally , if @xmath42 and @xmath42 , then we simply say that @xmath24 and @xmath37 are equivalent and will be denoted by @xmath43 or @xmath44 if the ring @xmath22 is clear from the context .
usually the critical group of a graph can be described in terms of a diagonal matrix called the smith normal form of @xmath18 .
moreover , is not difficult to see that if @xmath24 and @xmath37 are equivalent , then @xmath45 when the base ring @xmath22 is principal ideal domain ( pid ) , another description of the critical group of a matrix @xmath24 is given by @xmath46 where @xmath47 is the greatest common divisor of all the @xmath48 minors of @xmath24 .
this article is divided in two sections : in the first section , we find diagonal matrices over an arbitrary commutative ring with identity that are equivalent to some matrices that generalize the laplacian matrices of the path , the cycle , and the complete graph . in the second section
we apply the results obtained in the first section in the case when the base ring is the ring of integers and some subrings of matrices .
in particular we are able to calculate the critical group of the @xmath4-cones of the @xmath5-duplications of the path , the cycle , the complete graph , the bipartite complete graph with @xmath4 vertices in each partition , the bipartite complete graph with @xmath6 vertices minus a matching , etc . in the following ,
every multigraph will be connected and will have a distinguished vertex @xmath49 , called the _ sink _ of @xmath50 .
sometimes we will simply write @xmath1 instead of @xmath51 . in @xcite
it can be seen any unexplained term of graph theory .
in this section we will find diagonal matrices that are equivalent to some matrices that generalize the laplacian matrices of the path , the cycle , and the complete graphs .
after that , we will apply these results in order to calculate the critical group for several families of graphs . for
all @xmath52 and @xmath53 , let @xmath54 , @xmath55 , @xmath56 , and @xmath57 where @xmath58 is the identity matrix on order @xmath59 .
since the critical module of a matrix is invariant under equivalency classes , then in order to determine the critical module of a matrix , it is enough to find an equivalent diagonal matrix .
[ prin ] let @xmath53 such that the equation @xmath60 has solution in @xmath22 and @xmath61 polynomials in @xmath62 $ ] that satisfy the recurrence relation @xmath63 with initial values @xmath64 and @xmath65 . then ( i ) : : @xmath66 for all @xmath52 , ( ii ) : : @xmath67 for all @xmath52 , ( iii ) : : @xmath68 for all @xmath52 , and ( iv ) : : @xmath69 for all @xmath70 , where @xmath71 @xmath72 for all @xmath73 and @xmath74 , let @xmath75 where @xmath76 for all @xmath77 and @xmath78 is the matrix with all the entries equal to @xmath79 . note that @xmath80 now we will prove the following statement : [ reduction1 ] for all @xmath73 and @xmath74 @xmath81 let @xmath82 be a solution of the equation @xmath60
moreover , for all @xmath83 , let @xmath84 that is , @xmath85 and @xmath86 are a solution of the equation @xmath87 . since @xmath88 and @xmath89 , then @xmath90 .
applying claim [ reduction1 ] , we get that @xmath91 .
@xmath92 for all @xmath93 , @xmath94 , let @xmath95 also , let @xmath96 that is , @xmath85 and @xmath86 are a solution of the equation @xmath97 . since @xmath98 for all @xmath73 and @xmath99 , then @xmath100 for all @xmath73 .
thus @xmath101 therefore , @xmath102 because @xmath103 @xmath104 we begin proving the following statement : [ claimk ] if @xmath52 , then @xmath105 .
it turns out because @xmath106 where @xmath107 are elementary matrices . on the other hand , @xmath108 because @xmath109 and @xmath110 .
@xmath111 for all @xmath112 and @xmath83 , let @xmath113 and @xmath114 also , let @xmath115 since @xmath116 for all @xmath117 , @xmath118 for all @xmath119 , and @xmath120 , then @xmath121 moreover , since @xmath122 , then @xmath123 finally , using similar reductions we get that @xmath124 and @xmath125 in ( * ? ? ?
44 ) , a simpler proof of theorem [ prin ] ( iii ) when @xmath22 is a principal ideal domain can be found .
the next lemma give us some useful properties of the polynomials @xmath61 . if @xmath126 , then ( i ) : : @xmath127 , ( ii ) : : @xmath128 , ( iii ) : : @xmath129 , ( iv ) : : @xmath130 .
it follows using induction on @xmath59 .
note that , @xmath131 .
if @xmath22 is a principal ideal domain and @xmath132 , then the equation @xmath60 has a solution if and only if @xmath133 . [ prinpid ] let @xmath22 be a principal ideal domain , @xmath132 with @xmath134 , @xmath135 , and @xmath136 . then ( i ) : : @xmath137 for all @xmath52 , ( ii ) : : @xmath138 for all @xmath52 , ( iii ) : : @xmath139 for all @xmath52 , and ( iv ) : : @xmath140 for all @xmath70 , where @xmath141 let @xmath142 be either one of the matrices @xmath143 , @xmath144 , @xmath145 , or @xmath146 , then @xmath147 . on the other hand , since @xmath134 if and only if @xmath148 if and only if the equation @xmath149 has solution in @xmath22 .
then , we get the result applying theorem [ prin ] to @xmath150 .
in this section we will apply the equivalences of the matrices @xmath145 , @xmath146 , @xmath144 and @xmath143 obtained in the previous section to the cases when @xmath151 and @xmath152 are in the ring of integers and in the subring of matrices of the form @xmath145 where @xmath153 are in a commutative ring with identity @xmath22 . at this point
we need to introduce some definitions . given a simple graph @xmath0 and a natural number @xmath154 , the @xmath5-duplication of @xmath0 , denoted by @xmath155 , is the multigraph obtained from @xmath0 when we replace every edge of @xmath0 by @xmath5 parallel edges .
note that @xmath156 for any graph @xmath0 .
= [ inner sep=0pt , minimum width=4.5pt ] ( 0,0 ) + ( 45:1 ) node[draw , circle , fill = gray ] ( v1 ) + ( 135:1 ) node[draw , circle , fill = gray ] ( v2 ) + ( 225:1 ) node[draw , circle , fill = gray ] ( v3 ) + ( 315:1 ) node[draw , circle , fill = gray ] ( v4 ) ( v1 ) to ( v2 ) to ( v3 ) to ( v4 ) to ( v1 ) ( v1)+(0,0.25 ) node @xmath157 ( v2)+(0,0.25 ) node @xmath158 ( v3)+(0,-0.3 ) node @xmath159 ( v4)+(0,-0.3 ) node @xmath160 ( 0,0 ) node @xmath161 ; ( 5,0 ) + ( 45:1 ) node[draw , circle , fill = gray ] ( v1 ) + ( 135:1 ) node[draw , circle , fill = gray ] ( v2 ) + ( 225:1 ) node[draw , circle , fill = gray ] ( v3 ) + ( 315:1 ) node[draw , circle , fill = gray ] ( v4 ) ( v1 ) .. controls + ( 150:0.5 ) and + ( 30:0.5 ) .. ( v2 ) ( v1 ) .. controls + ( 210:0.5 ) and + ( -30:0.5 ) .. ( v2 ) ( v2 ) .. controls + ( 240:0.5 ) and + ( 120:0.5 ) .. ( v3 ) ( v2 ) .. controls + ( 300:0.5 ) and + ( 60:0.5 ) .. ( v3 ) ( v4 ) .. controls + ( 150:0.5 ) and + ( 30:0.5 ) .. ( v3 ) ( v4 ) .. controls + ( 210:0.5 ) and + ( -30:0.5 ) .. ( v3 ) ( v1 ) .. controls + ( 240:0.5 ) and + ( 120:0.5 ) .. ( v4 ) ( v1 ) .. controls + ( 300:0.5 ) and + ( 60:0.5 ) .. ( v4 ) ( v1)+(0,0.25 ) node @xmath157 ( v2)+(0,0.25 ) node @xmath158 ( v3)+(0,-0.3 ) node @xmath159 ( v4)+(0,-0.3 ) node @xmath160 + ( -0.7,0.7 ) node @xmath162 ; given a graph @xmath0 and a natural number @xmath163 , the @xmath163-cone of @xmath0 , denoted by @xmath164 , is the multigraph obtained from @xmath0 by adding a new vertex @xmath1 and adding @xmath163 parallel edges between @xmath1 and all the vertices of @xmath0 .
note that @xmath165 for any graph @xmath0 .
= [ inner sep=0pt , minimum width=4.5pt ] ( 0,0 ) + ( 45:1 ) node[draw , circle , fill = gray ] ( v1 ) + ( 135:1 ) node[draw , circle , fill = gray ] ( v2 ) + ( 225:1 ) node[draw , circle , fill = gray ] ( v3 ) + ( 315:1 ) node[draw , circle , fill = gray ] ( v4 ) ( v1 ) to ( v2 ) to ( v3 ) to ( v4 ) to ( v1 ) ( v1)+(0,0.25 ) node @xmath157 ( v2)+(0,0.25 ) node @xmath158 ( v3)+(0,-0.3 ) node @xmath159 ( v4)+(0,-0.3 ) node @xmath160 + ( -0.7,-0.5 ) node @xmath161 ; ( 5,0 ) + ( 45:1 ) node[draw , circle , fill = gray ] ( v1 ) + ( 135:1 ) node[draw , circle , fill = gray ] ( v2 ) + ( 225:1 ) node[draw , circle , fill = gray ] ( v3 ) + ( 315:1 ) node[draw , circle , fill = gray ] ( v4 ) + ( 0:0 ) node[draw , circle , fill = gray ] ( s ) ( v1 ) to ( v2 ) to ( v3 ) to ( v4 ) to ( v1 ) ( v1 ) to ( v3 ) ( v2 ) to ( v4 ) ( s)+(0,0.2 ) node @xmath1 ( v1)+(0,0.25 ) node @xmath157 ( v2)+(0,0.25 ) node @xmath158 ( v3)+(0,-0.3 ) node @xmath159 ( v4)+(0,-0.3 ) node @xmath160 + ( -0.7,-0.5 ) node @xmath166 ; the most direct application of the results obtained in the first section is when the base ring @xmath22 is the ring of integers and the matrices are the reduced laplacian matrix of the @xmath59-cone of the thick path , the thick cycle and the thick complete graph .
( 0,0 ) + ( 0,0 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( s ) + ( -1.8,-1.5 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v1 ) + ( -0.9,-1.5 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v2 ) + ( 0.9,-1.5 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v3 ) + ( 1.8,-1.5 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v4 ) + ( -0.3,-1.5 ) node[draw , circle , fill = black , inner sep=0pt , minimum width=1.6pt ] + ( 0,-1.5 ) node[draw , circle , fill = black , inner sep=0pt , minimum width=1.6pt ] + ( 0.3,-1.5 ) node[draw , circle , fill = black , inner sep=0pt , minimum width=1.6pt ] ( s ) to ( v1 ) ( s ) to ( v2 ) ( s ) to ( v3 ) ( s ) to ( v4 ) ( v1 ) to ( v2 ) ( v3 ) to ( v4 ) ( s)+(0,+0.3 ) node @xmath1 ( v1)+(0,-0.3 ) node @xmath157 ( v2)+(0,-0.3 ) node @xmath158 ( v3)+(0,-0.3 ) node @xmath168 ( v4)+(0.1,-0.3 ) node @xmath169 ( v1)+(0.65,0.8 ) node @xmath4 ( v1)+(1.55,0.6 ) node @xmath4 ( v1)+(2.1,0.6 ) node @xmath4 ( v1)+(2.95,0.8 ) node @xmath4 ( v1)+(0.6,0.2 ) node @xmath5 ( v3)+(0.4,0.2 ) node @xmath5 ; ( 0,0 ) + ( 0,0 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( s ) + ( 60:1.4 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v1 ) + ( 0:1.4 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v2 ) + ( -60:1.4 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v3 ) + ( -120:1.4 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v4 ) + ( -180:1.4 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v5 ) + ( -240:1.4 ) node[draw , circle , fill = gray , inner sep=0pt , minimum width=4pt ] ( v6 ) + ( 80:1.4 ) node[draw , circle , fill = black , inner sep=0pt , minimum width=1.6pt ] + ( 90:1.4 ) node[draw , circle , fill = black , inner sep=0pt , minimum width=1.6pt ] + ( 100:1.4 ) node[draw , circle , fill = black , inner sep=0pt , minimum width=1.6pt ] ( s ) to ( v1 ) ( s ) to ( v2 ) ( s ) to ( v3 ) ( s ) to ( v4 ) ( s ) to ( v5 ) ( s ) to ( v6 ) ( v1 ) to ( v2 ) to ( v3 ) to ( v4 ) to ( v5 ) to ( v6 ) ( s)+(0,0.3 ) node @xmath1 ( v1)+(0.2,0.2 ) node @xmath159 ( v2)+(0.3,0 ) node @xmath158 ( v3)+(0,-0.3 ) node @xmath176 ( v4)+(0.1,-0.3 ) node @xmath177 ( v5)+(-0.3,-0.2 ) node @xmath168 ( v6)+(-0.2,0.2 ) node @xmath178 ( s)+(40:0.7 ) node @xmath4 ( s)+(-15:0.7 ) node @xmath4 ( s)+(-80:0.7 ) node @xmath4 ( s)+(-140:0.7 ) node @xmath4 ( s)+(-195:0.7 ) node @xmath4 ( s)+(-260:0.7 ) node @xmath4 ( s)+(30:1.4 ) node @xmath5 ( s)+(-30:1.4 ) node @xmath5 ( s)+(-90:1.4 ) node @xmath5 ( s)+(-150:1.4 ) node @xmath5 ( s)+(-210:1.4 ) node @xmath5 ; [ completagorda ] for all @xmath167 , @xmath154 , and @xmath70 , let @xmath188 be the @xmath4-cone of the thick complete graph where all the edges have multiplicity equal to @xmath5 .
then @xmath189 where @xmath172 .
note that @xmath193 is a commutative ring with identity because @xmath210 for all @xmath211 . on the other hand , since @xmath212 , then @xmath193 is not a principal ideal domain because it has zero divisors .
now , we will apply theorem [ prin ] @xmath104 when the base ring is the subring of matrices with entries in @xmath193 to obtain a theorem that will be a powerful tool to calculate the critical group of several graphs . given @xmath216 and @xmath217 ,
let @xmath218 let @xmath221 and @xmath222 be as in claim [ claimk ] , then @xmath223 are elementary matrices and @xmath224 \oplus \left ( \begin{array}{cc } a&b\\ 0 & a+mb \end{array } \right),\end{aligned}\ ] ] where @xmath225 . in the last part of this paper , we will use theorem [ completamatrices ] to find the critical group of some graphs whose laplacian matrix is given by @xmath219 for @xmath226 .
the simplest case is when @xmath227 .
if @xmath219 for @xmath228 is the laplacian matrix of a graph , then the vertex set of the graph can be partitioned in two sets and the incidence structure between these sets is given by a matrix in @xmath229 . in this sense
, we will define the following families of graphs : note that @xmath238 is the bipartite complete graph with @xmath4 vertices in each partition and @xmath239 is the bipartite complete graph with @xmath6 vertices minus a matching .
the laplacian matrix of all these graphs can be represented by @xmath219 for some two by two matrices @xmath240 and @xmath241 .
( 0,-0.8 ) ( 0,0.8 ) arc ( 0:180:0.3 )
( -0.6,-0.8 ) arc ( 180:360:0.3 ) ; ( 0,0 ) + ( 0,-0.8 ) ( 0,0.8 ) arc ( 0:180:0.3 )
( -0.6,-0.8 ) arc ( 180:360:0.3 ) + ( -0.3,1.6 ) node[inner sep=0pt , minimum width=3.5pt , draw , circle , fill = gray ] ( v1 ) + ( -0.3,1.2 ) node[inner sep=0pt , minimum width=3.5pt , draw , circle , fill = gray ] ( v2 ) + ( -0.3,0.95 ) node[inner sep=0pt , minimum width=1pt , draw , circle , fill = black ] ( ) + ( -0.3,0.8 ) node[inner sep=0pt , minimum width=1pt , draw , circle , fill = black ] ( ) + ( -0.3,0.65 ) node[inner sep=0pt , minimum width=1pt , draw , circle , fill = black ] ( ) + ( -0.3,0.4 ) node[inner sep=0pt , minimum width=3.5pt , draw , circle , fill = gray ] ( v3 ) + ( -0.3,0 ) node[inner sep=0pt , minimum width=3.5pt , draw , circle , fill = gray ] ( v4 ) + ( 1.5,1.6 ) node[inner sep=0pt , minimum width=3.5pt , draw , circle , fill = gray ] ( v5 ) + ( 1.5,1.2 ) node[inner sep=0pt , minimum width=3.5pt , draw , circle , fill = gray ] ( v6 ) + ( 1.5,0.4 ) node[inner sep=0pt , minimum width=3.5pt , draw , circle , fill = gray ] ( v7 ) + ( 1.5,0 ) node[inner sep=0pt , minimum width=3.5pt , draw , circle , fill = gray ] ( v8 ) ( v2 ) .. controls + ( 252:0.4 ) and + ( 108:0.4 ) .. ( v3 ) ( v1 ) .. controls + ( 246:0.6 ) and + ( 114:0.6 ) .. ( v3 ) ( v2 ) .. controls + ( 246:0.6 ) and + ( 114:0.6 ) .. ( v4 ) ( v1 ) to ( v2 ) ( v3 ) to ( v4 ) ( v1 ) to ( v5 ) ( v2 ) to ( v6 ) ( v3 ) to ( v7 ) ( v4 ) to ( v8 ) + ( -2.4,0.8 ) node @xmath243 + ( -0.8,0.9 ) node @xmath244 + ( -2.3,-0.1 ) node @xmath245 + ( 0.3,0 ) ( 1.8,0.8 ) arc ( 0:180:0.3 )
( 1.2,-0.8 ) arc ( 180:360:0.3 ) + ( 0.2,-0.1 ) node @xmath236 + ( -5,0 ) node ; ( 3.07,0 ) node ; before using theorem [ completamatrices ] in order to calculate the critical group of the @xmath59-cones of graphs @xmath246 , we will introduce a theorem that gives an equivalent matrix of @xmath246 when the base ring is a general commutative ring with identity .
@xmath72 since @xmath252 for @xmath253 and @xmath254 , then by theorem [ completamatrices ] @xmath255 \oplus \left ( \begin{array}{cccc } a & 0 & 0&b\\ 0 & a & b&0\\ 0 & 0 & a&mb\\ 0 & 0 & mb&a \end{array } \right).\ ] ] moreover , since @xmath256 and @xmath257 then @xmath258 hence @xmath259 @xmath92 since @xmath260 for @xmath261 and @xmath254 , then by theorem [ completamatrices ] @xmath262 \oplus \left ( \begin{array}{cccc } a & -b & 0&b\\ -b & a & b&0\\ 0 & 0 & a&(m-1)b\\ 0 & 0 & ( m-1)b&a \end{array } \right).\ ] ] moreover , since @xmath263 @xmath264 @xmath265 and @xmath266 then @xmath267 @xmath104 since @xmath268 for @xmath269 and @xmath270 , then by theorem [ completamatrices ] @xmath271 \oplus \left ( \begin{array}{cccc } a - b & -b & b&b\\ -b & a - b & b&b\\ 0 & 0 & a+(m-1)b&(m-1)b\\ 0 & 0 & ( m-1)b&a+(m-1)b \end{array } \right).\ ] ] moreover , since @xmath272 @xmath273 and @xmath274 then the result is followed .
@xmath111 since @xmath275 for @xmath276 and @xmath277 , then by theorem [ completamatrices ] @xmath278 \oplus \left ( \begin{array}{cccc } a - b & b & b&0\\ b & a - b & 0&b\\ 0 & 0 & a+(m-1)b&b\\ 0 & 0 & b&a+(m-1)b \end{array } \right).\ ] ] moreover , since @xmath279 and @xmath280 then the result is followed .
when a graph @xmath0 is not regular , then there is not a straightforward way to define their matrix @xmath284 .
thus , we will define @xmath285 as @xmath286 , @xmath287 as @xmath288 , and @xmath289 as @xmath290 .
now , we have the following equivalent matrices of @xmath285 , @xmath287 , and @xmath289 .
@xmath72 since @xmath294 for @xmath295 and @xmath296 , then by theorem [ completamatrices ] @xmath297 \oplus \left ( \begin{array}{cccc } a&0&0&b\\
0&2a&b&b\\ 0&0&a&mb\\ 0&0&mb&2a+mb \end{array } \right).\ ] ] moreover , @xmath298 and @xmath299 @xmath92 since @xmath300 for @xmath301 and @xmath302 , then by theorem [ completamatrices ] @xmath303 \oplus \left ( \begin{array}{cccc } a&b&0&0\\ b&(m+1)a&0&b\\ 0&0&a&b\\ 0&0&b&(m+1)a+mb \end{array } \right)\ ] ] moreover , @xmath304 and @xmath305 @xmath104 since @xmath306 for @xmath307 and @xmath296 , then by theorem [ completamatrices ] @xmath308 \oplus \left(\begin{array}{cccc}a&-b&0&b\\-b&2a&b&b\\0&0&a&(m-1)b \\0&0&(m-1)b&2a+mb \end{array}\right).\ ] ] moreover , @xmath309 @xmath310 and @xmath311 since @xmath316 and @xmath317 with @xmath318 , then applying corollary [ bipartitacompleta ] @xmath72 to @xmath319 we get that @xmath320 on the other hand , @xmath321 where @xmath315 and we get the result . note that @xmath238 is the complete bipartite graph with @xmath4 vertices in each partition .
lorenzini in @xcite calculated that @xmath322 which agrees with the corollary [ bipartitacompletag1 ] for @xmath323 and @xmath324 .
also note that @xmath325 has @xmath326 invariant factors different to @xmath327 .
on the other hand , it is not difficult to see that @xmath360/r % \end{array } % \right ) % = \left ( \begin{array}{cc
} st / r&-l^2(n+t)/r^2\\ 0 & n(n+2l)/r \end{array } \right ) \sim_u u \oplus stn(n+2l)/ur^2,\ ] ] where @xmath351 and @xmath352 .
note that @xmath370 is the graph @xmath371 .
in general the expression for @xmath372 given in corollary [ bipartitacompletag5 ] does not give us the invariant factors of @xmath372 . also note that @xmath373 has @xmath326 invariant factors different to @xmath327 . | given a graph @xmath0 with a distinguished vertex @xmath1 , the critical group of @xmath2 is the cokernel of their reduced laplacian matrix @xmath3 . in this article
we generalize the concept of the critical group to the cokernel of any matrix with entries in a commutative ring with identity . in this article
we find diagonal matrices that are equivalent to some matrices that generalize the reduced laplacian matrix of the path , the cycle , and the complete graph over an arbitrary commutative ring with identity .
we are mainly interested in those cases when the base ring is the ring of integers and some subrings of matrices . using these equivalent diagonal matrices
we calculate the critical group of the @xmath4-cones of the @xmath5-duplications of the path , the cycle , and the complete graph . also , as byproduct , we calculate the critical group of another matrices , as the @xmath4-cones of the @xmath5-duplication of the bipartite complete graph with @xmath4 vertices in each partition , the bipartite complete graph with @xmath6 vertices minus a matching . |
the origin of ray bursts is still unknown and is currently the subject of a `` great debate '' in the astronomical community .
do the bursts have a galactic origin @xcite or are they cosmological @xcite ? and what is their distance scale ? in this paper ,
i do not attempt to answer the first question , but rather , i show that _ if _ one assumes that ray bursts are cosmological in origin , one can begin to answer the second question and place a constraint on the distance scale to the bursts .
this is because cosmological bursts are expected to trace the large scale structure of luminous matter in the universe @xcite .
the constraint comes from comparing the _ expected _ clustering pattern of bursts on the sky which will depend on their distance scale because of projection effects with that _ actually observed_. the observed angular distribution is in fact quite isotropic @xcite ; hence , only a lower limit to the distance scale can be placed because a sufficiently large distance will always lead to a sufficiently isotropic distribution on the sky . here
i use a powerful new likelihood method @xcite , which i had previously developed to analyze repeating of ray bursts in the batse 1b and 2b catalogs @xcite , to compare the observed counts in cells distribution in the new batse 3b catalog @xcite with that expected for bursts at cosmological distances . here
i will assume for simplicity that @xmath5 and @xmath6 , and that the large scale structure clustering pattern is constant in comoving coordinates .
the results are in fact insensitive to these assumptions because of the small redshifts that are involved .
i follow the usual convention and take @xmath7 to be the hubble constant in units of 100 km s@xmath8 mpc@xmath8 .
let @xmath9 be a large number of circular cells , each centered on a random position on the sky .
each cell is of fixed solid angle size @xmath10 , where @xmath11 is the angular radius of the cell .
i set the number of cells to be such that any part of the sky is covered , on average , by one cell ; hence , @xmath12 . let @xmath13 to be the number of these cells having @xmath14 ray bursts in them , out of the @xmath15 in the batse 3b catalog , where @xmath16 i then define the observed counts in cells distribution , @xmath17 , as the probability that a randomly chosen cell of size @xmath18 has @xmath14 bursts in it .
the counts in cells distribution contains information about clustering of ray bursts on scales comparable to the angular size @xmath11 of the cell .
i now define @xmath19 to be the counts in cells distribution that is expected if ray bursts are cosmological in origin and trace the large scale structure of luminous matter in the universe .
this expected distribution depends on only one unknown parameter , the effective distance @xmath20 to ray bursts ( which i define below ) , because the angular clustering pattern of bursts on the sky will depend by projection on this distance .
the _ likelihood _ @xmath21 measures how likely it is that the observed counts in cells distribution @xmath22 is drawn from the expected distribution @xmath19 . since @xmath19 depends on the unknown effective distance @xmath20 to ray bursts , the likelihood is really a measure of how likely a given value of @xmath20 is .
i find that @xcite : @xmath23 now the cumulative @xmath24 distribution of ray bursts seen by batse begins to roll over from a @xmath25 power
law for bursts fainter than @xmath26 .
since this is many times above threshold , it suggests that batse sees most of the source distribution and that this distribution is not spatially homogeneous @xcite .
i define @xmath20 as the comoving distance beyond which the source density drops appreciably .
it is not the distance to the very dimmest burst in the batse catalog , but rather the typical distance to most of the dim bursts in the sample ; thus , @xmath20 is the _ effective distance _ to the `` edge '' of the source distribution in the batse catalog .
i take the power spectrum which characterizes the large
scale clustering of ray burst sources to be the same as that determined from a redshift survey of radio galaxies @xcite .
this power spectrum is characteristic of moderately rich environments , and is intermediate between that of ordinary galaxies and clusters .
because the exact bias factor relating the clustering of ray burst sources to that of luminous matter is unknown , such an intermediate ansatz is reasonable . in any case
, the resultant distance limit depends only weakly on the bias factor ( roughly as the square root ) .
knowledge of the power spectrum permits a calculation of the expected angular clustering pattern , the expected counts in cells distribution @xmath19 , and finally the likelihood @xmath27 [ from equation ( 1 ) ] , all as a function of the effective distance @xmath20 to ray bursts @xcite .
i have included the smearing due to finite positional errors on the clustering on small scales @xcite .
indeed , each burst in the batse catalog is assigned a positional uncertainty @xmath28 corresponding to a @xmath29 confidence that the true burst position is within an angle @xmath28 to the position listed in the catalog .
i have chosen the cell size @xmath11 in order to maximize the sensitivity of detection , or signal to noise , given the strength of the signal expected . for a sample of 1122 bursts ( the total number of bursts in the batse 3b catalog ) with positional smearing of @xmath30 ( the median value in the 3b catalog ) , the signal to noise is maximized when
cells of @xmath31 are used @xcite .
figure 1 shows the likelihood of the batse 3b catalog data as a function of the effective comoving distance @xmath20 , calculated using cells of size @xmath31 .
the likelihood is normalized to that expected for an isotropic distribution on the sky . at large values of @xmath20 ( the maximum value allowed is @xmath32 mpc , the size of the horizon in a closed universe )
, the likelihood goes to unity , because by projection a sufficiently large distance will always lead to an isotropic distribution on the sky .
note also that there is no value of @xmath20 for which the likelihood is greater than 1 ; thus , the maximum likelihood value for @xmath20 is @xmath33 and the 3b data are consistent with isotropy . the solid line in fig .
1 shows the likelihood for a positional smearing of @xmath30 , corresponding to the median value in the 3b catalog . to illustrate the dependence of these results on positional errors ,
i also show ( dashed line ) the results for a larger positional smearing value quoted in the 3b catalog .
their best fit model gives a median positional error of @xmath34 .
] of @xmath35 ( with cells of size @xmath36 to maximize signal to noise ) .
small values of the effective comoving distance to ray bursts are unlikely , according to fig .
1 : i find , at the 95% confidence level , that for the 3b median positional error of @xmath30 , @xmath20 must be greater than @xmath0 mpc , corresponding to a redshift @xmath1 .
if the positional errors are larger than quoted and are better characterized by @xmath35 , these results are only slightly weakened : at the 95% confidence level , @xmath20 must be greater than @xmath37 mpc , corresponding to a redshift @xmath38 .
these limits are not sensitive to earlier assumptions on cosmology and clustering evolution since these only become important at higher redshifts . they are also conservative limits in that a constant median value for the positional errors was used , rather than the entire distribution of errors .
this is because the bright bursts , which ostensibly are nearer to us , are more clustered and are responsible for the bulk of the expected signal , but in fact have smaller errors than the median value .
the faint bursts , which are far away , are hardly clustered to begin with ( even before smearing ) , but have errors larger than the median value .
hence the expected clustering pattern has been smeared more by using a constant median value ( this permits a simpler calculation ) than by smearing using the entire distribution of errors . so the counts in cells statistic has been weakened somewhat and thus the quoted lower limits are in fact conservative .
if ray bursts are cosmological and trace the large scale structure of luminous matter in the universe , and their positional errors are as quoted in the 3b catalog , then the lack of any angular clustering in the data implies that the observed distance to the `` edge '' of the burst distribution must be farther than @xmath0 mpc .
since there are 1122 bursts in the catalog , an effective limit on the _ nearest _ burst to us can be placed by convoluting the likelihood as a function of @xmath20 ( fig .
1 ) with the nearest neighbor distribution of 1122 bursts inside a sphere of radius @xmath20 .
i find that the nearest burst must be farther than @xmath2 mpc at the 95% confidence level , and farther than @xmath39 mpc at the 99.9% level . at this level of confidence , then , none of the bursts that have been observed by batse are in nearby galaxies . a signature from the coma cluster or the `` great wall '' ( @xmath40 mpc ) is not likely to be seen in the data at present , since only a few bursts could have originated from these distances .
the median distance to the nearest burst is @xmath3 mpc .
since the brightest burst in the 3b catalog has a fluence of @xmath41 in rays , this implies that the total energy released in rays during a burst event is of order @xmath4 erg .
as the number of observed ray bursts keeps increasing , the distance limit will improve .
in fact , with 3000 burst locations , the clustering of bursts might just be detectable @xcite and would provide compelling evidence for a cosmological origin . if it is not detected , the redshift to the `` edge '' of the burst distribution would be put at @xmath42 or beyond .
i would like to acknowledge useful discussions with carlo graziani , don lamb , cole miller and bob nichol .
this research was supported in part by nasa through the _ compton _ fellowship program grant nag 5 - 2660 , grant nag 5 - 2868 , and contract nasw-4690 . | if ray bursts have a cosmological origin , the sources are expected to trace the large scale structure of luminous matter in the universe .
i use a new likelihood method that compares the counts in cells distribution of ray bursts in the batse 3b catalog with that expected from the known large
scale structure of the universe , in order to place a constraint on the distance scale to cosmological bursts .
i find , at the 95% confidence level , that the comoving distance to the `` edge '' of the burst distribution is greater than @xmath0 mpc ( @xmath1 ) , and that the nearest burst is farther than @xmath2 mpc .
the median distance to the nearest burst is @xmath3 mpc , implying that the total energy released in rays during a burst event is of order @xmath4 erg .
none of the bursts that have been observed by batse are in nearby galaxies , nor is a signature from the coma cluster or the `` great wall '' likely to be seen in the data at present . |
linked cluster expansions ( lces ) have a long tradition in statistical physics . originally applied to classical fluids , later to magnetic systems ( @xcite,@xcite,@xcite and references therein ) , they were generalized to applications in particle physics in the eighties @xcite . there
they have been used to study the continuum limit of a lattice @xmath0 field theory in 4 dimensions at zero temperature . in @xcite
they were further generalized to field theories at finite temperature , simultaneously the highest order in the expansion parameter was increased to 18 . usually the analytic expansions are obtained as graphical expansions . because of the progress in computer facilities and the development of efficient algorithms for generating the graphs , it is nowadays possible to handle of the order of billions of graphs .
the whole range from high temperatures down to the critical region becomes available , and thermodynamic quantities like critical indices and critical temperatures are determined with high precision ( the precision is comparable or even better than in corresponding high quality monte carlo results ) @xcite-@xcite .
an extension of lces to a finite volume in combination with a high order in the expansion parameter turned out to be a particularly powerful tool for investigating the phase structure of systems with first and second order transitions by means of a finite size scaling analysis @xcite .
linked cluster expansions are series expansions of the free energy and connected correlation functions about an ultralocal , decoupled theory in terms of a hopping parameter @xmath1 .
the corresponding graphical representation is a sum in terms of connected graphs .
the value of @xmath1 parametrizes the strength of interactions between fields at different lattice sites .
usually they are chosen as nearest neighbours , but also more general , less local couplings lead to convergent expansions ( @xcite and references therein ) under appropriate conditions on the decay of the interactions .
in contrast to the ultralocal terms of a generic interaction we will sometimes refer to hopping terms as non - ultralocal . in this paper we develop dynamical linked cluster expansions ( dlces ) . these are linked cluster expansions with hopping parameter terms that are endowed with their own dynamics .
such systems are realized in spin glasses or partially annealed neural networks with ( fast ) spins and ( slow ) interactions @xcite-@xcite .
they also occur in variational estimates for su(n)-gauge - higgs systems as we will show later in this paper . like lces
they are expected to converge for a large class of interactions .
formally dlces amount to a generalization of an expansion scheme from 2-point to point - link - point - interactions .
these are interactions between fields associated with two points and with one pair of points called link .
the points and links are not necessarily embedded on a lattice , and the links need not be restricted to nearest neighbours .
we develop a new theory in which a generalized notion of connectivity plays a central role .
standard notions of equivalence classes of graphs like 1-line irreducible and 1-vertex irreducible graphs have to be generalized , and new notions like 1-multiple - line irreducible graphs must be defined in order to give a systematic classification .
we describe algorithms for generating these classes .
this is essential , since the number of graphs rapidly increases with the order in the expansion parameter .
it typically exceeds the number of lce - graphs by more than an order of magnitude .
the paper is organized as follows . in section 2
we specify the models that admit a dlce .
we introduce and explain the idea behind the abstract notions of theory .
detailed definitions of , related notions and the computation of weights are given in section 3 .
section 4 treats the issue of renormalization in the sense of suitable resummations of graphs .
algorithms for their generation are described in section 5 .
applications to spin glasses and ( partially annealed ) neural networks are indicated in section 6 .
there it is of particular interest that dlces allow for the possibility of avoiding the replica trick . in section 7
we present results for the transition line of an su(2 ) higgs model .
these are obtained by dlces applied to gap equations that follow from convexity estimates of the free energy density .
the results are in good agreement with corresponding high precision monte carlo results @xcite .
section 8 contains a summary and outlook .
the reader who is more interested in our actual applications of dlces may skip sections 3 - 5 in a first reading .
these sections are of interest from a systematic point of view and essential if he is interested in applying dlces .
in this section we first specify the class of models for which we develop dynamical linked cluster expansions .
next we illustrate some basic notions of multiple - line graph theory , in particular the need for a new notion of connectivity .
by @xmath2 we denote a finite or infinite set of points .
one of its realizations is a hypercubic lattice in @xmath3 dimensions , infinite or finite in some directions with the topology of a torus .
@xmath4 denotes the set of unordered pairs @xmath5 of sites @xmath6 , @xmath7 , also called unoriented links , and @xmath8 a subset of @xmath4 .
we consider physical systems with a partition function of the generic form [ 2.zgen ] & & z(h , i , v ) + & & = u , with measures [ 2 m ] = _
x_0 d(x ) , u = _ l _ 1 du(l ) and action [ 2act ] s(,u , v ) = _ x _ s^((x ) ) + _ l _ 1s^1(u(l ) ) - _ x , y _ v(x , y)(x)u(x , y)(y ) , with non - ultralocal couplings & & v(x , y ) = v(y , x ) = 0 , + & & v(x , x ) = 0 .
for later convenience the normalization via @xmath9 is chosen such that @xmath10=0 $ ] . the field @xmath11 is associated with the sites @xmath12 and the field @xmath13 lives on the links @xmath14 , and we write @xmath15 for @xmath16 . for definiteness and for simplicity of the notation here we assume @xmath17 and @xmath18 . in our actual applications
@xmath19 is a 4-component scalar field and @xmath20 an su(2)-valued gauge field , or @xmath19 and @xmath20 both are ising spins , or the @xmath19s are the ( fast ) ising spins and the @xmath20s @xmath21 the ( slow ) interactions .
the action is split into two ultralocal parts , @xmath22 depending on fields on single sites , and @xmath23 depending on fields on single links @xmath24 . for simplicity
we choose @xmath23 as the same function for all links @xmath14 .
we may identify @xmath8 with the support of @xmath25 , _ 1 = \
{ l=(x , y ) v(x , y ) = 0 } .
the support of @xmath26 need not be restricted to nearest neighbours , also the precise form of @xmath22 and @xmath23 does not matter for the generic description of dlces , @xmath22 and @xmath23 can be any polynomials in @xmath19 and @xmath20 , respectively . the only restriction is the existence of the partition function . in one of our applications
@xmath22 will be a @xmath27-type theory with o(n ) symmetry , @xmath23 consists of a term linear or quadratic in @xmath20 .
note that the interaction term @xmath28 contains a point - link - point - interaction and generalizes the 2-point - interactions @xmath29 of usual hopping parameter expansions .
the effective coupling of the @xmath19 fields has its own dynamics governed by @xmath30 , the reason why we have called our new expansion scheme _
dynamical _ lce .
dynamical linked cluster expansions are induced from a taylor expansion of @xmath31 about @xmath32 , the limit of a completely decoupled system .
we want to express the series for @xmath33 in terms of connected graphs .
let us consider the generating equation [ 2.genequation ] w / v(xy ) & = & 1/2 < ( x ) u(x , y ) ( y ) > + & = & 1/2 ( w_h(x ) i(x , y ) h(y ) + w_h(x)h(y)w_i(x , y ) + & + & w_h(x)i(x , y)w_h(y ) + w_i(x , y)h(y)w_h(x ) + & + & w_h(x)w_h(y)w_i(x , y ) ) . here @xmath34 denotes the normalized expectation value w.r.t
. the partition function of eq .
( [ 2.zgen ] ) .
subscripts @xmath35 and @xmath36 denote the derivatives of @xmath33 w.r.t .
@xmath35 and @xmath37 , respectively .
next we would like to represent the right hand side of eq .
( [ 2.genequation ] ) in terms of connected graphs .
once we have such a representation for the first derivative of @xmath33 w.r.t .
@xmath25 , grapical expansions for the higher derivatives can be traced back to the first one . for each @xmath33 in eq .
( [ 2.genequation ] ) we draw a shaded bubble , for each derivative w.r.t . @xmath38 a solid line , called a @xmath19-line , with endpoint vertex @xmath39 , and for each derivative w.r.t . @xmath40 a dashed line , called a @xmath20-line , with link label @xmath16 .
the main graphical constituents are shown in fig . 1 .
two @xmath19-lines with endpoints @xmath39 and @xmath41 are then joined by means of a dashed @xmath20-line with label @xmath42 , if the link @xmath42 has @xmath39 and @xmath41 as its endpoints , i.e. @xmath16 . according to these rules eq .
( [ 2.genequation ] ) , multiplied by @xmath26 and summed over @xmath39 and @xmath41 , is represented by fig . 2 .
note that , because of the taylor operation , each solid line from @xmath39 to @xmath41 carries a factor @xmath26 . since the actual need for a new type of connectivity is not quite obvious from fig . 2 , because eq . ( [ 2.genequation ] ) does not contain higher than first order derivatives w.r.t .
@xmath40 , let us consider a term ( 15.0,3.0 ) ( 15.0,7.0 ) w_h(x)w_h(y)w_h(r)w_h(s ) w_i(x , y)i(r , s ) occurring in the second derivative of @xmath33 w.r.t .
@xmath26 , @xmath43 .
according to the above rules this term would be represented as shown in fig . 3a .
while the 2 vertices in the last term of fig .
2 are connected in the usual sense via a common ( solid ) line ( the dashed line with an attached bubble could be omitted in this case ) , the graph in fig .
3a would be disconnected in the old sense , since neither @xmath39 nor @xmath41 are line - connected with @xmath44 and @xmath45 , but -as a new feature of dlce graphs- the lines from @xmath39 to @xmath41 and from @xmath44 to @xmath45 are connected via the dashed lines emerging from a common bubble shown in the middle of the graph . as we see from fig .
3a , we need an additional notion of connectivity referring to the possibility of multiple - line connectivity . while the analytic expression is fixed , it is a matter of convenience to further simplify the graphical notation of fig .
3a at @xmath32 .
two possibilities are shown in fig .
3b and fig .
3c . to fig .
3b we later refer in the formal definition of the new type of multiple - line connectivity . in the familiar standard notion of connectivity two vertices of a graph
are connected via lines .
the vertices are line - connected .
already there , in a dual language , one could call two lines connected via vertices
. the second formulation is just appropriate for our need to define when two lines are connected .
the corresponding vertices mediating the connectivity of lines are visualized by tubes , in fig .
3b we have just one of them .
the tubes should be distinguished from the former type of vertices represented as full dots which are connected via bare @xmath19-lines . in fig .
3c we show a simplified representation of fig .
3b that we actually use in graphical expansions .
( 15.0,7.0 ) the derivative terms have to be evaluated at @xmath32 .
for @xmath32 we have a decomposition of @xmath33 according to w(h , i , v=0)= _ x _ 0 w^(h(x ) ) + _ l _ 1 w^1(i(l ) ) with [ 2.w0 ] z^(h ) = and [ 2.w1 ] z^1(i ) = .
in eq.s ( [ 2.w0],[2.w1 ] ) we have omitted any single site or single link dependence , because we assume that @xmath22 and @xmath23 are the same for all @xmath46 and @xmath47 , respectively .
therefore , at @xmath32 , the only non - vanishing derivatives of @xmath33 are [ 2.wd1 ] . w_h(x_1)h(x_2) ...
h(x_n ) _ v=0 = _
x_1,x_2, ... x_n and [ 2.wd2 ] . w_i(l_1)i(l_2) ...
l_m , but mixed derivatives w.r.t .
h and i vanish .
as anticipated in fig.s 3b and 3c , for @xmath32 we replace the dashed bubbles and graphically distinguish between bubbles with @xmath19-lines and bubbles with @xmath20-lines .
we define .
( 3.0,1.0 ) ( 0.0,0.0 ) ( 4.0,0.0 ) ( 0.0,0.0 ) ( 0.0,0.0)(1.0,0.8)(2.0,0.8 ) ( 0.0,0.0)(1.0,0.5)(2.0,0.5 ) ( 0.0,0.0)(1.0,-0.7)(2.0,-0.7 ) ( 1.2,0.2)(0.2,0)@xmath48
( 1.2,-0.1)(0.2,0)@xmath48 ( 1.2,-0.4)(0.2,0)@xmath48 } n = v^c_n = ( ) _ h=0 for a connected n - point vertex with @xmath49 bare @xmath19-lines emerging from it and .
( 4.0,1.0 ) ( 0.0,-0.6 ) ( 4.0,0.0 ) ( 1.0,1.5)(2.0,2.1)(3.0,1.5 ) ( 1.0,1.0)(2.0,1.6)(3.0,1.0 ) ( 1.0,0.0)(2.0,0.6)(3.0,0.0 ) ( 1.6,1.1)(0.2,0)@xmath48 ( 1.6,0.8)(0.2,0)@xmath48 ( 1.6,0.5)(0.2,0)@xmath48 ( 2.0,1.85)(0,1)0.2 ( 2.0,1.6)(0,1)0.2 ( 2.0,1.35)(0,1)0.2 ( 2.0,1.1)(0,1)0.2 ( 2.0,0.85)(0,1)0.2 ( 2.0,0.6)(0,1)0.2 ( 2.0,0.35)(0,1)0.2 ( 2.0,0.1)(0,1)0.2 } = m^1 c _ = ( ) _ i=0 for a connected @xmath50-line consisting of @xmath50 bare lines .
if @xmath51 , we often omit the dashed line .
if the bare lines of a @xmath50-line are internal @xmath19-lines , they get vertices attached to their endpoints , if they are external @xmath20-lines , no vertices will be attached .
let @xmath52 denote the lattice volume in @xmath3 dimensions .
the taylor expansion of @xmath33 about @xmath32 to second order in @xmath25 then reads [ 2.long ] w(h , i , v ) & = & w(h , i , v=0 ) + & + & _ x , y _ 0 v(x , y ) w_h(x ) w_h(y)w_i(x , y ) + & + & _ x , y , r , s _ 0 v(x , y )
v(r , s ) + & ( & 4 w_h(y)w_h(s)w_h(r)h(x)w_i(x , y)w_i(r , s ) + & + & 2 w_h(x)h(r)w_h(y)h(s)w_i(x , y)w_i(r , s ) + & + & 4 w_h(y)w_h(s)w_h(r)h(x)w_i(x , y)i(r , s ) + & + & 2 w_h(r)h(x)w_h(y)h(s)w_i(r , y)i(x , s ) + & + & w_h(x)w_h(y)w_h(r)w_h(s)w_i(x , y)i(r , s ) ) _ v=0 + & + & o(v^3 ) , where we have used that @xmath53 . for each @xmath33 in the products of @xmath33s we now insert eq.s ( [ 2.wd1]),([2.wd2 ] ) .
if we choose @xmath25 in a standard way as next - neighbour couplings v(x , y ) = 2k _ = 0^d-1 ( _ x+,y+ _ x-,y ) with @xmath54 denoting the unit vector in @xmath55-direction , eq .
( [ 2.long ] ) becomes in a graphical representation at @xmath56 [ 2.graph ] & = & ( 2k ) ( 2d ) ( 4.0,1.0 ) ( 0.0,0.0 ) ( 4.0,1.0 ) ( 0.0,0.2 ) ( 2.0,0.2 ) ( 0.0,0.2)(1,0)2.0 + & + & ( 2k)^2 \ { ( 2d)^2 ( 3.5,1.0 ) ( 0.0,0.0 ) ( 4.0,1.0 ) ( 0.0,-0.2 ) ( 0.0,0.6 ) ( 2.0,0.2 ) ( 0.0,-0.2)(1.0,-0.2)(2.0,0.2 ) ( 0.0,0.6)(1.0,0.6)(2.0,0.2 ) + ( 2d ) ( 4.0,1.0 ) ( 0.0,0.0 ) ( 4.0,1.0 ) ( 0.0,0.2 ) ( 2.0,0.2 ) ( 0.0,0.2)(1.0,-0.4)(2.0,0.2 ) ( 0.0,0.2)(1.0,0.8)(2.0,0.2 ) + & & + ( 2d)^2 ( 3.5,1.0 ) ( 0.0,0.0 ) ( 4.0,1.0 ) ( 0.0,-0.2 ) ( 0.0,0.6 ) ( 2.0,0.2 ) ( 0.0,-0.2)(1.0,-0.2)(2.0,0.2 ) ( 0.0,0.6)(1.0,0.6)(2.0,0.2 ) ( 0.9,0.2)(0,1)0.15 ( 0.9,0.4)(0,1)0.15 ( 0.9,0.6)(0,1)0.15 ( 0.9,0.0)(0,1)0.15 ( 0.9,-0.2)(0,1)0.15 ( 0.9,-0.4)(0,1)0.15 + ( 2d ) ( 4.0,1.0 ) ( 0.0,0.0 ) ( 4.0,1.0 ) ( 0.0,0.2 ) ( 2.0,0.2 ) ( 0.0,0.2)(1.0,-0.4)(2.0,0.2 ) ( 0.0,0.2)(1.0,0.8)(2.0,0.2 ) ( 1.0,0.2)(0,1)0.15 ( 1.0,0.4)(0,1)0.15 ( 1.0,0.6)(0,1)0.15 ( 1.0,0.0)(0,1)0.15 ( 1.0,-0.2)(0,1)0.15 ( 1.0,-0.4)(0,1)0.15 + & & + 2(2d ) ( 3.5,1.0 ) ( 0.0,0.0 ) ( 4.0,1.0 ) ( 0.0,0.0 ) ( 0.0,0.3 ) ( 2.0,0.0 ) ( 2.0,0.3 ) ( 0.0,0.0)(1.0,-0.5)(2.0,0.0 ) ( 0.0,0.3)(1.0,0.8)(2.0,0.3 ) ( 1.1,0.2)(0,1)0.15 ( 1.1,0.4)(0,1)0.15 ( 1.1,0.6)(0,1)0.15 ( 1.1,0.0)(0,1)0.15 ( 1.1,-0.2)(0,1)0.15 ( 1.1,-0.4)(0,1)0.15 } + & + & o(k^3 ) . for clarity , here we have written explicitly the topological symmetry factors and the lattice embedding numbers .
( usually graphs represent their full weights including these factors . ) note that the first two graphs of the second order contribution also occur in a usual lce with frozen @xmath20-dynamics , the next two differ by an additional dashed 2-line and the last one becomes even disconnected without the dashed line . as usual , graphical expansions for correlation functions , in particular susceptibilities ,
are generated from @xmath57 by taking derivatives w.r.t .
the external fields @xmath38 and @xmath40 .
graphically this amounts to attaching external @xmath19-lines and @xmath20-lines with in passing we remark that the conventional lce is included as a special case of the dlce , if the @xmath20-dynamics is `` frozen '' to some value @xmath58 , so that w^1(i ) & = & -s_1(u_0)+ i u_0 , + & = & u_0 , + & = & 0 , i.e. , no n - lines do occur with @xmath59 . in this case
it becomes redundant to attach dashed lines to bare lines . as mentioned above , in an lce only the first three contributions
would be left in eq .
( [ 2.graph ] ) .
the definition of a as it will be given here is adapted to the computation of susceptibilities , where the sum is taken over all possible locations of the fields .
the definition easily generalizes to correlation functions . for details of the standard definiton of graphs in the framework of linked cluster expansions and related notions we refer e.g. to @xcite . here , for convenience
, we briefly recall the very definition of a graph to point out the new properties of as defined below in this section .
a ( standard lce ) graph or diagram is a structure = ( _ , _ , e _ , _ ) , where @xmath60 and @xmath61 are disjoint sets of internal lines and vertices of @xmath62 , respectively .
@xmath63 is a map e _ : _ & & \{0,1,2 , } , + v & & e_(v ) that assigns to every vertex @xmath25 the number of external lines @xmath64 attached to it .
finally , @xmath65 is the incidence relation that assigns internal lines to their two endpoints .
a or is a structure = ( _ , _ , _ , e_^ ( ) , e_^(u ) , _ , _ ) .
@xmath66 , @xmath67 and @xmath68 are three mutually disjoint sets , _ & = & , + _ & = & , + _ & = & .
@xmath69 is a map e_^ ( ) : _ & & \{0,1,2 , } , + v & & e_^()(v ) that assigns to every vertex @xmath25 the number of bare external @xmath19-lines @xmath70 attached to @xmath25 .
every such @xmath19-line represents a field @xmath19 . the number of external @xmath19-lines of @xmath71 is denoted by @xmath72 .
similarly , @xmath73 is a map e_^(u ) : _ & & \{0,1,2 , } , + m & & e_^(u)(m ) that assigns to every multiple line @xmath74 the number of external @xmath20-lines @xmath75 attached to @xmath74 .
every such @xmath20-line represents a field @xmath20 associated with a lattice link .
the number of external @xmath20-lines of @xmath71 is given by @xmath76 .
furthermore , @xmath77 and @xmath78 are incidence relations that assign bare internal lines to their endpoint vertices and to their multiple lines , respectively .
we treat lines as unoriented .
the generalization to oriented lines is easily done .
more precisely , let @xmath79 be the set of unordered pairs of vertices @xmath80 with @xmath81 , @xmath82 .
( the bar implies unordered pairs , the prime the exclusions of @xmath83 , @xmath84 . ) as in the standard linked cluster expansion , self - lines are excluded .
every bare internal line is then mapped onto its pair of endpoints via _ : _ ^.
we say that @xmath25 and @xmath85 are the endpoint vertices of @xmath86 if @xmath87 .
if there is such an @xmath86 , @xmath25 and @xmath85 are called neighbours .
similarly , @xmath78 is a map _ : _ & & _ , + l & & _ ( l ) that maps every bare internal line to a multiple line . a multiple line @xmath88 is composed of bare internal lines @xmath86 which belong to @xmath74 in the sense that @xmath89 .
@xmath90 is the total number of bare internal lines belonging to @xmath74 . with @xmath91 , @xmath74
is called a @xmath50-line .
we always require that @xmath92 . on the other hand
, every bare internal line belongs to one and only one multiple line . for simplicity
we often identify a @xmath93-line with the only one bare line that belongs to it .
next we introduce some further notions that will be used later .
external vertices are vertices having external @xmath19-lines attached , _
, ext = \ { v _ e_^()(v)=0 } , whereas internal vertices do not , @xmath94 .
similarly , external multiple lines have external @xmath20-lines attached , _ , ext = \ { m _ e_^(u)(m)=0 } , and the complement in @xmath67 are the internal multiple lines , @xmath95 . for every pair of vertices @xmath81
, @xmath82 , let @xmath96 be the set of lines with endpoint vertices @xmath25 and @xmath85 , and @xmath97 the number of these lines .
thus @xmath96 is the set of lines @xmath25 and @xmath85 have in common . with @xmath70 denoting the number of external @xmath19-lines attached to @xmath84 , t__(v ) = _ w _ ^1(v , w ) + e_^()(v ) is the total number of bare lines attached to @xmath25 .
some topological notions and global properties of graphs will be of major interest in the following .
a central notion is the connectivity of a .
recall that we want to consider the dlce expansion of the free energy and of truncated correlation functions as an expansion in connected graphs .
as indicated in section 2 , the main generalization compared to the common notion of connectivity of a graph which is required here is that an additional type of connectivity is provided by multiple - lines . to define the connectivity of a @xmath71 , @xmath71
first is mapped to a ( standard ) lce graph @xmath98 to which the standard notion of connectivity applies .
there are various equivalent ways to define such a map .
we choose the following one .
* for every multiple - line @xmath88 define a new vertex @xmath99 . let @xmath100 and define @xmath101 as the union of the vertices of @xmath71 and the new set of vertices originating from the multiple - lines .
* for every bare internal line @xmath86 define two new internal lines @xmath102 and incidence relations ( l_1 ) & = & ( v_1 , w(_(l ) ) ) , + ( l_2 ) & = & ( v_2 , w(_(l ) ) ) , where @xmath103 and @xmath104 are the two endpoint vertices of @xmath42 .
the set of all lines @xmath102 , for all @xmath86 , is denoted by @xmath105 . *
define the external incidence relations : & & \{0,1,2 , } , + ( v ) & = & e_^()(v ) , , + ( v ) & = & e_^(u)(m ) , .
now , @xmath106 is defined by = ( , , , ) .
having defined the standard lce graph @xmath106 for any @xmath71 , we call @xmath71 multiple - line connected or just connected if @xmath106 is connected ( in the usual sense ) . in fig
. 4 we have given two examples for a connected ( upper graph ) and a disconnected ( lower graph ) . ( 15.0,7.0 ) the next important notion is the topological equivalence of two .
two [ top.equiv ] _ i = ( _ i , _ i , _ i , e^()_i , e^(u)_i , _ i , _
i ) , i=1,2 are called ( topologically ) equivalent if there are three invertible maps _ 1:_1 & & _ 2 , + _ 2:_1 & & _ 2 , + _ 3:_1 & & _ 2 , such that _ 2 _ 2 & = & _ 1 _ 1 , + _ 2 _ 2 & = & _ 3 _ 1 , and e_2^ ( ) _ 1 & = & e_1^ ( ) , + e_2^(u ) _ 3 & = & e_1^(u ) . here
@xmath107 means decomposition of maps , and _ 1 : ^ & & ^ + _ 1(v , w ) & = & ( _ 1(v),_1(w ) ) .
a symmetry of a @xmath108 is a triple of maps @xmath109 , @xmath110 and @xmath111 such that _ 2 & = & _ 1 , + _ 2 & = & _ 3 , and e^ ( ) _ 1 & = & e^ ( ) + e^(u ) _ 3 & = & e^(u ) . the number of these maps is called the symmetry number of @xmath71 . we denote by @xmath112 the set of equivalence classes of connected with @xmath113 bare internal lines , @xmath114 external @xmath19-lines and @xmath115 external @xmath20-lines
. furthermore we set _ e_1,e_2 : = _ l0 _ e_1,e_2(l ) .
a multiple line graph @xmath71 does not need to have a vertex .
if @xmath116 , we have @xmath117 as well .
if in addition @xmath71 is connected , @xmath67 consists of only one element , with all external @xmath20-lines attached to it .
( we anticipate that @xmath71 is 1-multiple - line irreducible ( 1mli ) by definition .
for the definition of 1mli cf .
section 4 below . ) the only graph of @xmath118 is given by = .
( 4.0,1.0 ) ( 0.0,0.0 ) ( 10.0,0.0 ) ( 0.9,0.6)(1.5,0.5)(2.1,0.6 ) ( 0.9,0.4)(1.5,0.3)(2.1,0.4 ) ( 0.9,-0.5)(1.5,-0.4)(2.1,-0.5 ) ( 0.9,-0.3)(1.5,-0.2)(2.1,-0.3 ) ( 1.5,-0.55)(0,1)0.15 ( 1.5,-0.35)(0,1)0.15 ( 1.5,-0.15)(0,1)0.15 ( 1.5,0.05)(0,1)0.15 ( 1.5,0.25)(0,1)0.15 ( 1.5,0.45)(0,1)0.15 ( 1.5,0.65)(0,1)0.15 ( 1.2,0.15)(0.2,0)@xmath48 ( 1.2,-0.05)(0.2,0)@xmath48 ( 1.2,0.05)(0.2,0)@xmath48 } e .
it represents the leading term of the susceptibility [ susc.0e ] _
0,e = _ l_1, ,l_e _ 1 < u(l_1 ) u(l_e ) > ^c and is given by @xmath119 .
the index @xmath120 in ( [ susc.0e ] ) stands for truncated ( connected ) correlation . by removal of a @xmath50-line @xmath88 we mean that @xmath74 is dropped together with all bare internal lines and all external @xmath20-lines that belong to @xmath74 .
this notion is explained in fig .
( it is used in section 4 for @xmath93-lines to define 1-particle irreducible ( 1pi ) and 1-line irreducible ( 1li ) . ) on the other hand , by decomposition of a @xmath50-line @xmath88 we mean that @xmath74 is dropped together with the external @xmath20-lines of @xmath74 , but all bare internal lines that belong to @xmath74 are kept in the graph , being identified now with @xmath93-lines .
this notion will be used below to define 1mli and renormalized .
it is illustrated in fig .
5b . similarly ,
decomposition of a vertex @xmath84 means to remove the vertex @xmath25 and to attach the free end of every line that entered @xmath25 before to a new vertex , a separate one for each line .
this notion is used to define 1-vertex - irreducible ( 1vi ) and renormalized vertex moments for . for
an example see fig .
( 15.0,7.0 ) in the last section we have defined and the notions of connectivity and equivalence of such graphs .
the definition is chosen in such a way that the series expansions of the free energy and of truncated correlation functions are obtained as a sum over equivalence classes of connected .
the number @xmath113 of bare internal lines of a @xmath71 counts the order in the expansion parameter @xmath26 to which @xmath71 contributes .
if @xmath26 is of the form [ susc.vform ] v(x , y ) = 2k _ z(x ) _ y , z , with @xmath121 any finite @xmath39-dependent set of lattice sites , the contribution of @xmath71 is a multiple of @xmath122 .
often used special cases are the nearest neighbour interactions [ susc.nn ] v(x , y ) = 2k _ = 0^d-1 ( _ x , y+ + _ x , y- ) and the uniform interaction [ susc.mf ] v(x , y ) = 2k ( 1 - _ x , y ) , which is used in models of spin glasses and partially annealed neural networks .
susceptibilities of the @xmath19 and @xmath20 fields will be represented as [ susc.susc ] _ e_1,e_2 & = & _ x_1, ,x_e_1 _ 0 _ l_1, ,l_e_2 _ 1 < ( x_1 ) ( x_e_1 ) u(l_1 ) u(l_e_2 ) > ^c + & & _ x_1, ,x_e_1 _ 0 _ l_1, ,l_e_2 _ 1 . _
h = i=0 + & = & _ l0 ( 2k)^l _ _
e_1,e_2(l ) w ( ) with lattice volume @xmath52 and dimension @xmath3 .
similar representations hold for higher moments @xmath55 .
the weight @xmath123 of a @xmath124 is given as the product of the following factors * for every vertex @xmath84 a factor v^c_n = ( ) _ h=0 , where @xmath125 is the total number of bare lines attached to @xmath25 .
* for every multiple line @xmath88 a factor [ susc.m1c ] m^1 c _ = ( ) _ i=0 , where @xmath91 , that is @xmath74 is a @xmath50-line , * a factor @xmath126 , where @xmath127 is the topological symmetry number of @xmath71 , * a factor counting the permutation symmetry of external @xmath19-lines , , * a factor counting the permutation symmetry of external @xmath20-lines , , * the lattice embedding number of @xmath71 , which is the number of ways @xmath71 can be embedded on a lattice of given geometry , e. g. on a hypercubic lattice .
to this end , the vertices of @xmath71 ( if any ) are placed onto lattice sites .
one arbitrary vertex is placed at a fixed lattice site , in order to account for the volume factor @xmath128 in ( [ susc.susc ] ) . a priori
there is no exclusion principle .
this means that any number of vertices can be placed at the same lattice site .
( this is sometimes called free embedding . )
two restrictions apply to the embeddings .
the first constraint results from the fact that a bare internal line represents a hopping propagator @xmath26 , with lattice sites @xmath39 and @xmath41 at which the two endpoint vertices of the line are placed at .
a reasonable computation of the embedding number takes into account the particular form of @xmath26 from the very beginning .
the second constraint is that bare lines of the same multiple - line have to be mapped on the same pair of sites . + for example , if @xmath26 is the nearest neighbour interaction ( [ susc.nn ] ) , two vertices which have at least one line in common are to be placed at nearest neighbour lattice sites .
on the other hand , a propagator @xmath26 of the form ( [ susc.mf ] ) implies a rather weak constraint in that @xmath39 and @xmath41 must be different , but otherwise can be freely placed over the lattice .
we remark that in case of a non - trivial internal symmetry ( such as considered in section 7 ) the expressions of eq.s ( [ susc.susc])-([susc.m1c ] ) must be modified appropriately .
in particular , the weight ( [ susc.m1c ] ) of a multiple - line does no longer take such a simple form .
( [ su2.labgraph ] ) below is an example for the case of a hopping term originating in an su(2 ) higgs model .
truncated correlation functions , susceptibilities and other moments are obtained as sums over that are connected .
their number rapidly grows with increasing order , that is with increasing number of bare internal lines .
the procedure of `` renormalization '' means that the connected moments are represented in terms of reduced ones .
the reduced moments are obtained by summation over classes which are more restricted than just by their property of being connected .
of course the number of graphs of such classes is smaller .
only the most restricted classes must be constructed .
the subsequent steps towards the moment computation are most conveniently done by operating analytically with the reduced moments . in particular , it is no longer necessary to generate all connected and the corresponding intermediate classes .
a connected @xmath71 is called 1-particle irreducible ( 1pi ) if it satisfies the following condition .
remove an arbitrary @xmath93-line of @xmath71 .
there is at most one connected component left that has external lines attached .
( this notion is the same as the one used in the context of feynman graphs . ) on the other hand , if in addition the remaining graph is still connected , then @xmath71 is called 1-line irreducible ( 1li ) . in many cases
it is sufficient to use only the second notion .
it is for instance sufficient that all vertices are constrained to have only an even number of lines attached , or more generally , if graphs and subgraphs with one external line are forbidden . for notational simplicity
we assume in the following that this is the case and henceforth refer only to the notion 1li .
the generalization to the case in which 1li and 1pi graphs must be distinguished goes along the same lines as for lces , which was discussed in @xcite . by @xmath129
we denote the subset of @xmath124 that are 1li .
1li - susceptibilities are defined as series in the hopping parameter similarly as in ( [ susc.susc ] ) by restricting the summation to 1li graphs ,
_ e_1,e_2 ^ 1li = _ l0 ( 2k)^l _ _ e_1,e_2 ^ 1li(l ) w ( ) .
susceptibilities are easily obtained in a closed form in terms of 1li - susceptibilities @xmath130 .
it can be shown that the @xmath130s can be obtained by an appropriate legendre transform .
for instance _ 2,0 & = & , + _ 2,1 & = & , where @xmath131 is the fourier transform of the hopping propagator @xmath26 , v(x , y ) = _
-^ e^-ik(x - y ) ( k ) . in lces
the second important resummation comes from so called vertex renormalizations .
this means partial resummation of graphs with specific properties such as having one external vertex only .
these sums then are considered as `` renormalized vertices '' replacing the vertices of graphs with complementary properties .
the procedure naturally leads to the notion of 1-vertex irreducibility ( 1vi ) and renormalized moments . in dlce
we follow this procedure .
the very definition of 1vi has to be modified slightly for because of the enhanced connectivity properties due to multiple - lines .
in addition , as a natural generalization , we supplement vertex renormalization by multiple - line renormalization .
a @xmath71 is called 1-vertex irreducible ( 1vi ) if it satisfies the following condition .
decompose an arbitrary vertex @xmath84 .
every connected component of the remaining graph has then at least one external line attached .
it can be a @xmath19-line or a @xmath20-line . we write _
e_1,e_2 ^ 1vi(l ) = \ { _ e_1,e_2 ^ 1li(l ) } for the set of equivalence classes of graphs that are both 1li and 1vi , with @xmath114 external @xmath19-lines , @xmath115 external @xmath20-lines and @xmath113 bare internal lines .
the renormalized vertex moment graphs are 1li graphs that have precisely one external vertex and no external multiple line , q_k(l ) = \ { _ k,0 ^ 1li(l ) }
. a @xmath71 is called 1-multiple - line irreducible ( 1mli ) if it satisfies the following criterion .
decompose an arbitrary multiple - line @xmath88 .
every remaining connected component has then at least one external line attached .
it can be a @xmath19-line or a @xmath20-line . we write _
e_1,e_2 ^ 1mli(l ) = \ { _ e_1,e_2 ^ 1li(l ) } .
the renormalized multiple - line moment graphs are graphs that are 1li and have precisely one external multiple - line , but no external vertex , r_k(l ) = \ { _ 0,k^1li(l ) } .
the equivalence classes of graphs that are both 1vi and 1mli are denoted by s_e_1,e_2(l ) = _ e_1,e_2 ^ 1vi(l ) _
e_1,e_2 ^ 1mli(l )
. with the renormalized moment graphs as defined above , the 1li - susceptibilities are now obtained in the form _
e_1,e_2 ^ 1li = _ l0 ( 2k)^l _ _ e_1,e_2(l ) ( ) .
the weights @xmath132 are given as a product of factors as described in the last subsection , with the following two exceptions .
* the vertex coupling constants @xmath133 are replaced by the renormalized vertex moments v^c_n v^c_n = _ l0 ( 2k)^l _ q_n(l ) w ( ) .
* the multiple line coupling constants @xmath134 are replaced by the renormalized multiple line moments m^1 c _ m^c_= _ l0 ( 2k)^l _ r_(l ) w ( ) . in the series representations above , the @xmath123
are computed according to the rules of subsection 3.2 .
in this section we describe a mechanism to generate the classes @xmath135 , @xmath136 and @xmath137 with @xmath138 internal lines .
similar as in lces the idea is to define appropriate classes of vacuum graphs ( no external lines ) .
the generation of the graphs of @xmath139 , @xmath140 and @xmath141 is then done by attaching external lines in various ways . by defining appropriate recursion relations , the order by order construction can then be restricted to the vacuum classes .
fortunately we can profit from graphs that have been generated already in standard lces .
this way we completely avoid a cumbersome recursive construction , but obtain the vacuum by operating on the lce vacuum graphs of the same order . the graphs of @xmath139 , @xmath140 and @xmath141 will be obtained through a sequence of simpler graph classes . at each step , any method of construction should satisfy the following two conditions .
* it should be surjective or complete .
at least one graph of every equivalence class should be generated .
* for every equivalence class precisely one representant should be kept .
all generated graphs of a particular class of graphs should be mutually inequivalent .
this requires a so called non - equivalence test for every newly generated graph .
the graph should be kept if and only if it is not equivalent to a graph already generated . in general
it must be compared with all that have been generated before .
if this is necessary , the non - equivalence test is rather time consuming .
the algorithm we describe below circumvents an extensive comparison of .
it is defined in such a way that two graphs can be equivalent only if they have their origin in the same graph of the prior graph class .
essentially we proceed in the following steps .
the starting point are the vacuum diagrams of @xmath142 as computed in @xcite .
@xmath142 is the set of ( standard ) lce equivalence classes of connected , 1li vacuum graphs with @xmath113 lines .
operating on @xmath142 , we generate all mutually inequivalent with @xmath113 lines that are connected , 1li and have no external lines ( neither @xmath19-lines nor @xmath20-lines ) , _
2(l ) _ 0,0 ^ 1li(l ) .
this is done in two steps . in a first step lines with the same endpoint vertices
are arranged in all possible ways into multiple lines .
we obtain a subset of @xmath143 which we denote by @xmath144 .
a @xmath145 belongs to @xmath144 , @xmath146 , if and only if all bare lines of @xmath71 that belong to the same multiple line have the same endpoint vertices .
the second step is to relax the additional constraint on @xmath144 .
this is achieved by resolving the vertices .
this means that every vertex is split into several ones , and that the bare lines that were attached to the original vertex before now are distributed over the new vertices in all possible ways . as a general constraint on the resolutions ,
the resulting should be 1li in order to stay in @xmath143 .
we impose an additional constraint on the resolutions , allowing only for so - called admissible vertex resolutions .
they are defined in such a way that two of @xmath147 can be equivalent only if they originate from the same graph of @xmath144 .
this considerably simplifies the non - equivalence test .
once we have obtained @xmath147 , the final classes @xmath135 , @xmath136 and @xmath137 are obtained by attaching external @xmath19-lines and @xmath20-lines in the appropriate way to all graphs of @xmath147 . the steps in the generation of the dlce graph classes
are shown schematically in fig .
( 15.0,2.5 ) ( 0.0,1.0)(2.0,0)@xmath142 ( 2.3,1.0)(1,0)0.9 ( 4.0,1.0)(2.0,0)@xmath148 ( 6.3,1.0)(1,0)0.9 ( 8.0,1.0)(2.0,0)@xmath143 ( 10.3,1.0)(1,1)0.9 ( 10.3,1.0)(1,0)0.9 ( 10.3,1.0)(1,-1)0.9 ( 12.0,2.0)(2.0,0)@xmath135 ( 12.0,1.0)(2.0,0)@xmath136 ( 12.0,0.0)(2.0,0)@xmath137 in the following we describe the construction in detail .
below we use the following notation . for
any finite set @xmath149 let @xmath150 denote the set of partitions of @xmath149 into mutually disjoint nonempty subsets of @xmath149 .
for every @xmath151 we have = _ p p. starting with the lce vacuum graphs of @xmath142 , we first of all arrange internal lines into multiple lines .
consider an arbitrary graph @xmath153 .
simultaneously for every pair of neighbouring vertices @xmath81 choose a partition @xmath154 , where @xmath155 is the set of common lines between the vertices @xmath25 and @xmath85 .
every @xmath156 belongs to a unique @xmath157 .
for every @xmath158 we introduce a multiple line @xmath159 such that all @xmath160 are precisely the bare internal lines that belong to @xmath159 .
this procedure defines the incidence relation @xmath161 for all @xmath86 and , along with that , a that belongs to @xmath144 . in fig .
7 we give examples for this construction . in the first one ,
an lce graph leads to 5 mutually inequivalent dlce gaphs , in the second one to 3 dlce graphs of @xmath148 . proceeding this way for all partitions of lines for all pairs of neighboured vertices of @xmath71 , and for all @xmath153
, we obtain the class of @xmath144 .
the non - equivalence check for every newly generated @xmath62 can be drastically restricted .
it is sufficient to compare @xmath62 only with those that have been constructed before from the same graph @xmath153 as @xmath62 results from .
other graphs of @xmath144 are inequivalent to @xmath62 .
( 15.0,7.0 ) the second step is to relax the constraint on @xmath163 that bare lines which belong to the same multiple line must have the same endpoint vertices .
this is achieved by vertex resolutions of all graphs of @xmath144 . to efficiently reduce the non - equivalence check of new graphs we allow only for so - called admissible vertex resolutions .
let @xmath164 , @xmath165 any vertex of @xmath62 and @xmath166 any partition of the set of lines @xmath167 entering @xmath25 .
we remove the vertex @xmath25 and draw for every @xmath158 a new vertex @xmath168 so that all lines @xmath160 enter the vertex @xmath168 rather than @xmath25 before its removal .
this procedure is called a vertex resolution of @xmath25 .
the resulting graph @xmath71 must be 1li in order to belong to @xmath143 .
the vertex resolution of @xmath25 is called admissible if in addition the ( standard lce ) graph @xmath169 is connected with @xmath169 defined in the following way .
for every @xmath170 there is precisely one @xmath171 with @xmath172 , and there is exactly one @xmath158 with @xmath160 . draw a vertex @xmath173 for every such multiple line , a vertex @xmath168 for every @xmath158 , and define a line @xmath174 with incidence relation ( ) = ( w(_(l ) ) , v(p ) ) .
this defines a graph @xmath175 with _ & = & \ { v(p ) p } \ { w(_(l ) ) l_v } + _ & = & \ { ( l ) l_v } .
a vertex resolution is called admissible if the graph @xmath169 is connected . roughly speaking this implies that no vertex drops off after vertex resolution
. fig .
8 shows two admissable and one non - admissible vertex resolutions ( the auxiliary graphs @xmath169 are not displayed ) .
( 15.0,7.0 ) the class @xmath143 is now generated as follows .
consider an arbitrary graph @xmath176 .
simultaneously for all vertices @xmath165 apply an admissible vertex resolution .
we obtain a @xmath145 . doing this for all possible resolutions of all vertices and for all @xmath176
, we get the complete graph class @xmath143 . again , due to the restriction to admissible vertex resolutions , the non - equivalence check of a newly generated graph @xmath145 can be restricted to graphs that have been generated from the same graph of @xmath144 as @xmath71 was . in fig .
9 we show an example for the application of admissible vertex resolutions . here
one graph of @xmath148 leads to eight graphs of @xmath143 , indicating the proliferation of dlce graphs .
note that the starting graph was only one of five of fig .
( 15.0,5.5 ) the final step is to construct the classes of @xmath180 , @xmath181 and @xmath182 out of the vacuum of @xmath183 .
this step is realized by attaching external @xmath19-lines and external @xmath20-lines in different ways to every @xmath145 .
@xmath179 is obtained by attaching @xmath184 external @xmath20-lines to just one multiple line @xmath88 , for all multiple lines @xmath74 and for all @xmath71 .
@xmath178 is obtained by attaching @xmath184 external @xmath19-lines to just one vertex @xmath84 , for all vertices @xmath25 and for all @xmath71 .
@xmath135 is obtained by attaching @xmath114 external @xmath19-lines to the vertices of @xmath71 and @xmath115 external @xmath20-lines to the multiple lines of @xmath71 , for all @xmath145 .
this is done under the constraint that the resulting have to be 1vi and 1mli .
again , the non - equivalence test of can be confined to pairs of graphs that have their origin in the same graph of @xmath143 . in fig .
10 we illustrate the attachment of external lines to a vacuum dlce graph of @xmath143 .
( 15.0,5.5 )
in this section we indicate applications of dlces to spin glasses and partially annealed neural networks with `` slow '' interactions and `` fast '' spins @xcite-@xcite , @xcite . for simplicity
we consider the sherrington - kirkpatrick ( sk ) model which is a spin glass model with infinite connectivity @xcite . following the notation of section 2 ,
the fast spins are represented by the @xmath185-field @xmath186 , @xmath187 , taking values in @xmath188 .
@xmath189 denotes the total number of spins .
the slow interactions are mediated by the @xmath20-field @xmath190 , @xmath191 .
the sk model is described by the partition function [ neur.z ] z_^^ = _
-^_i < j = 1^n du_(i , j ) , with @xmath9 some normalization that will be specified below .
the effective hamiltonian @xmath192 of @xmath20 is given by [ neur.h ] ( u ) = - + n _
i < j=1^n u_(i , j)^2 , where @xmath55 is a positive coupling constant
. @xmath193 is the partition function of a spin system [ neur.zspin ] z_(u ) = _ \ { _ i = 1 } ) at frozen spin - spin couplings @xmath190 and temperature @xmath194 , controlling the fast spin fluctuations . in ( [ neur.z ] ) it has been assumed that the equilibrium distribution of the slow variables is a boltzmann distribution governed by a second temperature @xmath195 .
this assumption is justified , if the time evolution of the @xmath20s is determined by a dissipative langevin equation ( for the precise conditions see e.g. @xcite ) .
such a langevin equation can be actually derived for the @xmath20s from an ansatz which is motivated by neural networks @xcite-@xcite .
one of the major quantities of interest is @xmath196 .
let us first rewrite @xmath197 in the form [ neur.zre ] z_n^ & = & _ -^_i < j = 1^n ( du_(i , j ) ) ) z_(u)^n + & & [ [ z_(u)^n ] ] , where we have introduced @xmath198 and @xmath199 .
the normalization has been chosen such that @xmath200 = 1 $ ] .
the limiting case @xmath201 with @xmath202 kept fixed corresponds to a quenching of the @xmath20s in a spin glass , in this limit the @xmath20 degrees of freedom decouple from the spin dynamics , while @xmath203 corresponds to a pure annealing in the disordered system ( see e.g. @xcite ) .
partial annealing then refers to a value of @xmath204 with @xmath205 .
in particular the limit @xmath206 with fixed @xmath202 and @xmath207 amounts to the replica trick in spin glasses @xcite . assuming that averaging @xmath208 $ ] and limes @xmath206 commute , we have [ neur.rep ] [ [ ( u ) ] ] = _ n0 .
the left hand side of ( [ neur.rep ] ) is calculated by means of the right hand side .
so far @xmath204 has been a positive real number . for integer @xmath209 , @xmath210 is the partition function of an n - times replicated system .
let us first consider this case .
we rewrite [ neur.zspin1 ] z_(u)^n = _
\{_i^(a ) } , with @xmath211 labelling the replicated spin variables , so that [ neur.zspin2 ] z_n^ & = & _ -^_i < j = 1^n du_(i , j ) _
\{_i^(a ) = 1 } , + s(u,^(a ) ) & = & -_a=1^n _
j=1^n u_(i , j ) _ i^(a ) _
j^(a ) + j n _
i < j u_(i , j)^2 .
linear terms in @xmath185 and @xmath20 may be included according to [ neur.zspin3 ] s_lin = - h _
a=1^n _ i=1^n _
i^(a ) + c _ i <
j=1^n u_(i , j ) with constant external fields @xmath212 and @xmath120 .
apparently @xmath213 has the form of models to which dlce applies , with a hopping term [ neur.hop ] s_hop(u,^(a ) ) = - _ a=1^n _ i
< j=1^n u_(i , j ) _ i^(a ) _
j^(a ) , a single link action [ neur.link ] s^1(u_(i , j ) ) = c u_(i , j ) + j n u_(i , j)^2 , and a single site action [ neur.site ] s^(_i^(a ) ) = - h _ a=1^n _
i^(a ) . application of dlce to @xmath214 and derived quantities yields their series expansions in @xmath207 .
so far , merely for simplicity , we have considered the sk model which is a spin glass model with infinite and uniform connectivity , in which all spins are coupled with equal variance @xmath215 = 1/(j n)$ ] . for integer @xmath216 ,
this system becomes exactly solvable in the thermodynamic limit corresponding to the limit @xmath217 , so that the leading order of a large-@xmath189 expansion becomes exact , cf .
e.g. @xcite .
in dlce this amounts to a resummation of appropriate tree graphs . in cases of finite or non - uniform connectivity
it is no longer possible to obtain solutions in a closed form .
examples are multilayered neural networks with couplings between adjacent layers , and spin glasses with short range interactions . models of this kind are suited for an application of dlces and will be studied in future work . finally we come to the limiting case of @xmath201 .
usually it is obtained by extrapolating the results obtained for @xmath209 towards @xmath201 .
for instance , @xmath218 $ ] is obtained from @xmath219 $ ] for @xmath209 .
this is a rather subtle point because the extrapolation is not unique without further assumptions . in the next subsection
we show how it is possible to completely avoid both the extrapolation and the replica trick and to directly compute the quantity of interest by means of dynamical linked cluster expansions .
the task is to appropriately identify the classes of contributing graphs .
first we adapt the notation to section 2 to include more general cases .
@xmath2 denotes the support of the spins , that is the set of lattice sites , with @xmath220 denoting their total number .
@xmath8 @xmath221 are the pairs of sites whose spins interact . in accordance with ( [ neur.zre ] ) , we write for the normalized link - average of a function @xmath222 [ sg.average ] [ [ f(u ) ] ] = u f(u ) with u & = & _
l_1 d(u(l ) ) , + d(u ) & = & _ 1 du , _
-^d(u ) = 1 .
it is convenient to introduce the single link expectation values < g(u ) > _ 1 d(u ) g(u ) and the generating function @xmath223 by < > _ 1 .
the way in which the replica trick can be avoided is examplified for the free energy density @xmath224 of the spin system averaged over the link couplings .
the partition function of the spin system for a given distribution of the link interactions @xmath225 is given by = _ sp , where @xmath226 and s_sp(,u ) & = & - _ x , y_0 v(x , y ) ( x ) ( y ) u(x , y ) , + & = & _ x_0 d(x ) . without loss of generality we identify the support of the interaction @xmath227 with the set @xmath228 of lattice sites where @xmath229 is supported , _ 1 = \
{ l=(x , y ) v(x , y)= 0}. for simplicity we assume @xmath26 to be of the form ( [ susc.vform ] ) , so that @xmath1 is a measure of the strength of the interactions @xmath26 .
the free energy density of the spin system allows for a series expansion in the standard lce sense , with the link field @xmath13 playing the role of a `` background field '' , [ sg.wsp_series ] w_sp(u ) = _ l0 ( 2k)^l _ ^sp_0(l)w^sp(,u ) . here
@xmath230 ( with @xmath231 ) denotes the set of equivalence classes of connected lce graphs with @xmath232 external lines and @xmath113 internal lines .
the spin - weights @xmath233 are of the form [ sg.wsp_weight ] w^sp(,u ) = r^sp ( ) _ _ _ 1^ _ l_1
u(l)^m(l ) .
the sum is taken over all non - vanishing lattice embeddings of the graph @xmath71 .
it runs over all maps of internal lines of the graph @xmath71 to pairs of lattice sites of @xmath8 that are consistent with the graph topology in the sense discussed in section 3 . for every @xmath14
, @xmath234 denotes the number of lines of @xmath71 that are mapped onto the link @xmath42 by the embedding .
all other factors that contribute to the weight are collected in the prefactor @xmath235 , including the inverse topological symmetry number of @xmath71 .
next we want to express @xmath236 $ ] as a series in @xmath1 by means of dlce . toward this end we set @xmath237 and insert the series ( [ sg.wsp_series ] ) with ( [ sg.wsp_weight ] ) into ( [ sg.average ] ) . at this stage
we are not concerned with question of ( uniform or dominated ) convergence and obtain [ sg.wspint ] [ [ w_sp(u ) ] ] & = & _ l0 ( 2k)^l _ _
0^sp(l ) u w^sp(,u ) + & = & _ l0 ( 2k)^l _ _ 0^sp(l ) r^sp ( ) _ _ _ 1^ _
l_1 < u(l)^m(l)>_1 .
the next step is to express the full single link expectation values in terms of the connected ones .
they are related by [ sg.conn ] < u^m > _ 1 = _ ( ) _
p < u^|p| > _ 1^c , where @xmath238 denotes the set of all partitions of @xmath239 into non - empty , mutually disjoint subsets of @xmath240 .
@xmath241 is the number of elements of the set @xmath242 .
the relation ( [ sg.conn ] ) is equivalent to the partition of all lines of @xmath71 that are mapped to the same lattice link into multiple lines , with every multiple line contributing a factor [ sg.wdiff ] < u^|p| > _ 1^c = .
_ i=0 = m^1 c_|p| . using ( [ sg.conn ] ) ,
( [ sg.wdiff ] ) we rewrite ( [ sg.wspint ] ) as [ sg.wsptodlce ] [ [ w_sp(u ) ] ] = _ l0 ( 2k)^l _ _ 0^sp(l ) r^sp ( ) _ ( _ ) ( _ p m^1 c_|p| ) ( _ _ 1^ _ l_1 1 ) .
the last summation in ( [ sg.wsptodlce ] ) is over all maps @xmath243 of the lines of @xmath71 to the lattice links of @xmath8 subject to the constraint that all lines that belong to the same multiple - line corresponding to some @xmath158 are mapped onto the same lattice link .
finally we rewrite ( [ sg.wsptodlce ] ) as a sum over . to this end
, we first observe that for every @xmath244 , every partition @xmath245 of the lines of @xmath71 into multiple - lines generates a @xmath246 in the obvious way .
let us denote by @xmath247 the subset of of @xmath248 that stay connected after decomposition of all multiple lines .
( these are the which stay connected in the usual graph theoretical sense , when the dashed lines are omitted . ) for every @xmath249 there is a unique @xmath250 and at least one @xmath251 such that @xmath252 .
let @xmath253 be the ( uniquely determined ) number of partitions @xmath251 with @xmath252 , and @xmath254 such an arbitrary partition .
( [ sg.wsptodlce ] ) then becomes [ sg.wdlce ] [ [ w_sp(u ) ] ] = _ l0 ( 2k)^l _ _ 0,0(l ) n_r^sp ( ( ) ) ( _ p
( ) m^1 c_|p| ) ( _ ( ) _ 1^ 1 ) .
the last bracket of ( [ sg.wdlce ] ) is the lattice embedding factor of the @xmath255 .
the second bracket from the right does not depend on the choice of @xmath254 and is the product of the multiple - line coupling constants as defined in section 3 .
finally , @xmath256 is precisely the remaining part of the weight of @xmath255 that was described in detail in section 3 , endowed with the correct inverse topological symmetry number of the @xmath255 ( because of the factor @xmath253 ) . in summary
, we obtain the series expansion of the link - averaged free energy density in terms of dlce graphs , [ sg.wseries ] [ [ w_sp(u ) ] ] = _ l0 ( 2k)^l _ _ 0,0(l ) w ( ) . the weight @xmath257 of a @xmath255 is defined and computed according to the rules given in section 3 .
( [ sg.wseries ] ) is the series representation of the link - averaged free energy density of the spin system , i.e. the free energy density of the @xmath201 replica system , in terms of dlce graphs .
it looks much like the series representation of the @xmath93-replica system , which is given by [ sg.w1series ] w_1-repl = _ l0 ( 2k)^l _ _ 0,0(l ) w ( ) according to the discussion of section 2 .
we recall that @xmath248 is the set of dlce vacuum graphs with @xmath113 bare lines that are connected in the generalized dlce sense . comparing ( [ sg.wseries ] ) and ( [ sg.w1series ] ) , the transition from @xmath203 to @xmath201 replicas is achieved by keeping only the subset @xmath258 of that are connected in the original ( lce ) sense .
we expect that the series ( [ sg.wseries ] ) are convergent for a large class of interactions @xmath30 and @xmath26 if the coupling constant @xmath1 is sufficiently small . for special interactions most of the yield vanishing contributions so that we can further restrict the sum to a subset of @xmath259 .
an example is given by the mean field type of interaction of the sk model where all pairs of sites are coupled with the same strength .
that is , in the notation of this subsection , v(x , y ) & = & k ( 1-_x , y ) , + _ 1 & & _ 1 + s^1(u ) & = & v u^2 . in the thermodynamic limit @xmath260
only those survive that consist exclusively of @xmath261-lines , and that do not contain any loop of @xmath261-lines .
the resulting tree structure provides the possibility of summing the series ( [ sg.wseries ] ) and of analytically continuing the result to the spin glass phase of large coupling @xmath1 .
work in this direction is in progress .
in this section we discuss applications of dlces to study the higgs transition within a variational estimate for the gauged su(2 ) higgs model .
the estimate serves as an effective description of the electroweak standard model .
we consider a 4-dimensional hypercubic finite temperature lattice @xmath2 of size @xmath262 , with @xmath263 the inverse temperature in lattice units and @xmath264 the spatial volume .
the lattice links are the set of nearest neighbour lattice sites , given by _ 1 = \ { ( x ; ) x_0 , = 0, ,3 } .
the gauge field @xmath265 is an su(2 ) valued field living on the lattice links @xmath8 .
it is convenient to parametrize such an su(2 ) matrix by [ su2.uparam ] u & = & _ 0(u ) _
2 + i ( u ) + & & u_0 + i with @xmath266 . the higgs field @xmath185 lives on the lattice sites @xmath2 .
its values are real multiples of su(2 ) , [ su2.phiparam ] ( x ) & = & _ 0((x ) ) _ 2 + i ( ( x ) ) + & & _
0(x ) + i ( x ) , with @xmath267 .
we say that @xmath185 is @xmath268 valued . as a special case of eq .
( [ 2.zgen ] ) we consider the partition function [ su2.zdlce ] z_ve1 = u , in which u = _
_ = 0 ^ 3 d_h(u(x ; ) ) , with @xmath269 the normalized haar measure on su(2 ) , and [ su2.phimeasure ] = _ x_0 d((x ) ) , where d ( ) & = & d^4 , + s^ ( ) & = & + ( - 1 ) ^2 + & = & ^2 + ( ^2 - 1 ) ^2 . finally , [ su2.svei ] s_ve1(u , ) & = & - _ x_0 \ { 4 _ link _ 0(u(x;0 ) ) + _ = 1 ^ 3 4 _ cube _ 0(u(x ; ) ) + & & + _ = 1 ^ 3 ( 2 ) + & & + _ 0((x ) ) } , depending on variational parameters @xmath270 , @xmath271 and @xmath272 @xmath21 , and on the hopping parameter @xmath273 .
the first two terms are non - gauge invariant substitutes for the wilson plaquette term of the gauge part of the su(2 ) higgs model , cf .
( [ su2.1.wilson ] ) below .
we distinguish timelike from spacelike directions .
timelike links are denoted as @xmath274 and spacelike as @xmath275 , @xmath276 .
the third term is a hopping term coupling @xmath185 and @xmath20 fields only in spatial directions .
the last term depends on the third variational parameter @xmath272 associated with the higgs field @xmath185 , while the remaining ultralocal action for @xmath185 has been absorbed in the measure . in the next subsection
we first indicate how we arrive at the action ( [ su2.svei ] ) within a variational estimate for the free energy of the su(2 ) higgs model .
we derive equations for the optimal choice of variational parameters and an equation -based on stability arguments- to determine the critical hopping parameter @xmath277 at which the higgs phase transition sets in .
these equations will depend on expectation values of n - point correlations that are most suitably evaluated with a dlce in three dimensions .
the generic graphical expansion exposed in sections 3 - 5 now has to be adapted to account for the internal symmetry .
this is somewhat involved so that we restrict our discussion to some remarks in section 7.2 , but postpone any details to a forthcoming paper @xcite . in section 7.3 we present some results for the critical line in comparision with other variational estimates and monte carlo simulations . by means of dlces we want to obtain an analytic estimate for the critical line in the su(2 ) higgs model that serves as an effective description of the higgs transition in the electroweak standard model .
because of the rather different methodical ansatz in this approach , for a common choice of parameters this estimate will provide an independent check of numerical results based on monte carlo simulations .
in addition , larger scalar field couplings are available in our approach and larger values for @xmath278 , the number of time slices .
( in our calculations @xmath279 or @xmath280 , while it was chosen as 2 or 3 in @xcite . ) larger @xmath278s are required for a finite size scaling control . to fix the notation we introduce the action of the su(2 ) higgs model as [ su2.1.higgs ] z & = & u , + s(u , ) & = & s_w(u ) + s_hop , + s_hop & = & - _ x_0 \ { _ = 0 ^ 3 ( 2 ) } .
the higgs self - interaction has been absorbed in the measure @xmath281 of eq .
( [ su2.phimeasure ] ) , @xmath282 is the hopping parameter term in 4 dimensions .
@xmath283 is the su(2 ) gauge invariant wilson action [ su2.1.wilson ]
s_w(u ) = _ x_0 _ _ ve ) . eq .
( [ su2.1.var ] ) follows from the convexity of the exponential function and holds independently of the specific choice of @xmath284 and @xmath285 , if the measure is positive definite and normalized .
@xmath52 is the 4-dimensional volume , @xmath286 denotes the true physical free energy density defined via @xmath284 , @xmath287 the trial free energy density defined via the second equality in ( [ su2.1.var ] ) .
@xmath288 refers to the original action in @xmath284 , in our case it is the su(2 ) higgs action , and @xmath289 to an ansatz for the action in @xmath285 depending on a generic set of variational parameters @xmath290 .
these parameters should be optimized so that the difference of the trial and the physical free energy density @xmath291 becomes minimal for @xmath292 at which @xmath287 takes its minimum .
the most naive ansatz for @xmath289 is a mean field ansatz in the spirit of a molecular field approximation , leading to a factorization of @xmath293 according to [ su2.1.zve0 ] z_ve0 ( , ) = z_link^4 l_0 v_3 ( ) z_site^l_0 v_3 ( ) .
the partition function depends on two variational parameters @xmath290 and @xmath272 and factorizes in a product over single link ( @xmath294 ) and single site ( @xmath295 ) partition functions , @xmath264 denotes the three - dimensional volume . in this ansatz
there is no space for implementing an asymmetry between temporal and spatial directions .
thus the results will be temperature independent by construction .
one can think of a variety of improvements of the molecular field ansatz , arguments against or in favour of the various versions will be given in @xcite .
here we only consider the ansatz which is most plausible from a physical point of view and leads to the best agreement with monte carlo results for comparable sets of parameters .
the ansatz treats the spacelike degrees of freedom of the hopping term in 3 dimensions beyond the mean field level ( that is with dlce ) , but the timelike degrees of freedom within a mean field approach , so that the partition function factorizes over the spacelike hyperplanes .
the reason is that the spatial hopping term is supposed to contain the nonperturbative properties of the full model that drive the higgs phase transition .
while a mean field approach for all variables will be too rough to produce high quality results for @xmath277 , it appears more reasonable for timelike variables , although a finite temperature effect on the higgs transition gets lost this way .
( we should remark that from results in a scalar @xmath0 theory in four dimensions the finite temperature effect on @xmath277 is expected to be anyway quite small @xcite . )
now it is easily checked that the choice of @xmath296 in eq .
( [ su2.svei ] ) leads to a factorization of @xmath297 according to [ su2.1.zvei ] z_ve1(_link,_cube , ) = z_cube(_cube,)^l_0 z_link(_link)^l_0 v_3 .
the partition function @xmath298 of the spacelike degrees of freedom reads [ su2.1.zcube1 ] z_cube = _
x_0^(3 ) ( d((x ) ) e^-s^((x ) ) _ = 1 ^ 3 d_h(u(x ; ) ) ) , with @xmath299 the 3-dimensional lattice and [ su2.1.zcube2 ] s_cube(u , ) & = & - _ x_0^(3 ) \ { _ = 1 ^ 3 4 _ cube _ 0(u(x ; ) ) + _ 0((x ) ) + & & + _ = 1 ^ 3 ( 2 ) } , while @xmath294 is an ultralocal one - link integral [ su2.1.z0 ] z_link = d_h(u ) with [ su2.1.s0 ] s_link(u ) = 4 _ link _ 0(u ) .
expectation values @xmath300 , @xmath301 , @xmath302 of observables @xmath303 refer to @xmath298 , @xmath294 and @xmath297 , respectively .
minimization of the trial free energy density @xmath304 in terms of these expecation values leads to three equations .
the first one @xmath305 is solved by @xmath306 in the symmetric phase .
the remaining two equations are given by [ su2.1.mfeq1 ] & & + | ( < _ 0(u ) > _ cube)^3 + ( < _ 0(u ) > _
link)^2 < _ 0(u ) > _ cube - _ cube + & & < _ 0(u ) > _ cube = 0 and [ su2.1.mfeq2 ] ( < _ 0(u ) > _ cube)^2 < _ 0(u ) > _ link - _ link < _ 0(u ) > _ link = 0 .
here we have used the shorthand notation [ su2.1.plaq ] < _ 0(u ) > _ cube & & < _ 0(u(x;1 ) ) > _ cube , + _ cube^1 , 2 & & < tr u(x;1 ) u(x+;2 ) u(x+;1)^-1 u(x;2)^-1 > _ cube + & - & ( < _ 0(u(x;1))>_cube)^4 , with @xmath307 . in ( [ su2.1.plaq ] ) we have used the lattice symmetries .
eq.s ( [ su2.1.mfeq1 ] ) , ( [ su2.1.mfeq2 ] ) should be solved for @xmath270 , @xmath271 , and @xmath272 as series in @xmath273 .
the stability condition for the symmetric minimum at @xmath306 is given as [ su2.1.stab ] & & 3 _ = 0 + & & + 12 | ( < _ 0(u ) >
_ cube)^3 + < _ 0(u ) > _ cube ( < _ 0(u ) >
_ link)^2 - _ cube + & & < _ 0(u ) > _ cube _
= 0 + & & + ( 2 ) 2 < _ 0(u ) > _ link < _ 0 ( ) > _ cube _
= 0 - 1 < _ 0 ( ) > _ cube _ = 0 + & & + & & < 0 , with < _ 0 ( ) > _ cube < _ 0((x ) ) > _ cube , @xmath307 . for an equality sign , ( [ su2.1.stab ] ) determines @xmath277 order by order in the expansion .
note the two type of expectation values @xmath308 , @xmath309 entering eq.s ( [ su2.1.mfeq1 ] ) , ( [ su2.1.mfeq2 ] ) , ( [ su2.1.stab ] ) .
while the single link expectation values @xmath308 can be evaluated exactly , the @xmath309s must be approximated .
the derivatives of @xmath310 w.r.t .
@xmath271 or twice w.r.t .
@xmath272 induce up to 5-point functions in @xmath20 and 6-point functions in four @xmath20s and two @xmath185s .
because of the bad signal / noise ratio monte carlo calculations of such connected correlations would not be feasible .
therefore this is the place , where we use a dlce to evaluate expectation values of the type @xmath311 ( all internal and configuration space indices suppressed ) order by order in @xmath273 . clearly , also in this scheme the price is high .
for example , the number of connected dlce graphs contributing to order @xmath312 to a 4-point function of 2 @xmath20s and 2 @xmath185s is about 100 .
not all correlations in ( [ su2.1.mfeq1]),([su2.1.mfeq2]),([su2.1.stab ] ) appear in the form of susceptibilities . in a product of 4 @xmath20s , for instance , the configuration space indices are fixed to the boundary of a plaquette .
such features must be noticed for calculating the graphical weights .
these remarks may indicate the complexity of the actual evaluation of eq.s ( [ su2.1.mfeq1 ] ) , ( [ su2.1.mfeq2 ] ) , ( [ su2.1.stab ] ) . here
we omit any further details on the list of contributing graphs and postpone them to @xcite .
before we quote some results , we comment on further subtleties , when dlces are adapted to internal symmetries of the hopping term .
in this section we focus on the ( gauge ) group @xmath313 , since we are interested in applications to the electroweak phase transition , but it should be obvious , how one could proceed along the same lines for a group @xmath314 with @xmath315 . in the development of the theory in sections 3 - 5
we had assumed that @xmath316 , @xmath317 to simplify the notation . in eqs .
( [ su2.uparam ] ) and ( [ su2.phiparam ] ) we had @xmath318 , @xmath319 .
let us write the hopping part of the action in the symmetrized form [ su2.hop1 ] s_hop = - _ x , y_0 v(x , y ) ( x)^u(x , y ) ( y ) , with hopping parameter v(x , y ) = 2 _ = 0 ^ 3 ( _ y , x+ + _ y , x- ) and c@=c u(x , x+ ) & u(x ; ) + u(x+,x ) & u(x;)^-1 + u(x , y ) & 0 , = 0, ,3 , using the parametrizations ( [ su2.uparam ] ) and ( [ su2.phiparam ] ) we have [ su2.hop2 ] & & ( x)^u(x , y ) ( y ) = ( x ) ( y ) u_0(x , y ) + & & + _ 0(y ) ( x ) ( x , y ) - _ 0(x ) ( y ) ( x , y ) + & & + ( x ) ( y ) ( x , y ) . note that in the parametrization in terms of @xmath320s and @xmath321s the @xmath22 part of the action is o(4 ) invariant , but @xmath296 is only o(3 ) invariant for @xmath322 or @xmath323 .
thus we have to distinguish between singlet @xmath324 and triplet @xmath325 indices in internal space .
recall the generating equations ( [ 2.genequation ] ) of the graphical expansion of a dlce in which @xmath326 had been expressed in terms of connected @xmath204-point functions , @xmath327 . using the @xmath328 symmetry it follows from eq .
( [ su2.hop2 ] ) that each derivative @xmath329 now has to be replaced by the differential operator [ su2.lab ] l_ab(x , y ) = _ ab + _ = 1 ^ 3 ( _ a , _ b,0 - _ b , _
a,0 + _ ab ) .
latin indices run from @xmath330 to @xmath331 , greek indices from @xmath93 to @xmath331 .
furthermore we have introduced the totally antisymmetric @xmath332-symbol which is zero if any of the three indices is zero and @xmath333 .
this implies for instance for the one nonvanishing term at @xmath32 in eq .
( [ 2.genequation ] ) a substitution according to [ su2.ww ] w_h(x)w_h(y)w_i(x , y ) w_h^a(x)w_h^b(y ) l_ab(x , y)w . according to the graphical rules
an @xmath204-line was generated by @xmath204 iterated derivatives w.r.t .
@xmath334 , evaluated at @xmath32 .
these derivatives now have to be replaced by a multiple application of @xmath113 yielding an n - line [ su2.labgraph ] ( 5.0,1.0 ) ( 0.0,-0.6 ) ( 4.0,0.0 ) ( 1.0,1.5)(2.0,2.1)(3.0,1.5 ) ( 1.0,1.0)(2.0,1.6)(3.0,1.0 ) ( 1.0,0.0)(2.0,0.6)(3.0,0.0 ) ( -0.2,0.7)(0.2,0)@xmath39 ( 4.2,0.7)(0.2,0)@xmath41 ( 0.6,1.5)(0.2,0)@xmath335 ( 3.3,1.5)(0.2,0)@xmath336 ( 0.6,1.0)(0.2,0)@xmath337 ( 3.3,1.0)(0.2,0)@xmath338 ( 0.6,0.0)(0.2,0)@xmath339 ( 3.3,0.0)(0.2,0)@xmath340 ( 1.6,1.1)(0.2,0)@xmath48 ( 1.6,0.8)(0.2,0)@xmath48 ( 1.6,0.5)(0.2,0)@xmath48 ( 2.0,1.85)(0,1)0.2 ( 2.0,1.6)(0,1)0.2 ( 2.0,1.35)(0,1)0.2 ( 2.0,1.1)(0,1)0.2 ( 2.0,0.85)(0,1)0.2 ( 2.0,0.6)(0,1)0.2 ( 2.0,0.35)(0,1)0.2 ( 2.0,0.1)(0,1)0.2 = _ = 0 ^ 3 ( _ y , x+ + _ y , x- ) ( 2)^n .
( _ i=1^n l_a_ib_i ) w^1(i ) _ i=0 where the @xmath93-link generating function @xmath223 is defined by = . for @xmath341 for all @xmath342 , @xmath204 applications of @xmath343 to @xmath344 reduce to @xmath345 ,
but otherwise we get a complicated mixing of singlet and triplet terms .
to evaluate products of @xmath113s systematically , it is again quite convenient to manipulate graphical rather than analytic expressions .
we represent one application of @xmath343 on @xmath344 as shown in fig .
( [ lab ] ) .
multiple applications of @xmath343 require an iteration of fig .
( [ lab ] ) .
( 15.0,4.0 ) twiddle lines are bare @xmath19-lines with a singlet index @xmath346 , crossed lines are bare @xmath19-lines with a triplet index @xmath347 .
the arrow in the last term stands for the @xmath348 symbol in @xmath343 , its orientation reflects the order of indices @xmath349 . this way one achieves a further graphical `` decomposition '' of an n - line into terms merely consisting of bare lines , from which the internal symmetry factors can be easily read off .
for example a 2-line is further decomposed into 11 terms , a 3-line into 29 and a 4-line into 142 .
this may indicate that the number of decomposed contributions rapidly increases with the number @xmath204 of bare lines collected to an n - line . in our actual calculations
up to order @xmath312 the maximal number of internal lines was 4 and the highest multiple line was a 6-line with 4 internal and 2 external u - lines . some of the results will be reported in the next section . in table 1
we compare results for @xmath277 obtained for 2 values of @xmath350 and @xmath351 and various lattice volumes within three methods .
the first line refers to a naive molecular field approximation with a partition function @xmath293 factorizing according to eq .
( [ su2.1.zve0 ] ) . because of the complete factorization in space and time directions the ansatz keeps only a trivial volume dependence and is temperature independent by construction .
the second line refers to monte carlo simulations of the su(2 ) higgs model in four dimensions @xcite for two lattice extensions @xmath352 and @xmath331 .
the third line shows results of the dlces with the variational ansatz ( ve1 ) presented in section 7.2 . again
this ansatz is independent of @xmath278 by construction , so that we get an identical result for @xmath353 .
the values for @xmath277 have been obtained by a linear regression in @xmath354 , @xmath355 denoting the order in @xmath273 of the dlce . for the regression we used the coefficients of @xmath356 , @xmath357 , @xmath358 and @xmath312 .
the value of @xmath277 at order @xmath312 was given by @xmath359 for @xmath360 and @xmath361 .
although the fourth order seems to be low for a series expansion , the agreement with the monte carlo results is not too surprising , if we remind the complexity of the expansion , which is manifest in the number of graphs contributing at fourth order to the various expectation values .
furthermore , for otherwise fixed parameters , we see a decrease of @xmath277 between the monte carlo and the dlce results for increasing @xmath278 .
this is in qualitative agreement with the expectation .
.[kc.1 ] results for @xmath277 for 2 values of @xmath362 and @xmath351 , various lattice extensions , and 3 methods .
the monte carlo results are taken from [ 15 ] , the mean field and dlce results , extrapolated to infinite order , from this paper .
( the error in the dlce results refers to the extrapolation of @xmath277 . )
[ cols=">,^,^ , < , < " , ] the partition function @xmath363 factorizes acording to [ su2.zveii ] z_ve2(_(0,string),_(s , link ) , ) = z_(0,string)(_(0,string),)^v_3 z_(s , link)(_(s , link))^3 l_0
v_3 , i.e. , a product of a partition function @xmath364 that couples fields with hopping parameter terms along a string in time direction , and @xmath365 , a partition function for a single spatial link . while expectation values w.r.t . @xmath365 can be evaluated exactly ( within the numerical accuracy ) , expectation values w.r.t .
@xmath364 are preferably evaluated with dlce . this way they do depend on @xmath278 , in principle they are sensitive to detect a temperature effect on @xmath277 .
so far the couplings along a string are only kept in time direction , while the spacelike directions are incorporated on a mean field level in @xmath365 .
the third variational ansatz ve3 therefore is chosen as a symmetrized version of ve2 in that the string now can extend in any of the 4 directions .
again we treat space - and timelike degrees of freedom differently , because we are interested in systems at finite temperature with temperature dependent coupling @xmath277 .
the number of variational parameters is larger , each of it depending on two indices telling us whether the parameter belongs to a single link action or to a string , and whether the string extends in temporal or spacelike directions . for both versions ve2 and
ve3 the actual @xmath278 dependence ( @xmath366 ) lies outside the systematic error , if we expand the series to order @xmath312 . for @xmath279
the fourth order is the minimal order in @xmath273 at which a finite temperature effect is visible in principle .
( for the graphs it should be possible to wind at least once around the torus in @xmath330-direction . ) at o@xmath367
the part of graphs for which the lattice embedding is sensitive to @xmath278 is about @xmath368 of the total number of graphs .
such a small `` signal@xmath369noise '' ratio produces an effect in the 7th or 8th digit , outside the error of the @xmath370 because of the truncation of the dlce series . by increasing the order of the expansion
, finite - temperature effects will become more pronouced . for higher orders an algorithmic implementation of the proliferating graphs becomes unavoidable .
in this paper we have introduced a new expansion scheme for 3-point interactions or , more precisely , for point - link - point interactions .
this scheme generalizes linked cluster expansions for 2-point interactions by including hopping parameter terms endowed with their own dynamics . in chapters 3 - 5 we have developed a theory with an additional new type of multiple - line connectivity .
we have introduced appropriate equivalence classes of graphs and discussed the issue of renormalization .
the main building blocks for an algorithmic generation of graphs have been constructed .
because of the fast proliferation of graphs already at low orders in the expansion , a computer aided implementation becomes unavoidable , if one is interested in higher orders of the expansion than we have computed so far . in chapter 6
we have indicated promising applications to spin glass systems .
in particular we have identified the dlce graphs contributing to the link - average @xmath371 $ ] of the spin free energy @xmath372 . in the past
such quantities often were only accessible by means of the replica trick . in section 7
we have outlined encouraging results for the transition line of the electroweak phase transition within the su(2)-higgs model .
the results have been obtained by dlces applied to gap equations that follow from variational estimates of the free energy .
they are in good agreement with corresponding high precision monte carlo results .
as we have seen , the main complications of dlces for su(n ) higgs systems come from the internal symmetry structure of the hopping term . in a forthcoming paper
we will demonstrate how one has to refine the graphical representation to calculate internal symmetry factors merely within a graphical expansion . in conclusion ,
dlces are involved , but practicable , at least with computer aided generation of graphs .
they provide an analytical tool to study systems in situations in which it has been impossible so far .
we would like to thank reimar khn ( heidelberg ) for pointing out to us ref.s @xcite-@xcite .
9 m. wortis , `` linked cluster expansion '' , in phase transition and critical phenomena , vol.3 , eds . c. domb and m.s .
green ( academic press , london 1974 ) . c. itzykson , j .-
drouffe , `` statistical field theory '' , vol.2 , cambridge university press , 1989 .
guttmann , `` asymptotic analysis of power - series expansions '' , in phase transition and critical phenomena , vol.13 , eds . c. domb and j.l .
lebowitz ( academic press ) .
m. lscher and p. weisz , nucl . phys .
* b300*[fs22 ] ( 1988 ) 325 .
t. reisz , nucl . phys .
* b450 * ( 1995 ) 569 .
t. reisz , phys . lett .
* 360b * ( 1995 ) 77 .
m. campostrini , a. pelissetto , p. rossi and e. vicari , nucl . phys .
* b459 * ( 1996 ) 207 .
s. zinn , s .- n . lai and m. e. fisher , phys .
* e54 * ( 1996 ) 1176 .
p. butera and n. comi , phys .
* e55 * ( 1997 ) 6391 .
h. meyer - ortmanns and t. reisz , jour . stat
* 87 * ( 1997 ) 755 .
a. pordt and t. reisz , int .
* a12 * ( 1997 ) 3739 - 3757 .
d. sherrington and s. kirkpatrick , phys .
lett * 35 * ( 1975 ) 1972 .
v. dotsenko , s. franz and m. mzard , j. phys .
* a27 * ( 1994 ) 2351 .
r. w. penney , a. c. c. coolen , and d. sherrington , j. phys . a : math .
* 26 * ( 1993 ) 3681 .
a. c. c. coolen , r. w. penney , and d. sherrington , phys . rev .
* b48 * ( 1993 ) 16116 .
z. fodor , j. hein , k. jansen , a. jaster and i. montvay , nucl .
* b439 * ( 1995)147 .
t. reisz , `` hopping parameter series construction for models with nontrivial vacuum '' , hep - lat/9802023 , submitted to nucl .
b. k. h. fischer and j. a. hertz , `` spin glasses '' , cambridge university press , cambridge 1991 .
j. zinn - justin , `` quantum field theory and critical phenomena '' , 3rd edition , oxford science publication , clarendon press , oxford 1996 .
h. meyer - ortmanns and t. reisz , in preparation . | dynamical linked cluster expansions are linked cluster expansions with hopping parameter terms endowed with their own dynamics .
this amounts to a generalization from 2-point to point - link - point interactions .
we develop an associated graph theory with a generalized notion of connectivity and describe an algorithmic generation of the new .
we indicate physical applications to spin glasses , partially annealed neural networks and su(n ) gauge higgs systems .
in particular the new expansion technique provides the possibility of avoiding the replica - trick in spin glasses .
we consider variational estimates for the su(2 ) higgs model of the electroweak phase transition .
the results for the transition line , obtained by dynamical linked cluster expansions , agree quite well with corresponding high precision monte carlo results . |
let @xmath0 be a non - cyclic finite group .
we write @xmath1 for the smallest number of conjugacy classes of proper subgroups of @xmath0 needed to cover it . in other words
, @xmath1 is the least @xmath8 for which there exist subgroups @xmath9 such that @xmath10 we say that the set of conjugacy classes @xmath11 is a _
normal covering _ for @xmath0 .
bubboloni and praeger @xcite have recently investigated @xmath1 in the case that @xmath0 is a finite symmetric or alternating group .
they show , for example , that if @xmath4 is an odd composite number then @xmath12 where @xmath13 is euler s totient function .
similar results are established for all values of @xmath4 , and for both @xmath14 and @xmath15 .
part of the motivation for their work comes from an application in number theory .
it is a well - known theorem of jordan that no finite group is covered by the conjugates of any proper subgroup .
a paraphrase of this statement is that @xmath16 for any finite group @xmath0 .
it is known that there exists a solvable group @xmath0 with @xmath17 for any @xmath18 @xcite .
it has been shown in @xcite that if @xmath19 then @xmath20 if and only if @xmath21 .
( notice that @xmath22 is undefined for @xmath23 , since the groups are cyclic in this case . )
other groups of lie type possessing a normal covering of size @xmath24 have been studied in @xcite and @xcite . in this paper
we give bounds on @xmath1 , where @xmath25 , for all values of @xmath4 . in some cases we are able to give an exact value .
our bounds extend without change to @xmath26 .
we introduce some notation .
we write @xmath27 for the integer part of a real number @xmath28 .
as already noted above , @xmath13 denotes euler s function .
we shall also use lehmer s _ partial totient function _ , which we define here .
let @xmath8 and @xmath29 be such that @xmath30 .
we define the partial totient @xmath31 to be the number of integers @xmath28 , coprime with @xmath4 , such that @xmath32 we give two separate upper bounds on @xmath1 .
[ thm : upperbounds ] let @xmath33 , and let @xmath34 be the number of prime factors of @xmath4 .
let @xmath35 be the distinct prime factors of @xmath4 , with @xmath36 .
let @xmath0 be a group such that @xmath25
. then 1 .
if @xmath37 then @xmath38 2 . if @xmath39 then @xmath40 a great deal of information is given in @xcite about the function @xmath41 , from which the following statement can be derived .
@xmath42 in which @xmath43 , where @xmath44 is the number of prime divisors of @xmath4 counted with multiplicity .
let @xmath45 be the size of the largest set of conjugacy classes of @xmath0 such that any pair of elements from distinct classes generates @xmath0 .
we call such a set an _ independent set of classes_. guralnick and malle @xcite have shown that @xmath46 for any finite simple group @xmath0 .
it is clear that whenever @xmath1 is defined , we have the inequality @xmath47 since if @xmath48 is a normal covering of @xmath0 , and if @xmath49 is a independent set of classes , then each element of @xmath48 covers at most one element of @xmath49 .
we establish two lower bounds for @xmath45 . by the observation of the previous paragraph ,
these also operate as lower bounds for @xmath1 .
[ thm : lowerbounds ] let @xmath33 , and let @xmath34 be the number of prime factors of @xmath4 .
let @xmath35 be the distinct prime factors of @xmath4 , with @xmath36 .
let @xmath0 be a group such that @xmath25 . 1 .
if @xmath50 then @xmath51 2 .
if @xmath52 , and if @xmath4 is not equal to @xmath53 or @xmath54 for any prime @xmath55 , then @xmath56 furthermore , if @xmath57 then @xmath58 the values @xmath59 are not amongst those for which the function @xmath60 is evaluated explicitly in @xcite
. however , theorem 10 of @xcite gives the following general estimate , @xmath61 where @xmath62 is the number of prime divisors of @xmath4 .
this yields the lower bound @xmath63 there are certain cases in which an upper bound for @xmath1 coincides with a lower bound for @xmath45 . in these cases
we must have @xmath64 , and we obtain a precise formula . [ thm : equalities ] let @xmath0 be a group such that @xmath25 . 1 . if @xmath65 , where @xmath55 is a prime and @xmath66 , and if @xmath6 , then @xmath67 2 .
if @xmath68 where @xmath55 and @xmath29 are distinct primes and @xmath69 , then @xmath70 3 . if @xmath71 where @xmath55 is a prime , then @xmath72 .
4 . if @xmath73 where @xmath55 is a prime , then @xmath74 .
certain cases of theorem [ thm : equalities ] will require independent treatment , as they arise as exceptional cases in the proof of theorem [ thm : lowerbounds ] .
theorem [ thm : upperbounds ] ( 1 ) , theorem [ thm : lowerbounds ] ( 2 ) , and theorem [ thm : equalities ] , taken together , imply that @xmath75 for all @xmath6 .
the upper bound is exact when @xmath4 is an odd prime .
( when @xmath76 it is known that @xmath20 ; see @xcite , or the remark after proposition [ prop : pa ] below .
it is also easy to show that @xmath77 in this case . )
it follows immediately that @xmath78 the lower bound for @xmath22 can be improved , as the following theorem indicates .
[ thm : pibound ] let @xmath0 be a group such that @xmath25
. then @xmath79 . from the first part of theorem [ thm : upperbounds ] and from theorem
[ thm : pibound ] , it is easy to show that @xmath80 it follows from the theorems which we have stated , that @xmath1 and @xmath45 are bounded above and below by monotonic functions which grow linearly with @xmath4 .
it appears that the situation for symmetric groups is similar .
it has been announced in @xcite , and will be demonstrated in a forthcoming paper @xcite now in preparation , that @xmath2 and @xmath81 are bounded above and below by linear functions of @xmath4 .
in fact the numbers @xmath2 and @xmath82 seem to be closely related ; in all cases where both are known exactly , they differ by at most @xmath83 . it is not hard to show , and it is worth remarking in this connection , that the upper bounds stated for @xmath1 in theorem [ thm : upperbounds ] are also upper bounds for @xmath2 , improving marginally on those of ( * ? ? ?
* theorem a ) .
it should also be noted that all of our bounds are independent of the field size @xmath29 .
we establish the upper bounds of theorem [ thm : upperbounds ] in section [ sec : coverings ] , by exhibiting explicit normal coverings of the necessary sizes .
this builds on work described in @xcite , in which coverings of @xmath84 by proper subgroups are constructed .
the two lower bounds of theorem [ thm : lowerbounds ] are proved in section [ sec : lowerbounds ] .
both are proved by exhibiting an independent set of classes .
this requires an account of overgroups of certain special elements in @xmath84 .
for such an account we rely on @xcite , which provides a classification of subgroups whose orders are divisible by primitive prime divisors of @xmath85 , for all @xmath86 .
the remaining cases of theorem [ thm : equalities ] are brought together in section [ sec : equalities ] . finally , theorem [ thm : pibound ]
is established in section [ sec : pibound ] .
its proof relies on work from the doctoral thesis of joseph dimuro @xcite , which extends the classification of @xcite to cover all @xmath87 . the classes of subgroups in our normal covering remain distinct , proper and non - trivial in the quotient of @xmath0 by @xmath88 .
this is true also of the classes of maximal overgroups which cover the conjugacy classes in our independent sets .
it follows that the bounds which we have stated for @xmath1 and for @xmath45 hold equally for @xmath89 and for @xmath90 .
we shall write @xmath91 for the space @xmath92 . throughout the paper , we assume that @xmath93 .
we begin by introducing the classes of subgroups which we shall need for our coverings .
proposition [ prop : efssss ] below contains standard information about certain subgroups of @xmath84 , and we shall not prove it here .
[ prop : efssss ] 1 .
let @xmath94 be a divisor of @xmath4 .
there exist embeddings of @xmath95 into @xmath84 .
all such embeddings are conjugate by elements of @xmath96 , and each has index @xmath94 in its normalizer in @xmath97 .
if @xmath94 is prime then the normalizer is a maximal subgroup of @xmath84 .
2 . suppose that @xmath98 , and let @xmath99 be a @xmath8-dimensional subspace of @xmath91 .
then the set stabilizer @xmath100 of @xmath99 in @xmath0 is a maximal subgroup of @xmath0 .
if @xmath101 is another @xmath8-dimensional subspace , then @xmath100 and @xmath102 are conjugate in @xmath0
. it will be convenient to have concise notation for these subgroups . 1 .
we refer to the maximal subgroups of proposition [ prop : efssss ] ( 1 ) as _ extension field subgroups _ of degree @xmath94 , and we write @xmath103 for the conjugacy class consisting of the intersections of all such subgroups with the group @xmath0 .
we refer to the subgroups of proposition [ prop : efssss ] ( 2 ) as _ subspace stabilizers _ of dimension @xmath8 , and we write @xmath104 for the conjugacy class consisting of all such subgroups
. the following technical lemma will be useful .
[ lem : subgroups ] 1 .
suppose that @xmath105 , and that @xmath106 stabilizes a @xmath8-dimensional subspace of @xmath91 .
then @xmath106 stabilizes a subspace whose dimension is @xmath107 .
2 . let @xmath105 , and let @xmath55 be a prime dividing @xmath4 . if @xmath106 lies in no extension field subgroup of degree @xmath55 , then it stabilizes a subspace of @xmath91 whose dimension is coprime with @xmath55
. suppose @xmath106 stabilizes a space @xmath99 of dimension @xmath8 .
then the transpose @xmath108 acts on the dual space @xmath109 , and stabilizes the annihilator of @xmath99 , which has dimension @xmath107 .
2 . if @xmath106 stabilizes no subspace whose dimension is coprime with @xmath55 , then every irreducible divisor of its characteristic polynomial has degree divisible by @xmath55 , and must therefore split into @xmath55 factors over @xmath110 .
suppose that the elementary divisors of @xmath106 are @xmath111 .
for each @xmath112 , let @xmath113 be an irreducible factor of @xmath114 over @xmath110 , and let @xmath115 have elementary divisors @xmath116 .
then it is not hard to see that any embedding of @xmath117 into @xmath84 must map @xmath118 to a conjugate of @xmath106 .
we are now in a position to exhibit some normal coverings of @xmath0 .
[ lem : coverings1 ] 1 .
let @xmath55 be a prime dividing @xmath4 .
then there is a normal covering @xmath119 for @xmath0 given by @xmath120 the size of @xmath119 is @xmath121 where @xmath122 this is minimized when @xmath55 is the smallest prime divisor of @xmath4 .
2 . let @xmath123 and @xmath124 be distinct prime divisors of @xmath4 .
then there is a normal covering @xmath125 for @xmath0 given by @xmath126 the size of @xmath125 is @xmath127 this is minimized when @xmath123 and @xmath124 are the two smallest prime divisors of @xmath4 .
the sizes of the sets @xmath119 and @xmath125 are easily seen to be as stated .
that @xmath119 is a covering follows immediately from lemma [ lem : subgroups ] .
so it remains only to prove that @xmath128 is a covering .
let @xmath129 , let @xmath130 be the characteristic polynomial of @xmath106 , and let @xmath131 be the irreducible factors of @xmath130 over @xmath132 , with degrees @xmath133 respectively
. then clearly there exist @xmath106-invariant subspaces @xmath134 such that @xmath135 for all @xmath112 , and such that @xmath136 whenever @xmath137 .
if any @xmath138 is divisible by neither @xmath123 nor @xmath124 , then @xmath106 is contained in a subspace stabilizer from one of the classes in @xmath125 .
so we assume that each @xmath138 is divisible by at least one of @xmath123 or @xmath124 .
suppose that @xmath139 is divisible by @xmath123 but not @xmath124 , and that @xmath140 is divisible by @xmath124 but not @xmath123 .
then @xmath141 is an @xmath106-invariant subspace , and its dimension is coprime with @xmath123 and @xmath124 ; so again , @xmath106 is in a subspace stabilizer from @xmath125 .
but if no such @xmath139 and @xmath140 can be found , then either all of the @xmath138 are divisible by @xmath123 , or they are all divisible by @xmath124 . in this case
, @xmath106 lies in an extension field subgroup either of degree @xmath123 or of degree @xmath124 .
we note that the argument of the last paragraph of this proof does not extend to the case of three primes , @xmath142 .
it is possible to find matrices whose invariant subspaces all have dimensions divisible by one of those primes , but which lie in no extension field subgroup . in the case that the primes are @xmath143 and @xmath144 , for instance , there are @xmath145-dimensional matrices whose irreducible invariant spaces have dimensions @xmath24 , @xmath146 and @xmath147 .
( another example is used in the proof of proposition [ prop:10p ] below . )
this is the explanation for the appearance of the two smallest prime divisors of @xmath4 in the first upper bound of theorem [ thm : upperbounds ] , which may at first seem a little curious .
the second upper bound of theorem [ thm : upperbounds ] is proved in a somewhat similar fashion .
[ lem : coverings2 ] let @xmath148 be the distinct primes dividing @xmath4 . then there is a normal covering @xmath149 of @xmath0 given by @xmath150 for @xmath39 , the size of @xmath149 is @xmath151 let @xmath129 .
suppose that @xmath106 is reducible , and that its smallest non - trivial invariant subspace has dimension @xmath8 . if @xmath152 then it is not hard to see ( for instance , by considering the irreducible factors of the characteristic polynomial ) that @xmath106 stabilizes at most one other proper non - trivial subspace , of dimension @xmath107 .
it follows that if @xmath55 is a prime dividing both @xmath4 and @xmath8 , then @xmath106 is contained in an element of @xmath153 .
it is now a straightforward matter to show that @xmath149 is a normal covering , and we omit further details .
the size of @xmath149 follows immediately from its definition .
recall that @xmath84 contains elements of order @xmath154 , often known as singer elements .
such elements stabilize no non - trivial proper subspace of @xmath91 .
the determinant of a singer element generates the multiplicative group of @xmath132 . in order to handle all groups @xmath0 in the range @xmath25 together , we define a parameter @xmath155 as follows .
@xmath156 let @xmath157 be a generator of the multiplicative group of @xmath132 .
then we have @xmath158 1 . for @xmath159 ,
let @xmath160 be a singer element with determinant @xmath157 in @xmath97 .
2 . for @xmath161 , define @xmath162 3 . for @xmath163 ,
define @xmath164 the reasons for defining @xmath165 as above will be clear from the following remark . 1 . since @xmath166 , we have @xmath167 .
it is clear from the definition of @xmath165 that @xmath168 , and hence that @xmath169 .
it follows easily that the actions of the matrices @xmath170 and @xmath171 are irreducible for all @xmath8 and @xmath172 .
therefore the module @xmath173 decomposes into precisely two irreducible summands , and @xmath174 decomposes into precisely three irreducible summands .
[ lem : overgroups ] suppose that @xmath175 .
let @xmath176 , and if @xmath177 then suppose that @xmath178 . let @xmath179 , and if @xmath177 then suppose that @xmath180 . 1 . if @xmath181 is a maximal subgroup of @xmath0 containing @xmath182 then @xmath181 is an extension field subgroup of prime degree . 2 .
if @xmath181 is a maximal subgroup of @xmath0 containing @xmath183 then @xmath181 is either an extension field subgroup whose degree is a prime divisor of @xmath184 , or else the stabilizer of a subspace of dimension @xmath8 or @xmath107 .
3 . let @xmath4 have at least @xmath146 distinct prime divisors .
if @xmath181 is a maximal subgroup of @xmath0 containing @xmath185 , then @xmath181 is the stabilizer of a subspace whose dimension is one of @xmath172 , @xmath186 , @xmath187 , @xmath188 , @xmath189 or @xmath190 . for all but one of the groups @xmath182 and @xmath183 under consideration , this is given in theorem 4.1 of @xcite ; though the result is stated there only for the groups @xmath84 and @xmath96 , the proof applies equally to intermediate subgroups .
the matrices @xmath182 and @xmath183 are there referred to as @xmath191 and @xmath192 respectively . for @xmath182 ,
the result is essentially that of kantor @xcite . for @xmath183 ,
the proof in @xcite relies on the existence of primitive prime divisors of @xmath193 , which is given by zsigmondy s theorem @xcite for all pairs @xmath194 except @xmath195 and @xmath196 ; the second of these exceptions accounts for the excluded case in the statement of the present lemma .
the argument uses the classification in @xcite , of subgroups of @xmath84 whose order is divisible by a prime divisor of @xmath197 , where @xmath198 .
the exceptional case which is not covered in @xcite is the group @xmath199 . in this case
we require a reference directly to the lists of @xcite .
we find that there are several irreducible subgroups whose order is divisible by the primitive prime divisor @xmath200 of @xmath201 ; we must show that none of these contains @xmath202 .
all of these subgroups are almost simple , and have a socle which is isomorphic either to one of the mathieu groups @xmath203 or @xmath204 , or to the unitary group @xmath205 , or to a linear group @xmath206 or @xmath207 .
( these subgroups may be found in table 5 ( lines 12 and 14 ) and table 8 ( lines 2 , 7 and 9 ) of @xcite . )
information about these groups may be found in @xcite . neither any of the groups themselves , nor any of their outer automorphism groups , have order divisible by @xmath208
. therefore an almost simple group of one of these types can contain no element of order @xmath209 , which is the order of the element @xmath202 . for the groups
@xmath185 we refer once again to the classification of @xcite , this time for matrix groups whose order is divisible by a primitive prime divisor of @xmath210 .
it is not hard to see that @xmath185 has no overgroups of classical type .
the condition that @xmath4 has @xmath146 prime divisors rules out the small dimensional sporadic examples contained in tables 17 .
other examples are ruled out because their order is less than @xmath210 , which is the order of the summand @xmath211 of @xmath185 .
we define a set of classes which will help us to establish a first lower bound for @xmath45 .
define a set @xmath212 of classes of @xmath0 by @xmath213\mid p|n,\ p\textrm{\ prime,\ } p < n/2\}\,\cup\ , \{\left[\sigma_k\right ] \mid k < n/2,\ \operatorname{hcf}(n , k)=1\},\ ] ] where @xmath214 $ ] denotes the conjugacy class of @xmath215 .
[ lem : setsize1 ] let @xmath6 , and let @xmath216 be the number of prime factors of @xmath4
. then @xmath217 where @xmath218 this is immediate from the definition of @xmath212 .
lemma [ lem : setsize1 ] , together with the following two lemmas , will imply the first part of theorem [ thm : lowerbounds ] .
[ lem : nocommonovergroups1 ] @xmath212 is an independent set of classes .
suppose that @xmath219 , or that @xmath220\notin \phi$ ] .
then lemma [ lem : overgroups ] provides full information about the maximal subgroups of @xmath0 which contain elements of @xmath212 , and it is easy to check that the result holds in this case .
next suppose that @xmath177 and @xmath220\in \phi$ ] .
( this implies that @xmath221 . ) lemma [ lem : overgroups ] gives full information about the maximal subgroups of @xmath0 covering elements of the classes in @xmath212 other than @xmath220 $ ] .
no class of subgroups contains elements of more than one such class , and it is easy to check that none covers the element @xmath222 itself . [
lem : lowerboundexception ] let @xmath223 where @xmath224 is a prime . then @xmath225 .
the proof of lemma [ lem : nocommonovergroups1 ] shows that in any normal covering of @xmath0 , the distinct classes in @xmath212 are covered by distinct classes of subgroups .
we add an extra conjugacy class to @xmath212 , namely the class represented by @xmath226 , where @xmath227 is a singer element in @xmath228 .
this element stabilizes no subspace of dimension @xmath8 for any @xmath8 coprime with @xmath4 ; nor does it stabilize a subspace of dimension @xmath24 or @xmath229 . therefore , by the second part of lemma [ lem : overgroups ] , if there is a covering of @xmath0 of size @xmath230 , then @xmath231 must lie in a subgroup in @xmath232 .
note that since @xmath24 and @xmath55 are coprime , @xmath233 has two irreducible summands of dimension @xmath55 .
it is not hard to show that these submatrices are not conjugate , and neither of them is reducible over @xmath234 ; it follows that @xmath233 is not contained in any embedding of @xmath235 into @xmath0 .
hence @xmath231 itself is not contained in an embedding of .
this completes the proof of the first part of theorem [ thm : lowerbounds ] .
we define a second independent set of classes which yields the second lower bound of theorem [ thm : lowerbounds ]
. we shall require the following lemma .
[ lem : bertrand ] let @xmath55 be a prime divisor of @xmath4 .
suppose that @xmath4 has at least @xmath146 distinct prime divisors , and that @xmath4 is not equal to @xmath236 or @xmath237 for any prime @xmath29 .
then there exists an integer @xmath238 such that @xmath239 , and such that @xmath238 is divisible by @xmath55 , and by no other prime divisor of @xmath4 . if @xmath240 then @xmath238 may be chosen so that it is not divisible by @xmath146 .
bertrand s postulate states that for every @xmath241 there is a prime @xmath242 such that @xmath243 .
the conditions on @xmath4 imply that @xmath244 .
so there is a prime @xmath245 such that @xmath246 if @xmath242 is not itself a prime divisor of @xmath4 , or if it is equal to @xmath55 , then we may take @xmath247 . on the other hand ,
if @xmath242 is a prime divisor of @xmath4 other than @xmath55 then clearly @xmath248 , and since we have assumed that @xmath244 , we have @xmath249 .
now we see that there exists @xmath250 equal either to @xmath251 or to @xmath252 , such that @xmath250 is not divisible by @xmath146 , and we may take @xmath253 .
let @xmath4 be a number with at least @xmath146 prime divisors , and not equal to @xmath53 or @xmath54 for any prime @xmath55 .
we define a set @xmath254 of classes of @xmath0 by @xmath255\mid j<(n-2)/4 , j\equiv 1 \bmod 3 \ } \\ & \quad \cup\ \{\left[\sigma_k\right ] \mid n/4 < k < n/2,\ \operatorname{hcf}(3n , k)=1\}\\ & \quad \cup \{\left[\sigma_{6b}\right ] \mid b < n/12 , \operatorname{hcf}(n,6b)=1\}\\ & \quad \cup\ \{\left[\sigma_{w_p}\right]\mid p|n , \textrm{\ $ p$ prime}\ } , \end{split}\ ] ] where @xmath238 is as constructed in lemma [ lem : bertrand ] , and where @xmath214 $ ] denotes the conjugacy class of @xmath215 .
to describe the size of the set @xmath254 we use lehmer s partial totient function @xmath256 , which was defined before the statement of theorem [ thm : upperbounds ] above . let @xmath4 have @xmath62 distinct prime divisors , where @xmath52 , and suppose that @xmath4 is not equal to @xmath53 or @xmath54 for any prime @xmath55 . 1 .
if @xmath24 or @xmath146 divides @xmath4 , then @xmath257 2 . if @xmath57 , then @xmath258 we write @xmath259 for the least integer not less than @xmath28 . the size @xmath106 of the set @xmath260\mid j<(n-2)/4 , j\equiv 1 \bmod 3 \}$ ] is @xmath261 , where @xmath262 . by examining residues modulo @xmath263 , it is not hard to show that @xmath264 , the first term in our sum .
the set @xmath265 \mid n/4 < k < n/2,\ \operatorname{hcf}(3n , k)=1\}$ ] clearly has size @xmath266 , immediately from the definition of the function @xmath256 .
the set @xmath267 \mid b
< n/12 , \operatorname{hcf}(n,6b)=1\}$ ] is empty if @xmath268 ; otherwise it has size @xmath269 . and clearly the set @xmath270\mid p|n , \textrm{\ $ p$ prime}\}$ ] has size @xmath62 as required . to establish the second lower bound in theorem [ thm : lowerbounds ]
, it will suffice to show that any normal covering for @xmath0 has size at least @xmath271 .
this is done in the following lemma .
[ lem : nocommonovergroups2 ] let @xmath4 have at least @xmath146 distinct prime divisors , and not equal to @xmath53 or @xmath54 for any prime @xmath55 . then @xmath254 is an independent set of classes .
lemma [ lem : overgroups ] describes the maximal subgroups of @xmath0 which contain elements of the classes in @xmath254 .
the elements @xmath185 lie only in members of @xmath272 or @xmath273 , where @xmath274 . notice that if @xmath275 then @xmath276 , and hence @xmath277 .
the elements @xmath183 , where @xmath8 is coprime with @xmath4 , lie only in members of @xmath104 or @xmath278 .
and the elements @xmath279 lie in subspace stabilizers and also in elements of @xmath153 .
it is easy to check that the values permitted for @xmath172 , @xmath8 , @xmath280 and @xmath238 ensure that no two elements of distinct classes in @xmath254 stabilize subspaces of the same dimension
. therefore no two classes in @xmath254 can be covered by a single class of subgroups .
in this section we establish the various claims of theorem [ thm : equalities ] .
we do this simply by comparing upper and lower bounds from earlier parts of the paper .
[ prop : pa ] if @xmath65 , where @xmath55 is a prime and @xmath66 , and if @xmath6 , then @xmath67 lemma [ lem : coverings1 ] and lemma [ lem : nocommonovergroups1 ] together tell us that @xmath281 but it is easy to check , using lemma [ lem : setsize1 ] , that @xmath282 , and that this number is as claimed in the proposition .
if @xmath76 , then the covering @xmath283 has size @xmath24 . since no finite group is covered by a single class of proper subgroups , it follow that @xmath20 in this case . [
prop : paqb ] if @xmath68 where @xmath55 and @xmath29 are distinct primes and @xmath69 , then @xmath284 as in the proof above , lemma [ lem : coverings1 ] with lemmas [ lem : nocommonovergroups1 ] and [ lem : lowerboundexception ] yield that @xmath285 where @xmath286 if @xmath223 ( or @xmath287 ) , and @xmath288 otherwise .
but we see that @xmath289 , with this number being as claimed in the proposition .
[ prop:6p ] if @xmath71 where @xmath55 is a prime , then @xmath290 in this case we have @xmath291 and it is easy to calculate that @xmath292 .
[ prop:10p ] if @xmath73 where @xmath55 is a prime , then @xmath293 if @xmath55 is @xmath24 or @xmath144 then the result follows from proposition [ prop : paqb ] ; if @xmath294 then it follows from proposition [ prop:6p ] .
so we may assume that @xmath295 .
then we have @xmath296 but in this case we see that @xmath297 whereas @xmath298 . to prove that the upper bound is sharp for @xmath45 , it will be sufficient to exhibit an element @xmath118 of @xmath0 which can not be covered by any class of subgroups containing an element of any conjugacy class in @xmath212 .
we define @xmath299 notice that @xmath300 is even , and coprime with @xmath144 and with @xmath55 .
it follows that @xmath118 does not stabilize a subspace of dimension coprime with @xmath4 .
but certainly @xmath118 lies in no extension field subgroup , and so it satisfies the required condition .
for a positive integer @xmath4 , let @xmath301 be the number of partitions of @xmath4 with exactly three parts . by an elementary counting
argument the following formula can be found for @xmath301 .
let @xmath305 be the set of partitions of @xmath4 into three parts having no common divisor greater than @xmath83 .
let @xmath306 .
then we have @xmath307 . by the mbius inversion formula ,
we obtain @xmath308 since @xmath309 we have @xmath310 now since the number of divisors of @xmath4 is less than @xmath311 , we obtain the following lemma .
the next lemma is the principal step in our proof .
it gives information about the maximal overgroups in @xmath0 , of an element of the form @xmath313 , where @xmath314 , @xmath280 and @xmath315 are coprime .
the proof relies on knowledge of the subgroups of @xmath84 whose order is divisible by a primitive prime divisor of @xmath85 , where @xmath316 .
an account of such subgroups is given in the doctoral dissertation of joseph dimuro @xcite ; this work extends the classification of @xcite , which deals with the case @xmath86 .
[ lem : overgroups2 ] let @xmath317 and let @xmath318
. for @xmath319 , with @xmath320 , and with @xmath321 coprime , let @xmath322 .
then every maximal overgroup @xmath181 of @xmath215 in @xmath0 is a subspace stabilizer , except possibly in the following cases .
we observe that @xmath91 may be decomposed as @xmath331 , where @xmath332 , @xmath333 and @xmath334 are @xmath215-invariant subspaces of dimensions @xmath314 , @xmath280 and @xmath315 respectively .
the action of @xmath215 on each of these summands is irreducible .
it follows that @xmath215 lies in the stabilizers of proper subspaces of at least @xmath335 different dimensions ; and so @xmath215 is covered by the class @xmath104 for at least @xmath335 values of @xmath8 . note that @xmath336 , and that @xmath337 divides the order of @xmath215 .
hence a maximal overgroup @xmath181 of @xmath215 must belong to one of the classes of groups mentioned in section 1.2 of @xcite .
we observe firstly that owing to our assumption that @xmath338 and @xmath339 , the subgroup @xmath181 can not be any of those in tables 1.11.9 of @xcite ; this immediately rules out several of the examples listed there .
we shall go through the remaining examples .
any element of a symplectic or orthogonal group is similar to its own inverse ; an element @xmath215 of a unitary group is similar to its conjugate - inverse @xmath341 , where @xmath342 is induced by an involutory field automorphism .
( see @xcite , section 2.6 , or ( 3.7.2 ) for groups in characteristic @xmath24 . )
if @xmath181 normalizes a symplectic or orthogonal group @xmath343 , then @xmath344 lies in @xmath343 itself , and so @xmath344 is similar to its own inverse .
then it is clear that @xmath345 is similar to its own inverse ( it does not matter here whether or not @xmath346 ) .
but this can not be the case since @xmath347 .
similarly , if @xmath181 normalizes a unitary group @xmath99 then @xmath348 lies in @xmath99 , and it follows that @xmath348 is similar to its conjugate - inverse .
but then it follows that @xmath349 is similar to its conjugate - inverse , and it is easy to show that this is not the case .
example 3 .
_ imprimitive examples .
_ here @xmath181 preserves a decomposition @xmath350 for @xmath351 .
let @xmath352 , so that @xmath353 .
recall that the @xmath354-module @xmath91 is the direct sum of @xmath146 irreducible submodules @xmath355 , @xmath356 , @xmath357 of dimensions @xmath314 , @xmath280 , @xmath315 respectively .
so @xmath358 has at most @xmath146 orbits on the set of spaces @xmath359 .
let @xmath242 be the smallest integer such that @xmath334 is contained in the direct sum of @xmath242 of the spaces @xmath359 .
we observe that @xmath360 , and so @xmath361 . without loss of generality
, we may assume that @xmath362 .
it is clear that @xmath101 is @xmath215-invariant .
let @xmath363 be the restriction of @xmath215 to @xmath101 .
then @xmath364 acts transitively on @xmath365 . since @xmath366 acts in the same way on each @xmath367 for @xmath368 , an upper bound for the order of @xmath363 is @xmath369 .
but since @xmath370 , and since @xmath371 by assumption , we see that @xmath372 if @xmath373 .
therefore we must have @xmath374 .
it follows that @xmath334 is a simple @xmath375-module .
now since @xmath366 commutes with the projections of @xmath101 onto its summands @xmath367 , we see that at least one of the spaces @xmath367 contains an @xmath366-invariant subspace of dimension @xmath315 .
so @xmath376 , and hence @xmath377 .
suppose that @xmath378 .
since @xmath366 has two fixed spaces of dimension @xmath250 , we see that @xmath379 , and that @xmath380 . if @xmath381 , then @xmath382 .
now we see that @xmath250 divides each of @xmath314 and @xmath383 .
since @xmath314 , @xmath280 , @xmath315 are coprime , it follows that @xmath384 . but
this implies that @xmath385 , which contradicts the assumption that @xmath371 .
so we may suppose that @xmath386 .
then it is not hard to show that @xmath332 has two irreducible summands as an @xmath387-module . but
this can occur only when @xmath388 , and this accounts for the first of the exceptional cases of the lemma .
example 4 .
_ extension field examples . _
if @xmath215 stabilizes an @xmath393-structure on @xmath91 , then @xmath394 lies in the image of an embedding of @xmath395 into @xmath84 .
now if this is the case then it is not hard , by considering the degrees of the eigenvalues of @xmath215 over the fields @xmath132 and @xmath393 , to show that @xmath242 must divide each of @xmath321 .
but this implies that @xmath389 , since @xmath321 are coprime .
example 5 .
_ tensor product decomposition examples .
_ here @xmath181 stabilizes a non - trivial tensor product decomposition @xmath396 .
the central product @xmath397 embeds into @xmath84 , and @xmath181 is the intersection of this group with @xmath0 . for @xmath398 and @xmath399 , we write @xmath400 for the corresponding element of @xmath401 .
suppose that @xmath409 , and let @xmath410 and @xmath411 be such that @xmath412 .
let @xmath413 . since the order of @xmath215 is coprime with @xmath29 , we see that @xmath414 is the identity on @xmath402 , and so @xmath415 for some @xmath416 .
the largest dimension of an irreducible @xmath417-subspace of @xmath91 is @xmath315 , and there are at most @xmath24 such subspaces .
we obtain the @xmath417-subspace decomposition of @xmath91 up to isomorphism by taking two copies of each summand of the @xmath418-subspace decomposition of @xmath419 .
it follows that there must be at least two summands of dimension @xmath315 , and hence that @xmath346 and that @xmath420 .
it follows also that the @xmath314 dimensional summand of @xmath215 splits into two summands as an @xmath421-module .
but it is not hard to see that this can occur only if @xmath422 , and so we have @xmath422 and @xmath423 .
this is the second exceptional case of the lemma .
example 8.(a ) _ permutation module examples . _ in this case @xmath424 is an alternating group @xmath425 for some @xmath426 .
then it is known that the order of an element in @xmath181 is at most @xmath427 where @xmath428 , by a result of massias @xcite . here
@xmath429 or @xmath430 . but
a routine calculation shows that the inequality @xmath431 holds for all @xmath432 , and for all @xmath318 .
( this inequality fails when @xmath177 and @xmath433 . )
example 11 .
_ cross - characteristic groups of lie type . _ the examples
not yet ruled out are contained in table 1.10 of @xcite . but every element of @xmath181 has order less than @xmath434 , which is less than @xmath435 for @xmath318 .
define a set @xmath436 of classes of @xmath0 by @xmath437\ } \cup \{[g_{\lambda } ] : \lambda \in p(n ) \}.\ ] ] let @xmath48 be a set of conjugacy classes of subgroups of @xmath0 which covers @xmath436 , of the smallest size such that this is possible . then clearly @xmath438 .
by the theorem of kantor @xcite mentioned in the proof of lemma [ lem : overgroups ] above , and by lemma [ lem : overgroups2 ] , we see that @xmath48 must contain a single class of extension field subgroups .
if @xmath439 and @xmath52 then each remaining elements of @xmath48 is either a class of subspace stabilizers , or else one of the classes of subgroups mentioned in the exceptional cases of lemma [ lem : overgroups ] .
each subspace stabilizer contains at most @xmath440 of the elements @xmath441 , and each of the exceptional classes contains at most @xmath442 .
now , using lemma [ lemma : g(n ) ] , we see that @xmath443 as required for the theorem . to remove the conditions that @xmath439 and that @xmath52 , it is enough to observe that the lower bound for @xmath45 given by theorem [ thm : lowerbounds ] is larger than @xmath444 in any case where either of these conditions fails .
99 j. h. conway , r. t. curtis , s. p. norton , r. a. parker and r. a. wilson , an atlas of finite groups , clarendon press , oxford , 1985 .
britnell , a. evseev , r.m .
guralnick , p.e .
holmes and a. marti , ` sets of elements that pairwise generate a linear group ' , j. combin .
theory , series a , 115 ( 3 ) ( 2008 ) , 442465 ; with corrigendum , j. combin .
theory , series a , 118 ( 3 ) ( 2011 ) , 11521153 .
daniela bubboloni , florian luca and pablo spiga , ` compositions of @xmath4 satisfying some coprimality conditions ' , preprint ( 2012 ) , http://arxiv.org/abs/1202.1670 .
daniela bubboloni and maria silvia lucido , ` coverings of linear groups ' , comm .
algebra 30 ( 5 ) ( 2002 ) , 21432159 .
d. bubboloni , m. s. lucido and th .
weigel , ` generic @xmath24-coverings of finite groups of lie type ' , rend .
sem . mat .
padova 115 ( 2006 ) , 209252 .
d. bubboloni , m. s. lucido and th .
weigel , ` @xmath24-coverings of classical groups ' , preprint ( 2011 ) , http://arxiv.org/abs/1102.0660v1 .
daniela bubboloni and cheryl e. praeger , ` normal coverings of finite symmetric and alternating groups ' , j. combin .
theory , series a , 118 ( 7 ) ( 2011 ) , 20002024 .
daniela bubboloni , cheryl e. praeger and pablo spiga , ` normal coverings of finite symmetric and alternating groups ii ' , in preparation .
eleonora crestani and andrea lucchini , ` normal coverings of solvable groups ' , arch .
98 ( 1 ) ( 2012 ) , 1318 .
joseph dimuro , _ on prime power elements of @xmath97 acting irreducibly on large subspaces _ ,
dissertation , university of south california , 2007 .
robert guralnick , tim penttila , cheryl e. praeger and jan saxl , ` linear groups with orders having certain large prime divisors ' , proc .
london math .
78 ( 1 ) ( 1999 ) , 167-214 .
robert guralnick and gunter malle , ` simple groups admit beauville structures ' , j. london math .
soc . , to appear . william m. kantor , ` linear groups containing a singer cycle ' , j. algebra 62 ( 1980 ) , 232234 .
d. h. lehmer , ` the distribution of totatives ' , canad .
j. math . 7 ( 1955 ) , 347357 .
jean - pierre massias , ` majoration explicite de lordre maximum dun lment du groupe symtrique ' .
fac.sci .
toulouse math . _ 5 ( 6 ) ( 1984 ) , 269280 .
g. e. wall , ` on the conjugacy classes in the unitary , symplectic and orthogonal groups ' , _ j. austral .
soc . _ 3 ( 1963 ) , 163 .
k. zsigmondy , ` zur theorie der potenzreste ' , monatsh .
fr math .
u. phys . 3 ( 1892 ) , 265284 . | for a non - cyclic finite group @xmath0 , let @xmath1 denote the smallest number of conjugacy classes of proper subgroups of @xmath0 needed to cover @xmath0 .
bubboloni , praeger and spiga , motivated by questions in number theory , have recently established that @xmath2 and @xmath3 are bounded above and below by linear functions of @xmath4 . in this paper we show that if @xmath0 is in the range @xmath5 for @xmath6 , then @xmath7 . we give various alternative bounds , and derive explicit formulas for @xmath1 in some cases .
normal covering , linear group 20d50 , ( secondary ) 20g40 . |
collective features in the microscopic dynamics often lead to the emergence of surprising and unexpected effects in the evolution of a physical system at the macroscopic level @xcite .
the collective cooling of many two - level particles to very low temperatures is discussed here as an example of such a macroscopic manifestation of microscopic collective behavior @xcite .
it is shown that the collective behavior of a large number of particles can produce much higher cooling rates than they could be achieved by means of individual cooling based on the spontaneous decay of the individual particles @xcite .
as in laser sideband cooling techniques for single two - level atoms @xcite , we consider an experimental setup , where red - detuned laser fields increase the excitation of the particles , thereby continuously reducing the number of phonons .
afterwards , the phonon energy can be removed constantly from the system .
this requires energy dissipation and yields an overall decrease of the von neumann entropy in the setup @xcite .
one possible decay channel is spontaneous emission from the excited states of the particles . during such a photon emission ,
a particle returns most likely into its ground state without regaining the phonon energy lost in the excitation process .
the net result is a _ conversion _ of the phonons , originally existing in the setup in the form of thermal energy , into photons escaping the system .
here we are interested in realizing much higher cooling rates than could be achieved with the help of the above described spontaneous decay of individual particles .
this is possible , when the time evolution of the system remains restricted onto a highly symmetric and strongly coupling subspace of states throughout the whole cooling process .
maximum cooling is obtained when the particles exhibit _ cooperative _ behavior in the excitation step as well as in the de - excitation step . to achieve this we assume as in ref .
@xcite , that the outward energy dissipation is conducted by an optical cavity . as shown in fig .
[ setup ] , the particles should be placed inside a resonant optical cavity with a relatively large rate for the leakage of photons through the cavity mirrors .
the particles can then transfer their excitation collectively into the resonator field , from where the energy leaves the system without affecting the number of phonons in the setup and without changing the symmetry of the involved states . [
cols="^ " , ] collective behavior of the system requires furthermore that the rabi frequency @xmath2 of the laser field for the cooling of a vibrational mode with frequency @xmath3 and the cavity coupling constant @xmath4 are for all particles ( practically ) the same .
initially , the particles should all be prepared in their ground state ( in the large @xmath0 limit fluctuations can be neglected ) . as shown below
, the collective states of the assembly of @xmath0 particles then experience a very strong coupling to the laser field as well as to the cavity mode @xcite . as a consequence ,
the number of phonons decreases exponentially with a with as large as @xmath1 times the single - particle coupling constants , which can be much higher than previously predicted cooling rates in comparable setups @xcite .
the proposed cooling scheme might be used to cool a large number of particles very efficiently .
so it should be applicable to the preparation of bose einstein condensates .
currently , these experiments mainly use evaporative cooling @xcite which systematically removes those atoms with a relatively high temperature from the trap .
consequently , only a small percentage of the initially trapped atoms is finally included in the condensate .
if one could instead cool all the atoms efficiently , yet at the same time avoid the loss of particles , it should become easier to experiment with large condensates .
cooling is also crucial for ion trap quantum computing , where the achievable gate operation times can depend primarily on the efficiency of the cooling of a common vibrational mode @xcite .
first cavity - cooling experiments involving many particles and observing enhanced cooling rates have already been performed @xcite . in this paper
, we consider some algebraic features of the dynamics ruling the collective cooling of many particles and further clarify some aspects of the underlying mechanisms , which have already been studied in ref .
a remarkable feature is , for example , the presence of a relatively large and negative coherence @xmath5 .
instead of solving the time evolution of the system explicitly , we avoid certain approximations made in ref .
@xcite by referring to the heisenberg picture .
the system considered in this paper and its collective dynamics may be a paradigmatic example for other applications of physical interest , like the evolution of a system undergoing a continuous phase transition , yet preserving some specific features during such an evolution .
the cavity coupling constant @xmath4 is for all particles the same , when the particles distribute themselves in the antinodes of the resonator field , as discussed in ref .
@xcite ( see fig . [ setup](a ) ) .
alternatively , a ring resonator could be used , as proposed in refs .
@xcite . in the following
we consider a collection of @xmath0 two - level particles ( atoms , ions or molecules ) with ground states @xmath6 and excited states @xmath7 ( see fig . [ setup](b ) ) .
the setup should be operated in a parameter regime , where @xmath8 here @xmath9 is the decay rate of a single photon in the cavity mode , @xmath10 is the spontaneous decay rate of a particle in the excited state and @xmath11 denotes the lamb - dicke parameter characterizing the steepness of the trap @xcite . as in ref .
@xcite , we discuss in this paper the two extreme cases , namely the cooling of common vibrational modes and the cooling of the individual phonon modes in the absence of common vibrational modes . in the following , @xmath12 denotes the annihilation operator for a phonon in the common vibrational mode with frequency @xmath3 , while @xmath13 is the annihilation operator for the cavity photons and @xmath14 is the lowering operator for particle @xmath15 . using the rotating wave approximation @xcite and going over to the interaction picture with respect to the interaction - free hamiltonian
, the time evolution of the system can be described by the hamiltonian @xmath16 one way to simplify this hamiltonian is to introduce the effective phonon annihilation operator @xmath17= 1 \ , , \end{aligned}\ ] ] which allows to write the hamiltonian ( [ hi0 ] ) as @xmath18 we now consider the operator @xmath19 for the annihilation of the phonons of mode @xmath3 of particle @xmath15 .
again we assume that the corresponding laser rabi frequencies @xmath2 are for all particles the same .
then the hamiltonian of the system equals in the interaction picture and in the rotating wave approximation @xmath20 proceeding as above , introducing the effective phonon annihilation operator @xmath21= 1\end{aligned}\ ] ] and using eq .
( [ par ] ) , the hamiltonian ( [ h0 ] ) becomes @xmath22 although this hamiltonian has some similarities with the hamiltonian ( [ hi ] ) , it describes a physically different situation . instead of coupling to a set of common vibrational modes
, each particle sees his own set of phonons .
spontaneous emission is described in the following by the master equation @xcite @xmath23 + \kappa \
, \big(c \rho c^\dagger - { \textstyle{1 \over 2 } } c^\dagger c \rho - { \textstyle{1 \over 2 } } \rho c^\dagger c \big ) \nonumber \\ & & + \gamma \sum_i \big(\sigma_i \rho \sigma_i^\dagger - { \textstyle{1 \over 2 } } \sigma_i^\dagger \sigma_i \rho - { \textstyle{1 \over 2 } } \rho \sigma_i^\dagger \sigma_i \big)\end{aligned}\ ] ] with @xmath24 being @xmath25 or @xmath26 , respectively , @xmath9 being the decay rate of a photon in the cavity and @xmath10 being the decay rate of the particle excited state . from this equation
one can easily see that the dissipation of cavity photons reduces the energy in the system without affecting the state of the particles .
cavity decay therefore does not disturb the collective behavior of the system .
however , this does not apply to the emission of photons from the particles , which can negatively interfere with the collective cooling process .
in case of the cooling of common vibrational modes , the setup consists effectively of _ two _ different subsystems .
one is the two - level particle system . as an effect of the emergence of the cooperative behavior ,
the collection of particles manifests itself as a _
bosonic _ system .
we will see below that a further consequence of this is the transition to a strong coupling regime , which in turn implies a much shorter time scale for the system evolution .
the other subsystem is the bosonic system of phonons and photons with a continuous _ conversion _ of phonons into photons .
we will see that it is convenient to consider a boson mode which is a superposition of them .
this is analogous to the field - atom polariton of hopfield @xcite .
the physical implication of the familiar commutator relations ( [ e2 ] ) is that the particles behave no longer like _
individual fermions_. instead , the time evolution generates the excitation of bosonic modes , namely collective dipole waves , with @xmath39 denoting the creation and annihilation operators of the associated quanta obeying the usual commutation relation ( [ e2 ] ) . as a consequence ,
the collection of the single two - level particles manifests itself as a _
bosonic _ system .
in other words , the ladder of equally - spaced dicke states approximates to a weakly - excited harmonic oscillator . using the operators ( [ col ] ) ,
the hamiltonian ( [ hi ] ) for the cooling of _ common _ vibrational modes can simply be written as @xmath41 where we have introduced the notation @xmath42 from this one sees that the time evolution of the system is mainly governed by the parameters @xmath43 , @xmath44 and @xmath9 , which scale as @xmath1 .
we thus have , as a consequence of the emergence of the particle collective behavior , the transition to the strong coupling regime , @xmath45 and @xmath46 .
the evolution of the system happens no longer on the time scale given by the parameters @xmath47 and @xmath4 but on a much shorter time scale defined by @xmath48 and @xmath49 .
another factor that contributes significantly to the described cooling mechanism is the _ interface _ or _
ambivalent _ role played by the particles with respect to the phonons and the photons in the setup .
the only difference between the particle - phonon and particle - photon interaction is the difference of the coupling constants , given by @xmath50 and @xmath49 , respectively . in some sense , the particles act as an engine transforming phonon energy into photon energy .
this last one is then dissipated outward through the coupling with the cavity which allows energy leakage .
however , in the absence of the leakage of photons through the cavity , the inverse transformation , photons to phonons , is also possible in principle . in such a situation ,
interesting interference and coherence effects arise , which we analyze below in detail .
indeed , a closer look at eq .
( [ xxx ] ) reveals that the hamiltonian for the cooling of common vibrational modes can alternatively be written as @xmath51 with @xmath52=1 \ , . ~~\end{aligned}\ ] ] instead of interacting with the phonons and photons separately , the particles see the boson mode with annihilation operator @xmath53 and number operator @xmath54\end{aligned}\ ] ] with @xmath55 physically , the creation of bosons corresponding to @xmath56 does not only affect the number of phonons and the number of photons in the system .
inevitably , it also creates a coherence between the @xmath57 and the @xmath13 subsystem .
this coherence @xmath5 provides a symmetric " channel for the phonon - photon energy transformation .
however , the leakage of energy outside the cavity perturbs such a symmetric phonon - photon balancing due to the @xmath5 action .
the system reacts by subsequent adjustments , trying to recover its lost balance .
crucial to such a re - adjustment mechanism is the difference in the time scales .
the time scale for reaching the quasi - stationary state is of the order @xmath58 , while the time scale for spontaneous decay is , as we see in the next section , of order @xmath59 .
the time evolution of the system turns out to be highly non - linear since the second and higher order derivatives of the physical observables are much larger than their first order derivatives .
the system reaches a quasi - stationary state on a time scale of the order @xmath58 .
the formation of this local equilibrium is solely governed by the effect of the hamiltonian @xmath60 .
let us therefore first consider the situation , where we can neglect spontaneous emission and @xmath61 and @xmath62 . in this case , the time evolution of the system , governed by the hamiltonian ( [ yyy ] ) , results in a redistribution of population between the bosonic subsystem described by @xmath63 and the particles described by @xmath64 .
to analyse this process we introduce the operators @xmath65 with the familiar @xmath31 commutators @xmath66 = { \rm i } \epsilon_{ijk } \
, l_k \ , .\end{aligned}\ ] ] with this notation , the hamiltonian ( [ yyy ] ) becomes @xmath67 and this , formally , simply generates a rotation around the 1-axis .
in such an algebraic picture , we can immediately conclude that the angular momentum @xmath68 and the total angular momentum @xmath69\end{aligned}\ ] ] are conserved during the time evolution of the system under the considered conditions .
especially , the conservation of @xmath70 implies the conservation of the total number of bosons , as accounted for by the number operator @xmath71 . when the number of particles in the excited state @xmath7 remains small compared to @xmath0 , the commutator relations ( [ e2 ] ) hold . in such a case
, we can assume that the expectation value of @xmath72 remains small .
neglecting @xmath64 in eq .
( [ comp ] ) implies @xmath73 using the definition ( [ a ] ) , this conservation law translates into @xmath74 which coincides with eq .
( 16 ) in ref .
@xcite and describes the continual balance between the rate of change of the expectation value of @xmath5 and the rate of change of the total number of phonons and photons in the system . to show that eq .
( [ cons ] ) does not violate the conservation of the number of particles in the setup , we remark that the system obeys a second conservation law .
considering again the heisenberg picture , we indeed find @xmath75 in the presence of many particles , with most of them remaining in their ground state , the expectation value of @xmath64 remains negligible .
this implies @xmath76 and , consequently , also @xmath77 @xcite . inserting this into eq .
( [ energy ] ) , we obtain the particle number conservation law @xmath78 despite being a coherence , the expectation value of @xmath5 acts like a population .
this allows the system to obey conservation of the particle number @xmath63 as well as conservation of @xmath79 by balancing the coherence @xmath5 accordingly . in the following , we study the conversion of phonons into cavity photons and creation of a non - zero coherence @xmath5 in more detail . in the absence of spontaneous emission , the system reaches after a very short time a stationary state with constant expectation values for @xmath80 , @xmath81 and @xmath5 . to calculate the corresponding values of the cavity photon number and coherence @xmath5 as a function of the phonon number , we use again the heisenberg picture , and obtain the second order differential equations @xmath82 these equations imply that the first order derivatives of the operators @xmath80 , @xmath81 and @xmath5 change on a time scale of the order @xmath58 , which is , for large @xmath0 , much faster than the time scale on which the time evolution of @xmath83 , @xmath81 and @xmath5 takes place .
since we are only interested in the time dependence of the phonon number @xmath80 , we can safely assume that the first order derivatives @xmath84 , @xmath85 and @xmath86 adapt adiabatically to the state of the system . setting the right hand sides of the differential equations ( [ sec ] ) equal to zero
, we find @xmath87 moreover , eq . ( [ m ] ) implies , due to the conservation laws ( [ cons ] ) and ( [ cons2 ] ) , that @xmath88 this means that , after a short time of order @xmath58 , the system reaches a quasi - stationary state with a constant ratio between the expectation value of the coherence @xmath5 and the number of photons inside the cavity , respectively , compared to the total number of phonons in the system . to a very good approximation ,
these ratios remains constant throughout the whole cooling process . in the absence of spontaneous emission , at most a redistribution of phonon energy
can occur in the system @xcite .
efficient cooling and the irreversible removal of energy from the system requires dissipation .
we therefore now consider the effect of dissipation in more detail .
using the master equation ( [ rho ] ) we find @xmath89 the leakage of photons through the cavity mirrors with decay rate @xmath9 not only decreases the number of photons in the cavity mode but also affects the size of @xmath5 . in the previous subsection
, we have seen that the system reaches a stationary state on a time scale of the order @xmath58 with @xmath81 and @xmath5 being a multiple of @xmath80 . using eq .
( [ trop ] ) , we find @xmath90 + { x \over y } \ , \big [ { \textstyle { 1 \over 2 } } \kappa \ , k_3 + { \textstyle { { \rm d } \over { \rm d}t } } k_3 \big ] \ , .
\nonumber \\ & & \end{aligned}\ ] ] this shows that on the time scale we are interested in , namely a time scale of the order @xmath59 , both , the operator @xmath81 and the presence of a negative valued operator @xmath5 , provide effective cooling channels in the system . inserting the results ( [ m ] ) and ( [ wednesday ] ) into eq .
( [ now ] ) we obtain indeed @xmath91 if @xmath92 denotes the phonon number expectation value @xmath93 and @xmath94 , we finally obtain @xmath95 \ , .\end{aligned}\ ] ] this describes exponential cooling of the atomic sample with a rate that can be as large as @xmath49 and @xmath96 ( cf .
( [ par ] ) ) .
the result ( [ rate ] ) coincides with eq .
( 17 ) in ref .
@xcite and is in good agreement with an exact numerical solution of the time evolution of the system ( see fig .
[ num](a ) ) .
spontaneous emission from excited atomic levels with decay rate @xmath10 is too slow to contribute to the cooling of the system , since most of the particles remain in their ground states and @xmath97 .
finally , we remark that the cooling process we describe here is not simply a redistribution of phonons into cavity photons and @xmath5 , which then decay into the environment with the spontaneous decay rates @xmath9 and @xmath98 , respectively .
instead , the leakage of photons through the cavity mirrors disturbs the equilibrium expressed by eqs .
( [ cons ] ) and ( [ cons2 ] ) perturbing the otherwise conserved quantities @xmath99 the coupling to the environment causes a _ dynamical response _ in the system , which generates a transition from a state with a fixed ratio of phonons to cavity photons ( cf .
( [ m ] ) ) into a state with no phonons and photons in the setup . in atom - cavity systems ,
the time evolution of the system is , on a very short time scale , usually dominated by spontaneous emission .
however , in the presence of many particles , the evolution is primarily governed by the hamiltonian @xmath25 , which drives the system into a quasi - stationary state characterised by the preserved quantities @xmath100 and @xmath101 within a time of the order @xmath58 .
the presence of the spontaneous decay rate @xmath9 , which scales as @xmath1 ( cf .
( [ par ] ) ) , disturbs this stationary state and causes the system to continuously assume new values for @xmath100 and @xmath101 . using eqs .
( [ m ] ) and ( [ ooo ] ) , we find @xmath102 while the relative size of the photon number @xmath103 and the coherence @xmath104 with respect to the number of phonons @xmath92 remain constant , their sum , as accounted for by @xmath100 and @xmath101 decrease exponentially in time due to the presence of dissipation , namely leakage of photons through the cavity mirrors .
this is typical for a system undergoing a phase transition .
there are many similarities but also many differences between the cooling of the common and the cooling of the individual phonon modes of particles .
for example , one can not simplify the hamiltonian @xmath26 using the collective lowering operator @xmath105 , as we did in eq .
( [ xxx ] ) .
however , there is still a high symmetry in the system and all particles are treated in exactly the same way . to take this into account
we introduce the vector operators @xmath106 with the usual scalar product , such that , for example , @xmath107 this notation allows us to write the hamiltonian ( [ h ] ) as @xmath108 which is of a similar form as the hamiltonian @xmath25 in eq .
( [ xxx ] ) . in analogy to section [ aaa ] ,
we proceed by introducing an effective annihilation operator @xmath109 for each particle @xmath15 with @xmath110=1 \ , . ~~\end{aligned}\ ] ] then the hamiltonian ( [ xxxx ] ) can be written as @xmath111 each particle couples individually to a new type of bosons , which are a superposition of a single phonon and a cavity photon .
the number operator accounting for all bosons @xmath109 equals @xmath112 \ , , \end{aligned}\ ] ] where we defined the coherence @xmath113 while @xmath114 counts the total number of phonons in the system , the expectation value of @xmath115 .
again , the coherence @xmath5 describes a certain `` symmetry '' between phonons and photons in the setup and accounts for a continual conversion of the two types of bosons into each other . neglecting spontaneous emission ,
i.e. assuming @xmath61 and @xmath62 , and using the hamiltonian ( [ xxxx ] ) , we find @xmath116 in the derivation of these equations , we neglected the operators @xmath117 and @xmath118 with @xmath119 since there are no interactions between particles and the phonons of other particles . from eq .
( [ trop2 ] ) we see that there are , as before , two conserved quantities in the system , namely @xmath120 one is associated with the total number of bosonic particles with annihilation operators @xmath109 , the other one counts the total number of phonons and photons in the setup and @xmath121 from the formal equivalence of the hamiltonians ( [ yyy ] ) and ( [ yyyy ] ) pointed out in the previous subsection , one might have had expected that the expectation value @xmath122 is preserved in the time evolution of the system . however
, this would only be the case if @xmath109 and @xmath123 commute with each other for @xmath119 and does not apply here .
to calculate the distribution of phonons , cavity photons and coherence @xmath5 in the system , which builds up in the absence of spontaneous emission , we now consider the second derivatives of the operators @xmath114 , @xmath124 and @xmath5 in the heisenberg picture .
neglecting again the population in the excited states , as accounted for by @xmath125 , we obtain @xmath126 setting the right hand side of these equations equal to zero and taking the conservation laws ( [ lee ] ) in the absence of dissipation into account , we find that the system possesses a stationary state with @xmath127 in contrast to the results in section [ fast ] , this equation describes a state with only a relatively small number of photons in the cavity mode compared to the total number of phonons in the setup .
however , as we see in the next subsection , the presence of a negative coherence @xmath5 of the same order of magnitude as @xmath114 provides a cooling channel , which is sufficient to obtain a cooling rate of the same order of magnitude as the cavity decay rate @xmath9 . as before , the effective removal of phonons from the setup requires leakage of photons through the cavity mirrors .
to take spontaneous emission into account we consider again the master equation ( [ rho ] ) and find @xmath128 which implies @xmath129 \nonumber \\ & & + { x \over y } \
, \big [ { \textstyle { 1 \over 2 } } \kappa \ , k_3 + { \textstyle { { \rm d } \over { \rm d}t } } k_3 \big ] \ , .\end{aligned}\ ] ] as already mentioned above , the number of photons in the cavity remains relatively small and does not contribute considerably to the cooling process . neglecting @xmath130 and using the results from the previous subsection , we obtain @xmath131 and @xmath132 \ , , \end{aligned}\ ] ] where @xmath133 and @xmath134 now describe the total number of phonons in the system , as accounted for by @xmath135 at time @xmath136 and @xmath137 , respectively .
as in the case of the cooling of common vibrational modes , cooling rates of the same order of magnitude as the cavity decay rate @xmath9 can be obtained , which can be as large as @xmath1 times the single particle coupling constants . however , there are some differences with respect to the case considered in section [ comm ] . in the present case ,
achieving such a high cooling rate and cooling the system to very low temperatures first requires the build up of a reasonably large coherence @xmath5 , which does not exist in the absence of any laser driving .
suppose , there are no initial correlations between the particles and their motional degrees of freedom
. then @xmath138 ( cf .
( [ trop3 ] ) ) .
moreover , the second order derivative of @xmath5 increases only on a time scale of order one ( cf .
( [ sec2 ] ) ) .
initially , the system is far away from the quasi - stationary state described in the previous subsection .
however , once @xmath5 reaches its equilibrium , the collective cooling process can begin . the coherence @xmath5 needs no longer to be established ; the system only has to adapt to the small changes of @xmath100 and @xmath101 caused by the leakage of photons through the cavity mirrors . for
a numerical solution of the time evolution of the system see fig .
[ num](b ) .
the emergence of collective dynamics in a system of a large number of particles manifests itself in some macroscopic features in the system behavior . as an example , we considered the problem of fast and efficient cooling of an assembly of @xmath0 two - level particles trapped inside a leaky optical cavity .
results obtained here confirm those derived in ref .
the particles are excited by red - detuned laser fields .
when the coupling constants are for all particles the same , a collective behavior emerges and the cooling rate can be as large as @xmath1 times the single - particle coupling constants .
the generation of cooperative behavior of the @xmath0 particles is crucial in the excitation step as well as in the de - excitation step .
the collective states of the assembly of @xmath0 particles then experience a very strong coupling to the laser field as well as to the cavity mode .
we have considered the case of particles sharing a common phonon mode and the case of individual phonon modes for each particle .
the two cases are similar and in both cases the collective cooling rate is very fast as compared to the single - particle cooling rate .
the two cases are , however , different for the behavior of the coherence @xmath5 : in the individual mode case , high cooling rates are achieved by first building up a reasonably large coherence @xmath5 , which does not exist in the absence of any laser driving .
initially , the system is far away from the quasi - stationary state , where the phonon and the photon populations are balanced through the @xmath5 action .
only when @xmath5 reaches its equilibrium , the collective cooling process can begin and the system can adapt to the small changes of @xmath100 and @xmath101 caused by the leakage of photons through the cavity mirrors . in the common mode case , there is no need for the pre - cooling " phase for a build up of the coherence @xmath5 . the system considered in this paper may be a paradigmatic example for other applications of physical interest , such as systems undergoing continuous phase transitions , yet preserving some specific features during their evolution .
systems presenting such a behavior might be of interest as well in biology .
this work was supported in part by the european union , coslab ( esf program ) , infn , infm and the uk engineering and physical sciences research council .
acknowledges funding as a james ellis university research fellow from the royal society and the gchq . 00
p. w. anderson , science , * 177 * , ( 1972 ) 393 .
a. beige , p. l. knight , and g. vitiello , _ cooling many particles at once _ , quant - ph/0404160 .
d. j. wineland and h. dehmelt , bull . am .
phys . soc . ,
* 20 * ( 1975 ) 637 .
j. i. cirac , m. lewenstein , and p. zoller , phys .
, * a51 * ( 1995 ) 1650 . g. morigi , j. eschner , and c. h. keitel , phys .
* 85 * ( 2000 ) 4458 .
a. bartana , r. kosloff , and d. j. tannor , j. chem .
* 106 * ( 1997 ) 1435 .
v. vuleti and s. chu , phys
, * 84 * ( 2000 ) 3787 .
p. horak and h. ritsch , phys .
, * a63 * ( 2001 ) 023603 . s. a. gardiner , k. m. gheri , and p. zoller , phys . rev . , * a63 * , 051603 .
v. vuleti , h. w. chan , and a. t. black , phys .
, * a64 * ( 2001 ) 033405 .
p. horak and h. ritsch , phys .
, * a64 * ( 2001 ) 033422 .
p. domokos and h. ritsch , j. opt .
, * b20 * ( 2003 ) 1098 .
a. t. black , h. w. hilton , and v. vuleti , phys .
* 91 * ( 2003 ) 203001 .
elssser , b. nagorny , and a. hemmerich , phys .
, * a67 * ( 2003 ) 051401 .
a. griessner , d. jaksch , and p. zoller , j. phys .
, * b37 * ( 2004 ) 1419 .
m. h. anderson , j. r. ensher , m. r. matthews , c. e. wieman , and e. a. cornell , science , * 269 * ( 1995 ) 198 .
a. beige , phys .
, * a69 * ( 2004 ) 012303 . h. w. chan , a. t. black , and v. vuleti , phys . rev .
lett . , * 90 * ( 2003 ) 063003 .
b. nagorny , th .
elssser , and a. hemmerich , phys .
* 91 * ( 2003 ) 153003 .
c. c. gerry and p. l. knight , _ introductory quantum optics _ , cambridge : cambridge university press ( 2004 ) , p. 257 .
this approximation introduces some errors when calculating the behavior of the system for times @xmath139 , with a restriction of achievable cooling rates from above .
nevertheless , collective cooling can be much more efficient than previously considered mechanisms .
s. m. barnett and p. m. radmore , _ methods in theoretical quantum optics _ , clarendon press ( oxford , 1997 ) .
j. j. hopfield , phys . rev . * 122 * , 1555 ( 1958 ) .
b. huttner , j. j. baumberg , and s. m. barnett , europhys .
* 16 * ( 1991 ) 177 .
r. h. dicke , phys . rev .
* 93 * , 99 ( 1954 ) .
f. t. arecchi , e. courtens , r. gilmore , and h. thomas , phys . rev . , * a6 * ( 1972 ) 2211 .
j. m. radcliffe , j. phys . ,
* a4 * ( 1971 ) 313 .
e. inn and e. p. wigner , proc .
* 39 * ( 1953 ) 510 . c. de concini and g. vitiello , nuc .
, * b116 * ( 1976 ) 141 .
m. n. shah , h. umezawa , and g. vitiello , phys .
, * b10 * ( 1974 ) 4724 .
assuming @xmath140 is also equivalent to neglecting the coherence operators @xmath141 and @xmath142 , which is consistent with our calculations in ref .
there we showed that @xmath143 and @xmath144 remain indeed zero to a very good approximation . | in a recent paper [ beige , knight , and vitiello , quant - ph/0404160 ] , we showed that a large number @xmath0 of particles can be cooled very efficiently using a bichromatic interaction .
the particles should be excited by red - detuned laser fields while coupling to the quantized field mode inside a resonant and leaky optical cavity .
when the coupling constants are for all particles the same , a collective behavior can be generated and the cooling rate can be as large as @xmath1 times the single - particle coupling constants . here
we study the algebraic structure of the dynamics and the origin of the collective cooling process in more detail . |
as a quantum system of interest interacts with its environment which is also considered being made up of a large number of the quantum particles , solving the dynamics of the total system is in fact an impossible task by now .
however , in many cases we are only interested in the dynamics of the quantum system of interest , thus we can investigate the reduced system by using the phenomenological environment as proposed by caldeira and leggett @xcite .
usually , in this method the environment is modelled by a bath of harmonic oscillators ( or other quantum particles ) and the influence of the bath on the quantum system is fully enclosed in the so - called spectral density function @xmath0 according to the characteristic of the environment one can choose the @xmath1 ohmic , sub - ohmic , super - ohmic forms and so on .
thus , the reduced dynamics of the quantum systems can be investigated in detail .
many such problems have been treated extensively and some significative results have been obtained in the field in past years @xcite .
the spectral density functions @xmath1 obtained from the microcosmic dynamical model plus its phenomenological environment ( mdmppe ) is in good agreement with that obtained from quasi - classical equations of motion satisfied by the system of interest .
we can obtain the effective spectral density function ( esdf ) following leggett as follows . setting @xmath2 taking cut - off frequency of the bath modes infinity , and supposing the spectral density function @xmath1 a continuous function
, we can obtain the esdf with @xmath3im@xmath4 $ ] ( see ref.@xcite ) , where @xmath5 represents the effects of arbitrary linear dissipative and/or reactive elements .
usually , @xmath5 can be obtained from the equation of motion of the system in fourier - transformed form , namely , @xmath6\left(\omega\right),$ ] where @xmath7 is a conservative potential .
this equation can be obtained from the scheme of mdmppe .
the esdfs from infinite cut - off frequency of baths for many models are obtained @xcite .
this method is in fact a universal one .
when the temperature of the systems are low enough , we can imagine that the dynamics of the quantum systems of interest obtained from the mdmppe should have some difference with that from quasi - classical equations of motions . in this case , the high - frequency modes of the bath are in fact suppressed , so the cut - off frequency may be not very large and it may be comparable to the characteristic frequency of the quantum system of interest .
so there are at least two differences between the dynamics of normal and low temperature for the quantum systems of interest .
the first is that the esdfs obtained from the mdmppe at low temperature should not be in accordance with the spectral density functions from quasi - classical equations of motion , and the second is that the system at low temperature must go through a non - markovian process @xcite other than the markovian one . with the progress of technique many quantum systems have been designed and manufactured . in particular , many qubit models for making quantum computers in future
are proposed in recent years .
most of these quantum systems should work in very low temperature @xcite .
the non - markovian effects of baths in the systems have been considered recently @xcite .
it is reported that in many chemical and biologic systems , the non - markovian effects may also be important @xcite . at very low temperature
, the cut - off frequencies of the bath should be finite . taking a finite cut - off frequency of the bath rather than the infinite one , we shall obtain another kind of the spectral density function . in the following
we denote the esdfs from infinite , and from finite cut - off frequencies of the bath for model @xmath8 by @xmath9 ( from infinite cut - off frequencies ) and @xmath10 ( from finite cut - off frequencies ) , where @xmath8 denotes the physical models .
a somewhat similar discussion on the problem in completely classical framework has been given in refs @xcite .
in this paper , we shall investigate the esdfs of four models by using the scheme of mdmppe at infinite and finite cut - off frequencies of baths . through the two sets of esdfs @xmath11 and @xmath12 we shall investigate the reduced dynamics of qubits in some practical physical models .
they are a quantum system in a bath ( model a ) ; a quantum system coupled to an intermediate harmonic oscillator ( iho ) and only the latter coupled to a bath ( model b ) ; a quantum system coupled to an iho and both of them coupled to their independent baths ( model c ) and a common bath ( model d ) .
in this section , we shall obtain two kinds of esdfs for the models a , b , c , and d when they are reduced to the original spin - boson form .
one kind of them are obtained from taking finite cut - off frequencies of baths and the other is gotten by setting the cut - off frequencies of baths infinity . after obtaining the spectral density functions we shall solve the reduced dynamics for the quantum systems of interest when they are set as the qubits .
the reduced dynamics governed by the two kinds of esdfs for the qubits will be compared . in order to compare the dynamics of the qubit governed by the different effective spectral functions in each model , we choose a same starting point @xmath13 for all of the models . in another word
, we always reduce them to the original spin - boson model with the same iho transformation and set @xmath14 in our derivations . because the model a is just the original spin - boson model ,
our work for reducing this model is just reducing it to itself .
so in the assumption of infinite cut - off frequency , the obtained esdf of the model should be equal to the @xmath15 , which will be shown in the following subsection a. models b , c , and d are different from the original spin - boson model , so when they are reduced to the original spin - boson form , the esdfs should be different from the starting function @xmath13 .
we have calculated the dynamics of the quantum system of interest governed by @xmath16 and @xmath12 with the same values of @xmath17 in each models of the paper , and they are certainly different .
however , even for the same model , the different esdfs @xmath16 and @xmath12 , may provide different damping for the same @xmath17 in the function @xmath18 , which will affect the dynamics of the quantum system of interest . do the different kinds esdfs affect the dynamics of quantum system of interest mainly through resulting in different damping or through their different forms ? to answer this problem , we introduce an effective coupling strength @xmath19 to calculate the dynamics as ref.@xcite .
the effective coupling constant is introduced from the physical deduce that the modes which has the same frequency as the characteristic frequency of the system @xmath20 has the major contribution to the system - bath interaction .
thus , by using the effective coupling constant we can obtain that , if the reduced dynamics governed by @xmath16 and @xmath12 are different , we can safely attribute the different forms of the @xmath16 and @xmath12 . for lack of space , in the following , we only plot the evolutions of elements of reduced density matrix in the same effective coupling strength @xmath21 for each models .
a quantum system embeds in a bath ( model a ) and a quantum system interacts with an iho which is coupled to a bath ( model b ) , are two basic models and many complicated models are developed from them .
so in this subsection , we use them to show the leggett s method for obtaining the esdfs @xmath22 , and @xmath23 .
then from the two sets of esdfs we investigate the reduced dynamics of qubits in the two models . model a : because the physical quantities of the quantum system of interest is not included in the esdf , we can deduce the esdfs of the model a with the hamiltonian as @xcite @xmath24 where @xmath25 is the momentum conjugate to coordinate @xmath26 . defining @xmath27 and using the dots for the time derivatives ,
we can obtain the classical equations of the motion as@xmath28 using the fourier transforms , we can write eqs.([1 - 2 - 1 ] ) and ( [ 1 - 2 - 2 ] ) as@xmath29 from eq.([1 - 3 - 2 ] ) we have@xmath30 insetting eq.([1 - 4 ] ) into eq.([1 - 3 - 1 ] ) , we have@xmath31q(\omega)=-u_{\omega}^{\prime}(q\left ( t\right ) ) .
\label{1 - 5}\ ] ] setting@xmath32 and @xmath33 we then have@xmath34 where ( see appendix)@xmath35 the expression of @xmath36 is given in the appendix . substituting eq.([1 - 9 ] ) into eq.([1 - 5 ] ) we have@xmath37q\left(\omega\right)\nonumber\\ & = -u_{\omega}^{\prime}\left(q\right).\end{aligned}\ ] ] if taking @xmath38 be infinite , we can easily obtain the spectral density function as @xmath39 this is the conventional ohmic spectral density function of bath in the original spin - boson model and it has been widely used in quantum dissipative systems . here
, we have only shown that , starting from @xmath13 , an esdf @xmath40 can be obtained .
the @xmath41 and @xmath40 are the same for the bath in the original spin - boson model .
the result shows that the leggett s method give a consistent result for the model .
if the model is not the original spin - boson model ( as following models b , c , and d ) , it can also be reduced to the original spin - boson model but the esdfs should not be equal to the @xmath15 even we start from the same @xmath13 . if taking @xmath38 be finite , we then obtain the esdf as @xmath42 here , @xmath43 the other forms of the spectral density functions , such as sub - ohmic and super - ohmic forms can be obtained similarly .
it shows that when the cut - off frequency is finite , we should add the term @xmath44 $ ] in the spectral density function as eq.([1 - 12 ] ) .
it is interesting that how does the added term influence the spectral density function ?
in fig.1 we plot @xmath45 $ ] in different cut - off frequencies of the bath .
it shows that the absolute values of the @xmath45 $ ] decreases with the increase of @xmath38 .
when @xmath46 , the @xmath47\rightarrow0 $ ] .
however , when the cut - off frequency of the bath is not big enough , the influence of @xmath45 $ ] should not be neglected in the description of the environment . in above derivation of the esdfs , the physical quantities for the quantum system of interest have not been involved , so the @xmath48 can control the dynamics of not only the harmonic oscillator as in eq.([1 - 1 ] ) but also other quantum systems . in order to check the differences of the influences that above two different esdfs @xmath49 and @xmath50 impose on the dynamics for the systems of interest
, we firstly investigate a simple quantum system , qubit . for the open qubit
the total harmiltonian reads @xmath51 this is one of the simplest open models of quantum system but its reduced dynamics can still not be solved exactly , and some approximations must be appealed to .
many investigations on this topic have been made in recent years @xcite . as we stressed that when the cut - off frequency is not too large to compare to the characteristic frequency of the qubit , the non - markovian effect is non - negligible .
the well established iterative tensor multiplication ( itm ) algorithm based on the quasiadiabatic propagator path integral ( quapi ) scheme is a good tool for solving the reduced dynamics of quantum system of interest .
the itm algorithm is a numerically exact one and is successfully tested and adopted in various problems of open quantum systems @xcite . for details of the scheme , we refer readers to previous works @xcite . to make the calculations converge we use the time step @xmath52 which is shorter than the correlation time of the bath and the characteristic time of the qubit . in order to include as much non - markovian effect of the bath as we can , and avoid the heavy calculation load we choose @xmath53 . here , @xmath54 is the memory times being included in the calculations . in this paper
we set @xmath55 k , and the initial state of the bath be the thermal equilibrium state , namely , @xmath56tr@xmath57 .
here , @xmath58 , @xmath59 is the boltzman constant .
as we calculate the evolutions of diagonal elements of the reduced density matrix @xmath60 , we set @xmath61 hz , @xmath62 , and the initial state of the qubit be @xmath63 .
as we calculate the evolutions of non - diagonal elements of the reduced density matrix @xmath64 , we set @xmath65 hz , @xmath66 , and the initial state of the qubit be @xmath67 .
here , @xmath68 and @xmath69 .
the @xmath70 and @xmath71 are the basis states of the qubit .
the comparison pictures on @xmath60 and @xmath64 in different esdfs ( @xmath72 , and @xmath73 ) and different cut - off frequencies of the bath ( @xmath74 and @xmath75 ) are plotted in fig.2 , where @xmath21 is set to be 0.004 .
it is shown that in the model a the qubit has shorter decoherence and relaxation times governed by the @xmath72 than by @xmath73 .
model b : there is another basic model that the quantum system of interest is not coupled to any bath but coupled to an iho which is coupled to a bath .
the hamiltonian of this model reads@xmath76 in the model , the pure hamiltonian of the quantum system of interest is @xmath77 .
the iho has mass @xmath78 and frequency @xmath79 , and is coupled by the quantum system of interest with strength @xmath80 .
the bath coupled to the iho is comprised of a set of harmonic oscillators , and their mass , momentum , coordinate , and coupling coefficients are denoted by @xmath81 .
the introduced @xmath82 is a controlling parameter for discussions .
it has been shown that the model b can be reduced to the model a through an esdf of the bath @xcite . setting the cut - off frequency of the bath
be infinite , we can obtain the esdf as @xmath83 which has been obtained in ref .
@xcite . here , @xmath84 . if the cut - off frequency is finite , the esdf becomes @xmath85 here , @xmath86 .
\label{2 - 3}\ ] ] if the quantum system of interest is a qubit the model b can be reduced to a spin - boson model as eq.([1 - 14 ] ) with esdf @xmath87 . by using the itm algorithm
we can also investigate the reduced dynamics of the qubit .
the evolutions of the diagonal and off - diagonal elements of the reduced density matrix of the qubit governed by @xmath88 and @xmath89 are plotted in fig.3 for the cases of @xmath90 and @xmath91 . here , we set @xmath92 , and @xmath93 , and other parameters are the same as fig.2 .
it is contrary to the case of model a , here the qubit has longer decoherence and relaxation times governed by the @xmath94 than by @xmath95 . as the cut - off frequency of the bath is not very large , the @xmath40 results in shorter decoherence and relaxation times for the qubit in the models a than the @xmath96 does .
however , in model b , to compare to the @xmath97 , the @xmath98 causes longer decoherence and relaxation times of the qubit , as the cut - off frequency of the bath is not very large .
when the cut - off frequency of the bath increases to 5 times ( for the model a ) and 25 times ( for the model b ) of the characteristic frequency of the qubit , the @xmath23 and the @xmath99 will lead to almost the same reduced dynamics of the qubit for the respective models . in this section
we investigate the other two models c and d. these two models are developed from above two basic models a and b. model c : the hamiltonian of the model c reads@xmath100 here , the quantum system of interest interacts with the iho and both of them are coupled to their independent baths .
the bath ( @xmath101 ) coupled to the iho and the bath ( @xmath102 ) coupled to the quantum system of interest are considered to be constructed with two sets of harmonic oscillators and those mass , momentums , coordinates , and coupling coefficients are denoted by @xmath103 @xmath104 @xmath105 @xmath106 ( @xmath107 for @xmath101 , and @xmath108 for @xmath102 ) .
as @xmath82 , the @xmath109 is also a controlling parameter for discussion . similar to above subsection we can obtain an esdf for the model as@xmath110 when the cut - off frequency of the bath is infinite . setting the cut - off frequency of the bath
be finite , we have the esdf for the model as @xmath111 if both @xmath80 and @xmath82 are equal to zero , namely the quantum system of interest have not any interaction to other systems except for being directly coupled to a bath ( where the bath is @xmath102 ) the model c will reduce to the model a , and from eqs.(21 ) and ( 22 ) , we can obtain @xmath112 when @xmath113 , the model c reduces to the model b which describes a quantum system of interest interacting with its environment through an iho , and from eqs.(21 ) and ( 22 ) , we have @xmath114 model d : model c describes that the quantum system of interest interacts with an iho and both of them are coupled to two independent baths .
however , when the quantum system is very closed to the iho , they may embed in one common bath , then the hamiltonian of the total system becomes @xmath115 ^{2}. \label{4 - 1}\ ] ] similarly , we can obtain an esdf from infinite cut - off frequency as @xmath116 e^{-\omega/\omega_{c}}.\label{4 - 4}\ ] ] setting the cut - off frequency of the bath be finite , we have @xmath117 \theta\left(\omega\right ) , \label{4 - 2}\ ] ] where@xmath118 @xmath119 if we set @xmath120 @xmath121 and @xmath122 the model d will reduce to the model a and from eqs.(26 ) and ( 27 ) , we have @xmath123 setting @xmath124 the model d will reduce to the model b and from eqs.(26 ) and ( 27 ) , we can obtain @xmath125 by using the esdfs @xmath126 and @xmath127 we can compare the reduced dynamics of the qubit in models c and d , as we have done in last subsection . we can plot above different spectral density functions as fig.4 . here , we are more interested in the question : in which model ( c or d ) the qubit has longer decoherence and relaxation times when the values of all of the parameters in the models are the same ? using the spectral density functions @xmath128 of the model c , and @xmath127 of the model d , we solve the reduced dynamics of the qubit in the two models .
the evolutions of the elements of the reduced density matrix of the qubit are plotted in fig.5 .
here , we set @xmath129 , @xmath130 , and @xmath131 .
other parameters are set with the same values as in fig.2 .
it is shown that the qubit in the model d has longer decoherence and relaxation times than it has in the model c at the same finite cut - off frequency of the baths .
the spectral density function is in fact a bridge to connect the microcosmic and macroscopical physical quantities in dissipative quantum theory . at low temperature ,
the excitations of the bath modes with high frequencies are suppressed .
so the spectral density function derived from infinite cut - off frequency of the bath will be out - of - true . in this paper
we investigate the esdfs @xmath132 for models ( @xmath133a , b , c , d ) at infinite and finite cut - off frequencies of baths from the microcosmic dynamical models by using the method initiated by leggett et .
as the cut - off frequency of the bath is comparable to the characteristic frequency of the quantum system of interest , the non - markovian effects may also be important . in this case
, it is improper to use any markovian approximations to investigate the the reduced dynamics of quantum system .
so we used a numerically exact itm algorithm based on the quapi scheme solving the reduced dynamics of qubits . in order to clearly compare the difference of the dynamics of quantum systems resulting from different esdfs @xmath134 and @xmath135
, we used the effective coupling constants calculating the dynamics of the qubit in each models and fixed other parameters such as the tempurature is set @xmath136k .
it is shown that a qubit ( the quantum system of interest ) in these models has different dynamics when they are governed by the spectral density functions @xmath134 and @xmath12 as the cut - off frequencies are not too large . the difference increase with the decrease of the cut - off frequency .
when the cut - off frequencies of the baths are larger about 5 times ( for the models based on model a ) and 25 times ( for the models based on model b ) than the characteristic frequency of the qubit , the two kinds of spectral density functions result in almost the same reduced dynamics .
in addition , we obtain that a qubit will have longer relaxation and decoherence times when it is coupled to an iho and both of them are coupled to a common bath rather than to their respective baths .
this project was sponsored by national natural science foundation of china ( grant no .
10675066 ) , natural science foundation of ningbo city ( grant no.2008a610098 ) and k c wong magna foundation in ningbo university .
in the appendix we give a proof of eq.([1 - 9 ] ) , namely prove equation @xmath137 setting@xmath138 and from fig.5 we see that there is not any singular point in the loop of @xmath139 .
so by using the cauchy s theorem we have@xmath140 for curve @xmath141 , because@xmath142 we have@xmath143 for curve @xmath144 , we replace pramaters @xmath145 and @xmath146 with @xmath147 and @xmath148 and setting @xmath149 because @xmath150 @xmath151 and@xmath152 we have@xmath153 for linear @xmath154 , the integral becomes @xmath155 setting @xmath156 we have@xmath157 here,@xmath158 with@xmath159 so we have@xmath160 from eqs.([a2 ] ) , ( [ a4 ] ) , ( [ a6 ] ) , and ( [ a14 ] ) we have the integral in linear @xmath161 as @xmath162 | in this paper we firstly obtain two kinds of effective spectral density functions by setting the cut - off frequencies of baths be infinite and finite .
secondly , we investigate the reduced dynamics of open qubits in four kinds of systems constructed with the basic spin - boson model .
it is shown that the qubit has different dynamics governed by the two kinds of spectral density functions .
in addition , we obtained that a qubit coupled to an intermediate harmonic oscillator has longer decoherence and relaxation times as they are coupled to a common bath than to their respective baths . in solving the dynamics of qubits we use a numerically exact algorithm , iterative tensor multiplication algorithm based on the quasiadiabatic propagator path integral scheme . |
dirac s somewhat unfortunate statement about interference between two ( or more ) photons never occurring has led to fruitful discussions clarifying the understanding of interference from a quantum mechanical point of view .
these days textbook presentations elucidate that single photon light fields behave just like classical states of light @xcite when used in a young s double - slit interference experiment @xcite . to exhibit non - classical features , measurements which detect two- or more photon _ coincidences _
, phase sensitive measurements such as homodyning and quantum tomography @xcite , and waiting - time distribution ( anti - bunching ) measurements can be performed @xcite .
single photon detection , however , i.e. intensity or first - order coherence measurements @xcite , are often held to be classical in character @xcite .
this conclusion stems from the analysis of young s interference patterns for single photon and glauber - coherent states of light @xcite , single photons emitted from two atoms @xcite and thermal fields @xcite . in these examples quantum and classical predictions for the form and visibility of interference patterns are identical @xcite .
some physicists have therefore concluded that `` experiments of this kind which measure the first - order correlation functions of the electromagnetic field do not distinguish between the quantum and classical theories of light '' @xcite .
one can even find more generalized statements such as `` for fields with identical spectral properties it is not possible to distinguish the nature of the light source from only the first - order coherence properties '' @xcite .
differences between classical and quantum expressions for first - order coherence phenomena are commonly attributed to different frequency modes only @xcite , but we will see that they can occur for monochromatic fields as well @xcite . to explain the perceived equivalence between the classical and quantum behaviour in first - order coherence phenomena some books refer to dirac s assertion regarding interference between two photons not occurring @xcite .
let us seek a more detailed explanation instead .
the purpose of this @xmath1 is twofold .
firstly i will give a simple general proof for the equivalence between the classical and quantum behaviour in first - order coherence phenomena for the conventional young s double - slit setup .
and then , i will derive the simplest possible example based on photon statistics which should show maximal violation of the classically expected behaviour : whereas in the classical case an interference pattern with perfect visibility is observed , one should find the same in one but zero visibility in another corresponding quantum case .
firstly , let us briefly reexamine young s double - slit : it is , assuming very narrow slits , a pedagogically appealing choice for explaining classical and quantum interference effects since it is an intuitive , geometrically simple , pure two - mode system . in its classic form
it has the drawback , though , of not representing the most general case since both slits are illuminated from the same source .
this limits the choice of states inside the interferometer to binomially distributed states of light which indeed can not show a behaviour different from classical states @xcite . the proof is given below .
giving up on this restriction allows us to tailor the quantum states needed for the non - classical behaviour of the first - order coherence we want to demonstrate here .
a young s double - slit setup can be cast into the general form of a two - mode interferometer illuminated with two _ different _ fields at its two input ports by being illuminated through a semitransparent beam - splitter positioned right between its two slits .
such a modified young s setup would correspond to a spatially constricted mach - zehnder interferometer . in order to avoid all issues regarding the spatial mode arrangement and parameterization of slit and screen locations @xcite we
will therefore , from now on , consider a general two - mode ( mach - zehnder ) setup as displayed in fig .
[ figure1 ] .
= 2.6 in = 1.4 in assuming that both beam - splitters are balanced and equal we choose the modes such that the action of the first beam splitter is described by @xcite @xmath2 the photons following channel _ 1 _ are delayed by a tunable phase @xmath3 and are subsequently mixed with the _ 2_-channel to form the detector modes @xmath4 and @xmath5 : @xmath6 see fig .
[ figure1 ] . for the case of a single photon entering the interferometer through mode
@xmath7 we thus receive the final state @xmath8 and the well known classical photo - detector response probabilities or intensities @xmath9 note that for a single photon entering through mode @xmath10 the role of @xmath11 and @xmath12 or the signs are exchanged .
a simple calculation furthermore confirms that this state is intensity balanced , that is , inside the interferometer intensities are equal @xmath13 let us summarize and generalize this result .
the input modes expressed in terms of the output modes are @xmath14 as can easily be read off eq .
( [ state.1.out ] ) .
the result for @xmath10 is @xmath15 for a conventional young s double slit setup the chances of a single photon to pass slit _ 1 _ are equal to it passing slit _ 2 _ , for every passing photon @xmath16 @xcite .
every fock - state component results in a symmetric binomial distribution of photons , since they randomly pass one hole or the other @xcite . in order to find the general expression for these binomial states ,
let us recast this result in the operator language employed in the first equation of ( [ mode.transformation ] ) .
we find that an arbitrary pure input state for a conventional young s setup has the form @xmath17 using eq .
( [ a.mode.in ] ) it immediately follows that @xmath18 where @xmath19 is the average photon number of @xmath20 in equation ( [ young.state.in ] ) . comparing this expression with ( [ signal.1.a ] )
we see that the textbook case of a conventional young s double - slit indeed always shows classical behaviour .
let us now study the general case of any pure states occupying modes @xmath7 and @xmath10 simultaneously . in order to get a better understanding for what we are in search of ,
let us however briefly remind ourselves of some basics of the theory of optical coherence .
the absolute value of the relative first - order coherence function @xmath21 is connected with the visibility @xmath22 of the interference pattern by @xcite @xmath23 in our case of equal intensities this yields the result that visibility equals first order coherence @xcite . with eq .
( [ signal.a.young ] ) we have , in other words , shown that in a young s double slit setup all light states @xcite appear to be first - order coherent , meaning @xmath24 everywhere @xcite .
since this is obviously the largest possible degree of first - order coherence the signatures of a modification due to quantum statistics can only lead to its _
suppression_. in order to study the general case of any pure states occupying modes @xmath7 and @xmath10 simultaneously we will , for simplicity , expand this state in terms of the modes _ 1 _ and _ 2 _ _ inside _ the interferometer .
@xmath25 with eqs .
( [ mode.transformation.2 ] ) we find @xmath26 and the corresponding expectation value for state ( [ general.state.in ] ) is @xmath27 of the state s density matrix only the diagonal terms contribute to the background intensity and only the first off - diagonal terms determine the interference pattern . to completely erase the interference pattern we want to get rid of the first off - diagonal terms . for single photon states that is obviously impossible if one maintains the restriction of balanced illumination of both interferometric paths . indeed assuming @xmath28 immediately recovers up to a phase
our classical result ( [ signal.1.a ] ) . adding a second photon changes this picture considerably
let us first consider the conventional young s setup again , a two - photon fock - state incident in mode @xmath7 impinging on the first beam - splitter @xmath29 leads to the inside state @xmath30 the amplitudes precisely match the corresponding weight factors in the general intensity expression ( [ expec.a.general ] ) rendering this state first - order coherent .
removal of the @xmath31-term will completely suppress this first - order coherence .
the two - photon state furthest deviating from the conventional young s mono - mode input is obviously the balanced bimodal input state @xmath32 which , indeed , does not show any first order interference .
this state is intensity balanced inside the interferometer .
it therefore only remains to be shown that this state s non - classical behaviour can not be attributed to some kind of randomness in phase .
guided by the discussion of the general intensity expression ( [ expec.a.general ] ) we can expect that a second - order coherence measurement @xcite ( two - photon coincidence detection with probability @xmath33 ) should give us an interference pattern of higher order since it connects diagonal density matrix elements with second off - diagonal elements .
the corresponding expectation value for the detection of two photons in channel @xmath34 has indeed perfect visibility @xmath35 here the prefactor @xmath36 normalizes the sum of all detection probabilities per shot to unity .
we note that the effective de broglie wavelength of this interference effect is halved with respect to the single photon detection case @xcite . for our present considerations
the most striking feature is the perfect visibility of this interference pattern which holds for @xmath33 , @xmath37 and @xmath38 as well .
these are the probabilities to detect two photons in coincidence in the respective channels indicated by the subscripts @xcite .
it is not difficult to show that a second - order pattern with perfect visibility can not coexist with a first - order pattern . with these perfect second - order visibilities state ( [ two.photon.a.b ] ) has to be free of random phase fluctuations .
we are led to consider a state that has a stable relative phase @xmath39 across both arms of the interferometer and illuminates both channels of the interferometer with equal intensity .
when we seek a classical description of this state inside the interferometer before it passes the phase shifter we have to choose a combination of coherent states @xmath40 of the form @xmath41 classically we would expect this state to be first - order coherent , quantum mechanically , however , _
the first - order coherence is suppressed .
_ it is clear from our discussion of the general intensity expression ( [ expec.a.general ] ) that , by moving to the second off - diagonal of the density matrix , we have found the simplest possible case of complete suppression of first - order coherence through quantum statistics .
in order to study the transition from classical to quantum behaviour let us look at states that linearly interpolate ( @xmath42 $ ] ) between the above extremes ( [ two.photon.a ] ) ( @xmath43 ) and ( [ two.photon.a.b ] ) ( @xmath44 ) @xmath45 these states first - order visibilities are @xmath46 , as can be seen from @xmath47 the respective channel intensities inside the interferometer are no longer balanced @xmath48 we will therefore have to compare classical and quantum behaviour using a channel intensity - independent measure , namely the ratio of first order coherence functions @xmath49 , where the subscripts stand for quantum and classical case . because of the stable relative phase @xmath50 across both arms of the interferometer and our assumed perfect mode match @xcite we expect @xmath51 @xcite . from eq .
( [ connection.visibility.g1 ] ) we can hence infer that the expected classical visibility only depends on the channel intensities , @xmath52 , which are given above . dividing the visibilities yields @xmath53 falls far below the classically expected value of 1 when we approach the bimodal input state ( [ two.photon.a.b ] ) at @xmath54 .
= 3.0 in = 1.4 in in the quantum - classical comparison leading to fig .
[ figure2 ] we assumed the same channel intensities @xmath55 and @xmath56 and stability of the cross - phase @xmath57 .
one can compare the classical to the quantum case employing a different set of assumptions .
we could for example keep the input intensities @xmath58 and @xmath59 matched and also vary the relative phase between modes @xmath7 and @xmath10 by , say , multiplying the latter with a phase factor : @xmath60 .
input state @xmath61 of eq .
( [ interpolation.state.in ] ) now depends on @xmath62 and @xmath63 and so does its classical counterpart @xmath64 here the glauber - coherent state amplitudes @xcite match the input intensities @xmath65 and @xmath66 of state ( [ interpolation.state.in ] ) .
the interference signals seen by detector @xmath34 for the phase tunable version of ( [ interpolation.state.in ] ) are @xmath67 and @xmath68 for its classical counterpart ( [ interpolation.state.classical ] ) .
the resulting visibilities are the pythagorean combinations of the trigonometric coefficients @xmath69 the quantum - statistical suppression @xmath70 @xmath71 is again strongest for @xmath54 .
the two classical models , just presented , are incompatible with each other since the second model s input intensities @xmath58 and
@xmath59 do not yield the first model s channel intensities @xmath55 and @xmath56 and vice versa .
consequently this discussion is to some extent arbitrary but it still serves to show that a physically reasonable classical model can not explain the quantum - statistical suppression of first - order coherence .
for experimental confirmation of the results presented here it is important to realize how easy it is to prepare the interpolation state , let us therefore rewrite eq .
( [ interpolation.state.in ] ) with @xmath72 as a mixing angle @xmath73 this form can be synthesized using spontaneous parametric down - conversion @xcite and passive linear devices .
let us think of a configuration where signal @xmath74 and idler @xmath75 photons of a single pair are mode - matched except for orthogonal polarization , say horizontal @xmath76 and vertical @xmath22. the initial state is thus in the state @xmath77 .
next the @xmath78-mode s polarization is turned such that it overlaps @xmath79 by an amount @xmath80 , subsequently the two spatial modes @xmath79 and @xmath78 are mixed with a polarization - insensitive beam splitter according to transformation ( [ mode.transformation.2 ] ) .
this yields the combination @xmath81 | 0 \rangle % \ ; , \label{prep.state.in.angle}\end{aligned}\ ] ] where the identifications @xmath82 , @xmath83 and @xmath84 have been made .
the presence of the @xmath85 photon allows us to identify the unwanted @xmath86-component which otherwise would pollute the contribution from the desired first two terms which constitute state ( [ interpolation.state.in.angle ] ) .
to conclude , we have given a general proof for the equivalence of the first - order coherence function for the classical and the quantum case measured with a conventional young s double - slit setup .
we have shown that this equivalence does not hold for arbitrary two - mode interferometers and have derived the simplest possible case for maximal quantum - statistical suppression of first - order coherence .
this case was analyzed by comparing it to two classical scenarios .
finally have we shown how to synthesize the quantum states used in this paper in order to allow for experimental confirmation of the predicted quantum - statistical suppression of first - order coherence .
i use the terminology of references @xcite and refer to @xmath87 as a _ first _ order function since it is of first order in the measured intensities , this is at variance with the terminology in other parts of the literature @xcite where it is refered to as a function of _ second _ order , in terms of amplitudes .
i will only consider pure states populating single spatio - temporal modes @xcite ( quasi - monochromatic , one polarization ) of sufficient transversal coherence length as to coherently illuminate both slits . in the case of paths overlapping at the beam splitters
i always assume perfect matching of these modes .
the generalization to mixtures of such pure states is straightforward @xcite . | classical optical interference experiments correspond to a measurement of the first - order correlation function of the electromagnetic field .
the converse of this statement : experiments that measure the first order correlation functions do not distinguish between the quantum and classical theories of light , does not always hold .
a counterexample is given .
# 1@xmath0#1 |
one of the central goals in contemporary observational cosmology is to detect the postulated background of primordial gravity waves predicted by inflation .
the most direct observational signature of these gravity waves is a particular pattern in the polarization of the cosmic microwave background ( cmb ) known as b - modes .
the amplitude of these gravity waves is typically parametrized in terms of the tensor - to - scalar ratio , @xmath6 ( see , e.g. * ? ? ? * and references therein for a thorough review on inflation ) . during the last few years
many experiments have been planned , built and fielded to measure @xmath6 , and the first relevant b - mode constraints have been already published by bicep ( @xmath7 , chiang et al .
2010 ) and quiet ( @xmath8 ; quiet 2011 , 2012 ) .
other ground - based and sub - orbital experiments are expected to vastly improve on these limits in the very near future . in order to make an actual detection of the inflationary gravity waves
, it is widely believed that a sensitivity of @xmath9 will be required . in terms of map - domain sensitivity
, this corresponds to a signal with an rms of a few tens of nk .
thus , not only will exquisitely sensitive detectors be needed , but also detectors with extremely low systematics .
however , the single most problematic systematic for future b - mode experiments is likely not to come from the instrument itself , but rather from the sky : non - cosmological galactic and extra - galactic foregrounds , for instance synchrotron and thermal dust , radiates with a temperature of several @xmath10 on large angular scales in the frequencies relevant for cmb measurements ( e.g. , * ? ? ?
* and references therein ) .
therefore , in order not to be foreground dominated , these foregrounds must very likely be suppressed by perhaps an order of magnitude or more . the only way to achieve this is by making multifrequency observations of the same fields of the sky , and
exploit the different frequency dependency of the various components to separate out the cosmological cmb signal from the non - cosmological foregrounds . as of today
, a very large fraction of the information we have about polarized foregrounds on large angular scales comes from the wmap satellite experiment , and in particular the lowest frequency channel at 23 ghz ( k - band ) .
this map is routinely used both for studies of foregrounds themselves , and as ancillary data for other experiments .
it is therefore critical to understand the systematic limitations inherent in these data . in this paper
we measure the spectral indices of four particularly bright compact objects ( tau a , virgo a , 3c273 and fornax a ) , with the goal of understanding some of the issued involved in spectral index estimation for cmb data in general : by considering high signal - to - noise objects with known properties , we have a clear a - priori prediction , and deviations from these expectations would indicate either model problems or systematic errors .
[ [ sky - maps - and - processing ] ] sky maps and processing + + + + + + + + + + + + + + + + + + + + + + + in this paper we consider the 7-year wmap sky maps @xcite , coadded over years and pixelized at a healpix resolution of @xmath11 , corresponding to @xmath12 pixels .
these data are available from lambda , including all necessary ancillary data , such as beam profiles and noise model .
most of our analysis is performed with the k- and ka - band data , although in one particular case we also consider the q - band data .
all analyses are carried out in antenna temperature units , and given that we will consider objects with steep synchrotron - like spectra we adopt effective frequencies of 22.45 ghz ( k - band ) , 32.64 ghz ( ka - band ) and 40.50 ghz ( q - band ) , respectively @xcite .
before one can estimate spectral indices across frequencies , it is necessary to bring all maps to a common angular resolution .
we therefore smooth all maps to an effective resolution of @xmath0 fwhm by first deconvolving the instrument beam and then convolving with a gaussian beam of the desired size .
note that the smoothing scale of @xmath0 is a particularly common value adopted in the literature , and the results presented here are therefore of wide interest .
estimation of uncertainties for all scalar quantities is done by forward monte carlo simulations .
that is , we add smoothed noise realizations to the actual wmap data based on the provided noise model , evaluate each statistic for each simulation , and then compute the resulting standard deviation over the ensemble .
although there already is a noise component present in the wmap data , this is identical for all simulations , and therefore do not contribute to the variance .
we emphasize , though , that uncertainties estimated in this manner are only statistical in nature , and do not account for systematic errors .
[ [ data - selection ] ] data selection + + + + + + + + + + + + + + in this paper we consider the four particularly bright point sources listed in table [ tab : image ] .
these were selected by thresholding the k - band polarization map , @xmath13 , at @xmath14 , and discarding all regions that either show obviously extended features or have a strong background .
this left us with tau a as the only near - galactic source , and three high - latitude sources ( virgo a , 3c273 and fornax a ) .
for further details on the polarization properties of these objects , see , e.g. , @xcite . only pixels in a @xmath0 radius disc around each source
were kept for analysis , although we also tried @xmath15 discs , obtaining consistent , but slightly more noisy , results .
note that a gaussian beam of @xmath0 fwhm ( @xmath16 ) has dropped off to 6% of its peak value at a distance of @xmath0 , and most of the volume is therefore contained within this radius .
[ [ data - model ] ] data model + + + + + + + + + + the low - frequency wmap polarization observations are strongly dominated by synchrotron emission which has a sharply falling spectrum .
we therefore approximate the total sky signal by a single power - law , resulting in the following data model , @xmath17 here @xmath18 is an @xmath19 matrix listing the stokes @xmath20 , @xmath21 and @xmath22 parameters for all relevant pixels at frequency @xmath23 columnwise , @xmath24 is a @xmath25 matrix denoting convolution with the common instrumental beam , @xmath26 denotes the true sky signal amplitude as measured at a reference frequency @xmath27 , and @xmath28 is ( smoothed ) instrumental noise .
all values are defined in antenna temperature units .
coordinates are defined according to the healpix convention @xcite .
given the data and model described in section [ sec : data ] , we estimate the polarization angle and fraction , the spectral index and the apparent shape of each source .
first , we note that each of the four objects considered here are well known in the literature , and have known polarization properties .
further , they are all known to be much smaller than @xmath0 in angular dimensions ( see table [ tab : image ] for precise details ) , and their polarization angles are known to be quite stable as a function of frequency .
( for example , the polarization angle of tau a is known to vary by only a few degrees over more than 10 decades in frequency ; e.g. , aumont et al . 2010 . )
we therefore assume that there is no real substructure within each source on the scales we consider .
lccccccccc tau a & @xmath29 & @xmath30 & @xmath31 & @xmath32 & @xmath33 & @xmath34 & @xmath35 & @xmath36 & @xmath37 + virgo a & @xmath38 & @xmath39 & @xmath40 & @xmath41 & @xmath42 & @xmath43 & @xmath44 & @xmath45 & @xmath46 + 3c273 & @xmath47 & @xmath48 & @xmath49 & @xmath50 & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath55 + fornax a & @xmath56 & @xmath57 & @xmath58 & @xmath59 & @xmath60 & @xmath61 & @xmath62 & @xmath63 & @xmath64 [ [ polarization - angle - and - fraction ] ] polarization angle and fraction + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + since our objects effectively are point sources with constant polarization angle , there should ( ideally ) be a single well - defined coordinate system in which all signal is aligned with the stokes @xmath21 parameter .
we search for this direction , @xmath1 , by minimizing the signal in the corresponding @xmath22 parameter , @xmath65 where @xmath66 and @xmath67 are the stokes parameters in galactic coordinates , and @xmath68 is the noise level . having rotated the data into the intrinsic polarization direction of the source , the polarization fraction is found simply by @xmath69 . [ [ observed - ellipticity - and - fwhm ] ] observed ellipticity and fwhm + + + + + + + + + + + + + + + + + + + + + + + + + + + + + although we smooth the data to a common angular resolution , and therefore should expect that the observations to have the desired fwhm , this is not true in practice due to beam asymmetries . to study the effective beam as a function of stokes parameters , we rotate the original map by a rotation angle @xmath1 into a new coordinate system @xmath70 , and consider all angles between 0 and @xmath71 in steps of @xmath72
. then , in this new coordinate system we fit an elliptical gaussian , @xmath73 , to the @xmath21 signal by minimizing @xmath74 where @xmath75 is the source amplitude , @xmath76 is the ellipticity , and @xmath77 is the direction of the major semiaxis .
( note that it is sufficient to consider only the @xmath21 component , because we rotate through all angles @xmath1 .
thus , @xmath78 corresponds to @xmath22 in the original system . )
[ [ spectral - indices ] ] spectral indices + + + + + + + + + + + + + + + + finally , we estimate spectral indices for both @xmath79 and @xmath80 using a standard tt - plot approach . for a single pixel with noiseless data , this approach is simply defined by @xmath81}{\log [ \nu_1/\nu_2]}.\ ] ] however , for multiple noisy observations more robust results are obtained by fitting a straight line , @xmath82 , to @xmath83 as a function of @xmath84 .
an additional advantage of this method is its insensitivity to absolute offsets in the data .
the spectral index is given as @xmath85 .
since both @xmath83 and @xmath84 have uncertainties , it is important to use a method that supports uncertainties in both directions with making the linear fit . in this paper
, we adopt the approach described by @xcite . as in the case of the beam parameters
, we also compute the spectral index as a function of polarization angle .
lcccccc tau a & @xmath86 & @xmath87 & @xmath88 & @xmath89 & @xmath90 & @xmath91 + virgo a & @xmath92 & @xmath93 & @xmath94 & @xmath95 & @xmath96 & @xmath97 + 3c273 & @xmath98 & @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 + fornax a & @xmath104 & @xmath105 & @xmath106 & @xmath107 & @xmath108 & @xmath109 estimation of uncertainties for spectral indices is a non - trivial issue , because the inherent systematic errors turn out to be significantly larger than the statistical .
we therefore add a systematic error term in quadrature to the statistical error .
the systematic error is defined by splitting the observed data points in two sets , according to whether @xmath83 is larger or smaller than @xmath110 , and estimate a new slope for each set .
the systematic error is taken to be the half difference between the two slopes .
in table [ tab : image ] we list the position and apparent ( beam - convolved ) size and shape of each of the four objects under consideration . the polarization fraction and angles , and spectral indices
are tabulated in table [ tab : polarization ] .
images of tau a are shown in figure [ fig : objects ] , both for k- ( left column ) and ka - band ( right column ) , and for stokes @xmath21 and @xmath22 parameters . in order to highlight the beam differences between these cases ,
we have first adopted a coordinate system which is offset by @xmath111 from the intrinsic polarization direction of tau a. this ensures a significant signal - to - noise in both @xmath21 and @xmath22 .
second , the color scale is tuned to highlight the tails of the instrumental beam , and scaled properly between the two frequencies taking into account the spectral index of tau a. the main results of this paper are shown in figures [ fig : beam_vs_psi ] and [ fig : beta_vs_psi ] . the first figure shows the beam parameters for tau a as a function of polarization orientation , and the second shows the spectral index as a function of polarization direction for all four sources . in the latter , the black points indicate the spectral index computed from k- and ka - bands , and ( for tau a only ) the red points show the spectral index between ka- and q - band . figure [ fig : indfit ] shows a subset of the tau a tt plots that are used for the k - ka calculations , corresponding to @xmath3 , @xmath112 and @xmath113 , respectively . as seen from the results shown in figure [ fig : beta_vs_psi ] ,
the polarized spectral index as measured by wmap between k- and ka - band at @xmath0 angular scale depends strongly on the coordinate system in which the index is computed .
for tau a , the derived index varies between , say , @xmath114 for a rotation angle of @xmath3 and @xmath115 for @xmath5 .
this effect is statistically highly significant , and it is robust with respect to instrumental noise and background levels ( see figure [ fig : indfit ] ) .
it therefore indicates the presence of a real systematic effect not taken into account in the present analysis . to understand these structures in greater detail , we construct an elliptical gaussian model of tau a based on the parameters listed in tables [ tab : image ] and [ tab : polarization ] at k- and ka - band , and estimate the spectral index from the resulting noiseless model , as for the real data .
the results are shown in figure [ fig : beta_vs_psi_sim ] .
clearly , the model faithfully reproduces the observed structures .
the only difference is a slight vertical offset , which is due to the fact that the measured spectral index ( reported in table [ tab : polarization ] ) is not a perfectly unbiased estimate of the true spectral index in the presence of systematics .
we can now use this model to understand the relative importance of the various systematic effects . to do so , we start out with an ideal model , adopting the observed k - band parameters also for ka - band , and set the ka parameters one - by - one to their true values . the results from this exercise are shown in figure [ fig : beta_vs_psi_systematics ] . here
we see that the most important systematic by far is the detector angle , and this effect alone reproduces the signal seen in figure [ fig : beta_vs_psi_sim ] very well .
the second most important effect is the beam ellipticity , which is at least three to four times smaller than the detector angle effect over most of the well - sampled regions of the polarization orientation .
other effects are small compared to these two .
we have studied four particularly bright polarized point sources in the 7-year wmap data , with the goal of understanding the effect of systematics on polarized spectral index estimation .
this topic is important for at least two reasons .
first , the wmap polarization sky maps represents the best currently available full - sky measurements of the polarized foregrounds at cmb frequencies . as a result
, they play a critical role in the analysis and optimization of existing and future b - mode experiments .
second , many ground - based and sub - orbital experiments use the wmap polarization measurements of tau a directly as a calibration source for both detector angles and absolute gain . in this paper
, we have found that the observed polarized spectral index of the relevant sources depends sensitively on the coordinate system in which the index is estimated .
for example , the spectral index of tau a is @xmath2 for a coordinate system rotated by @xmath116 relative to the intrinsic polarization direction of the source , while it is @xmath4 in a coordinate system rotated by @xmath112 .
the most significant contributor to this effect is the slightly different polarization angles of the k- and ka - band detectors , with some smaller contribution coming from beam asymmetries .
experiments that , directly or indirectly , use the k- and ka - band measurements of tau a as a calibrator source should take into account these systematic uncertainties when performing their analyses . finally , we note that the test described in this paper is very simple to implement , only takes a few cpu seconds to run , and have a very direct and intuitive interpretation
. we therefore expect other experiments to find it useful as a test of their own systematics , in particular when applied to tau a. this project was supported by the erc starting grant stg2010 - 257080 and a leverhulme visiting professorship for hke .
some of the results in this paper have been derived using the healpix @xcite software and analysis package .
liddle , a. r. , & lyth , d. h. 2000 , cosmological inflation and large - scale structure , by andrew r. liddle and david h. lyth , pp .
isbn 052166022x .
cambridge , uk : cambridge university press , april 2000 . | we study four particularly bright polarized compact objects ( tau a , virgo a , 3c273 and fornax a ) in the 7-year wmap sky maps , with the goal of understanding potential systematics involved in estimation of foreground spectral indices .
we estimate the spectral index , the polarization angle , the polarization fraction and apparent size and shape of these objects when smoothed to a nominal resolution of @xmath0 fwhm .
second , we compute the spectral index as a function of polarization orientation , @xmath1 . because these objects are approximately point sources with constant polarization angle , this function should be constant in the absence of systematics .
however , for the k- and ka - band wmap data we find strong index variations for all four sources .
for tau a , we find a spectral index of @xmath2 for @xmath3 , and @xmath4 for @xmath5 . on the other hand ,
the spectral index between ka- and q - band is found to be stable . a simple elliptical gaussian toy model with parameters matching those observed in tau a reproduces the observed signal , and shows that the spectral index is in particular sensitive to the detector polarization angle . based on these findings ,
we first conclude that estimation of spectral indices with the wmap k - band polarization data at @xmath0 scales is not robust .
second , we note that these issues may be of concern for ground - based and sub - orbital experiments that use the wmap polarization measurements of tau a for calibration of gain and polarization angles . |
experimental data on solar neutrinos ( @xmath16@xmath17)@xmath18 has accumulated rapidly in the last few years by the operation of several large detector facilities such as super - kamiokande ( sk ) , the sudbury neutrino observatory ( sno ) and kamland . the data from these detectors were precise and informative .
the neutral current signals ( nc ) from @xmath3b neutrinos in sno @xcite showed that @xmath16@xmath17 flavor was not conserved because the flavor - blind nc rate explicitly exceeded the charged current ( cc ) @xmath16@xmath17 signal rate .
the absolute nc rate was close to the @xmath3b @xmath16@xmath17 flux of the standard solar model ( ssm)@xcite . the ( @xmath16-electron ) scattering ( es ) signal in sk @xcite indicated the _ _ type _ _ of flavor conversion with a highly precise energy - independent es spectrum in the measured window 5 - 15 mev .
this result severely restricted the possible @xmath16@xmath17 models .
global analyses of solar @xmath16@xmath17 data pinpointed matter conversion ( msw effect ) in the large mixing ( mass ) ( lma - msw ) regime as the most favored model of solar neutrinos@xcite . finally , kamland observed oscillations of _ _ terrestrial _ _ _ _ anti__neutrinos@xcite .
the resulting parameters were approximately consistent with lma - msw neutrinos , as expected by the invariance of cpt symmetry .
the solar neutrino problem therefore appears resolved . in view of the preponderance of evidence of global consistency for the lma - msw model ,
why search for an alternate model ? in principle , however , no physical effect specific to lma has yet been observed to prove its _ _ uniqueness . _ _ the nc / cc result from sno@xcite , shows only that the @xmath16@xmath17 s oscillate but it does not specify the mechanism as lma - msw .
the lack of a uniqueness proof becomes a central issue ( not just a point of principle ) if an _ _ explicitly testable _ _ competitive model can be invented . in this work
i show that such a solar neutrino model can be constructed . the uniqueness of _ _ this _ _ model can be proved or disproved by a variety of effects at signal energies @xmath195 mev . in particular ,
in contrast to lma - msw , sizable shape distortions are predicted in the 35 mev window in es spectra even though the shape is still flat above 5 mev .
such an effect in solar neutrino spectra is predicted for the first time .
the 35 mev window is now experimentally observable for the first time in a liquid scintillation device such as the already operational kamland detector ; thus , the above prediction can be readily subjected to experimental test .
the physical mechanism of the present model is long wavelength oscillations ( lwo ) in vacuum , discovered by a new approach based on `` spectral engineering '' .
although the underlying parameters were previously detected in the general dragnet of global analysis programs @xcite the _ _ experimental effects _ _ of the model are new , unsuspected and distinctly different from those of generic vacuum oscillations .
the @xmath15 parameters of the model , very different from lma , imply manifest violation of cpt invariance in view of the kamland results . whether this is a fatal flaw is at present unclear from the standpoint of basic theory .
the suggested experimental tests of the new model , however , offer a decisive resolution of the question , thus they are doubly important .
the present work has its origins in a question posed in 1994 , if it was possible to conceive a neutrino oscillation model elusive enough to _
_ escape _ _ detection in experiments at the time .
the answer was yes , if solar neutrinos oscillated at only the lowest energies .
in very long wavelength ( vlwo ) oscillations in vacuum with @xmath11m@xmath20ev@xmath21ev@xmath13 ) , @xmath22b neutrinos are hardly converted ; neutrinos @xmath191 mev are converted , and explain the @xmath23ga results@xcite .
thus , the @xmath3b flux deficit observed in the @xmath24cl data can only be produced by an initial flux _ _ reduced astrophysically to @xmath251/3 .
_ _ this scenario , called `` just - so 2''@xcite , was the first model to be ruled out @xcite by the sno nc result the initial flux is apparently the full ssm flux , thus , _ _ conversion _ _ is centrally responsible for the flux deficit .
such an effect is available via lma - msw especially since it also supplies the crucial undistorted sk spectrum .
thus , conversion models at two extremes produce the observed undistorted @xmath3b spectrum in principle , via : 1 ) vlwo with no @xmath3b @xmath16@xmath17 flavor conversion but a flux reduction by a factor @xmath250.36 ; and 2 ) lma - msw with @xmath16@xmath17 survival of @xmath250.36 but the full initial ssm @xmath22b flux .
the question then arises whether a scenario could be engineered with a _
_ mixture _ _ of _ _ both _ _ flavor conversion _ _ and _ _ a smaller initial flux that leaves the high energy part of the @xmath3b @xmath16@xmath26 spectrum _ _ shape _ _ _ _ invariant _ _ so that all the experimental consequences ( at @xmath275 mev ) follow automatically . by definition
, this does need a smaller initial @xmath3b flux than the ssm / sno value , however , in the spirit of global analyses made so far , one could tolerate in principle a lower flux up to the 2@xmath28 lower limit of the sno nc result , i.e. , @xmath27 75% of the ssm flux .
the surprising initial discovery was that a shape invariant high energy @xmath22b neutrino spectrum could indeed be engineered to a high degree of precision ( within a few percent ) by a sensitive choice of the initial flux and mass - mixing parameters in a very narrow band @xmath11m@xmath13= 70 - 80 @xmath29ev@xmath13 , @xmath30 times larger than the vlow - just - so2 regime @xcite .
the oscillations in the new regime are thus termed long - wavelength or lwo in this paper .
the new mass - mixing band is very different from lma - msw regime . in view of the kamland result ,
one thus immediately faces the problem of violation of cpt invariance by neutrinos . nevertheless , work proceeded undeterred .
recent theoretical ideas appear to allow that the otherwise rigid invariance of cpt may be relaxed in principle in the neutrino sector@xcite .
fig.1 illustrates the central thesis of this work . it shows a series of @xmath31b @xmath32 spectral shapes engineered via lwo to reproduce at high energies , the rate and shape of the lma - msw ( thick red ) solution as closely as possible .
the lma shape is defined as the standard ssm shape ( thin blue ) scaled by the factor 0.365 ( 0.349(22 ) , the cc flux at sno ) .
the convenient definition `` lma = 0.365 ssm '' agrees very closely in the @xmath9(@xmath22b ) and the flat spectral shape observed above 5 mev .
the other spectra are derived with `` shape invariant '' lwo parameter combinations ( see box and caption of fig.1 ) chosen individually and interactively on - line .
the shapes change extremely sensitively with the three @xmath16@xmath17 parameters that tune up a viable solution systematically : the @xmath11m@xmath33 arranges the flavor survival p@xmath34 spectrum such that a roughly horizontal part energywise , is positioned in the @xmath275 mev regime ; the @xmath9(b)parameter scales it into the lma shape window ; the sin@xmath35 that strongly affects the curvature , flattens it into a nearly undistorted shape .
the final angles that expose the new model are in general , smaller than previously examined ; e.g. , solutions with the same @xmath11m@xmath36=76@xmath29ev@xmath36 but [email protected] , beyond the present range @xmath38 0.58 , failed earlier @xcite and in this work .
1 shows that the needed shape invariance can be achieved with precision above @xmath258 mev below which the lwo shapes fan out deviating from the lma - msw shape .
we aim at neutrino reaction signals such as es and cc@xmath27 5 mev using these input shapes .
thus , deviations in fig 1 between 5 and 8 mev appear problematic . in both es - sk and cc - sno that apply the cerenkov technique , the experimental signals of the electron kinetic energy t range above 5 mev . in the quest for lwo
@xmath16@xmath39 shapes that generate es / cc shapes that fit the data , the spectral smearing due to the inherent reaction as well as the detector energy resolution play a key role . in the end , for both cc and es , the overall spectral smearing is such that the dominant contribution to the es and cc signal comes from e(@xmath40 mev where the spectral match is close .
the possible deviations occur mainly near the 5 mev signal threshold where background is strong in both sk and sno .
this effect can be understood in the case of the es shape since the es signal at a given electron energy is derived from the whole neutrino spectrum above that energy , i.e. the reaction smearing effect is complete@xcite .
even a large deviation e.g. , at 56 mev is damped down considerably in the es spectrum . in the case of the cc spectrum t@xmath275 mev which implies e(@xmath16@xmath17)@xmath276.44 mev ,
the deviations look more worrisome since in this case one normally expects 1-to-1 correspondence of e(@xmath2 ) and t , thus no reaction smearing . in reality , there is considerable reaction smearing even for the cc signal since in the reaction @xmath16@xmath39 + d @xmath41p + e , the proton recoils can not be neglected@xcite .
the effect removes the e(@xmath16@xmath17)-t correspondence significantly ( although not to the extent in es ) . the final effect is shown in fig .
2 where the light blue curve for no recoil corrections peaks at 8 mev .
the recoil correction results in a large downshift to @xmath197 mev .
the detector resolution smearing produces a further downshift to @xmath256.5 mev .
thus , also the cc signal shape is determined largely by e(@xmath2)@xmath278 mev .
the plan of search for a new @xmath16@xmath26 model was as follows .
in the first step , the viable basic parameters of the lwo scenarios were fixed by matching predicted shapes and rates to the es - sk result , by far the most the stringent test .
the es spectral / rate match set the primary fix of the sin@xmath35 and @xmath9(b ) values for each selected @xmath11m@xmath33 in the viable band .
the match to the sno cc data followed with the same parameter set except for possibly very slight changes in the scaling @xmath9(b ) only . with a reasonable match to data on @xmath16@xmath17 ( @xmath3b ) , the matches to
low energy data from @xmath24cl and @xmath42ga were tested with the same final set of parameters reached above .
the upper limit of the @xmath11m@xmath33 band is @xmath43@xmath29ev@xmath36 , because , beyond this limit , the spectral deviations around 5 mev become too large to accommodate a spectral match to the data . at @xmath2570 @xmath29ev@xmath36 and below , the required scaling factor @xmath9(b ) dips @xmath190.75 , the 2@xmath28 limit of sno .
the procedure involved the calculation of the es spectra for lma and lwo scenarios normalized to that for the ssm spectrum with no conversion .
thus , the es signal for each scenario is : @xmath44/<\mathrm{w}(\lambda ) > } \ ] ] where w and z are the differential es cross - sections for e- and @xmath29/@xmath45 neutrinos respectively @xcite , averaged over the ssm @xmath16@xmath17 spectrum p(@xmath46 ) @xcite modified by the e - flavor survival function p@xmath34 that , in general , depends on e(@xmath16@xmath17 ) .
the @xmath47 in ( 1 ) indicate convolution with the detector resolution in sk@xcite . for the lma case ,
p@xmath48 is a _ _ constant = 0.365 _ _ , close to 0.349(22 ) ( relative to the ssm flux of @xmath49 cm@xmath36s@xmath51 ) , the value found in cc - sno .
then , w(@xmath46;p@xmath52)= w(@xmath46)@xmath530.365 and z(@xmath46;1-p@xmath52 ) = w(@xmath46)(1-p@xmath54 w(@xmath55 , because of the nearly constant value of z / w@xmath56 0.16 for solar @xmath2 energies @xmath273 mev .
this leads to r(es;lma ) 0.466 @xmath560.465(15 ) , the sk result @xcite relative to the ssm flux above ( the reason for choosing p@xmath34 ( lma)=0.365 ) . in the case of the lwo spectra
, the same procedure was used . here
p@xmath34 is energy dependent , given by the standard oscillation formula for solar baseline : @xmath57\ } } \ ] ] for each value of @xmath11m@xmath13 , using the convoluted spectra @xmath19w(@xmath46;p@xmath58z(@xmath46;(1-p@xmath59 and @xmath19w(@xmath60 , the es signals were calculated as in ( 1 ) .
this spectrum ( energy dependent in principle ) , was matched to the data points of the es - sk spectrum @xcite by varying the parameters @xmath9(es ) ( @xmath191 ) and sin@xmath35 for each value of @xmath11m@xmath33 in the range 70 - 80 @xmath29ev@xmath33 .
the best match was selected interactively for a given value of @xmath11m@xmath33 from the @xmath61of the fit of the r(es ) curve to the es - sk data points .
the results are shown in fig .
3 . the box in fig . 3 shows the values of the parameter combinations [ @xmath11m@xmath33 ; sin@xmath35 ; @xmath9(es ) ] , the `` best match '' [ @xmath62@xmath33 ] ( dof = 16 for 17 sk data points in the range 5 - 14 mev ) and the [ r(es ) area ] normalized to ssm for each curve including es - lma .
the es(lwo ) curves are seen to match the sk data , particularly those for @xmath11m@xmath33 = 75 - 78 @xmath29ev@xmath33 , even in the narrow slit allowed by the high statistical precision of the sk data . indeed , except for the first two data points at the lowest energies in fig . 3 that are the most susceptible to background , the @xmath64 values ( dof = 14 ) for t= 6.514 mev ( also given in the box in fig 3 ) for @xmath11m@xmath33 = 75 - 78 @xmath29ev@xmath33 jump to 1.05 to 1.15 with probabilities in the range of 35% compared to the best fit 50% value for @xmath62@xmath33=1 .
the rate matches are excellent ( see table 2 below ) .
our new model thus passes its most severe test .
a key reason is the significantly smaller mixing angles used here ( compared to previous analyses ) that flatten the es spectra sufficiently to be scaled into the es data window .
another key effect is detector resolution as shown in fig .
4 . the unconvoluted es(lwo ) curve ( in blue ) with a discouraging shape ( even without the low energy points ) reduces to the much more acceptable curve ( in red ) that includes the effect of the sk energy resolution .
the cc reaction @xmath65 has been applied at sno to measure directly the @xmath32 @xmath31b spectrum .
the theoretical cross sections for this reaction including the recoils of the two protons in the final state have been extensively tabulated@xcite .
the calculation of the cc spectra for lwo and lma , input @xmath16@xmath17 spectra must be summed over the contributions of @xmath16@xmath17 from 0 to e(@xmath16@xmath17)(max ) = t+ q ( = 1.44 mev ) for each signal electron kinetic energy t instead of just from the single @xmath16@xmath17 energy t+q .
the cross section c(@xmath46 ; p@xmath34 ) is first averaged over the input solar @xmath16@xmath17 spectrum p(@xmath46 ) modulated by the p@xmath34 of the oscillation scenario .
p@xmath34(lma ) = 0.365 .
p@xmath34(lwo ) is given by ( 2 ) .
for each value of @xmath11m@xmath33 the cc shapes were generated as : @xmath66 where p@xmath34 was calculated with the same mixing parameter fixed in the es - sk spectral match .
the @xmath47 indicates , as before , convolution with the sno detector resolution@xcite .
@xmath9(cc ) is the scale factor applied to match the lwo - cc spectrum to lma - cc spectrum measured at sno .
the shape matches aimed at fits ( by eye ) within a @xmath6712% band , the error band in the measured sno - cc spectrum@xcite .
5 shows the lwo - cc matches to the lma - cc spectrum ( thick red line marked ssm*0.365 ) .
the legend box shows in [ ] the signal areas under the curves in the window 513.5 mev .
the scale parameter @xmath9(cc ) is the true flux of @xmath16@xmath17(@xmath3b ) .
the spectral matches in fig .
5 are good with only a @xmath2510% deviation even at the lowest energies that are most susceptible to errors from background and the disentanglement of the relatively broad nc _
_ @xmath68-ray _ _ _ _ shower _ _ signal that occurs at the effective energy of @xmath255 mev instead of at its full energy of 6.15 mev@xcite .
the rate matches are also good ( see table 2 ) .
the best match lwo parameter group in the cc matches as well as overall , is [ @xmath11m@xmath13 ( @xmath29ev@xmath13)/sin@xmath69(cc ) ] = [ 77/0.57/0.82 ] .
we now consider how the low energy parts of the new model , constructed so far only with high energy @xmath3b solar neutrinos in mind , match the signals from the @xmath24cl and @xmath71ga detectors that respond not only to the low energy parts of the @xmath3b spectrum ( @xmath24cl q=0.814 mev ) but beyond , to the lowest energies of the pp @xmath16@xmath17 s ( @xmath71ga q=0.235 mev ) .
interaction rates ( in snu=10@xmath72 /target
nucleus / s ) were calculated using well - known target data and standard @xmath2 spectra from every solar source @xcite modulated by the lwo effect using ( 2 ) .
table 1 lists these results for @xmath24cl and @xmath71ga for the ssm as well as lwo with the same parameters fixed by the es and cc - matches above .
the values for the ssm agree with published values@xcite , checking the overall source data and the calculation methods .
table 1 shows the signal rates in the lines marked `` total snu '' in 4 different _ _ solar astrophysical _ _ scenarios .
these rates _ _ _ _ compare with the measured snu for cl and ga ( last line in each section ) .
the favored lwo groups are highlighted .
the first result is the line marked ssm , for the standard sun .
the lwo results for ga range between @xmath258082 snu that deviate by @xmath73 from the measured value of 70.8(44 ) snu@xcite .
the match of lwo rates with the cl data , 2.56(23 ) snu@xcite , is worse , deviating systematically from it by @xmath74 for every lwo parameter choice . a similar problem exists _ _ also for the lma - cl match _ _ ( see table 2 below ) .
the fact that the mismatch in cl is much worse than in ga indicates that problem lies in the spectral region between 15 mev .
neutrino spectral engineering can not solve this problem .
astrophysical spectral adjustments , on the other hand , could be resorted to , to a limited extent by changes in the solar fluxes .
the only negotiable fluxes in the framework of the ssm are the @xmath3b flux ( predicted with a large error @xmath2520% ) and the cno fluxes ( @xmath25100%)@xcite . two astrophysically sound remedies touching these solar sources are conceivable .
the lwo scenarios above are based inherently on the assumption @xmath9(b)@xmath75(b : ssm ) .
this idea must be astrophysically based , not stand as an ad hoc construct as tacitly viewed so far . the simplest way to reduce the @xmath3b flux
is to reduce the central temperature t of the sun using astrophysically sound scaling laws : @xmath9(b)@xmath76t@xmath77 and @xmath9(be)@xmath76t@xmath78@xcite . using the matched values of @xmath9(cc - b ) above ( @xmath257180% ) , the reduction in @xmath9(be ) can be calculated .
this correction is applied in the lines marked ssm(t ) in table 1 .
the correction helps to a small extent . the second remedy concerns the cno flux for which there is little astrophysical guidance .
global fits of all neutrino scenarios proposed so far are eminently compatible with 05% cno luminosity in the sun@xcite .
we therefore suggest that the cno flux is seriously reduced , viz .
the lines marked ( no cno ) reflect this assumption .
this correction helps substantially in the lwo - cl match .
finally both astrophysical modifications result in the rates marked ssm [ no cno(t ) ] .
the cl rates now match the lwo predictions within 1.65@xmath28 ( similarly , no - cno helps _ _ also the lma - cl _ _ match to move within 1.5@xmath28 , see table 2 ) the lwo ga match also improves to a much closer @xmath250.3@xmath28 .
we therefore accept and _ _ require _ _ these astrophysical modifications ; hence the designation `` astroparticle '' to our model .
table 1 indicates the reason why @xmath11m@xmath33 values below 75@xmath29ev@xmath33 become rapidly untenable .
the lwo oscillations become progressively faster as e(@xmath16@xmath17 ) decreases ( see fig .
1 ) . at the energy of the @xmath79be line ( 0.862 mev ) , the oscillation period changes rapidly with @xmath11m@xmath36 . below 75 @xmath29ev@xmath33
the be line strobes into a resonance , thus p@xmath34 jumps to @xmath251 .
the be contribution to the cl ( as well as the ga ) rate jumps , causing both to fall out of match . even though the lwo-@xmath3b matches are good for @xmath80m@xmath81ev@xmath13 the be factor limits the model at @xmath11m@xmath33 = 75 @xmath29ev@xmath33 as long as @xmath9(@xmath22b ) @xmath82 0.8 @xmath9(@xmath22b : ssm ) .
the discussion above has examined in detail the compatibility of the predictions of the new lwo - based astroparticle model of solar neutrinos to every experimental result available to date with sensitivities that span the entire solar neutrino spectrum . in table 2
we summarize the match of model to data quantitatively and individually .
the `` astroparticle lwo model '' is competitive with the ( astroparticle ) lma in the consistency with the global data .
the major _ _ _ _ difference is a 1.4@xmath28 mismatch with the sno _ _ absolute nc rate _ _ , a main reason why the lwo scenario was disfavored in global analyses such as @xcite even though it was found competitive in the absence of the nc result .
for the lma - msw , _ _ only _ _ two types of experimental data can provide new support .
the first is a substantial improvement in the precision of the absolute nc signal rate .
we note that the fundamental difference _ _ vis a vis _ _ the lwo model is a mere 20% lower @xmath3b @xmath16@xmath17 flux . the nc rate in sno is now known to @xmath6713% 1@xmath28@xcite .
this precision needs to improve markedly to exclude the possibility of a 20% lower flux by at least 3@xmath28 .
the second possibility is the only predicted effect specific to the lma viz . ,
a gradual rise of the cc signal from 0.365 ssm above 5 mev to 0.6 ssm below 1 mev in the region of the @xmath83be and pp neutrinos@xcite .
this regime , endemic to high background is difficult to access experimentally and the lma vs. lwo test requires sharp differentiation of somewhat comparable signals at 0.6 ssm and 0.45 ssm ( see last lines in table 1 ) respectively . the first opportunity for this test will arise when borexino@xcite , designed for very low background at low energies , starts operation in 2004 .
we now turn to experiments aimed specifically at exposing the lwo basis of the new model .
the basic premise of the present model is spectral shape invariance for e(@xmath16@xmath84 mev under the action of lwo . however , as seen in fig . 1 this action
is fundamentally accompanied by _ _
strong shape changes _ _ _ _ immediately below _ _ _ 5 mev . _ experimental tests accessible to the signal window @xmath19 5 mev , hitherto unexplored via direct solar neutrino detection , hold the key to a compelling demonstration of the present model since the lwo spectral predictions in this window are dramatically different from the relatively weak shape effects in the lma case .
this window is closed to the present large - scale cerenkov detectors sk and sno .
it is however accessible to a liquid scintillator ( ls ) based detector on the scale of at least 1 kton target mass ( similar to sno though much smaller than sk with 22 kton fiducial mass ) .
kamland is such a detector that is already operational@xcite .
borexino is another , that will soon be operational .
both are designed for low background neutrino spectroscopy at low energies , a key factor in considering signals below 5 mev .
we first consider kamland because interesting data can be expected from this device in the very near future .
let us first consider es signals below 5 mev .
6 shows the es spectra expected in a 1 kton ls detector .
these are theoretical curves unconvoluted with detector resolution since the superior energy resolution in a ls device results in little smearing of the spectral shapes .
shows that while the es spectra for lma and lwo match identically above 6 mev , they diverge significantly below this critical energy .
while it is obviously desirable to measure the full spectrum and determine the shape characteristics below 6 mev ( normalized to the `` flat '' part @xmath276 mev ) , a reliable preliminary indication of the effect can be obtained just from the relative rates in two broad windows below and above 6 mev .
table 3 lists the signal rates/2y live time for the two models in the windows 36 and 616 mev . in the lwo scenario ( with practically no differences between the viable solutions ) , the excess of counts at 36 mev is measurably higher than in lma , 484(39 ) vs. 370(38 ) , a @xmath85 effect . for the lower threshold at 2.5 mev background permitting
, the lwo effect can possibly be detected at the @xmath86 level .
kamland has been in operation now for nearly two years but it is not clear if data on signals in these windows are available .
information is therefore eagerly awaited . in kamland the background characteristics in the 25 mev regime , the key factor , is not known .
the relatively shallow depth of this detector could pose a high cosmogenic background interference . in borexino , with a larger overburden , this problem is less acute .
the disadvantage is that in the present design , the largest mass available in borexino is only 300 tons@xcite .
it is envisioned that the target mass can be upgraded to approach 1 kton in the future , when a definitive result on lwo can be attempted .
the es spectra in fig . 6 are interesting and informative especially since they may be observable already in kamland . however , the spectral effects are considerably less prominent than could be expected from fig .
1 because of the basic spectral smearing effect of the es reaction ( although the less serious detector smearing is avoided ) .
one can thus expect stronger effects in cc transitions with bound nuclei in the final state that ensure that the signal electron energy is uniquely connected to the @xmath2 energy ( unlike in the @xmath16@xmath17+d reaction ) .
other considerations in the choice of the reaction are : 1 ) a high rate and/or a large target mass since the @xmath3b @xmath16@xmath17 flux is small and 2 ) a low threshold .
the latter is important since an energy threshold @xmath273 mev is desirable in order to avoid background .
an ideal choice is the case of @xmath79li , long known as a candidate for direct solar neutrino spectroscopy @xcite but not so far applied in practice .
7 shows the nuclear data relevant to the inverse @xmath88(ib ) reaction @xmath16@xmath17+@xmath87li@xmath89be + e. the ib threshold is only 0.862 mev so that dominant part of the @xmath3b @xmath16@xmath17 spectrum can be observed above a @xmath90 mev signal threshold .
the ib transition is a superallowed transition with a precisely calculable rate , one of the highest known in beta decay . as a light nucleus with a high natural abundance ,
a relatively small mass of 30100 tons ( compared to 1 kton ) may suffice for a decisive result .
technical details are beyond the scope of this paper , but a suitable technology for the li - cc experiment appears possible .
fig . 8 ( left ) shows cc spectra observable in the li ib reaction .
almost the entire spectrum occurs above 3.5 mev , comfortably above the background threshold . with the thick red line for ssm*0.365=lma as the reference ,
the spectra directly display the essential action of fig . 1
, shape invariance above 8 mev and dramatic modulation at lower energies .
the prominence of the modulations is illustrated in fig .
8 ( right ) where the lwo spectra are normalized to lma which appears as a flat line .
the spectral deviations are strong and compelling .
table 4 lists the signal rates in two broad windows above and below 7 mev .
the yearly rates are high . in the low window ,
the lwo rates are @xmath2550% higher than the lma , whereas in the high window they are same ( within @xmath253% ) .
it would thus appear that the essential proof of the lwo basis of the present model could be obtained in one year s live time possibly with a target mass as small as 30 tons of li .
the lwo modulations become even stronger at very low energies in regime of the pp neutrinos . even in the relatively small window of a few hundred kev several periods of lwo modulations occur .
these modulations _ _ may _ _ be observable by a suitable cc transition with a sufficiently low threshold that can be implemented by a technology that offers sharp energy resolution .
the modulations at these energies are more sensitive to the neutrino parameters than at the high energy windows considered above
. thus it may be possible to determine them more precisely than the narrow lwo specifications already inherent in the present model .
an ideal choice for this application is the cc transition in @xmath91 in @xcite .
a practical real time pp neutrino detector based on indium is under development in the lens project @xcite .
the application of lens to the spectroscopy of the present neutrino model will be discussed elsewhere .
is the lma - msw model the last word in solar neutrinos ? it may well be .
the present work raises a serious possibility that it need not _ _ necessarily _ _ be .
we have made the _ _ prima facie _ _ case that a model based on lwo , a very different physics scenario , offers a credible alternative . only a 20% difference in a measured absolute signal rate stands between them .
the serious pause against the new model is , clearly , the possible violation of cpt invariance that may hold rigidly also in the neutrino sector in spite of recent suggestions to the contrary .
it is thus a classic case of decision by a unique physical result vs. discovery by consensus with a mass of different data .
experiments with the potential for compelling spectral impact are proposed in this work and may already be observable .
only they can tell which model is the prevalent one .
the impact of this decision goes well beyond the already rich field of solar neutrinos .
this science has led us deep into the sun s interior and beyond the standard model of particle physics .
is it poised to take us to yet more unknown territory opened by the breakdown of one of the most cherished symmetry principles in physics ?
i wish to express my sincere thanks to peter kurcynski who initiated me into the mysteries of ms - excel computing used extensively in this work and for helpul discussions .
i am grateful to k. kubodera and s.nakamura for generously supplying their tabular data in a format ready for use in this work .
i thank john bahcall , michel cribier , eligio lisi , sandip pakvasa and carlos penya - garay for their comments on the draft ms . * * d53**,1665 ( 1996 ; hep - ph/9510367 .
this paper and @xcite , first considered the effect of @xmath9(@xmath22b)@xmath92(@xmath22b : ssm ) which revealed the possibility of very low @xmath93 solutions .
the conclusions were only qualitative in the current context since these works considerably predate precision results from sk and sno .
these papers showed that shape distortions occur at pp - neutrino energies .
they did not point out the crucial medium energy shape effects brought out in the present paper .
# at e(@xmath94 ) @xmath27 5 mev there is practically no difference between true lma and [email protected] + * relative to bpssm @xmath9(@xmath22b ) = 5.05@xmath95cm@xmath13s + $ true lma - msw + * * no cno only | the large mixing ( mass ) ( lma ) -msw model of solar neutrinos ( @xmath0 ) is now widely held to be near definitive , based on global consistency with data . no physical effect , however , compels its _ _ uniqueness__. the present search for an explicitly testable competitive model was stimulated by a surprising finding the high energy part of the standard solar model ( ssm ) @xmath1b @xmath2 spectrum can be scaled very precisely to observed flux levels _ without measurable shape distortion _ via sensitive combinations of long wavelength flavor conversion in vacuum and a @xmath3b @xmath2 flux @xmath4(b : ssm ) .
pursuantly , a new `` astroparticle '' model with the relatively specific parameters @xmath5ev@xmath6 ev@xmath7 ) ; sin@xmath8 0.580.56 ; @xmath9(b ) = 0.830.81@xmath10b : ssm ) coupled with modest changes in the ssm , offers a viable solution consistent with data . because kamland has set @xmath11m@xmath12ev@xmath13 , sin@xmath14 for _ _ anti__@xmath15-e scattering signals in the window 35 mev , even though the spectrum is flat above 5 mev .
this window is accessible to experiment for the first time in kamland .
new experiments are proposed to observe the more dramatic charged - current spectral effects .
r. s. raghavan bell laboratories , lucent technologies , murray hill nj 07974 usa infn
laboratori nazionali del gran sasso , italy |
the presence of multiple time and length scales constitutes one of the major problems in computer simulations of matter . in simulations that faithfully capture the dynamic evolution of a system , the fastest particles in the system
dictate the required time resolution .
if different types of particles with widely varying diffusion rates are present , then the slower particles may be unable to explore the entire configuration space during the course of the simulation , leading to ergodicity problems . in the computational study of complex fluids , such as colloidal suspensions ,
this problem frequently occurs , since such systems typically contain particles of different sizes .
a concrete example is the ` nanoparticle haloing ' phenomenon discovered by lewis and coworkers @xcite .
this experimental work deals with a new approach to the stabilization of suspensions of micron - sized spherical particles , which tend to aggregate under the influence of their mutual van der waals attraction . conventional approaches to prevent this gelation , such as charge stabilization ( variation of the ph to alter the surface charge of the particles ) , lead to complications in certain applications , e.g. , in the formation of colloidal crystals , where the electrostatic repulsion prevents the close packing of the particles .
it was found that these complications can be avoided by generating an effective colloidal repulsion through the addition of highly - charged nanometer - sized particles to the suspension of the ( near - neutral ) colloids . a concrete explanation for the underlying mechanism leading to the effective repulsions is currently lacking .
evidently , a computational approach to this problem must involve both the microspheres and the nanoparticles , which typically differ by a factor 100 in diameter .
traditionally , the effective pair interaction between colloids ( potential of mean force ) is then calculated by `` integrating out '' the smaller species , e.g. , by simulating a system containing two spheres at fixed separation embedded in a sea of smaller particles @xcite . here
, we discuss a new simulation algorithm @xcite that is capable of explicitly incorporating both species and computing equilibrium properties of the suspension , without suffering from the disparity in time scales .
this algorithm can be viewed as a continuum version of the widely - used cluster algorithms for lattice spin models @xcite .
equilibrium monte carlo methods are aimed at obtaining static thermodynamic and structural properties by generating system configurations according to the boltzmann distribution .
nonphysical moves may be employed in the underlying markov process , which in principle offers an exquisite way to overcome the presence of multiple time scales .
thus , monte carlo algorithms are the method of choice for , e.g. , simulations of critical phenomena and phase transitions . in the context of size - asymmetric fluids ,
collective moves have been devised in which groups of particles are moved simultaneously .
unless these groups are identified in a careful way , which typically involves knowledge about the physical properties of the system , the monte carlo step will entail a large energy change and consequently have only a very small acceptance rate .
thus , while this approach has been used quite successfully in specific situations @xcite , it typically involves a tunable parameter and can not be generalized in a straightforward fashion .
the situation is rather different for spin models , for which swendsen and wang ( sw ) @xcite introduced a _ cluster algorithm _ in which groups of parallel spins are identified in a probabilistic manner , based upon the fortuin
kasteleyn mapping of the potts model onto the random - cluster model @xcite .
these groups can subsequently be flipped independently .
this rejection - free algorithm ( every completed cluster can be flipped without an additional evaluation of the resulting energy difference ) suppresses critical slowing down .
accordingly , its generalization to off - lattice fluids has been a widely - pursued goal . however , the sw algorithm as well as the even more efficient single - cluster variant introduced by wolff @xcite relies on invariance of the hamiltonian under a spin - inversion operation . for a fluid
, this translates into particle
hole symmetry , which is only obeyed for lattice gases .
consequently , efficient off - lattice cluster algorithms have only been designed for a small number of specific fluid models @xcite , and it is not clear how these methods can be generalized .
hard - sphere fluids constitute another special case . here every configuration without particle overlaps has the same energy .
dress and krauth @xcite devised a strategy to create such configurations by means of geometric transformations .
their approach is particularly advantageous in size - asymmetric mixtures , but can not be applied to systems with other interactions without supplementing it with a costly acceptance criterion .
nevertheless , a remarkable feature of this geometric cluster algorithm is that it relies on the invariance of the hamiltonian under geometric operations . here
, we exploit this property to formulate a cluster monte carlo algorithm that is applicable to arbitrary pair potentials without the imposition of an acceptance criterion .
the geometric cluster algorithm as formulated by dress and krauth starts from a configuration of particles , with periodic boundary conditions .
this configuration is rotated around an arbitrarily chosen pivot .
groups of particles that overlap between the original and the rotated configuration are exchanged between these configurations . in practice , it is more convenient to construct only a single cluster and to carry out a point reflection on the fly @xcite . for each cluster
a new pivot is chosen .
this constitutes the counterpart of the wolff algorithm for spin models @xcite . in the presence of a general , isotropic pair potential
@xmath0 the cluster construction proceeds as follows ( cf .
[ fig : geomc ] ) : 1 .
choose a random pivot point .
2 . choose the first particle @xmath1 at random and move it from its original position @xmath2 to its new position @xmath3 , via a point reflection with respect to the pivot .
3 . identify all particles that interact with @xmath1 in its original position or in its new position .
these particles @xmath4 are considered for a point reflection with respect to the pivot .
this reflection is carried out with a probability @xmath5 $ ] , where @xmath6 and @xmath7 .
note that @xmath8 solely depends on the interaction strength between @xmath1 and @xmath4 .
4 . repeat step 3 in an iterative fashion for each particle @xmath4 that is added to the cluster .
if @xmath4 is moved with respect to the pivot , then all its interacting neighbors that have not yet been added to the cluster are considered for inclusion as well .
the cluster construction is completed once all interacting neighbors have been considered .
it is instructive to compare this prescription to the wolff cluster algorithm , in which a cluster of parallel spins is grown from a randomly chosen initial spin .
parallel spins with a ferromagnetic coupling constant @xmath9 are added to the cluster with a probability @xmath10 .
the energy difference between the parallel and the antiparallel configuration indeed equals @xmath11 .
an antiparallel spin is never added to the cluster , in accordance with the fact that such a pair will be in a state of lower energy if only the first spin is flipped .
this algorithm is ergodic , since each particle can be moved over an arbitrarily small distance .
namely , there is a non - vanishing probability that a cluster consists of only a single particle and the pivot can be located arbitrarily close to the center of this particle .
detailed balance is proven by considering a configuration @xmath12 which is transformed into a new configuration @xmath13 by means of a cluster move .
the energy change @xmath14 results from every particle that is _ not _ included in the cluster but interacts with one or more particles that are part of the cluster .
each `` broken bond '' has a probability @xmath15 , so that the total probability of forming a cluster is given by @xmath16 the prefactor @xmath17 accounts for the probability of choosing a specific pivot and for the probability of creating a specific arrangement of bonds inside the cluster .
the total set @xmath18 of broken bonds can be divided into two subsets .
the broken bonds @xmath19 that lead to an increase @xmath20 in pair energy have a probability @xmath21 , whereas the broken bonds @xmath22 that lead to a decrease in pair energy have a probability equal to unity .
accordingly , the transition probability can be written as @xmath23 \;.\ ] ] the reverse move , in which the configuration @xmath13 is transformed into the configuration @xmath12 by moving a cluster that is constructed in the same way , has a probability @xmath24 where the sum runs over the same set @xmath18 of broken bonds as in eq .
( [ eq : xy ] ) , but the sign of all energy differences @xmath25 has been reversed ( indicated by @xmath26 ) . accordingly , the sum over @xmath20 in eq .
( [ eq : xy - exp ] ) is replaced by the negative sum over the complementary set @xmath27 , @xmath28 \;.\ ] ] the point reflection is a self - inverse operation , so that detailed balance is obeyed without the need to impose an acceptance criterion : @xmath29 = \frac{\exp(-\beta e_{y})}{\exp(-\beta e_{x } ) } \;.\ ] ] since energy differences are taken into account on a pair - wise basis during the construction of the cluster , rather than as a total energy difference after the cluster has been completed , the cluster is constructed in such a way that large energy differences are avoided .
the clusters are representative of the actual structure of the system . in order to demonstrate that the generalized geometric cluster algorithm ( gca ) indeed constitutes the off - lattice counterpart of the sw and wolff cluster algorithms , it is instructive to formulate a multiple - cluster variant .
this formulation yields a full decomposition of an off - lattice fluid configuration into _ stochastically independent clusters_. in the following , we demonstrate how it can be phrased as a natural extension of the single - cluster version .
first , a cluster is constructed according to the wolff version of the gca , with the exception that the cluster is only _ identified _ ; particles belonging to the cluster are marked but not actually moved .
the chosen pivot will also be used for the construction of all subsequent clusters in this decomposition .
these subsequent clusters are built just like the first cluster , except that particles that are already part of an earlier cluster will never be considered for a new cluster .
once each particle is part of a cluster the decomposition is completed and each cluster is moved with a probability @xmath30 .
although all clusters except the first are built in a _ restricted _ fashion , every individual cluster is constructed according to the rules of the wolff formulation of the gca .
the exclusion of particles that are already part of another cluster simply reflects the fact that every bond should be considered only once . if a bond is broken during the construction of an earlier cluster it must not be re - established during the construction of a subsequent cluster . in order to establish that this prescription is a true equivalent of the sw algorithm
, we prove that each cluster can be moved ( reflected ) independently while preserving detailed balance .
if only a single cluster is actually moved , this essentially corresponds to the wolff version of the gca , since each cluster is built according to the gca prescription .
the same holds true if several clusters are moved and no interactions are present between particles that belong to different clusters ( the hard - sphere algorithm is a particular realization of this situation ) . if two or more clusters are moved and broken bonds exist between these clusters , i.e. , a non - vanishing interaction exists between particles that belong to disparate ( moving ) clusters , then the shared broken bonds are actually preserved and the proof of detailed balance provided in the previous section no longer applies in its original form .
however , since these bonds are identical in the forward and reverse move , the corresponding factors cancel out .
this is illustrated for the situation of two clusters whose construction involves , respectively , two sets of broken bonds @xmath31 and @xmath32 .
each set comprises bonds @xmath19 that lead to an _ increase _ in pair energy and bonds @xmath22 that lead to a _ decrease _ in pair energy .
we further subdivide these sets into _ external _ bonds that connect cluster 1 or 2 with the remainder of the system and _ joint _ bonds that connect cluster 1 and 2 .
accordingly , the probability of creating cluster 1 is given by @xmath33 upon construction of the first cluster , the creation of the second cluster has a probability @xmath34 since all joint bonds in @xmath35 already have been broken . the factors @xmath36 and @xmath37 account for the probability of realizing a particular arrangement of internal bonds in clusters 1 and 2 , respectively .
hence , the total transition probability for moving both clusters ( upon fixing the pivot ) is given by @xmath38 \;.\ ] ] in the reverse move , the energy differences for all external broken bonds have changed sign , but the energy differences for the joint bonds connecting cluster 1 and 2 are the same as in the forward move .
thus , cluster 1 is created with probability @xmath39 \prod_{j \in \left\{l_1^{\rm joint}\right\ } } \exp[-\beta { { \mbox{{{\usefont{u}{psy}{m}{n}d}}}}}_j ] \;,\end{aligned}\ ] ] where the @xmath40 reflects the sign change compared to the forward move and the product over the external bonds involves the complement of the set @xmath41 .
the creation probability for the second cluster is @xmath42\ ] ] and the total transition probability for the reverse move is @xmath43 \;.\ ] ] accordingly , detailed balance is still fulfilled , @xmath44 = \exp\left [ - \beta ( e_y - e_x ) \right ] \;,\ ] ] in which @xmath45 and @xmath46 and @xmath47 and @xmath48 refer to the total internal energy before and after the move , respectively .
this treatment applies to any simultaneous move of clusters , so that each cluster in the decomposition indeed can be moved independently without violating detailed balance .
it is noteworthy that the probabilities for breaking joint bonds in the forward and reverse moves cancel only because the probability in the cluster construction factorizes into pairwise probabilities , as opposed to the probability for a multiple - particle move in a metropolis - type algorithm .
small ( diameter @xmath49 ) and @xmath50 large particles ( diameter @xmath51 ) at a total packing fraction @xmath52 . @xmath53 and @xmath54 represent the large large and large small correlation functions , respectively . ] as demonstrated in fig .
[ fig : wolff - sw ] for a binary mixture , the results obtained by means of the multiple - cluster geometric algorithm agree perfectly with those obtained using the single - cluster version .
the most striking feature of the generalized geometric cluster algorithm , apart from the fact that it creates clusters that can be moved in a rejection - free manner , is the speed at which it relaxes size - asymmetric mixtures .
we illustrate this here for a binary fluid mixture consisting of @xmath55 small and @xmath56 large spherical particles with size ratio @xmath57 . the particles are contained in a fixed volume , at equal packing fractions @xmath58 .
while @xmath59 is kept fixed , @xmath55 increases from @xmath60 to @xmath61 as @xmath62 is varied from @xmath63 to @xmath64 .
pairs of small particles and pairs involving a large and a small particle act like hard spheres .
however , in order to prevent depletion - driven aggregation of the large particles , they have a yukawa repulsion , @xmath65/(r/ \sigma_{\rm l } ) & \quad r >
\sigma_{\rm l } \ ; , \end{array } \right.\ ] ] where @xmath66 and the screening length @xmath67 . in the simulation
, the exponential tail is cut off at @xmath68 . as a measure of efficiency we consider the integrated autocorrelation time @xmath69 obtained from the energy autocorrelation function @xcite , @xmath70 for conventional mc calculations
, @xmath69 rapidly increases with increasing @xmath62 , because the large particles tend to get trapped by the small particles .
accordingly , an accurate estimate for @xmath69 could only be obtained for @xmath71 . by contrast , the generalized gca has an autocorrelation time that only weakly depends on the size ratio , as illustrated in fig .
[ fig : efficiency ] . at @xmath72
the resulting efficiency gain already amounts to more than three orders of magnitude . .
whereas the autocorrelation time per particle ( expressed in @xmath73s of cpu time per particle move ) rapidly increases with size ratio , the gca features only a weak dependence on @xmath62 . ] , for a monodisperse lennard - jones fluid .
( a ) reduced density @xmath74 ; ( b ) reduced density @xmath75 .
identical symbols compare to identical temperatures in both panels .
all temperatures are indicated in terms of @xmath76 .
see text for discussion.,title="fig : " ] + ( a ) , for a monodisperse lennard - jones fluid .
( a ) reduced density @xmath74 ; ( b ) reduced density @xmath75 .
identical symbols compare to identical temperatures in both panels .
all temperatures are indicated in terms of @xmath76 .
see text for discussion.,title="fig : " ] + ( b ) a crucial limitation of the generalized gca is that each cluster must only occupy a fraction of the entire system .
as observed by dress and krauth @xcite , the entire system typically will be occupied by a single cluster once the percolation threshold is reached in the combined system containing a configuration and its point - reflected counterpart . for the original system this leads , in three dimensions , to a practical upper limit in the packing fraction around @xmath77@xmath78 .
this number will vary as a function of size asymmetry and , if additional pair interactions are present , temperature .
it is therefore instructive to study the cluster - size distribution as a function of reduced density @xmath79 and temperature @xmath80 .
figure [ fig : density ] illustrates the cluster - size distributions as obtained by means of the multiple - cluster gca for a regular ( one - component ) lennard - jones fluid . in the top panel , @xmath81 .
already at temperatures that are far above the critical temperature @xmath82 , the cluster - size distribution starts to tend toward a bimodal form , indicative of the formation of large clusters . in the vicinity of the critical temperature
, the average cluster size has become very large . for comparison
, the bottom panel displays the cluster - size distributions at a twice smaller density , @xmath75 . in this case the bimodal shape does not appear until close to @xmath83 .
the properties suggest that the generalized gca , at least for the one - component lennard - jones fluid , will not suppress critical slowing the primary advantage of lattice cluster algorithms . for the off - lattice gca ,
this property is less essential , because of the speed - up it delivers for the simulation of size - asymmetric fluids over a wide range of temperatures and packing fractions .
nevertheless , we have investigated the integrated autocorrelation time for the energy at the critical point , as a function of linear system size . in fig .
[ fig : critdyn ] these times are collected for three algorithms .
( 1 ) conventional local - update metropolis algorithm ; ( 2 ) wolff version of the gca ; ( 3a ) sw version of the gca , in which each cluster is reflected with a probability @xmath84 ; ( 3b ) sw version of the gca , in which each cluster is reflected with a probability @xmath85 . just as for spin models , the single - cluster version outperforms the swendsen
wang type cluster decompositions .
however , all variants of the gca exhibit the same power - law behavior , which outperforms the metropolis algorithm by a factor @xmath86 .
it is important to emphasize that this acceleration may be due to the suppression of the hydrodynamic slowing down @xcite as a function of linear system size for a critical lennard - jones fluid , in units of particle sweeps , for three different monte carlo algorithms : local moves ( `` metropolis mc '' ) ; gca with swendsen
wang type cluster decomposition and probability 0.50 ( `` sw 50% '' ) and 0.75 ( `` sw 75% '' ) of moving each cluster ; single - cluster gca ( `` wolff '' ) . ] caused by the conservation of the density ( which may couple to the energy correlations @xcite ) .
another striking point is that already for moderate system sizes the generalized gca outperforms the metropolis algorithm , despite the time - consuming construction of large clusters [ cf .
[ fig : density](a ) ] which only lead to small configurational changes . ) in an environment of smaller particles ( diameter @xmath87 , packing fraction @xmath88 ) that experience a yukawa - type attractive interaction with the colloids .
the symbols represent data obtained by means of the hard - core gca @xcite ; the solid line was obtained from the generalized gca . ]
in order to illustrate the capabilities of the generalized gca , we have computed the pair correlation function of dilute colloidal particles ( packing fraction @xmath89 ) in an environment of particles with a five times smaller diameter ( packing fraction @xmath90 ) .
large and small particles act as hard spheres , but unlike pairs ( i.e. , large
small ) experience a yukawa attraction which promotes the accumulation of small particles around the colloids .
this system has been studied in ref .
@xcite by means of the original hard - core gca , in which clusters are moved according to an acceptance criterion , as proposed in ref .
this potentially greatly deteriorates performance , as entire clusters will be constructed that are subsequently rejected .
indeed , the authors report @xcite that the colloid pair correlation function @xmath91 had to be obtained via numerical differentiation of the integrated pair correlation function , rather than through direct sampling .
the generalized gca can handle this system without complication , as demonstrated in fig .
[ fig : amokrane ] . in order to obtain data comparable to the solid curve in this figure , the authors of ref .
@xcite utilized a polynomial fit to the integrated pair correlation function , which potentially leads to ambiguities .
we have discussed the generalized geometric cluster algorithm , a monte carlo method for the simulation of fluids by means of geometric operations .
this algorithm creates nonlocal multiple - particle moves that are capable of rapidly decorrelating fluid configurations that contain particles of widely different sizes .
the multiple - particle moves are constructed in such a way that the typical decrease in acceptance rate is avoided ; every proposed move is accepted without violating detailed balance .
it is anticipated that this algorithm will find widespread application , in particular in the simulation of complex fluids and suspensions in which the solvent is modeled as an implicit background .
potential generalizations include the treatment of particles with internal degrees of freedom , such as nonspherical particles , and other geometries , such as layered systems .
this material is based upon work supported by the national science foundation under grants nos .
dmr-0346914 and cts-0120978 . | we discuss a new monte carlo algorithm for the simulation of complex fluids .
this algorithm employs geometric operations to identify clusters of particles that can be moved in a rejection - free way .
it is demonstrated that this _ geometric cluster algorithm _ ( gca ) constitutes the continuum generalization of the swendsen
wang and wolff cluster algorithms for spin systems .
because of its nonlocal nature , it is particularly well suited for the simulation of fluid systems containing particles of widely varying sizes .
the efficiency improvement with respect to conventional simulation algorithms is a rapidly growing function of the size asymmetry between the constituents of the system .
we study the cluster - size distribution for a lennard - jones fluid as a function of density and temperature and provide a comparison between the generalized gca and the hard - core gca for a size - asymmetric mixture with yukawa - type couplings . |
signals having the power spectral density ( psd ) at low frequencies @xmath1 of the form @xmath2 with @xmath3 close to @xmath4 are commonly referred to as `` @xmath0 noise '' , `` @xmath0 fluctuations '' , or `` flicker noise .
'' power - law distributions of spectra of signals with @xmath5 , as well as scaling behavior are ubiquitous in physics and in many other fields @xcite . despite the numerous models and theories proposed since its discovery more than 80 years ago @xcite , the subject of @xmath0 noise remains still open for new discoveries .
most models and theories of @xmath0 noise are not universal because of the assumptions specific to the problem under consideration . a short categorization of the theories and models of @xmath0 noise is presented in the introduction of the paper @xcite .
see also recent experimental investigations and physical models of @xmath0 noise in condensed matter @xcite and a review by balandin @xcite .
often @xmath0 noise is modeled as the superposition of lorentzian spectra with a wide range distribution of relaxation times @xcite .
an influential class of the models of @xmath0 noise involves self - organized criticality ( soc ) . in 1987
et al _ @xcite introduced the notion of soc with the motivation to explain the universality of @xmath0 noise .
although paper @xcite is the most cited paper in the field of @xmath0 noise problems , it was shown later on @xcite that the mechanism proposed in @xcite results in @xmath6 fluctuations with @xmath7 and does not explain the omnipresence of @xmath0 noise .
the @xmath0 noise in the fluctuations of mass was first seen in a sandpile model with threshold dissipation by ali @xcite .
et al _ @xcite studied a one - dimensional directed model of sandpiles and showed that the exponent @xmath3 is exactly @xmath4 for noise in the spectrum of fluctuations of mass .
an example of @xmath0 noise in the classical sandpile model has been provided in @xcite .
another models of @xmath0 noise involve a class of maps generating intermittent signals .
it is possible to generate power - laws and @xmath0-noise from simple iterative maps by fine - tuning the parameters of the system at the edge of chaos @xcite where the sensitivity to initial conditions of the logistic map is a lot milder than in the chaotic regime @xcite .
manneville @xcite showed that iterative function can produce interesting behavior , power - laws and @xmath0 psd . in paper
@xcite a mechanism of intermittency exhibiting @xmath0 noise which occurs in nonlinear dynamical systems with invariant subspace and having the transverse lyapunov exponent equal to zero was considered .
intermittency as a mechanism of @xmath0 noise continues to attract attention @xcite . in many cases the physical processes
can be represented by a sequence of random pulses .
the mathematical way of generating power - law noise from a sequence of pulse has been discussed by halford @xcite .
the spectrum of the signal consisting of the pulse sequences which belong to the class of markov process , was investigated in @xcite . in those models
the power - law psd appears due to power - law distribution of pulse durations .
main objection to this approach is an apparent lack of physical mechanism generating durations over several orders of magnitude @xcite . on the other hand
, we can point out that the diffusion process gives rise to broad distributions of lifetimes .
it is known that for the unbiased random walk , the distribution of the first return times has a power - law form with the exponent @xmath8 @xcite .
another possible mechanism leading to broad distribution of lifetimes is owing to the formation of avalanches . in many situations ( for example in barkhausen noise @xcite ,
fluid invasion into disordered media @xcite , dislocation avalanches in microcrystals @xcite ) the volume of avalanches has power - law distribution and , in addition , the volume scales as a power - law function of the duration of the avalanche .
similar properties have also avalanches in the models of soc @xcite . one more way of obtaining @xmath0 noise from a signal consisting of pulses has been presented in @xcite .
it has been shown that the intrinsic origin of @xmath0 noise may be a brownian motion of the interevent time of the signal pulses , similar to the brownian fluctuations of the signal amplitude , resulting in @xmath9 noise .
recently , the nonlinear stochastic differential equations ( sdes ) generating signals with @xmath0 noise were obtained in @xcite ( see also papers @xcite ) , starting from the point process model of @xmath0 noise .
analysis of the long - range correlated bursting signals is presented in @xcite , as well .
the purpose of this paper is to show the connection between the nonlinear sdes generating signals with @xmath0 noise and signals consisting of random pulses with the power - law distribution of pulse durations .
as we will demonstrate , in both of these models @xmath0 spectrum appears due to the scaling properties of the signal .
in addition , the signal generated by sde can be approximated by rectangular pulses yielding the same distribution of signal intensity and the same power - law exponent in the psd .
although the models generating @xmath0 noise that we consider in this paper ( nonlinear sdes and random pulses with the power - law distribution of pulse durations ) are not unique , the method of their derivation from the scaling properties of the signal was not investigated before .
we obtain nonlinear sdes generating signals with @xmath0 noise starting _ not _ from the point process model , as has been done in @xcite , but _ from the scaling properties of the signal required to get @xmath0 noise_. this approach allows us to reveal new connections between those seemingly different models .
the paper is organized as follows : in we consider nonlinear sdes generating signals with @xmath6 psd and show that such sdes can be obtained by requiring a proper scaling . in
we analyze signals consisting of random pulses with the power - law distribution of pulse durations and power - law dependence of pulse height on the pulse duration .
we show that such pulses have the same scaling properties as the signal generated by sdes in . in
we produce the connection between the nonlinear sdes modeling and the rectangular pulses series more explicitly .
summarizes our findings .
nonlinear sdes generating signals with @xmath6 psd are derived in papers @xcite . in this section
we show these sdes can be obtained only from the scaling properties required by @xmath6 psd .
this new technique reveals more directly the origin of @xmath6 psd compared to the derivation starting from the point process model , as has been done in @xcite .
pure @xmath6 psd is physically impossible because the total power would be infinite .
therefore , we will consider signals with psd having @xmath6 behavior only in some wide intermediate region of frequencies , @xmath10 , whereas for small frequencies @xmath11 psd is bounded .
we can obtain nonlinear sde generating signals exhibiting @xmath0 noise using the following considerations .
wiener - khintchine theorem relates psd @xmath12 to the autocorrelation function @xmath13 : @xmath14 if @xmath15 in a wide region of frequencies , then for the frequencies in this region the psd has a scaling property @xmath16 when the influence of the limiting frequencies @xmath17 an @xmath18 is neglected . from the wiener - khintchine theorem
it follows that the autocorrelation function has the scaling property @xmath19 in the time range @xmath20 .
the autocorrelation function can be written as @xcite @xmath21 where @xmath22 is the steady state probability density function ( pdf ) and @xmath23 is the transition probability ( the conditional probability that at time @xmath24 the signal has value @xmath25 with the condition that at time @xmath26 the signal had the value @xmath27 ) . the transition probability can be obtained from the solution of the fokker - planck equation with the initial condition @xmath28 .
one of the ways to obtain the required property is for the steady state pdf to have the power - law form @xmath29 and for the transition probability to have the scaling property @xmath30 where @xmath31 is the scaling exponent , the meaning of which will be revealed below , equation . indeed , from equations and a change of variables it follows @xmath32 thus , the autocorrelation function has the required property with @xmath3 given by equation @xmath33 note , that according to equation the change of the magnitude of the stochastic variable @xmath34 is equivalent to the change of time scale @xmath35 . in order to avoid the divergence of steady state pdf the diffusion of stochastic variable @xmath27 should be restricted at least from the side of small values and , therefore , holds only in some region of the variable @xmath27 , @xmath36 . when the diffusion of stochastic variable @xmath27 is restricted , equation also can not be exact
however , if the influence of the limiting values @xmath37 and @xmath38 can be neglected for time @xmath24 in some region @xmath39 , we can expect that the scaling approximately holds in this time region . to get the required scaling of the transition probability
, the sde should contain only powers of the stochastic variable @xmath27 .
this will be the case if the coefficient in the noise term is the power - law depending , i.e. , proportional to @xmath40 .
the drift term then is fixed by the requirement for the steady state pdf .
thus we consider sde @xcite @xmath41 here @xmath42 is a standard wiener process ( the brownian motion ) and @xmath43 is the white noise intensity .
note that sde is the same as in papers @xcite , only here we obtained it from the consideration of the scaling properties , not starting from the point process model .
changing the variable @xmath27 in to the scaled variable @xmath44 or introducing the scaled time @xmath45 and using the property of the wiener process @xmath46 one gets the same resulting equation .
thus , change of the scale of the variable @xmath27 and change of time scale are equivalent , as in equation , and the exponent @xmath31 is @xmath47 from equation it follows that the power - law exponent in the psd of the signal generated by sde is @xmath48 in order to obtain a stationary process and avoid the divergence of steady state pdf the diffusion of stochastic variable @xmath27 should be restricted or equation should be modified .
the simplest choice of the restriction is the reflective boundary conditions at @xmath49 and @xmath50 .
another choice would be modification of equation to get rapidly decreasing steady state pdf when the stochastic variable @xmath27 acquires values outside of the interval @xmath51 $ ] .
for example , the steady state pdf @xmath52 with @xmath53 has a power - law form when @xmath36 and exponential cut - offs when @xmath27 is outside of the interval @xmath51 $ ] .
such exponentially restricted diffusion is generated by the sde @xmath54x^{2\eta-1}\rmd t + \sigma x^{\eta}\rmd w_t\label{eq : sde - restricted}\ ] ] obtained from equation by introducing additional terms in the drift .
the presence of the restrictions at @xmath49 and @xmath50 makes the scaling not exact and this limits the power - law part of the psd to a finite range of frequencies @xmath10 .
let us estimate the limiting frequencies .
taking into account the limiting values @xmath37 and @xmath38 , equation for the transition probability corresponding to sde becomes @xmath55 here @xmath37 , @xmath38 are the parameters of the transition probability .
the steady state distribution @xmath56 has the scaling property @xmath57 inserting equations and into equation we obtain @xmath58 this equation means that time @xmath24 in the autocorrelation function should enter only in combinations with the limiting values , @xmath59 and @xmath60 .
we can expect that the influence of the limiting values can be neglected and the scaling holds when the first combination is small and the second large , that is when time @xmath24 is in the interval @xmath61 .
then , using equation the frequency range where the psd has @xmath6 behavior can be estimated as @xmath62 however , numerical solutions of proposed nonlinear sdes show that this estimation is too broad , i.e. , the numerically obtained frequency region with the power - law behavior of psd is narrower than according to equation .
note , that for @xmath63 , i.e. , @xmath64 the width of the frequency region is zero , and we do not have @xmath6 power spectral density . and @xmath38 .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) the psd of such a signal .
the dashed ( green ) line shows the slope @xmath66 .
used parameters are @xmath67 , @xmath68 , @xmath69 , @xmath70 , and @xmath71.,title="fig:",scaledwidth=33.0% ] and @xmath38 .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) the psd of such a signal .
the dashed ( green ) line shows the slope @xmath66 .
used parameters are @xmath67 , @xmath68 , @xmath69 , @xmath70 , and @xmath71.,title="fig:",scaledwidth=33.0% ] and @xmath38 .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) the psd of such a signal .
the dashed ( green ) line shows the slope @xmath66 .
used parameters are @xmath67 , @xmath68 , @xmath69 , @xmath70 , and @xmath71.,title="fig:",scaledwidth=33.0% ] comparison of the numerically obtained steady state pdf and the psd with analytical expressions for sde with @xmath72 and @xmath68 is presented in figure [ fig : sde ] .
for the numerical solution we use the euler - maruyama approximation , transforming the differential equations to difference equations .
we can use constant time step , however at large values of @xmath27 the coefficients in the equations become large and thus require a very small time step .
more effective method of solution is to use a variable time step , decreasing with the increase of @xmath27 . as in @xcite
we choose the time step in such a way that the coefficient before noise becomes proportional to the first power of @xmath27 .
very similar numerical results one gets also by using the milstein approximation @xcite .
we see good agreement of the numerical results with the analytical expressions .
a numerical solution of the equations confirms the presence of the frequency region for which the power spectral density has @xmath6 dependence .
the @xmath0 interval in the psd in figure [ fig : sde ] is approximately between @xmath73 and @xmath74 and is much narrower than the width of the region @xmath75 predicted by equation .
the width of this region can be increased by increasing the ratio between the minimum and the maximum values of the stochastic variable @xmath27 .
as we see in figure [ fig : sde]a , the numerical calculations exhibit a structure of the signal consisting of peaks or bursts .
analysis @xcite reveals that the sizes of the bursts are approximately proportional to the squared durations of the bursts with the power - law distributions of the bursts durations and interburst time .
the exponent of the pdf of the interburst time approximately equal to @xmath8 has been obtained numerically @xcite and analytically @xcite .
in this section we consider pulse sequences with independent pulses .
the shapes of the pulses are characterized by an arbitrary large set of parameters @xmath76 , whereas the occurrence times of the pulses are described by a set of time moments @xmath77 .
the general form of the signal can be written as @xmath78 where functions @xmath79 determine the shape of individual pulses .
the pulse duration @xmath80 is included in the set of parameters @xmath76 or , more generally , is a function of the parameters , @xmath81 .
inter - pulse duration is @xmath82 .
such a pulse sequence is schematically shown in figure [ fig : pulses ] .
we assume that : ( i ) the pulse sequences are stationary and ergodic ; ( ii ) parameters @xmath76 of different pulses are independent ; ( iii ) all pulses are described by the same function @xmath79 ; ( iv ) the pulse parameters @xmath76 have the distribution @xmath83 .
the easies way to calculate psd of a pulse sequence is to start from the definition of the power spectral density directly @xcite .
however , in order to demonstrate connection with the previous section we will consider the expression for the autocorrelation function of the signal .
the autocorrelation function is given by the equation @xmath84 where @xmath85 the observation time interval and the brackets @xmath86 denote averaging over realizations of the pulse sequence . using the signal we can write @xmath87 where @xmath88 is the number of pulses during the observation time interval @xmath85 .
the autocorrelation function can be decomposed into two parts , the first part containing the autocorrelation of each pulse with itself and the second part containing all cross terms : @xmath89 in many cases @xmath6 psd is caused only by the first part .
for example , it is known that when the pulses occur randomly as a poisson process , the psd of the signal depends only on the shapes of the pulses , as given by carson s theorem @xcite .
conditions when a sequence of randomly occurring pulses leads to @xmath6 noise were investigated in @xcite .
note , that when inter - pulse duration @xmath82 is uncorrelated with the duration of the pulse , different pulses may be overlapping . even in case when the other terms in equation are nonzero ,
the first part can have different dependence on time @xmath24 and dominate for some time range @xmath90 .
therefore , initially we will consider only the first part of equation .
it can be written as @xmath91 where @xmath92 is the mean number of pulses per unit time .
since the pulse duration @xmath80 is a function of the parameters @xmath76 , the pdf of pulse durations is @xmath93 introducing the autocorrelation function of the pulses with the same duration @xmath94 , @xmath95 we can write equation in the form @xmath96 if the pdf of pulse durations has a power - law form @xmath97 and the autocorrelation function of the pulses with the same duration has the scaling property @xmath98 then it follows that the autocorrelation function @xmath13 has the required property with @xmath3 given by equation @xmath99 the meaning of the parameter @xmath100 will be revealed below , equations and .
note , that the scaling in equation is the same as described by equation ( 9 ) in @xcite , where the pulse area @xmath101 is used instead of pulse duration @xmath80 . in order to avoid the divergence of the pdf of pulse durations
, equation should hold only in some region of the pulse durations @xmath80 , @xmath102 . in this case
the scaling can not be exact .
however , if the influence of the limiting values @xmath103 and @xmath104 can be neglected for time @xmath24 in some region @xmath39 , we can expect that the scaling approximately holds for this time region .
one of the ways to get the required scaling of the autocorrelation function @xmath105 is to consider pulses having the same shape , only stretched in height and in time .
the signal consisting of such pulses was investigated in @xcite . for stretched pulses we can write @xmath106 where @xmath107 is the height of the pulse .
the function @xmath108 is nonzero only when @xmath109 . from equation
we obtain @xmath110 where @xmath111 is the mean squared amplitude of pulses having the same duration @xmath94 and the function @xmath112 is the autocorrelation function of pulse shapes .
if the height @xmath113 of the pulse is a power - law function of the pulse duration @xmath80 , @xmath114 then the scaling of the autocorrelation function holds .
it should be noted , that even when pulse height @xmath113 is proportional to @xmath115 , the coefficient of proportionality is not necessarily constant .
in particular , the sign of the pulses can be random .
only coefficient of proportionality for the average of the square of the pulse height @xmath116 should be constant .
if pulse height is a power - law function of the pulse duration , then the change of the magnitude of the pulse height @xmath117 is caused by the change of pulse duration @xmath118 . comparing this scaling property to equation
we see that the power - law exponent @xmath100 plays a similar role as @xmath119 , i.e. , @xmath120 the sign minus in equation appears because stretching the time , as in equation , is equivalent to the shortening of the pulse duration .
now we will investigate the influence of limiting pulse durations @xmath103 and @xmath104 . from the assumptions made above , equations , , and , we have that @xmath121 when @xmath102 . here
@xmath122 is the coefficient independent from @xmath80 . for time
@xmath123 we can write the autocorrelation function , according to equations and , as @xmath124 according to paper @xcite , physically reasonable pulses are square - integrable and have everywhere finite derivative of the autocovariance function with respect to time . then for large @xmath125 we can approximate @xmath126 and get @xmath127\ , , & $ 0<\beta<1 $ , \\
\nu b\left[c(0)\ln\tau_{\mathrm{max}}+c'(0)-c(0)\ln t\right]\ , , & $ \beta=1 $ , \\ \nu b\left[c(0)\frac{\tau_{\mathrm{max}}^{\beta-1}}{\beta-1}+t^{\beta-1 } \frac{c'(0)}{2-\beta}\right]\ , , & $ 1<\beta<2 $ .\\}\ ] ] thus , for @xmath128 and @xmath123 the term containing time @xmath24 has the scaling property the same as in equation , the limiting values of the pulse durations @xmath103 and @xmath104 do not influence the scaling of the autocorrelation function . on the other hand ,
if @xmath129 then the influence of @xmath104 becomes significant .
@llll signal&@xmath100&@xmath130&meaning of @xmath130constant area pulses&@xmath131&@xmath132&uniform distribution of pulse durationsconstant energy pulses&@xmath133&@xmath131&pulses of constant height&@xmath132&@xmath134&uniform distribution of inverse durationsgeometrically similar pulses&@xmath4&@xmath135 &
as has been pointed out in @xcite , the condition for @xmath0 spectrum , @xmath136 , can be easily satisfied .
the power - law dependence of the pulse height on the pulse duration can occur naturally .
various cases are listed in table [ tab : power - cases ] .
the value @xmath137 corresponds to the pulses of constant height ; @xmath138 corresponds to constant area pulses .
geometrically similar pulses have @xmath139 .
if the energy is proportional to square of the signal , the constant energy pulses correspond to @xmath140 .
since we have @xmath0 spectrum when @xmath141 , this spectrum occurs for constant area pulses ( @xmath138 ) and uniform distribution of pulse durations ( @xmath142 ) in a wide interval . for constant height pulses ( @xmath137 ) we have @xmath0 spectrum when the distribution of inverse durations @xmath143 is uniform , that is when @xmath144 .
signal consisting of overlapping constant height pulses has pdf of poisson distribution . on the other hand , pulses with @xmath145
can lead to power - law tails in the pdf of the signal .
let us consider rectangular pulses with the only random parameter being the pulse duration @xmath80 .
large signal intensities are due to pulses of large height , for which one can neglect the overlap between pulses .
each pulse of the height @xmath146 occurs with the probability @xmath147 and lasts for time @xmath80 .
thus the pdf of the signal intensity @xmath148 is ( see for analogy @xcite ) @xmath149 if the pdf of pulse durations has the power - law @xmath150 form and the height of the pulse depends on the pulse duration as @xmath151 , then from equation we obtain @xmath152 , where @xmath153 for the pure @xmath0 noise @xmath154 and we get the exponent @xmath68 . for the case of @xmath155 spectrum
we have the following relation between the exponent @xmath3 of the spectrum and exponent @xmath156 of the signal pdf : @xmath157 taking into account equation we see that relation is the same as by equation describing the power - law spectrum of the signal generated by the nonlinear sde . note that pdf of the signal intensity has the same power - law exponent @xmath68 also when @xmath0 noise is generated by the nonlinear sde . .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) the psd of such a signal .
the dashed ( green ) line corresponds to the spectrum obtained using wiener - khintchine theorem from the autocorrelation function .
the time intervals between the pulses are distributed according to poisson process with the average @xmath158 .
the parameters used are @xmath159 and @xmath160.,title="fig:",scaledwidth=33.0% ] .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) the psd of such a signal .
the dashed ( green ) line corresponds to the spectrum obtained using wiener - khintchine theorem from the autocorrelation function .
the time intervals between the pulses are distributed according to poisson process with the average @xmath158 .
the parameters used are @xmath159 and @xmath160.,title="fig:",scaledwidth=33.0% ] .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) the psd of such a signal .
the dashed ( green ) line corresponds to the spectrum obtained using wiener - khintchine theorem from the autocorrelation function .
the time intervals between the pulses are distributed according to poisson process with the average @xmath158 .
the parameters used are @xmath159 and @xmath160.,title="fig:",scaledwidth=33.0% ] typical signal for rectangular constant area pulses ( @xmath138 ) is shown in figure [ fig : fixed - area]a , the pdf of the signal is shown in figure [ fig : fixed - area]b and the psd in figure [ fig : fixed - area]c .
we see a good agreement of numerically obtained psd with the analytical estimation . in figure
[ fig : fixed - area]b we can see that the pdf has a power - law tail for large signal intensities . note that due to the overlapping of the pulses the pdf of the signal at smaller intensities is not a power - law and acquires a power - law tail only for larger intensities , where the overlap can be neglected .
this is in contrast with the sde , where the steady state pdf of the signal can have a power - law form for all values of the signal sufficiently far from limiting values @xmath37 and @xmath38 .
as it was shown in secs .
[ sec : sde ] and [ sec : pulses ] , the signals having @xmath6 psd and generated by the nonlinear sdes have similar scaling like the signals consisting from random pulses . in this section
we approximate nonlinear sde by a sequence of pulses and show that this approximation gives the same pdf of signal intensity and a power - law region in the psd with the same exponent
. however , other details of the approximated signal can be different : for example , the signal consisting from pulses does not exhibit the intermittent bursts characteristic for solution of sde .
let us consider sde together with reflective boundaries at @xmath49 and @xmath50 .
using euler - maruyama approximation with time step @xmath161 and replacing the stochastic differential equation with the difference equation we have @xmath162 here @xmath163 is a gaussian random variable with zero mean and unit variance .
variable time step @xcite @xmath164 results in the equation @xmath165 where @xmath166 is a small parameter .
the approximation of sde becomes better with decreasing @xmath167 .
it should be noted that equation has an universal form : @xmath168 enters only in the first power independent on the exponent @xmath169 . we can estimate the steady state pdf of @xmath168 in @xmath170-space by considering equation as an euler - maruyama approximation of the sde in @xmath170-space , @xmath171 taking into account reflective boundaries at @xmath37 and @xmath38 we get from the fokker - planck equation in @xmath170-space corresponding to equation that the steady state pdf has the power - law form , @xmath172 . from this steady state pdf and equation
it follows that pdf of @xmath173 has a power - law form @xmath174 with @xmath175 the same expression for the power - law exponent @xmath130 can be obtained from equations and . from equation
we obtain @xmath176 : @xmath177 where @xmath178 is a gaussian random variable with zero mean and unit variance . here
we used the fact that the sum of @xmath179 gaussian variables @xmath180 is a gaussian variable with the dispersion equal to @xmath179 .
we can conclude that @xmath176 does not differ significantly from @xmath168 as long as @xmath181 .
the maximal value of @xmath179 when @xmath176 is approximately equal to @xmath168 is @xmath182 .
the duration in which the stochastic variable @xmath27 does not change significantly is @xmath183 where @xmath100 is given by .
the duration @xmath80 , being proportional to @xmath173 , has the power - law pdf with the same exponent @xmath130 as the pdf of @xmath173 : @xmath184 here @xmath185 is normalization coefficient and @xmath186 from equation the value of the stochastic variable @xmath168 is connected with the duration @xmath80 by the relation @xmath187 therefore , we can approximate the signal generated by sde by rectangular pulses of random duration @xmath80 having the pdf of durations and pulse height @xmath188 related to the pulse duration @xmath80 by equation .
the pulses are not overlapping and immediately follow each other .
although the durations of adjacent pulses obtained from the signal generated by sde are correlated , for simplicity we will neglect this correlation .
the pdf of the signal @xmath27 constructed as such a pulse sequence has power - law form .
using equations and we get that the power - law exponent in the pdf @xmath189 is equal to @xmath190 , with @xmath156 appeared in sde . when pulses occur not randomly but follow each other , the other terms in equation are nonzero .
however , one can check that for some range of time @xmath24 the first part in equation dominates .
thus the psd of this pulse sequence has a power - law part with the exponent given by equation . using the value of the exponent @xmath100 from equation we get the power - law exponent in the psd .
the frequency range @xmath191 where psd of the signal consisting of pulses has power - law behavior coincides with inequalities .
thus , the proposed approximation of sde by the sequence of pulses correctly reproduces power - law parts of the pdf and the psd of the generated signal .
is given by equation with @xmath192 , @xmath193 , and @xmath194 and pulse heights are proportional to @xmath195 .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) comparison of the psd calculated using the signal generated by sde with the same parameters as in figure [ fig : sde ] ( black line ) and using the signal consisting of pulses ( gray line ) .
the dashed ( green ) line shows the slope @xmath0.,title="fig:",scaledwidth=33.0% ] is given by equation with @xmath192 , @xmath193 , and @xmath194 and pulse heights are proportional to @xmath195 .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) comparison of the psd calculated using the signal generated by sde with the same parameters as in figure [ fig : sde ] ( black line ) and using the signal consisting of pulses ( gray line ) .
the dashed ( green ) line shows the slope @xmath0.,title="fig:",scaledwidth=33.0% ] is given by equation with @xmath192 , @xmath193 , and @xmath194 and pulse heights are proportional to @xmath195 .
( b ) the pdf of the signal intensity .
the dashed ( green ) line shows the power - law with the exponent @xmath65 .
( c ) comparison of the psd calculated using the signal generated by sde with the same parameters as in figure [ fig : sde ] ( black line ) and using the signal consisting of pulses ( gray line ) .
the dashed ( green ) line shows the slope @xmath0.,title="fig:",scaledwidth=33.0% ] to illustrate the approximation of nonlinear sde by a sequence of pulses , in figure [ fig : cmp ] we compare the pdf and the psd of the signal consisting from pulses to the pdf and the psd of the signal shown in figure [ fig : sde ] .
the sde has parameters @xmath67 , @xmath68 , @xmath69 , @xmath70 , therefore , duration of the pulses has power - law pdf with @xmath192 ( according to equation ) and @xmath193 , @xmath194 .
the height of each pulse is proportional to the duration of the pulse to the power of @xmath140 , obtained from equation .
this value of @xmath100 means that each pulse has the same energy .
the signal consisting of such pulses is shown in figure [ fig : cmp]a .
as one can see , this signal looks rather different from the one shown in figure [ fig : sde]a .
large values of the signal in figure [ fig : cmp]a do not come in the intermittent bursts as in figure [ fig : sde]a .
this difference is caused by the assumption that the durations of different pulses are uncorrelated .
the pdf of the signal , shown in [ fig : cmp]b , is the same as in figure [ fig : sde]b .
comparison of the psds is shown in figure [ fig : cmp]c .
there is qualitative agreement between the psd of the signal generated by the nonlinear sde and the psd of the signal consisting of pulses .
the psd of the signal consisting of pulses has a power - law part in a different range of frequencies , from @xmath196 up to @xmath197 .
this difference from the expected range @xmath198 is caused by neglected other terms in equation .
in summary , we have demonstrated the connection between the nonlinear sdes generating signals with @xmath6 noise and signals consisting of random pulses with the power - law distribution of pulse durations .
the exponent @xmath130 of the power - law pdf of pulse durations and the exponent @xmath100 characterizing the dependence of the pulse height on the pulse duration are related to the parameters @xmath169 and @xmath156 of the sde by means of the equations and .
the signal generated by sde and corresponding signal consisting of rectangular pulses yield the same distribution of signal intensity and the same power - law exponent in the psd .
the appearance of @xmath6 spectrum and relationship between parameters can be obtained just by considering the scaling properties of the signals .
the revealed connection between different models of @xmath0 noise provides further insights into the origin and relationship between different models of @xmath0 noise .
10 url # 1#1urlprefix[2][]#2 ward l m and greenwood p e 2007 _ scholarpedia _ * 2 * 1537 | there are several mathematical models yielding @xmath0 noise . for example
, @xmath0 spectrum can be obtained from stochastic sequence of pulses having power - law distribution of pulse durations or from nonlinear stochastic differential equations .
we show that a couple of seemingly different models exhibiting @xmath0 spectrum are due to the similar scaling properties of the signals .
in addition , we demonstrate a connection between signals with the power - law behavior of the power spectral density generated by the nonlinear stochastic differential equations and modeled by a sequence of random different pulses .
an approximation of solutions of the nonlinear stochastic differential equations by the sequence of pulses correctly reproduces the power - law parts of the probability density function and of the power spectral density .
this connection provides further insights into the origin of @xmath0 noise .
_ keywords _ : stochastic processes ( theory ) , current fluctuations , stationary states |
observations of stars of various metallicities [ @xmath2 @xmath3 ( fe / h)/(fe / h)@xmath4 have shown that the abundance ratios be / fe and b / fe do not vary appreciably over a broad range of @xmath2 ( e.g. duncan et al . 1996 ) . for @xmath2 less than about 0.1 , most of the fe is thought ( truran & timmes 1994 ) to be produced in type ii supernovae ( snii ) .
limiting our considerations to this range of metallicities , the constancy of b / fe and be / fe strongly suggest that b and be are also produced by processes related to sniis .
the accepted paradigm for the origin of most of the be is production in spallation reactions of accelerated particles .
b , in addition to such reactions , is also produced in neutrino interactions in sniis ( woosley & weaver 1995 ) .
the observed be abundance as a function of metallicity can be approximated by @xmath5 ( duncan et al .
1996 ) which , with the solar fe / h=3@xmath610@xmath7 , implies that @xmath8 , independent of @xmath2 .
this constant ratio greatly simplifies the treatment of the evolution and thereby allows us to clearly exhibit the strong constraints that the be and b data place on cosmic ray acceleration in the early galaxy .
we can simplify the equations of galactic evolution by considering the total be , b and fe contents of the galactic gas instead of the abundance ratios of these elements relative to h. assuming that the removal rates of these trace elements from the gas due to incorporation into stars , as well as the return rates due to stellar mass loss are small relative to their production rates ( assumptions whose consistency for @xmath90.1 we demonstrate below ) , the be and fe contents ( measured in number of atoms ) can be written as @xmath10 where the @xmath11 s are production rates ( measured in atoms / sec ) . an integral similar to that for
be can also be written for b , but for now we limit our discussion to be because b could contain a significant non - cosmic ray contribution due to neutrinos . using the integral for fe , we can replace the time variable in the integral for be by the iron content fe .
we obtain @xmath12 since be / fe is observed to be constant as a function of @xmath2 , it will also be constant as a function of fe as long as long as @xmath2 is a monotonic function of fe . in this case
the solution of eq .
( 2 ) is @xmath13 = const .
the calculated ( woosley & weaver 1995 ) @xmath14fe yields per snii , averaged over the scalo ( 1986 ) imf , lead to a relatively constant @xmath14fe yield as a function of @xmath2 ( ranging from 0.09 to 0.15 m@xmath15 per snii for @xmath2 varying from 10@xmath16 to 0.1 ) . taking the @xmath14fe yield per snii equal to 0.1 m@xmath15 ( consistent with sn87a observations )
, we obtain @[email protected]@xmath610@xmath19 atoms / snii or 2.7@xmath610@xmath20 m@xmath15/snii , independent of @xmath2 .
to see when the removal of be and fe can be ignored , let @xmath21 be the removal rate of galactic gas ( mostly h ) due to star formation .
the rates @xmath22 of be and fe removal ( measured in g / sec ) are then 9(be / h)@xmath21 and 56(fe / h)@xmath21 , respectively .
assuming that all stars of mass @xmath010m@xmath15 form type ii supernovae , the snii formation rate is @xmath23 where @xmath24 is the scalo imf ( measured in number of stars per unit stellar mass ) . with
be and fe productions per snii of @xmath25 and 0.1 m@xmath15 , respectively , the be and fe production rates @xmath26 ( measured in g / sec ) are @xmath181.35@xmath27 and @xmath185@xmath28 . since @[email protected]@xmath610@xmath30@xmath2 and @xmath31@xmath183@xmath610@xmath7@xmath2 , the production to removal ratio for both be and fe becomes @xmath32 thus , for @[email protected] , @xmath34@xmath03 ( becoming @xmath18300 for @xmath2=10@xmath35 ) , allowing us to ignore the removal .
this important result is simply the consequence of the fact that , while the be and fe productions are independent of the metallicity of the ism , their removal is proportional to the metallicity .
we have shown ( ramaty et al . 1996 ) that the early galactic b / be data and the predicted b - to - be production ratio are consistent with a non - negligible contribution of @xmath36b production by neutrino spallation in sniis ( woosley & weaver 1995 ) . as both the b - to - be cosmic ray production ratio as well as the imf averaged @xmath36b production by neutrinos are nearly independent of metallicity , our results for be apply also for b.
be and b production by cosmic rays depends on a variety of parameters , but the required constant production , @[email protected]@xmath610@xmath19 atoms / snii independent of metallicity , sets strong constraints on these parameters .
we have developed a libeb production code ( ramaty et al .
1996 ) in which a population of accelerated particles of given composition , energy spectrum and total power @xmath38 is injected into an ambient medium of given composition .
we have taken into account all primary and secondary reaction paths leading to libeb from cno .
we assume that the transport of the accelerated particles in the ism is characterized by a single parameter , @xmath39 , the particle escape path length from the galaxy which we take to be energy independent .
the code employs updated nuclear cross sections and works equally well at nonrelativistic and ultrarelativistic particle energies . for the ambient medium composition we use solar abundances scaled with metallicity . for the accelerated particles we use three compositions : 1 .
ssz , identical to the ambient composition ; this would be the composition if the accelerated particles were accelerated directly out of the ism .
snii , ejecta of type ii supernovae for various @xmath2 ( 12 - 40 m@xmath15 progenitors , woosley & weaver 1995 ) , averaged over the scalo imf ; these compositions would be appropriate if the accelerated particles were accelerated directly from the ejecta of such supernovae .
the snii composition is very weakly dependent on metallicity ; p / o@xmath18100 , @xmath40/o@xmath1820 , c / [email protected] .
3 . snii@xmath41 , identical to snii but with p / o=@xmath40/o=0 .
we employ a shock acceleration spectrum modified with a high energy turnover , @xmath42@xmath43(@xmath44/@xmath45)exp(@xmath46 ) , were @xmath11 is the injection rate of the accelerated particles , @xmath47 and @xmath48 are momentum and kinetic energy , both per nucleon , @xmath49 is velocity , and @xmath50 is a parameter .
the deposited power that accompanies the nuclear reactions is @xmath38=@xmath51 .
1 shows @xmath52 as a function of @xmath2 ( panel a ) and @xmath50 ( panel b ) .
the be production per snii is @xmath53 , where @xmath54 is the mechanical energy of an snii available for acceleration and @xmath55 is the fraction of this energy going into the particles that produce the be .
for the snii@xmath41 composition , which contains no accelerated protons and @xmath40 particles , the be is produced only by accelerated cno interacting with ambient h and he ; thus @xmath52 is not dependent on the ambient medium metallicity ( fig .
this is also true for snii since the metallicity of this composition is high enough so that up to @xmath2=0.1 the contributions of accelerated protons and @xmath40 s is negligible .
thus , if @xmath56 is independent of @xmath2 ( a reasonable assumption ) , then both snii ( for @[email protected] ) and snii@xmath41 ( for all @xmath2 ) produce constant be yields per snii , as required by the data . on the other hand ,
for ssz , for which the protons and @xmath40 s make a major contribution , the be yield is increasing with metallicity ( fig .
1a ) , a result which is not compatible with the observed constant be / fe .
this means that the accelerated particles responsible for the be production should be accelerated from freshly nucleosynthesized matter before it mixes with the ambient ism .
that in the low metallicity era of our galaxy accelerated cno , and not protons and @xmath40 s , were responsible for the production of the be and b was pointed out by duncan , lambert & lemke ( 1992 ) and cass , lehoucq , & vangioni - flam ( 1995 ) . if @xmath57=10@xmath58 erg , we require that @[email protected]@xmath610@xmath60 to produce 3.6@xmath610@xmath19 be atoms per snii with an acceleration efficiency @[email protected] . this constraint provides another argument against ssz for which @xmath59@xmath332@xmath610@xmath35 when @[email protected] ( fig .
for snii , @[email protected]@xmath610@xmath60 implies that @xmath50@xmath050 mev / nucl ( fig .
1b ) , i.e. the accelerated particle spectrum must be relatively hard .
the energetic constraint on @xmath50 is less severe for the snii@xmath41 composition . from fig .
1b the minimum energy ( @xmath55@xmath57@xmath1810@xmath1 erg ) required to produce the be is obtained for the snii@xmath41 composition .
as this value is much lower than the energy per supernova imparted to the galactic cosmic rays ( a few times 10@xmath61 ergs ) , the be could be produced with less energy than needed to maintain the cosmic rays .
the particles could be accelerated by the reverse supernova shock in all sniis , a scenario which could lead to the preferential acceleration of metals .
thus as little as 3% of the total energy in accelerated particles could go into accelerating high metallicity snii ejecta to produce the bulk of the be and part of the b ( spallation by neutrinos producing the rest ) .
the remaining 97% would go into accelerating cosmic rays out of the low metallicity ism .
these cosmic rays produce very little be and b for z / z@xmath15 @xmath330.1 .
cass , m. , lehoucq , r. , & vangioni - flam , e. 1995 , nature , 373 , 318 duncan , d. k. , et al .
1996 , , 99 , 179 duncan , d. k. , lambert , d. l. , & lemke , m. 1992 , , 401 , 584 ramaty , r. , et al .
1996 , nuclei in the cosmos ( nucl . phys . ) in press scalo , j. m. 1986 , fund .
11 , 1 truran , j. w. , & timmes , f. x. 1994 , nuclei in the cosmos iii ( aip ) , 501 woosley , s. e. , & weaver , t. a. 1995 , , 101 , 181 | we derive constraints on the cosmic rays responsible for the be and part of the b observed in stars formed in the early galaxy : the cosmic rays can not be accelerated from the ism ; their energy spectrum must be relatively hard ( the bulk of the nuclear reactions should occur at @xmath030 mev / nucl ) ; and only 10@xmath1 erg / snii in high metallicity , accelerated particle kinetic energy could suffice to produce the be and b. the reverse snii shock could accelerate the particles . |
many questions of practical or scientific interest are based on combinatorial optimization problems whose numerical solution requires time resources growing exponentially with the system size , or more precisely , with the number of binary variables needed to encode the problem .
these include examples like planning and scheduling problems in various real - world applications , optimization of chip design , cryptografic systems in computer science , or glassy systems and random structures in physics .
all these problems are characterized by a non - trivial cost function , or energy , which has to be minimized over a large set of discrete degrees of freedom .
the hardest of these optimization problems are collected in a class called _ np - hard _ @xcite .
hardness refers in this context to the exponential growth of the computational resolution time which is observed for all known numerical algorithms . despite an extremely large number of known np - hard problems , and numerous approaches to solve them ,
no algorithms could be constructed by now which are able to solve such a problem in a time growing only polynomially with the system size .
this point supports the widespread conjecture , that no such effective algorithms are constructible . the numerical search for globally optimal solutions is thus restricted to relatively small systems .
once one has to solve larger systems , good polynomial - time algorithms are needed which construct low - cost configurations .
these are not guaranteed to be optimal , but in many cases they can serve as reasonable approximations @xcite . many of these algorithms are based on heuristic considerations , e.g. on expected correlations between local structures of the specific problem instance and its optimal solutions . exploiting these correlations can largely improve the performance of a heuristic procedure .
here we are using the example of vertex covers ( vc ) on random graphs .
it belongs to the basic np - hard problems @xcite and can be considered as a prototype optimization problem over a random structure . in every step of the presented heuristic algorithms , a vertex and possibly some environment is chosen randomly and covered locally optimal .
the local structure of the graph , in particular its vertex - degree distribution ( distribution of co - ordination numbers ) , can be exploited : vertices of high degree are more likely to be covered , those of small degree are more likely to remain uncovered .
we are using random graphs in order to get some information about the _ typical - case _ behaviour of the algorithm .
this is to be contrasted with the worst - case picture used in the traditional theory of computational complexity @xcite .
the study of heuristic algorithms is also interesting from a more theoretical point of view .
many randomized optimization or decision problems show characteristic phase transition , when the parameters of the randomness are tuned , see e.g. the special issues @xcite for an overview .
the analysis of algorithms is frequently used in theoretical computer science to construct bounds for these phase transitions , for some examples see e.g. @xcite and references therein . a different approach to these transitions
is given by applying techniques from equilibrium statistical mechanics , as was done successfully for some of the fundamental combinatorial problems like 3-satisfiability @xcite , number partitioning @xcite , or also vertex cover @xcite .
this paper is organized as follows : in section [ sec : model ] , we introduce the definition of vertex covers and independent sets , present the algorithms and review some important facts on random graphs .
the general dynamical equations for the graph evolution process are developed in section [ sec : dynamics ] , and they are solved for some cases in section [ sec : linear ] .
section [ sec : nonlinear ] is dedicated to approximations of the dynamical behaviour of cases which could not be solved exactly . in section [ sec : leaf ] , we present how the leaf - removal algorithm of bauer and golinelli @xcite fits into the presented scheme .
it is also generalized to cases where the original algorithm fails to construct a vertex cover .
the last section finally summarizes the results and gives an outlook to possible extensions .
let us start with the definition of vertex covers @xcite : take any graph @xmath0 with @xmath1 vertices @xmath2 and @xmath3 undirected edges @xmath4 .
vertex cover _
( vc ) is a subset @xmath5 of vertices such that for every edge @xmath6 there is at least one of its endpoints @xmath7 or @xmath8 in @xmath9 : @xmath10 we call the vertices in @xmath9 covered ( _ cov _ ) , whereas the vertices in its complement @xmath11 are called uncovered ( _ uncov _ ) . the definition of a vertex cover implies therefore that every edge has at least one covered end - point
. the full vertex set @xmath12 is of course a trivial vertex cover of any graph @xmath13 . in this case , all edges have two covered end - points , and at least some of the vertices can be set _
uncov _ without uncovering any edge .
the corresponding _ optimization problem _ consists in finding a vertex cover of smallest cardinality . this problem belongs , according to the standard book by garey and johnson @xcite , to the basic np - hard optimization problems .
therefore , it is expected to require a solution time growing exponentially in @xmath1 and @xmath3 .
the numerical solvability is consequently restricted to relatively small graphs .
vc is related to other well - known and widely used np - hard problems .
the first one is the _ independent set _ ( is ) problem .
an is is a subset @xmath14 of vertices such that for all @xmath15 we have @xmath16 .
so @xmath17 is obviously a vc for every is @xmath18 , and every maximal is is the complement of a minimal vc .
the _ independence number _ , defined as the maximal cardinality max@xmath19 of all iss , is consequently given by @xmath20 .
a _ clique _ is a fully connected subgraph .
so , if the vertex subset @xmath14 is an is in @xmath0 , it is a clique in the complementary graph @xmath21 .
finding the largest clique in one graph is equivalent to finding the largest independent set in the complementary graph .
there is also a physical problem which is equivalent to vertex cover , or more obviously to the independent set problem .
imagine the graph @xmath13 to be a lattice , and pack hard spheres of chemical radius 1 onto the vertices .
then , once a vertex is occupied by such a particle , all neighbouring vertices have to be empty .
this is exactly what defines an is .
the vertices which are not occupied by spheres thus form a vc .
this equivalence provides the basis of the statistical mechanics approach to minimal vertex covers on random graphs @xcite .
as already mentioned , the construction of a minimal vertex cover is np - hard , thus requiring exponential time resources .
it is therefore reasonable to develop good approximation algorithms running in polynomial time .
here we describe a class of linear time heuristic algorithms which are able to produce small , but in general suboptimal vcs . in this algorithm , vertices are sequentially assigned the values _
cov _ and _ uncov _ until the full graph is covered .
an assigned value is not changed any more .
this can be interpreted as a graph reduction process : once a vertex is covered , it can be removed from the graph , together with all incident edges . if a vertex is set _ uncov _ , all its neighbours have to be covered in order to cover the graph .
the central vertex , its neighbours and all covered edges can again be removed from the graph .
the graph thus becomes smaller and smaller , until no edges are left .
the size of the resulting vertex cover crucially depends on the order of the vertex selection and the decision to cover / uncover the selected vertex .
the main heuristic idea is simple : a vertex of high degree is more likely to be covered , a vertex of small degree is more likely to be uncovered @xcite . the simplest local information , namely the vertex degree , is thus correlated to the structure of small vertex covers , and can be exploited algorithmically .
this is done in the following way : given an initially uncovered graph @xmath0 and a non - negative integer @xmath22 , which we call the _ depth _ of the algorithm . in every algorithmic step a vertex @xmath7
is chosen randomly , and its nearest , 2nd - nearest ... and @xmath22th - nearest neighbours are selected .
all these vertices , together with all edges connecting two of it , form the induced subgraph @xmath23 .
this subgraph can be efficiently covered such that all @xmath22th nearest neighbours are set to _
cov_. for details see the algorithm presented below .
the full subgraphs , together with all edges connecting it to other vertices , is deleted from @xmath13 . for locally tree - like graphs ,
as e.g. random graphs or bethe lattices , the vertex covers of @xmath24 are especially simple : the @xmath22th neighbours of @xmath7 are covered , the @xmath25st are uncovered , the @xmath26nd are covered again , and so on , until @xmath7 itself is covered ( uncovered ) for even ( odd ) depth @xmath22 of the algorithm .
the main heuristic concerns now the question , how the central vertex @xmath7 is selected .
it exploits the above - mentioned correlation between vertex degree and covering state of an arbitrary vertex . for even depth @xmath22 , the central vertex
is covered by the above procedure @xcite .
so it is useful to choose more frequently vertices of high degree in the subgraph remaining after elimination of already considered vertices .
we therefore select a vertex of degree @xmath27 with some weight @xmath28 which is a monotonously growing function of @xmath27 . in the case of odd @xmath22 , the central vertex
is uncovered @xcite . here , the selection weight @xmath28 for vertices has to be a monotonously decreasing function of the vertex degree @xmath27 .
the algorithm is summarized as follows , inputs are the graph @xmath0 , the non - negative integer depth @xmath22 , a mapping @xmath29 , which is constantly set to @xmath30 initially , and the positive weight function @xmath31 : = = = = = = heuristic - vc(@xmath32 ) + * begin * + @xmath33 * then * + ; \{all edges are covered by vertices with @xmath34 } + select a vertex @xmath35 of current degree @xmath36 randomly with weight @xmath37 ; + @xmath38 + + @xmath39 ; + @xmath40 ; + heuristic - vc(@xmath32 ) ; + ; + + + @xmath41 ; + + @xmath42 ; + @xmath43 ; + @xmath44 ; + @xmath45 ; + @xmath46 for all @xmath22th neighbours of @xmath7 ; + @xmath47 + + select a vertex @xmath48 of maximal distance from @xmath7 ; + @xmath49 ; + @xmath50 for all neighbours @xmath51 of @xmath8 ; + @xmath52 ; + ; \{subgraph covered and removed from @xmath13 } + heuristic - vc(@xmath32 ) ; + ; + * end * ; + please note that the degree of free vertices may change whenever @xmath13 is reduced .
the algorithm always considers the current degree in the reduced subgraph , which equals the number of uncovered incident edges .
the algorithm therefore defines a markov process .
some special cases of the algorithm where already considered for finite connectivity random graphs : in @xcite , the case @xmath53 and @xmath54 was included into a complete backtracking algorithm .
the upper bound for the minimal vertex covers constructed in this way was rather poor , but it will be improved in sections [ sec : linear ] and [ sec : nonlinear ] by using better @xmath28 .
the case of @xmath55 was analyzed in @xcite using a different technique .
whereas being also quite unsatisfactory for small and intermediate average graph connectivities @xmath56 , this algorithm correctly reproduces the leading asymptotic behaviour for large @xmath57 . in @xcite ,
very surprising results where obtained for @xmath58 , where only vertices of degree 1 ( leafs ) are selected and uncovered , their neighbour is covered , and both are removed from the graph .
this algorithm is able to cover almost all edges for @xmath59 , thus producing a minimal vertex cover , but it stops for higher connectivities if no vertices of degree 1 are left , even if an extensive number of edges remains uncovered .
these two results suggest a promising generalization for the case @xmath60 : if we choose @xmath61 for all @xmath62 , this algorithm will work nearly as well as the leaf removal procedure for small connectivities , but it will also give the correct asymptotic behaviour for large @xmath57 .
the extreme case of choosing always a vertex of smallest current degree will work best on random graphs , but it goes beyond the analysis presented in this paper . in order to gain some insight into the typical behaviour of this algorithm
, we apply it to random graphs .
this subsection is dedicated to summarizing some interesting known results about these graphs , as far as they are important for our analysis . as can be expected from the algorithm presented above
, we will concentrate our attention to the distribution of vertex degrees . for a complete presentation see @xcite .
the original idea @xcite is to assign an equal probability to all graphs having the same numbers @xmath1 of vertices and @xmath3 of edges .
a random graph @xmath63 , with @xmath64 , is constructed in the following way : the vertex set is chosen to be @xmath65 . for all vertex pairs @xmath66 an edges is included into @xmath67 with probability @xmath68 .
the two vertices remain disconnected by a direct edge with probability @xmath69 .
this graph has on average @xmath70 edges , its mean vertex degree equals @xmath71 .
the most interesting case for vertex covers are graphs of finite average connectivity , i.e. the average vertex degree @xmath56 stays finite in the thermodynamic limit @xmath72 . in the above language , we have to fix @xmath73 .
the resulting degree distribution is far from uniform . for @xmath74 ,
a randomly chosen vertex has degree @xmath27 with probability @xmath75 i.e. the degree distribution approaches a poissonian in the thermodynamic limit . in our analysis
we also need the probability of finding a vertex of degree @xmath27 by following an arbitrary edge .
it is obviously proportional to @xmath76 , due to normalization we have @xmath77 we thus find a poissonian distribution of @xmath78 .
the average degree of vertices reached by following an edge equals @xmath79 , so there are on average @xmath57 additional edges .
random graphs undergo a percolation transition at average vertex degree @xmath80 . below this threshold ,
the graph consists of an extensive number of small connected components , each containing up to @xmath81 vertices . for @xmath82 ,
the number of small connected components of @xmath83 is still extensive , but there is also one macroscopic connected component of @xmath84 vertices .
this giant component grows with increasing @xmath57 , and exponentially approaches size @xmath1 when @xmath57 becomes large .
the concept of random graphs was recently generalized to random graphs of arbitrary degree distribution @xcite . there
, every graph of a given distribution @xmath85 ( not necessarily a poissonian ) is assigned the same probability .
these graphs can be easily generated : for all vertices @xmath86 , a degree @xmath36 is drawn randomly from @xmath85 .
if @xmath87 is even , we continue , if not , we repeat the above procedure .
then a large vertex set is created , containing every vertex @xmath7 exactly @xmath36 times .
now we sequentially select pairs of vertices and add these to the edge set @xmath67 , excluding only self - connections .
following again an edge , the reached vertex has degree @xmath27 with probability @xmath88 , with @xmath89 denoting the average degree . before analysing the behaviour of the presented heuristics
, we will give a short overview over known properties of minimal vertex covers on random graphs , see @xcite for the original presentation .
the analysis there was carried out for random graphs @xmath90 , using the mapping to a hard - sphere lattice gas described in subsection [ sec : vc ] . using the replica method ,
a grand - canonical approach was taken , including a chemical potential @xmath91 controlling the number of hard spheres . in the limit @xmath92
, the system tends to the closest packings , or , equivalently , to the minimal vertex covers .
assuming the validity of replica symmetry , it was found that minimal vcs contain a fraction of vertices given by @xmath93 with @xmath94 being the real branch of the lambert - w function defined as the inverse of @xmath95 .
it was also shown , that replica symmetry is locally stable for @xmath96 , whereas it is unstable for larger average connectivities , leading to broken replica symmetry .
the correctness of ( [ eq : minvc ] ) was recently shown to be exact in @xcite based on a leaf - removal algorithm , which , as mentioned above , can be understood as a variant of heuristic - vc .
an interesting insight into the structural properties of minimal vertex covers was given by identifying a covered and an uncovered _ backbone_. the first one is defined as the set of all vertices which are covered in all minimal vcs , the uncovered backbone unifies all vertices being _ uncov _ in all minimal vcs .
denoting their relative sizes by @xmath97 , the replica symmetric analysis leads to @xmath98 the remaining @xmath99 vertices belong neither to the covered nor the uncovered backbone , they change the covering state from one minimal vc to the next .
a strong correlation between degree distribution and backbone was observed : vertices of small degree tend to be more frequently in the uncovered backbone , whereas vertices of high degree can be found more likely in the covered backbone .
as mentioned in subsection [ sec : algo ] , this can be exploited in the heuristic algorithm by adapting the selection weights .
we assume that the input to the algorithm heuristic - vc is a random graph @xmath0 with @xmath1 vertices and degree distribution @xmath85 , and we concentrate on the graph reduction process for a moment .
the size of the constructed vertex cover will be calculated in subsection [ sec : vcsize ] . in every algorithmic step ,
a vertex is selected with weight @xmath28 depending only on its current degree @xmath27 .
then , this vertex and all its nearest neighbours , 2nd nearest neighbours , ...
@xmath22th nearest neighbours are removed from the graph . the edges incident to these vertices
are removed , too . following this procedure ,
a smaller graph is defined , and the algorithm is iterated .
this graph reduction process is markovian , because the action of each algorithmic steps depends only on the properties of the current reduced graph , more precisely on the current vertex degrees and neighbourhood relations .
let us further assume , that we start at ( algorithmic ) time @xmath100 , and every iteration step is counted as @xmath101 .
the dynamics will be described by rate equations for the vertex - degree distribution @xmath102 , or the number of vertices @xmath103 of degree @xmath27 , where @xmath104 denotes the remaining vertex number at time @xmath105 .
their dynamics can be decomposed into the following elementary processes ( where @xmath106 denotes the average over the current degree distribution @xmath102 ) : * _ removal of the central vertex _ : a central vertex of degree @xmath27 is selected with weight @xmath28 , i.e. with probability @xmath107 .
@xmath108 is thus reduced by one with this probability . *
_ removal of the 1st , 2nd , ... , @xmath22th nearest neighbours _ : according to the last item , the central vertex has on average @xmath109 neighbours . as the degrees of neighbouring sites are uncorrelated in random graphs , each of these has degree @xmath27 with independent probabilities @xmath110 .
random graphs are locally tree - like , @xmath78 of the @xmath27 edges of a 1st neighbour lead to 2nd nearest neighbours , i.e. the average number of 2nd neighbours equals @xmath111 .
this argument can easily extended to 3rd neighbours etc . * _ update of the connectivity of @xmath112st neighbours _ : the edges connecting @xmath22th and @xmath112st neighbours are removed from the graph , too .
the degree of every @xmath112st neighbour is thus reduced by one .
these processes are combined to evolution equations for the _ expected numbers _
@xmath108 of vertices with degree @xmath27 at time @xmath105 : @xmath113 the first line describes the deletion of vertices , the second the update of the degrees of all @xmath112st neighbours .
these equations are valid for the average trajectory , which is , however , followed with probability approaching 1 for @xmath114 .
macroscopic deviations appear only with exponentially small probability and are thus important for small @xmath1 only .
the quality of using the average trajectory is demonstrated in the inset of fig .
[ fig : depthzero ] .
there the trajectory of a single graph with @xmath115 vertices is found to excellently follow the analytical prediction .
using equations ( [ eq : rate ] ) , we can calculate also the evolution of the total numbers of remaining vertices , @xmath116 , and edges , @xmath117 : @xmath118^{m-1 } - \left [ \frac{{\langle}d ( d-1 ) { \rangle}_t}{{\langle}d { \rangle}_t}\right]^{k+1 } \right)\ .\end{aligned}\ ] ] as we are mainly interested in the behaviour of large graphs , @xmath119 , we may change to intensive quantities by writing @xmath120 . setting further @xmath121 , and replacing differences by derivatives in the thermodynamic limit , we find @xmath122 the graph reduction process is thus described by an infinite set of non - linear differential equations , where the non - linearity enters only through the time - dependent averages @xmath123 .
as we were starting with an ordinary random graph , these equations have to be solved under the initial condition @xmath124 where @xmath57 equals the initial average vertex degree . in section [ sec : nonlinear ]
we also need the dynamical equations for the moments @xmath125 of @xmath102 . multiplying the second of eqs .
( [ eq : dgl ] ) with @xmath126 and summing over all degrees yields @xmath127 please note that these equations do not contain any finite and closed subset of equations , because the evolution of any moment depends also on higher moments .
a similar approach was chosen in @xcite to analyze an algorithm constructing the maximal sub - graph having minimal degree @xmath128 , i.e. the so - called @xmath128-core .
pittel et al .
rigorously derived and solved a closed set of equations for @xmath129 and @xmath130 . an analogous reduction to a finite number of equations
will be constructed in section [ sec : leaf ] for the generalized leaf - removal algorithm , but it can not be achieved for the general case .
our approach resembles also the rate equation approach used in the area of growing networks @xcite .
note , however , that there the evolution of the number @xmath108 of vertices of degree @xmath27 depends only on the vertices having smaller degree , because edges are always added and never deleted .
so , in principle , the evolution equations can be solved by calculating first @xmath131 , then @xmath132 and so on . in the graph reduction process
the problem becomes more complicated since the evolution of @xmath133 depends also on @xmath134 , for arbitrary @xmath27 , but there is no maximal degree @xmath27 in the poissonian initial condition . before trying to solve equations ( [ eq : dgl],[eq : ic ] ) for specific choices of @xmath28 and @xmath22
, we will give general expressions for the number @xmath135 of vertices which are covered by the algorithm .
for the locally tree - like case of random graphs , the @xmath22th neighbours of the selected vertex are covered , the @xmath25st are uncovered etc .
so the covering state of the central vertex depends on the depth @xmath22 : for even @xmath22 , it will be covered ( if @xmath136 ) , for odd @xmath22 , it will be uncovered .
we therefore consider these two cases independently . for odd @xmath22 ,
the central vertex is almost always uncovered .
denoting the expected number of covered vertices at time t with @xmath137 , we thus find @xmath138 cf .
the first of equations ( [ eq : nm ] ) .
going again to the limit of large graphs , @xmath72 , this can be written as a differential equation for @xmath139 : @xmath140 once we know the solution of the graph dynamical equation ( [ eq : dgl ] ) , we can calculate the time @xmath141 where all edges are covered , @xmath142 , and integrate the last equation over the time interval @xmath143 $ ] . as all removed edges were covered by our algorithm ,
we thus have constructed a vertex cover of relative size @xmath144 .
as the described average trajectory is followed with probability one for @xmath145 , this @xmath144 gives an upper bound for the true minimal vertex cover size of the random graph under consideration .
for even @xmath22 , the central vertex is covered in general .
only if it is disconnected , i.e. if its degree equals zero , it is set to _
, see the algorithm heuristic - vc .
the last case happens with probability @xmath146 .
we thus conclude for @xmath135 @xmath147 or , in the limit @xmath72 , @xmath148 this equation can be integrated , once we know the solution of equations ( [ eq : dgl ] ) , and an upper bound to @xmath149 can be read off .
the problem in solving differential equations ( [ eq : dgl ] ) with initial conditions ( [ eq : ic ] ) is , that the poissonian shape of the degree distribution is , in general , not conserved under the dynamics . in such cases one has to keep track of all the individual probabilities @xmath102 for each possible degree @xmath27 . as the dynamics of @xmath102 depends on @xmath150 for all @xmath27 , and @xmath27
is not bound from above for the poissonian initial condition , it is not obvious how to construct a finite and closed subset of equations which can be solved separately , opening the door to the solution for all @xmath85
. there exist , however , some cases where the poissonian shape of the degree distribution is conserved , as can be shown explicitly by plugging a poissonian ansatz @xmath151 into ( [ eq : dgl ] ) and verifying , that the same equation for the average vertex degree @xmath152 is reproduced for arbitrary @xmath27 . the most general case for this behaviour is found for linear selection weights @xmath153 where @xmath154 are arbitrary non - negative real numbers . in this case , the graph can be totally specified by calculating @xmath155 and @xmath156 . their evolution can be read off from equations ( [ eq : nm ] ) , where the averages @xmath123 can be expressed via @xmath152 : @xmath157 in the limit @xmath72 the graph reduction dynamics is thus completely determined by the differential equations @xmath158 eliminating @xmath159 from the second equation , we end up with @xmath160 these equations have to be solved under the initial conditions @xmath161 and @xmath162 equations ( [ eq : dgllin],[eq : cdgl ] ) simplify further if we restrict it for a while to constant selection weights @xmath164 , i.e. @xmath163 . there we find @xmath165 in the second line , the equation for @xmath159 was already used to eliminate the complicated sum of powers of @xmath152 . using the initial conditions @xmath161 and @xmath162 , the second line results in @xmath166 and one of the two functions can be eliminated from the first of equations ( [ eq : dglconst ] ) .
we consequently find @xmath167 which is solved implicitly by @xmath168 the algorithm stops when all edges are covered , i.e. for @xmath169 .
this final time @xmath141 corresponds , in the original algorithmic language , to @xmath170 iterations of heuristic - vc , and is given by @xmath171 these results can be used in order to determine the relative size @xmath172 using the results of section [ sec : vcsize ] .
there we observed a difference between even and odd values @xmath22 of the depth of heuristic - vc according to the fact that the central vertex in one case is almost always covered , in the other case uncovered .
we therefore continue discussing these cases separately , starting with @xmath53 , going then to arbitrary odd @xmath22 , and discussing general even @xmath22 at the end of this section .
the simplest possible algorithm has depth @xmath53 : in every algorithmic step an arbitrary vertex is chosen and covered if its degree is non - zero , uncovered else .
the vertex and all incident edges are removed from the graph .
this simple algorithm was already analysed in @xcite as the heuristic underlying a complete backtracking algorithm .
the results given there can be easily reproduced , for @xmath53 the integration in eq .
( [ eq : ct ] ) can be trivially carried out .
we find a simple linear decrease of the average vertex degree , @xmath173 and the final time becomes @xmath174 as one vertex is removed in every algorithmic step .
the resulting size of the constructed vertex cover follows easily by integrating equation ( [ eq : xeven ] ) , with @xmath175 : @xmath176 cf .
this gives the very first and simplest upper bound on the true size of minimal vertex covers which is , however , not very good : for , e.g. , @xmath177 , we find @xmath178 compared to the true value @xmath179 , cf .
( [ eq : minvc ] ) . also the asymptotic behaviour for large average vertex degrees , @xmath180 ,
does not meet the exact asymptotic behaviour , which was evaluated by frieze @xcite to be @xmath181 to leading order .
this asymptotics is found to hold also for a slightly more complicated case : it is valid for all non - zero values of the depth , i.e. for @xmath182 .
let us , for simplicity , start with arbitrary odd values of the depth @xmath22 of the algorithm . for @xmath60 ,
the algorithm is equivalent to the one proposed by gazmuri @xcite : in every time step , an arbitrary vertex is chosen and set to _
uncov _ , all its neighbours are covered , and the whole cluster including all incident edges are removed from the graph .
gazmuri has already characterized the performance of this algorithm , using a different technique .
please remember that , according to equation ( [ eq : cdot ] ) , the time dependency of the mean vertex degree is given by @xmath183 this can be used to solve equation ( [ eq : xodd ] ) for the evolution of the number of covered vertices , which , using the poissonian distribution ( [ eq : poissonansatz ] ) , and integrating over @xmath105 reads @xmath184 changing variables from @xmath105 to @xmath185 , and plugging in equation ( [ eq : cdot1 ] ) , we find @xmath186 this expression gives the @xmath187-@xmath185-trajectory which is interestingly independent on the depth @xmath22 ( as long as @xmath22 is odd ) .
the only difference is given by the time dependencies @xmath188 as @xmath152 is @xmath22-dependent , cf . equation ( [ eq : ct ] ) .
the graph is completely covered when @xmath152 reaches zero .
the cardinality of this vertex cover is almost surely given by @xmath189 , i.e. @xmath190 independently on the ( odd ) depth @xmath22 of the algorithm .
for @xmath60 , gazmuri s bound is thus reproduced , and the leading order of the behaviour for large initial connectivities @xmath57 is correctly found .
the algorithm is , however , less successful for small and intermediate connectivities , as we will see in the following sections .
the case of even depth leads to more complicated expressions , which can not be evaluated explicitly .
the main problem is induced by the @xmath192-contribution in the evolution of @xmath139 , as given in equation ( [ eq : xeven ] ) .
after having applied the poissonian ansatz ( [ eq : poissonansatz ] ) , the latter reads @xmath193 hence we find for the relative vertex - cover size , and arbitrary even depth @xmath22 @xmath194 this integral can not be evaluated explicitly for arbitrary even @xmath22 .
we can , however , extract the asymptotic behaviour for @xmath195 . in this limit
, the terms of @xmath196 and @xmath197 can be neglected in both the numerator and the denominator in the last integral .
the corrections from the integration interval @xmath198 are of order 1 and thus suppressed by the prefactor @xmath199 , compared to the leading terms @xmath200 which coincide with the case of odd depth . after having discussed in great detail algorithms with @xmath202 , i.e. simple algorithms selecting central vertices completely at random , without regarding its degree
, we now turn to the case of linear @xmath28 . in this case , as already mentioned at the beginning of this section , the degree distribution still stays poissonian .
this can be understood intuitively in the following way : the case @xmath203 is equivalent to choosing an arbitrary edge with constant probability , and selecting one of the end - vertices . according to section [ sec : randomgraph ] , the chosen vertex will have degree @xmath27 with probability @xmath204 . the more general case @xmath205 , cf .
( [ eq : linear ] ) , corresponds to mixing this selection procedure ( weight @xmath206 ) with the uniform selection of vertices ( weight @xmath207 ) .
fixing @xmath206 to a non - zero value is sensible only for _ even depth _
values @xmath22 since these correspond to covering the central vertex of the cluster of radius @xmath22 .
as mentioned above , the vertex degree is correlated to the covering state in small vertex covers , so it does not make sense to preferentially select vertices of high degree and to uncover it subsequently .
the performance of the algorithm is thus only improved for even @xmath22 .
the strongest improvement is obtained for @xmath208 , i.e. for @xmath209 .
whereas eqs .
( [ eq : dgllin],[eq : cdgl ] ) lead to complicated coupled non - linear differential equations for @xmath152 and @xmath210 , which have to be solved numerically for general even @xmath22 , the case @xmath53 becomes very simple .
there , the equations read @xmath211 and are solved by @xmath212 the graph is covered for @xmath169 , which leads to vertex covers of size @xmath213 this size is always smaller than the one found for the very simplest algorithm ( @xmath214 ) , but stays worse than gazmuri s algorithm ( @xmath215 ) .
also the asymptotic behaviour is not correctly reproduced .
if we choose non - linear selection weights @xmath216 , the graph reduction dynamics deviates from the unrestricted ensemble of random graphs , and the degree distribution becomes non - poissonian .
the new distribution thus can not be described by the evolution of its mean value alone , and we have to solve all equations from ( [ eq : dgl ] ) simultaneously . in general , this can not be achieved analytically . to approximate the solution numerically
, we may cut the tail of @xmath102 , and solve only a finite number of equations .
this works fine for small initial values of @xmath217 because of the rapid decrease of the poissonian distribution . for larger values of @xmath57 however , the number of remaining equations becomes large , too .
so it would be better to find an approximation of @xmath102 depending only on a small number of parameters .
the dynamics of these parameters can be determined from the lowest moments @xmath218 with @xmath219 ranging from 1 to the number of parameters .
this section is dedicated to algorithms of depth @xmath53 , but non - linear selection weights @xmath220 .
every algorithmic step thus removes one vertex from the graph .
no vertices of degree 0 are selected ( @xmath221 ) ; all selected vertices thus have to be covered according to heuristic - vc .
we thus trivially have @xmath222 and the size of the constructed vertex cover is given by @xmath223 with @xmath169 .
note that this does not necessarily imply @xmath224 as only vertices of non - vanishing degree are selected . at time @xmath141 , the remaining graph consists of @xmath225 completely disconnected vertices .
as the initial condition @xmath226 of the graph reduction dynamics is given by a poissonian of mean @xmath57 , we are looking for a deformation of the poissonian distribution which allows to independently vary mean and variance , and thus to approximate the true @xmath102 . a simple possibility is given by the generalized binomial expression @xmath227^{(\mu(t)^{-1}-d ) } \ \left[c(t)\mu(t)\right]^d\ .\ ] ] for @xmath228 , this expression approaches a poissonian distribution of mean @xmath152 , the initial condition is thus characterized by @xmath162 and @xmath229
. please note , however , that ( [ eq : binomial ] ) does not necessarily describe a probability distribution , because @xmath230 becomes negative for certain @xmath27 if @xmath231 is not a positive integer .
we do not expect this to produce serious problems for the calculated averages , as long as the absolute value of these negative @xmath230 stays neglectable compared to the total normalization 1 .
this is exactly what happens in our case , as we will see below .
the moments of @xmath230 are given by @xmath232[1 - 2\mu(t)]\cdots[1-n\mu(t)]\ .\ ] ] the product @xmath233 can be expanded into a sum of pure powers , their mean values are determined by inverting this sum : @xmath234 the coefficients are given iteratively by @xmath235 , using the trivial identity @xmath236 and the convention @xmath237 for all @xmath238 .
given two non - trivial moments , e.g. @xmath239 , @xmath152 and @xmath240 can be calculated , and all other moments are determined .
we can thus approximate the dynamics of the degree distribution by considering of the dynamics of the first two moments of @xmath102 only .
the exact equations for @xmath241 and @xmath242 follow from eq .
( [ eq : dglmoment ] ) by fixing @xmath53 and @xmath243 , @xmath244 they contain higher moments of @xmath102 , but are closed approximately by using ( [ eq : moment1],[eq : moment ] ) : @xmath245 \right ) & = & c(t)^2 [ 1-\mu(t ) ] - \frac { \tilde \pi^{(\alpha+2)}(t)}{\tilde \pi^{(\alpha)}(t ) } + \frac { \tilde \pi^{(\alpha+1)}(t)}{\tilde \pi^{(\alpha)}(t)}\left ( 1 - 2 c(t ) [ 1-\mu(t ) ] \right)\end{aligned}\ ] ] these two equations can be easily converted to ordinary differential equations @xmath246 and @xmath247 which are , however , not analytically solvable for general values of @xmath248 .
they can instead be solved efficiently using numerical standard techniques .
the results are displayed in fig .
[ fig : depthzero ] together with numerical simulations obtain for large graphs .
we find that the binomial approximation works extremely well for small values of @xmath248 , systematically growing deviations appear for larger @xmath248 , cf .
[ fig : distr ] .
we also find that heuristic - vc is able to approximate the true minimal vertex - cover size up to a few percent , performing better for larger @xmath248 .
we observe , however , that even the vertex covers constructed for very large @xmath248 remain suboptimal , i.e. extensively larger than the minimal vcs .
the behaviour of eqs .
( [ eq : binomapprox ] ) for large @xmath162 can be extracted analytically by expanding the equations for @xmath249 ( @xmath250 ) to @xmath251 ( @xmath252 ) .
@xmath253 is a sum of terms of @xmath254 , @xmath255 and @xmath252 , and @xmath240 stays of @xmath252 due to its initial condition @xmath256 .
we thus find @xmath257 which is solved to leading orders by @xmath258 . from the vanishing of these leading orders we can read of the dominant contributions to the constructed vertex covers , which we conjecture to be exact also for the true dynamics : @xmath259 for larger @xmath248 ,
the asymptotics becomes better and better , but still does not reach the correct behaviour @xmath260 of the minimal vcs .
note that , for sufficiently large @xmath57 , gazmuri s depth - one algorithm outperforms the depth - zero algorithm for arbitrary @xmath248 .
we expect , however , that the correct asymptotic behaviour is reached by exponential selection weights @xmath28 , and hence in particular for the heuristic where always a vertex of maximal degree is chosen .
the best performance can be achieved using a generalization of the leaf - removal algorithm ( lr ) proposed by bauer and golinelli @xcite .
their algorithm is based on the following observation : consider a vertex of degree one , i.e. a vertex which has a single neighbour .
one of these two vertices has to be covered in order to cover also the connecting edge .
it is obviously better to cover the neighbour than the vertex of degree one .
iterating this procedure , there are two possible final situations : + ( i ) all edges are covered , and the constructed vertex cover is a minimal one .
+ ( ii ) there are uncovered edges , but no vertices of current degree one are left .
the algorithm stops without having constructed a vertex cover . +
this algorithm can be understood as a special case of heuristic - vc with depth @xmath60 and @xmath261 . following a completely different route , bauer and golinelli
have found a surprising result : for random graphs of average degree @xmath262 , their algorithm is able to cover almost all edges , and the predicted minimal vertex cover size coincides with the replica symmetric one found in @xcite . for larger average degrees ,
@xmath263 , case ( ii ) is valid , and a finite fraction of edges remains uncovered .
we re - derive this result below using our dynamical rate equations . in order to construct a small vertex cover also for higher - connectivity graphs ,
the algorithm has to be modified to @xmath264 for all @xmath265 .
the case @xmath266 , with @xmath267 , will therefore be analysed in sec .
[ sec : leafgeneral ] . for @xmath268
, this algorithm performs nearly as well as pure leaf - removal for small average vertex degrees . for large @xmath57
it finally becomes more and more similar to gazmuri s algorithm , being still better for any finite @xmath57 .
we will call this algorithm _ generalized leaf removal _ ( glr ) .
the best performance is , of course , obtained for @xmath269 , with @xmath270 .
there always a vertex of minimal degree is selected and uncovered , all neighbours are covered .
as long as the fraction @xmath271 of vertices of degree one is non - zero , the algorithm is equivalent to the leaf - removal procedure .
this is valid in particular also for @xmath262 , where ( almost ) minimal vcs are constructed .
the analysis of this algorithm goes , however , beyond the analysis presented here .
let us first concentrate on the leaf - removal algorithm ( lr ) in its original version , i.e. on depth @xmath60 and selection weights @xmath272 .
in every algorithmic step , exactly two vertices are removed from the graph @xmath13 , and exactly one vertex is covered .
we therefore conclude @xmath273 the degree distribution follows , according to ( [ eq : dgl ] ) , from the dynamical equations @xmath274 vertices of degree @xmath62 are only touched if they are first or second neighbours of a vertex of degree one , in which cases they are either covered and removed , or their vertex degree is reduced by one .
the degrees of neighbouring vertices are statistically independent , we thus expect @xmath102 to keep its poissonian shape for all @xmath62 .
in fact , the ansatz @xmath275 together with the global normalization @xmath276 \right)\ ] ] can be plugged into eqs .
( [ eq : lrdgl ] ) and leads to uniquely determined equations for @xmath277 and @xmath271 .
the latter is , for technical reasons , replaced by an equation for @xmath152 . using @xmath278 \nonumber\\ { \langle}d(d-1 ) { \rangle}_t & = & \gamma(t ) \kappa(t)^2\ , \end{aligned}\ ] ] these
read @xmath279 the initial conditions are @xmath280 .
the equation for the average vertex degree @xmath152 can be removed by observing @xmath281 which is solved by @xmath282 the solution of the two remaining equations is given implicitly by @xmath283 ^ 2 + 2w(c_0e^{\kappa(t ) } ) \right ) \nonumber\\
\gamma(t ) & = & \frac{w(c_0e^{\kappa(t ) } ) } { ( 1 - 2t)c_0}\end{aligned}\ ] ] as can be checked explicitly using eqs .
( [ eq : lrreduceddgl ] ) .
@xmath284 denotes again the lambert - w function , cf .
[ sec : minvc ] .
the graph is covered if this trajectory reaches @xmath169 , i.e. for @xmath285 . from the first of eqs .
( [ eq : lrsol ] ) we thus find the final time @xmath286 this result is only valid , if @xmath287 for all @xmath27 and all @xmath288 .
( [ eq : lrmoments ] ) , we find @xmath289 \nonumber\\ & = & \frac { \kappa(t)}{(1 - 2t)c_0 } \left ( \kappa(t ) - w(c_0
e^{\kappa(t ) } ) [ 1-e^{-\kappa(t ) } ] \right ) \nonumber\\ & = : & \frac { \kappa(t)}{(1 - 2t)c_0 } \phi ( \kappa(t ) ) \ .\end{aligned}\ ] ] the prefactor of @xmath290 is non - negative , it is thus sufficient to investigate @xmath290 for @xmath291 .
we have @xmath292 , and @xmath293 - e^{-\kappa}w(c_0 e^{\kappa})\ , \ ] ] i.e. @xmath294 . the monotonous function @xmath94 becomes larger than 1 for @xmath263 , i.e. in this case @xmath271 would approach zero from negative values .
this is a contradiction .
we therefore conclude @xmath295 and this value coincides with the relative size of a minimal vertex cover . for @xmath263
the algorithm gets stuck if @xmath296 is reached , no vertices of degree one are left , and a finite fraction of all edges remains uncovered . in order to overcome this problem
, we generalize the leaf - removal algorithm by modifying the selection weights to @xmath297 .
this algorithm interpolates between the algorithms of bauer and golinelli ( @xmath298 , exact minimal vcs for @xmath262 ) and the one of gazmuri ( @xmath299 , correct leading asymptotic behaviour for @xmath300 ) .
we thus expect that this algorithm shows a very good performance in the whole finite - connectivity region for large , but finite @xmath206 .
also in this case , the poissonian shape of @xmath102 remains correct for all degrees @xmath62 , and ansatz ( [ eq : lransatz ] ) together with the normalization constraint ( [ eq : norm ] ) remains valid . plugging everything into the dynamical equations ( [ eq : dgl ] ) , we directly arrive at @xmath301[c(t)+ap_1(t)]}{[1+ap_1(t ) ] c(t ) } \nonumber\\
n(t ) \dot \kappa(t ) & = & -\kappa(t ) \frac{[\gamma(t)\kappa(t)^2+c(t ) ] [ c(t)+ap_1(t)]}{[1+ap_1(t ) ] c(t)^2 } \nonumber\\ n(t ) \dot \gamma(t ) & = & \gamma(t ) \frac{c(t)+2a p_1(t)}{1+a p_1(t ) } -\kappa(t)\gamma(t ) \frac{c(t)+a p_1(t)}{1+a p_1(t)}\end{aligned}\ ] ] with @xmath302 $ ] .
these equations determine the exact graph reduction dynamics for generalized leaf removal , and can be solved numerically .
due to the non - zero selection weights for all degrees , these equations do not suffer from the appearance of negative values for certain @xmath102 , and the algorithm always constructs a vertex cover .
the number of covered vertices at algorithmic time @xmath105 can be calculated from eq .
( [ eq : xodd ] ) , which for our special choices of @xmath22 and @xmath28 reads @xmath303 the relative size of the finally constructed vertex cover is , with probability one , given by @xmath144 , with @xmath141 following from @xmath169 . in figure [ fig : glr ] , the results are presented for several values of @xmath206 and compared with gazmuri s resp .
bauer and golinelli s algorithms .
the solution of many combinatorial optimization problems requires exponential time resources and is thus restricted to relatively small system sizes . for larger systems , good and fast approximation algorithms are needed , which are frequently based on heuristic considerations concerning correlations between local problem structure and optimal problem solutions .
constructing a minimal vertex cover of a given graph belongs to the basic np - hard problems , and can be understood as a prototype combinatorial optimization problem . in this paper , we have therefore analysed linear - time algorithms for constructing small vertex covers of finite - connectivity random graphs . the applied heuristic exploits the observation that vertices of high vertex degree are more likely to be covered in minimal vcs , whereas those of small degree remain more frequently uncovered .
we have introduced and analysed mainly two types of algorithms , namely depth - zero and depth - one algorithms . in the first case ,
vertices are selected randomly and ( if connected to any other vertex ) they are covered .
we found that the performance of the algorithm can be largely improved by preferentially selecting vertices of high degree .
we also observed that depth - zero algorithms were outperformed by depth - one algorithms .
these select a vertex in every algorithmic step and uncover it .
all neighbours of the selected vertex must be covered consequently .
the best performance is achieved if the algorithm always selects a vertex of smallest degree . as was already found by bauer and golinelli @xcite , this procedure even outputs an ( almost ) minimal vc if it is applied to random graphs of average connectivity @xmath262 .
the algorithm constructs good approximations also for higher connectivities .
both types of algorithms can be interpreted as markovian graph reduction processes .
they are analytically characterized by the evolution of the degree distribution of the remaining uncovered subgraphs .
the dynamical equations were solved in some cases , in other cases approximations were necessary . *
the applied heuristic was restricted to considering the simplest local structure , namely the degree of the selected vertex .
depth - one algorithms can be improved by e.g. selecting a vertex of minimal degree but with maximal number of next - nearest neighbours . covering the nearest neighbours
thus results in a higher number of covered edges .
it would be interesting to extend the rate equations to this case .
* vc is used as a prototype optimization problem , but the approach can be generalized to other combinatorial problems defined over random structures , e.g. to graph coloring or satisfiability problems . for random 3-satisfiability , lower bounds for the sat / unsat threshold
are usually obtained using algorithms in the so - called card - game representation @xcite , which corresponds to @xmath304 in our analysis . *
our analysis was restricted to the typical time evolution of the degree distribution .
deviations appear with exponentially small probability - and are thus important for small systems .
these rare events can be systematically exploited by exponentially frequent restarts of the algorithm .
if a minimal vc is found with probability @xmath305 , we need @xmath306 restarts to almost surely construct a vc ( for all @xmath307 ) .
as observed recently for simple heuristics of vertex cover @xcite and 3-satisfiability @xcite , this random restart algorithm can be exponentially faster than exact standard procedures ( like backtracking ) . the performance can be improved further by using a more sophisticated heuristics . * also sophisticated complete algorithms , i.e. those that find an optimal solution for sure , use heuristic arguments for accelerating the combinatorial search .
the presented ideas may hence contribute to the analysis of such algorithms , and the insight may be used to exponentially speed up the numerical search .
garey and d.s .
johnson , _ computers and intractability _ , ( freeman , san francisco 1979 ) .
please note that `` approximation '' is not used in the sense of a polynomial - time approximation scheme , which guarantees to approximate the optimal solution within some given error for arbitrary samples .
here we mean low - cost configurations which may serve as reasonable solutions in practical applications .
special issue of art .
* 81 * ( 1996 )
. special issue of theor .
* 265 * ( 2001 ) .
gazmuri , networks * 14 * , 367 ( 1984 ) .
wormald , ann .
. prob . * 5 * , 1217 ( 1995 ) .
pittel , j. spencer , and n.j .
wormald , j. combinat .
b * 67 * , 111 ( 1996 ) .
j. franco , theor .
sci . * 265 * , 147 ( 2001 ) .
d. achlioptas , theor . comp .
sci . * 265 * , 159 ( 2001 ) .
r. monasson and r. zecchina , phys .
e * 56 * , 1357 ( 1997 ) .
r. monasson , r. zecchina , s. kirkpatrick , b. selman , and l. troyansky , nature * 400 * , 133 ( 1999 ) .
s. mertens , phys .
lett . * 81 * , 4281 ( 1998 ) .
m. weigt and a.k .
hartmann , phys .
* 84 * , 6118 ( 2000 ) a.k . hartmann and m. weigt ,
theor . comp .
sci . * 265 * , 199 ( 2001 ) .
m. bauer and o. golinelli , phys .
lett . * 86 * , 2621 ( 2001 ) .
m. bauer and o. golinelli , eur .
j. b * 24 * , 339 ( 2001 ) .
m. weigt and a.k .
hartmann , phys .
e * 63 * , 056127 ( 2001 ) .
m. weigt and a.k .
hartmann , phys .
lett . * 86 * , 1658 ( 2001 ) .
this works at least for the case where the graph contains no loops of length up to @xmath308 , which is allways true for the most inportant cases @xmath309 . also for random graphs , @xmath24 is almost allways a tree , and the above argument works with probability one ( for @xmath72 ) .
b. bollobs , _ random graphs _ , ( academic press 1985 ) .
p. erds and a. rnyi , publ .
inst . hung .
sci . * 5 * , 17 ( 1960 ) .
m. molloy and b. reed , rand .
struct . alg .
* 6 * , 161 ( 1995 ) .
newman , s.h .
strogatz , and d.j .
watts , phys .
e * 64 * , 026118 ( 2001 ) .
r. albert and a. l. barabasi , rev . mod .
74 , 47 ( 2002 ) .
krapivsky , s. redner , and f. leyvraz , phys .
lett . * 85 * , 4629 ( 2000 ) .
dorogovtsev , j.f.f .
mendes , and a.n .
samukhin , phys .
lett . * 85 * , 4633 ( 2000 ) .
a.m. frieze , discrete math . * 81 * , 171 ( 1990 ) .
a. montanari and r. zecchina , phys .
* 88 * , 178701 ( 2002 ) .
s. cocco and r. monasson , preprint cond - mat/0203012 . of the vertex
covers constructed by the depth - zero heuristic with @xmath310 ( full lines from top to bottom ) . for a comparison ,
the replica symmetric @xmath311 ( dashed line , exact for @xmath59 ) and the results @xmath312 of gazmuri s depth - one algorithm ( dotted line ) are added .
the inset shows the time dependent average vertex degree @xmath152 for the same values of @xmath248 ( full lines from top to bottom ) together with numerical data for a single random graph with @xmath313 .
this illustrates the quality of taking the average trajectory ( @xmath314 ) , as well as the quality of the binomial approximation ( @xmath315).,width=566 ] for heuristic - vc with depth @xmath53 , selection weights @xmath316 , and initial condition @xmath317 , for @xmath318 ( full , dotted , dashed lines ) .
the symbols are numerical data for a single graph of size @xmath319 and coincide extremely well with the binomial approximation represented by the lines ( lines are connecting data for integer @xmath27 and thus are guides to the eyes only ) .
the quality of the approximation is similarly good for all investigated initial @xmath57 .
+ ( b ) same as ( a ) , but with @xmath320 .
the coincidence between numerical data and binomial approximation is slightly worse than in ( a ) , but still very convincing .
the quality of the approximation increases with growing initial @xmath57 , supporting thus our conjecture that the asymptotic behaviour is correctly described by the approximation.,width=566 ] of the vertex covers constructed by generalized leaf removal .
the selection weight is @xmath321 with @xmath322 ( dashed lines from top to bottom ) . for a comparison , the result of the original leaf - removal algorithm
is represented by the full line : for @xmath59 , a minimal vc is found , whereas the algorithm fails completely to construct a vc for @xmath323.,width=566 ] | in this paper , the dynamics of heuristic algorithms for constructing small vertex covers ( or independent sets ) of finite - connectivity random graphs is analysed . in every algorithmic step ,
a vertex is chosen with respect to its vertex degree .
this vertex , and some environment of it , is covered and removed from the graph .
this graph reduction process can be described as a markovian dynamics in the space of random graphs of arbitrary degree distribution .
we discuss some solvable cases , including algorithms already analysed using different techniques , and develop approximation schemes for more complicated cases .
the approximations are corroborated by numerical simulations . |
planetary nebulae ( pne ) are of great importance in galaxy studies as _ unique _ observable proxies for stars in regions of low surface brightness .
they can be used to infer the spatial , kinematic , and chemical properties of the underlying stellar population , at distances of up to 100 mpc ( cf .
@xmath01 mpc for resolved stellar spectroscopy ) . here
i review the use of pn kinematics to probe mass distributions in galaxies , with particular attention to disk masses of spiral galaxies ( [ sec : disk ] ) , and halo masses of elliptical galaxies ( [ sec : ell ] ) .
pn kinematics are also useful for studying angular momentum , orbit properties , and substructure in galaxies topics beyond the scope of this review . in all such dynamical studies ,
it is important to know _
what is _ the underlying stellar population traced by the observed pne in external galaxies .
these ( bright ) pne are frequently assumed to originate with an `` old '' stellar population that comprises the bulk of the stellar mass in the galaxy .
this assumption has a great deal of theoretical uncertainty ( _ e.g. _ , ( * ? ? ?
* marigo 2004 ) ; ( * ? ? ?
* ciardullo 2005 ) ; ( * ? ? ?
* buzzoni 2006 ) ) , but empirically , it works remarkably well when pn numbers are compared to stellar photometry .
i will return to this issue later , especially in [ sec : iss ] .
a summary is in [ sec : summ ] .
there are unique pros and cons to studying pne in the milky way vs. extragalactically .
probably the biggest dynamical challenge is the pn distance uncertainty .
pn rotation curves have been derived to 14 kpc ( _ e.g. _ , ( * ? ? ?
* schneider & terzian 1983 ) ; ( * ? ? ?
* maciel & lago 2005 ) ) , supporting the presence of a dark halo .
but it may be more useful to turn the analysis around and use other mass constraints to derive the pn distance scale ( ( * ? ? ?
* phillips 2001 ) ) ; or to use pne to study the dynamical structure of the galaxy ( _ e.g. _ , ( * ? ? ?
* beaulieu et al . 2000 ) ) .
pn kinematics in spiral galaxies outside the milky way have not been studied much ( relative to elliptical galaxies ) because of possible contamination from other emission - line regions , and because of the availability of alternative mass tracers . however , there are some recent developments in this area , based upon observations summarized in table [ tab : tab1 ] .
.selected observations of planetary nebula velocities in external galaxies [ cols="<,^,^,^,^,>,^ " , ] the standard observational technique for extragalactic pn kinematics involves an imaging survey with `` off - band '' and `` on - band '' images , where `` on - band '' usually means a narrow - band filter centered around the 5007 line .
velocities are acquired by follow - up with a multi - object spectrograph .
some examples of this are wide - field imaging surveys of m33 and m83 , followed by fiber spectroscopy with velocities accurate to @xmath1 km s@xmath2 ( ( * ? ? ?
* ciardullo 2004 ) ; ( * ? ? ?
* herrmann & ciardullo 2006 ) ) .
another technique is to use a type of slitless spectroscopy called _ counter - dispersed imaging _
( cdi ) , in which pn detection and velocity measurement are done all in one step . used occasionally with existing instrumentation ( _ e.g. _ , ( * ? ? ?
* douglas 2000 ) ) , cdi has really come into its own with the custom - built planetary nebula spectrograph ( pn.s ; ( * ? ? ?
* douglas 2002 ) ) .
originally optimized for observation of 5007 only , the pn.s has a new arm for simultaneous imaging , due for commissioning in late 2006 .
this addition allows for improved pn detection reliability , better background contaminant rejection , and the capability of measuring @xmath3()/@xmath3 ( ) line ratios .
a spectacular success of the pn.s has been the wide - field kinematic survey of the local group spiral m31 , with 2615 pn velocities over 7 deg@xmath4 , encompassing much of the bulge , disk , and halo ( ( * ? ? ?
* ; * ? ? ?
* merrett 2003 , 2006 ) ) .
there is also an extensive traditional imaging+fiber survey ( ( * ? ? ?
* halliday 2006 ) ) .
comparing techniques , the pn.s is clearly more efficient , but has slightly worse velocity accuracy , and probably poorer astrometry for facilitating further spectroscopic study .
the most fundamental kinematic property of a spiral galaxy is the rotation curve , or circular velocity profile , which provides the cumulative mass profile via the relation @xmath5 .
it is infeasible to obtain @xmath6 from integrated - light stellar spectroscopy in the nearest spirals because of their large angular extent and rapidly decreasing surface brightness profiles .
with pn kinematics in m31 and m33 , @xmath6 is obtained to 46 disk scale - lengths @xmath7 , and found to agree with the flat rotation curves obtained from and co gas measurements ( modulo the asymmetric drift corrections ; ( * ? ? ?
* ciardullo 2004 ) ; ( * ? ? ? * merrett 2006 ) ) . this appears to rule out magnetic fields as an alternative explanation for the flat rotation curves ( _ e.g. _ , ( * ? ? ?
* battaner & florido 2005 ) ) . although measurements in large samples of spiral galaxies make it clear that @xmath6 is fairly constant at large @xmath8 , implying large discrepancies between the visible and total masses ( ( * ? ? ? * persic 1996 ) ) , there is considerable freedom for decomposing @xmath9 into its subcomponents .
thus there is a strong degeneracy between the mass - to - light ratio @xmath10 used for the stars , and the inferred properties of the dark matter ( dm ) halo ( _ e.g. _ , ( * ? ? ?
* bottema 1997 ) ) .
the degeneracy can be broken by measuring the velocity dispersion of the stellar disk , especially in the face - on case .
this is evident with the isothermal disk approximation , where the vertical velocity dispersion @xmath11 is directly proportional to the disk surface mass density @xmath12 , and the scale - height @xmath13 should be fairly constant : @xmath14 .
such dispersions have been measured in inner disks , but one can not assume that @xmath10 is the same in the outer disk and in these low surface brightness regions , pn kinematics are generally the only accessible stellar tracer . in m33 and m83 ,
the pn dispersion is found to decrease fairly slowly with radius , implying a disk mass scale - length much larger than the photometric @xmath7 ( see fig . [
fig : fig1 ] , left ; ( * ? ? ?
* ciardullo 2004 ) ; ( * ? ? ?
* herrmann & ciardullo 2006 ) ) .
this means that @xmath10 increases markedly with radius ( fig .
[ fig : fig1 ] , right)the opposite to what one expects from inside - out disk formation .
a similar effect is seen in m31 , where warp models are also tried out , but still do nt fit the observations ( ( * ? ? ?
* merrett 2006 ) ) .
the m33 pn measurements imply that the rotation curve is dominated by the disk in the inner parts , apparently ruling out the cuspy dark halo profiles predicted by @xmath15cdm theory .
however , the disk is not massive enough to account for _ all _ the `` missing mass '' , ruling out the cold gas dm theory ( ( * ? ? ?
* pfenniger 1994 ) ) .
another alternative is the mond gravitational theory , but with this , the m33 pn data point to a self - consistency problem ( ( * ? ? ?
* stubbs & garg 2005 ) ) .
it seems the only major model for spiral galaxy mass profiles which is not `` disproved '' is a dm halo with a shallow core but further work is clearly needed .
lenticular galaxies ( s0s ) are still mysterious objects .
it is not known whether they are more akin to spiral galaxies ( resulting from gas stripping and quenching ) , or to ellipticals ( resulting from major galaxy mergers ) .
these two basic structural - formational scenarios can be distinguished via their large - radius kinematics an observational goal we are now beginning to address with the pn.s ( e. noordermeer in prep ) .
elliptical galaxies are notoriously difficult to model , with their complicated internal structures and dearth of suitable kinematic tracers . with pn data
, there is the potential for engaging in global `` three - dimensional '' dynamical studies of ellipticals .
for example , on the 4-m william herschel telescope , the ifus sauron and oasis ( with ao ) can probe a galaxy s central parts from parsec to kpc scales , while the pn.s reaches out to tens of kpc .
such contiguous constraints over four decades in radius are important because of the coupling between triaxial orbits in the halo and the central supermassive black hole though there is an implicit assumption that the pne and stars share the same dynamics ( see [ sec : iss ] )
. early work on pne in elliptical galaxies produced small but tantalizing kinematic samples in their halos ( see table [ tab : tab1 ] ) .
the exception was the nearby ngc 5128 , which remains the best - studied early - type galaxy .
recent improvements in instrumentation and methods have led to an explosion of pn velocities ( again , see table [ tab : tab1 ] ) .
particularly important has been the exploration for the first time of ordinary @xmath16 ( @xmath17 ) ellipticals halos .
the panoply of modern observational techniques includes slitless dispersed imaging with the vlt+fors ( ( * ? ? ?
* mndez 2001 ) ; ( * ? ? ?
* teodorescu 2005 ) ) , whose 535 velocities in ngc 4697 remains by far the largest data - set in an ordinary elliptical .
three ellipticals have been studied with the pn.s as a prelude to a large , systematic sample ( ( * ? ? ?
* romanowsky 2003 ) ) .
pn kinematics have been measured in the giant virgo ellipticals m87 and ngc 4472 ( m49 ) using a `` masked - cdi '' hybrid technique useful given pre - imaging data with uncertain astrometry .
fabry - prot narrow - band scanning has been used successfully on several galaxies ( ( * ? ? ? * sluis & williams 2006))an approach which should become more potent with larger telescopes ( _ e.g. _ , ( * ? ? ? * bershady 2004 ) ) .
the standard mass indicator in ellipticals is the projected velocity dispersion profile @xmath18 , although it is very non - trivial to translate this into @xmath6 . in discussing
observed pn dispersions in galaxies , i will also mention the dispersions of globular clusters ( gcs ) when available , as these are complementary tracers of halo mass . in ngc 5128
, @xmath18 is fairly constant at @xmath19 km s@xmath2 for both the pne and gcs , out to @xmath0 10 @xmath20 ( effective radii ) .
this implies a dark halo , although perhaps not as massive as expected ( ( * ? ? ?
* peng 2004 ) ) . in the brightest cluster galaxy ngc 4472 , the dispersion is also flat for the pne and gcs to @xmath0 10 @xmath20 ( g. bergond in prep ) . in m87 ,
the elevated pn and gcs dispersion profiles reveal the effects of the massive virgo cluster core ( romanowsky & kochanek 2001 ) , but their very different slopes imply in this case a difference in the orbital properties between the stars and the gcs . in ngc 1399 ,
the pne and gcs also show a rising @xmath18 ( ( * ? ? ?
* arnaboldi 1994 ) ; y. schuberth in prep ) . this may be evidence of the fornax cluster core ( ( * ? ? ?
* saglia 2000 ) ) , or of transient heating from a galaxy interaction ( ( * ? ? ?
* napolitano 2002 ) ) .
more remarkable behavior is seen in the five observed ordinary ellipticals , where @xmath21 declines markedly with radius , with some galaxy - to - galaxy scatter in the slope ( fig .
[ fig : fig2 ] , left ) .
this suggests low dm content , as first inferred by @xcite from pn kinematics in ngc 3379 .
ngc 3379 is the only one of these galaxies so far with gc kinematics data ( 49 velocities ; ( * ? ? ?
* puzia 2004 ) ; ( * ? ? ?
* pierce 2006 ) ; ( * ? ? ?
* bergond 2006 ) ) .
the gc halo dispersion declines much more gradually than the pne , but the implication of this is not clear ( see [ sec : halom ] ) . before moving on to direct mass results ,
i consider a number of complications in deriving elliptical mass profiles from pn kinematics .
one is that foreground stars and background emission line sources can masquerade as pne due to low signal - to - noise or to limited wavelength coverage . with current pn.s data , background emitters
are rejected based on resolving the emission in space or wavelength , or on starkly outlying velocities .
the next issue is the well - known mass - anisotropy degeneracy , whereby the types of orbits have an effect on the projected dispersions similar to the impact of the mass distribution .
for example , low @xmath22 can be caused by low mass _ or _ by radially - biased orbits whose motions are largely in the plane of the sky .
the model curves shown in fig .
[ fig : fig2 ] ( left ) are for isotropic galaxies , but if the pn orbits in the halo have radial anisotropy , then the @xmath15cdm prediction could be substantially lowered .
fortunately , this degeneracy can be broken by examining the _ shape _ of the line - of - sight velocity distribution in addition to its _ width _ ( dispersion ) . with integrated - light stellar data
, this shape is characterized by high - order gauss - hermite polynomials , and for discrete pn velocity data , the shape can be inferred statistically .
such analyses can require @xmath0 1000 discrete data points even under the simplifying assumption of sphericity ( ( * ? ? ?
* merritt & saha 1993))but this discouragingly large number is really more relevant for sparse stellar and galaxy clusters . with elliptical galaxies ,
there are additional constraints that drastically reduce the uncertainties : the pne are assumed to be drawn from a well - known spatial distribution , and the stellar kinematics firmly establish the mass and anisotropy characteristics for the central half of the galaxy .
the diffuse stellar component is thus of great importance in exploiting the pn kinematics , which raises the important question of how well do the pne trace the diffuse stellar light ?
one effect may arise from radial metallicity gradients , which are thought to exist in the stellar bodies of ellipticals based on their color gradients .
a metallicity gradient would imply an oxygen gradient , and thus if the halo pne had systematically weaker emission , there would be a bias in the pn detections resulting in lower observed @xmath21 .
however , the evidence so far in nearby ellipticals is that halo metallicities have weakly negative or even _ positive _ gradients ( ( * ? ? ?
* gregg 2004 ) ; ( * ? ? ?
* mndez 2005 ) ; ( * ? ? ?
* rejkuba 2005 ) ; ( * ? ? ?
* walsh 2006 ) ) .
another potential problem is if the pne are `` contaminated '' by a young ( _ e.g. _ @xmath23 gyr ) population .
however , cmd and pn observations find no evidence for young halo populations ( ( * ? ? ?
* gregg 2004 ) ; ( * ? ? ?
* peng 2006 ) ) , and uv fluxes indicate that even in the few early - type galaxies with recent star formation ( @xmath24 gyr ) , this contributes only 12% of the stellar mass ( ( * ? ? ?
* yi 2005 ) ) .
a different approach is to directly test the link between the stars and pne : do they have the same spatial and kinematic distributions in regions of overlap ?
as mentioned earlier , various studies of pne in elliptical galaxies have not turned up any indications of population biases . even in m31 , with its spatially - varying star - formation history , there are no evident population effects such as correlations between kinematics and pn luminosity , or variations in the pn luminosity function with radius . in ngc 3379 ,
the surface density and kinematics of the stars and the pne match up well , although it is difficult to rule out small differences ( see fig .
[ fig : fig2 ] , right ) .
the one clear example of a subpopulation signature is in ngc 4697 , where
an east - west asymmetry of the bright pne implies an unmixed stellar component perhaps corresponding to a recent interaction with another galaxy ( ( * ? ? ?
* sambhus 2006 ) ) .
this feature coincides with the asymmetry of the observational set - up , so confirmation is needed of the result in this galaxy and in others .
dynamical models usually assume equilibrium , which will not hold in the outermost halo regions where crossing times are long .
however , equilibrium should be a good approximation out to @xmath0 10 @xmath20 ( see ( * ? ? ?
* mamon 2006 ) ) .
the final major source of uncertainty is the intrinsic shape of the galaxies . modeling flattened systems as spherical can skew mass inferences , and in particular , a large face - on disk could produce low velocity dispersions in projection .
however , this particular scenario does not work for ngc 821 and ngc 4697 , which are _ near - edge - on _ `` low dm '' galaxies .
having considered all these _ caveats _ , we turn to detailed modeling of ngc 3379 , with extensive stellar kinematics ( ( * ? ? ?
* statler & smecker - hane 1999 ) ) and an initial data set of @xmath0 100 pn velocities ( ( * ? ? ?
* romanowsky 2003 ) ) .
an `` orbit modeling '' method is used which constrains the anisotropy directly from the data , assuming spherical symmetry .
the resulting models do indeed have radial anisotropy in the halo , allowing for more dm than suggested by fig .
[ fig : fig2 ] ( left ) .
the @xmath25-band mass - to - light ratio inside 5 @xmath20 is @xmath26 ( in solar units ) , which is very similar to estimates for the stellar @xmath10 .
this result is not what one would expect from @xmath15cdm theory , but given the considerable uncertainties in the mass constraints and in theoretical predictions , @xmath15cdm can not be ruled out .
the other @xmath16 galaxies have not yet been modeled in such detail , but jeans model comparisons suggest similar results , with the exception of ngc 1344 .
while signs of dm from pn kinematics may be skimpy , virial - radius constraints ( from weak gravitational lensing , satellite dynamics , and galaxy cross - correlations ) do demonstrate the presence of massive halos around ordinary ellipticals .
thus the burning question is not whether ellipticals are `` naked '' galaxies without dm , but _
what is the radial distribution of dm _ that can satisfy all these constraints ? this would provide clues about the physics of galaxy formation and about the nature of the dm itself .
a meta - analysis of the available kinematic constraints suggests that dm halos in ellipticals have low concentrations ( ( * ? ? ?
* napolitano 2005 ) ) , as found in many other studies of late - type galaxies and galaxy clusters . however , these low concentrations may be less of a problem now as the consensus on the cosmological parameter @xmath27 is shifting toward much lower values .
perhaps this points toward the use of pne as precision cosmology tools .
@xcite argue that far from being a problem for @xmath15cdm , the pn data are a natural outcome . simulating the formation of elliptical galaxies via spiral mergers , they find familiarly declining @xmath28 profiles .
the physics of the simulations may not adequately represent the formation of galaxies in @xmath15cdm , but in any case , the study highlights three major effects that could contribute to the declines .
the first is varying radial anisotropy , which indeed is included in our direct modeling of ngc 3379 .
the second is that the triaxial structure of the galaxies causes variations in @xmath21 with the viewing angles .
the third is that _ if _ the observed pne stem from a young stellar population formed in the merger , then they are not a fair tracer of the bulk of the stars ( see [ sec : iss ] ) , and produce a steeper decline in @xmath21 . more definitive statements on pn - based mass profiles obviously await experiments with triaxial models ( _ e.g. _ , ( * ? ? ?
* de lorenzi 2006 ) ) .
the large pn.s program sample will also be able to average out viewing - angle effects .
of further use are independent mass constraints , such as gcs and x - ray emission .
as mentioned , in ngc 3379 , @xmath18 is very different for the pne and gcs , which can be largely explained through different spatial and anisotropy characteristics .
the gcs do suggest a more massive halo , but this appears to be contradicted by the kinematics of a large - radius gas ring implying a remarkably low - mass halo ( ( * ? ? ?
* schneider 1985 ) ) .
planetary nebula are powerful tools for exploring galaxy mass profiles , and with the advent of new instrumentation , the sample of galaxies with pn kinematics is now exploding . in spiral galaxies ,
fairly _ constant _ pn velocity dispersions imply surprisingly large disk mass scale - lengths . in ordinary ellipticals , _ declining _ pn dispersions suggest low - concentration dark matter halos .
an important issue still needing clarification for dynamical models is the connection between pne and their parent stellar population . | planetary nebula are useful kinematic tracers of the stars in all galaxy types .
i review recent observationally - driven developments in the study of galaxy mass profiles .
these have yielded surprising results on spiral galaxy disk masses and elliptical galaxy halo masses .
a key remaining question is the coupling between pne and the underlying stellar populations . |
it was observed by douglas and hull @xcite that d - branes on @xmath1 with a constant neveu - schwarz ( ns ) two - form potential @xmath2 give rise to an effective world - volume theory on a non - commutative torus .
even though this initial observation was re - considered and generalized by many authors @xcite , all the subsequent work is restricted to flat backgrounds . a perturbative analysis along the lines of @xcite
, on the other hand , shows that the quantization of world - volume geometries should be a much more general phenomenon which persists in the case of curved backgrounds . in this work we shall present the first non - perturbative ( in @xmath3 ) investigation of world - volume geometries in a curved string background with non - vanishing ns 3-form field @xmath0 . by the string s equation of motion . ]
an exact treatment of d - branes in curved backgrounds is possible within the framework of boundary conformal field theory .
here we illustrate the basic techniques and some general features of the resulting world - volume geometries in a particular example , namely the su(2 ) wzw theory , and study d - branes in the wzw model associated with the gluing condition @xmath4 .
we shall argue that their world - volumes may be regarded as fuzzy two - spheres when the level @xmath5 is sent to infinity , i.e. when the background becomes flat . for finite level , @xmath6 is non - zero and we shall find non - associative deformations of these fuzzy spheres , which are closely linked to the theory of quantum groups . while the infinite level result can be predicted from the semi - classical analysis in @xcite together with the general phenomenon of world - volume quantization in flat backgrounds @xcite , our results on the finite level provide a non - trivial extension of the standard rules .
apparently , many features of the world - volume geometry are not captured by the perturbative treatment of d - branes on group manifolds that was suggested recently in @xcite .
we shall follow a general procedure which allows us to extract world - volume geometry from the world - sheet description of any ( generalized ) d - brane , even when it is given in purely algebraic terms .
the essential input data are the operator product expansions ( ope ) of boundary fields ( open string vertex operators ) .
since they depend on the ordering of the operators , it is not surprising that the brane world - volume obtained in this way is a non - commutative space , in general .
we shall see that non - associativity may show up as well .
our approach is inspired by a project initiated by j. frhlich and k. in @xcite ( see also @xcite for earlier ideas in the same direction ) , where it was proposed to construct non - commutative target space geometries from opes of closed string vertex operators .
this was developed further in @xcite .
it appears , however , that non - commutative geometry emerges in a more natural way and on a more fundamental level in the open string case , cf .
the picture below .
our findings add to the growing evidence that brane physics surpasses classical geometry even though the emergence of a non - commutative world - volume need not necessarily mean that a d - brane behaves non - geometrically in the sense of the criterion formulated in @xcite .
this criterion rests on a comparison of low - energy effective field theories in the stringy and in the large - volume regime , and we do not attempt to test it in the present paper .
but we would like to point out that the structures contained in the non - commutative world - volume also form the main ingredient of the effective action of the brane . while we have chosen the su(2)@xmath7 example mainly because of its simplicity and because there exists a semi - classical curved background picture ,
it is also an important ingredient of the cft formulation of the neveu - schwarz 5-brane , see e.g. @xcite . given that questions like stability of the configuration can be clarified , our findings should be relevant for the geometry of d - branes in the presence of a stack of 5-branes .
similarly , our su(2 ) wzw results could be applicable in the study of branes on an ads@xmath8 string background , see e.g. @xcite .
before we show how one can read off fuzzy geometry from branes in the wzw model , let us briefly review the emergence of non - commutative spaces in the more standard case of branes in flat @xmath9-dimensional euclidean space @xmath10 , or on a flat torus @xmath11 . consider a d - brane which is localized along a @xmath12-dimensional hyper - plane @xmath13 in the target , with tangent space @xmath14 .
the conformal field theory associated with such a euclidean d - brane is defined on the upper half of the complex plane .
it contains an @xmath9-component free bosonic field @xmath15 subject to neumann boundary conditions in the directions along @xmath14 and dirichlet boundary conditions for components perpendicular to the world - volume of the brane . from the free bosons ,
one may obtain various new fields , in particular the open string vertex operators @xmath16 which can be inserted at any point @xmath17 on the real line . when there is no magnetic field on the brane , the ope of these u(1)-primaries reads ( with @xmath18 and for @xmath19 )
v_k_1(x_1 ) v_k_2 ( x_2 )
= ( x_1-x_2)^k_1k_2/2 v_k_1 + k_2(x_2 ) + , [ opeb0 ] where the dots indicate less singular non - primary contributions .
we can rewrite this relation by introducing the objects @xmath20(x ) \ : = \ \frac{1}{(2\pi)^{p/2 } } \int_{tv_p } d^pk \ \hat
f(k ) \ v_k(x)\ ] ] for each function @xmath21 with fourier transform @xmath22
. then the boundary ope ( [ opeb0 ] ) translates into a `` definition '' of pointwise multiplication of functions , v[f](1 ) v[g](0 ) = v[fg ] ( 0 ) + .
[ pointw]we have specialized to coordinates @xmath23 and @xmath24 for convenience , arbitrary insertion points can be recovered via conformal covariance .
the effect of switching on a @xmath2-field is described by adding the term s_b
= dz d|z b _
x^(z,|z ) |x^(z,|z ) [ bterm]to the action of the original theory without @xmath2-field .
one can easily see that this is a pure boundary term with no influence on the bulk properties of the theory .
it only changes the boundary conditions .
if we assume for definiteness that @xmath13 is spanned by the first @xmath12 coordinates @xmath25 , the new boundary conditions read ( with @xmath26 ) _ y x^(z,|z ) = b^ _ _ x x^(z , |z ) z = |z , = 1 , , p .
[ bglue ] this means that the ( exact ) free boson propagator becomes ( @xmath27 ) x^(x_1)x^(x_2)_b = - ( ^ + _ s^ ) |x_1-x_2| - i _
a^(x_1-x_2 ) [ xcorr]where @xmath28 and @xmath29 denote the symmetric resp . anti - symmetric part of the matrix @xmath30 . explicitly , _
[ magic]in particular , when @xmath2 is large we obtain @xmath31 , which means that @xmath29 is the poisson bi - vector corresponding to the symplectic form @xmath2 .
( [ xcorr ] ) immediately yields the boundary ope for a non - vanishing @xmath2-field , @xmath32 as before , this can be used to define a ( deformed ) product @xmath33 for functions through @xmath34(1 ) v[g](0 ) = v[f \star g ] ( 0 ) + \ldots\ ; , $ ] where now ( f g ) ( x ) : = e^i ^_a ^x_^y _
f(x ) g(y ) _ y = x .
[ moyal]this is the associative , non - commutative moyal - weyl product of functions @xmath35 on the world - volume @xmath13 of the brane . in the context of the derivation we have given ,
non - commutativity of @xmath33 arises because the ordering of boundary fields in general does matter , cf . the sign - term in eq .
( [ xcorr ] ) .
the algebra of functions with product ( [ moyal ] ) is , of course , the non - commutative brane world - volume uncovered by douglas and hull using a different approach .
it is a deformation of the ordinary algebra of functions , with deformation parameter(s ) given by ( the matrix ) @xmath29 . in @xcite ,
the term ( [ bterm ] ) was viewed as a bulk perturbation of the @xmath36 theory , i.e. techniques of conformal perturbation theory were applied to the operator @xmath37 being inserted into arbitrary correlation functions of the @xmath36 theory .
this perturbative analysis , which can be extended to arbitrary @xmath38-models ( at least in the case @xmath39 ) , leads to a string theoretic picture of kontsevich s quantization of poisson manifolds @xcite , see also the work of cattaneo and felder @xcite .
it clearly displays that the quantization of world - volume geometries should be expected beyond the case of constant @xmath2-fields .
this will be confirmed through our exact analysis of the wzw model ( see discussion of the limit @xmath40 below ) . as we remarked in the introduction
, new phenomena are bound to occur when @xmath41 does not vanish . in such cases ,
the classical world - volume of a brane comes equipped with some generalization of an ordinary poisson - structure , and there exists no general notion of `` quantization '' for such geometries .
hence , the investigation of branes in a non - vanishing ns 3-form field strength @xmath0 can teach us new lessons on how to quantize certain non - poisson geometries . in our example of branes on su(2 ) we shall recover some variants of well - known quantum group algebras .
our formulation of the simple example of flat branes in a constant @xmath2-field motivates the following _ general procedure _ :
when we want to associate non - commutative spaces to branes which are given as boundary conditions on the world - sheet , we take the ope of boundary fields ( open string vertex operators corresponding to internal excitations of the brane ) as a basic input .
then we choose a suitable subset of boundary fields ( e.g. primaries as above ) and use them as abstract generators of an algebra of `` functions '' on the ( non - commutative ) world - volume of the brane , with multiplication table given by the boundary ope ( projected onto the subset , and evaluated at @xmath23 and @xmath24 , say ) .
further comments on this general prescription will be given later , but now we would like to test it in the case of su(2 ) wzw models , where the semi - classical picture provides certain expectations as to how the `` quantized world - volume '' of branes should look like .
the su(2 ) wzw model at level @xmath5 describes strings moving on a three - sphere @xmath42 of radius @xmath43 , which is equipped with a constant ns 3-form field strength @xmath44 where @xmath45 denotes the usual volume form on the unit sphere , and @xmath46 are components of the 1-form @xmath47 . in superstring theory
, this geometry appears in the space transverse to a stack of @xmath5 ns 5-branes . these branes act as sources for @xmath5 units of ns 3-form flux through a three - sphere surrounding their ( 5 + 1)-dimensional world - volume .
the world - sheet swept out by an open string in @xmath42 is parametrized by a map @xmath48 from the upper half - plane h into the group manifold su(2)@xmath49 . from this field @xmath50 one obtains lie algebra valued chiral currents @xmath51 as usual .
we shall be interested in maximally symmetric d - branes on su(2 ) , which are characterized by the gluing condition @xmath52 along the boundary @xmath53 .
they were analyzed from a semi - classical point of view in @xcite , and we shall briefly recall the findings of this approach .
( for a detailed path integral description of branes in su(2 ) , see @xcite . )
we first decompose the tangent space @xmath54 at each point @xmath55su(2 ) into a part @xmath56 tangential to the conjugacy class through @xmath57 and its orthogonal complement @xmath58 ( with respect to the killing form ) . in @xcite ,
the following two basic observations were made : 1 . with gluing conditions of the type @xmath59 , the endpoints of open strings on su(2 )
are confined to conjugacy classes , i.e. @xmath60 2 . along the individual branes , i.e. along the conjugacy classes of su(2 ) , the gluing condition becomes @xmath61 except for two degenerate cases , namely the points @xmath62 and @xmath63 on the group manifold , the conjugacy classes are two - spheres in su(2 ) .
taking into account the usual correspondence between @xmath64 and the flat space coordinate @xmath65 , recalling that the metric on the three - sphere scales with @xmath5 , and comparing with the gluing conditions ( [ bglue ] ) , we infer that the d - branes associated with @xmath59 carry a non - vanishing 2-form potential ( b - field ) @xmath66 in the limit @xmath67 the three - sphere grows and approaches flat 3-space .
one can parameterize it by a parameter @xmath68 taking values in the lie algebra su(2 ) , such that @xmath69 . then
, the formula for the @xmath2-field reads @xmath70 this is the kirillov 2-form on the spheres in the algebra su(2)@xmath71 .
extrapolating formula ( [ magic ] ) to our curved background , we can construct a bi - vector @xmath72 introducing an orthonormal basis @xmath73 in su(2 ) , and the left- and right - invariant vector fields @xmath74 on the group manifold , one can give an elegant formula for the bi - vector @xmath29 , @xmath75 the schouten bracket of @xmath29 ( which generally characterizes the deviation from the jacobi identity ) is of the form @xmath76\ = \ \frac{1}{6}\ ; f_{abc}\ ; ( e^a_l- e^a_r)(e^b_l - e^a_r ) ( e^c_l - e^c_r)\ .\ ] ] here @xmath77 are the lie algebra structure constants , the same as those in the expression for the field strength @xmath6 .
this calculation makes sense for an arbitrary simple lie group . in general , the right hand side does not vanish and gives the obstruction for the jacobi identity . in the case of @xmath78su(2 ) , @xmath79 vanishes for dimensional reasons : it is a 3-vector tangent to the 2-dimensional conjugacy classes . in the infinite volume limit @xmath67
, the bi - vector @xmath29 becomes @xmath80 which is the kirillov - kostant poisson bi - vector .
consequently , the geometry of the limiting theory @xmath81 is very close to the well - known situation of flat branes in a flat background with constant @xmath2-field , and we expect that the world - volume algebras of our branes in the wzw model will be quantizations of two - spheres . for finite @xmath5 , however , the background is curved and carries a non - vanishing ns 3-form @xmath6 .
this will result in a non - associative deformation of the @xmath82 theory .
since the three indices of the new object @xmath6 can relate three - fold products with different positions of brackets , the violation of associativity will turn out to be rather mild .
the semi - classical extension of the above analysis shows that , for fixed gluing conditions , only a finite number of su(2 ) conjugacy classes satisfy a dirac - type flux quantization condition @xcite .
these `` integer '' conjugacy classes are the two points @xmath62 and @xmath63 along with @xmath83 of the spherical conjugacy classes ( those passing through the points @xmath84 for @xmath85 ) . the wzw model on the upper half - plane
is known in enough detail to support and specify the rather crude arguments of the previous subsection by an exact cft analysis .
in fact , for the situation we are dealing with ( gluing conditions @xmath59 in a `` parent '' cft on the full complex plane with diagonal modular invariant partition function ) , cardy @xcite was able to list all @xcite possible boundary conditions .
there exist @xmath86 of them , differing in the bulk field one - point functions ( brane charges ) and labeled by an index @xmath87 . without entering a detailed description of these boundary theories @xcite , we recall that their state spaces have the form [ partdec ] _
= _ j n_^j ^j where @xmath88 , @xmath89 , denote irreducible highest weight representations of the affine lie algebra @xmath90 , and where @xmath91 are the associated fusion rules .
note that only integer spins @xmath92 appear on the right hand side of ( [ partdec ] ) .
there exists a variant of the state - field correspondence which assigns a boundary field @xmath93 to each element @xmath94 ( see e.g. @xcite ) .
in particular , the su(2 ) wzw boundary theory labeled by @xmath95 contains su(2)-multiplets associated to primary boundary fields , namely @xmath96 and @xmath97 .
all these boundary fields are defined for arguments @xmath17 on the real line and their correlators have , in general , no unique analytic continuation into the upper half - plane . in the flat target case
, we chose u(1)-primaries as generating elements of the world - volume algebra .
now , it is more appropriate not to break the group symmetry by hand and , therefore , to keep the full su(2)-multiplets @xmath98 . for a fixed order @xmath99 of arguments on the real line ,
the ope of two such boundary fields reads [ boundope ] ^i_i(x ) ^j_j(y ) ~ _ k , k ( x - y)^h_i+h_j - h_k lll i & j & k + i & j & k c^,_ijk
^k_k(y ) , where @xmath100 is the conformal dimension of @xmath101 and @xmath102 $ ] denote the clebsch - gordan coefficients of the group su(2 ) .
the latter simply compensate for the different transformation behavior of the fields on the left and right hand side under the action of the zero - mode subalgebra of @xmath103 .
hence , the non - trivial information in ( [ boundope ] ) is contained in the new structure constants @xmath104 . in a consistent theory
, these must obey sewing constraints , which were first analyzed by lewellen in @xcite ; see also @xcite .
recently , these constraints were reconsidered by runkel @xcite for the a - series of virasoro minimal models .
his findings carry over to su(2 ) wzw models on the upper half - plane and show that the only possible solution to the sewing constraints is given by the fusing matrix @xmath105 of the wzw theory , [ cval1 ] c^,_ijk = f_k ll & + i & j _ .
it is one of the fundamental results on the relation between quantum groups and conformal field theory ( see e.g. @xcite ) that the fusing matrix of the wzw model is obtained from the @xmath106 symbols of the quantum group algebra @xmath107 according to [ cval2 ] f_k ll & + i & j _
= lll i & j & k + & & _ q q = e^ . in the limit @xmath108 , the @xmath106 symbols of the quantum group algebra approach those of the classical algebra @xmath109 , thus the structure constants @xmath110 of the boundary ope become @xmath106 symbols of the group su(2 ) when the level @xmath5 is sent to infinity . note that in this limit , the conformal dimensions @xmath111 tend to zero so that the opes ( [ boundope ] ) of boundary fields become regular as in a topological theory .
we are now prepared to follow the procedure sketched at the end of section 2 and to read off the world - volume geometry of branes in the su(2)-wzw model .
so let us think of the boundary fields @xmath112 as being assigned to elements @xmath113 of some vector space , and let us use the operator product expansion ( [ boundope],[cval1],[cval2 ] ) to define a multiplication by the prescription [ fsopek ] y^i_i y^j_j = _ k , k lll i & j & k + i & j & k c^,_ijk y^k_k . as in ( [ boundope ] ) , the summation on the right hand side runs from @xmath114 to a maximal spin @xmath115 .
first , we shall investigate this product in the limiting case @xmath116 , where it produces a familiar algebraic structure .
passing to finite levels leads to the following two changes : there is a @xmath5-dependent deformation of structure constants @xmath117 , cf .
( [ cval2 ] ) , and the range of the summation in ( [ fsopek ] ) becomes a function of the level , @xmath118 .
we shall separate these two phenomena by looking at an intermediate case where @xmath5 is non - rational and where we omit the @xmath5-dependent restriction on the @xmath119-summation . _
infinite level @xmath81_:recall that , in the case of infinite level , the structure constants @xmath117 in eq .
( [ fsopek ] ) are given by the @xmath106 symbols of the group su(2 ) .
the semi - classical analysis showed that @xmath120 , so we expect the world - volume algebra to be associative .
indeed this can be confirmed using the biedenharn - elliot ( or pentagon ) relation for the @xmath106 symbols , along with the fact that @xmath106 symbols of the form ( [ cval2 ] ) vanish whenever @xmath121 .
hence , for infinite level our relations define an infinite set of associative algebras @xmath122 with finite linear bases consisting of dim@xmath123 elements . since the dimension of each of these algebras is a perfect square , one may already suspect that they are full matrix algebras , i.e. that @xmath124 with @xmath125 . to describe the isomorphism , we first note that @xmath126 admits an action of the group su(2 ) by conjugation with group elements evaluated in the @xmath127-dimensional representation of su(2 ) . under this action , the su(2)-module @xmath126 decomposes into a direct sum of irreducible representations @xmath128 , [ fsdec ] m_n ( ) _ j=0^n-1 v^j .
only integer @xmath92 appear , so this agrees with the decomposition of the state space @xmath129 in eq .
( [ partdec ] ) for boundary wzw models at sufficiently large ( or infinite ) level @xmath5 .
thus , we can identify our elements @xmath130 with a basis of the spaces @xmath128 .
the isomorphism ( [ fsdec ] ) allows to work out multiplication rules for any two such basis elements from the multiplication of @xmath131-matrices .
the result @xcite turns out to coincide with our formula ( [ fsopek ] ) , which shows that @xmath132 and @xmath133 are indeed isomorphic as associative algebras .
the non - commutative spaces @xmath132 are known as _ fuzzy spheres _ and are obtained when one quantizes functions on a two - sphere with the usual poisson structure ( see e.g. @xcite and references therein ) .
the two - spheres may also be identified with co - adjoint orbits of su(2 ) . according to kirillov
, their quantization gives all representations of the lie algebra su(2 ) or of its universal enveloping algebra @xmath134 .
note that the size @xmath135 of our matrices agrees with the number of components for an su(2)-multiplet of spin @xmath95 .
hence , through the investigation of maximally symmetric branes on su(2 ) at @xmath81 , we have recovered kirillov s theory of co - adjoint orbits .
_ finite non - rational level @xmath5_:let us stress that this case does not appear among the exact boundary theories above ( for non - compact wzw models , it is the generic situation ) .
we include it here merely as an intermediate step before presenting the structure for finite integer level @xmath5 . to be more precise
, we consider the algebras spanned by @xmath130 with relations ( [ fsopek ] ) in which the structure constants @xmath117 are given by the @xmath106 symbols ( [ cval2 ] ) of the quantum group algebra @xmath136 , but with summation over the same range as in the case @xmath81 .
the resulting algebras @xmath137 with @xmath138 not a root of unity cease to be associative .
but they are still quasi - associative in the sense that y^i_i ( y^j_j y^k_k ) ( ^i_in ^j_jm ^k_kl ) ( ) = ( y^i_n y^j_m ) y^k_l [ qassoc ] where the @xmath139 denote representations of @xmath140 and where @xmath141 is drinfeld s `` re - associator '' @xcite .
the proof of this statement is sketched in the appendix . when we perform a standard quasi - classical limit ,
commutators are replaced by the brackets corresponding to the bi - vector @xmath29 . for a general compact simple lie
group @xmath29 fails to satisfy the jacobi identity .
this corresponds to the leading non - vanishing term in the @xmath142-expansion of the re - associator @xmath143 , @xmath144 where @xmath73 is , as above , an orthonormal basis in the lie algebra , and @xmath77 are the corresponding structure constants . when applied to the relation ( [ qassoc ] ) , the lie algebra generators @xmath73 act by the adjoint vector fields @xmath145 . in the case of @xmath78su(2 ) this leads to vanishing of the first order correction to the associativity law .
this is in accordance with vanishing of @xmath146 $ ] in this case .
note that even in the su(2 ) case higher order corrections to the associativity law do not vanish .
let us briefly mention that our quasi - associative algebras @xmath137 are closely connected to associative deformations of the fuzzy sphere which employ the clebsch - gordan coefficients of the deformed @xmath136 instead of their classical analogs .
some details on these algebras and their associativity can be found in the appendix . for now ,
let us only remark that they are factors of the quantum spheres introduced by podle in @xcite .
their relation to our algebras @xmath137 is based on the fact that one can obtain the clebsch - gordon maps of classical lie algebras from their @xmath147-deformed counterparts with the help of drinfeld s `` twist element '' @xmath148 .
the latter provides the following factorization formula for the re - associator : @xmath149 where @xmath150 denotes the co - product of @xmath140 . combining these two roles of the twist element @xmath105
, one can show that our algebras @xmath137 are `` twist equivalent '' to associative factors of a podle sphere or , more explicitly , to the same matrix algebras @xmath151 as in the case of infinite level .
hence , we simply recover the representations for the usual @xmath147-deformation of @xmath140 at generic values of the deformation parameter . _
finite integer level @xmath5_:the associated algebras @xmath152 are spanned by the generators @xmath153 with the label @xmath92 chosen from the set @xmath154 .
multiplication of these elements is defined through eq .
( [ fsopek ] ) with structure constants @xmath117 now given by the @xmath106 symbols of @xmath136 at the root of unity @xmath155 .
in addition , the summation on the right hand side is now restricted to run from @xmath114 to @xmath156 . viewed as su(2)-modules ,
the linear spaces @xmath152 decompose as follows : @xmath157 s^2_{\ik/2 - \a } & \ \mbox { for } \
\ \frac{\ik}{4 } \leq \a \leq \frac{\ik}{2 } \end{array } \right .
\ \ .\ ] ] again , the algebras @xmath152 are only quasi - associative , and they provide examples of the geometries considered in @xcite . using the concept of representations introduced in @xcite , it is not difficult to show that each of the quasi - associative algebras @xmath158 possesses precisely one indecomposable representation on a vector space @xmath159 of dimension @xmath160 \ik - 2\a + 1 & \
\mbox { for } \ \ \frac{\ik}{4 } \leq \a \leq \frac{\ik}{2 } \end{array } \right . \
\ .\ ] ] according to our previous discussion , the algebras @xmath158 and their representations on @xmath161 , generalize kirillov s theory of co - adjoint orbits to quantum groups at roots of unity .
in other words , the algebras @xmath158 we obtain are `` quantizations '' of integer conjugacy classes on su(2 ) .
summing over all possible brane sectors , i.e. over the index @xmath162 , we construct a deformed universal enveloping algebra . of course
, quantum group algebras were constructed within the framework of chiral conformal field theory before , see e.g. @xcite .
as long as we avoid roots of unity , our new derivation from boundary conformal field theory reproduces well - known algebraic structures .
differences between the two approaches occur only when @xmath147 is a root of unity . in that case ,
boundary conformal field theory improves upon the old constructions in two respects .
first of all , the theory gives `` physical '' representations exclusively so that there is no need for additional truncations .
furthermore , the dimensions @xmath163 of the representation spaces are invariant under the simple current symmetry which interchanges @xmath95 and @xmath164 .
when we increase the level @xmath5 , the radius of the three - sphere grows and we can fit more and more branes into the background . at the same time , the 3-form field strength decreases and the world - volume algebras become `` more associative '' while their non - commutativity survives .
this is to be compared to the non - commutative targets obtained in @xcite from closed strings : the @xmath67 limit of these targets is simply the classical group su(2 ) .
the different behavior of closed and open string geometry may be explained as follows : both closed and open strings feel the presence of the ns 3-form field @xmath6 at finite level .
open strings are also sensitive to the concrete choice of a 2-form potential @xmath2 , while closed strings `` see '' only its cohomology class . in the flat space limit @xmath81 ,
the cohomology becomes trivial while @xmath2 itself stays non - zero and is responsible for non - commutativity on the brane .
we have derived non - commutative world - volume algebras for d - branes in the su(2 ) wzw model , using a general scheme that can be applied to arbitrary branes given as conformal boundary conditions , including supersymmetric cases . in the process , we have seen how abstract objects from the cft description , like cardy s boundary states and runkel s ope coefficients , acquire a geometrical meaning if in terms of non - commutative ( and sometimes non - associative ) spaces .
the su(2 ) wzw model provides just the simplest example of a string background with a non - vanishing 3-form field strength @xmath6 , but we think that it illustrates quite nicely much of the behavior one should expect from more complicated backgrounds . in particular , the discussion of su(2 ) branes carries over to boundary wzw models with other structure groups g ( at least in the compact case ) and leads to a quantization of integer conjugacy classes in g. it might be interesting to investigate also branes that are not maximally symmetric , i.e. where the gluing conditions respect only a subalgebra of the maximal chiral symmetry algebra @xcite .
boundary cft yields world - volumes independently of whether limiting classical pictures are available or not , and it actually provides more structure than a mere set of non - commutative algebras .
connes program @xcite shows that , in order to talk about the geometry of a non - commutative space , it is necessary to fix further `` spectral data '' , including a hilbert space on which the ( associative ) world - volume algebra and a generalized dirac or laplace operator act .
how these data can be extracted from a cft has been discussed , for the bulk case , in @xcite .
the importance of the laplace operator , which is related to the conformal hamiltonian @xmath165 , can also be seen in the context of our definition of non - commutative world - volumes : in order to re - derive the ope of boundary operators from the algebraic structure of the world - volume , the spectrum of conformal dimensions must be known , cf . the remark after eq .
( [ pointw ] ) . in a cft on the upper half - plane ,
additional structure is available , e.g. in the form of boundary condition changing operators which induce transitions between two different boundary conditions @xmath166 .
the ope of the boundary fields @xmath167 with boundary condition changing operators gives rise to bi - modules @xmath168 over the world - volume algebras of the two associated branes . in the case of d
- branes on a group manifold , these bi - modules allow to construct tensor products for representations of the associated quantum group .
opes involving two boundary condition changing operators provide even more data , namely a full braided tensor category . some comments on our general scheme to extract a world - volume algebras from the boundary cft description of branes are in order .
it involves a choice of `` generating elements '' among the boundary fields . from a pure cft perspective
, one could restrict to primary operators only , or one could work with all boundary operators and thus with an infinite - dimensional world - volume . in a sense , the latter algebra would include all internal excitations of the `` static '' space defined using primary fields .
the wzw case , where it proved natural to keep the full group multiplets associated with primary boundary fields , suggests that there are distinguished `` intermediate '' choices . for a large class of cfts ,
the appropriate generalization of the lowest - dimension spaces of wzw models is likely to be given by the special subspaces introduced in @xcite ; see also @xcite . placing the cft into a string theory context
can remove the arbitrariness and provide clear guidelines as to which world - volume generators to select from the boundary fields : string theory contains additional parameters like @xmath3 , and the relevant generators of the world - volume algebra are those surviving in some limiting regime .
e.g. in the flat background case , one can remove all higher excitations by sending @xmath3 to zero while keeping the @xmath2-field finite ; see @xcite and also @xcite .
it may be possible that a number of interesting limits exists ; then one expects that the world - volume of a brane can look very different in different regimes , and that full string theory can `` interpolate '' between those geometries .
the next task would be to calculate the effective action on the in general non - commutative world - volume of the brane .
the lowest - order terms are , of course , already given by our `` multiplication table '' ( the ope coefficients ) . in principle , higher - order contributions can be computed from the same data , but in practice one still needs to integrate over world - sheet moduli . in the context of the douglas - hull model ,
the effective field theories were found to be non - commutative supersymmetric gauge theories with some amount of non - locality @xcite .
seiberg and witten could show that these models are equivalent to ordinary gauge theories on a flat brane @xcite .
it remains to be seen whether classical structures are stretched further when more general cft backgrounds are taken as a starting point .
perhaps it is worthwhile to compare the induced field theories with existing models on fuzzy geometries ( see e.g. @xcite )
. it would also be interesting to investigate further the relation between world - volume non - commutativity as introduced in @xcite and non - commuting moduli as discovered by witten @xcite .
both phenomena can be traced back to failures in locality properties of boundary fields
see @xcite for the case of moduli so that there exists a direct connection between the brane s intrinsic `` fuzziness '' and the way it `` perceives '' its ambient target .
+ * acknowledgements : * we would like to thank i. brunner , c. chu , r. dijkgraaf , m. douglas , j. frhlich , j. fuchs k. , o. grandjean , p. ho , j. hoppe , c. klimk , n. landsman , g. moore , a. polychronakos , g. reiter , a. schwarz , c. schweigert , s. shatashvili , i.t .
todorov and especially j. teschner for useful and stimulating discussions .
is grateful to the daad for support and to the aei potsdam for hospitality .
* note added : * after this work was completed , another approach to the geometry of branes in wzw models based on exact cft methods was presented in @xcite .
here we collect some basic material on clebsch - gordan maps , @xmath106-symbols and the ( quasi-)associativity of various algebras mentioned in the main text .
let us denote by @xmath169 the irreducible representation of @xmath136 with spin @xmath170 . by definition ,
clebsch - gordan maps @xmath171 intertwine between the actions of @xmath136 on the product module @xmath172 and the irreducible module @xmath173 .
@xmath106 symbols enter the theory through the basic relation c_q(mk|l ) ( c_q(ij|m ) ^k ) = _ p _ q c_q(ip|l ) ( ^i c_q(jk|p ) ) .[eqa1]they obey a number of fundamental equations . for our purposes , the biedenharn - elliot ( pentagon ) relation is the most important one . with the spin labels set to the values that we need below , it implies _ m _
q y^k_k . [ qfuzzy ] the clebsch - gordan coefficients on the right hand side are obtained from the maps @xmath177 once we have selected a basis in each representation space @xmath178 .
associativity of this algebra is rather easy to prove with the help of eqs .
( [ eqa1 ] ) and ( [ eqa2 ] ) : ( y^i_i y^j_j ) y^k_k & = & _ l , l , m , m _ q y^l_l + & = & y^i_i ( y^j_j y^k_k ) for the special case @xmath174 this computation proves the associativity of the world - volume algebra in the limit @xmath116 . when the level @xmath5 is finite and non - rational , however , the defining relation for our algebra @xmath137 from sect . 4 employs the _ undeformed _ clebsch - gordan maps along with the deformed @xmath106 symbols . hence ,
using relation ( [ eqa1 ] ) for @xmath174 , we generate an undeformed @xmath106 symbol in our computation above .
the latter can not be absorbed with the help of the pentagon identity , since we have to deal with a product of one undeformed and two deformed @xmath106 symbols . _
q c(ip|l)(^ic(jk|p ) ) + ( ^-1 ) ^ijk = ( ^i ^j ^k ) & & ( ^-1 ) : v^i v^j v^k v^i v^j v^k . note that this relation involves clebsch - gordan maps of the lie algebra and @xmath147-deformed @xmath106-symbols at the same time .
@xmath143 allows to modify the proof we have given for the associativity of the algebra ( [ qfuzzy ] ) such that we obtain the quasi - associativity property ( [ qassoc ] ) .
a relation between our quasi - associative algebra @xmath137 and the associative @xmath147-deformation of the fuzzy sphere can be established with the help of drinfeld s twist element @xmath105 . by definition
, it maps the deformed and undeformed clebsch gordan maps onto each other , @xmath180 this property becomes crucial in showing that the quasi - associative algebra for non - rational @xmath5 is `` twist - equivalent '' to the associative @xmath147-deformed fuzzy sphere .
some details on the notion of twist equivalence can be found e.g. in section 7.3 of @xcite .
douglas , c. hull , _ d - branes and the noncommutative torus _ , j. high energy phys .
9802 ( 1998 ) 008 , hep - th/9711165 y .- k.e .
cheung , m. krogh , _ noncommutative geometry from d0-branes in a background @xmath2-field _ , nucl .
( 1998 ) 185 , hep - th/9803031 + f. ardalan , h. arfaei , m.m .
sheikh - jabbari , _ mixed branes and m(atrix ) theory on noncommutative torus _ , hep - th/9803067 ; , j. high energy phys .
9902 ( 1999 ) 016 , hep - th/9810072 + h. garca - compen , _ on the deformation quantization description of matrix compactifications _ , nucl .
* b541 * ( 1999 ) 651 , hep - th/9804188 c. chu and p. ho , _ noncommutative open string and d - brane _ , nucl .
* b 550 * ( 1999 ) 151 , hep - th/9812219 v. schomerus , _ d - branes and deformation quantization _ , j. high energy phys . 9906 ( 1999 ) 030 , hep - th/9903205 a.yu .
alekseev , v. schomerus , _ d - branes in the wzw model _ , hep - th/9812193 h. garca - compen , j.f .
plebaski , _ d - branes on group manifolds and deformation quantization _ , hep - th/9907183 j. frhlich , k. , _ conformal field theory and the geometry of strings _ , crm proceedings and lecture notes vol .
* 7 * ( 1994 ) 57 , hep - th/9310187 a.h .
chamseddine , j. frhlich , _ some elements of connes non - commutative geometry , and space - time geometry _ , in : chen ning yang , a great physicist of the twentieth century , c.s .
liu and s .- t .
yau ( eds . ) , international press , 1995 , hep - th/9307012 j. frhlich , o. grandjean , a. recknagel , _ supersymmetric quantum theory , non - commutative geometry , and gravitation _ , les houches lecture notes 1995 , hep - th/9706132 o. grandjean , _ non - commutative differential geometry _ , ph.d .
thesis , eth zrich , july 1997 i. brunner , m.r .
douglas , a. lawrence , c. rmelsberger , _ d - branes on the quintic _ , hep - th/9906200 c.g .
callan , j.a .
harvey , a. strominger , _ world sheet approach to heterotic instantons and solitons _ , nucl .
* b359 * ( 1991 ) 611 + s .- j .
rey , _ the confining phase of superstrings and axionic strings _ , phys .
* d43 * ( 1991 ) 526 + s. frste , d. ghoshal , s. panda , _ an orientifold of the solitonic fivebrane _ , phys .
* b411 * ( 1997 ) 46 , hep - th/9706057 + m. bianchi , y.s .
stanev , _ open strings on the neveu - schwarz pentabrane _ , nucl .
phys . * b523 * ( 1998 ) 193 , hep - th/9711069 a. giveon , d. kutasov , n. seiberg , _ comments on string theory on ads@xmath181 _ , adv
. theor .
* 2 * ( 1998 ) 733 , hep - th/9806194 + j. de boer , h. ooguri , h. robins , j. tannenhauser , _ string theory on ads@xmath181 _ , j. high energy phys .
9812 ( 1998 ) 026 , hep - th/9812046 + d. kutasov , n. seiberg , _ more comments on string theory on ads@xmath181 _ , j. high energy phys . 9904 ( 1999 ) 008 ,
hep - th/9903219 j. teschner , _ operator product expansion and factorization in the @xmath182-wznw model _
, hep - th/9906215 m. kontsevich , _ deformation quantization of poisson manifolds i _ , q - alg/9709040 a.s .
cattaneo , g. felder , _ a path integral approach to the kontsevich quantization formula _ , math/9902090 k. , _ conformal field theory : a case study _
, hep - th/9904145 a.yu .
alekseev , a. malkin , e. meinrenken , _ lie group valued moment maps _
, j. diff .
geom . * 48 * ( 1998 ) 445 j.l .
cardy , _ boundary conditions , fusion rules and the verlinde formula _ , nucl .
* b324 * ( 1989 ) 581 g. pradisi , a. sagnotti , y.s .
stanev , _ completeness conditions for boundary operators in 2d conformal field theory _ , phys .
* b381 * ( 1996 ) 97 , hep - th/9603097 a. recknagel , v. schomerus , _ d - branes in gepner models _ , nucl .
* b531 * ( 1998 ) 185 , hep - th/9712186 d.c .
lewellen , _ sewing constraints for conformal field theories on surfaces with boundaries _ , nucl .
* b372 * ( 1992 ) 654 i. runkel , _ boundary structure constants for the a - series virasoro minimal models _
, nucl . phys .
* b549 * ( 1999 ) 563 , hep - th/9811178 l. alvarez - gaum , c. gomez , g. sierra , _ quantum group interpretation of some conformal field theories _ , phys .
* b220 * ( 1989 ) 142 j. hoppe , _ diffeomorphism groups , quantization and @xmath183 _ , int .
j. mod .
phys . * a4 * ( 1989 ) 5235 j. madore , _ the fuzzy sphere _ , class .
( 1992 ) 69 v.g .
drinfeld , _ quasi - hopf algebras and knizhnik -zamolodchikov equations _ , in : problems of modern quantum field theory , proceedings alushta 1989 , research reports in physics , springer verlag , 1989 + v.g .
drinfeld , _ quasi - hopf algebras _ , leningrad math . j. vol .
* 1 * ( 1990 ) no .
6 p. podle , _ quantum spheres _ , lett .
* 14 * ( 1987 ) 193 g. mack , v. schomerus , _ action of truncated quantum groups on quasi - quantum planes and a quasi - associative differential geometry and calculus _ ,
* 149 * ( 1992 ) 513 v. schomerus , _ construction of field algebras with quantum symmetry from local observables _ ,
* 169 * ( 1995 ) 193 , hep - th/9401042 g. moore , n. reshetikhin , _ a comment on quantum group symmetry in conformal field theory _ , nucl .
* b328 * ( 1989 ) 557 a.yu .
alekseev , s. shatashvili , _ from geometric quantization to conformal field theory _
, commun .
* 128 * ( 1990 ) 197 ; , commun .
* 133 * ( 1990 ) 353 j. frhlich , t. kerler , quantum groups , quantum categories , and quantum field theory , lecture notes in mathematics vol . 1542 , springer verlag , 1993 j. fuchs , c. schweigert , _ orbifold analysis of broken bulk symmetries _ , phys .
lett . * b447 * ( 1999 ) 266 , hep - th/9811211 ; , hep - th/9902132 ; , hep - th/9908025 + l. birke , j. fuchs , c. schweigert , _ symmetry breaking boundary conditions and wzw orbifolds _ , hep - th/9905038 ; a. connes , noncommutative geometry , academic press , 1994 w. nahm , _ quasi - rational fusion products _ ,
j. mod .
* b8 * ( 1994 ) 3693 , hep - th/9402039 a.yu .
alekseev , a. recknagel , v. schomerus , _ generalization of the knizhnik - zamolodchikov equations _ , lett
. math .
* 41 * ( 1997 ) 169 , hep - th/9610066 n. seiberg , e. witten , _ string theory and noncommutative geometry _ , hep - th/9908142 a. connes , m.r .
douglas , a. schwarz , _ noncommutative geometry and matrix theory : compactification on tori _ , j. high energy phys . 9802 ( 1998 ) 003 , hep - th/9711162 m. li , _
comments on supersymmetric yang - mills theory on a noncommutative torus _ , hep - th/9802052 + m. berkooz , _ non - local field theories and the non - commutative torus _
* b430 * ( 1998 ) 237 , hep - th/9802069 i.ya .
arefeva , i.v .
volovich , _ noncommutative gauge fields on poisson manifolds _ , hep - th/9907114 h. grosse , c. klimk , p. prenajder , _ towards finite quantum field theory in non - commutative geometry _ , int .
j. theor .
* 35 * ( 1996 ) 231 - 244 , hep - th/9505175 ; , lecture notes clausthal 1995 , hep - th/9510177 + u. carow - watamura , s. watamura , _ noncommutative geometry and gauge theory on fuzzy sphere _ , hep - th/9801195 e. witten , _ bound states of strings and d - branes _ ,
* b460 * ( 1996 ) 335 , hep - th/9510135 a. recknagel , v. schomerus , _ boundary deformation theory and moduli spaces of d - branes _ , nucl .
* b545 * ( 1999 ) 233 , hep - th/9811237 a. recknagel , v. schomerus , _ moduli spaces of d - branes in cft - backgrounds _ , hep - th/9903139 a.yu .
alekseev , h. grosse , v. schomerus , _ combinatorial quantization of the hamiltonian chern - simons theory i _ , commun .
* 172 * ( 1995 ) 317 , hep - th/9403066 g. felder , j. frhlich , j. fuchs , c. schweigert , _ the geometry of wzw branes _
, hep - th/9909030 | the geometry of d - branes can be probed by open string scattering .
if the background carries a non - vanishing b - field , the world - volume becomes non - commutative .
here we explore the quantization of world - volume geometries in a curved background with non - zero neveu - schwarz 3-form field strength @xmath0 . using exact and generally applicable methods from boundary conformal field theory
, we study the example of open strings in the su(2 ) wess - zumino - witten model , and establish a relation with fuzzy spheres or certain ( non - associative ) deformations thereof .
these findings could be of direct relevance for d - branes in the presence of neveu - schwarz 5-branes ; more importantly , they provide insight into a completely new class of world - volume geometries . + + |
a new field of research has emerged in the past decade in the context of solid - state chemistry and physical chemistry .
it is the science of metal organic frameworks ( mof s ) .
metal organic frameworks , also called _ porous coordination polymers _
( pcp s ) , are hybrid crystalline porous materials consisting of metallic species connected to one another with organic linkers .
they display an extremely large range of crystal structures and host - guest properties , which makes them an important class of materials with potentially major impact in adsorption / separation technologies of strategic gas linked with energy supply and environmental problems .
the combination of tunable porosity , the functionalization of the internal surface together with the structural flexibility of the host opens the way to an extremely rich host - guest chemistry , putting this class of materials in a unique position .
one fascinating aspect of hybrid frameworks is the ability of a subclass of structures to behave in a remarkable guest - responsive fashion
. these so - called _ soft porous crystals _
( spcs)@xcite exhibit a variety of large amplitude dynamic behaviors of their frameworks in response to external stimuli of weak intensity ( light , electric field , gas exposure , etc . ) .
the change in the spc channels in response to the external constraint is reversible and maintains the crystalline character of the solid .
as an example , one may cite the mil-53 type frameworks@xcite which exhibit guest - induced structural phase transitions upon gas adsorption and desorption , called `` breathing '' transitions .
the bistable behavior of this system is controlled by the gas pressure that acts as the external stimulus .
molecular simulation of adsorption in flexible porous solids is a challenging field,@xcite and several approaches have been developed over the time to circumvent the difficulty of direct simulation in the osmotic ensemble , in which the temperature @xmath0 , mechanical strain @xmath1 , host framework @xmath2 and adsorbate chemical potential @xmath3 are imposed .
firstly , many published works trying to shed light onto the interplay between adsorption and structural changes have relied on an indirect approach , in which regular grand canonical monte carlo ( gcmc ) simulations ( i.e. at constant chemical potential , temperature and volume ) are performed on a number of possible host structures .
adsorption isotherms computed using this approach are then compared with experimental data , and any steps observed between `` rigid host '' isotherms can be ascribed to adsorption - induced structural transitions .
this was successfully used to demonstrate that hydrogen adsorption could trigger flexibility in zif-8,@xcite as well as in early simulation studies of the breathing of mil-53(cr ) upon co@xmath4 adsorption.@xcite while in most cases these gcmc simulations were performed on rigid adsorbents , some authors have used a forcefield accounting for the flexibility of the framework , describing local deformations of the host material even though the overall unit cell volume is preserved.@xcite a second approach to guest - induced flexibility of soft porous crystals is to perform a series of @xmath5 molecular simulations at increasing loading @xmath6 .
dubbeldam et al .
have showcased a methodology @xcite where the material s unit cell , loaded with adsorbate , is energy - minimized .
this allows one to follow the zero - kelvin response of the material s structure to guest adsorption , and describes elastic deformation ( generated by the so - called adsorption stress ) as well as structural transitions .
others groups have performed @xmath7 simulation either without adsorbate @xcite or with a varying number of guest molecules,@xcite as in the case of h@xmath4o in mil-53(cr).@xcite compared to these indirect approaches , studies using direct molecular simulation in the osmotic ensemble are scarce .
maurin and coworkers @xcite have reported the use of an hybrid molecular dynamics / monte carlo scheme ( md / mc ) @xcite for the description of co@xmath4 adsorption in mil-53(cr ) .
this scheme includes , as part of a @xmath8 monte carlo simulation , shorts runs of @xmath5 molecular dynamics in which the unit cell of the system can deform .
these authors demonstrated that such a scheme presents severely limiting problems of convergence towards equilibrium , resulting in only some of the structural transitions being reproduced during an adsorption desorption cycle .
recently , those authors have shown that a careful and thorough calibration of force field parameters could enable them to witness the second transition during adsorption@xcite .
in essence , direct simulation in the osmotic ensemble does not easily allow one to efficiently overcome free energy barriers between very different states of the system .
it can , however , be successful if the different host structures have similar energy and volume , as is the case of silicalite-1 , where the three possible phases of the material differ only by 0.6% in unit cell volume.@xcite finally , a last class of simulation methods for adsorption in flexible nanoporous materials relies on the use of free energy methods ( also called density of state calculations ) .
these techniques are based on the computation of the full free energy landscape as a function of one or more order parameters .
the thermodynamic behavior of the system can then be fully determined from this landscape .
for example , miyahara et al .
studied the adsorption - induced structural transitions of a so - called jungle gym model system through the reconstruction of grand free energy profiles by thermodynamic integration as a function of subnet displacement.@xcite by choosing both the unit cell loading and cell vectors as order parameters , one can reconstruct free energy landscapes directly from adsorption isotherms in a rigid host .
similar reconstruction has been successfully applied to the mil-53 materials by an analytical approach that uses fits of experimental isotherms.@xcite in this work , we extend these ideas and present a new monte carlo simulation method using a non - boltzmann sampling algorithm to perform direct molecular simulation of adsorption in flexible nanoporous solids .
this method is compared to standard gcmc calculation and direct osmotic simulation .
the adequate ensemble to study adsorption in flexible materials is the osmotic ensemble introduced by brennan and madden@xcite in the context of polymer - solvent mixtures .
the control parameters for this ensemble are the temperature , @xmath0 , the mechanical constraint exerted on the material , @xmath1 , the chemical potential of the adsorbed gas , @xmath3 , and the number of particles of the host framework , @xmath2 .
the osmotic grand potential is written as follows : @xmath9 if the host exhibits structural phase transitions , i.e. if it exists in an equilibrium between a number of metastable structures , the host degrees of freedom may be decoupled from the rest of the systems variables .
one can then decompose the osmotic grand potential as a sum of the free energy of the host material and the grand canonical potential for the guest molecules.@xcite this was used extensively to study the mil-53(al ) adsorption induced breathing , mainly through an analytical scheme.@xcite the osmotic potential is then written as : @xmath10 theoretical models based on this decomposition of the osmotic potential have met success in allowing the interpretation of data obtained from either adsorption
desorption experiments@xcite or molecular simulation in the grand canonical ensemble.@xcite however , direct molecular simulation is constrained by severe fundamental limitations , which were highlighted in the previous section .
both monte carlo and molecular dynamics - based techniques typically can not overcome free energy barriers coupling adsorption and host deformation , unless the host phases are very close energetically and geometrically .
here we present a molecular simulation methodology able to bypass these limitations by relying on non - boltzmann sampling based on the use of a wang landau algorithm .
landau algorithm was originally developed to calculate density of states on the fly in the canonical ensemble,@xcite by performing a monte carlo simulation with a non - boltzmann acceptance probability of @xmath11 , where @xmath12 is the density of states of the system .
this technique was later extended to other ensembles and the calculation of free energy as a function of other order parameters ( energy , volume , number of particles , reaction coordinate , etc.).@xcite we propose here that the wang landau can be applied to the calculation of the free energy in the osmotic ensemble , as a function of the order parameters involved in the adsorption - induced structural transitions : the unit cell loading @xmath6
( i.e. the quantity of adsorbed particles ) and the host unit cell volume @xmath13 .
the density of state also depends on temperature @xmath0 and number of particles of the host framework , @xmath2 .
since these thermodynamics variables are fixed in all our study , there are omitted in our notation in the following .
the knowledge of the extended density of states @xmath14 of a given system allows to calculate all of its thermodynamic properties .
the osmotic thermodynamic potential can thus be expressed as : @xmath15\ ] ] while the osmotic wang landau free energy for a given chemical potential @xmath3 and external pressure @xmath1 is : @xmath16\ ] ] thus , once the two - dimensional extended density of states for the system has been calculated for a given host and a fixed temperature , the osmotic potential as well as the landau free energy can be derived from this density of state for any set of parameters @xmath17 . in turn ,
any observable may then be similarly computed _ a posteriori _ , as a function of chemical potential and mechanical pressure , from a single extended density of state calculation .
we now turn to the practical calculation of this two - dimensional density of states during a monte carlo simulation . during a `` standard ''
monte carlo simulation using boltzmann acceptance probabilities , the probability to generate a configuration with @xmath6 adsorbed particles and a volume @xmath13 is proportional to @xmath14 .
the idea behind the wang
landau algorithm is to modify the acceptance probabilities of the insertion / deletion and volume change mc moves so that the simulation will homogeneously visit all states .
this means that acceptance probabilities are multiplied by the ratio of the density of state between the old and new states : @xmath18\\[1ex ] \mathcal p(n{\ensuremath{_{\text{ads}}}},v\rightarrow n{\ensuremath{_{\text{ads}}}}+1,v ) & = & \textrm{min } \left [ 1 , \frac{q(n{\ensuremath{_{\text{ads}}}},v)}{q(n{\ensuremath{_{\text{ads}}}}+1,v)}\cdot \frac{v}{(n{\ensuremath{_{\text{ads}}}}+1)\lambda^3 } \textrm{e}^{-\beta \delta e } \right]\\[1ex ] \mathcal p(n{\ensuremath{_{\text{ads}}}},v\rightarrow n{\ensuremath{_{\text{ads}}}}-1,v ) & = & \textrm{min } \left [ 1 , \frac{q(n{\ensuremath{_{\text{ads}}}},v)}{q(n{\ensuremath{_{\text{ads}}}}-1,v)}\cdot \frac{n{\ensuremath{_{\text{ads}}}}\lambda^3}{v } \textrm{e}^{-\beta \delta e } \right ] \ ] ] where @xmath19 is the de broglie wavelength . the density of state ( dos ) @xmath14 is discretized and updated on the fly during the monte carlo simulation .
the volume @xmath13 is a continuous variable , so it is discretized into a series of bins @xmath20 $ ] ; for each value of @xmath6 ( which is naturally discrete ) and each bin of @xmath13 , the computer stores and updates the value of @xmath21 . because it is initially unknown , it is set uniformly to 1 at the beginning of the simulation .
then , at each monte carlo step , the dos of the current @xmath22 state is multiplied by a factor @xmath23 .
simulation proceeds until all states have been uniformly visited , which is indicated by a flat histogram of visited states in @xmath22 space .
when the histogram is flat , the accumulated density of state has converged with an accuracy proportional to @xmath24 .
then , the histogram is cleared , the factor @xmath23 is decreased ( we used the @xmath25 scheme in this study ) , and the simulation is allowed to continue for a new wang landau iteration , with this new value of @xmath23
. convergence of the dos depends on the overall number of wang
landau iterations performed and the convergence of the dos in the course of each iteration . to assess the flatness of the histogram , which determines the latter , one of two schemes is typically used .
first , one can either consider a minimum visit criterium , i.e. consider the histogram is flat when all bins were at least visited a specific number of times ( we chose 500/@xmath26 in this study ) .
a second criterion is to require that the least visited and most visited bins of the dos are within 20% of the mean number of visits .
both approaches yielded the same result on our system , and we there used first requirement ( minimum number of visits per bin ) , which is less demanding in terms of computational cost.@xcite furthermore , @xmath23 has to be initially high enough to allow the system to rapidly visit all states , so we chose its initial value as @xmath27 .
the wang landau free energy profiles obtained thusly converged after 23 wang
landau iterations , with a final value of @xmath28 .
the two - dimensional wang
landau sampling in @xmath22 space as outlined above forces a broad sampling of the system s configurational space , overcoming free energy barriers coupling adsorption and framework deformation .
however , the dimensionality of the order parameter space and the stringent conditions imposed on the convergence of the density of state mean the methods has a very expensive computational cost .
one possible approach to reduce
the cpu cost is that the problem is _ embarrassingly parallel _ in nature : independent simulations can be run on disjunct sets of the order parameter space , and the full density of state later reconstructed from all simulations by continuity .
but such lengthy calculations are not in fact necessary , as other schemes can be used to reconstruct a full 2d free energy profile from 1d wang landau calculations .
two reconstruction schemes of the total osmotic density of state have been realized ( see fig . [ reconstruction ] ) .
the first one consists in computing the one dimensional dos at fixed adsorption loading @xmath29 , and reconstructing the 2d dos using a density of state calculation at a given reference volume @xmath30 ( @xmath31 ) .
the second scheme is to calculate the dos for several fixed volume @xmath13 as a function of @xmath6 ( @xmath32 ) , and then reconstructing the density of states using the knowledge of a density of state at a given @xmath6 .
both reconstructions leads to the same result within statistical uncertainties .
however , the second approach presents a strong advantage since one can use the dos at @xmath33 .
indeed , taking advantage of the fact that @xmath34 , it follows that : @xmath35 the use of the known host free energy function as a basis for reconstruction speeds up calculations .
moreover it allows to recalculate the 2d free energy profiles for various host free energy profiles without re - running any new calculations since the @xmath36 term is not included directly in the simulations , but only used in post - processing data to reconstruct free energy surface .
this is of great help in investigating the effect of the host material s properties on adsorption - induced breathing .
in order to focus our study on the efficiency of various molecular simulation methods for soft porous crystals , we have limited the computational effort involved in each individual simulation by working on a very simple model mimicking the geometrical and physical properties of the mil-53 family of `` breathing '' materials .
the model , depicted in fig [ snap ] , is made of unidimensional diamond - shaped pores separated by walls of lennard - jones particles .
the corner to corner distance is denoted by @xmath37 , and the angle between two walls is @xmath38 ; the channels are parallel to the @xmath39 axis .
both values are chosen to mimic the experimental cell parameters of the np and lp phase of the mil-53(al ) : @xmath37 is taken as 10 and @xmath38 is either 42.6 ( np phase ) or 75.1 ( lp phase ) .
the simulation cell in the perpendicular dimension is fixed to 10 .
all simulations were then performed on a @xmath40 supercell with periodic boundaries conditions .
wall - fluid and fluid - fluid interactions were both described by 126 shifted lennard - jones potentials , with a spherical cutoff at a distance of 7.5 .
fluid - fluid interaction parameters are taken as @xmath41 and @xmath42 k. the wall - fluid interactions were tuned so that the adsorption enthalpy in the np phase be more favorable than in the lp phase , which is the _ sine qua non _ condition for breathing.@xcite the wall - fluid lennard - jones parameters were varied in the range of @xmath43 , and @xmath44 k. mc simulations in the nvt ensemble with a single adsorbate molecule were performed to calculate np and lp adsorption enthalpies .
results ( plotted in fig .
s1 ) show that the adsorption enthalpy difference is maximum for @xmath45 .
we chose a @xmath46 value of @xmath47 k to get reasonable adsorption at pressures comparable to experiment on the real material .
a free energy term as a function of the opening @xmath38 was added in order to account for the bistable nature of the material s framework .
the free energy landscape @xmath48 was defined as a smooth biparabolic potential , as shown in fig .
[ fhost ] .
the free energy difference between np and lp structures was taken from ref . and fixed at 7.5 kj.mol@xmath49 at 300 k. unless otherwise specified , the free energy barrier was arbitrary chosen as 15 kj.mol@xmath49 .
this simple system is expected to behave only qualitatively as the real material it is inspired from , rather than reproducing any experimental data .
this model does not include the rotation of organic linkers for example , and the only internal degree of freedom of the host considered at this stage is the opening angle . before studying the full thermodynamics of the system in the osmotic ensemble
, we first performed a study in the grand canonical ensemble .
there , the host was considered as rigid and was fixed in either the np phase , or the lp phase , and adsorption isotherms were computed at 300 k using standard gcmc simulations for each phase .
the resulting adsorption isotherms are shown as dashed lines in fig .
[ iso ] ; they depict a behavior typical of the experimental data on mil-53 breathing . in particular , the henry constant is significantly higher in the np phase and the maximum adsorbed amount is lower , which are the _ sine qua non _ conditions for breathing .
the adsorbate density maps in the np and lp phases have been analyzed ( see fig .
the lp phase features two different types of adsorption site .
the osmotic free energy difference between the lp and np phases can be computed by integrating the gcmc adsorption isotherms:@xcite @xmath50 finally , using this thermodynamic model , we can produce equilibrium isotherms that feature vertical steps corresponding to lp@xmath51np and np@xmath51lp structural transitions , as depicted on fig .
[ iso ] ( for @xmath52 k ) .
the respective transition pressures calculated in this approximation are 0.063 and 46.2 bar , respectively .
adsorption in this simple model does trigger breathing at 300 k demonstrating its ability to capture at the microscopic level the features of a breathing framework .
we can thus reconstruct a full thermodynamic picture of the system from gcmc isotherms , although one needs to be mindful of the fact this formulation remains incomplete .
indeed , the simulations were performed in the grand canonical ensemble , missing all effects of flexibility and local deformation of the framework .
combining them only gives a representation in a sub - osmotic ensemble , that only takes into account two possible sets of cell parameters corresponding to those of the empty np and lp phases . explicitly considering the host flexibility is paramount when describing the adsorption - induced flexibility of soft porous crystals . in order to evaluate the convergence and efficiency of direct monte carlo simulations for highly flexible materials
, we performed a mc simulation of adsorption in our mil-53 model material in the osmotic ensemble .
to do so , the simulation included the standard gcmc moves ( insertion , deletion and translation ) as well as moves involving changes of the unit cell parameters . taking into account the constraints due to the nature of the material , the length @xmath37 of the model was considered fixed ( consistent with the rather rigid nature of the organic linkers in the real material ) and the volume change steps thus only affected the opening angle , @xmath38 , of the material , homogeneously for all pores in the simulated supercell .
`` adsorption '' and `` desorption '' isotherms were then computed , which corresponded to a series of osmotic ensemble simulations at increasing ( resp . decreasing ) chemical potential @xmath53 ,
each simulation starting from the final configuration of a previous run at the immediately lower ( resp .
higher ) value of @xmath53 .
all simulations were started from the lp phase : the unit cell was initially empty for adsorption , or full preloaded with adsorbate for desorption calculations .
each simulation was carried for 100 millions mc steps , with chemical potentials corresponding to a total pressure ranging from 0.01 to 1000 bar .
several simulations were performed using different mc moves set up , more specifically , different insertion and structure change moves probabilities were tested , ranging from 10% to 20% .
the combination of 20% insertion / deletion move , 20% structure change move and 60% guest molecules translation yields the more transitions between the large and the narrow pores forms .
the acceptance probabilities of insertion / deletion moves ranges from 10% at low pressure to around 1% at high guest loading , while the acceptance of structure change mc move consistently stay around 5% , although most of the accepted moves are of a low amplitude . the number of transition between the closed and the open form is strikingly low for a vast majority of the simulation which presents np - lp switches only at very specific pressure range as depicted in fig . [ xlp ] . at @xmath54 bar , one only witness a low number of very short transition to the np phase . at @xmath55 bar , about 20 np@xmath56lp transitions are noticed during a simulation of 100 million mc steps . at @xmath57 bar , only one transition occurs from the lp to the np after a few millions steps , and the system is stuck in the closed form until the end of the simulation run ( i.e. @xmath58 steps ) .
during desorption , no significant volume change is observed till 0.64 bar where there is only one lp@xmath51np transition during 100 million steps . as pressure lowers ,
the system behaves exactly as it does during adsorption , i.e. presenting multiple transitions . upon adsorption ,
the first breathing transition np@xmath51lp is obviously observed although the low number of large fluctuation gives doubt about the accurate convergence toward the true thermodynamical equilibrium . the second transition has never been observed in any simulations we have tried . this transition would require a complex change in configurational space involving changes in particles coordinates , guest quantity , and cell parameters .
the opening change steps only scale the adsorbate position according to the new cell parameters , and fail to take into account such necessary configurational changes .
the reason the first transition is observed is due to the low adsorption loading at which the transition occurs , which renders the configurational change much less substantial , contrary to the second transition that occurs at higher pressure , hence higher loading . during desorption simulation ,
results are essentially the same and only the low pressure transition is spotted , although the pictures highly hinge on the initial condition .
a simulation starting from an initial configuration in the lp phase is forced to be in the expected state at high pressure , but one does not witness the closing of the structure until the pressure is orders of magnitude below the equilibrium transition .
such an hysteresis is only due to the incapacity of reaching a statistically significant thermodynamic equilibrium using a direct simulation in the osmotic ensemble .
the inability of this raw implementation to reproduce the second transition of the breathing phenomenon is in tune with previous published results .
as it was expected , the monte carlo scheme we used to simulate the system in the osmotic ensemble is equivalent to the npt molecular dynamics used by maurin and his coworkers.@xcite it appears that there is an inherent obstacle to the simulation of this system in the osmotic ensemble , and one would need to recourse to complex mc moves combining guest molecules insertion / deletion , adsorbate coordinates translation , and host framework structure change , with no guaranty of convergence within an accessible computational time without resorting to complex statistical bias . to overcome the convergence difficulties in direct osmotic simulations , we implemented the non - boltzmann sampling method described in the `` theory '' section , based on the wang
landau algorithm .
we have thus calculated the full density of states in the @xmath59 space , from which the full thermodynamic behavior of the system can be deduced . given the geometry of the material studied
, sampling was performed using the diamond - shaped channel angle , @xmath38 , as an order parameter instead of the unit cell volume @xmath13 .
thus , a series of wang landau calculations were performed at fixed values of @xmath38 , going from 35 to 86 , to obtain free energy profiles @xmath60 .
each of these 52 wang landau calculations performed 23 iterations , with the last ( and longest ) iteration featuring from 250 millions to 1 billion mc steps . from these
, the full 2d density of states was calculated using the reconstruction scheme described previously . the uncertainty on the landau free energy is then expected to be of the order of @xmath61j.mol@xmath62 . in the case of a simulation of a more complex and realistic system
, the number of wang landau cycles could be reduced while still getting reasonable accuracy .
we report in fig .
[ dos ] the calculated osmotic - ensemble free energy landscapes as a 2d - function of the opening , @xmath38 , and the number of adsorbed guest molecules , @xmath6 , for a set of increasing chemical potentials @xmath53 corresponding to external gas pressures in the range 0.001 to 100 bar .
the landscape is characterized by the presence of two wells corresponding to metastable forms of the host material .
a saddle point is found between the two minima that corresponds to the optimal pathways between the two phases .
these broad features are direct consequences the bistable nature of the host , taken into account by the biparabolic nature of free energy profile ( fig .
[ fhost ] ) . at low pressure
, this host free energy dominates the 2d landscape , and favors the lp form as the most stable at 300 k. at slightly higher gas pressure , the minimum of the np region shifts and now corresponds to a non - zero adsorbate loading , while that of the lp phase does not .
this is in keeping with the np phase s higher affinity for the guest , as established by the comparison of their henry constants .
in addition to 2d free energy surfaces , the evolution of lp and np free energy , i.e. the free energy of the local minima in the free energy surfaces , as a function of gas pressure is shown in fig .
[ fnplp ] .
it can there be seen that the adsorption - induced stabilization provided by this adsorption leads the np phase to become more stable than the lp one , above @xmath63 bar .
this crossing corresponds to the equilibrium of the first breathing transition lp@xmath51np ; this pressure is in line with the results of both the direct osmotic simulations and the grand canonical thermodynamic integration .
finally , with further increase in the external gas pressure , the np form s loading reaches saturation , while adsorption in the lp phase now increases markedly , stabilizing this latter .
the free energy difference between both phases eventually start to decrease at approximately 3 bar , and the second breathing transition np@xmath51lp is observed at an equilibrium pressure of 50.9 bar .
the evolution of the opening angle extracted from the minimum of the free energy landscape as a function of the gas pressure is shown in fig .
[ xlp ] and compared to direct osmotic simulation results .
the first transition matches with direct osmotic simulation results .
the second transition clearly stands out in the dos formulation while , in the direct simulation in the osmotic ensemble , it is only observed during desorption simulation and at a pressure widely shifted from the equilibrium value ( see fig .
[ xlp ] ) .
our free energy calculation allows to overcome the convergence breakdown in direct osmotic simulation and clearly reveals both transition .
these simulations also agrees on the local contraction / expansion of the material upon adsorption in each structural phases .
the opening angle of the lp phase is plotted on the whole gas pressure range in fig .
[ alphalp ] disregarding whether the lp phase is a stable or metastable state .
starting from the equilibrium angle of the empty material , the opening angle slightly decreases upon gas adsorption , reaches a minimum and significantly increases at higher gas pressure .
such contraction - dilation phenomenon upon adsorption is a classical behavior observed in many porous material.@xcite it is worth mentioning that such local _ flexibility _ is linked to the elastic constant of the phase but is not intrinsically related to the large flexibility ( or structural transitions ) observed in this bi - stable material.@xcite we then compared the transition pressures obtained with our new approach to the values predicted by our analytical model applied to gcmc isotherms .
for both transitions , a discrepancy is observed : 0.051 bar ( wang landau ) vs. 0.062 bar ( gcmc ) for the first transition , and 50.9 bar vs. 46 bar for the second one .
these differences can be ascribed to adsorption induced elastic deformations of the lp and np phase , which are accounted for in the osmotic ensemble ( see fig .
[ alphalp ] ) but not in simple gcmc adsorption isotherms .
indeed , we calculated the wang
landau landscapes for a series of modified host free energy functions , in which both the lp and np phase were rendered less compliant ( their elastic constant @xmath64 was multiplied by a factor ranging from @xmath65 to @xmath66 ) .
these simulations showed an evolution of the transition pressures towards the values obtained from the model applied to gcmc isotherms ( see fig .
the two methods are thus fully consistent . from the full free energy landscape , we have thus analyzed the relative stability of the metastable states , and found it to be different from the results one would have obtained from either direct osmotic molecular simulation ( which fail to give equilibrium transition pressures ) and gcmc isotherms ( which fail to account for local framework deformations ) . moreover , we can also look at the full picture of the breathing phenomenon in the @xmath22 space . on the free energy surface
, we can determine a `` reaction path '' for the breathing transitions , and in particular locate the transition state of the each transition , which corresponds to the saddle point of the surface .
this enables us to estimate the free energy barriers involved in the np@xmath51lp and lp@xmath51np transitions .
the existence of this barrier can account for the hysteretic nature of the system , trapping it in a metastable state .
earlier studies have shown that this free energy barrier influences the dynamics of the breathing transition.@xcite we present in figure [ hyst ] the evolution of the free energy barrier for both the closing and opening transitions as a function of external gas pressure , computed from the analysis of the osmotic free energy landscape . in the figure
, we compare these barriers to a threshold value of 18.75 kj.mol@xmath49 over which the transition is considered impossible , and under which the transition occurs immediately . a theoretical framework to study
this phenomenon is still lacking and this value remains arbitrary ( other barrier thresholds were used to compute the corresponding adsorption - desorption hysteresis loop and are shown in fig .
s4 ) , but it enables us to give a simple interpretation of the hysteresis loop on the basis of free energy barriers involved in the transitions . on this basis , we describe here a simple adsorption - desorption cycle , starting in the empty lp phase . upon adsorption of guest molecules , the free energy barrier of the lp@xmath51np transition
lowers until it reaches the critical value at which the system switches to the now much more stable np phase .
this necessarily happens at a pressure higher than the equilibrium transition , as expected .
as pressure further rises , the np phase eventually ends up saturated in guest molecules and the saddle point ( and lp ) free energy lower faster than the np minimum , up to a point where the np@xmath51lp free energy barrier meets the threshold .
this transition again occurs at higher pressures than the equilibrium between both phases . then , upon desorption , the free energy barrier of the lp@xmath51np decreases to the point where it gets below the free energy threshold and the material closes itself at a pressure lower than the equilibrium , creating a hysteresis loop .
the same phenomenon appears for the last reopening transition , and we see that adsorption and desorption branches display different breathing transitions , creating the two hysteresis loops shown in fig .
[ hyst ] and fully compatible with experimental results on breathing mof s .
this hysteresis in the adsorption - desorption loop is highly dependent on the free energy activation value chosen for the analysis , and the assumption made that the threshold remains the same for both the opening and closing transitions as well as being independent of the pressure .
regardless of those hypotheses , this density of state formulation provides a qualitative interpretation of the hysteresis nature of the adsorption - desorption in flexible porous materials
we have developed a new monte carlo simulation method for the simulation of soft porous crystals , based on non - boltzmann sampling in \{guest loading , volume } space using the wang landau algorithm .
this simulation approach can be used to fully characterize the adsorption properties and the material s response to adsorption at thermodynamic equilibrium .
in addition , it produces the full free energy landscape of the material as a function of both deformation and loading , and can thus be used to better understand the hysteretic nature of adsorption - induced structural transitions .
we showcased this new method on a simple model of the mil-53 family of breathing materials , highlighting its advantages compared with the conventional simulation techniques used in the field ( grand canonical monte carlo of rigid structures , and direct simulations in the osmotic ensemble ) .
this method will be applied to atomistic models of real materials using realistic host forcefields in future work .
the authors thank fabien cailliez , alex neimark , and alain fuchs for discussions and insightful comments .
they acknowledge funding from the agence nationale de la recherche under project `` soft - crystab '' ( anr-2010-blan-0822 ) .
37ifxundefined [ 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 * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop
_ _ ( , ) @noop * * , ( ) @noop * * , ( ) adsorption isotherms computed at @xmath52 k with gcmc simulations in the np ( dashed red ) and lp ( dashed blue ) rigid structures . the full black line is a composite adsorption isotherm , featuring the breathing transitions calculated using an analytical model in the osmotic ensemble.,width=302 ] mean opening angle as a function of external gas pressure during osmotic monte carlo simulations of adsorption ( green ) and desorption ( magenta ) , at @xmath52 k. the behavior for the system at thermodynamic equilibrium is displayed as a dashed black line ( dos calculation).,width=302 ] reconstruction schemes of the total osmotic density of state , either by directly computing from a monte carlo density of state calculation in the osmotic ensemble ( left ) , or using several wl simulations at fixed adsorption loading @xmath6 , and reconstructing them using a density of state calculation at a given reference volume @xmath30 ( center ) , or using several wl simulations at fixed volume @xmath13 , and reconstructing them using the density of state of the bare host material , i.e. at @xmath68 ( right ) . , scaledwidth=60.0% ] evolution of the osmotic free energy landscape in @xmath69 space as external gas pressure increases at 300 k : at 0.01 bar ( top left ) , 0.05 bar(top right ) , 1 bar ( middle left ) , 10 bar ( middle right ) , 50 bar ( bottom left ) and 100 bar ( bottom right).,scaledwidth=85.0% ] top panel : free energy barrier for the np @xmath51 lp and lp @xmath51 np breathing transitions ( in red and blue respectively ) , as a function of external gas pressure , compared to a threshold of 18.75 kj.mol@xmath49 ( dashed black line ; see text for details ) . lower panel : computed adsorption ( green ) and desorption ( orange ) isotherms showing hysteresis loops ; the isotherm at thermodynamic equilibrium is shown as a dashed black line.,width=377 ] | soft porous crystals are flexible metal organic frameworks that respond to physical stimuli such as temperature , pressure and gas adsorption by large changes in their structure and unit cell volume . while they have attracted a lot of interest , molecular simulation methods that directly couple adsorption and large structural deformations in an efficient manner are still lacking .
we propose here a new monte carlo simulation method based on non - boltzmann sampling in \{guest loading , volume } space using the wang landau algorithm , and show that it can be used to fully characterize the adsorption properties and the material s response to adsorption at thermodynamic equilibrium .
we showcase this new method on a simple model of the mil-53 family of breathing materials , demonstrating its potential and contrasting it with the pitfalls of direct , boltzmann simulations .
we furthermore propose an explanation for the hysteretic nature of adsorption in terms of free energy barriers between the two metastable host phases . |
since the discovery of atmospheric neutrino oscillations @xcite , more recently that of solar neutrino oscillations @xcite , and still more recently confirmation of these two types of neutrino oscillations with accelerator- and reactor - produced neutrinos @xcite , many models have been proposed in the literature to explain the mass and mixing parameters associated with these oscillations @xcite .
( we have cited just the most recent references for those experiments and several review articles for the models proposed . )
some models restrict their scope to the lepton sector , while others such as grand unified models are more ambitious and attempt to explain the masses and mixings in both the lepton and quark sectors .
the textures for the mass matrices obtained in the models may be simply postulated at the outset , derived from the observed mixing matrix with a diagonal charged lepton mass matrix , obtained from a certain unification group in conjunction with a well - defined family symmetry , or rely solely on the clebsch - gordon coefficients of that group . until now
, all viable models for the lepton sector have only had to explain the apparent near maximal @xmath12 mixing angle for the atmospheric neutrinos and the less than maximal @xmath13 large mixing angle ( lma ) solution for the solar neutrinos , while satisfying the two observed mass squared differences , @xmath14 and @xmath15 , and the upper bound for the still unobserved @xmath16 mixing angle . on the other hand , the neutrino mass hierarchy ( normal or inverted ) , the majorana vs dirac nature of the neutrinos , the appropriate values for @xmath16 and the cp - violating phases ( one dirac and two majorana ) remain undetermined experimentally . with the next more precise round of neutrino oscillation experiments which will restrict some of the unknowns , many of the more quantitative models proposed
are not expected to survive the more stringent tests of their predictions . in particular , determination of the neutrino mass hierarchy will eliminate models based on an approximately conserved @xmath17 lepton number @xcite if the hierarchy is observed to be normal , while the conventional type i seesaw models will be strongly disfavored if the hierarchy is observed to be inverted @xcite . moreover , small as opposed to very small predictions for @xmath16 pose a critical test for many models .
it is thus of utmost importance to sharpen the predictions for each model , so that the list of presently viable models can be winnowed down as much as possible when the new data is obtained .
one particularly interesting class of models is that based on the @xmath0 grand unification group .
there are two kinds of minimal models in this class : those based on higgs representations with dimensions @xmath18 , and possibly also @xmath19 and/or @xmath20 @xcite ; and those based on @xmath21 , and @xmath22 representations @xcite .
the former type generally has symmetric and/or antisymmetric matrix elements , while the latter type generally involve lopsided matrices for the down quarks and charged leptons .
the matrix textures are derived either from the clebsch - gordon coefficients required or from some abelian or non - abelian family symmetry .
these two types of @xmath0 models tend to predict @xmath16 angles either near the present upper chooz bound of @xmath23 , or in the @xmath24 range , respectively .
the purpose of this paper is to sharpen the allowed values for the mixing parameters of the lopsided @xmath0 model proposed some time ago by babu , barr , and the author @xcite .
upon further refinement by the latter two authors @xcite , the model has become very predictive with all four quark mixing parameters and nine quark and charged lepton masses determined by just eight of the twelve model input parameters .
the four remaining input parameters characterize the right - handed majorana mass matrix with three of them complex in general . in most of the previous papers based on this model
, the authors took the latter parameters to be real and adjusted them to give good fits to the then known neutrino mass and mixing data . here
we allow all three to be complex and obtain scatter plots of the predicted neutrino mixing parameters . in so doing , we obtain some bounds on @xmath16 and @xmath12 which , if violated , will rule out the model which has been highly successful todate . by introducing two very small additional parameters in the dirac neutrino mass matrix ,
we find that the observed baryon asymmetry can be obtained through resonant leptogenesis in the model . in order to obtain the latter predictions ,
the bounds obtained on the neutrino mixing parameters and cp phase are even more tightly constrained .
with maximal cp violation in the majorana mass matrix and two of the three parameters equal , the mixing angles and phase become highly correlated with each other and the baryon asymmetry generated in the model . in sec .
ii we briefly review some of the model details and present the results of the parameter searches in sec .
a summary of the constraints found in the model is presented in sec .
iv , where we also briefly comment on the predictions of the model found by jankowski and marbury @xcite for the @xmath25 lepton flavor - changing branching ratio .
their prediction is also crucial for the viability of the model and has a chance of being tested by the meg collaboration @xcite even before the values of the mixing angles are further restricted by new reactor and/or long baseline experiments .
the @xmath0 grand unified model as initially proposed by babu , barr and the author @xcite , and later updated by barr and the author to apply to the large mixing angle ( lma ) neutrino oscillation solution has already been described in great detail in the literature @xcite .
here we just summarize the salient features and refer the reader to those references for more information .
the model is based on the minimum set of higgs fields which solves the doublet - triplet splitting problem as proposed by barr and raby @xcite .
this requires just one @xmath26 whose vacuum expectation value ( vev ) points in the @xmath27 direction , two pairs of @xmath28 s which stabilize the solution , along with several higgs fields in the @xmath29 representations and higgs singlets .
the higgs superpotential exhibits the @xmath30 symmetry which is used for the flavor symmetry of the grand unified model .
the combination of vevs , @xmath31 and @xmath32 break @xmath0 to the standard model , where the latter two vevs point in the @xmath33 singlet direction .
the electroweak vevs arise from doublets in the @xmath34 and @xmath35 representations of @xmath33 in the combinations @xmath36 \langle h_d \rangle & = & \langle \overline{5 } ( { \bf 10}_h)\rangle \cos \gamma + \langle \overline{5}({\bf
16}'_h)\rangle \sin \gamma,\\ \end{array } \label{eq : higgsmixings}\ ] ] while the mixing orthogonal to @xmath37 gets massive at the grand unified theory ( gut ) scale .
matter superfields appear in three chiral @xmath38 s , two pairs of vectorlike @xmath38 and @xmath39 , two @xmath40 s and several @xmath41 s .
the mass matrices follow from froggatt - nielsen diagrams @xcite in which the superheavy fields are integrated out , with the flavor symmetry allowing only certain particular contributions to their mass matrix elements .
the dirac mass matrices for the up quarks , down quarks , neutrinos and charged leptons are found to be @xmath42 m_n = \left(\matrix { \eta & \delta_n & \delta'_n \cr \delta_n & 0 &
\epsilon \cr \delta'_n & - \epsilon & 1\cr } \right)m_u,\ & m_l = \left(\matrix { 0 & \delta & \delta ' e^{i \phi } \cr \delta & 0 & \sigma + \epsilon \cr \delta ' e^{i\phi } & - \epsilon & 1\cr } \right)m_d.\\
\end{array } \label{eq : dmatrices}\ ] ] in the convention where the left - handed fields label the rows and the left - handed conjugate fields label the columns .
the entries in the first row and first column are the most uncertain , especially for the up quark and dirac neutrino mass matrices , for they arise from higher order terms involving several integrations out of the massive neutrino lines . as originally proposed , the parameters @xmath43 and @xmath44 were absent . in a later application of the model to leptogenesis @xcite
, we have observed that the baryon asymmetry lies closer to the preferred value of @xmath45 @xcite , if we allow @xmath43 and @xmath44 to take on nonzero values in the dirac neutrino matrix @xmath46 , which are small enough in @xmath47 , however , so as not to upset the good results in the quark sector .
we shall consider both possibilities for these two parameters in the searches to be presented .
the lopsided nature of the down quark and charged lepton mass matrices arises from the dominance of the @xmath48 @xmath49 contribution to the electroweak symmetry breaking over that of the @xmath29 .
the pronounced lopsidedness with @xmath50 readily explains the small @xmath51 quark mixing and near maximal @xmath52 atmospheric neutrino mixing for any reasonable right - handed majorana neutrino mass matrix , @xmath53 .
all nine quark and charged lepton masses , plus the three ckm angles and cp phase , are well - fitted with the original eight input parameters defined at the gut scale to fit the low scale observables after evolution downward from @xmath54 gev @xcite : @xmath55 the effective light neutrino mass matrix , @xmath56 , is obtained from the conventional type i seesaw mechanism @xcite once the right - handed majorana mass matrix , @xmath53 , is specified .
while the large atmospheric neutrino mixing @xmath57 arises primarily from the structure of the charged lepton mass matrix @xmath58 , the structure of the right - handed majorana mass matrix @xmath53 determines the type of @xmath59 solar neutrino mixing , so that the solar and atmospheric mixings appear to be essentially decoupled in the model .
the lma solar neutrino solution is obtained with a special form of @xmath53 @xcite , which can be explained by the structure of the froggatt - nielsen diagrams : @xmath60 where the parameters @xmath61 and @xmath62 are those introduced in eq .
( [ eq : dmatrices ] ) for the dirac sector .
the vanishing @xmath63 subdeterminant of @xmath53 arises naturally if the yukawa couplings are universal at the gut scale .
the new parameters @xmath64 and @xmath65 are undetermined but expected to be of @xmath66 , since the first row and first column of @xmath53 have been properly scaled by powers of the very small @xmath62 parameter . as such
, the right - handed majorana mass matrix exhibits a very strong hierarchy .
given the right - handed majorana mass matrix above and @xmath67 , the seesaw formula results in @xmath68 & = & -\left(\matrix { 0 & \frac{1}{a - b } \epsilon & 0\cr \frac{1}{a - b } \epsilon & \frac{b^2-c^2}{(a - b)^2 } \epsilon^2 & \frac{b}{b - a } \epsilon\cr 0 & \frac{b}{b - a } \epsilon & 1\cr } \right)m^2_u/\lambda_r .
\end{array } \label{eq : mnui}\ ] ] note that the very strong hierarchy in @xmath53 can nearly balance the large hierarchy in @xmath46 to yield a rather mild mass hierarchy for the @xmath56 light neutrino mass matrix .
in fact , if @xmath69 were to hold , which has a texture similar to that in eq .
( [ eq : mr ] ) , the seesaw formula would yield a light neutrino mass matrix exactly proportional to the identity matrix resulting in a three - fold mass degeneracy .
the author has argued in ref .
@xcite that a normal hierarchy solution is much more stable than an inverted hierarchy solution for such type i seesaw models , and our results apply for this kind of solution .
barr has recently proposed more general possibilities which result in such a mild neutrino mass hierarchy @xcite .
the parameter @xmath70 can be chosen real and to a large degree sets the scale for the atmospheric neutrino mixing mass squared difference . in most previous studies of the model , aside from that of leptogenesis , we chose the remaining three parameters , @xmath64 and @xmath65 , to be real .
in order to determine the range of predictions for the neutrino mixing angles , we now allow all three of them to be complex .
in order to carry out the search of the parameter space for a stable normal hierarchy solution , we shall require that the oscillation parameters determined by the model lie in the following presently allowed ranges at the 90% confidence level @xcite-@xcite , @xcite : @xmath71 \delta m^2_{21 } & = ( 7.4 - 8.5 ) \times 10^{-5}\ { \rm ev^2 } , & \quad \tan^2 \theta_{12 } & = 0.33 - 0.50,\\[0.1 in ] \delta m^2_{31 } & \simeq \delta m^2_{32 } , & \quad \sin^2 2\theta_{13 } & \leq 0.16.\\ \label{eq : nudata } \end{array}\ ] ] the allowed variation in @xmath11 corresponds to the present bound of @xmath72 @xcite .
we find that the following bounds for the magnitudes of the input parameters in @xmath53 cover the above experimentally allowed ranges : @xmath73 where we have set @xmath74 gev as explained below .
we then use monte carlo techniques to throw points in the complex @xmath64 and @xmath65 planes which uniformly cover the disks whose radii are indicated above .
it should be noted that in this paper , we shall neglect any running of the neutrino mass and mixing parameters from the gut scale down to the low scales .
this has become more or less standard procedure in the literature and is less susceptible to corrections when @xmath75 is low , the neutrino mass hierarchy is normal , and the two low - lying neutrino mass eigenstates have nearly opposite @xmath76-parities @xcite .
all three conditions hold in the model under consideration .
in fact , we can use the convenient add - on program package for mathematica called `` reap '' provided by antusch et al .
@xcite to test the assumption that the evolution effects are negligible for our case . for this purpose , what is required in addition to the mass matrices in eq .
( [ eq : dmatrices ] ) is specification of the appropriate higgs vevs , @xmath77 and @xmath78 in eq .
( [ eq : higgsmixings ] ) , together with the two input scaling masses , @xmath79 and @xmath80 ; from these the corresponding yukawa coupling matrices can be determined . since @xmath81 receives a contribution only from @xmath82 , the 33 element of the dirac matrix @xmath46
is given by @xmath83 where @xmath84 is the yukawa coupling matrix for the dirac neutrino mass matrix and @xmath85 with @xmath86 gev as usual .
on the other hand , @xmath87 has two contributions from @xmath88 and @xmath89 , but only the first contributes to the 33 element of the charged lepton mass matrix @xmath58 ; hence @xmath90 where @xmath91 is the yukawa coupling matrix for the charged leptons and @xmath92 is the projection of @xmath37 onto the vev of @xmath93 . with @xmath94 as suggested earlier to provide a sufficiently lopsided charged lepton mass matrix , and the values for @xmath79 and @xmath80 in eq .
( [ eq : inputparam ] ) , we find that @xmath95 y'_l & \equiv & y_l \cos \gamma = 0.0293 ( m_l / m_d ) , \end{array } \label{eq : yukawas}\ ] ] where we have defined @xmath96 as the effective charged lepton yukawa coupling matrix due to the @xmath97 suppression .
we assume that the vevs , @xmath98 and @xmath99 , are held constant while @xmath84 and @xmath96 evolve from the gut scale to the weak scale according to the renormalization group equations . by inserting this information into the reap package of antusch et al .
@xcite , we can obtain the evolved results for the neutrino mixing parameters .
we find for various test cases that the evolved mixing angles and phases are changed by less than @xmath100 from their gut scale values , while the light neutrino masses are reduced by @xmath101 .
the latter discrepancy can be corrected if we simply reduce the right - handed majorana scaling factor @xmath70 by @xmath101 from @xmath102 gev to @xmath103 gev , with essentially no alteration in the mixing angles and phases . with this understanding of the role played by evolution of the yukawa couplings , we simply ignore the evolution and proceed to use the gut scale parameters given earlier .
scatter plots are shown in fig . 1 for the original choice of @xmath104 and @xmath105 . the atmospheric and solar neutrino mixing regions are shown in figs .
1(a ) and ( b ) , where @xmath14 and @xmath15 are plotted against @xmath106 and @xmath107 , respectively , the atmospheric and solar mixing angles in terms of the neutrino mixing matrix elements . for the atmospheric neutrino mixing , we observe from fig .
1(a ) that while @xmath70 sets the scale for the mass squared difference , there is a spread of points for @xmath14 in the allowed interval @xmath108 ; cf .
a value of @xmath74 gev has been chosen , in order to center the allowed region predicted by the model on the experimentally allowed region as determined by the super kamiokande data . for their best fit point of @xmath109 and @xmath110
, the model prefers a nearly maximal mixing angle of @xmath111 , in agreement with their finding .
note that the highest concentration of model points also occurs in this region surrounding their best fit point .
for the solar neutrino mixing depicted in fig .
1(b ) , we observe a nearly uniform spread in @xmath15 over the region allowed in eq .
( [ eq : nudata ] ) by the present data . on the other hand , for the solar mixing angle
there is a slightly greater concentration of points toward the smaller allowed values for @xmath112 .
although it has been customary to present the neutrino mixing data in @xmath113 vs. @xmath114 mixing planes as in figs .
1(a ) and ( b ) , some authors @xcite have suggested that @xmath115 is a more sensitive measure of the mixing angle , especially for angles close to a maximal mixing of @xmath116 .
of particular interest is whether @xmath12 lies below @xmath116 or above @xmath116 , if the mixing is not exactly maximal . with a normal mass hierarchy as suggested by our model , the former signifies that the highest neutrino mass eigenstate corresponds more to a @xmath117 flavor composition rather than to a @xmath118 composition , while the opposite is true for the latter possibility . in figs .
1(c)-(f ) we also present results for ( c ) @xmath119 , ( d ) @xmath120 , ( e ) @xmath121 , and ( f ) the baryon asymmetry , @xmath9 , vs. @xmath11 .
we see from these plots that there is just a slight preference for values of @xmath11 greater than 0.5 than for those less than 0.5 . concerning the solar neutrino mixing angle distribution in fig .
1(c ) , the full range of @xmath122 is populated , but values below 0.4 are slightly preferred to those above 0.4 in keeping with remarks made about fig .
1(b ) . nearly all the allowed points for @xmath120 in fig .
1(d ) are found to lie below 0.012 and above @xmath123 , with those values above 0.001 most densely populated .
interestingly enough , the present best fits for @xmath120 are 0.006 and 0.004 from three - neutrino mixing analyses with all oscillation data considered @xcite .
the model seems to be rather insensitive to the leptonic cp - violating phase @xmath121 in this more general search of the complex values for @xmath64 and @xmath65 , for the full range from @xmath124 to @xmath125 is essentially uniformly populated in fig .
finally we note that values for the baryon asymmetry in fig .
1(f ) are concentrated in the @xmath126 range , which are much too small to explain the observed baryon asymmetry in the universe . in fig .
2 , the corresponding scatter plots are presented for the choice @xmath127 and @xmath128 .
this latter choice has been made as a result of a protracted search , so as to maximize the calculated baryon asymmetry as we shall see , without spoiling the results in the quark sector . in the previous study reported in the last paper of reference @xcite , the corresponding values taken were @xmath129 and @xmath130 .
but without altering the parameter , @xmath62 , we observed a maximum of @xmath9 which fell short of the presently observed baryon asymmetry .
the atmospheric neutrino mixing plane in fig .
2(a ) exhibits a somewhat stronger preference for lower values of @xmath14 than in fig .
1(a ) , while the solar mixing angle distributions in figs .
2(b ) and ( c ) exhibit a slightly greater preference for smaller solar mixing angles . from fig .
2(d ) we see the upper bound on @xmath120 is predicted to lie lower than before in fig .
1(d ) and is now given by @xmath131 .
the distribution is also somewhat more diffuse and less concentrated near the upper limit .
again no preferred values of the cp - violating phase @xmath121 are observed in fig .
2(e ) , but the baryon asymmetry is shifted to higher values in the broader range , @xmath132 to @xmath133 in fig .
the scatter plots presented in figs . 1 and 2 were obtained by throwing 400,000 sets of points for the parameters @xmath64 and @xmath65 in the ranges given in eq .
( [ eq : bounds ] ) and demanding that the calculated mixing parameters lie in the experimental ranges quoted in eq .
( [ eq : nudata ] ) .
but the question arises whether more sparsely scattered points can be obtained outside the apparent boundaries .
this is of special interest for the baryon asymmetry , where in fact one isolated point in fig .
2(f ) does show up with @xmath134 , well outside the major clustering but still below the observed value of @xmath135 . rather than throw 10 - 100 times as many points , we searched for @xmath53 input parameters which favor larger values of @xmath9 .
we found such a region defined by constraining the parameters @xmath64 and @xmath65 to lie near or along their imaginary axes in the complex planes , still bounded by the magnitudes given in eq .
( [ eq : bounds ] ) . in figs .
3(a ) and ( b ) , we present the scatter plots in the atmospheric and solar neutrino mixing planes obtained by throwing 100,000 sets of points for positive and negative imaginary values of @xmath64 and @xmath65 bounded by the magnitudes of eq .
( [ eq : bounds ] ) , again for the new values of @xmath136 and @xmath44 . while the solar mixing plane of fig .
3(b ) is very similar to that of fig .
2(b ) , the atmospheric mixing plane in fig .
3(a ) is dramatically affected as the spread of points now lie along the arc shown .
the upper branch corresponds to values of @xmath137 and the lower branch to values of @xmath138 .
the small secondary upper branch is obtained when @xmath139 , while the major upper branch is found when @xmath140 . further insight into the restricted results are obtained from figs . 3(c)-(f ) .
bands develop in all four plots of ( c ) @xmath119 , ( d ) @xmath120 , ( e ) @xmath121 , and ( f ) @xmath9 vs. @xmath11 .
one of the most striking results is that @xmath120 is now limited to the restricted range ( 0.0018 , 0.0035 ) . as seen in fig .
3(f ) , 29 points have appeared in the @xmath141 range . in fig .
4 similar plots for the mixing planes and distributions are shown , where 200,000 sets of points for imaginary values of @xmath64 and @xmath65 were thrown with the restriction @xmath142 applied . in all 84 points survive this cut .
we now see that the upper branch in the atmospheric neutrino mixing plane of fig .
4(a ) is essentially eliminated , for @xmath143 is required for most points to obtain such large values of @xmath9 .
the scatter points in the solar neutrino mixing plane is little affected aside from the density of points as seen in fig .
4(b ) . the bands in figs .
3(c ) , ( d ) and ( e ) become extremely narrow in figs .
4(c ) , ( d ) , and ( e ) when the cut @xmath142 is imposed . a strong correlation arises between the solar mixing angle in @xmath119 and the atmospheric mixing angle in @xmath11 ; the larger the former , the smaller the latter .
the upper bands in figs .
4(d ) and ( e ) are obtained when both parameters @xmath144 and @xmath145 have positive imaginary values and yield a positive dirac phase @xmath121 , while the lower bands arise from negative imaginary values for @xmath144 and @xmath145 .
note that for the restriction @xmath142 , the leptonic cp phase lies in the two ranges , @xmath146 . the distribution of points in @xmath9 and @xmath11 which survive the cut of @xmath142 is shown in fig .
several points with @xmath147 are observed to be clustered around @xmath148 .
for these few points , values of @xmath149 seem to be preferred .
this suggested that we construct scatter plots with all three parameters , @xmath64 and @xmath65 taken to be imaginary with the restriction @xmath150 .
the corresponding plots are shown in fig . 5 , where 100,000 sets of points are thrown and 381 points survive .
it is rather remarkable that the atmospheric neutrino mixing plane in fig .
5(a ) is restricted to a short arc centered at @xmath151 and with @xmath152 lying in the range ( 0.985 , 0.995 ) , as @xmath153 with @xmath154 for our choice of @xmath70 .
the solar mixing plane in fig .
5(b ) is also somewhat truncated at the lower end of the spectrum as now @xmath155 . the bands in figs
. 5(c)-(e ) are similar to those in fig .
4(c)-(e ) but again somewhat truncated by the maximum value of @xmath12 .
the relevant values of @xmath120 now lie in the interval ( 0.0020 , 0.0031 ) .
the cp phase is bounded by @xmath156 in fig .
5(f ) . with the strong correlation of the mixing angles noted above
, @xmath9 is now directly related to the mixing angle @xmath12 in fig .
5(f ) , as the random scattering in fig .
4(f ) has disappeared .
the observed baryon asymmetry value of @xmath135 is obtained for @xmath157 , which implies @xmath158 , @xmath159 and @xmath160 . for this case the sum of @xmath13 and the cabibbo angle is given by @xmath161 , very close to the value of @xmath116 which is referred to as quark - lepton complementarity in the literature @xcite . on the other hand , maximal atmospheric neutrino mixing with @xmath162 corresponds to @xmath163 , @xmath164 , @xmath165 , and @xmath166 . as an example of a complete solution
, we present the results for a special case which arises for the choice of right - handed majorana neutrino parameters : @xmath167 with @xmath168 gev , @xmath169 , and @xmath170 .
one obtains @xmath171 with the following mixing parameters @xmath172 where @xmath173 and @xmath174 are the majorana phases in the majorana phase matrix , @xmath175 . for this case
, @xmath176 mev , @xmath177 mev , @xmath178 mev , @xmath179 gev , @xmath180 gev , @xmath181 gev , and @xmath182 gev where @xmath183 is the width of the lightest right - handed state .
further study reveals that the narrow band structures in fig . 5 , obtained with @xmath64 and @xmath65 pure imaginary and @xmath150 , arise because @xmath145 is limited to the very narrow ranges , @xmath184 , for all 381 points which satisfy the mixing parameter constraints imposed in eq .
( [ eq : nudata ] ) . on the other hand ,
the ratio @xmath185 varies in the range @xmath186 along each band as @xmath11 decreases from 0.55 to 0.45 .
the pair of slightly displaced bands in fig .
5(d ) is associated with the choices of positive or negative imaginary values for @xmath144 and @xmath145 .
clearly the imposition of these restrictions for the @xmath187 parameter set has greatly enhanced the probability that the baryon asymmetry will lie in or near to its observed value .
this makes for a rather compelling scenario within the model .
the question arises as to how likely is it that such restricted values should apply for the majorana mass parameters .
it is easy to see from the froggatt - nielsen diagrams for the majorana mass matrix that if only one higgs vev is responsible for breaking lepton number , by the @xmath188 family symmetry of the model only one diagram will contribute to each matrix element , the structure of @xmath53 will be purely geometrical and its determinant will vanish , cf .
ref . @xcite . with @xmath189
, only one diagram contributes to each of the 11 , 13 , and 31 matrix elements , while two or more diagrams will be required for the 12 and 21 elements .
this is the simplest possibility for a nonvanishing determinant , as required for the seesaw mechanism .
the fact that all three parameters are purely imaginary makes for a maximum cp violation in the right - handed majorana mass matrix as all other parameters are real . to understand the apparent difficulty in reaching values in the observed range of @xmath45 , we note the following .
the scenario of resonant leptogenesis @xcite arises naturally in the model , for the two heavy right - handed majorana neutrinos have nearly equal masses and nearly opposite cp - parity .
however , one must impose the condition that the mass difference of these two heavy majorana states be at least as large as the half - width of either state for the resonance formula to make sense .
this forced us to raise the magnitudes of @xmath136 and @xmath44 in the dirac neutrino mass matrix . for smaller values of these parameters
than those chosen in the relevant figures , fewer points reach the @xmath190 range of @xmath9 , as the range of @xmath11 is shifted lower with more points occuring below 0.50 than above 0.50 .
it is interesting that in going from the set of matrices with the original choice of @xmath136 and @xmath44 in fig . 1 to the new choice of values in figs . 2 - 5 , the @xmath191 mixing parameter in the quark mixing matrix
is raised from 0.2220 to 0.2240 , in better keeping with the new observed value which is preferred for unitarity reasons . the one negative effect in making this change
is that the up quark mass is raised from 3.5 mev to 5.5 mev after evolution from the gut scale , which is above the preferred range .
while the condition for thermal resonant leptogenesis is easily satisfied for such right - handed majorana masses as in the example given above , the overproduction of gravitinos in the early universe is a problem in the supergravity scenario of susy breaking ; this can be alleviated if the susy breaking occurs via the gauge - mediated scenario @xcite . as illustrated in ref .
@xcite , this problem can be completely avoided and satisfactory leptogenesis and baryon asymmetry obtained , if the model is expanded to include three additional intermediate mass scale singlets which lead to a double seesaw mechanism . however , with this expanded seesaw mechanism many more parameters are introduced , and the model becomes much less predictive .
from the above results we can conclude that the model with a normal mass hierarchy can easily fit the presently observed neutrino mixing data .
as originally proposed , eight of the twelve model parameters have been fixed by the quark and lepton masses and quark mixing data , while the remaining four ( three complex and one real ) parameters of the right - handed majorana mass matrix are varied .
the real @xmath70 parameter centers the @xmath14 model spectrum on the observed @xmath192 confidence level interval given in eq .
( [ eq : nudata ] ) . for the best fit point of @xmath193 ,
the predicted atmospheric neutrino mixing angle is nearly maximal , corresponding to @xmath194 in agreement with the observed best fit value of 1.00 .
no such preference for a particular value of @xmath119 in the allowed range is found .
the predicted values for @xmath120 are of special interest .
there we found a range of @xmath195 which lies noticeably lower than that predicted by most other models where values very close to the upper bound of 0.04 are more generally expected @xcite .
this is true not only for the nearly conserved @xmath17 models , but also for the @xmath0 models with symmetric @xmath196 and @xmath197 higgs representations .
the lower values predicted in our model have relevance for the future study of neutrino oscillations with reactor beams .
if @xmath198 disappearance is observed very close to the present chooz bound by the next generation reactor experiments , the @xmath0 model under consideration will be ruled out . on the other hand ,
it is interesting to note that if future reactor experiments do not see some @xmath198 disappearance , the model predicts it will certainly be seen with superbeam and/or neutrino factory experiments .
while there remains much interest in determining the dirac phase @xmath121 for @xmath76 violation , it is rather surprising that the model is able to accommodate almost any value . with the original dirac neutrino mass parameters , the resonance leptogenesis inherent in the model generated a baryon asymmetry which falls more than two orders of magnitude short of the observed value . in an effort to increase the effect ,
the original very small @xmath62 parameter was adjusted and two new equally small parameters , @xmath43 and @xmath44 , were added to the dirac neutrino mass matrix . this helped to recover two missing orders of magnitude for @xmath9 , provided the three complex parameters in the right - handed majorana mass matrix were chosen very near to or on their imaginary axes . by selecting only those points which satisfied the observed neutrino mixing results and
for which @xmath199 , we observed that the mixing angles lie along very narrow bands .
these bands were further dramatically tightened by choosing not only @xmath64 and @xmath65 purely imaginary , but restricting @xmath65 to be equal to @xmath144 .
for this situation a limited range of @xmath153 was found over which @xmath119 falls from 0.50 to 0.38 , @xmath200 varies from 0.0031 to 0.0020 , and the dirac cp phase varies from @xmath201 . on the other hand , a narrow band also developed for the baryon asymmetry due to the strong mixing angle and cp phase correlations , as it rises from @xmath202 with increasing @xmath11 .
the appearance of such narrow bands is directly related to the very narrow acceptable range for @xmath145 , while the ratio @xmath185 varies along the length of each band .
we argued that the @xmath150 restriction can be understood in terms of one froggatt - nielsen diagram for each of the 11 , 13 , and 31 matrix elements of @xmath53 , while two such diagrams can contribute to the 12 and 21 elements .
all these results are of considerable interest and will severely test the model in the future .
since the light neutrino mass spectrum has a normal hierarchy , the model predicts no neutrinoless double beta decay should be observed in the near future .
in fact , for all choices of the model parameters which lead to acceptable ranges for the mixing parameters of eq .
( [ eq : nudata ] ) , we find the effective mass is @xmath203 @xcite , three orders of magnitude below the present bound @xcite . in conclusion ,
we address briefly the issue of lepton flavor violation in the decay process @xmath25 .
jankowski and maybury @xcite have studied this particular class of @xmath0 models with lopsided charged lepton mass matrices in order to determine the expected branching ratio for the above decay , under the assumption that the supersymmetric grand unified model breaks directly to the constrained minimal supersymmetric standard model . given the combined constraints on the cmssm parameters from direct searches and from the wmap satellite observations
@xcite , they find that for a low value of @xmath50 preferred by the model , the branching ratio should be of @xmath204(@xmath205 ) and very close to the present upper limit of @xmath206 @xcite . hence the @xmath207 decay should be observed in the new meg experiment @xcite which is expected to start taking data in 2006 .
this prediction is based on the slepton masses being in the 125 - 150 gev range , while the neutralino masses are around 300 - 500 gev .
these ranges are also favored in cmssm models by the recent study of the best chi - squared fits to all the data with @xmath208 by ellis et al .
on the other hand , minimal @xmath0 models leading to symmetric or antisymmetric mass matrix elements favor a smaller lepton - violating branching ratio of @xmath204(@xmath209 ) or less @xcite .
the coming meg experiment which is expected to be sensitive down to the @xmath210 branching ratio level may thus be able to rule out one of the two classes of minimal @xmath0 models before the new reactor experiments can be launched .
+ the author thanks uli haisch and enrico lunghi for their help with the scatter plots .
the author also thanks the theory group at fermilab for its kind hospitality .
fermilab is operated by universities research association inc . under contract
de - ac02 - 76ch03000 with the department of energy .
ahmed et al .
( sno collab . ) , phys .
lett . * 92 * , 181301 ( 2004 ) ; b.t .
cleveland et al .
( homestake collab . ) , astrophys .
j. * 496 * , 505 ( 1998 ) ; j.n .
abdurashitov et al .
( sage collab . ) , phys .
lett . * 83 * , 4686 ( 1999 ) ; w. hampel et al .
( gallex collab . ) , phys .
b * 447 * , 127 ( 1999 ) ; m. altmann et al .
( gno collab . ) , phys .
b * 490 * , 16 ( 2000 ) . for several reviews , see s.f . king , rept .
. phys . * 67 * , 107 ( 2004 ) ; g. altarelli and f. feruglio , new j. phys .
* 6 * , 106 ( 2004 ) ; r.n .
mohapatra , new j. phys .
* 6 * , 82 ( 2004 ) ; s.m .
barr and i. dorsner , nucl .
b * 585 * , 79 ( 2000 ) . for the earliest and some of the latest such models cf .
s. petcov , phys.lett .
b * 110 * , 245 ( 1982 ) ; r. barbieri , l. hall , d. smith , a. strumia , and n. weiner , jhep * 9812 * , 017 ( 1998 ) ; a. joshipura and s. rindani , eur .
j. c * 14 * , 85 ( 2000 ) ; r.n .
mohapatra , a. perez - lorenzana , and c. de s. pires , phys .
b * 474 * , 355 ( 2000 ) ; s.t .
petcov and w. rodejohann , phys .
d * 71 * , 073002 ( 2005 ) ; w. grimus and l. lavoura , j. phys .
g * 31 * , 683 ( 2005 ) .
babu and r.n .
mohapatra , phys .
lett . * 70 * , 2845 ( 1993 ) ; mu - c .
chen and k.t .
mahanthappa , phys .
d * 62 * , 113007 ( 2000 ) ; t. fukuyama and n. okada , jhep * 0211 * , 011 ( 2002 ) ; h.s .
goh , r.n .
mohapatra , and s.p .
ng , phys . lett .
b * 570 * , 215 ( 2003 ) ; m. bando and m. obara , prog .
. phys . * 109 * , 995 ( 2003 ) ; g.g .
ross and l. velasco - sevilla , nucl .
b * 653 * , 3 ( 2003 ) ; c.s .
aulakh , b. bajc , a. melfo , g. senjanovic , and f. vissani , phys .
b * 588 * , 196 ( 2004 ) ;
m. bando , s. kaneko , m. obara , and m. tanimoto , phys .
b * 580 * , 229 ( 2004 ) ; b. dutta , y. mimura , and r.n .
mohapatra , phys .
b * 603 * , 35 ( 2004 ) .
c.h . albright and s.m .
barr , phys .
d * 58 * , 013002 ( 1998 ) ; k.s .
babu , j.c .
pati , and f. wilczek , nucl .
b * 566 * , 33 ( 2000 ) ; t. blazek , s. raby , and k. tobe , phys .
d * 62 * , 055001 ( 2000 ) ; r. kitano and y. mimura , phys .
d * 63 * , 016008 ( 2001 ) ; z. berezhiani and a. rossi , nucl .
b * 594 * , 113 ( 2001 ) .
p. minkowski , phys .
b * 67 * , 421 ( 1977 ) ; t. yanagida , in _ proceedings of the workshop on unified theory and the baryon number of the universe _ , tsukuba , japan , 1979 , p. 95 ; m. gell - mann , p. ramond , and r. slansky , in _ supergravity _ ,
( p. van nieuwenhuizen and d.z .
freedman , eds . ) , north holland , amsterdam , 1979 , p. 315 ; s.l .
glashow , in _ proceedings of the 1979 cargese summer institute on quarks and leptons _ , ( m. levy , j .- l .
basdevant , d. speiser , j. weyers , r. gastmans , and m. jacob , eds . ) , plenum press , new york , 1980 , p. 687 ; r.n .
mohapatra and g. senjanovic , phys .
. lett . * 44 * , 912 ( 1980 ) ; j. schechter and j.w.f .
valle , phys .
d * 22 * , 2227 ( 1980 ) . | the bounds on the neutrino mixing angles and cp dirac phase for an @xmath0 model with lopsided mass matrices , arising from the presence of @xmath1 and @xmath2 higgs representations , are studied by variation of the one real and three unknown complex input parameters for the right - handed majorana neutrino mass matrix .
the scatter plots obtained favor nearly maximal atmospheric neutrino mixing , while the reactor neutrino mixing lies in the range @xmath3 with values greater than @xmath4 most densely populated . a rather compelling scenario within the model follows , if we restrict the three unknown complex parameters to their imaginary axes and set two of them equal .
we then find the scatter plots are reduced to narrow bands , as the mixing angles and cp phase become highly correlated and predictive .
the bounds on the mixing angles and phase then become @xmath5 , @xmath6 , @xmath7 , and @xmath8 . moreover , successful leptogenesis and subsequent baryogenesis are also obtained , with @xmath9 increasing from @xmath10 as @xmath11 increases from 0.45 to 0.55 .
fermilab - pub-05 - 021-t + hep - ph/0502161 + may 2005 + |
quasars have been used as efficient probes of high - redshift clustering because they are known to favour rich environments .
quasars may also trace large - scale structures at early epochs ( @xmath5 ) in the form of large quasar groups ( lqgs ) , which have comparable sizes to the largest structures at the present epoch .
the largest of these is the clowes - campusano lqg of 18 quasars at @xmath6 , with a maximal extent of 200@xmath7mpc ( clowes & campusano 1991 ) . to examine the galaxy environment of quasars in this lqg ,
we have conducted an ultra - deep optical ( to @xmath8 ) study of a @xmath9 field containing 3 quasars from this lqg , with additional @xmath10 imaging obtained for subfields around 2 of them .
any associated passively - evolving galaxies can be identified from their extremely - red optical - ir colours ( @xmath11 ) .
we find a 3.5@xmath12 excess of @xmath13 galaxies in the @xmath14 field centred on the z=1.226 lqg quasar 104420.8 + 055739 ( haines et al . 2000 ) .
this excess is due entirely to a factor @xmath1511 overdensity of @xmath2 galaxies , which must have @xmath16 to explain their colour .
we observe clear red sequences at and @xmath17 ( figure 1 ) , comparable to red sequences observed in other @xmath18 clusters , and indicating a population of 1518 passively - evolving massive ellipticals at the quasar redshift .
we have produced a density map ( figure 2 ) of galaxies with the same optical colours and magnitudes as the red sequence galaxies to estimate the full extent of any associated clustering beyond the @xmath10 image ( shown as box ) .
there is clearly significant substructure , with the quasar located between two groups of red sequence galaxies and further clustering can be seen extending over the whole structure is suggestive of being a cluster in the early stages of formation through the coalescence of subclusters .
the quasar is located within a concentration of blue ( @xmath4 ) galaxies , with @xmath19 clearly at @xmath20 , having @xmath21 , indicating a region of enhanced star - formation , similar to those found around several @xmath22 lqg quasars ( hutchings , crampton & johnson 1995 ) .
the concentration appears extended to form a _ band _ which bisects the two groups of red sequence galaxies .
we propose that this band of recent star - formation , and the quasar itself , have been triggered by the merging of the two groups , as the galaxies interact with the colliding icms .
the two other lqg quasars are in poor environments , although a background @xmath23 quasar lies on a cluster periphery .
quasars at these redshifts ( @xmath24 ) are found either on the peripheries of clusters ( see also sanchz & gonzlez - serrano 1999 ) or in relatively undistinguished environments .
clowes , r. g. , & campusano , l. e. 1991 , , 249 , 218 haines , c. p. , clowes , r. g. , campusano , l. e. , & adamson , a. j. 2000 , + submitted hutchings , j. b. , crampton , d. , & johnson , a. 1995 , , 109 , 73 sanchz , s. f. , & gonzlez - serrano , j.i .
1999 , , 352 , 383 | we report significant associated clustering in the field of a @xmath0 quasar from the clowes - campusano lqg in the form of both a factor @xmath1 overdensity of @xmath2 galaxies , and red sequences of 1518 galaxies at @xmath3 indicative of a population of massive ellipticals at the quasar redshift .
the quasar is located between two groups of these galaxies , with further clustering extending over .
a band of @xmath4 galaxies bisects the two groups of red sequence galaxies , and we suggest that the merging of these two groups has triggered both this band of star - formation and the quasar . # 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
since the discovery of the first extrasolar planet around a solar - type star , 51 pegasi by @xcite , by now ( october 2006 ) 209 exoplanets have been found .
these planets form 179 exoplanetary systems , among which there are 21 multiple systems .
the main detection technique of exoplanets is based on the doppler shift of the hosting star s spectral lines .
this method is adequate to discover giant planets , but it is not yet capable to detect earth - sized planets .
however , the detection and investigation of earth - like exoplanets can be of great importance for the further improvements of formation theories of our solar system , where both rocky earth - like , and gaseous jupiter - like planets have been formed .
another major question which has a special interest not only for the scientific community is whether life may have evolved in some exoplanetary systems .
the region of an exoplanetary system , where water - based life might evolve , is called the habitable zone ( hz ) .
simply saying , the hz of a planetary system is that region , where liquid water can exist on the surface of a planet @xcite .
the dynamical stability of hypothetical earth - like planets in the hzs of several extrasolar planetary systems has been the subject of intensive research ( see for instance * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
according to these studies many systems can host terrestrial exoplanets in their hzs .
recently , there have been much efforts for the construction of space probes aiming at the detection of terrestrial exoplanets in the near future .
such is corot , to be launched in november 2006 , and kepler with a launch in 2008 .
these missions will use transit photometry as detection technique .
the transit of an unseen exoplanet in front of the hosting star s disc is expected to result in an observable dimming of the star s light intensity .
there are also other plans , such as nasa s tpf and sim , or esa s darwin and gaia missions in the next decade aiming at studying all aspects of exoplanets . by using data obtained from the observations ,
the orbital parameters of an exoplanet can be derived only within some error limits . in order to improve the orbital parameters
certain constraints should be used . for instance , if an earth - like planet were discovered in an extrasolar planetary system , where an already detected giant planet is known to orbit the star , such a constraint could be the dynamical stability of the terrestrial planet .
this means , that those orbital solutions for the terrestrial planet , which result in unstable orbits , are unlikely and therefore should be excluded .
the dynamical stability of earth - like planets can be established in several ways , for example ( i ) by exploring the stable and unstable regions of the phase space of each extrasolar planetary system separately , or ( ii ) by using stability maps computed in advance for a large set of orbital parameters . in this paper
we present such stability maps and show how to apply them to the detected exoplanetary systems .
the second method has the advantage that the stability properties of a terrestrial planet can be easily established when the orbital parameters of the giant planet of an exoplanetary system are modified .
very often the orbital parameters of the giant planets are uncertain , and due to the accumulation and improvements of the observational data , they are changed quite frequently .
if the first approach is chosen , one has to re - explore the phase space of the individual exoplanetary system after each modification of the orbital parameters of the giant planet .
however , this is not necessary when the second method is used , since the stability properties of the investigated planetary system can be easily re - established from the already existing stability maps . as a first step towards general stability maps
, we compiled a stability catalogue for a dynamical model consisting of a star , a giant planet , and a small earth - like planet .
we present this catalogue in this paper .
section 2 describes the dynamical model and the applied numerical methods .
details of the catalogue and the stability maps are given in sections 3 and 4 .
section 5 shows how to apply the catalogue to individual exoplanetary systems . by using this catalogue , the dynamical reality of the orbital elements of a newly discovered terrestrial planet in an extrasolar planetary system
can be easily determined .
this catalogue can also be used to establish the stability properties of the hzs of known exoplanetary systems .
this is shown in section 6 .
the majority of planetary systems , detected up to now , consists of a star and a giant planet revolving about it in an eccentric orbit .
therefore we use a relatively simple dynamical model ( known as the elliptic restricted three - body problem ) , where two massive bodies , called the primaries , move in elliptic orbits about their common center of mass , and a third body of negligible mass moves under their gravitational influence ( for details see * ? ? ?
* ) . in our application
the primaries are a star and a giant planet , and the third body is a small earth - like planet , regarded as massless .
we note , however , that among the extrasolar planetary systems with long observational baselines , there is a high rate of multiple planet systems .
thus a more convenient model for our stability investigations would be the restricted @xmath1-body problem with @xmath2 giant planets ( @xmath3 ) .
the presence of additional giant planets may enhance the instabilities , induced by just one massive planet , and make the hz of a system more unstable .
the instabilities may appear due to mean motion resonances between the massive giant planet and the earth - like planet .
thus , by mapping these resonances , we can find the most significant birthplaces of instabilities in the habitable zones . on the other hand , the dynamical model with one giant planet
also offers the most convenient way to display the most important mean motion resonances as a function of the mass ratio of the star and the giant planet , and of the eccentricity of the giant planet . in this paper
we mapped these resonances , and the effects of additional giant planets will be studied in a future work . to explore the stable and unstable regions of the motion of the third body
, the equations of motion have to be integrated for a large set of initial conditions and for a long time for each orbit .
this can be very time consuming , even in the case of a huge computing potential . in order to study efficiently the stability properties of dynamical systems , several numerical techniques and chaos detection methods
have been developed recently . using the working hypothesis , that chaotic behaviour means instability through chaotic diffusion of trajectories in the phase space
, these methods can be applied to explore the stable and unstable regions of dynamical systems .
we are , however , aware of the fact , that chaotic behaviour does not necessarily mean instability . there could be orbits exhibiting chaotic behaviour , but stable in nekhoroshev - sense , e.g. remaining in a bounded region of the phase space for time intervals equal to the age of the investigated planetary system .
however , if other destabilizing effects , such as the presence of yet undetected giant planets are present in the exoplanetary systems , chaotic behaviour may result in really unstable orbits . in other words , in high dimensional dynamical systems unstable orbits
emanate only from the chaotic regions of the phase space . exploring these regions we can find the birth domains of unstable orbits .
however , in order to decide the stability character of an orbit , one has to use long - term numerical integration of the equations of motion . on the other hand , in the case of a large set of initial conditions
the fast chaos detection methods allow us to explore qualitatively the dynamical structure of the extrasolar planetary systems .
our investigations are based on such chaos detection methods . as a main tool we used the method of the relative lyapunov indicators ( rli ) @xcite , and in some cases the fast lyapunov indicators ( fli ) @xcite .
we also applied the maximum eccentricity method ( mem ) ( see for instance * ? ? ?
* ) , as an independent check of the rli and the fli by direct numerical integration of the equations of motion .
the knowledge of the eccentricity of stable orbits is essential for the problem of the development of a biosphere on a terrestrial planet in the habitable zone of a star .
for example , recent estimations for the maximum eccentricity for the earth after detailed atmosphere models
are @xmath4 . regarding the rli and the fli , our choice is motivated by the complementary character of the two chaos detection methods .
the rli measures the difference between the convergence of the finite time lyapunov indicators to the maximal lyapunov characteristic exponent of two initially very close orbits .
this method is extremely fast in determining the ordered or chaotic nature of individual orbits , even to detect orbits with very small lyapunov characteristic exponents ( e.g. sticky orbits , * ? ? ?
* ) , as well as to distinguish between ordered and chaotic regions of the phase space .
according to our experiments , gained in different dynamical problems , it is enough to integrate the two initially close orbits for a few hundred times of the longest orbital period of the investigated system . as to the fli
, their time evolution shows clearly the ordered or chaotic character of an orbit , and the method is very effective in detecting stable resonant orbits .
concerning the computation of the fli , the integration time is several tens of thousands orbital periods . in the mem
we compute by numerical integration the largest value of the orbital eccentricity , reached by an orbit during a given , long time interval ( @xmath5 orbital periods in this paper ) .
these methods have been successfully applied to several exoplanetary systems to determine their dynamical stability properties : for the rli see @xcite , for the fli @xcite , and for the mem @xcite .
we compiled a catalogue of dynamical stability , adapted to a system consisting of a star with mass @xmath6 , a giant planet with mass @xmath7 , and a small test planet with negligible mass .
the total mass of the star and the giant planet is taken as unit mass , thus @xmath8 is the mass parameter of the problem , which has been changed between ( i ) @xmath9 and @xmath10 in steps of @xmath11 , ( ii ) between @xmath12 and @xmath13 in steps of @xmath14 , and ( iii ) between @xmath15 and @xmath16 in steps of @xmath17 .
this results in 23 values for @xmath18 .
the giant planet was assumed to move around the star in an elliptic orbit , whose plane is taken as reference plane , with semi - major axis @xmath19 , eccentricity @xmath20 , argument of periastron @xmath21 , and mean anomaly @xmath22 . without the loss of generality , the semi - major axis @xmath19 can be taken as unit distance , thus @xmath23 .
we assumed that @xmath24 .
the eccentricity @xmath20 was changed between 0.0 and 0.5 with a stepsize of @xmath25 . during our computations
we also changed @xmath22 between @xmath26 and @xmath27 with @xmath28 both for inner and outer orbits of the test planet ( with respect to the orbit of the giant planet ) .
additionally , we changed @xmath22 for inner orbits of the test planet between @xmath26 and @xmath29 with @xmath30 too .
we assumed that the test planet moves in the orbital plane of the giant planet , and we always took @xmath31 , @xmath32 , @xmath33 as initial orbital elements of the test planet .
the case when the orbit of the test planet is inclined with respect to the giant planet s orbit will also be studied in future work .
we changed the semi - major axis @xmath34 of the test planet in two different intervals : ( a ) for orbits of the test planet
inside " the orbit of the giant planet we changed @xmath34 between 0.1 and 0.9 with a stepsize of @xmath35 , and ( b ) for outside " orbits between 1.1 and 4.0 with a stepsize of @xmath36 .
the catalogue consists of 552 stability maps , corresponding to each ( @xmath37 ) pair of the 23 values of @xmath18 and 16 values of @xmath22 for inner orbits ( 368 maps ) and 8 values of @xmath22 for outer orbits ( 184 maps ) of the test planet . for
a given ( @xmath37 ) pair the stability map is prepared in such a way , that for each set of the described initial conditions the rli ( or fli ) value of the corresponding orbit of the test planet is computed and visualized in the ( @xmath38 ) parameter plane . due to the very small stepsize in @xmath34 and @xmath20 , each stability map corresponds to more than @xmath39 initial conditions , thus providing a very fine resolution . in the following section we show some stability maps from the catalogue , containing the hzs of some exoplanetary systems as examples .
the complete catalogue is available at ` http://astro.elte.hu/exocatalogue ` and ` http://www.astro.univie.ac.at/adg ` .
in fig . [ fig1 ] we displayed a case , where the semi - major axis of the test planet s orbit is larger than that of the giant planet ( @xmath40 , and @xmath41 , @xmath42 ) .
light regions correspond to low values of the rli ( @xmath43 ) , and thus to ordered , dynamically stable motion of the test planet .
dark shades correspond to large values of the rli ( @xmath44 ) and chaotic behaviour of the test planet .
the dark unstable regions on the left are due to the proximity of the giant planet .
the figure is dominated by v - shaped gray stripes , corresponding to outer mean motion resonances between the test planet and the giant planet .
these resonances , marked at the top of the figure , can represent either ordered ( stable for infinite time ) , or weakly chaotic ( which may become unstable after very long time ) behaviour . with the increase of @xmath20
many resonances overlap , giving rise to strongly chaotic and thus very unstable behaviour .
the reason for this is that by increasing @xmath20 the apocenter distance of the giant planet also increases , and the giant planet perturbs more strongly the outer test planet .
the figure also shows , for two planetary systems , the positions of the zero age hz , which we call the _ classical _ hz , and the _ present
_ hz . we recall , that the habitable zone ( hz ) is that region around a star , where water can exist in liquid phase on the surface of a planet by virtue of stellar irradiance .
the boundaries of the classical hz are calculated by assuming the physical properties of a zero age main sequence star , while the boundaries of the present hz depend also on the age of the hosting star @xcite .
the borders of the classical hzs for a large number of extrasolar planetary systems are given by @xcite , while the borders of the present hzs can be found in the more recent work of @xcite .
we note that according to the latter work the present hzs are shifted outwards with respect to the location of the classical hzs .
[ fig2 ] shows a stability map for the same initial conditions as in fig .
[ fig1 ] , but resulting from the fli computations .
it has the same overall structure , as the map obtained by the rli .
note , however , the complementary character of the results obtained by the two methods .
the rli detects the chaotic separatrices of the resonances ( dark v - shaped stripes in fig .
[ fig1 ] ) , while the fli finds the stable resonant orbits inside the resonances ( light stripes inside the resonances in fig .
[ fig2 ] ) .
note , that for the given initial conditions ( @xmath42 ) almost all resonances can have stable orbits , but the 1:3 resonance is chaotic according to both methods , a well known result for the 3:1 inner resonance @xcite .
[ fig3 ] has been calculated by the mem , again for the same initial conditions as in figs .
[ fig1 ] and [ fig2 ] .
it can be seen , that similarly to the rli and the fli , the mem also indicates the mean motion resonances . since in the mem orbits
were integrated for @xmath5 periods of the giant planet , we used it as a test of the two other chaos detection methods
. figs .
[ fig4 ] and [ fig5 ] show stability maps for outer orbits of the test planet , computed by the rli for @xmath45 and @xmath46 , respectively .
it can be seen , that for larger values of @xmath18 the chaotic regions are more extended , due to the stronger perturbing effect of the giant planet . in figs .
[ fig6 ] [fig10 ] we display cases where the semi - major axis of the test planet is smaller than that of the giant planet ( @xmath47 ) .
two panels for @xmath41 ( figs .
[ fig6]a and [ fig6]b ) show the results for two different starting positions of the giant planet , @xmath42 ( figs .
[ fig6]a ) and @xmath48 ( figs .
[ fig6]b ) . here
again mean motion resonances , this time inner ones , between the test planet and the giant planet dominate the stability maps .
inside the resonances the stable or chaotic behaviour of the test planet depends on the initial angular positions of the two planets .
this is clearly visible , when we compare the two panels . in fig .
[ fig6]a the giant planet starts from the pericenter , in fig .
[ fig6]b from the apocenter , while the initial position of the test planet in both cases is at @xmath49 , corresponding to the direction of the giant planet s pericenter . comparing the two figures
it can be seen that the location of the resonances does not change , since this depends on the ratio of the semi - major axes of the two planets .
however , the character of the resonances becomes different . in fig .
[ fig6]b several resonances ( 5:2 , 5:3 , 3:2 ) are more stable than in fig .
this is due to the fact that the relative initial positions determine the places of conjunctions of the two planets .
if they meet regularly near the pericentre of the giant planet , the motion of the test planet becomes chaotic , while it can remain regular if the conjunctions take place near the apocenter of the giant planet .
the effect of the initial phase difference between the planets is important , therefore we prepared stability maps for more initial values of the mean anomaly @xmath22 of the giant planet
. the essential role of the phases for computations of orbits in mean motion resonances for terrestrial planets in exoplanetary systems has already been stressed by @xcite , where the giant planet was initially in periastron or apastron , and the phase of the terrestrial planet has been set to 12 different initial angles of the mean anomaly .
( a detailed treatment of the geometry of the resonances can be found in * ? ? ?
* ) . by studying the series of figs
[ fig6 ] - [ fig10 ] , one can follow the effect of the increase of the mass parameter on the stability structure . for larger mass of the giant planet ,
thus for larger perturbations , the chaotic region is more extended , as expected .
we note also a peculiar feature of figs .
[ fig6]a , [ fig7 ] [ fig10 ] . in the strongly chaotic regions on the right , which are due to the proximity of the giant planet
, there are two almost parallel light stripes , corresponding to satellite - type orbits around the giant planet .
these orbits are less chaotic ( or they may be ordered ) , than those emanating from the surrounding region . between these stripes
the dark gap corresponds to the perihelion distance of the giant planet , therefore in orbits originating from here , the test particles suffer close encounters with the giant planet , or they even can collide with it .
in this section we show how to use the catalogue to determine the stability of hypothetical earth - like planets in exoplanetary sytems . as an example we consider the case of hd10697 , where @xmath50 au , @xmath51 , @xmath52 .
[ fig9 ] shows a stability map , calculated for @xmath53 for inner orbits of the test planet .
this corresponds to the minimum mass of the giant planet ( minimum masses are used throughout ) .
the stability of a small planet ( starting with @xmath31 ) in the system hd10697 can be studied along the line @xmath51 .
one can see , that for small semi - major axes , @xmath54 the parameter space is very stable . when @xmath55 several resonances appear , among which the most important are the @xmath56 , @xmath57 , @xmath58 , and @xmath59 resonances . for @xmath60
a strongly chaotic region appears .
the classical hz of this system is between @xmath61 au @xcite .
thus in fig .
[ fig9 ] the scaled classical hz is situated between @xmath62 ( shown as a rectangle , elongated in horizontal direction ) .
one can see , that the inner part of the classical hz contains ordered regions , but stripes of certain resonances are also present .
the outer part of the classical hz is in the strongly chaotic region . in table
[ tbl-1 ] we list the orbital parameters of the exoplanetary systems displayed in figs . [ fig1][fig10 ] , and we characterize the stability properties of their hypothetical terrestrial planets in their classical and present hzs given by @xcite and @xcite . we give also the boundaries of the stable and the strongly chaotic regions . as sources for the parameters we used the webpages ` http://www.exoplanets.org ` and jean schneider s catalogue ( ` http://exoplanet.eu ` ) , denoted by e and js , respectively , in table 1 and in the figures , when there were significant differences between the data . in the column mmrs we display the highest and the lowest order mean motion resonances of the given exoplanetary system ( not only in their hz ) , the other resonances between them can be taken from the corresponding figure . when the highest order resonance is above @xmath63 , we call it as h.o.r . in the column chz / phz
we give a characterization of the stability properties of the classical and present hzs .
the notation s means a fully stable hz , where no mean motion resonance exists .
when the hz contains a few resonances , we consider it partly stable ( ps ) . if many resonances fill in the hz , we call it marginally stable ( ms ) . strongly chaotic ,
thus very unstable domains are denoted by sc .
the uncertainty in the eccentricity @xmath20 of a giant planet can affect the stability of the hz significantly .
the uncertainty in @xmath20 is indicated on the stability maps by the width of the rectangle of the classical hz .
( in the case of the present hz , marked by parentheses , we did not show the uncertainty in @xmath20 , since it is the same . )
when the uncertainty in @xmath20 is large , the stability properties at the top and at the bottom of the rectangle of the classical hz can be quite different .
for instance , in the case of the system hd 30177 it can be seen from fig .
10 , that for small values of @xmath20 there are a few resonances in the classical hz , thus it is partly stable , while for high values of @xmath20 the hz is full of resonances , therefore it is only marginally stable .
the habitability of individual exoplanetary systems has been studied by many authors .
as we have already mentioned , the orbital parameters of these systems are changed quite frequently , so it is not an easy task to compare directly our results with the results of other authors using older or different orbital data . on the other hand , there are only a few works studying the habitability of a large portion of the known systems .
such global investigations have been made by @xcite studying the habitability of 85 known exoplanetary sytems , and more recently by @xcite investigating the present habitability of 152 exoplanetary sytems . in these studies the gravitational influence of the known giant planet on the hz is estimated mainly by some multiple of its hill radius .
however , the stability properties of the hz of an exoplanetary system are also affected by mean motion resonances .
if the hz of a system contains only a few resonances , the ordered regions between them can provide stability for an earth - like planet .
if the resonances cover densely the hz , the long term stability of an earth - like planet is more questionable ; either the resonances overlap each other generating a strongly chaotic region , or chaotic diffusion , acting inside a mean motion resonance , can destabilize the motion of a hypothetical small planet .
we note that the presence of additional , yet undetected giant planets in an extrasolar planetary system can enhance the depletion of the mean motion resonances through their possible secular interactions with the known giant planet , similarly to the formation of the kirkwood gaps in our solar system @xcite . in our stability maps we investigated the dynamical stability of the classical hz given by @xcite , and the present hz calculated by @xcite .
studying the dynamical stability of the classical and the present hzs of 15 extrasolar planetary systems ( displayed in table 1 ) , we have found quite good agreement with the results of @xcite and @xcite .
we emphasize , however , that our stability maps also exhibit the resonant structure of the hzs , which is essential to establish the real stability properties of an earth - like planet . in what follows
we characterize the hz of some individual exoplanetary systems .
hd 52256 and hd 121504 : : : depending on the data source , the classical hz of the system hd 121504 is either fully stable ( source e ) , or partly stable ( source js ) , containing a few resonances ( fig .
[ fig1 ] ) .
the classical hz of the system hd 52256 is partly stable , but contains more resonances and smaller stability regions for the data js , than for e. the present hzs of these systems are located in the very regular regions of the @xmath64 parameter plane with a few resonances in the case of hd 52256 for the data
our results are in good agreement with the results of @xcite and @xcite concerning the extension of the stability regions
. these systems could harbor stable habitable earth - like planets .
+ and @xmath42 .
the classical hzs of two exoplanetary sytems ( with orbital parameters taken from two different data sources , denoted by e and js , see section 5 ) are also indicated as elongated rectangles , whose width corresponds to the uncertainty in the eccentricity of the giant planet s orbit of each system .
the inner borders of the present hzs are displayed as opened parentheses .
the outer borders are out of the investigated range of the @xmath64 plane .
as described in sections 5 and 6 , it can be seen , that depending on the data source the classical hz of the system hd 121504 is either fully stable ( source e ) , or partly stable ( source js ) , containing a few resonances .
the system hd 52265 is partly stable , but contains more resonances and smaller regions of stability for the data js , than for e. the present hzs are located in the very regular regions of the @xmath64 plane . ]
+ and @xmath42 .
light shades indicate stable regions , dark means chaoticity .
the very light stripes correspond to stable resonant orbits in the mean motion resonances .
note the complementary character of figs .
[ fig1 ] and [ fig2 ] , described in the text . ] + and @xmath42 . here
again lighter shades mean orbits , where the eccentricity remains relatively small .
the changes in the maximum eccentricity of the test planet indicate clearly the positions of the mean motion resonances . ]
hd 8574 : : : the classical hz of this system is in the strongly chaotic region for both data sources e and js ( fig .
[ fig4 ] ) .
@xcite found that @xmath65 of the present hz of the system can survive the gravitational influence of the giant planet , thus it could contain a stable habitable planet .
our approach clearly shows that the present hz of the system is full with overlaping mean motion resonances , and only partly stable for the data e , and marginally stable for the data
thus the chance for the system to host a stable habitable planet is small .
+ and @xmath42 . a comparison with fig .
5 reveals the increase of the chaotic region for increasing @xmath18 . as described in section 6
, the classical hz of the system hd 8574 is strongly chaotic for both data sources e and js .
the present hz of the system ( denoted by parentheses ) is partly stable for e , and marginally stable for js . ]
70 virginis and hd 178911 b : : : the classical hz of 70 vir is partly in the strongly chaotic region ( fig . [ fig5 ] ) , while its remaining part is full with overlaping mean motion resonances and thus only marginally stable .
this means that the classical hz of this system is unlikely to host habitable earth - like planets .
the classical hz of hd 178911 b is partly stable , and it can contain stable earth - like planets .
our results are in good agreement with the results of @xcite .
the present hzs of these systems are in the ordered regions of the @xmath64 plane , thus these systems can host earth - like habitable planets in stable orbits .
these results agree well with the results of @xcite .
+ and @xmath42 . as described in section 6 , the classical hz of 70 vir is partly in the strongly chaotic region , while its other part is marginally stable .
the classical hz of the system hd 178911 b is partly stable .
the present hz of hd 178911 b ( denoted by parenthesis ) is stable , while at 70 vir it is partly stable , containing higher order mean motion resonances ( 1:8 , 1:9 , etc .
, not shown in the figure ) .
] eps eri , hd 114729 , and gj 777 a : : : the classical hzs of eps eri and hd 114729 are marginally stable ( fig .
[ fig6 ] ) .
there is only a small chance to find earth - like planets here .
@xcite found that only 17% of the present hz of eps eri can have stable orbits , and nowhere in hd114729 does so . in the case of gj 777 a , due to the large uncertainty in @xmath20 of the giant planet ( @xmath66 ) ,
the classical hz can be stable ( for @xmath67 ) , or partly stable ( for @xmath68 ) .
there are no computed stability data for @xmath69 , however , the stable region will certainly decrease for larger values of @xmath20 .
the present hz of gj 777 a for @xmath68 is partly in the strongly chaotic region of the @xmath64 plane , while its remaining part is filled by overlaping mean motion resonances and is marginally stable . for @xmath70 , the present hz of gj 777 a is partly stable .
@xcite found that 64% of the present hz can survive the gravitational influence of the giant planet .
however , even though resonances were included in the development of their fast method of assessing stability @xcite , the method does not permit detailed consideration of the resonances in the individual systems to which it was applied . considering the classical hz , this system can host stable earth - like planets , as has also been found by @xcite and @xcite . + , @xmath42 ; ( b ) @xmath41 , @xmath48 .
the comparison of the two figures shows the effect of the relative starting positions of the two planets ( for more details see the text in section 4 ) .
as described in section 6 , the classical hzs of eps eri and hd 114729 are marginally stable . due to the large uncertainty of the giant planet s eccentricity ,
the classical hz of gj 777 a can be stable ( for @xmath70 ) , or partly stable ( for @xmath68 ) .
the hz of the solar system is placed on the map under the approximate condition that there is only one giant planet , jupiter in the system .
it is reassuring that the hz of the solar system is in the stable region .
the present hz of gj 777 a for @xmath68 is partly in the chaotic region , while its other part is marginally stable , and for @xmath70 it is partly stable.,title="fig : " ] , @xmath42 ; ( b ) @xmath41 , @xmath48 .
the comparison of the two figures shows the effect of the relative starting positions of the two planets ( for more details see the text in section 4 ) .
as described in section 6 , the classical hzs of eps eri and hd 114729 are marginally stable . due to the large uncertainty of the giant planet s eccentricity
, the classical hz of gj 777 a can be stable ( for @xmath70 ) , or partly stable ( for @xmath68 ) .
the hz of the solar system is placed on the map under the approximate condition that there is only one giant planet , jupiter in the system .
it is reassuring that the hz of the solar system is in the stable region .
the present hz of gj 777 a for @xmath68 is partly in the chaotic region , while its other part is marginally stable , and for @xmath70 it is partly stable.,title="fig : " ] hd 70642 : : : the classical hz of this system can be stable , or partly stable , depending on the eccentricity of the giant planet ( fig . [ fig7 ] ) .
the present hz of the system contains a few mean motion resonances , but between them there are ordered regions , which can host stable earth - like planets .
this result has also been found by @xcite . + and @xmath42 . due to the large uncertainty of the eccentricity of the giant planet ,
the classical hz of the system hd 70642 is either stable , or partly stable .
the present hz of this system is partly stable within the lower and upper limits of @xmath20 . ]
hd 72659 : : : the classical hz of this system contains only a few mean motion resonances ( fig . [ fig8 ] ) , thus it can host stable earth - like planets .
this is in good agreement with the results of @xcite and @xcite .
@xcite found , that @xmath71 of the present hz is not disrupted gravitationally by the giant planet .
this is true , if one considers only the effect of the strongly chaotic region .
however , the inner part of the present hz is full with mean motion resonances and is marginally stable .
thus the place for a stable earth - like planet is much smaller , than @xmath71 of the whole present hz , when the effects of the mean motion resonances are taken into consideration .
however , the presence of an earth - like planet can not be excluded entirely .
+ and @xmath42 .
the classical hz of the system hd 72659 is partly stable .
the outer part of the present hz of the system is chaotic , while its inner part is marginally stable .
we placed also the classical hz of the multiple planet system 47 uma on the figure for comparison , by taking into account from its two giant planets the more massive component b. ] hd 50554 : : : both the classical and the present hz of this system are entirely in the strongly chaotic region of the @xmath64 plane ( fig .
[ fig9 ] ) .
thus this system can not host habitable earth - like planet .
@xcite came to the same conclusion . on the other hand
, there could be ( retrograde ) satellite - type orbits in the system .
14 her : : : considering the classical hz , this system can host stable earth - like planets , however , mean motion resonances fill densely the classical hz ( fig .
[ fig9 ] ) . as to the present hz ,
the place for a stable earth - like planet is much smaller , with one part in the chaotic region , and the other is filled by many resonances .
our results are in good agreement with the results of @xcite .
hd 10697 : : : the classical hz of this system contains a strongly chaotic region , while its inner part between the mean motion resonances contains ordered , thus stable regions as well ( fig . [ fig9 ] ) .
the present hz of this system is in the strongly chaotic region , thus it can not harbour stable earth - like planets .
this is also found by @xcite .
+ and @xmath42 . along the line @xmath51
, corresponding to the system hd 10697 , the parameter space is very stable for @xmath72 .
for @xmath73 the parameter space is strongly chaotic . between @xmath74
there are several resonances .
the scaled classical hz of the system hd 10697 is between @xmath75 , thus its inner part is partly stable , but its outer edge is strongly chaotic .
so is its present hz .
as described in section 6 , both the classical and the present hz of the system hd 50554 are strongly chaotic .
interestingly , there could be ( retrograde ) satellite - type orbits in this system .
the classical hz of 14 her is partly stable , while its present hz is chaotic on one side , and marginally stable on the other . ]
hd 30177 : : : the classical hz of this system may host earth - like planets , since between the mean motion resonances there are ordered regions on the @xmath64 parameter plane ( fig .
[ fig10 ] ) . considering the present hz ,
its outer part is in the strongly chaotic region , while the inner part of it may contain stable earth - like planets .
this result agrees well with the result of @xcite . + and @xmath42 . due to the large uncertainty in the eccentricity of the giant planet , the classical hz of the system
hd 30177 is either partly stable ( at the bottom of the rectangle ) , or marginally stable ( at the top of the rectangle ) .
the present hz is strongly chaotic for large values of @xmath20 , while its inner part is partly or marginally stable . ]
thus we can conclude , that the effect of the mean motion resonances can not be neglected , when studying the habitability of the known extrasolar planetary systems . in this sense our work complements the already existing studies , such as @xcite .
in this paper we described a catalogue of stability maps of hypothetical terrestrial planets in exoplanetary systems with one giant planet . by using these maps
, one can easily study the stability properties of possible earth - like planets and the habitable zones in known or future exoplanetary systems .
the novelty of our work is the invariance of our results with respect to the frequent changes in the computed orbital parameters of the extrasolar planetary systems .
thus any new orbital determination ( leading to new elements ) can immediately be checked without any time - consuming computations .
this catalogue could also be useful for observing astronomers in determining the orbital parameters of terrestrial planets detected in the future by space instruments ( corot , kepler ) using transit photometry .
[ cols="<,^,^,^,^,^,^,^ , < " , ]
zss , s , and b thanks the support of the hungarian scientific research fund ( otka ) under the grants d048424 and t043739 .
ep - l wishes to acknowledge the support by the austrian fwf ( hertha firnberg project t122 ) .
this research has also been supported by the austrian - hungarian scientific and technology cooperation under the grant a-12/04 , and also benefited fom the issi team `` evolution of habitable planets '' .
the numerical integrations were done on the niidp supercomputer in hungary .
we thank the referee for useful suggestions .
99 asghari n. et al . , 2004 , a&a , 426 , 353 bois e. , kiseleva - eggleton l. , rambaux n. , pilat - lohinger e. , 2003 , apj , 598 , 1312 dvorak r. , pilat - lohinger e. , funk b. , freistetter f. , 2003 , a&a , 410 , l13 rdi b. , dvorak r. , sndor zs .
, pilat - lohinger e. , funk b. , 2004 , mnras , 351 , 1043 froeschl c. , lega e. , gonczi r. , 1997 , cemda , 67 , 41 guzzo m. , lega e. , froeschl c. , 2002 , physica d , 163 , 1 jones b.w . , sleep p.n . , chambers j.e . , 2001 , a&a , 366 , 254 jones b.w . , underwood d.r . ,
sleep p.n . , 2005 ,
apj , 622 , 1091 jones b.w . , sleep p.n . , underwood d.r .
, 2006 , apj , 649 , 1010 kasting j.f .
, whitmire d.p . ,
reynolds r.t . , 1993 ,
icarus , 101 , 108 mayor m. , queloz d. , 1995 , nature , 378 , 355 menou k. , tabachnik s. , 2003 , apj , 583 , 473 morbidelli a. , moons m. , 1993 , icarus , 103 , 99 moons m. , morbidelli a. , 1995 , icarus , 114 , 33 murray c.d . , dermott s. , 1999 , solar system dynamics .
cambridge university press , cambridge pilat - lohinger e. , dvorak r. , 2002 , cemda , 82 , 143 sndor zs .
, rdi b. , efthymiopoulos c. , 2000 , cemda , 78 , 113 sndor zs . , rdi b. , szll a. , funk b. , 2004 , cemda , 90 , 127 szebehely v.g .
, 1967 , theory of orbits . academic press , new york wisdom j. , 1983 , icarus , 56 , 51 | in the near future there will be launched space missions ( e.g. corot , kepler ) , designed to detect earth - like extrasolar planets .
the orbital elements of these ( still hypothetic ) planets will contain some uncertainties , that can only be eliminated by careful dynamical investigations of the hosting planetary systems .
the proportion of extrasolar planetary systems with one known giant planet is high ( @xmath0 ) , therefore as a first step we have investigated the possible existence of terrestrial planets in these systems . in this paper a development of a stability catalogue of the habitable zones of exoplanetary systems is reported .
this catalogue is formed by a series of stability maps , which can help to establish , where earth - like planets could exist in extrasolar planetary systems having one giant planet . after a description of the dynamical model and the numerical methods , details of the stability maps
are discussed .
an application of the stability catalogue to 15 known exoplanetary systems is also shown , and a characterization of the stability properties of their habitable zones is given .
[ firstpage ] astrobiology celestial mechanics methods : numerical ( stars : ) planetary systems . |
a cubature formula with high accuracy is an important tool for numerical computation and has various applications .
one of the applications is to construct polynomial hyperinterpolation , introduced by sloan @xcite , which is an approximation process constructed by applying the cubature formula on the fourier coefficients of the orthogonal projection operator . a cubature formula of degree @xmath4 with @xmath5 nodes with respect to a measure @xmath6 supported on a set @xmath7 takes the form @xmath8 where @xmath9 , called weights , are ( positive ) numbers , @xmath10 is a set of points , called nodes , @xmath11 with @xmath12 , and @xmath13 denoted the subspace of @xmath1-variate polynomials of total degree @xmath14 restricted to @xmath7 . for a cubature formula of degree @xmath4 to exist ,
it is necessary that @xmath15 there are improved lower bounds of the same order in terms of @xmath3 .
a challenging problem is to construct cubature formulae with fewer nodes , that is , with the number of nodes @xmath5 close to the lower bound .
in this paper we consider the case that the measure is given by the product chebyshev weight function @xmath16 on the cube @xmath17^d$ ] . for @xmath18 , the gaussian quadrature formula of degree
@xmath4 needs merely @xmath19 points .
our main result is a new family of cubature formulae that uses @xmath20 many nodes .
for @xmath21 these formulae are known to have minimal number of nodes .
for @xmath22 they are still far from the lower bound , but they appear to be the best ones that are known at this moment . we refer to section 2 for further discussions .
we present numerical tests on these cubature formulae in three variables and also apply them to constructing polynomial hyperinterpolation operator in three variables . for every function
@xmath23 the @xmath24-orthogonal projection of @xmath25 on @xmath26 is @xmath27 where @xmath28 is a @xmath1-dimensional point , @xmath29 is a @xmath1-index of length @xmath30 @xmath31 is any @xmath24-orthonormal basis of @xmath26 with @xmath32 of total degree @xmath30 ( concerning the theory of multivariate orthogonal polynomials , we refer the reader to the monograph @xcite ) .
clearly , @xmath33 for every @xmath34 . given a cubature formula of degree @xmath35 , we obtain from ( [ proj ] ) the polynomial approximation of degree @xmath3 by the _
discretized fourier coefficients _
@xmath37 where @xmath38 and thus @xmath39 for every @xmath34 .
this is the hyperinterpolation operator .
it satisfies the basic estimate : for every @xmath23 , @xmath40 where @xmath41 , so that it converges in mean .
the convergence rate can be estimated by a multivariate version of jackson theorem ( see , for example , @xcite ) , which shows that @xmath42 for @xmath43 , @xmath44 .
it becomes an effective approximation tool in the uniform norm when its operator norm ( the so - called lebesgue constant ) grows slowly ( cf .
the hyperinterpolation has been used effectively in several cases : originally for the sphere @xcite , and more recently for the square @xcite , the disk @xcite , and the cube @xcite .
we will use our new cubature formulae to construct a hyperinterpolation operator of three variables for the chebyshev weight function on the cube .
we show that the computation can be carried out efficiently using the 3-dimensional fft and that the algorithm can be completely vectorized .
we will also present numerical results on hyperinterpolation of several test functions .
the paper is organized as follows . in section 2
we construct new cubature formulae and report results of numerical tests , where comparisons are made with tensor - product gauss - chebyshev formulae .
hyperinterpolation in three variables is considered in section 3 , where we show how to compute it effectively and report the results of numerical tests . finally in section 4
, we obtain a new ( nontensorial ) clenshaw - curtis type formula in the cube by integrating the hyperinterpolant in section 3 and show that it has a clear superiority over tensorial clenshaw - curtis and gauss - legendre cubature on nonentire test integrands , a phenomenon known for 1-dimensional and 2-dimensional clenshaw - curtis formulae ( see @xcite ) .
we consider cubature formula for the product chebyshev weight function ( [ chebw ] ) , which is normalized so that its integral over @xmath45^d$ ] is @xmath46 . for @xmath18 , we write @xmath47 .
let @xmath48 denote the space of polynomials of total degree @xmath49 in @xmath1 variables .
we write @xmath50 if @xmath18 .
the gaussian quadrature formula for @xmath51 takes the form @xmath52 for @xmath53 , a cubature formula of degree @xmath4 needs at least ( cf .
@xcite ) @xmath54 many nodes .
cubature formulae that attain this lower bound can be constructed for the product chebyshev weight @xmath55 ( see @xcite and the references therein ) by studying common zeros of associated orthogonal polynomials . in @xcite ,
these cubature rules were derived by an elementary method which depends on a factorization of the gauss - lobatto quadrature into two sums , over _ even _ indices and _ odd _ indices , respectively .
this factorization method was also used for @xmath56 in @xcite and yields a cubature formula of degree @xmath4 for @xmath57 with roughly @xmath58 many nodes .
a close inspection of the factorization method shows that it actually allows us to derive cubature formulae of degree @xmath4 for @xmath57 with roughly @xmath59 many nodes .
this number of nodes is substantially less than @xmath60 of the product cubature formula or @xmath61 of the formulae in @xcite , although it likely far from optimal as seen from .
we start with the gauss - lobatto formula for @xmath51 on @xmath45 $ ] .
it takes the form @xmath62 which again holds for all @xmath63 .
we proceed to factor this rule into two terms .
the factorization depends on whether @xmath3 is even or @xmath3 is odd .
define @xmath64 and define @xmath65 where we use the superscripts @xmath66 and @xmath67 to signify that the sum is taken over even indices or odd indices , respectively .
evidently , the quadrature becomes @xmath68 by definition .
the chebyshev polynomials , @xmath69 , are orthogonal with respect to @xmath51 on @xmath45 $ ] , @xmath70 the following elementary lemma plays a key role in constructing cubature formulae on @xmath45^d$ ] .
[ lem1 ] for @xmath71 and @xmath72 , @xmath73 * proof . *
the proof follows from elementary trigonometric identities .
for example , for @xmath74 , an elementary calculation shows that @xmath75 from which @xmath76 for @xmath77 follows immediately .
the case when @xmath78 is a multiple of @xmath3 follows from the first equal sign of the above equation without summing it up .
similarly , @xmath79 from which the stated result follows .
the proof for @xmath80 is similar and is omitted for brevity .
0.5 cm let @xmath81 , that is , @xmath82 for a function @xmath83 , we define the sum @xmath84 as a @xmath1-fold multiple sum in which @xmath85 is applied to the @xmath78-th variable of @xmath25 .
let us define @xmath86 for each @xmath87 , we then define @xmath88 since the sum introduces a symmetry among @xmath87 , there are @xmath89 distinct @xmath90 sums .
[ mainth ] for @xmath91 and each @xmath87 , the cubature formula @xmath92^d } f(x ) w_d(x ) dx = 2^{d-1}i_{n , d}^\a f\ ] ] is exact for @xmath93 and its number of nodes , @xmath5 , satisfies @xmath94 * proof . * for @xmath95 let @xmath96 , which is a polynomial of total degree @xmath97 .
it suffices to establish for @xmath98 , since this set is an orthogonal basis of @xmath48 .
in this case we have @xmath99^d } t_k(x ) w_d(x ) dx = 2^{d-1}\left [ i_n^{\a_1 } t_{k_1}\cdots i_n^{\a_d } t_{k_d}+ i_n^{{\widetilde}{\a}_1 } t_{k_1}\cdots i_n^{{\widetilde}{\a}_d } t_{k_d } \right]\;.\ ] ] by the orthogonality of @xmath100 , the left hand side is equal to 1 if @xmath101 and zero if @xmath102 . from the definition of @xmath103 and @xmath104 ,
it is evident that @xmath105 .
hence , for @xmath106 , the right hand side is equal to @xmath107 , verifying the equation for @xmath106 .
assume now @xmath108 .
if one of @xmath109 , then @xmath110 by lemma [ lem1 ] .
we are left with the case that @xmath111 for all @xmath112 . since @xmath113
, there can be at most one @xmath114 .
furthermore , @xmath115 shows that there is exactly one @xmath116 .
thus the right hand side becomes @xmath117 , which is zero as @xmath118 and @xmath119 according to the lemma [ lem1 ] .
0.5 cm for the case of @xmath21 , theorem [ mainth ] contains two distinct cubature formulae for @xmath120 , respectively , whose number of nodes are either equal to @xmath121 in or @xmath122 , those are the ones that have appeared in @xcite , and later in @xcite , as mentioned earlier .
for @xmath123 , there are 4 distinct formulae for @xmath124 , respectively .
for @xmath125 , the number of nodes is @xmath126 for @xmath127 and @xmath128 for @xmath129 , respectively . in order to demonstrate the effectiveness of the new cubature formula , we present in figures 1 - 2 numerical results of ( [ cuba - d ] ) with @xmath130 on the integrals of six text functions with respect to the product chebyshev measure on the 3-cube .
the first three functions are analytic entire ( a polynomial , an exponential and a gaussian ) , whereas the other three are less smooth : one analytic but not entire ( a 3-dimensional version of the runge test function ) , one @xmath131 nonanalytic , and one @xmath132 .
these functions are analogues of test functions for algebraic cubature in dimension 1 and 2 , see @xcite .
we compare them with two natural choices for cubature on a tensor product domain : the tensor - product gauss - chebyshev and gauss - chebyshev - lobatto formulae .
the results , obtained with matlab ( cf .
@xcite ) , demonstrate the superiority of the new formula in all cases , especially for the less smooth functions , in terms of number of function evaluations .
it should be pointed out that , however , the superiority for the less smooth functions arises for _ even _
@xmath3 ( a sort of parity phenomenon ) .
other numerical tests ( not reported for brevity ) have shown that the cubature formula has the same behavior for @xmath133 .
a natural question associated with cubature formulae is polynomial interpolation .
let @xmath134 denote the set of the nodes of the cubature formula ( [ cuba - d ] ) .
the interpolation problem looks for a polynomial subspace , @xmath135 , of the lowest degree such that @xmath136 has a unique solution in @xmath135 . in the case of @xmath21 , this problem
is completely solved in @xcite , where @xmath135 is a subspace of @xmath137 which includes @xmath138 , and compact formulae of the fundamental interpolation polynomials are also given there .
for @xmath139 , however , the problem is much harder , since the number of nodes of our cubature is far from @xmath140 .
for example , if @xmath123 , then @xmath141 , whereas our cubature has @xmath142 many nodes .
the problem essentially comes down to study the polynomial ideal that has @xmath134 as its variety ( see @xcite ) .
a simpler approach to polynomial approximation via these new nodes is given by hyperinterpolation , as described in the introduction .
in the next section we shall apply such a method in the 3-dimensional case .
we now use cubature formula ( [ cuba - d ] ) to construct hyperinterpolation as in ( [ hyper ] ) for the 3-cube @xmath143 ^ 3 $ ] . in this case
, @xmath144 is the product chebyshev orthonormal basis ( cf .
@xcite ) , i.e. @xmath145 where @xmath146 and @xmath147 .
moreover , let @xmath148 be the set of @xmath149 chebyshev - lobatto points , and @xmath150 , @xmath151 its restriction to even and odd indices , respectively .
then , @xmath152 with @xmath153 , see ( [ sigmatilde ] ) .
the weights of the cubature formula ( [ cuba - d ] ) for @xmath154 , are @xmath155 note that , since @xmath156 the polynomial @xmath157 in ( [ hyper ] ) is not interpolant .
now , defining @xmath158 we can write @xmath159 where @xmath160 this shows that the 3-dimensional coefficients array @xmath161 is a scaled discrete cosine tranform of the 3-dimensional array @xmath162 where we eventually pick up only the @xmath163 hyperinterpolation coefficients corresponding to @xmath164 . a fast implementation of hyperinterpolation is now feasible ( for example in matlab ) , via the fft .
indeed , we have written a matlab code ( see @xcite ) , completely vectorized by several implementation tricks , whose kernel can be summarized as follows : 0.2 cm * algorithm : fast total degree hyperinterpolation in the 3-cube * * construct the hyperinterpolation point set @xmath10 as union of the two subgrids in ( [ hypernodes ] ) ; * compute the cubature weights in ( [ hyperweights ] ) ; * compute the 3-dimensional array @xmath165 at the complete grid @xmath166 by ( [ f ] ) ( notice that @xmath25 is evaluated _ only _ at @xmath10 ) ; * compute the 3-dimensional array of coefficients @xmath161 by three nested applications of the 1-dimensional @xmath167 operator ; * select the coefficients @xmath161 corresponding to the triples @xmath168 such that @xmath164 .
we recall that there is a simple way to approximate a function in the 3-cube by tensor - product of polynomials of degree @xmath3 , that is , by a tensor - product discrete chebyshev series ( ultimately a tensor - product hyperinterpolant ) .
such an approximation uses @xmath169 function evaluations , and @xmath169 coefficients .
in contrast , let us stress again the following facts on our total - degree hyperinterpolation of degree @xmath3 in the 3-cube : 0.2 cm * remark * * the number of hyperinterpolation nodes , or function evaluations , is equal to @xmath170 ; * the number of hyperinterpolation coefficients is @xmath171 . in order to compare the performances of total - degree and tensor - product hyperinterpolation in the 3-cube , we show , in the following figures , the hyperinterpolation errors versus both the number of nodes and the number of coefficients on the six test functions already used in section 2 , and we choose again @xmath172 , see ( [ hypernodes ] ) .
the errors are relative to the maximum deviation of the function from its mean and are computed on a uniform control grid . since the computation of the coefficients via
the fft has roughly the same cost for both kinds of hyperinterpolation , we have chosen the number of function evaluations as a measure of computational cost for the construction , and the number of coefficients as a measure of the compression capability of the algorithms .
the situation here is in some sense opposite to that of figures 1 - 2 . indeed , total - degree appears superior to tensor - product hyperinterpolation on the smoothest functions , but not on the less smooth ones . as it is natural from the observation above , the behavior of total - degree hyperinterpolation in terms of number of coefficients is better than that in terms of number of nodes ( function evaluations ) .
in the recent paper @xcite , perusing an idea already present in @xcite , it has been shown how hyperinterpolation allows us to construct new cubature formulae . given @xmath173 and @xmath23 , we can approximate the integral of @xmath174 in @xmath6 as @xmath175 where the generalized `` orthogonal moments '' @xmath176 and the cubature weights @xmath177 are defined by @xmath178 observe that the cubature formula ( [ prodcub ] ) is _ exact _ for every @xmath179 , and that @xmath176 are just fourier coefficients of @xmath180 with respect to the @xmath24-orthonormal basis @xmath144 .
let all the assumptions for the construction of the cubature formula ( [ prodcub ] ) be satisfied , and in particular let @xmath173 .
then the sum of the absolute values of the cubature weights has a finite limit @xmath181 notice that ( [ lim ] ) ensures that the sum of absolute values of the weights is bounded , and thus by recalling that @xmath182 is a projection operator on @xmath183 we obtain the polya - steklov type ( cf .
@xcite ) convergence estimate @xmath184 where @xmath185 denotes the error of the best polynomial approximation of degree @xmath3 to @xmath25 in the uniform norm . now , applying ( [ prodcub])-([momwei ] ) in the case @xmath186 ( since then @xmath187 ) we obtain , via hyperinterpolation , a cubature formula for the standard lebesgue measure from an algebraic cubature formula for another measure ( absolutely continuos with respect to the former ) .
the specialization of this approach to the @xmath46-dimensional chebyshev measure gives ultimately the popular clenshaw - curtis quadrature formula @xcite .
an extension to dimension 2 has been studied in @xcite . here
we apply the method in dimension 3 , obtaining a new nontensorial clenshaw - curtis - like cubature formula in the 3-cube . in figures 7 - 8 we display the relative errors of such a formula for @xmath172 ( cf .
( [ hypernodes ] ) ) on the six test functions already used above , compared with those of the tensor - product clenshaw - curtis , gauss - legendre , and gauss - legendre - lobatto formulae .
the numerical results have been obtained with matlab , using @xcite for the gaussian formulae and @xcite for the tensor - product clenshaw - curtis formula .
in particular , we see that with the entire test functions nontensorial clenshaw - curtis cubature is more accurate than the tensor - product version , but less accurate than the other two tensor - product formulae . on the other hand , in the less smooth cases the nontensorial clenshaw - curtis formula is better than all the other three , especially for _ odd _ hyperinterpolation degrees @xmath3 , which correspond to use @xmath149 _ even _ in ( [ cuba - d ] ) ( again a sort of parity phenomenon , cf .
figure 2 ) .
this behavior echos that of 1-dimensional and 2-dimensional clenshaw - curtis formulae ( see @xcite ) .
other numerical tests ( not reported for brevity ) have shown that the other versions of the nontensorial clenshaw - curtis formula , corresponding to @xmath188 in ( [ hypernodes ] ) , produce essentially the same results .
o. hansen , k. atkinson and d. chien , on the norm of the hyperinterpolation operator on the unit disk and its use for the solution of the nonlinear poisson equation , i m a j. numer .
, published online 20 march 2008 . | a new algebraic cubature formula of degree @xmath0 for the product chebyshev measure in the @xmath1-cube with @xmath2 nodes is established . the new formula
is then applied to polynomial hyperinterpolation of degree @xmath3 in three variables , in which coefficients of the product chebyshev orthonormal basis are computed by a fast algorithm based on the 3-dimensional fft .
moreover , integration of the hyperinterpolant provides a new clenshaw - curtis type cubature formula in the 3-cube . |
the soudan-2 experiment is currently taking data using its fine - grained iron tracking calorimeter of total mass 963 tons . this detector images non - relativistic as well as relativistic charged particles produced in atmospheric neutrino reactions .
it is operating underground at a depth of 2100 meters - water - equivalent on level 27 of the soudan mine state park in northern minnesota ( northwest of sudbury ) .
the calorimeter s modular design enabled data - taking to commence in april 1989 when the detector was one quarter of its full size ; assembly of the detector was completed during 1993 .
data - taking has continued with 85% live time , even though dynamite blasting has been underway nearby for the minos cavern excavation since summer 1999 .
the data exposure as of this conference is 5.40 fiducial kiloton - years ( kty ) .
results presented here are based upon a 5.1 kty exposure .
the tracking calorimeter operates as a slow - drift ( 0.6 cm/@xmath12s ) time projection chamber .
its tracking elements are meter - long plastic drift tubes which are placed into the corrugations of steel sheets .
the sheets are stacked to form a tracking lattice of honeycomb geometry .
a stack is packaged as a calorimeter module and the detector is assembled building - block fashion using these modules @xcite .
the calorimeter is surrounded on all sides by a cavern - liner active shield array of two or three layers of proportional tubes @xcite .
topologies for contained events in soudan 2 include single track and single shower events ( mostly @xmath0 and @xmath1 quasi - elastic reactions ) and multiprong events .
flavor - tagging proceeds straightforwardly : an event having a leading , non - scattering track with ionization @xmath13 compatible with muon mass is a candidate charged current ( cc ) event of ; an event having a prompt , relatively energetic shower prong is a candidate @xmath1 cc event . recoil protons of momenta greater than approx .
350 mev/@xmath14 are imaged by the calorimeter , allowing a much - improved measurement of the incident neutrino direction , especially for sub - gev quasi - elastic reactions .
we measure the atmospheric neutrino @xmath0/@xmath1 flavor ratio - of - ratios @xmath16 using single track and single shower events which are fully contained within the calorimeter ( all hits more than 20 cm from the nearest surface ) .
these samples contain mostly quasi - elastic neutrino reactions , but include a background of photon and neutron reactions originating in cosmic ray muon interactions in the surrounding cavern rock .
the latter rock events " are mostly tagged by coincident hits in the active shield , however some are unaccompanied by shield hits and constitute a background .
the amount of zero - shield - hit rock background in a neutrino event sample is estimated by fitting event vertex - depth distributions to a combination of tagged - rock and @xmath15 monte carlo distributions . as expected
, the fits show the background to be mostly confined to outer regions of the calorimeter .
details of our analysis procedures for quasi - elastic events can be found in refs .
@xcite .
the track and shower event samples for our 5.1 kty exposure are summarized in table [ tab : results ] .
our full detector monte carlo simulation of atmospheric neutrino interactions is based on the 1996 bartol flux for the soudan site @xcite .
.soudan-2 track and shower event samples from the 5.1 fiducial kiloton - year exposure .
[ cols="<,^,^",options="header " , ] [ tab : hires ] the atmospheric monte carlo ( mc ) sample represents 28.2 kiloton years of exposure .
the mc event rates displayed in table [ tab : hires ] have been normalized to the @xmath1 data sample .
this normalization is equivalent to a reduction of the bartol neutrino fluxes by 21% .
the assumption implicit with this adjustment is that the @xmath1 sample is devoid of oscillation effects
[ fig : fig01 ] , [ fig : fig02 ] and [ fig : fig03 ] show hires distributions with this normalization in place . fig .
[ fig : fig01 ] shows the distributions of these samples in cosine of the zenith angle . for @xmath1 events , the shape of the distribution for data ( fig .
[ fig : fig01]a , crosses ) coincides with that predicted by the monte carlo ( dashed histogram ) for null oscillations .
the distribution of @xmath0 data however , falls below the mc prediction in all bins ( fig .
[ fig : fig01]b ) with the relative dearth being more pronounced for @xmath0 s incident from below horizon . distributions in @xmath17 for hires events are shown in fig .
[ fig : fig02 ] . for null oscillations
this variable distributes according to a ` phase space ' which reflects the neutrino points - of - origin throughout the spherical shell volume of the earth s atmosphere .
that is , down - going @xmath15 s populate the peak at lower @xmath17 from 0.0 to 2.0 .
neutrinos incident from / near the horizon occur within the dip region extending from 2.0 to 2.6 , while upward - going neutrinos populate the peak at higher values .
[ fig : fig02]a shows that , allowing for statistical fluctuations , the @xmath1 data follows the shape of the null oscillation mc distribution .
in contrast , the @xmath0 data ( fig .
[ fig : fig02]b ) falls below the null ocillation mc for all but the most vertically down - going flux .
to convert results of our atmospheric neutrino simulation generated under the no - oscillation hypothesis into simulated neutrino oscillation data , we apply to every mc event an @xmath4-dependent weight representing the probability of survival for a given @xmath8 and @xmath7 . an exploratory matchup of @xmath0 data with trial oscillation scenarios is shown in fig .
[ fig : fig03 ] . for the mixing angle @xmath7 fixed at unity
, we plot the mc distribution weighted for @xmath0 to @xmath3 oscillations with differing @xmath8 settings . with @xmath18 ( fig .
[ fig : fig03]a ) , the oscillation prediction lies above the data in most @xmath4 bins . with @xmath19 ( fig .
[ fig : fig03]b ) , the upgoing @xmath15 flux is now better described by the mc , although the expectation for horizontal and down - going neutrinos remains somewhat high . with @xmath20 , a rough agreement overall
is achieved ( fig .
[ fig : fig03]c ) . going to higher @xmath8
, we find that for @xmath21 the oscillation - weighted mc falls below the data in most bins ( fig .
[ fig : fig03]d ) .
these hints concerning the parameters regime preferred by the data are borne out by a more considered analysis , as we now show . to determine the neutrino oscillation parameters @xmath8 and @xmath7 from our data , we construct a @xmath22 function over the plane - of - parameters . for points
@xmath23 in the physical region of the @xmath24 plane , we fit the mc expection to our data at each point .
the mc flux normalization , @xmath25 as well as @xmath26 and @xmath27 , is a free parameter : @xmath28 we assume that the oscillation affecting our data is purely @xmath0 into @xmath3 and that the @xmath1 data is unaffected . the @xmath22 is summed over data bins containing our selected ( hires ) @xmath0 and @xmath1 samples , where @xmath29 are @xmath0 @xmath17 bins , with @xmath30 containing all the @xmath1 events .
the denominator @xmath31 accounts for finite statistics in the neutrino monte carlo and for uncertainty in the rock background in the @xmath15 data .
not yet included are error terms which address systematic errors in the analysis , however preliminary examination shows statistical errors to be the dominant error source in the analysis .
the mc counts @xmath32 for the @xmath33 bin are constructed using oscillation weight factors .
we find the location of minimum @xmath34 , and plot @xmath35 which is @xmath36 .
the @xmath37 surface thereby obtained is shown in fig .
[ fig : fig04 ] . a crater region of low @xmath22 values is clearly discerned , at the bottom of which is a relatively flat basin .
the lowest point @xmath38 occurs at values @xmath7 = 0.90 ,
@xmath39 , with flux normalization @xmath25 = 0.78 .
an additional structure is the @xmath37 ridge which occurs at large mixing angle and for @xmath8 above 10@xmath40 ev@xmath41 . for oscillation solutions in this regime , depletion in the downward - going @xmath0 neutrino flux with sub - gev energies
is predicted for @xmath0 to @xmath3 oscillations by equation ( 1 ) arising from the first oscillation minimum .
our hires events have sufficient resolution to show such an effect if it would be present . however , no pronounced depletion is observed , and so the @xmath22 has a high value there . to find the region allowed for the oscillation parameters by our data at 90% confidence level ( cl )
, we use the method of feldman and cousins @xcite . at each of 2500 points @xmath42 on a grid spanning the physical region of the plane parameters ,
we run 1000 simulated experiments . for each of the simulated sets , we find @xmath43 such that @xmath44 is less than @xmath43 for 90% of the simulated experiments at @xmath23 . the surface defined by local @xmath45 over the oscillation parameters plane
is shown in fig .
[ fig : fig05 ] .
note that the surface is not a plane at @xmath45 = 4.61 , but rather has a concave shape .
the central shaded portion is approximately @xmath37 = 4.6 , however the outlying regions have @xmath37 values which are lower . at each point over the physical region , if @xmath46 is less than @xmath47 , then @xmath23 belongs to the allowed region of the 90% cl contour .
the region allowed by our data at 90% cl is shown by the shaded area in fig .
[ fig : fig06 ] .
although @xmath38 occurs at the location depicted by the solid circle , the relatively flat basin of our @xmath37 surface extends to lower @xmath8 values .
superk has reported their best fit @xmath8 value to be @xcite ; our data is compatible with that as well as with somewhat higher @xmath8 values .
we plan to include more data in the above analysis . an additional data sample is comprised of @xmath0 flavor events which are partially contained . for each event of this category
, the primary vertex is required to be @xmath48 80 cm ( one hadronic interaction length ) from exterior surfaces of the calorimeter , and the final state must contain a non - scattering , exiting track with ionization compatible with a muon mass . this sample is useful because the assignment of @xmath0 flavor is reliable to better than 98% , and because the events are relatively energetic and consequently point well " to the incident @xmath15 direction .
the mean energy for neutrinos which initiate pces is estimated to be 4.7 gev , to be compared to a mean energy of 1.3 gev for @xmath0 hires events .
the mean angular deviation of the reconstructed @xmath0 direction versus the true direction ( in monte carlo ) is @xmath49 .
unfortunately the number of @xmath0 pces is low , less than one - third the population of our @xmath0 hires sample . in order to isolate pce two - prong and multiprong topologies ,
the data events ( with mc events interspersed throughout ) are processed through a software filter ; those which pass are scanned .
the filter is designed to eliminate downward - stopping muons which have endpoint decays . in soudan-2 ,
an electron from a muon decay near rangeout can give rise to a small shower of @xmath50 10 hits from the end of a muon track ; the topology is roughly akin to that of a neutrino - induced two - prong .
consequently care must be taken to avoid remnant up - down asymmetry in pces introduced by the filter .
the problem is avoided by requiring that there be @xmath48 20 hits from a pce vertex which are additional to the muon track . to mitigate against cosmic - ray induced backgrounds we require that any hit in the active shield which is coincident with a pce , must be clearly associated with the exiting muon track .
occasionally it happens that a charged pion ejected from the cavern rock is incident upon the calorimeter .
if the pion penetrates by more than an interaction length and then scatters inelastically , it can mimic a @xmath15 pce topology .
background events of this type are removed by requiring the net momentum of the hadronic system of a pce to lie within the same hemisphere which contains the candidate muon track . with these selections
we obtain 31 @xmath0 events for which the cosmic ray induced background is less than one event . the @xmath0 pce rate predicted by our monte carlo with no oscillations
is 40 events .
the distribution of pce data in @xmath51 can be compared , as done previously for contained hires @xmath0 events , to representative oscillation scenarios having @xmath7 = 1.0 . in fig .
[ fig : fig07 ] we observe that , compared to the data ( crosses ) the oscillation prediction ( dashed histogram ) is relatively high for @xmath18 ( fig .
[ fig : fig07]a ) . however the prediction for @xmath21 ( fig .
[ fig : fig07]d ) is too low relative to the data . figs .
[ fig : fig07]b , c suggest that the scenario preferred by the data lies in the regime between @xmath52 .
in the near future , we will include the partially contained @xmath0 events into our @xmath22 fit to @xmath53 .
additionally , a sample of upward - stopping muon events initiated by neutrino reactions below the detector has been isolated and will be analyzed for oscillation effects .
and of course we will continue to accumulate and analyze new data .
our goal is to keep the soudan-2 detector running and tuned for the change - of - beams incident at our underground site
. the change will be from atmospheric @xmath15 s to fermilab @xmath0 s in fall 2003 , at which time the detector will serve the minos experiment @xcite .
mann , plenary talk at the xix int .
symposium on lepton and photon interactions at high energies , stanford university , august 1999 , hep - ex/9912007 ; t. kafka , _ in : _
taup99 - proceedings of the sixth int .
workshop on topics in astroparticle and underground physics , college de france , paris , france , september 1999 , nucl .
b ( proc . suppl . )
* 87 * ( 2000 ) 186 . | neutrino interactions recorded in a 5.1 fiducial kiloton - year exposure of the soudan-2 iron tracking calorimeter are analyzed for effects of neutrino oscillations .
using contained single track and single shower events , we update our measurement of the atmospheric @xmath0/@xmath1 ratio - of - ratios and find @xmath2 . assuming this anomalously low r - value is the result of @xmath0 flavor disappearance via @xmath0 to @xmath3 oscillation , we select samples of charged current events which offer good resolution , event - by - event , for @xmath4 reconstruction .
oscillation - weighted monte carlo events are fitted to these data events using a @xmath5 function summed over bins of log@xmath6 .
the region allowed in the ( @xmath7 , @xmath8 ) plane at 90% cl is obtained using the feldman - cousins procedure : @xmath9 and @xmath10 .
a small but relatively energetic sample of partially contained @xmath0 events has also been isolated .
their distribution in log@xmath11 relative to null oscillation monte carlo is compatible with @xmath0 to @xmath3 oscillation scenarios within the parameters region allowed by our contained events . |
the essence of radiotherapy with heavy charged particles is its intrinsic capability of three - dimensional dose formation with a bragg peak at the beam range , which can be precisely controlled by adapting the beam energy incident on a patient . in conventional broad - beam delivery,@xcite
a range shifter degrades excessive beam energy and a range compensator fills target - depth deficit varying in the field to best conform the spread - out bragg peak to a planning target volume.@xcite a range shifter is typically a composite of plates of precisely controlled thickness and a range compensator is a physical filter custom - made for an individual field .
their material should be uniform , stable , machinable , inexpensive , and ideally equivalent to water , which is the reference material for dosimetry .
these requirements are also valid for solid phantom materials , which are used when water - phantom dosimetry is not possible or convenient . to estimate the effective thickness of a material by the resultant range shift in water , an effective density
is assigned for the material . however , this approximation may cause dosimetric errors due to alteration of the radiation because the effective density differs for multiple scattering and nuclear interactions.@xcite dosimetric water equivalence , which is evaluated by uniformity of effective densities for relevant interactions , is essential to energy degrading of charged - particle beams for range shifting , range compensation , and dosimetry .
some common materials used for this purpose are high - density polyethylene ( hdpe),@xcite polymethyl methacrylate ( pmma),@xcite and synthetic resin of acrylonitrile , butadiene , and styrene ( abs).@xcite in general , hdpe is inexpensive and close to water in density , pmma is good at dimensional stability and contains oxygen which is dominant in water , and abs is available in various forms in industry as chemical wood . for proton and ion - beam dosimetry ,
pmma has been commonly used as a water - equivalent phantom material.@xcite in the past , water equivalence of general plastics ( pmma , hdpe , ps , and ptfe ) , tissue - substitute plastics ( a150 and commercial products ) , and other solid materials ( graphite , bone , aluminum , and copper ) have been studied for energy degrading and calorimetry of proton and ion beams by experiment and monte carlo simulation.@xcite those studies primarily focused on fluence - correction factor , which is the ratio of dose in water to dose - to - water in material of interest at a same water - equivalent depth . while the fluence - correction factor is essential for dosimetry with non - water phantom , generalization for a variety of modulated treatment beams may be difficult due to varied dose contribution from secondary particles and atomic recoils . for carbon - ion radiotherapy ,
the radiation components need to be handled separately due to variability of relative biological effectiveness with particle charge and energy,@xcite for which the fluence - correction factor for total dose would not suffice . instead
, survival of carbon ions that attenuate with depth to form a bragg peak could be clinically more relevant . in this regard ,
we evaluate water equivalence of some plastics for energy degrading of carbon - ion beams .
in addition to hdpe and pmma , we theoretically evaluated two other oxygen - rich plastics , polyethylene terephthalate ( pet ) and polyoxymethylene ( pom ) , among which we experimentally tested hdpe , pmma , and pom with carbon - ion beams .
we excluded abs as it varies in composition of carbon , hydrogen , and nitrogen .
table [ tab:1 ] summarizes the relevant properties of these plastics and water,@xcite in which the densities may vary with degree of polymerization .
incidentally , @xmath0 is the electron density , where @xmath1 is the atomic mass unit .
.[tab:1 ] material properties ( elemental composition , oxygen weight fraction @xmath2 , typical density @xmath3 , number of electrons per atomic mass unit @xmath4 , and mean excitation energy @xmath5 ) excerpted from ref .
( also available at http://pdg.lbl.gov/[the particle data group website ] ) except for an empirical hdpe density . [ cols="<,<,^,^,^,^",options="header " , ] carbon - ion attenuation ratio @xmath6 measured as bragg - peak reduction for range shift @xmath7 by hdpe ( circles ) , pmma ( squares ) , and pom ( triangles ) for 290-mev ( open ) and 430-mev ( filled ) beams along with analytical model curves ( solid lines ) . ]
as the density of polymer plastics depends on manufacturing condition , the effective density must be experimentally determined before its clinical use to an accuracy better than 1% .
therefore , the stopping - power similarity to water ( @xmath8 ) for hdpe is not essentially an advantage .
the similarity is only relevant when geometrical dose properties are of interest , such as field width and penumbra size . for applications that allow or require regular air gaps in a phantom , higher @xmath9 values are acceptable or
may even be preferable .
for example , yajima _ et al . _ developed a multilayer ionization chamber composed of 3-mm and 4-mm pmma plates interleaved with 1-mm air gaps,@xcite which would reduce the density by factor @xmath10 resulting in mean stopping - power ratio @xmath11 .
if pom is used instead , the @xmath12 value will be 1.06 , which will further be adjusted to 1.00 by extending the air gaps to 1.26 mm .
beam blurring due to multiple scattering in a range compensator should ideally be minimized . in this
regard , hdpe with @xmath13 is superior to pmma ( 0.890 ) , pet ( 0.953 ) , pom ( 0.968 ) , and water ( 1 ) although their effective differences will be generally marginal .
for example , when a carbon - ion beam of 20-cm avaiable range traverses a plastic plate for 10-cm range shift , the induced beam blurring in 30-cm distance will be 1.9 mm for hdpe , 2.1 mm for pmma , 2.2 mm for pet , and 2.2 mm for pom , using eqs .
( [ eq:5 ] ) and ( [ eq:6 ] ) .
the dosimetric water equivalence of the plastics that we tested here originated from the similarity with water in oxygen content and should be also valid for protons and other ions .
therefore , pom will be generally a good dosimetric material if its high density is tolerable .
for some dosimetric applications that require material transparency , such as water tank with optical dosimeter - alignment system , pet may desirably substitute for pmma .
we evaluated dosimetric water equivalence of four common plastics , hdpe , pmma , pet , and pom , by uniformity of effective densities for carbon - ion - beam interactions . among them , pom was the best and virtually indistinguishable from high - density water , which would be ideal for range control and preferable for dosimetry with regular air gaps such as for multilayer ionization chamber . for applications that require transparency such as water tank with optical dosimeter - alignment system , pmma was verified to be reasonably water equivalent and pet would be even better .
the poorest was hdpe with a large fractional attenuation ratio of 0.45% per cm range shift , although with marginal superiority for reduced multiple scattering .
analytical model calculations agreed with measurements within 1% for carbon - ion - survival ratio of plastic to water .
the water equivalence of plastics was highly correlated with their oxygen content , as expected from the composition of water .
haettner _ et al . _ measured the number of carbon ions , @xmath14 , attenuating with depth @xmath15 in water for @xmath16 mev and 400 mev.@xcite we compiled their data to construct an energy - independent universal formula as a function of residual range in water , @xmath17 , where @xmath18 is the range - equivalent depth . using the measured @xmath19 curve for each energy , the @xmath18
was tentatively set to the maximum - gradient depth . using the data points for @xmath20 ,
a straight line was fitted and its intercept at @xmath21 was defined as the number of stopped carbon ions , @xmath22 .
the @xmath18 was then redefined with submillimeter adjustment to meet @xmath23 .
the measurements showed natural exponential behavior for @xmath24 cm , for which we determined the mean free path using the 400-mev dataset . while there would naturally be some difference between the energies in the small region of @xmath25 cm due to marginal variation of range straggling , systematic deviation was observed unexpectedly in the larger region of @xmath26 cm .
as the exponential behavior held in the intermediate region of @xmath27 for the 200-mev dataset , the deviation may be attributed to increased measurement difficulty for the 400-mev beam .
therefore , we rescaled the 400-mev dataset to match the 200-mev dataset in the exponential region of @xmath24 cm and we adopted the straight line fitted to the 200-mev dataset for the @xmath28 cm region .
the resultant universal formula for carbon - ion attenuation is @xmath29 which is shown in fig .
[ fig:4 ] along with the measurements rescaled as described above .
number of carbon ions , @xmath14 , per number of stopped carbon ions , @xmath22 , as a function of residual range in water , @xmath30 , measured by haettner _
et al._@xcite for @xmath16 mev ( @xmath31 ) and 400 mev ( @xmath32 ) with the fitted function ( solid line ) . ] t. kanai , m. endo , s. minohara , n. miyahara , h. koyama - ito , h. tomura , n. matsufuji , y. futami , a. fukumura , t. hiraoka , y. furusawa , k. ando , m. suzuki , f. soga , and k. kawachi , `` biophysical characteristics of himac clinical irradiation system for heavy - ion radiation therapy , '' int . j. radiat .
. phys . * 44 * , 201210 ( 1999 ) .
n. kanematsu , t. inaniwa , and y. koba , `` relationship between electron density and effective densities of body tissues for stopping , scattering , and nuclear interactions of proton and ion beams , '' med .
phys . * 39 * 10161020 ( 2012 ) .
y. takada , t. himukai , k. takizawa , y. terashita , s. kamimura , h. matsuda , k. hotta , r. kohno , m. komori , and t. kanai , `` the basic study of a bi - material range compensator for improving dose uniformity for proton therapy , '' phys .
biol . * 53 * 55555569 ( 2008 ) . c. brusasco , b. voss , d. schardt , m. krmer , and g. kraft , `` a dosimetry system for fast measurement of 3d depth - dose profiles in charged - particle tumor therapy with scanning techniques , '' nucl .
methods b * 168 * 578592 ( 2000 ) .
h. palmans , j. e. symons , j - m .
denis , e. a. de kock , d. t. l. jones , and s. vynckier,``fluence correction factors in plastic phantoms for clinical proton beams , '' phys .
biol . * 47 * 30553071 ( 2002 ) .
u. schneider , p. pemler , j. besserer , m. dellert , m. moosburger , j. de boer , e. pedroni , and t. boehringer , `` the water equivalence of solid materials used for dosimetry with small proton beams , '' med . phys .
* 29 * 29462951 ( 2002 ) .
l. al - sulaiti , d. shipley , r. thomas , a. kacperek , p. regan , and h. palmans , `` water equivalence of various materials for clinical proton dosimetry by experiment and monte carlo simulation , '' nucl .
instrum . meth .
a * 619 * 344347 ( 2010 ) .
l. al - sulaiti , d. shipley , r. thomas , p. owen , a. kacperek , p. regan , and h. palmans , `` water equivalence of some plastic - water phantom materials for clinical proton beam dosimetry , '' appl .
. isotop . * 70 * 10521055 ( 2012 ) .
sigmund , a. schinner , and h. paul , `` errata and addenda for icru report 73 , stopping of ions heavier than helium , journal of the icru vol . 5 no . 1 ( 2005 ) , '' available at http://www.icru.org/testing/reports/errata-and-addenda-icru-report-73-stopping-of-ions-heavier-than-helium[the icru website ] ( 2009 ) .
l. sihver , c. h. tsao , r. silberberg , t. kanai , and a. f. barghouty , `` total reaction and partial cross section calculations in proton nucleus ( @xmath33 ) and nucleus nucleus reactions ( @xmath34 and @xmath33 ) , '' phys .
c * 47 * 12251236 ( 1993 ) .
m. torikoshi , s. minohara , n. kanematsu , m. komori , m. kanazawa , k. noda , n. miyahara , h. ito , m. endo , and t. kanai , `` irradiation system for himac , '' j. radiat .
( tokyo ) * 48 * ( suppl .
a ) a15a25 ( 2007 ) .
b. schaffner , t. kanai , y. futami , m. shimbo , and e. urakabe , `` ridge filter design and optimization for the broad - beam three - dimensional irradiation system for heavy - ion radiotherapy , '' med .
* 27 * 716724 ( 2000 ) . | purpose : : beam range control is the essence of radiotherapy with heavy charged particles . in conventional broad - beam delivery , fine range adjustment is achieved by insertion of range shifting and compensating materials . in dosimetry , solid phantoms are often used for convenience .
these materials should ideally be equivalent to water . in this study , we evaluated dosimetric water equivalence of four common plastics , hdpe , pmma , pet , and pom . methods : : using the bethe formula for energy loss , the gottschalk formula for multiple scattering , and the sihver formula for nuclear interactions , we calculated the effective densities of the plastics for these interactions .
we experimentally measured variation of the bragg peak of carbon - ion beams by insertion of hdpe , pmma , and pom , which were compared with analytical model calculations .
results : : the theoretical calculation resulted in slightly reduced multiple scattering and severely increased nuclear interactions for hdpe , compared to water and the other plastics . the increase in attenuation of carbon ions for 20-cm range shift
was experimentally measured to be 8.9% for hdpe , 2.5% for pmma , and 0.0% for pom while pet was theoretically estimated to be in between pmma and pom .
the agreement between the measurements and the calculations was about 1% or better .
conclusions : : for carbon - ion beams , pom was dosimetrically indistinguishable from water and the best of the plastics examined in this study . the poorest was hdpe , which would reduce the bragg peak by 0.45% per cm range shift , although with marginal superiority for reduced multiple scattering
. between the two clear plastics , pet would be superior to pmma in dosimetric water equivalence . |
neutron stars represent the most dense form of matter known before black hole formation .
the equation of state of such material is not well understood but typical expected size scales of @xmath2 km for a @xmath3 object give @xmath4 similar to that of the last stable orbit around a schwarzschild black hole .
the gravitational potential seen by the accreting material is then very similar in neutron stars and black holes , but with the obvious difference that the material may then fall seamlessly and invisibly from the edge of the disc through the event horizon in the black holes ( but see e.g. krolik , hawley & hirose 2005 for an alternative view ) , but must impact onto the solid surface for the neutron stars .
this boundary layer emission should give a clear distinction between the two types of systems , yet in practice it can be quite hard to discriminate between the disc accreting , low magnetic field neutron stars and galactic black holes ( gbh ) . both show hard spectra at low mass accretion rates ( e.g. barret & vedrenne 1994 ; barret et al .
2000 ) while at higher mass accretion rates the black holes can show low temperature , optically thick comptonized emission from a corona which can look similar to an optically thick boundary layer in the neutron stars systems ( the black hole ` very high state ' , done & gierliski 2003 , hereafter dg03 ) .
nonetheless , dg03 showed that there are clear differences between the black hole and neutron star systems .
they used the huge database now available on these objects from the _ rossi x - ray timing explorer _ ( _ rxte _ ) to build a picture of the x - ray spectral evolution as a function of overall mass accretion rate ( as measured by luminosity , parameterized in units of the eddington rate , @xmath5 ) .
they used data from gbh and both subclasses of disc accreting , low magnetic field neutron stars namely atolls and z sources ( see e.g. hasinger & van der klis 1989 ) .
the atolls and gbh cover the same luminosity range in @xmath5 , from @xmath6 to @xmath7 , so their accretion flows should be directly comparable ( z sources are typically more luminous , and may have residual magnetic fields : hasinger & van der klis 1989 ) .
dg03 showed that the gbh as a class were all consistent with the same spectral evolution as a function of mass accretion rate , as were the atolls , but that this behaviour _ differed substantially _ between the two types of system even though they covered the same range in @xmath5 .
dg03 interpreted this in the light of the recent advances in accretion flow models , which suggest that at low mass accretion rates the optically thick , cool , inner disc can evaporate into an optically thin , hot flow ( e.g. an advection dominated accretion flow : narayan & yi 1996 or a jet dominated accretion flow : falcke , kording & markoff 2004 ) .
these models are currently controversial as they conflict with the occasional detection of extremely broad iron lines in gbh spectra in this state ( miller et al 2002 ; miniutti , fabian & miller 2004 ; miller et al 2006 ) , though these extreme widths could also be an artifact of complex absorption ( done & gierliski 2006 ) .
nonetheless , the truncated disc / hot inner flow models are very attractive as they can qualitatively explain a wide range of observed properties in x - ray binaries .
this includes the spectral evolution of the gbh , which can be broadly explained by the transition radius between the disc and hot flow progressively decreasing as the mass accretion rate increases , while the atolls are consistent with an identical accretion flow behaviour but with the addition of the expected boundary layer / surface emission ( dg03 )
. these observational differences between gbh and atolls are consistent with the existence of the event horizon , the most fundamental predicted property of a black hole ( see also mcclintock , narayan & rybicki 2004 ; narayan & heyl 2002 ; garcia et al .
2001 ; sunyaev & revnivtsev 2000 ) . however , the sample of atolls used by dg03 was fairly small , with only 4 objects ( 1 of which was an accreting millisecond pulsar ) .
this contrasts with their gbh selection , which included all available objects which were not heavily absorbed . here
we present a much wider study using all the available atolls in the _ rxte _ database which are not heavily absorbed to investigate their behaviour in more detail .
we find that while all the atolls ( and millisecond pulsars ) broadly show the same spectral evolution , there are some subtle differences . some of these can be connected to spin / inclination effects , but there are also some differences in transition behaviour which correlate with long term mass accretion rate .
we speculate that these may be linked through the jet being affected by any remnant polar magnetic field , which in turn depends on the accretion rate .
we searched the _ rxte _ database for low magnetic field , disc accreting neutron star systems .
we do not include z sources ( sco x-1 , gx349 + 2 , gx17 + 2 , cyg x-2 , gx51 , gx340 + 0 e.g. kuulkers et al 1996 , cir x-1 , lmc x-2 , hasinger & van der klis 1989 ) as we were interested in the range @xmath8 to compare with the galactic black holes from dg03 .
we also exclude all objects with heavy absorption , defined as @xmath9 @xmath10 , as above this the spectral decomposition becomes more uncertain .
given that most systems are in the galactic plane , this removes many objects ( 4u 162449 , gro j174428 , grs 1747312 , sax j1747.02853 ) .
the absorption criteria also excludes objects where there is substantial intrinsic absorption .
this can be continuous , completely blocking our view of the continuum source so that the only scattered flux is seen ( the accretion disc corona sources ) , or the absorption can be episodic , giving discrete dips in the lightcurve ( the dippers ) .
we exclude all accretion disc corona sources ( 2s 0921 - 630 , 4u 1624 - 49 , 4u 1822 - 371 , ms 1603 + 206 ) , and dippers where the absorption events are so frequent that most observations are affected by them ( exo 0748 - 676 ) , but we do include dippers where the absorption is limited to contaminating only a few observations ( which we remove , see next section ) .
details of all included sources are shown in table [ tab : sources ] . [ cols="^,^,^,^,^,^,^,^,^,^ " , ]
the standard products spectra ( and associated background and response files ) from the proportional counter array ( pca ) , available from the high energy astrophysics science archive research centre ( heasarc ) database were gathered for each source , giving a total of over 3000 spectra . following dg03 we fit these with a model of the disc emission and a comptonized continuum , described by diskbb ( mitsuda et al .
1984 ) and thcomp ( zdziarski et al . 1996 ) in xspec , together with a smeared edge and gaussian line to approximately mimic the reflected spectrum . the absorption was fixed at the galactic value for each object ( see table [ tab : sources ] ) .
this model gave a good fit to all the spectra ( @xmath11 ) .
following dg03 , the best - fitting model to each spectrum was integrated to form intrinsic soft and hard colours ( i.e. absorption corrected and independent of the instrument response ) , defined as flux ratios in the bandpasses 46.4/34 and 9.716/6.49.7 kev , respectively .
the unabsorbed model was also used to estimate the bolometric flux in the 0.011000 kev bandpass , and this converted into a luminosity using the distance estimate given in table [ tab : sources ] . as discussed by dg03 ,
the colours are fairly robust to changes in the model spectra as long as the source is not heavily absorbed and the advantage of using intrinsic colours ( as opposed to the more widely used instrument colours ) is that many objects can be directly compared on the same plot .
the bolometric fluxes are more uncertain , as the model is extrapolated outside of the observed energy bandpass .
nonetheless , this model corresponds to expected physical components , and we fix a lower limit to the disc temperature of @xmath12 kev and an upper limit to the comptonizing temperature of @xmath13 kev so the continuum components can not produce arbitrarily large luminosities in the unobserved energy ranges ( dg03 ) .
we identify and exclude data contaminated by dips and x - ray bursts by using the standard 1 lightcurves ( 0.125s time resolution ) corresponding to each standard 2 spectra .
we used the intrinsic r.m.s .
variability as a tracer of these .
all lightcurves with r.m.s above 50 percent were inspected for bursts , while the 16-s rebinned lightcurves were checked for dips if their r.m.s
. was above 20 percent .
observations were also excluded if they were contaminated by galactic ridge emission , seen as a softening of the spectra and large iron line contribution at the lowest luminosities in objects with low galactic latitude : see e.g. wardziski et al .
( 2002 ) for an example of this in the black hole gx339 - 4 .
[ fig : landr ] shows the resulting colour - colour and colour - luminosity diagram for all 34 atolls and 6 millisecond pulsars combined together .
one interesting thing to note is that several atoll sources approach or even exceed eddington luminosities ( 4u 1735 - 44 ; 4u 1744 - 26 ( gx 3 + 1 ) ; 4u 1758 - 20 ( gx 9 + 1 ) ; xte j1806 - 246 : gladstone 2006 ) , showing that the high luminosity sources are not necessarily z sources ( hasinger & van der klis 1989 )
[ fig : landr](a ) shows that while the atolls show the same overall behaviour as claimed by dg03 , they also show subtle but significant differences .
this is most evident in terms of the spectral evolution during the transition between the hard ( island ) and soft ( banana ) spectral states , which forms the middle branch of the total z - shaped track on the colour - colour diagram ( even though these are atolls , not z sources , see gierliski and done 2002 ; muno , remillard & chakrabarty 2002 ) .
we have highlighted this by picking out in red those objects where the transition makes an almost vertical track on the colour - colour diagram ( hereafter called verticals ) , while the blue symbols show those where the transition from island to banana starts much further to the right on the colour - colour diagram , but ends at approximately the same place , making a much more diagonal middle branch ( hereafter called diagonals ) .
this difference in spectral evolution is also picked out in the colour - luminosity diagram as a different luminosity for the hard - soft transition . while the distances are generally rather uncertain , this correlation between the behaviour on the colour - colour and colour - luminosity diagrams give some confidence in the reality of the luminosity difference .
we stress that this is not primarily due to hysteresis , the well known effect in both black holes and neutron stars where the luminosity at which the hard - soft transition occurs varies considerably in the same object ( e.g. nowak 1995 , maccarone & coppi 2003 ) . here ,
the transition luminosity is varying between different objects .
we show this difference in transition behaviour by plotting the colour - colour and colour - luminosity diagrams for each individual source which shows a clear state transition .
those making a vertical transition are shown in fig .
[ fig : verticals ] , while those making a diagonal transition are shown in fig .
[ fig : diagonals ] . not all objects have data covering a transition .
some , such as all the millisecond pulsars , are only seen in the hard ( island ) state ( see fig . [
fig : msps ] ) , others e.g. 4u 1758 - 20 are seen only in the soft ( banana ) branch .
of the ones which do make transitions , showing both hard and soft spectra , the effects of data windowing mean that not all have observations covering the transition period .
nonetheless , the size of the _ rxte _ database mean there are 12 sources which do have enough data to constrain the shape of the transition on the colour - colour diagrams .
the remaining sources , where there is insufficient transition data available at present are included in the background points plotted on each image , but not plotted individually here ( they can be seen in gladstone 2006 ) .
the majority of sources have well defined transition luminosities ( to within a factor 23 ) , irrespective of whether the transition is from hard to soft i.e. on the rising part of the light curve , or from soft to hard i.e. with a decreasing flux . however , there are two sources where the transitions have a much large scatter in their properties , namely 4u 160852 and 4u 1908 + 005 ( aql x-1 ) .
the colour - colour and colour - luminosity diagrams for these are plotted as the bottom two panels in fig .
[ fig : verticals ] ) , but it is not clear from the data that these sources should be classed as verticals , so here we examine them in more detail .
these two systems are also the only two ( apart from the millisecond pulsars ) which show large scale outbursts . in order to investigate their scatter in transition
behaviour we select simple outbursts from the light curve , where there is a clear fast rise followed by a monotonic decay . from this data
we build a new , simplified colour - luminosity diagram ( fig .
[ fig : hysteresis ] ) .
the black triangular points show the island state seen on the rise , which looks very similar to the msps ( fig .
[ fig : msps ] ) .
we then exclude all of the transition data on the rise / peak of each outburst , and plot the remaining banana branch and transition back to island state on the decay as the blue square points .
plainly the transition luminosity is much larger in the rise to outburst than during the decay .
this effect of hysteresis is also seen in most galactic black holes ( e.g. maccarone & coppi 2003 ) .
equally plainly , the decay shows a clear vertical track , confirming the identification of these as verticals .
hence we speculate that the msps are similar objects , and that if the msp outburst ever reached high enough luminosities to make a transition then this transition would display hysteresis and the colour - colour track during the outburst decay would be vertical .
none of the diagonals show marked hysteresis , though 4u 170544 and ks 173126 may show a small effect in that their top branch ( island state ) extends to higher luminosities than seen during the transition ( fig .
[ fig : diagonals ] ) .
a similar small hysteresis effect may also be present in the vertical 4u 091954 .
there are subtle differences in the spectral evolution of the atolls on the colour - colour and colour - luminosity diagrams : there is the distinction between the transition track ( verticals and diagonals ) , and then within the diagonals there is a difference in spectral hardness of the banana branch , and within the verticals there are objects which show large scale hysteresis . here
we examine possible explanations for these effects .
a range of inclinations is expected for these sources , and this could be important for the observed spectrum if the intrinsic emission is not isotropic .
some degree of anisotropy is certainly expected from the accretion disc due to its planar geometry , and as long as the boundary layer has a different angular dependence then the overall spectrum will change as a function of inclination . to estimate the effects of inclination we use the spectral model consisting of the disc emission ( diskbb ) and the optically thick boundary layer emission , which we model by thermal comptonization ( comptt ) .
dg03 showed that the banana branch could be roughly characterized by a disc varying in temperature in the range 1.01.5 kev together with an equal luminosity boundary layer with @xmath14 = 3 kev and @xmath15 = 5 ( see also revnivtsev & gilfanov ( 2006 ) ) .
we use these parameters to model the shape of the banana branch .
we first assume that the boundary layer is isotropic , while the disc normalization varies as @xmath16 but with angle averaged luminosity equal to that of the boundary layer .
thus the apparent ratio of disc to boundary layer flux will change as the inclination angle varies . in fig .
[ fig : inclination](a ) we show how the banana branch moves on the colour - colour diagram for inclinations of @xmath17 , @xmath18 and @xmath19 .
the change is mainly only in the soft colour , with a greater proportion of disc emission shifting the start of the banana branch to lower soft colours .
plainly this can not account for either of the two subtle spectral effects seen , but seems to have more potential to explain the different positions of the banana branch amongst the diagonals than the difference between verticals and diagonals .
examination of some of the individual banana branch spectra of the diagonals showed that the comptonization temperature of the boundary layer is also changing , and in a way which is correlated with @xmath20 .
this could indicate that the boundary layer emission is itself anisotropic , perhaps with a temperature gradient so that it is hotter close to the equatorial plane where the disc hits the neutron star .
[ fig : inclination](b ) shows the effect of changing the temperature of the boundary layer from 2.5 , 3.0 and 3.5 kev , for inclination angles of @xmath17 , @xmath18 and @xmath19 , respectively .
this matches very well with the range of banana branches seen in the diagonals .
testing this would require knowledge of the inclination , but very few sources have good enough orbital determinations to constrain this
. however , any source exhibiting dips must be at fairly high inclination ( @xmath21 ) , and hence would be expected to have a banana branch starting at fairly large hard and soft colours .
this can be seen in the case of 4u 1724307 ( see fig [ fig : diagonals ] ) , a dipping source whose banana branch ranges from 1.651.75 in soft colour and 0.7 to 0.8 in hard colour
. there are three other dipping sources , 4u 170430 , 4u 174637 and 4u 191505 ( not included in this work ) , that also support this idea , with each exhibiting large values of both hard and soft colours in the banana state ( for details , see gladstone 2006 ) .
burst oscillations and khz qpos have finally given observational constraints on lmxb neutron star spins ( e.g. the review by van der klis 2000 ) .
for our sample , the inferred range is from 330 to 619 hz .
however , this is unlikely to be the origin of the difference in transition track as the spins seem fairly evenly distributed between the two classes .
however , spin could also give a difference in the overall luminosity / temperature of the boundary layer .
lower spin gives a higher relative speed between the inner edge of the disc and the surface , leading to a lower @xmath20 ratio and a higher boundary layer temperature , as required for the different banana branch hard colours seen .
thus it seems likely that some combination of spin and inclination effects are responsible for the different banana branches seen in the diagonals , but neither is likely to explain the origin of the two different types of transition behaviour . of the subsample of systems with data covering the transition , the publically available asm and/or pca galactic bulge long term light curves
show large scale transient outbursts only in 4u 160852 , 4u 1908 + 005 and all the millisecond pulsars ( see table 1 ) .
this is very intriguing as 4u 160852 and 4u 1908 + 005 are both verticals .
however , examination of the long term lightcurves of the other verticals clearly shows that these objects do not undergo dramatic outbursts , so this can not be the origin of the vertical / diagonal distinction . however , 4u 160852 and 4u 1908 + 005 are also the only objects to show clear large - scale hysteresis .
thus it seems possible that hysteresis is linked to large amplitude outbursts .
there are no counterexamples i.e. no systems that show major disc outbursts ( which push the accretion rate high enough to make a spectral transition ) which do not show hysteresis .
all the millisecond pulsar outbursts remain in the hard state , and all other atoll sources with enough data to deliniate the transition and hence constrain hysteresis do not exhibit major outbursts .
this hypothesis can be tested on the black hole lmxb systems . unlike the atolls , none of these sources are persistant ( mcclintock & remillard 2006 ) .
this distinction in disc instability behaviour between black holes and neutron stars can be broadly explained in the context of the hydrogen ionization models .
neutron stars are lower in mass than black holes , so for the same companion star to overflow its roche lobe requires a smaller binary orbit for a neutron star compared to a black hole .
smaller orbital separation means a smaller disc due to tidal truncation .
the size of the disc , together with the mass loss rate from the companion star ( determined by its evolutionary state ) determines the temperature of the coolest part of the disc .
thus the smaller neutron star systems are less likely to have a disc which can drop below the hydrogen ionization temperature required to trigger the disc instability ( king & ritter 1998 ; king , kolb & szuszkiewicz 1997 ; dubus et al 1999 )
. it is well known that most black hole systems show large scale hysteresis ( e.g. maccarone & coppi 2003 ) .
thus they support the idea above that hysteresis is linked to large amplitude outbursts .
even more support comes from the one obvious exception , which is cyg x-1 .
this does not show hysteresis during its hard / soft transitions , and is of course a persistent ( hmxrb ) source ( e.g. maccarone & coppi 2003 ; dg03 ) .
maccarone & coppi ( 2003 ) also suggested that hysteresis could be linked to large scale outbursts , but with only cyg x-1 as an counterexample , the connection to outbursts is not quite so unambiguous .
cyg x-1 is also one of the few known galactic black holes with a high mass companion , so there is the possibility that its accretion structure is somewhat different due to lower angular momentum material from a stellar wind . by contrast , with the neutron stars being generally stable to the hydrogen ionization trigger for the disc outbursts , there are many counterexamples , all with low mass companions .
we can also rule out the alternative origin for hysteresis discussed by maccarone & coppi ( 2003 ) , namely that it is produced simply by a difference in behaviour in low mass x - ray binaries between a hard to soft transition on the rising part of the light curve , and a soft to hard transition as the flux decreases .
most of the sources shown in figs .
[ fig : diagonals ] and [ fig : verticals ] make the transition in both directions , and do not show large scale hysteresis ( though there are smaller scale effects , consistent with the difference in environment : meyer & meyer - hoffmeister 2005 ) . instead , hysteresis is seen where the mass accretion rate changes dramatically ( from @xmath22 to @xmath23 ) .
we speculate that the accretion flow is able to access non - equilibrium states during the rapid changes in accretion disc structure caused by the disc instability .
thus the neutron stars clearly show a one to one correlation between the dramatic large amplitude outbursts triggered by the disc instability and hysteresis , consistent this being the origin of hysteresis .
this predicts that the millisecond pulsar outbursts should also show hysteresis if their accretion rate at the outburst peak ever goes high enough to trigger a spectral transition ( in which case we would expect them also be verticals , along with 4u 160852 and 4u 1908 + 005 ) . however , there is no such one to
one correlation between outbursts and transition behaviour ( vertical / diagonal ) , so we explore further aspects of the systems below .
table [ tab : sources ] shows the binary parameters where these are known .
we first explore whether there is any correlation between transition behaviour and binary period ( which is a tracer of companion type , and also of superburst behaviour : cumming 2003 ) . for the verticals ,
the periods span a large range , from 1 hour for the persistent sources 4u 0614 + 091 and 4u 091954 through to 19 and 288 hours for the transients 4u 160852 and 4u 1908 + 005 .
this distinction is as expected from the disc instability model ( wide binary implies a much larger disc , so a cooler outer edge which is more likely to trigger the hydrogen instability ) .
if the millisecond pulsars are also verticals then these fill in the period distribution from 40 minutes to 4 hours.there is very little data to compare this to the diagonals , as only 2 have periods ( 11 minutes and 3.9 hours ) , but the broad span seen from the verticals encompasses most of the binary periods seen in atolls , so binary period alone is unlikely to be the origin of the transition track dichotomy .
the distinction between millisecond pulsars , which plainly retain a residual magnetic field channelling the flow , and other atolls which do not show pulsations is most probably due to the long timescale mass accretion rate ( cumming , zweibel & bildsten 2001 ) .
plausibly , high accretion rates can bury the magnetic field below the neutron star surface , but the very low average mass accretion rate in the millisecond pulsars is insufficient to bury the field ( cumming et al .
thus there is a potential physical mechanism for the long term mass accretion rate to change the properties of the accretion flow .
we estimate the long - term mass accretion rate , @xmath24 , ( or corresponding average luminosity , @xmath25 ) for all of the systems from the observed x - ray emission .
the pca data gives estimates for @xmath26 through spectral fitting , but the light curves are highly incomplete , so the data can not simply be used as an indicator of the long term average mass accretion rate . by contrast , the _ rxte _ all sky monitor ( asm ) gives an almost continuous light curve for every bright x - ray source in its field of view , but its lack of spectral resolution means that going from count rate to @xmath27 is highly uncertain . hence we combine the two approaches .
we use the pca light curve to define the average @xmath25 during these observations , then select the simultaneous asm points to find the average asm count rate during the pca observations .
the ratio of this to the full asm light curve gives the correction for the incompleteness of the pca observations .
this approach works well unless the source becomes very faint , in which case contamination of the asm by other nearby sources and/or galactic ridge emission can be a problem .
the only sources for which this is an issue are the transients i.e. the verticals 4u 160852 and 4u 1908 + 005 and all the millisecond pulsars .
the outbursts of 4u 160852 and 4u 1908 + 005 are so bright that they dominate the average of the asm lightcurves . however , this is not the case for the millisecond pulsars , so for these we use a different approach .
here we use the pca data alone , which have good outburst coverage , assuming that the source intensity is negligible in the periods outside the known outbursts .
the results of both the millisecond pulsars and atoll sources are shown in table [ tab : lbol ] .
ccc source name & source type & @xmath28 + + igr j00291 + 5934 & msp & 1.4 @xmath29 + xte j0929314 & msp & 1.4 @xmath30 + xte j1751305 & msp & 5.2 @xmath29 + xte j1807294 & msp & 2.7 @xmath30 + xte j1808369 & msp & 6.7 @xmath29 + xte j1814338 & msp & 1.6 @xmath30 + + 4u 0614 + 091 & v & 3.2 @xmath31 + 4u 091954 & v & 3.1 @xmath31 + 4u 160852 & v & 6.6 @xmath31 + 4u 1908 + 005 & v & 3.7 @xmath31 + slx 1735 - 269 & v & 1.6 @xmath31 + + 4u 163653 & d & 4.4 @xmath32 + 4u 170243 & d & 7.8 @xmath31 + 4u 170544 & d & 1.9 @xmath32 + 4u 1724307 & d & 7.3 @xmath31 + 4u 172834 & d & 1.9 @xmath32 + 4u 1820303 & d & 1.3 @xmath33 + ks 173126 & d & 3.4 @xmath32 + + it can be seen that there is a systematic difference between the three classes of sources .
the millisecond pulsars have the lowest @xmath28 , then the verticals , then the diagonals .
thus it seems possible that this is the origin of the difference in transition properties .
the millisecond pulsars low @xmath34 allows the field to diffuse out of the crust , and to be strong enough to collimate the flow and produce pulsations .
one possibility might be that while the high @xmath35 of the diagonals suppresses the field entirely , the medium @xmath34 in the verticals allows some field to diffuse out .
but the verticals show no trace of pulsation , so the magnetic field can not collimate any significant part of the flow .
one way around these pulsation limits is to separate the field and accretion flow .
any non - spherical accretion flow will predominantly bury the field in the region of the flow , leaving the field to escape in regions with little accretion .
even the hot accretion flow envisaged for the hard island states favours the equatorial plane , with a geometry which is more like a thick disc than a truly spherical flow ( e.g. narayan & yi 1995 ) .
this nicely circumvents the pulsation limits , but then there is no interaction between the magnetic field and the flow , and so no physical mechanism to change the behaviour of the transition . however , one aspect of the system that a polar magnetic field could affect is the jet .
the most recent numerical simulations of the accretion flow magnetohydrodynamics show a causal link between the jet and accretion flow ( hawley & krolik 2006 ; mckinney 2006 ) , so this might give an indirect link between the magnetic field and accretion flow properties . one way to test this is to look at the radio emission from these systems .
while theoretical models of jets are not well developed , we can use the observed gbh behaviour as a template .
these show a clear correlation between radio and x - ray luminosity in their low / hard state , showing that @xmath36 is important in determining the power of the jet ( gallo et al .
2006 ) . hence to see whether there are differences in the jet emission between the millisecond pulsars , verticals and diagonals we need to compare the radio emission at the same @xmath36 .
hysteresis gives potential problems , so ideally the comparison would be between persistent verticals and diagonals in the hard island state at the same @xmath37 .
however , there is very little data to make this comparison , with only 4u 0614 + 091 and 4u 172834 for the persistent verticals and diagonals , respectively , and these do not overlap in @xmath37 ( migliari & fender 2006 ) . thus we speculate that the origin of the difference in transition behaviour between the verticals and diagonals is due to the presence of some magnetic field at the pole in the verticals which affects the accretion flow indirectly through jet formation , by contrast to the diagonals which have higher mass accretion rates , sufficient to bury the field everywhere .
the atolls and millisecond pulsars are consistent with showing the same overall spectral evolution with changing mass accretion rate , but there are some subtle differences .
the spectral shape of the soft ( banana ) branch shows subtle variations from object to object , most probably due to combination of inclination and spin changing the ratio of the observed disc to boundary layer luminosity , and boundary layer temperature . however , there are also clear differences in behaviour during the hard / soft transition which point to a more fundamental distinction .
the data shows two types of sources , those where the transition makes a vertical track on the colour - colour diagram , and occurs at @xmath0 , and those which make a diagonal track on the colour - colour diagram with the transition at @xmath38 .
there are hysteresis effects in individual sources which introduce dispersion in the transition luminosity , but these are large _ only _ for the outbursting atolls ( which are both verticals ) . splitting these outbursts into the rapid rise and slow decay phases
show that the rapid rise looks like the millisecond pulsars ( so they are probably also verticals ) while the slow decay looks like the persistent verticals .
thus it seems likely that large scale hysteresis effects are only seen in sources where the disc structure changes rapidly due to the onset of the hydrogen ionization instability .
this is also consistent with the observed black hole behaviour .
the association of the millisecond pulsars with verticals suggests that the difference in transition is ultimately linked to the surface magnetic field , and indeed , all the verticals have long term mass accretion rates which are smaller than those of the diagonals , though not as small as those of the millisecond pulsars . thus the verticals could have some small b field which is able to affect the inner accretion flow , but this must be indirect as otherwise these systems also would show pulsations .
we speculate that the physical link between the magnetic field ( predominantly polar ) and accretion flow ( predominantly equatorial ) may be due to the changes in the jet , which would be testable with more radio data on these sources .
this research has made use of data obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center . | we systematically analyze all the available x - ray spectra of disc accreting neutron stars ( atolls and millisecond pulsars ) from the _ rxte _ database .
we show that while these all have similar spectral evolution as a function of mass accretion rate , there are also subtle differences .
there are two different types of hard / soft transition , those where the spectrum softens at all energies , leading to a diagonal track on a colour - colour diagram , and those where only the higher energy spectrum softens , giving a vertical track .
the luminosity at which the transition occurs is _ correlated _ with this spectral behaviour , with the vertical transition at @xmath0 while the diagonal one is at @xmath1 .
superimposed on this is the well known hysteresis effect , but we show that classic , large scale hysteresis occurs only in the outbursting sources , indicating that its origin is in the dramatic rate of change of mass accretion rate during the disc instability . we show that the long term mass accretion rate correlates with the transition behaviour , and speculate that this is due to the magnetic field being able to emerge from the neutron star surface for low average mass accretion rates .
while this is not strong enough to collimate the flow except in the millisecond pulsars , its presence may affect the inner accretion flow by changing the properties of the jet .
= -0.5 cm [ firstpage ] accretion , accretion discs x - rays : binaries , atoll sources |
titanium carbide is one the hardest transition metal binary carbides known , which renders it resistant to wear and makes it a favorable material for applications such as drill bits and cutting tools . with a melting point of over 3000@xmath0c , it is thermally quite stable .
however , transition metal binary carbides are brittle , difficult to machine , and highly susceptible to thermal shock .
the ternary carbides known as @xmath14 ( max ) phases @xcite ( where @xmath15 = 1,2 , or 3 and @xmath16 is a transition metal , @xmath17 is a group a element mostly from groups 13 and 14 , and @xmath18 is either c or n ) often overcome some of these shortcomings , while still being heat tolerant . as a class , the max phases have unusual yet attractive and often unique combinations of properties that bring together some of the best attributes of ceramics and metals @xcite . like metals , they are excellent electric and thermal conductors , with exceptional thermal shock resistance and damage tolerance @xcite . in some cases , they are creep @xcite , oxidation @xcite , and fatigue @xcite resistant .
furthermore , they are elastically quite stiff , yet readily machinable @xcite . in this work
we are interested in the max phases in the ti
al c system .
both ti@xmath3alc [ fig .
[ fig:523struct](a ) ] and ti@xmath4alc@xmath3 [ fig .
[ fig:523struct](b ) ] have been relatively well studied .
the crystal structure of ti@xmath3alc was first solved in the 1960s @xcite ; ti@xmath4alc@xmath3 was discovered several decades later in 1994 @xcite . of the @xmath19 60 max phases known to date , ti@xmath3alc and ti@xmath4alc@xmath3 are particularly attractive in terms of high temperature applications .
they are two of the most lightweight and oxidation resistant max phases @xcite , and the accessibility and relative low cost of their raw materials render them the most promising for up - scaling and industrialization .
alc , ( b ) ti@xmath4alc@xmath3 , and ( c ) ti@xmath2al@xmath3c@xmath4 showing lattice parameters @xmath10 and @xmath9 and the unique ti , al , and c lattice sites as constrained by symmetry , along with the refined atomic position @xmath20 parameters . ]
in addition to ti@xmath3alc , henceforth referred to as 211 , and ti@xmath4alc@xmath3 , henceforth referred to as 312 , there is also a ti@xmath2al@xmath3c@xmath4 or 523 phase , which is in a category of higher - order max phases .
the latter were first reported in 2004 in the ti
c system ( _ i.e. _ ti@xmath2si@xmath3c@xmath4 and ti@xmath21si@xmath3c@xmath2 ) by palmquist et al @xcite .
wilhelmsson et al .
@xcite reported the existence of ti@xmath2al@xmath3c@xmath4 in 2006 . in both cases
, these phases were only observed in local stacking sequences in transmission electron microscopy , tem , micrographs .
more recently , the stacking sequence of ti@xmath2al@xmath3c@xmath4 was characterized using x - ray diffraction , xrd , in two studies @xcite . in our study @xcite ,
the characterized sample contained @xmath22 wt.% ti@xmath2al@xmath3c@xmath4 .
in another study @xcite , only a small amount of ti@xmath2al@xmath3c@xmath4 was observed and neither weight nor volume fractions were reported .
we note in passing that the first structure proposed by wang _
@xcite is totally wrong and unsubstantiated by the results shown in that paper ( see ref .
it is also crucial to note that our paper was submitted a few weeks before that of ref .
interestingly , ref .
@xcite was submitted , reviewed , and accepted in exactly one week and published soon thereafter .
it was only after our paper was published that the same group , working with a composition that was only 19.7 wt% ti@xmath2al@xmath3c@xmath4 , made the case that the space group was @xmath23 @xcite .
like most other max phases , ti@xmath3alc and ti@xmath4alc@xmath3 both have layered hexagonal structures [ figs . [ fig:523struct](a ) and [ fig:523struct](b ) , respectively ] belonging to space group @xmath24 ( no .
194 ) . in both structures ,
c layers ( ti@xmath3c for ti@xmath3alc and ti@xmath4c@xmath3 for ti@xmath4alc@xmath3 ) are interleaved between layers of al . the ti@xmath2al@xmath3c@xmath4 phase , on the other hand [ fig .
[ fig:523struct](c ) ] , consists of alternating ti@xmath3c ( `` 211-like '' ) and ti@xmath4c@xmath3 ( `` 312-like '' ) layers interleaved between al layers .
due to the shift in stacking sequence , three formula units need to be included in a unit cell .
the lattice parameters are thus @xmath10=@xmath25 and @xmath9=@xmath26 @xcite .
while xrd and tem have been used to characterize the room temperature lattice parameters of ti@xmath2al@xmath3c@xmath4 and verify its stoichiometry and stacking sequence @xcite , a complete structure refinement to determine bond lengths has not yet been performed .
furthermore , its thermal expansion coefficients , tecs , have not been measured .
previously , high - temperature neutron diffraction , htnd , was used to investigate the high - temperature stability of ti@xmath3alc @xcite and ti@xmath4alc@xmath3 @xcite up to 1550@xmath0c , but in refs . @xcite and @xcite ,
only the temperature - dependent phase fractions were studied with no reports of high - temperature crystallographic data . in @xcite ,
the lattice parameters and tecs were reported for ti@xmath4alc@xmath3 , but the bond lengths have yet to be reported for both ti@xmath3alc and ti@xmath4alc@xmath3 . the original aim of this work was to determine the crystal structure parameters of phase - pure ti@xmath3alc . however , for reasons that are not fully understood , during processing , ti@xmath3alc decomposed to yield ti@xmath2al@xmath3c@xmath4 and some ti@xmath4alc@xmath3 @xcite . and while obtaining the relevant crystal structure parameters on such a multiphase sample is not ideal we decided to proceed with this htnd study nevertheless for the following reasons : a. to date , neither we nor others have been able to synthesize phase - pure ti@xmath2al@xmath3c@xmath4 , and the results obtained here are better than no results ; b. given the very similar crystal structures , elastic properties , and tecs among the three phases ( see below ) , the deviations from phase - pure behavior due to the presence of other phases should be small ; c. careful rietveld analysis can deconovolute the contributions from each phase . these comments notwithstanding , to estimate the error in the various parameters as a function of temperature , we compare the results obtained on the multiphase sample with a predominantly single - phase sample of ti@xmath4alc@xmath3 and
show that the differences , for the most part , are equal to ( or less than ) the experimental uncertainty .
herein we use htnd on a sample consisting of 38(@xmath271 ) wt.% ti@xmath2al@xmath3c@xmath4 ( `` 523 '' ) , @xmath28 wt.% ti@xmath3alc ( `` 211 '' ) , @xmath29 wt.% ti@xmath4alc@xmath3 ( `` 312 '' ) , and @xmath30 wt.% of an additional intermetallic tial phase to determine the temperature - dependent crystal structures of all three ti
al c carbide phases .
we report on the temperature evolution of the lattice parameters , isotropic thermal atomic displacement parameters , adps , and bond lengths during both heating and cooling for ti@xmath3alc and ti@xmath4alc@xmath3 .
we verify our results by comparing the room temperature lattice parameters and lattice expansions to previous studies of predominantly single - phase ti@xmath3alc and ti@xmath4alc@xmath3 samples @xcite .
we also present htnd results for a predominantly single - phase ti@xmath4alc@xmath3 sample for comparison .
the sample used herein was prepared using pre - reacted ti@xmath3alc powders that were commercially obtained ( kanthal , hallstahammar , sweden ) .
the powders were cold isostatic pressed , ciped , at 200mpa and heated at a rate of 300@xmath0c h@xmath13 to 1500@xmath0c , then sintered for 2 h under a hydrogen atmosphere .
the ti@xmath4alc@xmath3 sample was prepared by hot pressing , hping , pre - reacted ti@xmath3alc ( kanthal , hallstahammar , sweden ) and titanium carbide ( alfa aesar , ward hill , ma ) in a 1:1 ratio to make 3:1:2 stoichiometry of ti : al : c .
powders were ball milled for 24 h , placed in a graphite die , and heated in a graphite - heated hot press under a vacuum of 10@xmath13 torr at a rate of 500@xmath0c h@xmath13 to 1400@xmath0c . it was held for 4 h under a pressure of @xmath31 mpa before cooling . in both cases , bulk samples 9 mm in diameter and 3 cm high were used for the htnd experiments .
the htnd experiments were conducted on the high - pressure preferred orientation ( hippo ) neutron diffractometer @xcite at the lujan neutron scattering center , los alamos national laboratory .
for both the multiphase ti
al c sample and ti@xmath4alc@xmath3 , bulk samples were placed in a vanadium , v , holder ( 9 mm diameter , 0.15 mm wall thickness ) , mounted in an ill - type high - temperature vacuum furnace with a v setup , and heated at a rate of 20@xmath0c min@xmath13 .
data were collected every 100@xmath0c during heating from 100@xmath0c to 1000@xmath0c . for the multiphase sample only ,
data were collected every 200@xmath0c upon cooling as well . at each measurement , the temperature was held constant during data collection and neutrons were detected with 42 panels of @xmath32he detector tubes arranged on five rings with nominal diffraction angles of 39@xmath0 , 60@xmath0 , 90@xmath0 , 120@xmath0 , and 144@xmath0 . to allow texture analysis ,
the sample was measured at rotation angles of 0@xmath0 , 22@xmath0 , and 45@xmath0 around the vertical axis for each temperature , with a count time of 15 min per orientation for a total count time of 45 min per temperature . a texture analysis using the material analysis using diffraction , maud , code @xcite showed a mild fiber texture for all temperature points . as the texture was weak ,
the data from the three rotations and the detector rings were integrated , resulting in one histogram per detector bank . due to the large detector coverage of hippo ,
this procedure is similar to spinning the sample to randomize the preferred orientation .
the neutron time - of - flight data were analyzed as texture - free powders with the rietveld method using the general structure analysis system , gsas @xcite .
only the higher resolution data from the 90@xmath0 , 120@xmath0 , and 144@xmath0 banks were used in the analysis .
the maud refinements , which incorporated preferred orientation , gave parameters that were within error bars of those determined without including texture by rietveld refinement with gsas .
therefore , all results reported herein are from the gsas refinements assuming random texture .
the _ gsaslanguage _ refinement script @xcite was used to ensure that identical refinement strategies were employed for all temperatures . the instrument alignment ( difc parameter in gsas )
was fixed for the backscattered ( 144@xmath0 ) detector bank , which has the highest resolution , and refined for the 90@xmath0 and 120@xmath0 bank for the lowest temperature run at 100@xmath0c . for subsequent runs ,
difc was fixed for all three banks .
refined parameters were 16 background parameters of gsas background function # 1 , lattice parameters of all phases , phase fractions , @xmath33 profile parameter for peak width , atomic positions , and isotropic thermal motion parameters .
the atomic positions in ti@xmath3alc [ @xmath20 coordinate of ti ; see fig .
[ fig:523struct](a ) ] and in ti@xmath4alc@xmath3
[ @xmath20 coordinates of ti@xmath34 and c ; see fig . [ fig:523struct](b ) ] were both refined in the @xmath24 space group . for ti@xmath2al@xmath3c@xmath4
, there are six unique atomic sites : ti@xmath34 , ti@xmath35 , and c@xmath34 with 312-like stacking , ti@xmath36 and c@xmath35 with 211-like stacking , and one al site between 211- and 312-stacked octahedra throughout the cell [ see fig . [ fig:523struct](c ) ] . the best space group to represent these positions
was found to be @xmath37 ; therefore , the @xmath20 coordinates of ti@xmath34 , ti@xmath36 , al , and c@xmath34 were refined according to the constraints induced by the trigonal @xmath37 space group .
in addition to the three max phase carbides , peaks corresponding to an intermetallic that crystallizes like @xmath38-tial , with tetragonal space group @xmath39 @xcite , were found .
refinement of the site occupancy factor on the ti site led to 50% ti and 50% al antisite defects , yielding a stoichiometry of ( ti@xmath8al@xmath8)al .
given the nearly null scattering intensity of the mixed site , and the low phase fraction of this phase , the isotropic thermal motion parameters were constrained together to reduce the number of variables .
first - principles phonon calculations were used to calculate the anisotropic mean - squared atomic displacements .
the dft calculations were performed using the projector - augmented wave ( paw ) method @xcite , as implemented in the vienna ab initio simulation package ( vasp ) code @xcite . the generalized gradient approximation ( gga ) exchange - correlation functional of perdew - burke - ernzerhof ( pbe )
was used @xcite with a cutoff of 500 ev .
the total energy was converged to 10@xmath40 ev , with a @xmath41-point grid of @xmath42 for ti@xmath3alc and ti@xmath4alc@xmath3 and @xmath43 for t@xmath2al@xmath3c@xmath4 .
the real - space force constants were calculated using density functional perturbation theory ( dfpt ) @xcite implemented in the vasp code .
the phonon frequencies were calculated from the force constants with phonopy @xcite , and the adps were calculated from the frequencies and eigenvectors of the force constant matrix .
further details on the adp calculations can be found in ref .
1% range . ] .profile agreement factors rfactors for rietveld refinements of neutron diffraction data collected during heating and cooling for the multiphase ti
al c sample .
[ cols="<,^,^,^",options="header " , ] @xmath44 bond expansion : @xmath45 from least - squares fit of @xmath46 vs. @xmath47 .
while the overall expansions and anisotropies in the three max - like phases are comparable , the most interesting aspect of this work is the relationship between bond length evolution and the stacking of the octahedra among the three phases . in the literature ,
it is fairly well established , both experimentally and theoretically , that the @xmath16c bonds adjacent to the @xmath17 layers ( _ i.e. _ ti@xmath34-c in fig .
[ fig:523struct ] ) are shorter than those in the stoichiometric binary @xmath48 , while the ones that are not , ( _ viz .
_ ti@xmath35c , in fig .
[ fig:523struct ] ) are longer . figure [ fig:523bonds](c ) and the ti
c lengths in table [ table:523bonds ] are fully consistent with this general conclusion .
not surprisingly , the ti c bond lengths in the 211 slab in the 523 phase are almost identical to those of the 211 phase [ fig . [ fig:523bonds](c ) ] .
similarly , the ti
c bond lengths in the 312 phase are very similar to those of the 312 slabs in the 523 phase [ fig . [ fig:523bonds](c ) ] .
this applies not only to the absolute ti
c bond lengths values but also to their thermal expansions , which are quite comparable as well [ fig .
[ fig:523struct](c ) ] .
it should be noted that the longest bonds in the 312-stacked octahedra in ti@xmath2al@xmath3c@xmath4 are slightly shorter than @xmath49 in ti@xmath4alc@xmath3 , while the ti c bonds in the 211-stacked octahedra in ti@xmath2al@xmath3c@xmath4 are slightly longer than those in ti@xmath3alc [ fig . [ fig:523bonds](c ) ] .
this suggests that the structure is slightly more uniform than the individual 211- and 312-stacked phases due to the interleaved nature of the stacking sequences .
these comments notwithstanding , it is clear from fig . [
fig:523bonds](c ) that the same structural units behave similarly .
these results are gratifying because they indirectly validate our rietveld analysis .
the situation for the ti - al bonds is not as clear . since the difference between the 211 and 312 phases is the number of ti
c octahedra between al layers , it is expected that only the ti
c bonds would be significantly affected by the change in ti
c stacking while the al ti bonds should be similar among the three phases .
however , we find that the al ti bonds are clearly affected by stoichiometry [ fig . [ fig:523bonds](d ) ] : @xmath50 is significantly longer in the 211 phase [ black diamonds in fig .
[ fig:523bonds](d ) ] than in the 312 phase [ red crosses in fig .
[ fig:523bonds](d ) ] , while the opposite is true of those bonds in the 523 phase ; _ i.e. _ , @xmath51 in ti@xmath2al@xmath3c@xmath4 [ compare blue circles and blue squares in [ fig:523bonds](d ) ] .
the reasons for this state of affairs are not fully understood , but are likely related to the following observations : a. the ti
c bonds are relatively stiff building blocks of the individual 312 and 211 units , as evidenced by the fact that they stay relatively the same size as in the original ti@xmath4alc@xmath3 and ti@xmath3alc phases when the stacking sequences are interleaved .
the dimension within the crystal that thus has the most flexibility is the al ti bond .
therefore , it is likely that the al ti bond plays a role as an effective `` compensating spring '' in the structure to minimize the crystal energy .
furthermore , this role would be different and most probably more dominant in the more complex ti@xmath2al@xmath3c@xmath4 higher - order phase .
b. among the possible factors that could be compensated for in the flexible al ti bond discussed above are those related to constraints on the lattice parameters especially on the @xmath9-lattice parameter , which essentially determines the al ti bond length , given that the ti
c octahedra are rigid blocks . in a sample with multiple competing phases ,
it is likely that these effects are prominent and manifest themselves in the al ti bond . c. the al atoms in ti@xmath3alc and ti@xmath4alc@xmath3 lie in a mirror plane within the structures , while al is not constrained to mirror symmetry between the ti
c atoms in the ti@xmath2al@xmath3c@xmath4 phase ( see fig . [ fig:523struct ] ) .
therefore , the changes in the al ti distances in the 211-stacked and 312-stacked structures that occur when they are interleaved to form 523 may be a consequence of the symmetry break .
these comments notwithstanding , it is important to note that the average ti al bond length in 523 ( 2.863 ) is equal the average of the al ti bond lengths in 312 and 211 ( also 2.863 ) .
the average expansions of those bonds are also similar ( see below ) . while more work is needed to fully understand the al ti bond length behavior , it can be reasonably concluded that the dimensions of the ti c units are consistent for a given stacking , regardless of whether they are interleaved in a higher - order phase or in a conventional max phase . based on this fact and the inconsistency of the al ti bonds , it is further speculated that the al ti bonds serve to compensate other energy minimization factors for the crystal , especially those related to symmetry and lattice constraints . in ti@xmath3alc and ti@xmath4alc@xmath3 , the al
ti bonds show the highest expansion [ see table [ table:523exp ] and figs . [ fig:523exp](b ) and [ fig:523exp](c ) ] . to our knowledge , there are no previous reports of temperature - dependent bond lengths in any of the ti
al c max phases to which to compare our results . however , in a previous htnd study of the nitride ti@xmath3aln , the al ti bond also showed a higher expansion rate than the ti
n bond @xcite .
similarly , a higher expansion rate was observed for the @xmath17@xmath16 bonds than the @xmath16c bonds in ti@xmath4sic@xmath3 @xcite and cr@xmath3gec @xcite .
this result is also consistent with a high - pressure xrd study of ti@xmath4alc@xmath3 , where the al ti@xmath34 bond was the most compressible , while the ti@xmath34c and ti@xmath35c bonds were more rigid @xcite . in ti@xmath2alc@xmath3 , the expansion rate of the al ti@xmath34 bond
14.6 @xmath12 k@xmath13 is the highest of all the bonds in the sample , but the al ti@xmath36 bond expansion is unexpectedly low , at 5.5 @xmath12 k@xmath13 ( table [ table:523bonds ] ) .
nonetheless , the average bond expansion in 523 ( 10.5 @xmath12 k@xmath13 ) is still similar to the 211 and 312 average ( 11.3 @xmath12 k@xmath13 ) .
also note that the error bars for the al ti bond expansions in ti@xmath2al@xmath3c@xmath4 are the highest of those for the bonds in all phases ( see fig .
[ fig:523exp ] ) .
this uncertainty further suggests that the al ti bond behavior is flexible within the structure and indicates other crystal imperfections and/or symmetry and lattice dimension effects , as discussed above . the results in fig .
[ fig:523adp ] show that , like all other max phases studied to date , the @xmath17 atom al in this case is a rattler in that it vibrates with a significantly higher amplitude than the other atoms in the structures .
the high adps of al , both calculated and experimental , relative to the ti and c atomic displacement values [ figs .
[ fig:523adp](a)-(c ) ] are also consistent with the relatively weaker al bonding evidenced by the higher al ti expansion rates , at least in ti@xmath3alc and ti@xmath4alc@xmath3 ( table [ table:523bonds ] ) and the flexibility of the al ti interaction , as discussed above .
previous htnd studies of ti@xmath4sic@xmath3 @xcite , ti@xmath4gec@xmath3 @xcite , ti@xmath3aln @xcite , cr@xmath3gec @xcite , and ti@xmath52aln@xmath4 @xcite have shown the same rattling phenomenon for the a - group element .
a comparison of the vibrational behavior of al with two other htnd studies of the al - containing nitrides ti@xmath3aln and ti@xmath52aln@xmath4 ( fig . [ fig : u - al - all ] ) further suggests that this `` rattling '' effect is independent of stacking sequence . the results shown in figs . [ fig:523adp ] and [ fig : u - al - all ] also clearly indicate that from a theoretical point of view , the adps of the three max phases should be very comparable .
given that the al atoms in ti
n nitrides are also predicted to behave similarly to one another but different from the carbides in their vibrational amplitudes ( see lines for ti@xmath3aln and ti@xmath52aln@xmath4 in fig .
[ fig : u - al - all ] ) , the dft calculations indicate that the adps of al should not be greatly influenced by stoichiometry .
interestingly , the agreement between theoretical and experimental isotropic adps for the al atoms is quite good in all five compounds plotted in [ fig : u - al - all ] .
the agreement for the other atoms is less good for reasons that are not well understood , but are typical of the max phases @xcite .
rietveld analysis of htnd data for a bulk sample containing ti@xmath2al@xmath3c@xmath4 [ 38(@xmath271 ) wt.% ] , ti@xmath3alc [ 32(@xmath271 ) wt .
% ] , ti@xmath4alc@xmath3 [ 18(@xmath271 ) wt .
% ] , and ( ti@xmath8al@xmath8)al [ @xmath7 wt.% ] has shown that ti@xmath2al@xmath3c@xmath4 exhibits similar thermal expansion and thermal motion parameters as ti@xmath3alc and ti@xmath4alc@xmath3 .
the thermal expansions for ti@xmath2al@xmath3c@xmath4 in the @xmath10- and @xmath9- directions , respectively , are @xmath53 = 9.1(@xmath11 ) @xmath12 k@xmath13and @xmath54 = 9.8(@xmath11 ) @xmath12 k@xmath13 . in all three phases , the average expansion rates of all the al
ti bonds are higher than the average ti c bond expansions .
ti@xmath2al@xmath3c@xmath4 consists of alternating layers of 312- and 211-like stacking , where the 312 layers are similar to ti@xmath4alc@xmath3 and the 211 layers are similar to ti@xmath3alc in dimensions and bond expansions .
the al atoms in all three phases vibrate with higher amplitudes than the ti and c atoms .
this work shows that ti@xmath2al@xmath3c@xmath4 exhibits similar properties to ti@xmath4alc@xmath3 and ti@xmath3alc , two of the most promising max phases , which indicates that phase purity can be more relaxed in processing when considering applications .
in addition , this work shows that further studies on ti@xmath2al@xmath3c@xmath4 can lead to enhanced property optimization and engineering for ternary carbides in the ti
al c system .
this work has benefited from the use of the lujan neutron scattering center at lansce , which is funded by the u.s .
department of energy s office of basic energy sciences .
los alamos national laboratory is operated by los alamos national security llc under doe contract de - ac52 - 06na25396 .
this work was also supported by the army research office ( w911nf-07 - 1- 0628 ) .
the authors would also like to thank d. j. tallman and dr .
t. el - raghy for providing the samples used in this work .
55ifxundefined [ 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 * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop `` , '' ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) | herein we report on the thermal expansions and temperature - dependent crystal structures of select ternary carbide max phases in the ti
al c phase diagram in the 1001000@xmath0c temperature range .
a bulk sample containing @xmath1 wt.% ti@xmath2al@xmath3c@xmath4 ( `` 523 '' ) , @xmath5 wt.% ti@xmath3alc ( `` 211 '' ) , @xmath6 wt.% ti@xmath4alc@xmath3 ( `` 312 '' ) , and @xmath7 wt.% ( ti@xmath8al@xmath8)al is studied by rietveld analysis of high - temperature neutron diffraction data .
we also report on the same for a single - phase sample of ti@xmath4alc@xmath3 for comparison .
the thermal expansions of all the max phases studied are higher in the @xmath9 direction than in the @xmath10 direction .
the bulk expansion coefficients
9.3(@xmath11 ) @xmath12 k@xmath13 for ti@xmath2al@xmath3c@xmath4 , 9.2(@xmath11 ) @xmath12 k@xmath13 for ti@xmath3alc , and 9.0(@xmath11 ) @xmath12 k@xmath13 for ti@xmath4alc@xmath3 are comparable within one standard deviation of each other . in ti@xmath2al@xmath3c@xmath4 , the dimensions of the ti
c octahedra for the 211-like and 312-like regions are comparable to the ti
c octahedra in ti@xmath3alc and ti@xmath4alc@xmath3 , respectively .
the isotropic mean - squared atomic displacement parameters are highest for the al atoms in all three phases , and the values predicted from first - principles phonon calculations agree well with those measured . |
a lot of attention has been focussed on the interacting electron problem in the last several decades , whereas the interacting boson problem has been considered more often in the framework of specific applications only .
however , there are a number of important situations where the elementary excitations are either intrinsically bosonic in character or else can usefully be viewed in terms of bosonic models .
@xmath0 is an example of the former situation,@xcite while quantum spin systems,@xcite granular superconductors,@xcite and flux lines in type
ii superconductors@xcite are examples of the latter .
therefore it is important to understand in detail the features of model boson systems , in much the same way that one studies the hubbard , anderson , and t - j hamiltonians for correlated fermions . in this paper
we consider a lattice model of interacting bosons , the bose hubbard ( bh ) hamiltonian : @xmath1 here @xmath2 is a boson annihilation operator at site @xmath3 , and @xmath4 .
the transfer integral @xmath5 sets the scale of the energy , and @xmath6 is the chemical potential .
@xmath7 , and @xmath8 are on site , near neighbor , and next near neighbor boson
boson repulsions . the interactions @xmath7 , and @xmath8 promote the formation of `` solid '' order , where the boson occupations fall into regular patterns , at special densities commensurate with the lattice .
the hopping matrix element @xmath9 favors mobile bosons , and consequently a superfluid phase at @xmath10 . in what follows the nature of the correlation functions
will be studied as we change the hamiltonian parameters and the density @xmath11 . when @xmath12 , the bh model maps onto the quantum spin1/2 hamiltonian @xmath13 the field @xmath14 . since @xmath15 , ordering of the density corresponds to finite wave vector ising type order .
similarly , @xmath16 so that superfluidity maps to ferromagnetic ordering in the xy plane . one of the things we shall be interested in in this work is the possibility that density and superfluid order are not mutually exclusive .
indeed , at @xmath17 , the special point @xmath18 corresponds to the heisenberg hamiltonian where ising and xy order coexist .
it has been suggested by various authors@xcite that the addition of further terms like @xmath8 or @xmath19 could stabilize this `` supersolid '' from a special symmetry point to a broader area of the phase diagram .
precisely at the heisenberg antiferromagnet ( af ) , the effect of a field @xmath19 is known : it breaks the full rotational symmetry and selects ordering in the xy plane since the spins can more easily take advantage of the field energy .
this argument has been used to suggest why doping favors the superconducting over the cdw state in the negative@xmath20 hubbard model@xcite where an analogous `` supersolid '' symmetry exists at half filling . while there have been many mean field ( mf ) studies of the spin hamiltonian eq . 2
, there have been to date only a few numerical studies @xcite .
monte carlo studies of interacting quantum boson and spin models provide a useful , exact method to study the nature of the correlations on finite lattices . combined with finite size scaling methods
, they can be used to extract information concerning the thermodynamic limit .
boson simulations are somewhat easier than related path integral methods for interacting electron systems , since they can utilize algorithms which scale linearly with the lattice size and can reach essentially arbitrarily low temperatures .
this paper is organized as follows : in section ii and iii we determine analytically the mf phase diagram , extending past work by considering additional types of order , and describe spin wave calculations of the dispersion relations in the various phases . in section iv we provide numerical results for the soft core model , extending our earlier studies@xcite . in section
v we describe new results for the hard core phase diagram .
conclusions are presented in section vi .
previous work established the mf phase diagram of the spin hamiltonian considering only the possibility of superfluidity and nel type ordering of the density.@xcite at half filling , or equivalently at zero magnetization @xmath21 , for @xmath22 and @xmath23 the spins form a nel state , corresponding to a checkerboard bose _
solid _ with an ordering vector @xmath24 . for @xmath25 a ferromagnetic phase is formed , with a net moment @xmath26 and @xmath27 .
this phase corresponds to a _ superfluid _ , and is also stable for arbitrary @xmath28 and @xmath29 away from half filling . a fully polarized magnetic phase in a strong magnetic field @xmath30 , where only @xmath27 , corresponds to a _ mott
insulator _ with precisely one boson per site . as the solid and the superfluid phases possess different broken symmetries
, one could expect that the transition between them is first order .
however , a rather different scenario has also been put forward , suggesting that the presumably first order transition is split up into two distinct second order transitions , where the two order parameters vanish at separate points@xcite in the regime between the two transitions _ both _ order parameters are non
zero , hence it has been termed a _
supersolid_.@xcite this intriguing possibility is the subject of the investigations reported in this paper .
the mean field analysis indeed finds such a supersolid phase @xcite , although in the hard core limit longer range forces ( @xmath31 ) are needed to stabilize it .
however , recently it was claimed that this conclusion changes in the soft core case , and a supersolid phase exists with nearest neighbor interaction alone@xcite . finally , recent studies on the related heisenberg model with competing first and second neighbor couplings @xmath32 and @xmath33 established the possibility of additional phases : a collinear phase , with alternating lines of up and down spins , at large @xmath34@xcite , and a disordered phase at intermediate values of @xmath34 .
@xcite these differing results clearly call for a reinvestigation of the problem
. the mf phase diagram of the spin hamiltonian eq .
2 worked out by matsuda and tsuneto@xcite , and described above , allowed only for a two sublattice magnetic ordering of the spins corresponding to a nel solid . representing the spins by classical vectors of length @xmath35 we extend earlier mf analyses@xcite for the case of a square lattice to include also the possibility of a collinear phase which is expected to form for intermediate to large next near neighbor repulsion @xmath29 ( see fig . 1 ) .
assuming that the spins are ordered in the xz plane , the mf energies per spin , @xmath36 and @xmath37 , of the nel and collinear spin configurations are given by @xmath38 @xmath39 and @xmath40 are the angles between the spin direction and the @xmath41axis on sublattice @xmath42 and @xmath43 , respectively . @xmath44 and @xmath45 are the corresponding angles in the collinear phase on even and odd rows .
the different phases are identified as follows : @xmath46 we performed the mf analysis in the same spirit as in refs.@xcite .
one proceeds by minimizing @xmath36 and @xmath37 separately with respect to the angles @xmath47 and @xmath48 , respectively .
then the results for fixed magnetic field @xmath30 are translated to fixed magnetisation , i.e. boson density .
finally , we compare the energies of the different phases to obtain the complete mf phase diagram of the spin hamiltonian eq .
2 . explicitly , for two sublattice nel type ordering we find the following phases for @xmath22 and @xmath23 : @xmath49 where @xmath50 is the magnetisation of the system .
for @xmath51 there is no nel order and for @xmath52 the mf ground state is always superfluid .
similarly we analyze the phase diagram following from minimizing @xmath37 for the ordered collinear spin structures corresponding to an ordering wave vector @xmath53 or @xmath54 . at half filling the collinear solid ( see fig .
1 ) is realized for arbitrary values of the near neighbor repulsion @xmath28 .
the reason is that at half filling the energy for the collinear solid is @xmath55 , i.e. independent of @xmath28 due to the cancellation of @xmath56 energies for near neighbor sites on the same and neighboring rows .
away from half filling only the superfluid minimizes @xmath37 for @xmath57 . for @xmath58
a collinear supersolid appears in the phase diagram and the boundary between the superfluid and the collinear supersolid is determined by @xmath59 which is again independent of @xmath28 . for @xmath58 the collinear supersolid phase occurs in a density strip of width @xmath60 around half filling .
given the mf solution for @xmath36 and @xmath37 separately , a comparison for the energies of the different phases allows to map out the complete mean field phase diagram of the spin hamiltonian eq . 2 .
e.g. at half filling , @xmath61 , we have to compare @xmath62 the resulting phase diagram is shown in fig . 2 .
interestingly , for @xmath63 increasing @xmath29 drives two transitions : first increasing @xmath29 frustrates the nel solid and leads to a transition to a superfluid .
increasing @xmath29 further stabilizes collinear order and leads to a transition from a superfluid to a collinear solid.@xcite away from half filling , @xmath64 , no solids , neither nel nor collinear are mf solutions .
instead , transitions occur between the superfluid , and the nel and collinear _ supersolid _ phases .
the boundaries between the different phases are given by @xmath65 finite doping leads to a rigid shift of the phase boundary lines obtained at half filling with the solid replaced by supersolid phases . for @xmath66
this still allows for two transitions with increasing @xmath29 , from a nel supersolid to a superfluid to a collinear supersolid . the @xmath28@xmath29 phase diagram for a fixed magnetisation @xmath67
is shown in fig .
in addition , figs .
4 and 5 show the phase boundaries in the @xmath29 , @xmath68 plane for a fixed value of @xmath28 and in the @xmath28 , @xmath68 plane for a fixed value of @xmath29 , respectively .
recently it was claimed that a finite core repulsion @xmath69 qualitatively changes this picture.@xcite supersolids were found to exist even _ at half filling _ , moreover _ without the next nearest neighbor repulsion @xmath8_. to study these claims we extend the mf analysis by introducing an approximate soft core representation allowing the spin length @xmath35 to be a variational parameter and adding a term @xmath70 to the hamiltonian .
the minimization of the ground state energy is now done separately with respect to @xmath71 and @xmath72 .
we expand the ground state energy around the superfluid phase , and consider the eigenvalues corresponding to small spatial modulations of the density and superfluid order parameter , in effect generating a ginzburg landau type expression . the superfluid collinear supersolid transition is studied by writing @xmath73 and expanding to second order in the ( small ) fluctuations @xmath74 and @xmath75 .
the expectation value @xmath76 of the ground state energy per site takes the form @xmath77\ , , \nonumber\\ e_{sf}&=&-8s^2 + 4(v_1+v_2)m^2+{1\over 8}v_0(4s^2 + 4m^2 - 1)^2 \ , .
\label{eq : eq12}\end{aligned}\ ] ] the ground state energy is the sum of eigenvalues of a matrix in the @xmath78 space . first we solve for @xmath79 at fixed number of particles , i.e. fixed @xmath68 , in the superconducting state where @xmath80 , and obtain @xmath81 a zero eigenvalue of the energy matrix signals the phase transition . the condition for the vanishing of the determinant can be solved for @xmath29 for arbitrary @xmath68 @xmath82 which gives the phase boundary between the superfluid and the collinear supersolid . with the same procedure
the phase boundary between the superfluid and the nel supersolid is at @xmath83 $ ] . as in the hard core case
the phase diagram displays nel and collinear supersolid , and superfluid phases . at half filling the supersolid phases vanish , and two insulating solids are direct neighbors to the superfluid , in contrast with the result of ref .
this result is independent of @xmath84 , _
i.e. _ it is true both in the soft and hard core limits , in agreement with the above hard core mf calculation .
the analyses of the spin wave fluctuations which exist in the literature@xcite are in disagreement .
the spectrum has been found to be either linear@xcite or quadratic@xcite _ at _ the solid supersolid phase boundary .
this dependence is crucial for numerical studies , as it determines the dynamical critical exponent @xmath41 and thereby the appropriate finite size scaling of the lattice . to settle the issue
, we redo the linear spin wave theory analysis for the spin model of eq . 2 and determine the spectrum in the superfluid , the nel solid and the nel supersolid .
again we assume that the spins are ordered in the xz plane with an angle @xmath85 to the @xmath41direction . on each sublattice
the spin quantisation axis is rotated to align the spins along the local direction of the magnetisation by @xmath86{\hat{\bf s}}_{i\in a(j\in b)}\ , .
\label{eq : eq15}\ ] ] to diagonalize the spin hamiltonian eq . 2 in terms of the rotated spins
@xmath87 we introduce spin raising and lowering operators @xmath88 and @xmath89 on sublattice @xmath42 by @xmath90 which obey the usual bosonic commutation relations in the large @xmath35 limit .
similarly , operators @xmath91 and @xmath92 are introduced on sublattice @xmath43 .
after fourier transformation this leads , up to a constant energy shift , to the linear spin wave hamiltonian @xmath93 \label{eq : eq17}\end{aligned}\ ] ] neglecting higher order terms in @xmath89 and @xmath92 . due to the two sublattice structure
the @xmath94sum is restricted to half of the brillouin zone , i.e. to momenta with @xmath95 . in the superfluid and the supersolid phase the @xmath94dependent coefficients in eq .
17 are given by : @xmath96\hskip0.3 cm , \hskip0.3 cm h_{41}= h_{31}+2\gamma_{\bf k}^{(1 ) } \label{eq : eq18}\end{aligned}\ ] ] where @xmath97 and @xmath98 .
the coefficients of the first order terms are required to vanish@xcite which leads to the conditions @xmath99 these two equations determine the angles @xmath39 and @xmath40 for a given value of the magnetic field @xmath30 .
the solutions of eqs .
19 determine the phase diagram of the model .
these equations fully coincide with the ones obtained by minimizing the free energy in the previous section .
we have already used eq .
19 to eliminate the magnetic field in the expressions for the coefficients @xmath100 and @xmath101 in eq .
18 of the spin wave hamiltonian . however , for the nel solid and the mott insulator phase where both @xmath102 and @xmath103 the elimination is not possible and instead @xmath100 and @xmath101 are given by @xmath104 eq .
17 is diagonalized by a generalized bogoliubov transformation using the equation of motion @xmath105_-$ ] with @xmath106 . in the boson language
the bogoliubov transformation involves coupled density and phase modes . as a result
we obtain the spin wave dispersion in the form @xmath107\left[h_{31}+h_{41}\right]+ \left[h_{33}+h_{34}\right]\left[h_{31}-h_{21}\right]\big)\cdot \right .
\nonumber \\ & & \left.\hskip0.6cm\cdot\big(\left[h_{33}-h_{34}\right]\left[h_{31}+h_{41 } \right]+\left[h_{11}+h_{21}\right]\left[h_{31}-h_{21}\right]\big)\big \rgroup^{1/2}\right\rbrace \label{eq : eq21}\end{aligned}\ ] ] typical dispersions are shown in fig . 6 in the different phases and on the phase boundaries .
we now overview the dispersion relations in the four phases . in the nel
solid which is realized for @xmath108 and @xmath109 the spin wave dispersion is given by @xmath110 thus , there are two excitation branches in a halved magnetic brillouin zone
. both branches are gapped . in the superfluid
there is a goldstone mode of linear @xmath111 dependence at small @xmath111 , and a well developed minimum around @xmath112 .
taking the continuum limit carefully identifies this with the roton part of the helium dispersion .
explicitly , with @xmath113 , the dispersion in the extended zone is given by @xmath114\right ) .
\label{eq : eq23}\ ] ] _ iii ) _ in the mott insulator phase for fields @xmath115 all spins are aligned along the magnetic field direction . there is a single gapped mode in the extended @xmath116 brillouin zone with the dispersion given by @xmath117 _ iv ) _ finally , in the supersolid phase one has a gapless linear mode , and a gapped one , again in the halved magnetic zone . to clarify the physics of the transitions we concentrate on the dispersion at @xmath118 and @xmath119 at the phase boundaries . at the supersolid
nel solid transition the critical mode is the goldstone mode at small @xmath111 . at the critical magnetic field , @xmath120 , which determines the nel solid to supersolid boundary by the vanishing of the gap of the lower excitation branch of the solid , we perform the small @xmath111 expansion for @xmath121 from eq . 22 . for @xmath122 where the solid exists at half filling we obtain @xmath123 this means that the linear mode of the supersolid softens into a quadratic one at the boundary signalling the destruction of superfluidity before lifting off into a gapped mode inside the solid phase .
this yields a quantum critical exponent @xmath124 .
this value of @xmath41 agrees with that of chester@xcite and cheng@xcite , but differs from that of liu and fisher@xcite , who obtain @xmath125 .
we feel , however , that the softening of the goldstone mode is a physically realistic picture , supporting our result . at the generic superfluid to nel supersolid transition the critical mode is at @xmath126 . inside the superfluid phase
the roton minimum is at this wavevector .
however in the solid , because of the zone halving , this roton minimum is folded back to @xmath127 . in the superfluid where @xmath128 we study the small @xmath111 expansion of the single mode in the neighborhood of @xmath129 and @xmath130 .
( note that the nel supersolid is only realised for @xmath131 . )
@xmath132 where @xmath133 $ ] . at the boundary to the nel supersolid which is reached at a magnetic field
@xmath134 the mean field conditions in eq .
19 tell that exactly at the transition the roton gap @xmath135 disappears : _ the solidification is signalled by the softening out of the roton mode of the superfluid_. the dispersion relation of the rotons also changes from a quadratic to a linear minimum , hence @xmath125 .
two remarks are in order here .
first , recalling the original landau argument about superfluidity it is clear that a vanishing roton energy leads to a vanishing critical velocity .
so , while the superfluid order parameter remains finite through the supersolid transition , the critical velocity collapses to zero .
_ inside _ the supersolid phase it again assumes a finite value , as the second excitation branch becomes gapped .
second , one can raise the question of how this picture is going to be modified in the absence of an underlying lattice .
in this continuum limit the modes which go soft are located at a finite magnitude of @xmath111 , i.e. on a _ ring _ in momentum space .
this means that the phase space for these excitations is much larger than for the usual goldstone modes , which are centered around @xmath136 .
it then is possible that these excitations may give rise to a fluctuation induced first order transition instead of the second order one taking place on the lattice@xcite .
similar expansions can be used to study the case of half filling . in this particle
hole symmetric case , not surprisingly both transitions have @xmath125 . in a recent monte carlo
study@xcite the same @xmath41 value was used in choosing the lattice size to study both the superfluid supersolid and supersolid solid transitions , whereas we find @xmath125 and @xmath124 , respectively off half filling .
it appears that our results call for the repetition of the numerical simulations with different @xmath41 factors when @xmath137 .
finally at high fields , at the superfluid to mott insulator transition the goldstone mode softens out again , leading to @xmath124 , in agreement with earlier field theoretical predictions@xcite and numerical simulations@xcite .
we have repeated the spin wave calculation for the collinear ordering for an ordering wave vector @xmath138 . in this case
the coefficients of the hamiltonian eq .
17 outside the collinear solid and the mott insulating phases are given by @xmath139\cos(k_y)+{v_2\over 2}\sin\theta_{r1}\sin\theta_{r2}\ , \gamma_{\bf k}^{(2 ) } \nonumber \\ h_{41}=h_{31}+\cos(k_y)\hskip4.3 cm \ , .
\label{eq : eq27}\end{aligned}\ ] ] the mf conditions are read off from the vanishing of the terms linear in the spin wave operators as before @xmath140 for the superfluid and the mott insulating phase the spin wave dispersions are obtained identical to the ones derived above in eq . 23 and eq . 24 . in the collinear
solid with @xmath141 the coefficients @xmath100 and @xmath101 are replaced by @xmath142 the two gapped modes in the collinear solid for magnetic fields @xmath143 and @xmath144 follow as @xmath145 as for the nel supersolid , the collinear supersolid has one gapless linear mode at small @xmath111 and a gapped one in the halved magnetic brillouin zone which in the case of collinear ordering with wave vector @xmath146 is determined by @xmath147 .
the transition from the superfluid to the collinear supersolid is now driven by the softening of the roton mode at @xmath146 . the dynamical exponent is again @xmath125 .
also the exponents at the superfluid to collinear solid and at the solid to supersolid transition are identical to the exponents found for the nel ordering transitions .
* results at half filling * in this section we describe the results of numerical simulations , and compare them with the picture gained from the analytical considerations .
our monte carlo calculations are performed using a path integral representation on the bh partition function by discretizing the inverse temperature @xmath148 into @xmath149 intervals , @xmath150 . a description of the technical details is contained in @xcite .
in order to characterize the phase diagram , we measure the boson winding number to determine the superfluid density @xmath151 .
we also measure the density density correlations @xmath152 and their fourier transform , the structure factor @xmath153 .
@xmath154 our normalization of the structure factor is such that if @xmath152 exhibits long range order , @xmath155 will be proportional to the lattice volume @xmath156 , where @xmath157 is the linear extent in the spatial dimension .
if @xmath152 exhibits only short range order , @xmath155 will be lattice size independent . here
@xmath158 are the possible ordering wavevectors of the solid phase . at weak coupling or high temperatures
, @xmath152 exhibits only short range order . for @xmath159 small , @xmath152 is enhanced but very rapidly decays to its uncorrelated value @xmath160 . however , at low temperatures for sufficiently large interactions , the density density correlations show long range oscillations . the associated structure factor
@xmath153 evolves from being rather featureless to exhibiting a sharp peak at @xmath161 as @xmath162 increases , and a peak at @xmath163 or @xmath54 as @xmath8 increases . for our 2d system , for sufficiently large @xmath162
, we expect a transition in the ising universality class .
that is , @xmath164 is finite . in fact , if @xmath165 we have @xmath166 . but even for a zero temperature phase transition such as would occur at the heisenberg point of the hard core model , one will still observe `` long range order '' at finite @xmath167 when the diverging correlation length exceeds the spatial lattice size as @xmath167 is lowered . in such instances , of course , a careful study of finite size effects is required to draw conclusions concerning the existence of long range order . here
we always report results for temperatures such that @xmath168 so that observables have taken on their ground state values .
we have checked the scaling behavior to be sure that the ground state is genuinely ordered , when so claimed .
7 shows the superfluid density @xmath151 and structure factor @xmath169 as a function of @xmath162 for @xmath170 and @xmath171 .
we see that at @xmath172 there is a phase transition from a superfluid to a solid phase .
the transition on the @xmath173 lattice shown is already rather sharp ; finite size rounding in the raw data for the structure factor and superfluid density near the transition point is further reduced as one goes to @xmath174 lattices .
that one has true diagonal long range order in the solid phase is confirmed by the fact that the structure factor scales linearly with the lattice volume . indeed , at @xmath175 , @xmath176 is almost precisely @xmath177 times as large on the @xmath174 lattice than the @xmath178 .
there does not appear to be any window of coexistence between the superfluid and solid phases at half filling . to within limits set by rounding
, the transition points for @xmath35 and @xmath151 coincide almost precisely .
we can make this statement more quantitative by performing the appropriate scaling analysis on the data .
for example , we have plotted @xmath179 and @xmath180 versus @xmath162 for different values of the exponent ratios @xmath181 .
curves for different lattice sizes should cross at the same critical value of @xmath162 for the appropriate choices of @xmath181 .
a complication is that the imaginary time lattice size must be scaled as the appropriate power of the spatial extent , and the dynamic exponent @xmath41 could be different for the two transitions . making the simplest assumption that @xmath41 is the same , however , as was already suggested by the raw data
, this scaling procedure shows that the transition points for the two observables are within @xmath182 of each other . while the structure factors do indeed cross nicely , the superfluid density curves come together rather than pass through each other .
this seems to be a rather generic feature of simulations of the bose hubbard model@xcite as opposed to related conserved current models @xcite . as @xmath8
is increased , @xmath152 shows a similar transition from featureless uncorrelated behavior to long range order , although in this case @xmath8 favors the formation of a `` striped '' collinear phase with alternating lines of occupied and empty sites .
the structure factor @xmath153 develops a peak at @xmath183 or @xmath184 . in order to determine whether @xmath8 can drive a supersolid phase at half filling
, we turn on @xmath8 close to the point where the transition between superfluid and solid occurs in fig . 7 . the density @xmath185 .
we show in fig .
8 a plot of @xmath151 and @xmath153 for @xmath186 and @xmath187 .
we see that @xmath8 drives the nel solid into a superfluid , and then at yet larger values causes the formation of a striped solid phase .
again , the plots suggest that there is no supersolid phase at @xmath185 .
scaling plots similar to those constructed at @xmath171 do not reveal any evidence for distinct critical points for superfluid and solid transitions to within our numerical accuracy .
we can put data from figs . 7 and 8 together with similar runs for different sweeps of @xmath162 and @xmath8 to obtain the ground state phase diagram of the soft core bh model at @xmath170 and @xmath185 .
this is shown in fig .
9 . at weak couplings we have a superfluid phase , while at strong couplings there are two possible solids : checkerboard and striped .
a strong coupling analysis predicts a phase boundary between the solid phases at @xmath188 the superfluid phase extends out along this line in a very robust manner , as opposed to the situation in 1d , where the superfluid window was rather narrow.@xcite this is a consequence of the highly degenerate nature of the strong coupling ( @xmath165 ) ground state along the line @xmath189 . as can easily be seen , not only do the nel and checkerboard solids have the same energy , but an infinite number of defect states are degenerate as well for @xmath190 .
for example in a horizontally aligned collinear solid a whole column can be shifted up and down without energy cost .
@xcite this large degeneracy stabilizes superfluidity , even at large coupling .
we will comment further on this point when discussing the hard - core phase diagram .
* results off half filling * although it does not appear that the bh model exhibits a supersolid phase at @xmath185 , we can see the coexistence of diagonal and off diagonal long range order when the filling is shifted away from @xmath185 . in fig .
10 we show @xmath151 and @xmath176 for the same parameters as fig .
7 except now @xmath191 .
we see that although @xmath151 declines significantly when the solid forms , the excess boson density @xmath192 ( the magnetization @xmath68 is spin language ) remains mobile in the solid background .
indeed , simulations at different densities ( we found supersolids out to dopings of 0.675 ) show that the tail in @xmath151 is precisely proportional to @xmath75 .
11 shows the analogous plot for a striped supersolid .
note that we have here separately displayed @xmath193 and @xmath194 .
as expected , the superfluid density in the @xmath195 and @xmath196 direction is correlated with the direction in which the striped solid channels run , as determined by the ordering wavevector @xmath183 or @xmath184 .
if we had separately measured @xmath197 and @xmath198 on the two sublattices of the checkerboard solid , we would have found an analogous symmetry breaking .
the nonzero value of @xmath199 is closely related to the appearance@xcite of a nonzero order parameter @xmath200 in the language of the spin hamiltonian eq . 2 .
if we were to use finite size scaling techniques to locate the precise phase boundaries , it would be necessary to scale the imaginary time length @xmath149 as a power of the spatial length @xmath201 , where @xmath41 is the dynamic critical exponent .
as we have earlier described , it may be that different values of @xmath41 are associated with the two transitions off half filling , in which case the finite size scaling analysis is much more delicate.@xcite we do not see the necessity of such a study here , since the supersolid phase occupies an extended portion of the phase diagram , and its existence is not predicated on proving the distinctness of two transition points .
_ figures 10 and 11 provide compelling evidence for the existence of a supersolid phase .
our physical picture of this supersolid is one in which @xmath185 of the bosons freeze into a rigid solid structure , while the remaining @xmath75 remain mobile .
as we have seen , a signal of long range order then is present in both the diagonal @xmath202 and off diagonal @xmath203 channels .
_ we have conducted our simulations of the bh hamiltonian in the canonical ensemble , and have presented our results by specifying the density @xmath204 rather than the chemical potential @xmath6 . in describing the nature of the phase diagram it is important to note that due to the existence of a gap in the solid phases , the @xmath6@xmath204 relation is non trivial .
if the gap is nonzero , when we dope our system even slightly away from half filling , the chemical potential is shifted by a considerable amount . in the language of the spin hamiltonian , eq . 2 , a sizeable field @xmath19 is required to change the magnetization of the gapped ising phase . in fig .
12 we illustrate this point by drawing the @xmath205@xmath206 phase diagram .
a sweep at constant chemical potential reveals a supersolid _
window_. a sweep at fixed density skirts the pure solid and remains in the supersolid phase .
this is why we see in figs .
10 and 11 a supersolid for an extended region @xmath207 rather than in some narrow region between phases exhibiting a single type of order . if we now examine densities @xmath208 , we find qualitatively similar results : a superfluid phase gives way to a striped supersolid phase as @xmath162 increases . by these measures , hole or
particle doping appears qualitatively similar .
the same is true of the checkerboard supersolid , where results for hole doping are entirely reminiscent of the analogous particle doped case .
in fact , however , something rather different does go on with particle and hole doping . in fig .
13 we show the ground state energy as a function of doping for @xmath170 , @xmath209 , and @xmath210 . for these parameters
, as we have seen , we have a striped supersolid off half filling and a striped solid at @xmath211 . the change in slope of @xmath212 at @xmath211 reflects a jump in the chemical potential which is , in fact , just the gap in the solid phase.@xcite .
there is nothing particularly unusual here .
the strange feature occurs for the checkerboard case . in fig .
14 we show the ground state energy as a function of doping for @xmath170 , @xmath209 , and @xmath171 .
the fact that @xmath213 is concave down for @xmath214 indicates an instability to phase separation .
previous studies@xcite have suggested the possibility of phase separation in systems with attractive boson interactions .
however , we do not have these lennard jones type potentials here , only purely repulsive ones .
it is not immediately apparent why mobile holes ( or particles ) in a rigid solid background should segregate themselves . a possible explanation , however , is as follows : consider an isolated doped hole in a checkerboard solid . in order to move to another site of the same sublattice , it must pass through an intermediate site on the opposite sublattice , a state of energy @xmath215 .
thus the hole s effective hybridization is @xmath216 .
( this sort of argument has previously been used to predict the shape of the phase boundary in the one dimensional extended bh hamiltonian , in good agreement with simulations.@xcite ) if two holes are near each other , the intermediate state is lower in energy , so the effective hybridization is increased .
this suggests a possible mechanism for phase separation : increased mobility of holes which propagate coherently .
of course , the increase in @xmath217 is partially offset by the entropy cost of confining one hole near the other .
unfortunately , there appears to be an analogous increase in @xmath217 for doped particles which are proximate , so this reasoning does not explain the fact that @xmath218 is concave down for @xmath208 only .
nevertheless , the simulations provide compelling evidence for a lack of particle - hole symmetry . in principle , one can also examine the issue of phase separation through anomalies in @xmath153 for small @xmath94 .
however , our use of the canonical ensemble makes this approach non - trivial .
further work on the question of phase separation is needed .
we now examine the phase diagram in the hard core case .
this is important to do for a number of reasons .
first , it allows us to make a connection to the spin model limit , eq . 2 .
second , as we have seen at @xmath170 , some of the interesting transitions occur at @xmath162 and @xmath8 values which are getting rather large , while we expect in most physical situations that the on site @xmath84 should be substantially greater than the near neighbor interactions .
one consequence of this , is that the doped bosons in the supersolid phase for our soft core model could move on the occupied sublattice , since the cost of @xmath84 was less than the coordination number @xmath41 times the near neighbor interaction strengths . in the hard core model such multiple occupancies are forbidden , and we want to make sure that our conclusions are not affected by this change . fig .
15 shows results for the superfluid density and structure factors for the half filled case .
we sweep @xmath8 at fixed @xmath219 .
a nel phase appears at small @xmath8 . for larger @xmath8
the superfluid phase appears before making a transition into a collinear solid for yet larger @xmath8 .
if @xmath162 is sufficiently small , the nel phase at weak @xmath8 is eliminated , and the system remains superfluid down to @xmath171 .
data for this and other sweeps is summarized in fig .
16 where the resulting ground state phase diagram is shown .
note that we find the superfluid
nel solid transition at @xmath171 occurs at a value @xmath162 close to @xmath220 , which is the result expected based on the mapping to the spin model , eq . 2 .
as in the soft core case the weak coupling superfluid extends out along the @xmath221 strong coupling boundary between the two solid phases . unlike analogous studies@xcite in 1d , this superfluid wedge is difficult to close , a phenomenon which we earlier explained by the large degeneracy of competing solid phases along the strong coupling line . we have conducted simulations along the line @xmath221 and find that the superfluid density vanishes at @xmath222 .
interestingly , there is no inset of solid order at this point .
this needs further study , for example to understand if some disordered dimer phase might exist in this regime , in analogy with related spin systems .
17 is a plot for a doped lattice with @xmath223 .
the main difference is that , as in the soft core case , there is a superfluid tail after the structure factor exhibits the transition into the solid phase .
that is , there is a supersolid in the hard
core case as well .
as expected , doping inhibits somewhat the formation of crystalline order , so that stronger couplings are required to induce the crystalline order as is seen by comparing the doped phase diagram , fig .
18 , with fig .
16 , the phase diagram at half filling .
despite considerable rounding of the transitions , scaling analyses conclude that the regions where @xmath35 is large are indeed ordered .
finally , figs .
19a , b show the ground state energy as a function of filling for the hard core model at @xmath224
( nel solid ) , and @xmath225 ( collinear solid ) , respectively .
the data are qualitatively similar to the soft core model . in the collinear solid @xmath212
is concave up , with a change in slope at @xmath211 which is the gap . in the nel solid @xmath212
shows a tendency for phase separation .
up to now we have focussed on the ground state phase diagram of the bh hamiltonian .
it is interesting to consider also the behavior of the system at finite temperatures . here
the motion of doped bosons in the bh model which we have studied with our simulations has a close connection with the idea of `` defectons '' in a solid@xcite where quantum tunneling caused by the finiteness of the de boer parameter delocalizes lattice defects at low temperature .
it is also of interest to study the behavior of the diffusion constant @xmath226 for the full range of temperature .
here we expect that defects are localized at high @xmath167 , and @xmath226 first decreases exponentially as @xmath167 is lowered in this classical regime .
@xmath226 should then exhibit a plateau as quantum diffusion takes over , and ultimately increase again as delocalization occurs . while they focus largely on the behavior of single defectons , andreev and lifshitz also consider the possibility of long range coulomb interactions causing localization into a `` defecton superlattice '' .
our insulating checkerboard solid is in fact an illustration of this .
the bose hubbard model with only on site @xmath84 , has no solid phase at @xmath185 , but when @xmath162 is turned on , an ordered lattice does form .
we have focussed here on zero temperature , the finite temperature phase diagram of the 2d bh model would be interesting to study as well .
the solid transitions are in the ising universality class , and hence have a finite @xmath164 .
similarly one expects a kosterlitz thouless type finite critical temperature for the superfluid transition . as for the topology of the phase diagram ,
several possibilities have been explored by liu and fisher@xcite .
one intriguing case is the appearence of a tetracritical point ; where the three ordered phases ( superfluid , supersolid and solid ) come together , giving way to the disordered phase with further increase of the temperature .
this happens within a limited , but finite range of parameters on the mean field level .
the corresponding scaling theory was developed by nelson and fisher @xcite .
other alternatives include a supersolid phase which exists only at finite temperatures , and that the tetracritical point is split into bicritical points.@xcite we hope to take up some of the issues in a further publication . in the path integral representation of the bh partition function used in our simulations , the particle number conservation leads to boson `` world lines '' propagating in the original 2 spatial dimensions plus an additional imaginary time direction which runs from 0 to @xmath148 .
this picture has been used to suggest close analogies between the physics of vortices in type ii superconductors@xcite and the phase diagram of the 2d bose
hubbard hamiltonian .
frey , nelson , and fisher,@xcite have recently discussed both thermally driven and quantum phase transitions , for example as caused by the introduction of defects or interstitials into the abrikosov lattice , in these vortex systems .
this also has close connections with the results we have discussed here .
in this paper we have considered quantum phase transitions in the bose hubbard hamiltonian .
we identified several phases : solid and supersolid phases with nel and collinear patterns , furthermore a superfluid and a mott type insulating phase .
the phase diagram has been determined analytically and the spin wave spectrum has been calculated .
the dynamical critical exponents at each transitions were calculated and preexisting controversies were settled .
our numerical work utilizing quantum monte carlo methods provided a detailed study of the different phases . concerning the phase diagram the existence of supersolid phases
has been forcefully confirmed .
these phases exist only off half filling , in accordance with the mean field results , but in disagreement with some recent claims .
the possibility of phase separation in the model has been investigated as well , and provides evidence for a violation of the previously assumed particle hole symmetry of the model .
we acknowledge useful discussions with d. arovas .
this work was supported by the national science foundation grant dmr 92 - 06023 , and by los alamos national laboratory under lacor grant no .
uc-94 - 2-a-213 .
apk gratefully acknowledges support through a habilitation scholarship of the deutsche forschungsgemeinschaft ( dfg ) .
gtz acknowledges the kind hospitality at the university of karlsruhe , where his work was supported by the sonderforschungsbereich 195 of the dfg .
much of the numerical work was performed on a connection machine 5 at thinking machines corporation .
an analysis of the related problem of possible ground states of the 2d ising model with competing interactions is contained in u. brandt , _
z. phys . _ * b53 * , 283 ( 1983 ) .
we thank j. freericks for bringing this reference to our attention . * fig .
3 : * hard core mean field phase diagram ( bold lines ) away from half filling for @xmath228 from comparing the energies of superfluid , nel and collinear supersolid .
thin lines indicate the phase boundaries at half filling @xmath61 ( see fig .
2 ) . *
fig . 6 : * spin wave dispersions in the ( a ) nel solid , ( b ) at the nel solid
nel supersolid boundary , ( c ) in the nel supersolid , ( d ) at the nel supersolid superfluid boundary , and ( e ) in the superfluid . in all plots @xmath29 and @xmath28 are fixed to @xmath231 , @xmath232 and the magnetic field @xmath233 ( @xmath30 in the text ) is varied .
* fig . 7 : * the superfluid density @xmath151 and @xmath176 as a function of @xmath162 for @xmath170 and @xmath171 . the density @xmath185 and @xmath234 .
the transitions in @xmath151 and @xmath176 appear to occur at roughly the same value of @xmath162 . *
12 : * the qualitative @xmath10 phase diagram of the bose hubbard model is illustrated .
a sweep at constant @xmath6 ( full arrow ) could cut across the phase boundaries as shown , revealing a supersolid window , while a sweep of constant density ( dashed arrow ) remains in the supersolid phase at strong coupling .
the mott insulating phase at large @xmath6 has density one boson per site . | the bose - hubbard model exhibits a rich phase diagram consisting both of insulating regimes where diagonal long range ( solid ) order dominates as well as conducting regimes where off diagonal long range order ( superfluidity ) is present . in this paper
we describe the results of quantum monte carlo calculations of the phase diagram , both for the hard and soft core cases , with a particular focus on the possibility of simultaneous superfluid and solid order .
we also discuss the appearance of phase separation in the model .
the simulations are compared with analytic calculations of the phase diagram and spin wave dispersion . |
paper considers the task of accurate image retrieval on a large scale . given a query image , we aim at finding the similar images from the database . a number of retrieval models have been proposed in this scenario , among which local feature based model , _
i.e. _ , bag - of - words ( bow ) @xcite , has obtained widespread applications .
nevertheless , traditional bow is lack of spatial layout and may encounter quantization errors . besides , local feature
based models usually lead to huge storage consumption , which is limited in real - world applications featured by the rapid expansion of the image volume . on the contrary ,
holistic representations such as hsv represent an image via global vectors and obtain good scalability , while the retrieval accuracy is less desirable . generally speaking ,
local descriptors and holistic representations demonstrate distinct strengths in finding similar images .
local features are capable of capturing local image patterns or textures , while holistic representations delineate overall feature distributions in images .
thus , a variety of methods @xcite are proposed to integrate the strengths of local and holistic features to yield more satisfying retrieval results .
@xcite propose a graph based fusion approach to merge and rerank multiple retrieval sets .
although this method achieves desirable accuracy , there are some disadvantages , _ e.g. _ , being not robust to the dynamic changes of the dataset or being unable to find reciprocal neighborhoods .
zheng _ et al .
_ @xcite propose a coupled multi - index framework to perform feature fusion at indexing level .
specifically , color and sift features are coupled into a multi - dimensional inverted index to filter out false positive sift matches .
the _ c - mi _ seems to achieve higher accuracy accompanied with less query time , while the storage consumption is still remarkable when the image dataset rapidly expands .
another representative work is the query - adaptive late fusion method proposed in @xcite . to a certain query ,
@xcite extracts five different image descriptors and each of them is used to search the whole dataset independently .
the initial score curves obtained via these representations are normalized by the references and later used to evaluate the effectiveness of the corresponding descriptors .
this query - adaptive method does have competitive retrieval performance , while it heavily relies on the number of to - be - fused features and needs considerable amount of storage to save these features , as well as the reference curves .
different from the above fusion methods utilizing multiple image descriptors to search the entire database simultaneously , c2f proposed in this paper is a two - layer fusion procedure working from coarse to fine .
the first stage aims at filtering out distractors from the dataset , which helps enhance the retrieval efficiency and reduce the memory consumption to be encountered in the next step .
the subsequent operation focuses on refining the rank list from the filtered candidates , which further improves the retrieval accuracy ( see figure [ fig : example ] for an illustration ) . to be more specific ,
we firstly exploit the holistic descriptors to filter out distractor images that are dissimilar to the query from the database , and thus the number of candidate images used to compared with the query can be greatly reduced .
this process is efficient and the retrieved candidates usually appear globally similar .
then , according to the similarity scores obtained via holistic features , we acquire @xmath0 candidate images which are more similar to the query when compared with other images in the database .
moreover , for each candidate image , we design an adaptive weight by holistic similarity scores to measure its similarity degree to the query . finally , accurate image search is conducted on candidate image pool via local features . in this step ,
adaptive weights of the candidate images are used in refining the initial retrieval list . the main contribution of the proposed framework consists in the coarse - to - fine fusion of retrieval sets given by different methods , which has two merits : 1 ) by filtering out distractors from the database , the number of candidate images to be used for comparison with the query can be greatly reduced .
2 ) given a certain query , the relative weight of each candidate can be automatically evaluated via global similarity scores , which helps ensure the stability of the retrieval performance .
we validate the performance of c2f on two public datasets , _
i.e. _ , ukbench and holidays datasets .
the evaluation shows that our method compares favorably with the recent state of the art .
the remainder of the paper is organized as follows : section 2 highlights the related works ; we describe the formulation details of c2f in section 3 ; section 4 provides comprehensive experimental results to validate the superiority of c2f ; finally , section 5 concludes the paper and states directions for future work .
a variety of works have been exposed to improve the image retrieval performance . in this section ,
we briefly introduce several closely related methods . to obtain a discriminative image representation with local features ,
bow based methods are usually adopted . in bow model ,
a codebook is generated off - line by unsupervised clustering algorithms , such as k - means , hierarchical k - means @xcite , and approximate k - means @xcite .
each local feature is quantized to one or a few visual words by approximate nearest neighbor algorithms
. then the image is represented by a high - dimensional histogram over the visual codebook .
this process is usually accompanied by the quantization error @xcite . to correct quantization defects ,
hard assignment can be replaced with schemes such as soft assignment @xcite , multiple assignment @xcite , sparse coding @xcite , _
etc_. nevertheless , quantizing a single feature to several visual words @xcite will introduce more storage burden and higher search complexity .
representatively , @xcite proposes to generate a @xmath1-bit binary signature to each local feature with the hashing approach via a smaller codebook , for example 20,000 .
the true matches are defined as those local features that are not only quantized to the same visual word but also have small hamming distance between their binary signatures . on the other hand , @xcite proposes a compact feature based clustering ( cfc ) to represent images .
@xcite improves the retrieval performance by automatically learning robust visual features and hash functions .
@xcite proposes a discriminative light unsupervised learning network to learn a compact image representation .
@xcite employs a two - layer hierarchical scheme to extract the global information and statistics of the local feature set of image datasets .
there is no doubt that methods based on compact holistic features are efficient in computation and memory usage , as well as more suitable to large - scale dataset .
since holistic features tend to be less invariant than local features and more sensitive to image transformations , their retrieval precision is often lower compared to local feature based methods .
as a consequence , it would be desirable if one method can effectively combine the complementary cues of local and holistic features .
rank aggregation @xcite is a solution to fuse local and holistic cues at the rank level , but the effectiveness of rank aggregation is discounted when there is no intersection among the top candidates retrieved by the local and holistic feature based methods , which does occasionally occur . to overcome this weakness , @xcite proposes an undirected - graph based fusion approach to further enhance the retrieval precision . built on this undirected - graph , @xcite proposes a directed graph which is robust to outlier distraction .
@xcite propose a graph - based optimization framework to leverage category independent object proposals for logo search in a large scale image database .
@xcite proposes a semantic - aware co - indexing algorithm to jointly embed local invariant features and semantic attributes into the inverted indexes .
@xcite introduces a simple color signature generation procedure and embeds local color features into the inverted index to provide color information by bag of colors . in @xcite , a couple multi - index framework
is proposed to perform feature fusion at indexing level by couple complementary features into a multi - dimensional inverted index .
@xcite proposes a late fusion at score level . particularly , the effectiveness of each to - be - fused feature is estimated in an unsupervised , query - adaptive manner .
@xcite proposes a bag - of - words based deep neural network for image retrieval task , which learns high - level image representation and maps images into bag - of - words space . from the above analysis
, it can be seen that considerable efforts have been made to improve the retrieval accuracy via fusion schemes , such as graph - based @xcite , query - adaptive late fusion @xcite .
usually , with more heterogeneous features , higher accuracy can be obtained , while more information needs to be stored and higher computational complexity incurred , correspondingly .
therefore , how to maintain a good balance between accuracy and complexity is an important issue when focusing on large - scale data .
different from the above fusion methods , in this work , we propose a two - layer fusion framework to rank database images from coarse to fine ( c2f ) .
this method improves retrieval efficiency and accuracy reliably .
in this section , we present the detailed structure of the proposed c2f aiming at accurate image retrieval on large - scale databases .
c2f is a two - layer fusion framework which consists of three key components .
1 ) it filters distractors from the whole database with holistic representations , and thus the number of images to be used for comparison with the query can be greatly reduced .
2 ) for a certain query , we design an adaptive weight for each selected candidate image according to the similarity scores obtained via holistic features .
3 ) we further refine the candidate set with local features , considering the impact of adaptive weights .
the mechanism of c2f is presented in figure [ fig : framework ] . in the next subsections
, we will elaborate each component of the framework .
we assume that there are a total of @xmath2 color images in the database @xmath3 . using holistic features , such as hsv , each database image
is represented as a @xmath4-dimensional histogram vector @xmath5 , and the image set can be defined as @xmath6 . in order to reduce the impact of bins with large values , hsv histograms of both query and dataset images are l1-normalized and square scaled , similar to the rootsift @xcite : @xmath7 where @xmath8 is a coefficient and its value is set to 0.5 .
to select candidate images from the dataset , we adopt the cosine distance between vectors of the query and the database images as similarity measurement : @xmath9 where @xmath10 is the query , @xmath11 is a database image , and @xmath12 represents the similarity score obtained by global feature .
the cosine similarity between the query and all the dataset images can be defined as : @xmath13 particularly , a larger cosine score corresponds to higher similarity with the query .
note that there exists large variationa among the database image .
thus , we hold the opinion that most database images have significant differences from the query . hence , the holistic image representation can be used to quickly filter out these noise images . for each query ,
@xmath0 candidates that share similar information with the query are adaptively selected from the original database .
different queries commonly have different candidate images .
after the filtering stage , @xmath0 candidate images more similar to the query are selected from the original image database .
the similarity of the query with different candidates varies extensively .
it is important to design an adaptive weight for each candidate image to evaluate its similarity level with the query . considering that holistic representations , _
i.e. _ , hsv histograms , delineate overall color feature distributions in images , the cosine values computed in sec .
[ method : choose ] are used as the basis of learning weights . in particular , the similarity scores of the query with @xmath0 candidates are represented as @xmath14 .
the motivation behind the weighting method is that the larger cosine values correspond to higher importance of the candidate images . to overcome the dictatorship incurred by the candidates with larger scores , the cosine similarity values of the corresponding candidate images are min - max normalized to @xmath15 $ ] : @xmath16
then , the normalized scores are used as the adaptive weights of the corresponding @xmath0 candidate images .
@xmath17 experimental results show that adaptive weights significantly improves the accuracy of image retrieval on both holidays and ukbench datasets .
the performance of c2f with and without wights are compared in table [ fig : weights ] .
given the @xmath0 candidate images and their corresponding weights of a certain query , we focus on utilizing local features to find out ground truths .
firstly , we extract sift descriptors from an image by hessian - affine detector @xmath18 \in r^{d\times m}$ ] .
the image contains @xmath19 local descriptors in @xmath20 dimensions .
then , a codebook trained on filckr60k dataset by approximate k - means ( akm ) @xcite clustering algorithm is used to quantize sift features . specifically , each local feature @xmath21 of an image is assigned to its nearest visual word of the codebook , represented by the i d of the corresponding visual word .
@xmath22 next , to enhance the discriminative validity of the visual words , a @xmath23-bit binary signatures feature is generated for each sift feature @xmath24 $ ] , which encodes the location of the sift descriptor within the voronoi cell .
the distance between two descriptors @xmath25 and @xmath26 lying in the same cell is approximated by the hamming distance of their corresponding binary signatures .
@xmath27 at this point , a sift descriptor is represented by @xmath28 and @xmath29 , and a standard inverted index is generated .
then the matching function is defined as : @xmath30 where @xmath31 is the hamming distance defined in e.q .
[ gongshi10 ] , and @xmath32 is a fixed hamming threshold .
the similarity score between the query and the candidate image is : @xmath33 where @xmath19 is the number of local feature in the query image @xmath10 , @xmath25 and @xmath26 is the local feature of the query and dataset image , respectively , l represents that the score is computed via local features .
considering the similarity degrees measured by holistic representations , the sorted list of the @xmath0 candidate images obtained via local features are refined according to the adaptive weights learned in sec .
[ method : weight ] .
the final similarity scores of the query and the candidate images are defined as : @xmath34 where @xmath35 is the final similarity score between the query and the dataset image d , @xmath36 is the weight obtained by the intuitively or steadily weighted method .
overall , we propose a two - layer fusion framework which takes advantage of global and local cues and ranks database images from coarse to fine . considering the efficiency and low storage consumption of holistic features
, we use the hsv histograms to coarsely search the original database and filter out noise images .
then , accurate image search is conducted on candidate images , combining the adaptive weights learned via global similarity scores to further enhance the retrieval accuracy .
experimental results in sec .
[ exper : state - of - the - arts ] demonstrate that the retrieval accuracy of c2f is competitive to state of the art . in order to illustrate the following analysis
more clearly , we provide a brief description of these parameters : @xmath37 is the number of dataset images , @xmath38 is the number of query images , @xmath39 is the number of candidate images , and @xmath40 is the number of features to - be - fused .
the rising processing time of the image retrieval mainly caused by the increasing number of candidate images , which are used for comparison with the query . in this paper
, we propose a two - layer fusion method which takes advantage of global and local cues and ranks database images from coarse to fine . the total comparison number in c2f
is @xmath41 , which is a relative small value compared with other fusion methods @xcite .
for example , the comparison number in @xcite is @xmath42 , especially , @xmath40 is set to 5 in the experiment . correspondingly , the shorten of the comparisons in c2f will certainly contribution to the efficiency and the storage consumption . the memory storage consumed by c2f
is quite smaller than that of @xcite .
this is because @xcite need to save multiple features of all images of the database , while c2f stores holistic features of all database images and local features of only @xmath0 candidate images . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] to further evaluate the performance of c2f on the large - scale image datasets , we merge holidays with imagenet 1@xmath19 . in the experiments , we use different numbers of distractors to test the scalability of the proposed two - layer fusion framework .
specifically , we set the number of candidates @xmath0 to 1000 , and adopt adaptive weights to enhance the retrieval accuracy .
table [ table : large - scale ] shows the retrieval results with different numbers of distractor images .
undoubtedly , the performance of the performance of c2f gradually decreases with the augment of distractors in the database .
when the number of distractor images is 0 , c2f obtain best accuracy . while when all 1 m distractors of imagenet are added into the holidays database , the map of c2f reduce significantly , which is 12.63% lower than that of without distractors .
moreover , we do not enhance the corresponing candidates when the distractors increase , so c2f can significantly enhance the efficiency and reduce the memory consumption at the cost of some accuracy loss .
it is well known that bow incorporated with hamming embedding can achieve desirable performance . meanwhile
, a major bottleneck in bow based methods is the length of the inverted list for it grows almost linearly with the database size .
a traverse on the inverted index generated by the whole database can be extravagantly expensive .
this problem appears more severe if we consider a typical case where the benchmark contains millions of images , for the inverted list can be as long as gbs .
it means that for each query feature , a naive full comparision with all images of the dataset takes considerable time . table [ table : memory ] and table [
table : efficiency ] give the average query memory consumption and the corresponding time on holidays and ukbench datasets with the increasing number of candidates , respectively . with the same candidate number ,
memory expended on ukbench is far less than that on holidays due to the different size of images , while incremental tendency reflected on two benchmarks is consistent .
similarly , from table [ table : memory ] and table [ table : efficiency ] , we observe that the average query time also significantly increases with the growing number of candidates .
this is because that more candidates means larger inverted index , namely , more search time to one query .
thus , an effective filter mechanism is very necessary for large - scale image retrieval , and a moderate @xmath0 candidate set is important to guarantee the tradeoff between the retrieval accuracy and memory consumption .
figure [ fig : failure ] presents some false retrieval results through c2f on holidays dataset .
as shown in figure [ fig : failure ] , these inaccurate images have the following two characteristics : 1 ) the image itself contains ambiguous information . taking the query in the upper left corner as an example , we want the returned top images contain the waterfall , while in fact , the query involves abundant greenery information .
this is why neither global representation nor local feature can acquire satisfactory results .
2 ) in the experiments , c2f utilizes hsv histograms delineating overall color feature distributions in images to filter out distractors , so the top ranked candidates share similar color composition .
this similar error will affect the query accuracy in the next stage , see the queries on the upper right corner and the lower left corner of the figure [ fig : failure ] .
in this paper , a two - layer fusion method is proposed , which takes advantage of global and local cues and ranks database images from coarse to fine .
the main purpose of c2f is to reduce memory consumption and the computational complexity , without compromising the retrieval accuracy . to achieve this goal
, c2f adopts holistic representation to filter out noisy images of the benchmark and choose the images with high similarity scores as candidates .
particularly , for each candidate , an adaptive weight is learned via the holistic similarity scores .
then retrieval is conducted on candidate set by taking the adaptive weights into account .
comprehensive experiments are conducted to evaluate the accuracy and scalability of the c2f . with the same holistic and local descriptors ,
the accuracy of c2f on holidays is 3.31% higher than the query - adaptive fusion method @xcite . in future work
, we will further explore more effective holistic representations and design more adaptive weighted functions . | this paper addresses the problem of large - scale image retrieval .
we propose a two - layer fusion method which takes advantage of global and local cues and ranks database images from coarse to fine ( c2f ) .
departing from the previous methods fusing multiple image descriptors simultaneously , c2f is featured by a layered procedure composed by filtering and refining . in particular , c2f consists of three components .
1 ) distractor filtering . with holistic representations ,
noise images are filtered out from the database , so the number of candidate images to be used for comparison with the query can be greatly reduced .
2 ) adaptive weighting . for a certain query
, the similarity of candidate images can be estimated by holistic similarity scores in complementary to the local ones .
3 ) candidate refining .
accurate retrieval is conducted via local features , combining the pre - computed adaptive weights .
experiments are presented on two benchmarks , _
i.e. , _ holidays and ukbench datasets .
we show that our method outperforms recent fusion methods in terms of storage consumption and computation complexity , and that the accuracy is competitive to the state - of - the - arts .
shell : bare demo of ieeetran.cls for ieee journals image retrieval , coarse - to - fine , holistic representation , local feature . |
let @xmath0 be a simple graph of order @xmath1 . for any vertex @xmath14 , the open neighborhood of @xmath15 is the set @xmath16 and the closed neighborhood is the set @xmath17=n ( v ) \cup \{v\}$ ] . for a set @xmath18 ,
the open neighborhood of @xmath19 is the set @xmath20 and the closed neighborhood of @xmath19 is the set @xmath21=n ( s ) \cup s$ ] .
a set @xmath22 is a total dominating set if every vertex of @xmath3 is adjacent to some vertices of @xmath2 , or equivalently , @xmath23 .
the total dominating number @xmath5 is the minimum cardinality of a total dominating set in @xmath4 .
a total dominating set with cardinality @xmath5 is called a @xmath24-set .
an @xmath8-subset of @xmath3 is a subset of @xmath3 of cardinality @xmath8 .
let @xmath25 be the family of total dominating sets of @xmath4 which are @xmath8-subsets and let @xmath26 .
the polynomial @xmath27 is defined as total domination polynomial of @xmath4 .
graph @xmath4 obtained by point - attaching from @xmath28.,width=264 ] let @xmath4 be a connected graph constructed from pairwise disjoint connected graphs @xmath9 as follows . select a vertex of @xmath10 , a vertex of @xmath11 , and identify these two vertices . then
continue in this manner inductively .
note that the graph @xmath4 constructed in this way has a tree - like structure , the @xmath13 s being its building stones ( see figure [ figure1 ] ) .
usually say that @xmath4 is obtained by point - attaching from @xmath28 and that @xmath13 s are the primary subgraphs of @xmath4 .
a particular case of this construction is the decomposition of a connected graph into blocks ( see @xcite ) . as an example , the @xmath1-barbell graph @xmath29 with @xmath30 vertices , is formed by joining two copies of a complete graph @xmath31 by a single edge ( figure [ bar_n ] ) . actually , this graph is a specific kind of point - attaching of two complete graphs @xmath31 and the graph @xmath32 .
observe that the total domination polynomial of @xmath1-barbell graph is @xmath33 this formula obtain easily from counting the total dominating sets of @xmath29 .
and @xmath34 , respectively.,width=453,height=75 ] calculating the total domination polynomial of a graph @xmath4 is difficult in general , as the smallest power of a non - zero term is the domination number @xmath5 of the graph , and determining whether @xmath35 is known to be np - complete . so presenting a closed formula for the total domination polynomial of any kind of point - attaching graphs is difficult , but for certain classes of graphs
, we can find a closed form expression for the total domination polynomial . in this paper , we consider some particular cases of these graphs and study their total domination polynomials . in section 2
, we consider graphs which obtain by a special point - attaching of a graph @xmath36 and @xmath37 copies of graphs @xmath38 .
we prove that all graphs whose total domination polynomial have just two roots @xmath39 are in this form .
also we study the total domination polynomial of some kind of generalized friendship graphs in this section . in section 3 ,
we investigate the total domination polynomial of cactus chains .
in this section , we consider graphs constructed from primary subgraphs and study their total domination polynomial .
some kind of these graphs have interesting properties . in the subsection 2.1
, we prove that a special kind of graphs from primary subgraphs have exactly two total domination roots . in subsection 2.2
we study the total domination polynomial of the generalized friendship graph .
graphs whose certain polynomials have few roots can sometimes give interesting information about the structure of the graph .
the characterization of graphs with few distinct roots of characteristic polynomials ( i.e. , graphs with few distinct eigenvalues ) have been the subject of many researchers @xcite .
also the first authors has studied graphs with few domination roots in @xcite .
let @xmath36 be an arbitrary graph of order @xmath1 and consider @xmath1 copies of graph @xmath38 . by definition , the graph @xmath40 is obtained by identifying each vertex of @xmath36 with an end vertex of a @xmath38 .
see figure [ h(3 ) ] . to obtain the total domination polynomial of @xmath40 , we need the following result .
.,width=139,height=120 ] @xcite[3 ] let @xmath4 be a connected graph of order @xmath41 .
then @xmath42 if and only if @xmath4 is @xmath43 , @xmath44 or @xmath40 for some connected graph @xmath36 .
[ hop ] for any graph @xmath36 of order @xmath1 , we have @xmath45 .
* proof .
* let @xmath2 be a total dominating set of @xmath40 of size @xmath46 in figure [ h(3 ) ] .
obviously @xmath47 . to choose @xmath48 ( @xmath49 ) other vertices of @xmath50
, we have @xmath51 possibilities .
so we have the result.@xmath52 now , we state and prove the following result . [ 2roots ] let @xmath4 be a graph
. then @xmath53 if and only if @xmath54 for some graph @xmath36 of order @xmath1 . *
* ( @xmath55 ) it follows from theorem [ hop ] .
( @xmath56 ) let @xmath4 be a graph with @xmath53 .
thus @xmath57 and @xmath4 has no isolated vertex .
since @xmath58 , by theorem [ 3 ] , every component of @xmath4 is a cycle @xmath43 , @xmath44 or a @xmath40 for some connected graph @xmath36 .
since @xmath59 and @xmath60 does not divide @xmath61 , we conclude that there exists a graph @xmath36 such that @xmath62 and the proof is complete .
@xmath52 the characterization of graphs whose graph polynomials have few roots have been an interesting problem and studied well in the literature ( @xcite ) .
also there is a conjecture in @xcite which states that every integer total domination roots is in the set @xmath63 .
so finding the graphs whose total domination polynomial have these few roots can be a good start for solving this conjecture .
theorem [ 2roots ] characterize all graphs whose total domination polynomial have just two distinct roots @xmath64 and @xmath65 .
now we consider another kind of point - attaching graphs and study their total domination polynomials .
the generalized friendship graph @xmath66 is a collection of @xmath1 cycles ( all of order @xmath67 ) , meeting at a common vertex ( see figure [ dutch ] ) .
the generalized friendship graph may also be referred to as a flower @xcite . for @xmath68
the graph @xmath66 is denoted simply by @xmath69 and is friendship graph .
the total domination polynomial of @xmath69 and its roots studied in @xcite . here , first we compute the total domination number of @xmath70 . and
@xmath70 , respectively.,width=453,height=113 ] for any @xmath71 , we have @xmath72 .
* proof .
* let @xmath73 be vertex set of @xmath70 that adjacent by @xmath74 ( common vertex in @xmath70 ) .
then @xmath75 is a total dominating set for @xmath70 ( see figure [ dutch ] ) and the set @xmath76 of size less than or equal @xmath1 is not total dominating set for @xmath70 , therefore @xmath72.@xmath52 the following theorem is useful for finding the recurrence relations of the domination polynomials of graphs .
the vertex contraction @xmath77 of a graph @xmath4 by a vertex @xmath78 is the operation under which all vertices in @xmath79 are joined to each other and then @xmath78 is deleted ( see @xcite ) .
@xcite [ thm ] 1 . for any vertex @xmath78 in the graph @xmath4
we have + @xmath80,x)-(1+x)p_u(g)$ ] , + where @xmath81 is the polynomial counting the total dominating sets of @xmath82 which do not contain any vertex of @xmath79 in @xmath4 .
2 . let @xmath83 be two non - adjacent vertices of @xmath4 with @xmath84 .
then + @xmath85,x)$ ] .
3 . if @xmath83 be two vertices of @xmath4 with @xmath17\subseteq n[u]$ ] .
then + @xmath86,x)$ ] .
4 . let @xmath87 and @xmath17=n[u]$ ] .
then + @xmath88,x)$ ] .
now we state and prove a recurrence relation for the total domination polynomial of @xmath70 .
[ fn4 ] for any @xmath89 , we have @xmath90 $ ] , with initial value @xmath91 .
* proof .
* consider graph @xmath70 and @xmath78 , @xmath15 as shown in figure [ f_n,4 ] . by theorem [ thm ]
, we have @xmath92 \\&\hspace*{2.5cm}+x^2d_t(p_3,x)^{n-1 } \\&\hspace*{2.2cm}=(x^2 + 2x)d_t((f_{n,4}\setminus u)/v , x)+(x^3 + 2x^2)d_t(p_3,x)^{n-1 } \\&\hspace*{1.9cm}\stackrel{part ( i)}{=}(x^2 + 2x)[d_t(f_{n-1,4},x)+xd_t(f_{n-1,4},x)-(x+1)p_w((f_{n,4}\setminus u)/v ,
x ) ] \\&\hspace*{2.5cm}+(x^3 + 2x^2)d_t(p_3,x)^{n-1 } \\&\hspace*{2.1cm}=x(x+1)(x+2)d_t(f_{n-1,4},x)-x(x+1)(x+2)p_w((f_{n,4}\setminus u)/v , x ) \\&\hspace*{2.5cm}+(x^3 + 2x^2)d_t(p_3,x)^{n-1}. \end{aligned}\ ] ] since @xmath93 , so we have the result.@xmath52 , width=491,height=162 ]
in this section , we consider another kind of point - attaching graphs and study their total domination polynomials .
these kind of graphs are important in chemistry .
a cactus graph is a connected graph in which no edge lies in more than one cycle .
consequently , each block of a cactus graph is either an edge or a cycle . if all blocks of a cactus @xmath4 are cycles of the same size @xmath8 , the cactus is @xmath8 -uniform .
a triangular cactus is a graph whose blocks are triangles , i.e. , a @xmath94 -uniform cactus .
we call the number of triangles in @xmath4 the length of the chain . obviously , all chain triangular cactus of the same length are isomorphic .
hence , we denote the chain triangular cactus of length @xmath1 by @xmath95 ( see figure [ gn ] ) . by replacing triangles in this definitions by cycles of length @xmath96
we obtain cacti whose every block is @xmath97 . in this section
we shall study the total domination polynomial of cactus chains . in this subsection we shall study the total domination polynomial of chain triangular cactus . to do this ,
we consider graph @xmath98 as shown in figure [ gn ] , which is also a kind of point attaching graphs .
note that @xmath98 is a point attaching of @xmath95 and @xmath32 .
first we state and prove the following theorem : * proof .
* consider the graph @xmath98 as shown in figure [ gn ] .
since @xmath104 is isomorphic to @xmath105 , by theorem [ thm](i ) , we have @xmath106,x)-(x+1)p_u(g_n ) \\&\hspace*{1.6cm}=(x+1)d_t(t_n , x)+x^2d_t(g_{n-2},x)-(x+1)d_t(g_{n-1},x ) . \end{aligned}\ ] ] note that @xmath107 .
@xmath52 * proof .
* consider the graph @xmath95 and its vertex @xmath78 as shown in the figure [ gn ] . by theorem [ thm](iii ) , we have @xmath109,x)+d_t(t_{n}\setminus n[\{u , w\}],x ) ] \\&\hspace*{1.5cm}=(x+1)d_t(g_{n-1},x)+x^2[d_t(g_{n-2},x)+d_t(g_{n-3},x ) ] . \end{aligned}\ ] ] note that @xmath110 is isomorphic to @xmath111.@xmath52 1
. consider graph @xmath123 and its vertex @xmath78 in figure [ qn ] .
by theorem [ thm](iii ) , @xmath124,x)+2d_t(q_n(1)\setminus n[\{u , w\}],x ) ] \\ & \hspace*{2.1cm}=xd_t(q_{n}+e , x)+x^2[d_t(q_{n-1}x)+2d_t(q^{'}_{n-1},x ) ] .
\end{aligned}\ ] ] by applying theorem [ thm](iv ) on @xmath125 , we have + @xmath126 .
+ so we have result .
2 . consider the vertex @xmath78 as shown in figure [ qn ] . by theorem [ thm](iii ) , we have @xmath127,x)+d_t(q_n(2)\setminus n[\{u , w\}],x ) ] \\ & \hspace*{2.1cm}=xd_t(q_{n-1}(1),x)+x^2[d_t(q^{'}_{n-1},x)+d_t(q_{n-1},x ) ] .
\end{aligned}\ ] ] by using part @xmath128 in the above equation , we have result .
3 . consider graph @xmath129 in figure [ qn ] . by theorem [ thm ] , @xmath130,x )
\\ & \hspace*{1.7cm}=(x+1)d_t(q_{n}(1),x)+x^2d_t(q_{n-1},x ) .
\end{aligned}\ ] ] using part @xmath128 in the above equation , we have result .
4 . by theorem
[ thm](iii ) , we have @xmath131,x)+d_t(q_n^\delta \setminus n[\{u , w\}],x ) ] \\ & \hspace*{1.7cm}=(x+1)d_t(q_{n}(1),x)+x^2[d_t(q^{'}_{n-1},x)+d_t(q_{n-1},x ) ] .
\end{aligned}\ ] ] so by using part @xmath128 in the above equation , we have result.@xmath52 * proof . * with regards to figure [ q_n ] and theorem [ thm ]
, we have @xmath135,x)+d_t(q_n\setminus n[\{u , w\}],x ) \\&\hspace*{1.6 cm } = d_t(q_{n-1}(2),x)+xd_t(q_{n-1}^\delta , x)+x^2[d_t(q_{n-2}^{'},x)+d_t(q_{n-2},x ) ] \end{aligned}\ ] ] now by lemma [ le ] results is obtained.@xmath52 e.r .
van dam , _ graphs with few eigenvalues , an interplay between combinatorics and algebra , _ center dissertation series 20 , thesis , tilburg university , 1996 .
van dam , _ nonregular graphs with three eigenvalues , _
j. combin .
theory ser .
b 73 ( 1998 ) 101 - 118 . | let @xmath0 be a simple graph of order @xmath1 .
the total dominating set is a subset @xmath2 of @xmath3 that every vertex of @xmath3 is adjacent to some vertices of @xmath2 .
the total domination number of @xmath4 is equal to minimum cardinality of total dominating set in @xmath4 and denoted by @xmath5 .
the total domination polynomial of @xmath4 is the polynomial @xmath6 , where @xmath7 is the number of total dominating sets of @xmath4 of size @xmath8 .
let @xmath4 be a connected graph constructed from pairwise disjoint connected graphs @xmath9 by selecting a vertex of @xmath10 , a vertex of @xmath11 , and identify these two vertices .
then continue in this manner inductively .
we say that @xmath4 is obtained by point - attaching from @xmath12 and that @xmath13 s are the primary subgraphs of @xmath4 . in this paper
, we consider some particular cases of these graphs that most of them are of importance in chemistry and study their total domination polynomials .
department of mathematics , yazd university , 89195 - 741 , yazd , iran + [email protected] + * keywords : * total domination number , total domination polynomial , total dominating set . * ams subj . class .
: * 05a18 ; 11b 73 , 05c12 . |
recent progress in microfluidic devices has led to the development of novel separation strategies that take advantage of the unprecedented control on the geometry and chemistry of the stationary phase at scales that are comparable to the size of the transported species @xcite .
a fundamental problem that is at the core of several of the proposed separation devices is the transport of brownian particles through entropy barriers .
a representative example is the motion of a suspended particle in a channel with periodically varying cross section .
the purely - diffusive transport in the absence of an external force has been studied extensively @xcite , and a well - known approach is to reduce the dimensionality of the problem via the fick - jacobs approximation . in this approximation
the motion in the cross section is reduced to an entropic contribution to the longitudinal transport @xcite .
the case of biased diffusion in the presence of a driving force has also received considerable attention due to its relevance to separation devices .
hnggi and coworkers examined the validity of a direct extension of the fick - jacobs equation to describe the biased motion of a brownian particle in a narrow channel of varying cross section .
the authors showed that using an ad - hoc position - dependent diffusivity provides a good approximation , for pclet numbers below a critical value @xcite .
dorfman and coworkers @xcite also investigated the biased motion of a brownian particle in a periodic channel .
specifically , they performed an asymptotic perturbation analysis to obtain simple expressions for the macroscopic transport coefficients that remain valid at relatively large pclet numbers . here
, we also consider the transport of brownian particles confined to a channel of periodically varying cross section but , in the present case , the confinement is induced by a potential energy landscape and not the solid boundaries of a channel .
this type of spatial confinement occurs , for example , in the transport of suspended particles in microfluidic channels .
depending on the density of the particles , the van der waals and electrostatic forces between the particle and the channel walls , and the chemical composition of the media , the particles could become confined to the secondary minima of the particle - wall interaction potential in the vertical direction @xcite . in that case
, the presence of a periodic pattern on the bottom wall , such as that created by the deposition of thin metal stripes perpendicular to the flow , would lead to the confinement of the brownian particles to a periodically varying channel parallel to the wall , analogous to the one considered here .
we shall show that the effect on the average velocity of the particles ( and the probability distribution in general ) induced by this type of _ soft _ confinement of the suspended particles to an energy minima is the same as that caused by geometric confinement between solid walls .
understanding the effect that this type of confinement has on the transport of suspended particles is particularly important for the development of recently proposed separation techniques in microfluidic devices that are based on partitioning @xcite .
let us consider the transport of brownian particles in a potential that is periodic in the @xmath0-direction , @xmath1 , and that confines the particles in the @xmath2-direction , that is @xmath3 for @xmath4 . in the limit of negligible inertia effects the motion of the particles
is described by the smoluchowski equation for the probability density @xmath5 , @xmath6 the probability flux , @xmath7 , is given by @xmath8 where @xmath9 is a uniform external force in the @xmath0-direction , @xmath10 is the viscous friction coefficient , and we have used the stokes - einstein equation to write the diffusion coefficient as a function of @xmath10 , @xmath11 . in order to obtain the asymptotic distribution of particles within a single period of the potential we first introduce the reduced probability density and the reduced probability current ( see refs .
@xcite or the analogous approach presented in ref .
@xcite ) , @xmath12 the reduced probability is obtained by solving the smoluchowski equation with periodic boundary conditions in @xmath0 . in particular , the long - time asymptotic probability density , @xmath13 , is governed by the equation ( dropping the tilde ) , @xmath14 the far - field condition in @xmath2 is a vanishingly small probability density and flux due to the confining potential , @xmath15 { } 0 .
\label{eqn : flux}\ ] ] finally , the reduced probability is obtained by imposing the periodic boundary conditions in @xmath0 , @xmath16 and the normalization condition , @xmath17 in the case of narrow or _ slender _ channels , we assume that the characteristic length scale in the cross - section , i. e. perpendicular to the channel centerline , is given by @xmath18 , where @xmath19 is the slenderness ratio .
we can then write the governing equation and boundary conditions using the following dimensionless variables , @xmath20 , @xmath21 , @xmath22 , where @xmath23 is the boltzmann constant and @xmath24 is the absolute temperature , and the re - scaled probability density @xmath25 .
for the sake of simplicity all bars are dropped after nondimensionalization . the governing equation for the reduced probability ( eq .
( [ eqn : governing_original ] ) ) then becomes : @xmath26 + \frac{\partial}{\partial z}\left[-\frac{\partial v}{\partial z}p_{\infty}-\frac{\partial p_{\infty}}{\partial z}\right ] = 0 , \label{eqn : governing}\ ] ] where the pclet number is defined as @xmath27 .
simple inspection of this equation shows that the leading order approximation for @xmath19 is the local equilibrium in @xmath2 , which corresponds to @xmath28 , due to the no - flux far - field condition . in general
, an accurate description of the long - time transport of brownian particles can be obtained from the first two moments of the asymptotic probability , which describe the average velocity and the broadening of the distribution .
applying macrotransport theory we can write the average velocity in terms of the asymptotic probability distribution @xcite , @xmath29dz.\ ] ] the dispersion coefficient @xmath30 , can also be calculated from the asymptotic distribution via the so - called @xmath31field , which is the solution of the following differential equation @xcite , @xmath32=\epsilon^{2}p_{\infty}\langle v\rangle .
\end{split}\ ] ] the boundary conditions for the @xmath31field are , @xmath33{}0 , \\
\nonumber b(x=1,z)-b(x=0,z)=-1.\end{aligned}\ ] ] finally , the dispersion coefficient is given in terms of @xmath34 by @xmath35dz.\ ] ]
we apply the standard long - wave asymptotic analysis to obtain an approximate solution to the problem described above .
first , we propose a solution to the stationary probability distribution in the form of a regular perturbation expansion in the slenderness parameter , @xmath36 analogously , we write a regular perturbation expansion for the probability flux , @xmath37 substituting these expansions into eq .
( [ eqn : governing ] ) it is straightforward to determine the governing equation for the leading order terms , @xmath38 the corresponding leading order boundary and normalization conditions , derived from eqs .
( [ eqn : flux]-[eqn : normalization ] ) , are : @xmath39 then , integrating eq .
( [ eqn : governing_0 ] ) and taking into account the no - flux condition we obtain : @xmath40 where @xmath41 is an unknown function that is to be determined from the second order @xmath42 balance of eq .
( [ eqn : governing ] ) , @xmath43 .
\label{eqn : governing_1}\ ] ] the corresponding boundary and normalization conditions are : @xmath44 substituting the solution obtained for @xmath45 into eq .
( [ eqn : governing_1 ] ) we obtain @xmath46 .
\label{eqn : p1b0}\ ] ] in steady state , the total flux in the @xmath0-direction is constant along the channel .
therefore , by integrating the equation above over the cross - section , and taking into account the zero flux condition in the @xmath2-direction , we obtain @xmath47=0 , \label{eqn : b0}\ ] ] where @xmath48 we note that @xmath49 plays the role of the width @xmath50 of a solid channel , that is , the integral in the previous equation becomes equal to the width of the channel if we model the rigid boundaries by a potential field that is zero ( infinite ) inside ( outside ) the channel .
in fact , replacing @xmath49 by @xmath50 in eq .
( [ eqn : b0 ] ) we obtain eq . ( 25 ) in ref .
this suggests that the confining potential @xmath51 has similar effects on the transport of brownian particles to those induced by a channel with solid walls . before we proceed to solve the equation for @xmath41 it is also interesting to compare it with the equation obtained from the direct extension of the fick - jacobs approximation in the presence of an external force .
in fact , in the narrow geometry that we are considering here , we can assume that the particle will reach local equilibrium in the cross section fairly rapid compared to its diffusive or convective motion along the channel .
this separation of time scales suggests that the conditional probability @xmath52 could be approximated by the equilibrium distribution @xmath53 .
then , the total probability density takes the form , @xmath54 where @xmath55 is the marginal probability distribution , @xmath56 considering the long time limit , we can then use the same approximation for the asymptotic distribution , @xmath57 finally , we can project the problem into the longitudinal direction by substituting this expression into the governing equation for the probability density and integrating over the cross - section , @xmath58i(x)\right\}=0 .
\label{eqn : fj}\ ] ] this is the form of the fick - jacobs equation used in the presence of an external force in previous studies @xcite . it is interesting to point out that , by making the substitution @xmath59 , the previous equation becomes identical to eq .
( [ eqn : b0 ] ) , which was derived from the asymptotic analysis . therefore , the leading order term of the asymptotic analysis is equivalent to the the proposed extension of the fick - jacobs approximation in the presence of an external field .
then , the validity of the ad - hoc position - dependent effective diffusivity proposed in ref .
@xcite could , in principle , be tested by extending the present approach to higher orders in the slenderness parameter .
we now go back to equation ( [ eqn : b0 ] ) to obtain the general solution for @xmath41 , @xmath60 where @xmath61 and @xmath62 are the constants of integration .
the leading order contributions to the average velocity and the dispersion coefficient can then be calculated from the probability distribution .
first , we obtain the general expression for the average velocity following a derivation analogous to that presented in ref .
@xcite ( appendix a ) for the particle current , @xmath63^{-1}.\ ] ] in order to calculate the dispersion coefficient given in eq .
( [ d * ] ) we first propose a regular perturbation expansion for the @xmath64-field , analogous to those proposed for the probability density and flux , @xmath65 substituting the expansion for both the probability density and the @xmath64-field into eq .
( [ eq : b ] ) it is straightforward to show that the zeroth - order balance is an homogeneous equation that is satisfied by an arbitrary function @xmath66 . the governing equation for @xmath66 is obtained from the @xmath42 balance , after integrating over the cross - section and taking into account the no - flux boundary condition for the @xmath64-field in the @xmath2-direction , @xmath67 the remaining boundary condition is , @xmath68 the above two equations determine @xmath69 uniquely to within an arbitrary additive constant , which does not affect the calculation of the effective dispersion coefficient @xcite , @xmath70
in this section , we discuss a specific example of the confining potential that illustrates our previous results and allows direct comparison with numerical results . consider the transport of brownian particles confined to a narrow channel by a parabolic potential of periodically varying width , @xmath71 where @xmath72 is the position of the center of the channel , which we assume to be constant ( straight centerline ) , and @xmath73 determines the minimum opening of the channel .
the slenderness of the geometry is given by @xmath74 , which is the ratio of the characteristic width of the potential in the @xmath2-direction , @xmath18 , to the length of one period , @xmath75 .
using the same dimensionless variables introduced before and dropping again all bars for the sake of simplicity , we obtain , @xmath76 where @xmath77 it is clear then that the particles will be confined in the @xmath2 direction by a parabolic potential and that the width of the confining region is determined by @xmath78 .
in fact , the equilibrium distribution of particles is given by the boltzmann distribution with variance @xmath79 . in fig .
[ fig : channel ] we plot the equipotential lines @xmath80 , with @xmath81 and @xmath82 . in equilibrium , approximately @xmath83 of the particles are confined to the region enclosed by these lines . for a periodic potential @xmath84 with @xmath85 and @xmath82.,width=384 ] for this specific potential
we obtain @xmath86 , and @xmath87 where @xmath88 and @xmath89 are integration constants .
in particular , @xmath88 is the first constant of integration of eq .
( [ eqn : b0 ] ) , which corresponds to the total flux in the @xmath0-direction , @xmath90 .
thus , imposing periodicity and the normalization condition we obtain the leading order probability density , @xmath91 and average velocity , @xmath92 finally , the leading order term of the marginal probability distribution , corresponding to the distribution in the heuristic extension to the fick - jacobs approximation , is given by @xmath93\ ] ] note that in the limiting case in which diffusive transport is dominant ( vanishingly small pclet number ) we recover the equilibrium boltzmann distribution , @xmath94 on the other hand , in the limit of deterministic motion we obtain , @xmath95 in this case , there is no boundary layer developing due to the large magnitude of the driving force , as it would be the case in the presence of solid walls and permeating forces @xcite .
therefore , the diffusive fluxes are negligible compared to convective transport , and the average velocity tends to its bulk value , something that is also clear from eq .
( [ avv ] ) .
in addition , the integral of the asymptotic distribution over a cross section is uniform , which is also clear from eq .
( [ eqn : b0 ] ) .
note , however , that this limit is valid if the slenderness condition satisfies @xmath96 , as discussed in ref .
we performed brownian dynamics simulations of particle transport under confinement by a potential landscape to examine the previous asymptotic results in more detail .
the motion of brownian particles in a viscous solvent in the limit of vanishingly small inertia is governed by the overdamped langevin equations , @xmath97 and @xmath98 where @xmath99 is a zero - mean gaussian white noise with the correlation @xmath100 .
the value of the slenderness parameter used in the numerical simulations is @xmath101 . for different values of the pclet number ( @xmath102 ) .
the solid lines indicate the asymptotic results and the symbols correspond to the results of the brownian dynamics simulations.,width=384 ] in figure [ fig : p0x_both ] we show the marginal probability distribution @xmath103 in a single period of the potential , for different values of the pclet number @xmath104 .
the leading order term of the distribution , given by eq .
( [ marginal ] ) , agrees well with the results of the brownian dynamics simulations for all pclet numbers .
the figure also shows that the marginal probability distribution tends to a uniform distribution as the pclet number increases .
this indicates that , as discussed before , the effect of the potential on the particle distribution , as well as the contribution of diffusive transport to the average velocity of the particles , decreases as the pclet number increases .
figure [ fig : v_both ] shows that the average velocity of the particles obtained in the simulations is accurately described by the leading order term in the perturbation expansion .
finally , in figure [ fig : d_both ] we compare the effective dispersion coefficient obtained from the leading order term in the @xmath64-field ( numerically solving eq .
( [ eqn : bfield ] ) by means of finite differences ) with that computed directly from the simulations .
we observe good agreement between the simulations and the asymptotic analysis .
we have investigated the transport of brownian particles driven by a uniform external force .
the particles are confined to a narrow periodic channel by a parabolic potential in the cross - section .
we used asymptotic methods to obtain the leading order solution of the two - dimensional smoluchowski equation for the long - time probability distribution of particles reduced to a single period of the potential .
we first showed that the leading order analysis reproduces a previously proposed extension of the fick - jacobs approximation to biased transport .
we thus provide a systematic method to improve on the fick - jacobs approximation through higher order analysis .
we also showed that the leading order equation is equivalent to that obtained for solid channels and thus demonstrated that a confining potential has analogous effects on the distribution of particles and their transport parameters , such as the average velocity and the asymptotic dispersion coefficient .
we then analyzed the case of a cosine variation in the aperture of the confining channel and compared the results with brownian dynamics simulations .
we compared the long - time marginal distribution as well as its first moments ( average velocity and dispersion coefficient ) and obtained excellent agreement with the first order in the asymptotic analysis over a wide range of pclet numbers . | we investigate the transport of brownian particles in a two - dimensional potential under the action of a uniform external force .
the potential is periodic in one direction and confines the particle to a narrow channel of varying cross - section in the other direction .
we apply the standard long - wave asymptotic analysis in the narrow dimension and show that the leading order term is equivalent to that obtained previously from a direct extension of the fick - jacobs approximation .
we also show that the confining potential has similar effects on the transport of brownian particles to those induced by a solid channel .
finally , we compare the analytical results with brownian dynamics simulations in the case of a sinusoidal variation of the width of the parabolic potential in the cross - section .
we obtain excellent agreement for the marginal probability distribution , the average velocity of the brownian particles and the asymptotic dispersion coefficient , over a wide range of pclet numbers . |
the metal - insulator transition ( mit ) in two dimensions ( 2d ) has been a subject of intensive research in recent years [ 1 ] but the physics behind this phenomenon is still not understood .
it has been established , however , that it occurs in the regime where both coulomb ( electron - electron ) interactions and disorder are strong .
the competition between these two effects has been suggested [ 2 ] to lead to glassy dynamics ( electron or coulomb glass ) .
therefore , it has been proposed that the 2d mit can be described alternatively as the melting of the wigner glass [ 3 ] , or the melting of the coulomb glass [ 4 ] .
the existence of a large number of metastable states in a glass results in fluctuations of conductivity @xmath6 with time ( conductivity noise ) so that mesoscopic samples should be more suitable for studies of glassy properties .
indeed , investigations of metallic spin glasses have shown [ 5 ] that mesoscopic measurements are necessary in order to provide detailed understanding of glassy ordering and dynamics . here
we report the results of transport and noise studies in mesoscopic samples , which demonstrate for the first time the existence of a glassy phase in a 2d system in semiconductor heterostructures .
the measurements were performed on a 1 @xmath7 m long , 90 @xmath7 m wide rectangular n - channel si metal - oxide - semiconductor field - effect transistor ( mosfet ) with the peak mobility of only @xmath8m@xmath9/vs at 4.2 k. other sample details have been given elsewhere [ 6 ] .
the fluctuations of current @xmath10 ( _ i .
e. _ @xmath6 ) as a function of time were measured using a low - noise current preamplifier and a standard two - probe lock - in technique at @xmath11 hz with the constant excitation voltage @xmath12 a few @xmath7v . by changing the gate voltage @xmath13
, the electron density @xmath0 was varied between @xmath14@xmath15 and @xmath16@xmath15 , whereas temperature @xmath17 ranged between 0.13 k and 0.80 k. it was established that the current fluctuations were not correlated with ( _ i. e. _ not due to ) the fluctuations of temperature or applied voltage . in addition
, another sample from the same wafer was measured in a four - terminal configuration at 0.25 k , and it was determined that the contact resistances and the contact noise were negligible .
1 ( left ) shows the time dependence of the relative fluctuations @xmath18 ( where @xmath19 denotes averaging over time ) for a fixed @xmath0 at different @xmath17 . at the lowest @xmath20 k ,
fluctuations are as high as 250 @xmath21 , and drop rapidly with increasing @xmath17 .
a similar decrease in fluctuations was observed with increasing @xmath0 , as discussed in more detail below .
the time - averaged conductivity @xmath22 was calculated for time intervals of several hours , for all @xmath0 and @xmath17 .
1 ( right ) depicts @xmath22 as a function of @xmath17 at selected @xmath0 .
similar to the behavior of various high - mobility 2d systems [ 1 ] , the temperature coefficient of conductivity @xmath23 changes sign when @xmath24 . even though the corresponding density @xmath25@xmath15 is much higher due to a large amount of disorder in our samples , the effective coulomb interaction is still comparable to that in other 2d systems ( @xmath26 , @xmath27 ratio of coulomb energy to fermi energy ) .
the densities @xmath28 , where @xmath29 , have been usually [ 1 ] identified with the critical density @xmath30 for the metal - insulator transition . however , a thorough analysis of @xmath31 at low @xmath0 shows [ 6 ] that , in our case , @xmath32@xmath15 , which is more than a factor of two smaller than @xmath28 .
the difference between @xmath30 and @xmath28 in our samples is attributed to a larger amount of disorder .
the root - mean - square ( rms ) fluctuations @xmath33 ( calculated over the frequency bandwidth from ( 10 hours)@xmath34 to ( 6 seconds)@xmath34 ) increase with @xmath0 , ranging between @xmath35 at all @xmath17 .
2 ( left ) shows the relative rms , @xmath36 , as a function of @xmath0 at three different @xmath17 .
while fluctuations seem to be independent of density at high @xmath0 , a dramatic increase of @xmath36 is observed with decreasing @xmath0 below @xmath37@xmath15 . even though an increase in @xmath17 causes a substantial reduction of @xmath36
, the onset of large fluctuations does not seem to depend on temperature .
the noise was studied in more detail using normalized power spectra @xmath38 , most of which were obtained in the @xmath39 hz bandwidth .
they were found to follow the well - known power - law frequency dependence @xmath40 ( hooge s law ) [ 7 ] . in all measurements
, the device noise was extracted from the total measured noise by subtracting the background noise present with no current flowing , _ i. e. _ by setting @xmath41 .
the power spectrum of the background noise was always several orders of magnitude smaller than sample noise and had no @xmath42-dependence ( white noise ) .
2 ( center ) shows @xmath43 as a function of @xmath0 at @xmath44 hz and @xmath45 . at high @xmath0 ,
@xmath43 does not depend on @xmath0 within the scatter of data . for @xmath46 , however , an exponential rise of @xmath43 by six orders of magnitude is observed with decreasing @xmath0 .
this striking increase of the slow dynamic contribution to @xmath6 is consistent with the behavior of @xmath36 ( fig .
2 , left ) .
in fact , since @xmath47 , it is clear that the observed enormous increase of the relative rms as @xmath0 is reduced below @xmath2 ( fig .
2 , left ) reflects a dramatic slowing down of the electron dynamics .
this is attributed to the freezing of the electron glass .
2 ( right ) shows another striking feature of our data .
the exponent @xmath3 , which describes the frequency dependence of the noise power @xmath40 , exhibits a sudden jump at @xmath48 . while @xmath49 for @xmath50 , @xmath51 for @xmath46 ; similar large exponents @xmath3 have been observed in metallic spin glasses [ 8 ] .
in general , it is possible to obtain such high values of @xmath3 if telegraph noise is superimposed on flicker ( @xmath52 ) noise [ 9 ] .
however , our data exhibit many abrupt jumps rather than two - level switching ( see fig . 1 , left ) . for @xmath46 , a rapid _ decrease _ of @xmath43
is also observed with increasing @xmath17 ( inset to fig .
2 , center ) , consistent with other studies on si mosfets [ 10 ] at higher @xmath17 ( @xmath53 k ) .
the observed @xmath17-dependence of noise disagrees with the models of thermally activated charge trapping - detrapping fluctuations [ 9,11,12 ] , noise generated by temperature fluctuations [ 13 ] , and in the vicinity of the anderson transition [ 14 ] , all of which predict an _ increase _ of noise with increasing @xmath17 .
models of noise in the hopping regime [ 15 ] predict either a @xmath17-independent noise or a power - law decrease of noise with @xmath17 . in our experiment
, however , @xmath43 follows an exponential rather than power - law @xmath17-dependence . on the other hand ,
the observed @xmath54 does agree with the behavior found in mesoscopic spin glasses [ 16,8,17 ] .
measurements of the time dependence of @xmath6 in low - mobility si mosfets have shown that , with decreasing @xmath0 , a sudden , enormous increase in the low - frequency conductivity noise and a sudden shift of the spectral weight towards lower frequencies occur at a well - defined density @xmath2 .
in addition , a dramatic increase of noise with decreasing @xmath17 have been observed for @xmath46 .
similar behavior in metallic spin glasses was attributed to spin glass freezing [ 16,8,17 ] . for @xmath55 ,
we have also observed long relaxation times and history dependent behavior characteristic of a glassy phase , but these effects will be described in detail elsewhere .
these observations are interpreted as signatures of the freezing of the electron glass at @xmath56 .
the metal - insulator transition occurs at the density @xmath57 indicating the existence of a metallic electron glass , as predicted in [ 18 ] .
the authors are grateful to ibm , yorktown heights for sample fabrication , and to v. dobrosavljevi and j. jaroszyski for useful discussions .
this work was supported by nsf grant dmr-0071668 and by an nhmfl in - house research program grant .
99 - 0pt see e. abrahams , s. v. kravchenko , and m. p. sarachik , rev .
phys . * 73 * , 251 ( 2001 ) , and references therein .
j. h. davies , p. a. lee , and t. m. rice , phys .
* 49 * , 758 ( 1982 ) ; see also m. pollak and a. hunt in _ hopping transport in solids _ , edited by m. pollak and b. i. shklovskii ( elsevier , amsterdam , 1991 ) , and references therein
. s. chakravarty _
_ , phil . mag . b * 79 * , 859 ( 1999 ) .
a. a. pastor and v. dobrosavljevi , phys .
lett . * 83 * , 4642 ( 1999 ) .
see m. b. weissman , rev .
phys . * 65 * , 829 ( 1993 ) .
s. bogdanovich and d. popovi , submitted to phys .
lett . ; cond - mat/0106545 .
f. n. hooge , phys .
a * 29 * , 139 ( 1969 ) .
j. jaroszyski _
et al . _ ,
* 80 * , 5635 ( 1998 ) .
see m. b. weissman , rev .
mod . phys . * 60 * , 537 ( 1988 ) . c. j. adkins and r. h. koch , j. phys .
c * 15 * , 1829 ( 1982 ) . see p. dutta and p. m. horn , rev .
mod . phys . * 53 * , 497 ( 1981 ) . c. t. rogers and r. a. buhrman , phys .
53 * , 1272 ( 1984 ) .
r. f. voss and j. clarke , phys .
b * 13 * , 556 ( 1976 ) .
lett . * 69 * , 3555 ( 1992 ) ; o. cohen and z. ovadyahu , phys .
b * 50 * , 10442 ( 1994 ) .
b. i. shklovskii , solid state comm . * 33 * , 273 ( 1980 ) ; v. i. kozub , solid state comm .
* 97 * , 843 ( 1996 ) ; sh .
kogan , phys .
b * 57 * , 9736 ( 1998 ) .
n. e. israeloff _
et al . _ ,
lett . * 63 * , 794 ( 1989 ) .
g. neuttiens _ et al .
b * 62 * , 3905 ( 2000 ) . v. dobrosavljevi and a. a. pastor , this volume .
figure 1 left : relative fluctuations of @xmath6 _ vs. _ time for @xmath58@xmath15 at several different @xmath17 ( lowest @xmath17 at bottom , highest @xmath17 at top ) .
different traces have been shifted for clarity .
right : @xmath59 _ vs. _
@xmath17 for different @xmath0 .
the data for many other @xmath0 have been omitted for clarity . @xmath28 and @xmath2 are marked by arrows .
they were determined as explained in the main text .
left : relative rms @xmath60 _ vs. _ @xmath0 at different @xmath17 .
@xmath30 , @xmath2 , and @xmath28 are marked by arrows .
center : the normalized noise power @xmath43 at @xmath61 hz _ vs. _
@xmath0 at @xmath20 k ; inset : @xmath43 _ vs. _ @xmath17 for three different @xmath62@xmath63 .
right : the exponent @xmath3 ( @xmath64 ) _ vs. _ @xmath0 . | large fluctuations of conductivity with time are observed in a low - mobility two - dimensional electron system in silicon at low electron densities @xmath0 and temperatures . a dramatic increase of the noise power ( @xmath1 ) as @xmath0
is reduced below a certain density @xmath2 , and a sharp jump of @xmath3 at @xmath4 , are attributed to the freezing of the electron glass at @xmath5 .
the data strongly suggest that glassy dynamics persists in the metallic phase .
glassy dynamics , metal - insulator transition 71.30.+h , 73.50.td , 73.40.qv |
in these lines i intend to show that there exists consistent evidence pointing to the need of revision and further study of what seem at present a settled issue , namely the independence of the speed of electromagnetic radiation on the motion of its source . the main point in
the evidence is the range disagreement during the earth flyby of the spacecraft near in 1998 .
its range was measured near the point of closest approach using two radar stations of the space surveillance network ( ssn ) , and compared with the trajectory obtained from the deep space network ( dsn)@xcite . as for the range , the two measurements should match within a meter - level accuracy ( the resolution is 5 m for millstone and 25 m for altair ) , but actual data showed a difference that varies linearly with time ( with different slopes for the two radar stations ) up to a maximum difference of about 1 km , i.e. more than 100 times larger than the accuracy of the equipment used ( see figure 10 of @xcite ) .
further , when near crossed the orbits of gps satellites ( orbital radius 26,600 km ) the measured range difference was 650 m , that is , a time difference of 2 @xmath0s .
is it reasonable that any standard gps receiver performs better than dsn or ssn ?
there has not been a complete explanation for the range discrepancy .
it is very difficult to find any physical reason that may produce this anomaly , for any physical disturbance of the path of the spacecraft should manifest equally in ssn and dsn measurements .
guruprasad@xcite proposed an explanation that points to a time lag in the dsn signals , but the model is , at best , within 10% of the measured data and , more important , it fails to explain an important feature , that is , the different slope for the two radars . if we assume that systems are working properly , then the measured range difference ( time lag ) could be due to different propagation time of the employed signals .
additional points in the evidence come from anomalies related to the tracking of spacecrafts , present in both doppler and ranging data .
the pioneer anomaly@xcite and the flyby anomaly@xcite refer to small residuals of the differences between measured and modeled doppler frequencies of the radio signals emitted by the spacecrafts .
although these residuals are very small ( less than 1 hz on ghz signals ) the problem is that they follow a non - random pattern , indicating failures of the model . according to the temporal variation of those residuals
the pioneer anomaly exhibits a main term , an annual term , a diurnal term and a term that appears during planetary encounters .
it should be clarified that a few years ago an explanation of the pioneer anomaly was published@xcite .
however , it is a very specific solution that applies only to the main term of the pioneer spacecraft anomaly , but left unresolved many other anomalies , including those of the spaceships cassini , ulysses and galileo ; the annual term ; the diurnal term ; the increases of the anomaly during planetary encounters ; the flyby anomaly ; and the possible link between all them ( it is hard to think that there are so many different causes for the mentioned anomalies ) . for all this , i believe that the issue can not be closed as it stands .
as a matter of fact , the range difference between ssn and dsn , @xmath1 , is perfectly fitted with @xmath2 where @xmath3 is a vector range pointing from the spacecraft to the radar , @xmath4 the spacecraft velocity relative to the radar , and @xmath5 the speed of light .
figure [ range ] shows this fit and its comparison with measured data .
the orbital and measured data were taken from@xcite .
although the exact location of the radar stations are unknown , the fit is statistically significant for both radar stations ( @xmath6 ) including the first outliers points .
it reproduces the ( almost ) linear dependence with time during the measured interval , and the two different slopes for millstone and altair stations due to their different locations .
since range measurements are based on time - of - flight techniques , the validity of ( [ deltar ] ) means that the electromagnetic waves ( microwave ) of the dsn and ssn travel at different speeds .
specifically , in the radar frame of reference , if the ssn waves travel at @xmath5 , then the dsn waves travel at @xmath5 plus the projection of the spacecraft velocity in the direction of the beam , in sharp contrast with the second postulate of special relativity theory ( srt ) . in view of the above result one may ask what is established , at present , about the relation of the speed of electromagnetic radiation ( light for short ) to the motion of the source . in order to elaborate this point the following questions are of relevance : 1 .
are there _ simultaneous _ measurements of the speed of light from different moving macroscopic sources ( not moving images ) with different velocities ? ; 2 . since ballistic ( emission ) theories are ruled out ( see , for example , desitter @xcite@xcite , brecher @xcite and alvager et al @xcite ) , how else could the speed of light depend on the source movement ?
how is it possible that there is a first order difference in @xmath7 in spacecraft range measurements , while at the same time there are many experiments on time dilation that are consistent with srt to second order in @xmath7 ( see , for example,@xcite ) ? ; 4 . if the velocity of light depend on the velocity of the source , why has this not been observed in other phenomena in the past ? ; in answer to the previous questions , so far as the author is aware , there is no known experimental work that simultaneously measures the speed of light from two different sources ( not images ) , or that simultaneously measures the speed of light and that of its source .
for example , in the work by alvager et al,@xcite the speed of light is measured at a later time ( @xmath8 200 ns ) than the emission time , and there is no measurement of the speed of the source at the time of the _ detection _ of the light .
note that measurements involving moving images produce different results from those produced by mobile sources .
for example , under srt , a moving source is affected by time dilation while a moving image is not .
therefore , to ensure the independence of the speed of light from its source movement , it is essential to have two sources with different movements .
although controversial and beyond the scope of the this note , time dilatation phenomena may be of different physical origin from first order terms , as it may be inferred from the work of schrdinger @xcite .
thus , measurements of time dilatation phenomena in accordance with srt , does not necessarily imply the independence of the speed of light with the movement of the source .
the experiments mentioned above @xcite@xcite@xcite @xcite only rule out ballistic theories in which radiation maintains the speed of the source at the time of _ emission _ , but do not rule out other ideas , like faraday s 1846@xcite .
in order to remove the ether , faraday introduced the concept of vibrating rays@xcite , in which an electric charge is conceived as a center of force with attached rays that extend to infinity .
the rays move with their center , but without rotating .
according to this view , the phenomenon of electromagnetic radiation corresponds to the vibration of these rays , that propagates at speed @xmath5 relative to the rays ( and the center ) .
that is , the radiation remains linked to the source even after emitted .
today we could describe the interaction as a kind of entanglement between the charge and the photon .
a framework for the electromagnetic phenomena according to faraday s ideas was developed .
it was called vibrating rays theory ( vrt ) @xcite in reference to faraday s vibrating rays . under faraday s idea
, the velocity of radiation at a given epoch will be equal to _ c _ plus the velocity of the source at the _ same _ epoch , in contrast with ballistic theories in which the emitted light retains the speed of the source at the _ emission _ epoch . in this sense
the radiation is always linked to the charge at every time after the emission .
consequently , the measured doppler effect corresponds to the speed of the source at the time of _ reception _ , as well .
further , a difference between active and passive reflection is expected , since the latter is still related to the original source according to vrt .
the deep space network ( dsn ) works with the so called active reflection ( the spacecraft re - emits in real time a signal in phase with the received signal from earth ) , while the space surveillance network ( ssn ) works with passive radar reflection . in consequence ,
the downlink signal from the aproaching spacecraft will propagate faster that the reflected one . using the available orbital data@xcite we found that , under vrt , the theoretical time - of - flight difference between active and passive reflection gives exactly the same range disagreement as ( [ deltar ] ) , see part 6 of @xcite .
the pioneer anomaly refers to the fact that the received doppler frequency differs from the modeled one by a blue shift that varies almost linearly with time , and whose derivative is @xmath9 where @xmath10 is the frequency difference between the measured and the modeled values . in the case of a source with variable speed ,
the main difference in doppler ( to first order ) between vrt and srt , is that srt relates to the speed of the source at the time of _ emission _ , while vrt relates to the speed of the source at the time of _ reception_. precisely , this difference seems to be present in the spacecraft anomalies .
if vrt is valid , it automatically invalidates all calculations and data analysis of spacecraft tracking which are based on srt .
so , it is not easy to make a direct comparison between the expected results from srt and vrt .
however , to see whether or not the main features predicted by vrt are present in the measurements , we can evaluate the residual by simulating a measured doppler signal assuming that light propagates in accordance to vrt but analyzed according to srt .
calling @xmath11 the emission time of the downlink signal from the spacecraft toward earth and @xmath12 the reception time at earth , the first order difference of the doppler shift between vrt and srt is ( see @xcite part 4 ) @xmath13 where @xmath14 and @xmath15 represent the velocities of the spacecraft at the corresponding epoch , @xmath16 is the unit vector from the spacecraft to the antenna , and @xmath17 the proper frequency of the signal .
that is , the velocity used in the srt formula is that at the time of _ emission _ while according to vrt is that corresponding at the time of _
reception_. since the spacecraft slows down as it moves away , then @xmath18 , therefore the difference corresponds to a small blue shift mounted over the large red shift , as it has been observed in the pioneer anomaly .
it should be noted that this difference appears because of the active reflection produced by the onboard transmitter . in case of a passive reflection ( for example , by means of a mirror )
the above difference vanishes .
an estimate of the order of magnitude of [ df ] is obtained by using that the variation of the velocity of the spacecraft between the time of emission and reception is approximately @xmath19 where @xmath20 is a mean acceleration during the downlink interval .
an estimate for the duration of the downlink is simply @xmath21 where @xmath22 is a mean position of the spacecraft between @xmath11 and @xmath12 , therefore @xmath23 since @xmath24 where @xmath25 is the gravitational constant , and @xmath26 the mass of the sun , then , the time derivative becomes @xmath27 if the difference ( [ ddfdtvrt ] ) is interpreted as an anomalous acceleration we get @xmath28 that is , the so - called anomalous acceleration is @xmath7 times the actual acceleration of the spacecraft . using data from horizons web - interface @xcite for the spacecraft ephemeris , some characteristic value for @xmath29 can be obtained .
consider the anomalous acceleration detected at the shortest distance of the cassini spacecraft during solar conjunction in june , 2002 .
the spacecraft was at a distance of @xmath30 au moving at a speed of @xmath31 km / s .
the anomalous acceleration given by ( [ aa ] ) is @xmath32 m / s@xmath33 of the same order of the measured one ( @xmath34 m / s@xmath33 ) .
also , the closest distance at which the pioneer anomaly has been detected was about @xmath35 au .
the anomalous acceleration predicted by ( [ aa ] ) at that distance is @xmath36 m / s@xmath33 of the same order as the measured one .
the anomaly given by ( [ aa ] ) decreases in time in a way that has not been observed .
note , however , that according to markwardt @xcite the expected frequency at the receiver includes an additional doppler effect caused by small effective path length changes , given by @xmath37 where @xmath38 is the rate of change of effective photon trajectory path length along the line of sight .
this is a first order effect that can partially hide the difference between srt and vrt .
therefore , a more careful analysis should take into account the additional contribution of ( dfpath ) in ( [ aa ] ) .
further , other first order effects may appear , for example , by a slight rotation of the orbital plane .
since orbital parameters are obtained by periodically fitting the measurements ( mainly due to spacecraft maneuvers or random perturbations ) with theoretical orbits , there is no straightforward way to weight the importance of this effect in ( [ aa ] ) . in other words , data
acquisition and analysis may hide part of the vrt signature .
apart from the residual referred to in the preceding paragraph there is also an annual term .
according to anderson et al @xcite the problem is due to modeling errors of the parameters that determine the spacecraft orientation with respect to the reference system .
anyway , levy et al @xcite claim that errors such as errors in the earth ephemeris , the orientation of the earth spin axis or the stations coordinates are strongly constrained by other observational methods and it seems difficult to modify them sufficiently to explain the periodic anomaly .
the advantage of studying the annual term over the main term , is that the former is less sensitive to the first order correction mentioned above , and , for the case of pioneer , also to the thermal propulsion correction@xcite .
clearly , the earth orbital position does not modify those terms .
as before , the annual term is explained by the difference between the velocity of the spacecraft at the time of emission and that at the moment of detection , which depends on whether the spacecraft is in opposition or in conjunction relative to the sun .
when the spacecraft is in conjunction , light takes longer to get back to earth than in opposition . the time difference between emission and reception
will be increased by the time the light takes in crossing the earth orbit . specifically , taking into account the delay due to the position of earth in its orbit , in opposition equation ( [ t3t2 ] ) should be written as @xmath39 while in conjunction it would be @xmath40 where @xmath41 is the mean orbital radius of earth .
therefore , an estimate of the magnitude of the amplitude of the annual term is @xmath42 for the case of pioneer 10 at 40 au we get @xmath43 and at 69 au @xmath44 in good agreement with the observed values .
using data from horizons web - interface @xcite a more complete analysis of the time variation of @xmath10 has be performed .
the residual ( that is , simulated doppler using vrt but interpreted under srt ) during 12 years time span is plotted in figure [ anual ] .
also the dumped sine best fit of the 50 days average measured by turyshev et al @xcite is plotted showing an excellent agreement between measurements and vrt prediction .
the negative peaks ( i.e. , maximum anomalous acceleration ) occur during conjunction when the earth is further apart from the spacecraft , and positive peaks during opposition .
also , the amplitude is larger at the beginning of the plotted interval and decreases with time , as it was observed@xcite@xcite .
like the pioneer anomaly , the earth flyby anomaly can be associated to a modeling problem , in the sense that relativistic doppler includes terms that are absent in the measured signals .
the empirical equation of the flyby anomaly is given by anderson et al@xcite , which , notably , can be derived using vrt , as is done in part 6 of @xcite .
consider the case of near tracked by 3 antennas located in usa , spain , and australia ( a full description of the tracking system is found in a series of monographs of the jet propulsion laboratory@xcite ) .
the receiving antenna was chosen as that having a minimum angle between the spacecraft and the local zenith . using available orbital data ,
a simulated doppler signal has been calculated using vrt .
thus , the simulated residual is obtained by subtracting the theoretical srt doppler , from the vrt calculation .
we observed , however , that the term that contains the velocity of the antennas , that is @xmath45 is not enough to completely remove the first order ( in @xmath46 ) earth signature ( @xmath47 is the velocity of the antenna , 1 refers to the emission epoch and 3 to the reception epoch , as in @xcite part 4 ) .
this is so because the velocity of the antennas is not uniform and the evaluation of the emission time is different for vrt and srt .
then , a small , first order related term remains .
anyway , since orbital parameters are obtained by periodically fitting the measurements to theoretical orbits , thus a similar procedure is needed for vrt . by doing
so the first order term is removed .
the only difference between orbits adjusted by srt and vrt is a slight rotation of the orbit plane , as mentioned above .
note that , in the case of range disagreement , two different orbital ajustment are needed by dsn and ssn due to the different propagation speed . in consequence
, it will be impossible to fit a simultaneous measurement , as it seems to happen with the range disagreement .
the final result shows that each antenna produces a sinusoidal residual with a phase shift at the moment of maximum approach .
therefore , if we fit the data with the pre - encounter sinusoid a post - encounter residual remains and vice versa . in figure [ flyby
] are simultaneously plotted the result of fitting the residual by pre - encounter data ( right half in red , corresponding to figure 2a of @xcite ) and by post - encounter data ( left half in blue , corresponding to figure 2b of @xcite ) .
note that the simulated plots are remarkably similar to the reported ones , including the amplitude and phase ( i.e. , minima and maxima ) of the corresponding antenna .
the fitting of post - encounter data ( blue ) can be improved by appropriately setting the exact switching times of the antennas ( which are unknown to the author ) .
the flyby doppler residual exhibits a clean signature of the vrt theory .
in this work i have presented observational evidence favoring a dependence of the speed of light on that of the source , in the manner implied in faraday s ideas of vibrating rays .
it is remarkable and very suggestive that , as derived from faraday s thoughts , simply by relating the velocity of light and the corresponding doppler effect with the velocity of the source at the time of detection , is enough to quantitatively and qualitatively explain a variety of spacecraft anomalies .
also , it is worth mentioning that a formulation of electromagnetism compatible with faraday s conception is possible , as shown in @xcite , which is also compatible with the known electromagnetic phenomena . this new formalism introduces a striking concept : that both instantaneous interaction ( static terms ) and retarded interaction ( radiative terms ) are simultaneously present .
finally , under vrt the manifestation of the movement of the source in the speed of light is more subtle than the naive @xmath48 hypothesis ( @xmath49 is a constant , @xmath50 ) usually used to test their dependence@xcite .
thus , it is also of fundamental importance the fact that , from the experimental point of view , it is very difficult to detect differences between vrt and srt , as discussed in @xcite , which is also manifest in the smallness of the measured anomalies , and in the non clear manifestation of the effect in usual experiments and observations .
for example , it produces a negligible effect on satellite positioning systems , see part 7 of@xcite .
i am aware of how counterintuitive these conceptions are to the modern scientist , but also believe that , given the above evidence , a conscientious experimental research is needed to settle the question of the dependence of the speed of light on that of its source as predicted by vrt , and that has been observed during the 1998 near flyby . as a closure , i recall fox s words regarding the possibility of conducting an experiment on the propagation of light relative to the motion of the source : `` _ nevertheless if one balances the overwhelming odds against such an experiment yielding anything new against the overwhelming importance of the point to be tested , he may conclude that the experiment should be performed .
i am thankful to fernando minotti who read this paper and improved the manuscript significantly , although he may not agree with all of the interpretations provided in this paper .
g. antreasian and j. r. guinn , aiaa paper no .
98 - 4287 presented at the aiaa / aas astrodynamics specialist conference and exhibit ( boston , august 10 - 12 , 1998 ) .
v. guruprasad , _ epl _ * 110 * , 54001 ( 2015 ) .
j. d. anderson , p. a. laing , e. l. lau , a. s. liu , m. m. nieto , and s. g. turyshev , _ phys .
lett . _ * 81 * , 2858 - 2861 ( 1998 ) .
j. d. anderson , j. k. campbell , j. e. ekelund , j. ellis , and j. f. jordan , _ phys .
lett . _ * 100 * , 091102 ( 2008 ) .
s. g. turyshev , v. t. toth , g. kinsella , s .- c .
lee , s. m. lok , and j. ellis , _ phys .
rev . lett_. * 108 * , 241101 ( 2012 ) w. desitter , _
z. phys . _ * 14 * , 429 ( 1913 ) .
w. desitter , _ z. phys . _
* 14 * , 1267 ( 1913 ) .
k. brecher , _ phys .
* 39 * , 1051 ( 1977 )
. t. alvager , f. j. m. farley , j. kjellman , and i. wallin , _ phys .
* 12 * , 260 ( 1964 ) .
b. botermann et al , _ phys .
lett . _ * 113 * , 120405 ( 2014 ) .
e. schrodinger ; _ ann .
der physik _ * 77 * , 325 - 336 ( 1925 ) .
m. faraday , _ phil . mag .
_ * 28 * , 345 ( 1846 )
. l. bilbao , l. bernal , f. minotti ; `` vibrating rays theory , '' arxiv:1407.5001 [ physics.class-ph ] ( 2014 ) .
http://ssd.jpl.nasa.gov/. c. b. markwardt , `` independent confirmation of the pioneer 10 anomalous acceleration , '' arxiv : gr - qc/0208046 v1 ( 2002 ) .
j. d. anderson , p. a. laing , e. l. lau , a. s. liu , m. m. nieto , and s. g. turyshev , _ phys .
d _ * 65 * , 082004 ( 2002 ) .
a. levy , b. christophe , s. reynaud , j - m .
courty , p. brio , and g. mtris , `` pioneer 10 data analysis : investigation on periodic anomalies , '' in journes scientifiques de la sf2a , paris , france pp.133 - 136 , 2008 .
hal-00417743
g. turyshev et al , arxiv : gr - qc/9903024 v2 ( 1999 ) .
descansoteam , jet propulsion laboratory , california institute of technology .
( accessed july 2014 ) . j. g. fox , _ am .
_ * 30 * , 297 ( 1962 ) .
) is plotted ( full lines , millstone in blue and altair in red ) . for millstone ,
the error bars refer to the uncertainties in the extraction of the data from figure 10 of @xcite , rather than to its tracking error ( 5 m ) , while for altair , the accuracy is 25 m ] | data from spacecrafts tracking exhibit many anomalies that suggest the dependence of the speed of electromagnetic radiation with the motion of its source .
this dependence is different from that predicted from emission theories that long ago have been demonstrated to be wrong . by relating the velocity of light and the corresponding doppler effect with the velocity of the source at the time of detection , instead of the time of emission , it is possible to explain quantitatively and qualitatively the spacecraft anomalies . also , a formulation of electromagnetism compatible with this conception is possible ( and also compatible with the known electromagnetic phenomena ) . under this theory
the influence of the velocity of the source in the speed of light is somewhat subtle in many practical situations and probably went unnoticed in other phenomena . |
the critical collapse of scalar fields gives rise to a new class of thought experiments in general relativity @xcite .
it has been suggested that a weak singularity may be visible to a distant observer during the collapse of a scalar field @xcite . with the presence of such potentially problematic phenomena
, one can ask to what degree cosmic censorship is violated .
the spirit of the cosmic censorship conjecture is that the evolution of spacetimes as seen by distant observers in asymptotically flat regions is determined uniquely by the einstein equations .
if visible infinite curvature arises in the evolution from regular initial data , one might have cause for concern about the classical completeness of einstein s theory of gravity .
the critical collapse is a fine - tuned limit , and `` strong cosmic censorship '' has been formulated to exclude such rare cases with zero measure in the space of initial conditions .
it is nevertheless instructive to understand the nature of this singularity , and how such a limit can be taken .
recently , martin - garcia and gundlach @xcite have numerically constructed a self - consistent discretely self - similar spacetime which they argued to be related to the evolutionary critical collapse solutions .
they proposed the existence of a future cauchy horizon emanating from the critical collapse event , on which new data must be specified .
the authors proceeded to find a unique way of specifying this data which can result in a regular future spacetime . in this paper
, we consider the problem from a different perspective . instead of searching for the critical solution of einstein equations by first taking the limit of discrete self - similarity , we study the constructive sequence of non - critical global spacetimes .
we then search for a unique limit as we approach criticality .
posed in such a way , the existence of a cauchy horizon would be very unexpected : each spacetime in the limit sequence has no cauchy horizon , so why would it form in the limit ? to study the problem , we developed a characteristic code which can track a scalar field collapse interior to the horizon . in section
[ sec : paradox ] , we outline a conceptual paradox in the search for a global critical spacetime structure . in section [ sec : spherical ] , we derive the scalar field equations of motion with spherical symmetry . we proceed to solve these equations numerically in section [ sec : code ] .
the numerical results are presented in section [ sec : results ] , where we explain our proposed solution to the apparent paradox .
consider a collapsing scalar field with amplitude characterized by a parameter @xmath2 . for small amplitudes @xmath3 , the field collapses and re - expands . for large amplitudes @xmath4 , a black hole forms .
an interesting question is to examine the behavior when @xmath5 .
one can consider two limits , one from below and one from above .
if we consider a sequence of subcritical spacetimes with collapsing scalar fields as they approach @xmath6 from below , we might expect the global structure of the resulting spacetime to be minkowski , i.e. a triangle shown in the left panel of fig .
[ fig : comp ] .
if we consider the limit of a sequence of supercritical spacetimes from above , we expect each stage to have a global schwarzschild structure . as the parameter decreases , the mass of the resulting black hole approaches zero , and one expects the limiting spacetime to resemble a schwarzschild spacetime with zero mass , shown in the right panel of fig .
[ fig : comp ] . in this limit , @xmath1 coincides with the horizon and becomes null ( and infinitely redshifted ) .
the global structure looks quadrangular , quite unlike the argued spacetime in the other limit .
this suggests several possible interpretations .
perhaps the two limits are different , and the limiting spacetime depends on the direction from which the limit was taken . or one of the limits is only an incomplete description of spacetime , and might be extensible to the same global structure . or perhaps the limit is not convergent from either direction , and oscillates in such a way that more data must be specified on a spontaneously formed cauchy horizon @xcite .
our study suggests a slightly different physical interpretation of the global structure .
critical spacetimes are only known as numerical solutions , which makes questions about global structure hard to answer .
it is most easily studied in characteristic coordinates which follow light ray propagation @xcite . in the subsequent section
we will describe our formulation and implementation of the numerical procedures .
the spherically symmetric @xmath7-dimensional spacetime metric can be written as @xmath8 where @xmath9 is the metric of a unit @xmath10-dimensional sphere with curvature @xmath11 , and the two - manifold metric @xmath12 is conformally flat .
the dynamics of the scalar field collapse are described by the reduced action @xmath13
\big\},\nonumber\end{aligned}\ ] ] where integration and differential operators are with respect to the flat two - metric @xmath14 .
variation of the above action with respect to the fields @xmath15 , @xmath16 and @xmath17 gives equations of motion [ eom ] @xmath18 while the two constraint equations are recovered by variation with respect to the ( flat ) metric @xmath19 for the particular case of spherically symmetric scalar field collapse in four dimensions ( @xmath20 ) , the equations of motion ( [ eom ] ) can be simplified by introducing auxiliary field variables @xmath21 and @xmath22 : [ evolve ] @xmath23 this form of the dynamical equations is better suited for numerical evolution .
we discretize and evolve the collapsing scalar field spacetime in double - null coordinates @xmath24 , where the radial characteristics of the wave equations are made explicit : the outgoing characteristic propagates at constant @xmath25 in the direction of increasing @xmath26 , while the incoming characteristic propagates at constant @xmath26 in the direction of increasing @xmath25 .
this approach has a number of advantages over some of the more traditional spacetime slicings .
the characteristic code only propagates information along characteristics at a numerical speed equal to the true characteristic speed .
the numerical domain of dependence is no larger than the physical domain of dependence , and the code still maintains full ( second order ) accuracy .
horizons are not particularly special as far as ingoing null characteristics are concerned .
this allows us to follow collapse all the way to the singularity .
even as an outgoing characteristic hits a singularity and the floating point numbers denormalize , this does not affect any of the other characteristics , which can be integrated to fill the whole maximally extended spacetime determined by the initial data . to illustrate this point , in fig .
[ fig : bh ] we present penrose and kruskal diagrams of a spacetime with a large black hole formed in the collapse of the scalar field wavepacket
. the two - dimensional metric @xmath27 has a residual gauge freedom under redefinition of the null coordinates @xmath28 , @xmath29 .
these two free functions ( of a single variable ) are used to define the coordinate @xmath26 on the initial slice and to map the central point @xmath1 to a straight line in the @xmath30 plane : @xmath31 the second gauge condition is particularly convenient , since it places the central point at a known location on the grid when discretizing .
the initial conditions are specified on a surface of constant @xmath25 by giving a scalar field profile @xmath32 .
together with the gauge choice ( [ gauge ] ) , this determines the rest of the variables .
in particular , @xmath33 is obtained by integrating the outgoing ( @xmath34 ) constraint equation ( [ constraint ] ) @xmath35 the integration is implemented as a fourth order runge - kutta algorithm with a fixed step .
although strictly speaking it is not necessary , as the evolution code is second order , it is no more complicated than a second order integrator would be .
although the constraint equations follow from the dynamical equations , their use might be required for stability @xcite .
using the outgoing constraint equation for evolution ( rather than just for initial conditions ) is not a good idea , however .
it becomes degenerate on the apparent horizon , where the outgoing light rays become ( marginally ) trapped @xmath36 .
the incoming constraint equation is perfectly fine , though , and could be followed all the way to the center of the spacetime , whether it is singular or not . by introducing an auxiliary field variable @xmath37 , the incoming constraint equation ( [ constraint ] ) can be written in a form that is very simple to integrate @xmath38 to integrate the incoming constraint , one would need to know the values of @xmath39 on the initial slice @xmath40 .
these can be found by integration of the equation @xmath41,\ ] ] which is obtained by combining the constraints with the evolution equation ( [ evolve]a ) to solve for @xmath42 on the initial slice .
having discussed our gauge choice , initial conditions and constraint equations , we now come to the discretization of the evolution equations .
the covariant differential operators in the evolution equations ( [ evolve ] ) are with respect to a flat metric , so in null coordinates they are written simply as @xmath43 where @xmath44 denotes any one of the three dynamical variables in equations ( [ evolve ] ) .
we discretize by finite differencing on a rectilinear @xmath30 grid with equal spacing @xmath45 , which to second order accuracy gives @xmath46 , \\
( \nabla x)^2_\times & = & -\frac{1}{2 \delta^2 } \left[(x_{++ } - x_{--})^2 - ( x_{-+ } - x_{+-})^2\right ] , \nonumber\end{aligned}\ ] ] where the differential operators are evaluated at the center of a grid cell ( see fig . [
fig : discrete ] ) .
the code takes a step by using discretized evolution equations to find the values of the fields at the @xmath47 grid point ( shaded node in fig .
[ fig : discrete ] ) .
the only non - trivial operation involved in this is finding @xmath48 accurate to second order , which is done by solving equation ( [ evolve]a ) discretized in the following fashion : @xmath49 the rest is then straightforward , as the right hand sides of equations ( [ evolve]b , c ) become known . as the code advances to the next slice of constant @xmath25 , the very first point on the grid ( @xmath50 )
is the center of the spacetime ( @xmath1 ) and has to be treated specially .
the asymptotic form of the evolution equations at @xmath1 @xmath51 is used to calculate the values of the fields in the center .
we tested the code on the collapse of a scalar field with various initial conditions .
in particular , we used pulse @xmath52 and kink @xmath53 , & v < \frac{3}{4 } \smallskip\\ 2p , & v \ge \frac{3}{4 } \\ \end{array } \right.\ ] ] field profiles in our simulations .
both profiles are fairly smooth functions with field energy density having compact support on the initial slice .
this avoids the interference of long - range tails in the initial data on the late - time evolution .
all runs shown in this paper used 65536 uniformly spaced grid points .
the critical scaling of black hole mass is shown in fig .
[ fig : scaling ] .
it agrees well with the literature @xcite .
our achievable dynamic range shows scaling over three orders of magnitude in black hole mass without the use of adaptive mesh refinement .
this is sufficient for our study .
further improvements in dynamic range can be sought after using adaptive mesh refinement techniques or adopting an initial gauge which places ingoing null rays on a grid more densely near the collapse point @xcite .
as the grid resolution is increased , the double arithmetic precision and round - off errors become the main obstacles in the quest for higher dynamical range . fig .
[ fig : subcrit ] shows the penrose diagram of a subcritical spacetime just below the threshold of black hole formation .
the lines of constant radius @xmath54 are drawn as solid curves , and the proper time along such lines are drawn as dashed curves .
most of the scalar field mass is shed before @xmath55 , but the field oscillating on ever smaller scales creates near - singular curvature in the center of spacetime .
when we examine the penrose diagram of a supercritical collapse as shown in fig .
[ fig : supercrit ] , we see the formation of a horizon near @xmath56 .
the code evolves well inside the event horizon , and we can identify the apparent horizon in the interior of the black hole .
the lines of constant @xmath54 become spacelike beyond the apparent horizon .
all the larger radii outside the black hole pile up at the horizon in this diagram .
we can pose the question if the spacetime described by fig .
[ fig : supercrit ] could possibly be the same as fig .
[ fig : subcrit ] . our original gauge choice from equation ( [ gauge ] ) fixes the @xmath1 line to coincide with @xmath50 .
it does not have to be this way ; the null coordinates leave the possibility for a global gauge change @xmath28 .
one can try to `` un - pile '' the lines of constant @xmath54 observers for the supercritical collapse spacetime in fig .
[ fig : supercrit ] by defining a gauge change @xmath57 such that the second @xmath54 curve coincides with that in a subcritical collapse spacetime of fig .
[ fig : subcrit ] .
the result is shown in fig .
[ fig : stretch1 ] .
this stretching is rather sudden and throws the @xmath1 line into an almost null direction , but all the other finite @xmath54 s appear to coincide .
this is non - trivial , since the residual gauge freedom is a single function of one variable , which we used to fix one value @xmath54 .
if the spacetimes are equivalent , the other @xmath54 and @xmath58 should fall into place , and similarly if the spacetimes are different they should diverge .
we indeed see that the proper time and outer radii move into place , as would be expected in a convergent spacetime .
it is also clear that the @xmath1 line is not convergent , and the limiting spacetime experiences infinite curvature at that one point . to address the future of this singular event
, we can look at the spacetime of the limiting sequence of super- and subcritical collapses .
we now discuss the global structure of the critical spacetime .
our first question is the nature of the @xmath1 line in the course of the singular collapse event .
an observer moving at some finite radius @xmath54 sees the collapse of a scalar field to a point , and a rebound .
this is true for both the supercritical and subcritical collapse , since even when a black hole forms , the majority of the initial scalar field energy escapes , and the black hole only contains an ever smaller fraction of the initial mass as the parameter is tuned to criticality .
long after the field rebounds , the observer can measure the gravitational redshift to adjacent interior radii , and concludes that there is no redshift for most radii . in the supercritical scenario , there are significant redshifts at scales @xmath59 .
but for any fixed radius observer , as one takes the limit of @xmath60 , the sphere of influence of the ever diminishing black hole mass shrinks to zero , and the spacetime converges pointwise to minkowski at @xmath0 .
the convergence is not uniform , since for any given redshift difference @xmath61 between the observer and a fiducial interior radius , one can construct a radius @xmath62 inside which the redshift is larger than @xmath61 .
in the two limits @xmath63 and @xmath64 , the spacetimes converge pointwise everywhere except for the line @xmath1 .
a simple analogy is a singular weak field star .
consider a sequence of spacetimes with a single star of mass @xmath65 and radius @xmath66 .
the newtonian potential outside the star is @xmath67 for @xmath68 . to simplify the argument
, we make the potential continuous and constant interior to its radius @xmath69 . for small values of @xmath61 ,
the spacetime is in the weak field regime everywhere to order @xmath61 .
if we take a sequence of such space times with fixed @xmath61 and decreasing @xmath65 , the maximal curvature @xmath70 increases without bound .
there appears to be an illusory naked singularity in that limit .
we can ask what the limiting spacetime looks like as @xmath71 .
the obvious answer would be empty minkowski space .
the convergence to this space is pointwise , but not uniform . in the limit , @xmath72 .
@xmath73 converges to 0 pointwise , but not uniformly , so in the limit , @xmath74 everywhere except for @xmath1 , and it is undefined at that point .
this is a removable singularity since we can define @xmath75 , which is a unique and physically acceptable solution .
we suggest that the future of the singular collapse is likewise regular everywhere , with a removable singularity at @xmath1 in the future of the collapse event . the apparent paradox in the penrose diagram seen in fig .
[ fig : comp ] arises from the gauge choice at @xmath1 .
this one line is poorly defined in the limit , since it becomes singular .
the left panel describes the physical spacetime in the critical collapse limit , and the right one is related by a singular gauge change .
our analysis takes a very different approach from martin - garcia and gundlach @xcite .
these authors solve an elliptic equation satisfying an ansatz of discretely self - similar critical collapse .
we took a limit of a series of hyperbolic initial value problems .
those authors found a possibility of specifying new cauchy data in the elliptic solution , which is never an option for our evolutionary approach .
the qualitative solution for their unique regular extension looks similar to our critical spacetime .
we have implemented a characteristic code in double - null coordinates and used it to study the global spacetime structure of a critical collapse of a scalar field .
we reproduce the standard critical behavior and universal scaling .
we compare the limiting spacetime from the subcritical and the supercritical collapse limits .
these two limits appear qualitatively different . based on the numerical simulations , we conjecture that the two limits converge pointwise to the same spacetime , but not uniformly .
in particular , the @xmath1 line is not convergent , but all other points appear to converge pointwise .
the two apparently different solutions then only differ by a gauge change .
the apparent naked singularity in the upper limit obtained by the sequence of spacetimes with ever decreasing black hole mass becomes a removable singularity in the limit .
we conclude that the nature of the limiting critical collapse space time requires a careful definition of the order in which limits are taken , since the convergence is not uniform . for any collapse parameter @xmath2 within the critical value @xmath76
, one can find sufficiently small radii within which the supercritical and subcritical solutions differ .
conversely , for any fixed @xmath0 , one can find a value @xmath77 where the super and sub critical space times agree to some tolerance @xmath61 at radius @xmath54 .
there is a unique limit , with no new cauchy data that emits from the singular event .
we would like to thank david garfinkle for helpful comments and tara hiebert for interactions during the early stages of the project .
m. w. choptuik , _ universality and scaling in gravitational collapse of a massless scalar field _ , phys .
lett . * 70 * , 9 ( 1993 ) . c. gundlach , _ critical phenomena in gravitational collapse _ ,
rept . * 376 * , 339 ( 2003 ) [ gr - qc/0210101 ] .
w. israel , _ the formation of black holes in nonspherical collapse and cosmic censorship _ , canadian journal of physics * 64 * , 120 ( 1986 ) .
l. m. burko , _ the singularity in supercritical collapse of a spherical scalar field _ ,
d * 58 * , 084013 ( 1998 ) [ gr - qc/9803059 ] .
j. m. martin - garcia and c. gundlach , _ global structure of choptuik s critical solution in scalar field collapse _ , phys .
d * 68 * , 024011 ( 2003 ) [ gr - qc/0304070 ] .
d. garfinkle , _
choptuik scaling in null coordinates _ , phys .
d * 51 * , 5558 ( 1995 ) [ gr - qc/9412008 ] . c. gundlach and j. pullin , _ instability of free evolution in double null coordinates _
, class .
* 14 * , 991 ( 1997 ) [ gr - qc/9606022 ] .
s. hod and t. piran , _ fine - structure of choptuik s mass - scaling relation _ , phys .
d * 55 * , 440 ( 1997 ) [ gr - qc/9606087 ] . | we examine the global structure of scalar field critical collapse spacetimes using a characteristic double - null code .
it can integrate past the horizon without any coordinate problems , due to the careful choice of constraint equations used in the evolution .
the limiting sequence of sub- and supercritical spacetimes presents an apparent paradox in the expected penrose diagrams , which we address in this paper .
we argue that the limiting spacetime converges pointwise to a unique limit for all @xmath0 , but not uniformly .
the @xmath1 line is different in the two limits .
we interpret that the two different penrose diagrams differ by a discontinuous gauge transformation .
we conclude that the limiting spacetime possesses a singular event , with a future removable naked singularity . |
we have developed a model ( grasil , _ http://grana.pd.astro.it_ ) for the sed of galaxies including dust effects @xcite .
we consider a geometry more complex and realistic than previous treatments , and take into account 3 distinct environments where stellar radiation interacts with dust ( figure [ cartoon2f_mc ] ) : ( 1 ) * molecular clouds ( mcs ) + young stars * ( clumped component ) , ( 2 ) * diffuse ism ( cirrus ) + older stars * ( diffuse component ) , ( 3 ) agb envelopes ( included in the ssp library , bressan et al . , 1998 ) . therefore both stars and dust have a _ realistic 2-phase distribution_. the association between young stars and mcs is modelled to mimic the birth of stars in mcs and their gradual escape as they grow old : the fraction of each ssp energy radiated inside mcs is decreased with its age ( figure [ cartoon2f_mc ] ) .
it is the first time that such an _ age - dependent or selective extinction _ is modelled ( see granato et al .
, 2000 for interesting applications ) .
the dust model consists of normal grains in thermal equilibrium , small grains fluctuating in temperature and pah molecules .
we have extended our model to the radio .
the thermal radio emission is proportional to the production rate of ionizing photons : @xmath0 the non thermal radio emission is the sum of a dominating term , proportional to the supernova rate ( calibrated on the galactic value ) and one due to supernova remnants ( based on the @xmath1d relation ) : @xmath2 see bressan ( these proceedings ) and bressan et al .
( 2001 ) for details .
as an example , the fit to the starburst galaxy m82 is shown in figure [ m82r ] .
note that the inclusion of the radio emission potentially ( with the adopted calibration ) helps in constraining the starburst age , due to the different timescales involved in the ir ( ongoing sfr , last few myrs ) and radio ( recent sfr , a few @xmath3 yrs ago ) emission .
the sub - mm to radio spectral index has been recently used to estimate the redshift of sub - mm selected galaxies ( e.g. carilli & yun 2000 ) . exploiting our model it is possible to investigate the dependence of this index on the evolutionary status of galaxies . in figure [ qzarp ]
the @xmath4m-@xmath5 ghz spectral index is plotted vs redshift using the model fitting the sed of arp220 ( figure [ m82r ] , this seems to be representative of the scuba galaxies , e.g. barger et al . , 1999 ) , but for different ages of the starburt .
the redshift dispersion around z=2 ( the typical value estimated for scuba galaxies ) is @xmath6 . to reduce this dispersion
, we propose to use a radio - radio spectral index ( e.g. @xmath7-@xmath5 ghz ) , almost independent of z , in combination with the previous one . indeed , with the linear combination shown figure [ qzarp ] , the @xmath8 due to the evolution becomes 0.5 around z=2 .
recently we have started to include in our model the stellar contribution to the x - ray emission of galaxies .
our starting point has been the work by van bever & vanbeveren ( 2000 ) .
they find that the dominating stellar components in the x - ray for starburst galaxies are massive binaries with a black hole or a neutron star as a primary , and an ob star as a secondary either in the h - shell burning phase or on the main sequence , and pulsars .
we have included these components , and the contribution from supernova remnants , in the ssp library .
the main uncertainties in the predicted output are ( figure [ sspx ] ) : the binary fraction , the minimum mass for bh formation and the initial spin period @xmath9 of pulsars ( whose @xmath10 ) .
a further uncertainty is the sed to assigne to each component . a first attempt for the stellar x - ray sed expected from the uv - to - radio fitting model for m82 , is shown in figure [ sspx ] . | we present our model to interpret the sed of galaxies . the model for the uv to sub - mm
sed is already well established @xcite .
we remind here its main features and show some applications .
recently we have extended the model to the radio range @xcite , and we have started to include the x - ray emission from the stellar component . |
during the past few years , phase - coherent quasi - particle transport in hybrid superconducting structures has emerged as a new field of study , bringing together the hitherto separate areas of superconductivity and mesoscopic physics [ altshuler , lee and webb 1991 , buot 1993 ] .
the seeds of this new field of research were sown for a variety of different reasons .
the search for novel devices led several groups to embark on experimental programmes aimed at exploring the properties of hybrid semiconducting - superconducting structures . during the 1980s
there had been great progress in understanding transport properties of normal sub - micron conductors and therefore towards the end of that decade , it seemed natural to ask how these were affected by the presence of superconductivity .
partly as a consequence of formidable technological problems to be overcome in growing hybrid nanostructures , the first experimental results did not emerge until the summer of 1991 , when two groups [ petrashov and antonov 1991 , kastalskii et al 1991 ] independently submitted papers on phase - coherent transport in sub - micron superconducting structures .
the first of these reported an * increase * in resistance in metallic conductors due to the onset of superconductivity and the second the discovery of a zero - bias anomaly in the dc resistance of a superconductor - semiconductor ( s - sm ) contact .
the former is still the subject of theoretical investigation [ wilhelm et al 1997 , seviour et al 1997 ] , while the latter is now well - understood . in a phase - coherent normal - superconducting ( n - s ) structure ,
the phase of quasi - particles as well as cooper pairs is preserved and transport properties depend in detail on the quasi - particle states produced by elastic scattering from inhomogeneities and boundaries .
a key physical phenomenon , which arises in the presence of superconductivity is the possibility that an electron can coherently evolve into a hole and vice versa .
this phenomenon , known as andreev scattering [ andreev 1964 ] , occurs without phase breaking and is describable by a variety of theoretical techniques .
the effect of superconductivity on transport across a n - s interface is of course an old subject . in lowest order ,
the classical tunneling hamiltonian approach ignores andreev scattering and predicts that the dc conductance @xmath0 is proportional to the density of states .
later [ shelankov 1980 , blonder , tinkham and klapwijk ( btk ) 1982 , blonder and tinkham 1983 , shelankov 1984 ] it was pointed out that the contribution to the sub - gap conductance from andreev scattering can be significant and a theory of a clean n - i - s interface was developed , which showed that for a delta - function barrier , there is indeed a marked deviation from tunneling theory , but as the barrier strength is increased , the result of classical tunneling theory is recovered .
btk theory applies to a one - dimensional n - i - s system or , by summing over all transverse wavevectors , to 2 or 3 dimensional systems with translational invariance in the plane of the barrier and yields for the current @xmath1 through the contact @xmath2 where @xmath3 and @xmath4 are andreev and normal reflection coefficients listed in table ii of [ blonder et al 1982 ] , @xmath5 the fermi function and @xmath6 a measure of the area of the junction . in the presence of disorder or other inhomogeneities ,
this must be replaced by the more general expressions outlined in sections ii and iv below .
prior to 1991 , experiments on n - i - s point contacts have been in broad agreement with btk theory , exhibiting a conductance minimum at zero voltage @xmath7 and a peak at @xmath8 , where @xmath9 is the superconducting energy gap . in the experiment of kastalskii et al . [
1991 ] , the dc current through a nb - ingaas contact is measured as function of the applied voltage . at the interface , depending on the semiconductor doping level , a schottky barrier naturally forms so that the system behaves like a superconductor - insulator - normal ( s - i - n ) structure .
( an exception to this is inas , which does not form a schottky barrier at an n - s interface . ) according to btk theory , as the barrier strength increases the sub - gap conductance should vanish .
in contrast , the experiment revealed an excess sub - gap conductance at low bias , whose value was comparable with the conductance arising when the superconducting electrode is in the normal state .
this zero bias anomaly ( zba ) was later observed by nguyen et al . [
1992 ] in an experiment involving inas - alsb quantum wells attached to superconducting nb contacts and by using high transmittance nb - ag ( or al ) contacts of varying geometry , xiong et al . [
1993 ] were able to observe the evolution from btk to zba behaviour . in an experiment by bakker et al .
[ 1994 ] involving a silicon - based two - dimensional electron gas ( 2deg ) contacted to two superconducting electrodes , a gate voltage was also used to control the strength of the zba and in [ magnee et al 1994 ] an extensive study of the zba in nb / si structures was performed . since these early experiments , a great deal of effort has been aimed at observing andreev scattering in ballistic 2degs , including [ van wees et al 1994 , dimoulas et al 1995 , marsh et al 1994 , takayanagi and akazaki 1995(a ) , takayanagi , toyoda and akazaki 1996(a ) ] kastalskii et al [ 1991 ] attributed the excess conductance to a non - equilibrium proximity effect , in which superconductivity is induced in the normal electrode , giving rise to an excess pair current .
initially this phenomenon was seen as separate from andreev reflection , but subsequent theoretical developments have shown that the distinction between the proximity effect and andreev reflection is artificial . as will become clear later , the zba arises from an interplay between andreev scattering and disorder - induced scattering in the normal electrode .
andreev scattering is sensitive to the breaking of time reversal symmetry and as a consequence the conductance peak is destroyed by the introduction of a magnetic field .
zero bias anomalies constitute the first of a small number of paradigms of phase - coherent transport in hybrid n - s structures .
a second paradigm is the observation of re - entrant [ van wees et al 1994 , charlat et al 1996 ] and long - range [ courtois et al 1996 ] behaviour signalled by the appearance of finite - bias anomalies ( fbas ) in the conductance of high - quality n - s interfaces . at high temperatures @xmath10 and bias - voltages @xmath11 , where for a n - metal of length @xmath12 and diffusion coefficient @xmath13 , @xmath14 , both the zba and fba conductance peaks decay as @xmath15 and @xmath16 . for a clean interface there
also exists a conductance maximum at @xmath17 , @xmath18 and therefore at low - temperature and voltage a re - entrance to the low - conductance state occurs .
an interesting feature of this phenomenon is the long - range nature of the effect , which typically decays as a power - law in @xmath19 , where @xmath20 .
this behaviour is in sharp contrast with the exponential decay of the josephson effect and has been observed in a number of experiments . in a t - shaped ag sample with al islands at different distances from the current - voltage probes [ petrashov et al 1993(b ) ] , a long - range proximity effect was observed , in which the influence of the island extended over length scales greater than the thermal coherence length @xmath21 .
similar behaviour was also observed [ petrashov et al 1994 ] in ferromagetic - superconductor hybrids made from ni - sn and ni - pb . in an experiment involving a square cu loop in contact with 2 al electrodes
courtois et al [ 1996 ] clearly identified both the short- and long - range contributions to phase - coherent transport . in this interferometer experiment , they observed a phase - periodic conductance decaying as a power - law in @xmath22 , in parallel with a josephson current which decays exponentially with @xmath23 . the above re - entrance phenomenon is also observed in a third paradigm of phase - coherent transport , which arises when a normal metal is in contact with two superconductors , with order parameters phases @xmath24 and @xmath25 , whose difference @xmath26 can be varied by some external means .
prior to the experimental realisation of these structures , the electrical conductance of such andreev interferometers was predicted to be an oscillatory function of @xmath27 .
spivak and khmelnitskii [ 1982 ] and altshuler and spivak [ 1987 ] identified a high temperature @xmath28 , weak localisation contribution to the conductance of a disordered sample , whose amplitude of oscillation was less than or of order @xmath29 .
for an individual sample , the period of oscillation was found to be @xmath30 , but for the ensemble average a period of @xmath31 was predicted .
nakano and takayanagi [ 1991 ] and takagi [ 1992 ] examined a clean interferometer in one - dimension and again predicted a @xmath30-periodic conductance with an amplitude of oscillation less than or of order @xmath29 .
lambert [ 1993 ] examined a disordered conductor in the low - temperature limit @xmath32 and identified a new contribution to the ensemble averaged conductance with a periodicity of @xmath30 .
this @xmath30 periodicity is a consequence of particle - hole symmetry , which also guarantees that at zero temperature and voltage , the conductance should possess a zero phase extremum [ lambert 1994 ] .
prior to experiments on such devices , the generic nature of this prediction was confirmed in numerical simulations [ hui and lambert 1993(a ) ] encompassing the ballistic , diffusive and almost localised regimes and in a tunnelling calculation of the ensemble averaged conductance by hekking and nazarov [ 1993 ] .
the first experimental realisations of andreev interferometers came almost simultaneously from three separate groups . in march of 1994 , de vegvar et al [ 1994 ] showed results for a structure formed from two nb electrodes in contact with an al wire .
they found a small oscillation @xmath33 with a sample specific phase in the @xmath30 periodic component , suggesting that the ensemble averaged conductance should have a periodicity of @xmath31 , in agreement with spivak and khmelnitskii [ 1982 ] .
however in contrast with all subsequent experiments , no zero phase extremum was observed . in april /
may of that year , pothier et al [ 1994 ] produced an interferometer involving two tunnel junctions , which showed a @xmath30-periodic conductance , with a zero phase maximum and a low - bias , low - temperature amplitude of oscillation of order @xmath34 , which decayed with increasing temperature . in may 1994 ,
van wees et al [ 1994 ] , [ see also dimoulas et al 1995 ] produced the first quasi - ballistic inas 2deg interferometer , with high transparency n - s interfaces .
this experiment showed the first re - entrant behaviour in which the amplitude of oscillation @xmath35 varied from @xmath36 at zero voltage , ( where a minus sign indicates a zero phase minimum and a + sign a zero - phase maximum ) passes through zero at a bias of order 0.1mv , reaches a maximum at a bias of order @xmath17 and then decays to zero at higher voltages .
( for a detailed study see [ den hartog et al , 1996 ] ) . unlike the josephson current which decays exponentially with @xmath37 , these conductance oscillations decayed only as a power - law . the first experiment showing an amplitude of oscillation greater than @xmath29
was carried out by petrashov et al [ 1995 ] . here ,
silver or antimony wires in the shape of a cross , make two separate contacts with superconducting al and the phase difference between the contacts is varied using either an external field applied to a superconducting loop or by passing a supercurrent throught the al .
the amplitude was found to be @xmath38 for ag and @xmath39 for sb , and exhibited a periodicity of @xmath30 . in this experiment ,
the phase difference @xmath27 was varied both by passing a magnetic flux through an external superconducitng loop and by passing a supercurrent through a straight section of the superconductor , thereby emphasising that precise the manner in which the order parameter phase is controlled is not important .
these experiments were crucial in demonstrating that in metallic samples , the ensemble averaged conductance is the relevant quantity and therefore a quasi - classical description is relevant .
it is perhaps worth mentioning that with hindsight , an earlier experiment reporting large - scale oscillations in a sample with two superconducting islands [ petrashov et al 1993(a ) ] can be regarded a precursor to these interferometer experiments .
however the phase of the islands was not explicitly controlled , making an interpretation of the 1993 experiments more difficult .
a fourth and more recent paradigm of phase - coherent transport is the appearance of negative multi - probe conductances in structures where andreev transmission of quasi - particles is a dominant process [ allsopp et al 1994 ] .
the first experiments reporting this behaviour were carried out by hartog et al [ 1996 ] , using a diffusive inas 2deg . these probe individual coefficients in the current - voltage relations and demonstrate fundamental reciprocity relations arising from time - reversal and particle - hole symmetry .
finally a fifth paradigm is the suppression of electrical conductance by superconductivity in metallic systems without tunnel barriers .
experimentally this phenomenon has been observed in several structures , involving both non - magnetic ( ie silver ) and magnetic ( ie nickel ) n - components in contact a superconductor via clean interfaces . in the experiment of petrashov and antonov [ 1991 ]
the conductance of a ag wire on which several pb islands are deposited , decreases by 3% when the islands become superconducting , this corresponding to a conductance decrease of @xmath40 .
these experiments reveal two regimes : as the field decreases below @xmath41 , an initial resistance increase occurs .
the resistance remains field insensistive until a lower field is reached ( corresponding to the order of a flux quantum through the sample ) , at which point a further conductance suppression occurs . in the experiment by petrashov
et al [ 1993b ] involving a t - shaped ag sample with al islands at different distances from the current - voltage probes , a superconductivity induced change in the resistance by up to 30% was observed , but it was found that for different samples , the resistance could either increase or decrease . although superconductivity - induced conductance suppression in metallic samples was predicted a number of years ago [ hui and lambert , 1993(b ) ] these experiments at first sight appeared to conflict with quasi - classical theories , which universally predict that the normal - state , zero - temperature , zero - bias conductance @xmath42 is identical the conductance @xmath43 in the superconducting state .
this effect is addressed in [ hui and lambert 1993b , claughton et al 1995 , wilhelm et al 1997 , seviour et al 1997 ] .
the main aim of this review is to outline the quasi - classical and multiple scattering theories needed to describe the dc electrical conductance of phase - coherent hybrid n - s structures . for this reason we shall not discuss thermodynamic phenomena such as the josephson effect in any detail , despite the fact that ground - breaking experiments using clean [ takayanagi and akazaki , 1995(b , c , d ) ] superconducting quantum s-2deg - s point contacts show quantization of the critical current as predicted by furasaki et al [ 1991 ] and for shorter junctions by beenaker and van houten [ 1991 ] .
there are several notable theoretical papers addressing andreev scattering in such structures , including [ van wees et al 1991 , bagwell 1992 , furasaki et al 1994 , gusenheimer and zaikin 1994 , zyuzin 1994 , hurd and wendin 1994 and 1995 , bratus et al 1995 , chang et al 1995 , koyama , takane and ebisawa 1995 and 1996 , levy yeyati et al 1996 , martin - rodero 1996 , wendin and shumeiko 1996 , reidel et al 1996 , volkov and takayanagi 1996(b ) ] . similarly to restrict the length of this review , recent theories of thermoelectric coefficients
[ bagwell and alam 1992 , claughton and lambert 1996 ] and shot noise will not be discussed [ datta 1995 ] , nor will we discuss work on coulomb effects in superconducting islands [ eiles et al 1993 , lafarge et al 1993 , tuominen et al 1992 and 1993 , hergenrother et al 1994 , black et al 1996 , hekking et al 1993 ] . while andreev interference effects are generic phenomena , their manifestation in a given experiment is sensitive to many parameters . for the purpose of writing this review ,
it is therefore convenient to adopt a simple classification of experimental arrangements sketched in figure 1 .
the generic structure shown in figure 1a represents our first class of n - s - n hybrids and has many realisations . for convenience , we distinguish these from a second class of n - s hybrids of the kind shown in figure 1b and 1c , in which the superconductor effectively forms part of an external reservoir and is not simply part of the scattering region .
figure 1d indicates a third class of n - ss-n structures , involving two ( or more ) separate superconductors s and s@xmath44 , with respective order parameter phases @xmath27 , @xmath45 .
as noted above , transport properties of such andreev interferometers are periodic functions of the phase difference @xmath46 .
figure 1e shows a fourth class of s - n - s structures , in which two superconducting reservoirs are connected to a normal scattering region . in this case
, the structure forms a josephson junction and in contrast with all other structures shown in figure 1 , in the linear response limit , the dc conductance measured between the superconducting reservoirs is identically zero . in this case , the relevant dc quantity is the current - phase relation and the associated critical current .
clearly one could also measure the josephson current between the two superconductors in the n - ss-n structure of figure 1c and therefore the above classification is intended to label the measurement being made , rather than the device being measured .
structures of the form 1e will not be discussed . for the most part ,
the theoretical descriptions discussed below have finessed problems of self - consistency , by computing measured quantities as a function of the superconducting order parameter pairing field @xmath47 induced by making contact with a piece of superconductor . as an example , figure 1a shows a normal mesoscopic scattering region in contact with a superconducting island which plays the role of an externally controllable source of @xmath48 , in much the same way that the coils of a magnet are an external source of magnetic field .
the coils are not of primary interest and in many cases , neither is the superconductor .
it is assumed that parameters characterizing the superconductor are given and the key question is how does superconductivity influence transport through the scattering region .
of course once the influence of superconductivity is understood , transport properties can be used to probe the symmetry and spatial stucture of the order parameter , as suggested by cook et al . [
1996 ] . furthermore ,
in the presence of large currents which modify the order parameter , a complete self - consistent treatment is necessary , as described for example in [ bruder 1990 , hara et al 1993 , barash et al 1995 , canizres and sols 1995 , chang et al 1995 , gyorffy 1995 , martin and lambert 1995 and 1996 , riedel et al 1996 ] .
in this section we review the multi - channel current - voltage relations for a disordered phase - coherent scatterer connected to normal reservoirs , obtained for two normal probes by [ lambert 1991 ] and extended to multi - probes by [ lambert hui and robinson , 1993 ] .
the coefficients ( denoted @xmath49 and @xmath50 below ) are directly accessible experimentally and can be combined to yield a variety of transport coefficients , the simplest of which is the electrical conductance of a normal - superconducting interface . to avoid time - dependent order parameter phases varying at the josephson frequency , which would render a time - independent scattering approach invalid ,
these are derived under the condition that all superconductors share a common condensate chemical potential @xmath51 .
the derivation of the fundamental current - voltage relation presented in [ lambert 1991 ] follows closely the multi - channel scattering theory developed during the 1980s for non - superconducting mesoscopic structures [ bttiker 1986 , buot 1993 ] . in the normal state , this approach yields , for example , the multi - channel landauer formula [ landauer 1970 ] for the electrical conductance @xmath52 where @xmath53 is the transmission coefficient of the structure . historically the above formula was not accepted without a great deal of debate and as we shall see in the following section , contradicts the corresponding expression used by practitioners of quasi - classical theories , where the alternative expression @xmath54 is employed , with ( in one - dimension ) @xmath55 the reflection coefficient .
in fact the above two expressions refer to a two - probe and to a four - probe measurements , respectively , and the crucial lesson from the debate surrounding these equations is that transport coefficients such as the electrical conductance are secondary quantities .
more fundamental are the current - voltage relations describing a given mesoscopic structure . equations ( [ a1 ] ) and ( [ a2 ] ) are not valid in the presence of andreev scattering , because charge transport and quasi - particle diffusion are no longer equivalent . for example when a quasi - particle andreev reflects at an n - s interface , the energy and probability density of the excitation are reflected back into the normal conductor , whereas a charge of @xmath56 is injected into the superconductor . thus charge flows into the superconductor , even though the excitation does not and as a consequence , a current - voltage relation should be used , which takes into account this charge - energy separation . for a scattering region connected to @xmath12 normal reservoirs , labelled @xmath57 , it is convenient to write this in the form @xmath58 where @xmath59 is the current flowing from reservoir @xmath60 and the coefficients @xmath49 will be discussed in detail below . in the linear - response limit
, this reduces to @xmath61 the above expressions describe reservoirs at voltages @xmath62 , connected to a scattering region containing one or more superconductors with a common condensate chemical @xmath51 and relates the current @xmath63 from reservoir @xmath60 to the voltage differences @xmath64 , where @xmath65 .
the @xmath66 formula describes a wide variety of experimental measurements and underpins many subsequent theoretical descriptions of disordered n - s interfaces and inhomogeneous structures . for this reason , after discussing the relationship between the coefficients @xmath49 , @xmath50 and the scattering matrix , we shall examine the two - probe formula in some detail and illustrate its application to some generic experimental measurements .
the @xmath66 analysis of [ lambert 1991 ] is based on the observation that in the absence of inelastic scattering , dc transport is determined by the quantum mechanical scattering matrix @xmath67 , which yields scattering properties at energy @xmath68 , of a phase - coherent structure described by a hamiltonian @xmath69 . if the structure is connected to external reservoirs by open scattering channels labelled by quantum numbers @xmath70 , then this has matrix elements of the form @xmath71 .
the squared modulus of @xmath71 is the outgoing flux of quasi - particles along channel @xmath70 , arising from a unit incident flux along channel @xmath72 . adopting the notation of [ lambert , hui and robinson 1993 ] , we consider channels belonging to current - carrying leads , with quasi - particles labelled by a discrete quantum number @xmath73 ( @xmath74 for particles , @xmath75 for holes ) and therefore write @xmath76 , where @xmath77 labels all other quantum numbers associated with the leads . with this notation , the scattering matrix elements @xmath78
satisfy the unitarity condition @xmath79 , the time - reversibility condition @xmath80 and if @xmath68 is measured relative to the condensate chemical potential @xmath81 , the particle - hole symmetry relation @xmath82^*$ ] .
( for convenience we adopt the convention of including appropriate ratios of channel group velocities in the definition of @xmath83 to yield a unitary scattering matrix . ) for a scatterer connected to external reservoirs by @xmath12 crystalline , normal leads , labelled @xmath84 , it is convenient to write @xmath85 , @xmath86 , where @xmath87 is a channel belonging to lead @xmath88 . with this notation , the quantities entering the current - voltage relation are of the form @xmath89 , \label{a5}\ ] ] which is an expression for the coefficient for reflection ( @xmath90 ) or transmission ( @xmath91 ) of a quasi - particle of type @xmath92 in lead @xmath93 to a quasi - particle of type @xmath73 in lead @xmath60 . for @xmath94
, @xmath95 is an andreev scattering coefficient , while for @xmath96 , it is a normal scattering coefficient . since unitarity yields @xmath97 where @xmath60 and @xmath93 sum over all leads containing open channels of energy @xmath68 ,
this satisfies @xmath98 where @xmath99 is the number of open channels for @xmath73-type quasi - particles of energy @xmath68 in lead @xmath60 , satisfying @xmath100 .
similarly particle - hole symmetry yields @xmath101 and time reversal symmetry @xmath102 having introduced the scattering coefficients @xmath103 , the coefficients @xmath49 of the fundamental formula ( [ a3 ] ) are given by @xmath104 with @xmath105 + 1\}^{-1}$ ] the distribution of incoming @xmath73-type quasi - particles from lead @xmath93 .
, the following notation is employed @xmath106 , @xmath107 , @xmath108 etc . ]
equation ( [ a3 ] ) yields the current - voltage characteristics of a given structure at finite voltages , provided all scattering coefficients are computed in the presence of a self - consistently determined order parameter and self - consistent values of all other scattering potentials . at finite temperature , but zero voltage , where @xmath109 , it reduces to equation ( [ a4 ] ) , with @xmath50 given by @xmath110[n_i^+(e)\delta_{ij}-p_{ij}^{++}(e , h ) + p_{ij}^{-+}(e , h)]\label{a10},\ ] ] where @xmath5 is the fermi function and equation ( [ a8 ] ) has been used . at finite voltages , but zero - temperature , it reduces to @xmath111\label{a29}.\ ] ] finally at finite voltages , the differential of equation ( [ a3 ] ) with respect to @xmath112 ( with @xmath51 and all other potentials held constant ) yields @xmath113 where at finite temperature , @xmath114 -\sum_\beta p_{ij}^{\alpha\beta}(e , h ) [ -\beta{\partial f_j^\beta(e)\over\partial e}]\ } \label{a31}.\ ] ] and at zero temperature , @xmath115[\delta_{ij}n_i^+(e_i)+p_{ij}^{-+}(e_j , h)-p_{ij}^{++}(e_j , h ) ] \label{a32}.\ ] ] where where @xmath116 .
it is worth noting that replacing @xmath68 by @xmath117 and utilizing the particle - hole symmetry relation ( [ a8 ] ) allows equation ( [ a10 ] ) to be rewritten in the form @xmath110[n_i^-(e)\delta_{ij}-p_{ij}^{--}(e , h ) + p_{ij}^{+-}(e , h)]\label{a19},\ ] ] which demonstrates that particles and hole are treated on an equal footing in equations ( [ a10 ] ) and ( [ a19 ] ) .
furthermore in view of the symmetries ( [ a8 ] ) , ( [ a9 ] ) the reciprocity relation @xmath118 is satisfied .
while the above notation is convenient for arbitrary @xmath12 , it perhaps obscures the simplicity of the final result and therefore in the literature , several alternative notations have been employed . for the case of @xmath66 normal probes , where the scattering matrix has the structure @xmath119 it is convenient to write @xmath120 , @xmath121 , @xmath122 and @xmath123 . with this notation , the sub - matrices @xmath124
have the form @xmath125 the matrix @xmath126 ( @xmath127 ) is a matrix of amplitudes describing the reflection of @xmath92-type quasi - particles from reservoir 1 ( 2 ) into @xmath73-type quasi - particles travelling back into reservoir 1 ( 2 ) . similarly @xmath128 ( @xmath129 ) is a matrix of amplitudes describing the transmission of @xmath92-type quasi - particles from reservoir 1 ( 2 ) into @xmath73-type quasi - particles of reservoir 2 ( 1 ) . using these sub - matrices , the @xmath66 current - voltage relation can be written @xmath130 where @xmath131 with @xmath132 and @xmath133 similarly the zero temperature differential conductance ( [ a32 ] ) becomes @xmath115[\delta_{ij}n_i^+(e_i)+r_a(e_j)-r_0(e_j ) ] \label{a32a}.\ ] ] as noted in [ lambert 1991 ] , the two - probe current - voltage relation ( [ a15 ] ) can be used to derive generalisations of both ( [ a1 ] ) and ( [ a2 ] ) . to further illustrate the versatility of the two - probe theory
, we now apply it to some typical experimental measurements . for convenience
in what follows , we set @xmath29 equal to unity .
0.3 cm * example 1 . the normal limit .
* consider the structure of figure 1a , where a scattering region containing superconducting inclusions is connected to normal reservoirs at potentials @xmath134 , @xmath135 .
in the normal limit , the condensate potential @xmath136 must disappear from the fundamental current - voltage relation ( [ a3 ] ) .
as shown below , this occurs , because in the absence of andreev scattering , unitarity of the scattering matrix implies that @xmath137 .
hence equation ( [ a15 ] ) reduces to @xmath138 which is simply the landauer formula ( [ a1 ] ) .
0.3 cm * example 2 .
experiments where @xmath139 .
* figure 1b shows an experiment in which the superconductor and reservoir 2 are held at the same potential . in this case ,
equation ( [ a15 ] ) yields @xmath140 \label{a21}.\ ] ] this experimental configuration is of the type used in tunneling experiments , aimed at probing the proximity effect in the vicinity of an n - s boundary [ gueron et al 1996 ] .
this result describes any of the structures shown in figure 1 , provided @xmath141 and is a generalization to disordered and inhomogeneous structures of the boundary conductance formula derived by [ blonder , tinkham , klapwijk 1982 ] .
0.3 cm * example 3 .
experiments where @xmath142 . *
an example of such an experiment is shown in figure 1c , where under sub - gap conditions , the presence of a long superconductor ( of length greater than the superconducting coherence length ) prevents the transmission of quasi - particles from reservoir 1 to reservoir 2 and vice versa . in this case , combining the unitarity condition @xmath143 with equation ( [ a15 ] ) yields @xmath144 \label{a22}.\ ] ] in common with the current - voltage relation from which it derives , equation ( [ a22 ] ) is valid in the presence of disorder and inhomogeneities and in the presence of an arbitrary number of superconducting inclusions of arbitrary geometry .
0.3 cm * example 4 .
experiments where @xmath145 .
* such a situation is illustrated in figure 1a . in this case , inverting equation ( [ a15 ] ) yields @xmath146 where @xmath147 .
hence the two - probe conductance @xmath148 takes the form @xmath149 as an example of this formula , we note that in the zero - temperature limit , where all quantities are evaluated at zero energy , equation ( [ a24 ] ) can be written [ lambert 1993 , lambert , hui and robinson 1993 ] @xmath150 for a symmetric scatterer , where primed quantities equal unprimed quantities , this reduces to @xmath151 , whereas in the absence of transmission between the reservoirs , the resistance @xmath152 reduces to a sum of two resistances @xmath153 .
it should be noted that a combination of particle - hole symmetry and unitarity yield at @xmath154 , @xmath155 and therefore equation ( [ a25 ] ) is symmetric under an interchange of primed and unprimed coefficients .
0.3 cm * example 5 .
experiments where @xmath156 . * as a final example , consider the experiment sketched in figure 1d where reservoir 2 acts as a voltage probe , with @xmath157 . in this case equation ( [ a15 ] ) yields for the ratio of the voltages @xmath158 where from equation ( [ a16 ] ) , the coefficient @xmath159 is positive . in contrast , the coefficient @xmath160 is necessarily negative for a normal system , but in the presence of andreev scattering can have arbitrary sign .
hence superconductivity can induce voltage sign - reversals which are not present in the normal limit .
this feature was first predicted within the context of negative four - probe conductances [ allsopp et al 1994 ] and has been confirmed in recent experiments by the groningen group [ hartog et al 1996 ] .
the above formulae relate measurable quantities to scattering matrix elements and therefore to end this section we briefly introduce the bogoliubov - de gennes equation [ de gennes 1989 ] , which forms a basis for computing the scattering matrix @xmath83 .
the bogoliubov - de gennes equation arises during the diagonalization of the mean - field bcs hamiltonian , which for a non - magnetic , spin - singlet superconductor takes the form @xmath161 where @xmath162 is a constant , @xmath163 and @xmath164 are field operators , destroying and creating electrons of spin @xmath165 at position @xmath166 and @xmath167 to each positive eigenvalue @xmath168 of @xmath69 , with eigenvector @xmath169 satisfying @xmath170 there exists a corresponding negative eigenvalue @xmath171 with eigenvector @xmath172 .
consequently @xmath173 is diagonalized by the transformation @xmath174 where to avoid overcounting , only one of @xmath175 or @xmath176 is included in the sum over @xmath70 .
it should be noted that whereas the bogoliubov - de gennes equation ( [ a38 ] ) refers to a model in which the variable @xmath166 varies continuously , in a tight - binding model , often used in numerical simulations , the corresponding bogoliubov - de gennes equation is @xmath177 where @xmath178 ( @xmath179 ) indicates the particle ( hole ) wavefunction on site @xmath60 and @xmath180 labels a neighbour of @xmath60 .
as discussed in [ hui and lambert 1990 ] and [ lambert , hui and robinson 1993 ] , for a scattering region connected to two crystalline normal leads , the bogoliubov - de gennes equation may be solved by means of a transfer matrix method , which yields a transfer matrix @xmath181 satisfying @xmath182 where @xmath183 @xmath184 ) refer to vectors of outgoing ( incoming ) plane - wave amplitudes on the left and @xmath185 , ( @xmath186 ) to corresponding amplitudes on the right , each plane - wave being divided by the square root of its longitudinal group velocity to ensure unitarity of @xmath83 . whereas @xmath181 connects plane wave amplitudes in the left lead to amplitudes in the right lead , the @xmath83-matrix connects incoming amplitudes to outgoing amplitudes and satisfies @xmath187 once @xmath181 is known , the s - matrix can be constructed .
indeed if @xmath83 is written as @xmath188 then @xmath181 has the form @xmath189 from which the following inverse relation is obtained , @xmath190 an alternative method of evaluating coefficients in the above current - voltage relations is provided by the recursive green s function method , which uses gaussian elimination to compute the green s function on sites located at the surface of the interface between external normal leads and the scatterer . given the surface green s function , scattering coefficients can be obtained from generalised fisher - lee relations [ fisher and lee 1982 ] , derived by [ takane and ebisawa 1992(a ) ] and later rederived in [ lambert 1993 , lambert , hui and robinson 1993 ] .
this recursive technique is identical to the decimation " method employed by [ lambert and hui 1990 ] and is essentially an efficient implementation of gaussian elimination .
a third method of evaluating the coefficient @xmath191 was derived by [ beenakker 1992 ] for the case where there is perfect andreev reflection at the boundary of a clean superconductor .
the resulting formula expresses @xmath191 in terms of scattering properties of the normal state , as discussed in section v below and facilitates the application of random matrix theory to n - s structures [ beenakker 1997 ] .
in the previous section , the multiple scattering approach to dc transport was reviewed and in the following section the method of quasi - classical green s function will be discussed . whereas the multiple scattering method is a stand - alone approach , the quasi - classical technique must be supplemented by boundary conditions describing quasi - particle scattering at the interface between different metals .
these boundary conditions are provided by scattering theory and therefore provide a bridge between the two approaches . in the literature , quasi - classical boundary conditions have not been discussed using the language of modern scattering theory and as a consequence are restricted to planar surfaces with translational invariance in the transverse direction . in this section we offer a rederivation of zaitsev s boundary conditions [ zaitsev 1984 ] , which not only fills this gap , but also yields a more general condition applicable to non - planar interfaces . as noted in section 2 ,
the starting point for a multiple scattering description is the scattering matrix @xmath83 , with matrix elements @xmath192 connecting incoming to outgoing scattering channels , and satisfying equation ( [ 3.1 ] ) .
clearly @xmath193 is a functional of all physical potentials entering the hamiltonian @xmath69 , as well as a function of e. since @xmath69 is hermitian , quasi - particle probability is conserved , which yields @xmath194 and @xmath195 furthermore time reversal symmetry yields @xmath196 and @xmath197 hence @xmath198 and @xmath199 it should be noted that the the unitary matrix @xmath83 in equation ( [ 3.1 ] ) and the matrix @xmath181 in equation ( [ 3.2 ] ) connects open channels to open channels only .
if evanescent states which decay at large distances from the scatterer are included in the outgoing states on the left - hand - side of ( [ 3.1 ] ) and states which grow at large distances are included on the right , then the corresponding scattering matrix @xmath200 and transfer matrix @xmath201 will not satisfy equations ( [ 3.6 ] ) and ( [ 3.7 ] ) .
the matrices @xmath83 and @xmath181 should then be constructed by eliminating closed channels from @xmath200 and @xmath201 . using the above definitions , the boundary conditions derived by zaitsev are readily extended to the case of multiple scattering channels .
the problem to be solved is that of connecting the green s function @xmath202 in a region @xmath93 to the green s function(s ) in all other regions connected to the scatterer .
although the following analysis is easily generalized to many leads attached to a scattering region , for simplicity of notation we restrict the discussion to only two leads . for the purpose of this section ,
it is convenient to write the channel index @xmath70 in the form @xmath203 , where @xmath204 labels a lead , @xmath205 labels all tranverse quantum numbers , @xmath206 for a right - going ( or right - decaying ) wave and @xmath207 for a left - going ( or left - decaying wave ) .
if @xmath208 denotes a position in a crystalline lead @xmath93 of constant cross - section , then a plane wave of unit flux incident along channel @xmath70 can be written @xmath209 where @xmath210 is the transverse mode associated with channel @xmath70 and @xmath211 is a plane wave of unit flux , with longitudinal wavevector @xmath212 . similarly for @xmath213 , and @xmath214 , @xmath208 located within the leads
, the green s function has the form @xmath215 } \over{\sqrt{v_{(i,\sigma , p)}(e)v_{(j,\sigma',p')}(e ) } } } , \label{3.13}\ ] ] where @xmath216 is equal to the sign of @xmath217 . for @xmath213 and a fixed value of @xmath216 ,
the green s function is simply a wavefunction and therefore the matrix @xmath218 with matrix elements @xmath219 satisfy relations of the form of equation ( [ 3.2 ] ) .
first consider the form of the green s function when @xmath220 and @xmath221 . when viewed as a function of @xmath222 , for @xmath223 ,
one has @xmath224 on the other hand , when viewed as a function of @xmath225 , using the conjugate equation for the green s function ( which involves the time reversed hamiltonian , one obtains for @xmath226 and @xmath227 , @xmath228 hence after eliminating the off - diagonal terms , @xmath229 , etc .
, we obtain the general boundary condition relating the green s functions on the left of the scatterer , to the green s functions on the right : @xmath230 equation ( [ 3.16 ] ) is a generalisation of zaitsev s boundary condition to the case of a non - planar barrier , which in general may contain impurities and break time - reversal symmetry .
yields an identical result and therefore the boundary condition is independent of the choice of @xmath216 . ]
since the sum on the right - hand - side of equation ( [ 3.13 ] ) includes evanescent channels , the general boundary condition involves the matrix @xmath201 .
however at large distances from the scatterer , evanescent channels can be ignored and therefore @xmath201 can be replaced by @xmath181 . in view of equation ( [ 3.11 ] ) , if time - reversal symmetry is present , the boundary condition simplifies to @xmath231 in the work of zaitsev , where @xmath205 represents the transverse momentum and all matrices are diagonal , equation ( [ 3.17 ] ) is satisfied for each separate @xmath205 and in the notation of zaitsev has the form @xmath232 as an example , taking the trace of equation ( [ 3.18 ] ) yields ) by @xmath233 and exploit the ciclyc property of the trace . ] @xmath234 which demonstrates that the anti - symmetric part of the quasi - classical green s function is continuous across a scattering region . more generally , equation ( [ 3.17 ] ) yields the corresponding relation @xmath235 similarly generalisations of all other boundary conditions can be expressed as traces over the matrices @xmath236 . indeed multiplying both sides of ( [ 3.17 ] ) by each of the matrices
@xmath237 and taking the trace of the resulting four equations yields @xmath238= -{\rm tr}[t^{-1}_{22}t_{21}a^{11}_{++}+t_{12}t^{-1}_{22}a^{22}_{-- } ] = { \rm tr}[ra^{11}_{++}-r'a^{22}_{-- } ] \label{3.21}\ ] ] @xmath239= -{\rm tr}[t^{-1}_{11}t_{12}a^{11}_{--}+t_{21}t^{-1}_{11}a^{22}_{++ } ] = { \rm tr}[r^\dagger a^{11}_{--}-{r'}^\dagger a^{22}_{++ } ] \label{3.22}\ ] ] @xmath240= { \rm tr}[-t^{-1}_{12}t_{11}a^{11}_{++}+t_{22}t^{-1}_{12}a^{22}_{++ } ] = { \rm tr}[{r^\dagger}^{-1}a^{11}_{++}+{r'}^{-1}a^{22}_{++ } ] \label{3.23}\ ] ] @xmath241= { \rm tr}[-t^{-1}_{21}t_{22}a^{11}_{--}+t_{11}t^{-1}_{21}a^{22}_{-- } ] = { \rm tr}[r^{-1}a^{11}_{--}+{r^{\prime\dagger}}^{-1}a^{22}_{-- } ] \label{3.24}\ ] ] subtracting ( [ 3.21 ] ) from ( [ 3.22 ] ) yields , in view of ( [ 3.20 ] ) , @xmath242= { \rm tr}[r'a^{22}_{-- } -{r'}^\dagger a^{22}_{++ } ] \label{3.25}\ ] ] and subtracting ( [ 3.23 ] ) from ( [ 3.24 ] ) yields @xmath243 = { \rm tr}[{r^{\prime\dagger}}^{-1}a^{22}_{--}-{r'}^{-1}a^{22}_{++ } ] \label{3.26}\ ] ] adding ( [ 3.21 ] ) to ( [ 3.22 ] ) and ( [ 3.23 ] ) to ( [ 3.24 ] ) yields @xmath244- { \rm tr}[a^{22}_{+-}+a^{22}_{-+}]= { \rm tr}[ra^{11}_{++ } + r^\dagger a^{11}_{-- } ] -{\rm tr}[r'a^{22}_{--}+ { r'}^\dagger a^{22}_{++ } ] \label{3.27}\ ] ] and @xmath244 + { \rm tr}[a^{22}_{+-}+a^{22}_{-+ } ] = { \rm tr}[{r^\dagger}^{-1}a^{11}_{++}+r^{-1}a^{11}_{-- } ] + { \rm tr}[{r'}^{-1}a^{22}_{++ } + { r^{\prime\dagger}}^{-1}a^{22}_{-- } ] \label{3.28}.\ ] ] for the translationally invariant case considered by zaitsev , writing @xmath245 @xmath246 , @xmath247 , @xmath248 and @xmath249 , one obtains from either of equations ( [ 3.25 ] ) and ( [ 3.26 ] ) @xmath250 from ( [ 3.27 ] ) and ( [ 3.28 ] ) , @xmath251 @xmath252 and from ( [ 3.19 ] ) @xmath253 where @xmath254/2,\ ] ] @xmath255/2\ ] ] and @xmath256/2.\ ] ] equations ( [ 3.29 ] ) and ( [ 3.32 ] ) , which are a limiting case of the more general relation ( [ 3.17 ] ) are identical to the boundary conditions of zaitsev and will be discussed in the next section .
the quasi - classical approach to the theory of superconductivity , initiated by eilenberger [ 1968 ] and larkin and ovchinnikov [ 1968 ] and developed by several authors [ usadel 1970 , eliashberg 1971 , larkin and ovchinnikov 1973,1975 ] has been largely used to analyse transport phenomena in dirty hybrid systems .
this approach can be used to study thermodynamic and kinetic properties of superconductors , whose dimensions significantly exceed the fermi wavelength @xmath257 . for the purpose of reviewing recent theoretical work , we briefly summarize the quasi - classical green s function approach , following mainly [ larkin et ovchinnikov 1986 ] .
there exist in the literature several excellent reviews on the subject including [ rammer and smith 1986 ] and we refer the reader to these for a more detailed exposition .
the derivation of the equation for quasi - classical green s function starts from the dyson equation for the matrix green s function @xmath258 in the non equilibrium keldysh formalism [ keldysh 1964 ) ] .
@xmath259 where @xmath260 following standard notation , we indicate with a `` hat '' matrices in nambu space and define retarded , advanced and keldysh green s functions @xmath261 @xmath262 @xmath263 where @xmath264 @xmath265 in these expressions , @xmath266 and @xmath267 , which allow equation ( [ 4.1 ] ) to be written in the less compact form @xmath268 the equation conjugate to equation ( [ 4.1 ] ) is @xmath269 and taking the difference between equations([4.1 ] ) and ( [ 4.8 ] ) yields @xmath270=0 . \label{4.9}\ ] ] this equation is simplified by going to the center - of - mass and relative coordinates ( @xmath271 ) and ( @xmath272 ) defined as @xmath273 fourier transforming with respect to @xmath274 and @xmath275 , and introducing the quasi - classical green s function defined by @xmath276 where @xmath277 is the energy measured with respect to the fermi level and @xmath278 is the unit vector in the direction of the momentum @xmath279 . on the assumption that the self - energy depends weakly on the energy @xmath280 , one sets @xmath281 in @xmath282 , which yields @xmath283 + i\left [ { \check { \sigma } } , { \check g}\right]=0 .
\label{4.11}\ ] ] where the curly brackets denote the anticommutator .
this is the equation for the quasiclassical green s function first derived by eilenberger [ 1968 ] .
@xmath284 is a block - diagonal matrix with diagonal block entries @xmath285 . in deriving equation ( [ 4.8 ] ) , the non - homogenous term present in the dyson equation ( [ 4.1 ] ) was eliminated and as a consequence , equation ( [ 4.11 ] ) determines @xmath286 only up to a constant .
as discussed in [ shelankov 1985 , zaitsev 1984 ] , a useful normalization condition for @xmath286 is @xmath287 in terms of the quasiclassical green s function , the physical current is given by @xmath288 where @xmath289 is the free single - particle density of states per spin , @xmath290 the usual pauli matrix and @xmath291 the fermi velocity . here
@xmath292 indicates the average over the angle @xmath293 formed by @xmath278 and the direction of @xmath294 , i.e. , @xmath295 .
it should be noted that , because of the angular average , only the antisymmetric part of @xmath296 enters the expression for the current . in the dirty limit , in the case of an isotropic scattering potential with elastic scattering time @xmath297
, the effect of non - magnetic impurities can be described by the self - energy @xmath298 and equation ( [ 4.11 ] ) can be considerably simplified [ usadel 1970 ] . in this case one
expands @xmath286 in spherical harmonics keeping only the s- and p - wave terms @xmath299 with @xmath300 and @xmath301 not depending on @xmath302 and @xmath303 . by inserting equation ( [ 4.15 ] ) in equation (
[ 4.11 ] ) , @xmath301 is expressed in terms of @xmath300 @xmath304 and for @xmath300 one obtains a diffusion - like equation @xmath305 -\partial_t \lbrace { \check { \tau}}_z , { \check g}_0 \rbrace=0 \label{4.17}\ ] ] with @xmath306 the diffusion coefficient and @xmath307 the mean free path .
eqs.([4.11 ] ) and ( [ 4.17 ] ) are valid for bulk systems and in the presence of boundaries , must be supplemented by boundary conditions which connect the green functions evaluated in different regions .
these boundary conditions have turned out to be crucial to the recent development of transport theory in hybrid structures . for this reason
, we present a brief discussion of these in the following subsection . in this subsection
, we discuss the boundary conditions for the quasiclassical green s function [ zaitsev 1984 ] , given by equations ( [ 3.29 ] ) to ( [ 3.32 ] ) of section iii . the antisymmetric functions @xmath308 and @xmath309 are continuous across the boundary , while the symmetric ones @xmath310 and @xmath311 experience a jump determined by the transparency of the barrier .
the size of the jump vanishes for perfectly transmitting interfaces .
the function @xmath286 is the quasi - classical green s function satisfying equation ( [ 4.11 ] ) , whereas the green s function @xmath312 arises from reflected waves at the boundary .
it is therefore necessary to derive an equation of motion for @xmath312 to be used together with equation ( [ 4.11 ] ) .
we will not show this derivation here , because in actual calculations , one only uses the green s function @xmath286 .
zaitsev has shown , that the final result is of the form @xmath313 which must be used in conjunction with the normalization condition @xmath314 the boundary conditions for the antisymmetric components are already decoupled with respect to the @xmath315 and @xmath316 . to decouple the symmetric components , we express eqs.([4.46]-[4.47 ] ) in terms of the symmetric and antisymmetric parts to yield @xmath317 and @xmath318 by manipulating equation ( [ 4.48 ] ) one obtains @xmath319 which is the extra condition to be used together with equations ( [ 3.30]-[3.31 ] ) . by means of equations ( [ 3.30]-[3.31 ] )
one expresses @xmath320 and @xmath321 in terms of @xmath322 and @xmath323 and substitute in equation ( [ 4.50 ] ) so that the final boundary condition reads @xmath324=(t/4)({\check g}_s^2{\check g}_s^1-{\check g}_s^1 { \check g}_s^2 ) \label{4.51}\ ] ] where @xmath325 , @xmath326 are the reflection and transmission coefficient of the barrier . as a simple example
, we consider the normal case , when the green s function is a two - by - two matrix @xmath327 with @xmath328 , @xmath329 being the usual distribution function entering the boltzman kinetic equation . by observing that @xmath330=4\left ( \begin{array}{c c } 0 & f_{1s }
-f_{2s } \\ 0 & 0 \\ \end{array } \right);~~ ( { \check g}_1-{\check g}_2)^2={\check 0};~~ { \check g}_{a}^2={\check 0 } \label{4.53}\ ] ] the boundary condition ( [ 4.51 ] ) assumes the form @xmath331 it is worth noticing that this result could have been obtained directly from simple counting arguments of the form @xmath332 in view of equation ( [ 4.13 ] ) for the current , equation ( [ 4.54 ] ) yields for the conductance of a tunnel junction in the normal case , @xmath333 where @xmath334 . in the general case ,
equation ( [ 4.51 ] ) can be simplified in the low transparency limit . if @xmath335 , then @xmath336 and the boundary condition equation ( [ 4.51 ] ) reduces to @xmath337 . \label{4.57}\ ] ] in the dirty limit , @xmath338 and equations ( [ 4.16]-[4.17 ] ) yield the boundary condition [ kuprianov and lukichev 1988 ] @xmath339 which after multiplying equation ( [ 4.57 ] ) by @xmath51 and taking the angular average yields @xmath340 . \label{4.59}\ ] ] by defining a conserved `` super '' current @xmath341 ( cf . equation ( [ 4.13 ] ) ) one finally arrives at the following boundary condition @xmath342 , \label{4.61}\ ] ] where @xmath343 is the drude electrical conductivity .
equation ( [ 4.61 ] ) is the desired boundary condition to be used together with the diffusion equation ( [ 4.17 ] ) in the presence of boundaries .
eq.([4.61 ] ) is strictly valid in the case of small barrier transparency .
such restriction has been recently somehow relaxed by lambert , raimondi , sweeney and volkov [ 1997 ] . in the next subsection
we show how to compute the current - voltage relation of the two fundamental elements of an hybrid system , namely a tunnel junction and a diffusive region . *
( i ) a tunnel junction . * consider the current through a tunnel junction as given by the commutator on the right - hand side of equation ( [ 4.61 ] ) .
the physical current is obtained from the keldysh component of equation ( [ 4.61 ] ) @xmath344_k = { \hat g}_2^r { \hat
g}_1+{\hat g}_2{\hat g}_1^a- { \hat g}_1^r{\hat g}_2-{\hat g}_1{\hat g}_2^a\ ] ] where we have dropped the subscript `` s '' .
the normalization condition , @xmath345 , allows us to choose @xmath346 where the matrix @xmath347 can be taken to be diagonal [ larkin and ovchinnikov 1975 ] @xmath348 as a result , multiplying by @xmath349 and taking the trace , yields ( cf . equation ( [ 4.13 ] ) ) @xmath350 where @xmath351\ ] ] @xmath352\ ] ] @xmath353\ ] ] @xmath354.\ ] ] due to the normalization condition
@xmath355 , we have @xmath356 where @xmath357 , @xmath358 and @xmath27 is the phase of the superconducting order parameter .
hence @xmath359 where @xmath360\ ] ] and @xmath361 ( f^z_1 -f^z_2).\ ] ] in equation ( [ 4.63 ] ) , @xmath362 is the josephson current , while @xmath363 is sometimes referred to as quasi - particle and interference current . to clarify equation ( [ 4.63 ] ) , consider the case where the two regions on the left and right of the barrier are at equilibrium and @xmath364 , so that the josephson current vanishes .
if the distribution functions @xmath365 have their equilibrium form , @xmath366 then the current through the junction becomes @xmath367 where @xmath368 at @xmath369 , this reduces to @xmath370 equation ( [ 4.64 ] ) shows that conductance of the tunnel junction is renormalized by a term @xmath371 which depends on the amount of superconducting pairing on the two sides of the junction . in the limit of normal systems , @xmath372 , and
one recovers the conductance of the normal state ( cf .
equation ( [ 4.56 ] ) ) . 0.3 cm * ( ii ) a diffusive region . * 0.3 cm we derive now the equivalent of equation ( [ 4.64a ] ) for a diffusive region in the absence of inelastic processes . in this case
the equation for the `` hat '' components of the green s function ( [ 4.17 ] ) reads @xmath376- ( { \hat g}^r\partial_{\bf r } { \hat g}^r ) ( \partial_{\bf r } { \hat f})- ( \partial_{\bf r } { \hat f})({\hat g}^a\partial_{\bf r } { \hat g}^a ) = 0.\ ] ] the last equation yields @xmath347 , once @xmath377 and @xmath378 have been determined .
choosing @xmath379 and @xmath380 , yields , after multiplying by @xmath349 and taking the trace , @xmath381=0 . \label{4.65}\ ] ] if we consider a diffusive region of length @xmath12 , the distribution function depends only on the longitudinal coordinate @xmath382 and @xmath383 has the form @xmath384 where @xmath385 using the formula for the physical current ( [ 4.13 ] ) one obtains @xmath386 to use equation ( [ 4.66b ] ) one has to solve the equation for @xmath377 . since @xmath387 , one can write @xmath388 and @xmath389 , which yields @xmath390 equation ( [ 4.67 ] ) is a non - linear equation , for which approximate analytical [ zaitsev 1990 , volkov et al . 1993 , volkov 1994 , zaitsev 1994 ] and numerical [ zhou et al , 1995 , yip 1995 , nazarov and stoof 1996 , golubov et al 1997 ] solutions have been obtained . for the purposes of the present discussion
, we note that at zero energy , the solution may be obtained easily and takes the form @xmath391 since @xmath392 the current ( [ 4.66b ] ) becomes @xmath393 where we have used the fact that for @xmath394 , @xmath395 .
the values @xmath396 and @xmath397 are determined by attaching the diffusive region to electrodes which are assumed to be in equilibrium .
for a normal system @xmath398 , while for a superconducting one @xmath399 and @xmath400 . as a result
the resistance of a diffusive region does not depend on the nature of the electrodes .
nazarov recognized [ nazarov 1994 ] that the simplicity of the results of equations ( [ 4.64],[4.68 ] ) can be generalized to an arbitrary circuit with normal and superconducting electrodes .
in fact , the evaluation of the conductance can be reduced to the use of a set of simple rules , which we now state .
a circuit is made of dispersive elements ( diffusive regions and tunnel junctions ) connected to one another and to the electrodes by nodes . to each circuit element
is associated a `` scalar '' current , @xmath93 , which is just the physical current and a `` vector '' current @xmath401 . to each node
is associated a `` spectral '' unit vector @xmath402 .
the vector current flowing through a dispersive element depends on the spectral vectors at the two ends .
it is useful to imagine the spectral vector as the spherical coordinate in the northern hemisphere .
the rules are : \i ) conductance in the presence of superconducting electrodes is the same as in normal circuits , with renormalized tunnel junction conductances .
the renormalization factor is given by the scalar product of the spectral vectors at the two sides of the junction .
\ii ) spectral vector of a normal electrode is at the north pole ( @xmath403 ) , while that of a superconducting electrode lies on the equator , with the longitude given by the phase of the order parameter .
\iii ) vector current is perpendicular to both the spectral vectors at the ends of a given element element .
if @xmath27 is the angle between the spectral vectors at the two ends of a given elements element , the magnitude of the vector current is @xmath404 and @xmath405 for diffusive and tunnel junction elements , respectively .
\iv ) vector current is conserved at each node ( generalized kirchoff rule ) .
these rules will be applied to a n - i - s interface in the following section .
in this section , we show how the multiple - scattering approach and quasi - classical theory can be used to describe the simplest example of phase - coherent structure , namely a n - s interface .
the simplest problem of a one - dimensional n - i - s system was examined by [ blonder , tinkham , and klapwijk 1982 ] .
this solution is also valid in higher dimensions , provided there is translational invariance in the direction perpendicular to the electronic motion . to see this ,
imagine a two - dimensional system with a finite width in y direction and infinite in the x direction .
the motion along y is quantized , and for fixed @xmath406 , the problem becomes one - dimensional and the bogolubov - de gennes equations reduces to @xmath407 where @xmath408 is the effective chemical potential in a given channel .
for a pairing potential with a step - like spatial variation of the form @xmath409 the bogolubov - de gennes equations can be solved by matching wave functions . for @xmath410
the solution @xmath411 for an incident electron from the left is @xmath412 and for @xmath413 @xmath414 the coherence factors @xmath415 and @xmath136 correspond to a particle - like excitation of energy @xmath68 and momentum @xmath416 , while @xmath417 and @xmath418 correspond to a hole - like excitation at the same energy and with momentum @xmath419 , where @xmath420 and @xmath421 ^ 2+\delta ^2=[(\hbar ^2/2 m ) \overline{q}^2+\overline\mu ] ^2+\delta ^2.$ ] also @xmath422 with @xmath423 , @xmath424 , where @xmath425 and @xmath426 for @xmath427 , while for @xmath428 we have @xmath429 to solve for the scattering coefficients @xmath430 , @xmath431 , and @xmath432 one needs matching conditions at the interface @xmath433 .
these are the usual continuity condition for @xmath411 and their derivatives and read @xmath434 where the term in @xmath435 in the last two equations of ( [ 5.7 ] ) allows for a delta - like potential @xmath436 in order to reproduce the btk model of an n - i - s interface .
the system ( [ 5.7 ] ) completely determines the four scattering coefficients .
the detailed solution is of course straightforward , but the resulting expressions look rather cumbersome , so it is convenient to resort to andreev s approximation of setting @xmath437 , which amounts to neglecting terms of the order @xmath438 writing @xmath439 yields @xmath440 for @xmath441 , where @xmath442 , this yields @xmath443 and in view of unitarity , @xmath444 . for @xmath445
one obtains @xmath446 the above treatment of a planar interface has been extended by a number of authors , including [ riedel and bagwell 1993 ] . in one dimension and at zero temperature
, the boundary conductance of equation ( [ a21 ] ) reduces to the btk result @xmath447 which for @xmath441 becomes @xmath448 , while for @xmath449 , @xmath450 and one recovers the landauer expression ( 2 ) , @xmath451 .
it is interesting to compare the sub - gap conductance @xmath452 with that in the normal state @xmath453 this allows one to express @xmath43 in terms of the dimensionless normal state conductance @xmath454 : @xmath455 which is a consequence of the fact that the superconducting electrode in this approximation only enters as a boundary condition . for a disordered n - region in contact with a superconductor the result ( [ 5.14 ] ) can be extended to higher dimensions by introducing the eigenvalues @xmath456 of the transmission matrix @xmath457 of the n - region and summing equation ( [ 5.14 ] ) over all @xmath70 [ beenakker 1992 ] , to yield equation ( [ 5.19 ] ) below . for @xmath458 , equation ( [ 5.11 ] )
predicts that the low - energy conductance is suppressed by a factor @xmath459 , as compared to the @xmath460 case .
already for values of @xmath461 this implies an almost vanishing sub - gap conductance . at the gap energy ,
the expression for @xmath191 achieves a peak value of unity within andreev approximation , with @xmath191 decaying to zero for energies above the gap .
this picture clearly resembles that of the tunneling approach , where the conductance is controlled by the supeconducting density of states , but is at variance with the experiment of [ kastalskii et al .
it was soon recognized that the low - energy peak observed by kastalskii et al was due to andreev scattering [ van wees , de vries , magne , and klapwijk , 1992 ] , but it was not possible to explain it using the theory discussed above , because the interplay between scattering due to disorder in the normal region and andreev scattering at the n - s interface is crucial .
the first insight into the zba was based on a detailed description using quasi - classical green function methods [ zaitsev 1990 , volkov and klapwijk 1992 , volkov , klapwijk and zaitsev 1993 ] .
hekking and nazarov [ 1993,1994 ] , obtained similar results using a tunneling approach and beenakker and co - workers [ beenakker et al 1994 ; marmorkos et al 1993 ] confirmed these results using multiple scattering methods . to illustrate how these different methods explain the zba , the following sub - section briefly summarizes quasi - classical predictions for the zero - energy conductance and in sub - section c
, the multiple scattering approach is discussed .
to apply the circuit rules of quasi - classical theory[nazarov 1994 ] to a n - i - s system , we consider two resistances in series ; the first is associated with the normal disordered region ( n ) , while the second is associated with the tunnel junction ( i ) .
we assume that the normal region progressively widens in a macroscopic electrode and that the superconductor plays the role of the other electrode . in this circuit
there are three spectral vectors .
the vector associated with the normal electrode is @xmath462 and that associated with the superconducting electrode is @xmath463 , both or which are fixed ( cf .
rule ii of circuit theory ) .
the third spectral vector is located at the node connecting the diffusive element and the tunnel junction and lies in the @xmath464 plane @xmath465 . according to the circuit rules ,
one computes the conductance of this simple circuit in the usual way , but the resistance of the tunnel junction is renormalized ( cf .
rule i of circuit theory ) @xmath466 where @xmath467 and @xmath468 are the conductances of the normal diffusive region and of the tunnel junction in the normal state , respectively .
the angle @xmath293 is determined from the conservation of the vector current ( cf .
rules iii - iv of circuit theory ) and satisfies the equation @xmath469 this equation has two simple limits : i ) @xmath470 , @xmath471 , @xmath472 ; ii ) @xmath473 , @xmath474 , @xmath475 .
this shows that , by varying the parameter @xmath476 , the conductance switches from a quadratic to a linear dependence on the tunnel junction conductance .
a two - particle process behaves effectively as a one - particle process ( @xmath477 ) .
the smallness of the probability for a two - particle transmission is compensated by the relative increase of disorder - induced scattering on the normal side .
a detailed comparion between eq.([5.15 ] ) and a numerical s - matrix investigation can be found in claughton , raimondi and lambert [ 1996 ] . to illustrate the multiple scattering approach to the zba
, we now show how equation ( [ 5.15 ] ) may be derived using random matrix theory [ beenakker et al 1994 ; for a more detailed introduction to random matrix theory applied to quantum transport see also stone , mello , muttalib and pichard [ 1991 ] and the recent review by beenakker 1997 ] .
the starting point is the following scaling equation for the eigenvalue density @xmath478 [ p.a .
mello and j .- l .
pichard 1989 ] , @xmath479 where @xmath480 is related to the transmission eigenvalue @xmath456 of the n - th transverse mode by @xmath481 and the average @xmath292 is over the appropriate ensemble . here
@xmath482 , where @xmath12 is the size of the system and @xmath77 the mean free path .
@xmath483 is the number of open channels in the normal region .
the above equation yields the ensemble averaged eigenvalue density for a diffusive system and it is physically equivalent to the quasiclassical theory outlined in the previous section . once @xmath484 is known , the conductance of the n - i - s system is given by ( cf .
eq.([5.14 ] ) ) @xmath485 with respect to the scaling variable @xmath83 , the presence of a barrier at the n - s interface , acts as an `` initial condition '' , instead of a `` boundary condition '' .
in fact , at the interface , where @xmath486 , the eigenvalue density in a mode - independent approximation , is given by @xmath487 where @xmath181 is the transmission coefficient of the barrier . to solve the evolution equation ( [ 5.17 ] ) with the initial condition ( [ 5.20 ] )
one introduces the auxiliary function @xmath488 @xmath489 which is analytic in the @xmath490-complex plane with a cut along the positive real axis . in terms of @xmath491 , the evolution equation ( [ 5.17 ] )
becomes @xmath492 writing @xmath493 and @xmath494 yields @xmath495 which is eulers equation of an ideal two - dimensional fluid .
the real and imaginary parts of @xmath496 are the cartesian components of the velocity field @xmath497 .
the solution can then be obtained directly in terms of the initial condition @xmath498 , as @xmath499 the initial condition ( [ 5.20 ] ) for the eigenvalue density translates for @xmath496 in @xmath500 which in the limit of small transparency barrier , i.e. , @xmath335 is approximated by ( cf .
eq.([4.57 ] ) ) @xmath501 to evaluate the conductance , we note that in terms of the function @xmath496 , the equation ( [ 5.19 ] ) becomes @xmath502 by using ( [ 5.25 ] ) together with ( [ 5.28 ] ) and writing the real and immaginary part of @xmath496 at @xmath503 , we obtain the pair of equations @xmath504 the first equation requires @xmath505 , and by defining @xmath506 , we rewrite the second equation @xmath507 by recalling that @xmath508 and @xmath509 are the conductances of the disordered normal region and of the tunnel barrier , we recover the equation for the conservation of the vector current ( [ 5.16 ] ) obtained within circuit theory .
the expression for the conductance ( [ 5.26 ] ) may now be obtained by noticing that @xmath510 and setting @xmath503 .
insertion of equation([5.31 ] ) into equation([5.26 ] ) , leads to an expression for the conductance identical to that given by circuit theory ( cf .
eq.([5.15 ] ) ) .
the above ideas have been further developed in a number of treatments of andreev scattering in chaotic and resonant stuctures , including [ altland and zirnbauer 1995 , berkovits 1995 , beenakker et al 1995 , brouwer and beenakker 1995(a , b ) , bruun et al 1995 , claughton et al 1995(b ) , argman and zee , 1996 , fraum et al 1996 , melsen et al 1996 , slevin et al 1996 ]
as well as the zba discussed in the previous section , a number of recent experiments have revealed a non - monotonic behaviour of the voltage and temperature dependence of the conductance . in this section , we use the quasi - classical approach to highlight the origin of this phenomenon . for simplicity
, we consider the case of a normal diffusive wire located in the region @xmath511 between a normal and a superconducting reservoir .
as explained in section iv(c ) the equation for the green s function @xmath512 , which is conveniently written @xmath513 must be solved with boundary conditions @xmath514 and @xmath515 .
the voltage dependent conductance is then obtained from @xmath516 where @xmath517 is the conductance in the normal state .
using the identity @xmath518 , equation ( [ 6.2 ] ) becomes @xmath519 when @xmath520 is a small quantity , one can expand the denominator of equation ( [ 6.3 ] ) to yield @xmath521 which allows us to obtain approximate solutions in both the low- and high - energy limits . in the limit @xmath522
where @xmath523 , one can treat the term linear in the energy as a perturbation and write @xmath524 . when inserted into equation ( [ 6.1 ] ) , this yields @xmath525 in the opposite limit of large energies , @xmath526 , one notices that among the solutions of the original non - linear equation , there exists a subset satisfying the following equation @xmath527 solutions of equation ( [ 6.6 ] ) are also solutions of the original equation ( [ 6.1 ] ) .
the solution of equation ( [ 6.6 ] ) reads @xmath528 where @xmath529 at the normal reservoir , the value of the solution can not be imposed , though in the limit of large energies , the r.h.s . of (
[ 6.7 ] ) is exponentially small and the solution satisfies the condition that the pairing function must vanish .
we also note that in both the low- and high - energy limits , the real part of @xmath530 is a small quantity and therefore the assumption made at the beginning is justified . with these two solutions
we may finally write @xmath531 @xmath532 with @xmath533 and @xmath534 numerical constants .
the above analysis illustrates the key ingredients required to obtain a reentrant effect , which was noted prior to the experiments by [ artemenko , volkov and zaitsev 1979 ] .
the effect originates from the presence of a distribution function which is spatially not in equilibrium and consequently will be most relevant in situations where the voltage drop is distributed along the wire , rather than dropping predominantly across the tunnel junction .
the fact that the conductance variation appears to be a small quantity both in the low energy and in the high energy limit , controlled respectively by the small parameters @xmath535 and @xmath536 , is not accidental
. it has been noticed that the conductance correction resembles the fluctuation correction in the theory of paraconductivity , the so - called maki - thompson term [ volkov 1994(a ) ] . in the high energy limit
, the proximity effect induces cooper pairs in the normal metal , which act like a superconducting fluctuation above the critical temperature . at low energy ,
the induced pair wave function is not small , so at first sight the correction should be sizeable .
however the `` longitudinal '' component of the induced pair wave function gives a contribution which is equal to the normal state conductance and is not temperature dependent , while the `` transverse '' component , which is small at low energies , gives the additional temperature dependent contribution . on general grounds ,
one expects that the low energy @xmath537 dependence should be universal , while the high energy behaviour may differ from the inverse square root law , depending on the quality of the interface and on the effective dimensionality of the mesoscopic wire .
we now discuss briefly the origin of the long - range proximity effects , observed in the experiments of [ petrashov et al .
1993,1995 , dimoulas et al .
1995 , courtois et al . 1996 ] , where the conductance of a mesoscopic loop between two superconducting electrodes was measured .
in such structures , the total current is the sum of two terms .
the first is a josephson current between the two superconducting electrodes and decays exponentially as @xmath23 in agreement with known theory of superconducting weak links [ likharev 1979 ] .
the second was shown to be due to a phase - periodic conductance with amplitude decaying as the inverse of the temperature .
a detailed theory of the effect has been provided by [ volkov , allsopp and lambert 1996 and volkov and takayanagi 1996(a ) ] using both quasi - classical and multiple scattering methods .
the different behaviour of the two contributions to the current occurs because the josephson current has a thermodynamic origin and is obtained by integrating over a range of energies of the order of @xmath538 , while the second contribution giving rise to the reentrant behaviour is kinetic in nature . in the structure considered in the experiments ,
a non - vanishing value of the pair wave function @xmath539 is induced in the normal region .
away from the superconductor the pair wave function decays exponentially as @xmath540 where @xmath541 .
the contribution to the josephson critical current comes from an integral over the energy involving the product of anomalous green s function @xmath542 and @xmath543 .
since this product contains green s functions with poles on the same side of the real axis , one can deform the contour of integration and transform the integral over the energy to a sum over matsubara frequencies @xmath544 . as a result ,
the main contribution comes from energies of the order of the temperature . at a distance l ,
the josephson current decays exponentially as @xmath23 where @xmath545 . by introducing the thouless energy @xmath546 , the condition for a vanishing josephson current becomes @xmath547 .
in addition to the above , the proximity effect gives rise to an additional contribution to the normal state conductance , which involves products of the type @xmath548 . as a consequence , the integral over the energy
can not be transformed to a sum over matsubara frequencies .
( this expresses the fact that the main contribution to the integral comes from energies of the order of the thouless energy . ) as a consequence , if the condition @xmath547 is satisfied , then the pair wave function spreads over the entire sample .
when a quasi - particle andreev reflects from a normal - superconducting interface , the phase of the outgoing excitation is shifted by the phase of the superconducting order parameter . consequently if a phase - coherent normal conductor is in contact with two superconductors with order parameter phases @xmath24 and @xmath25 , transport properties will be oscillatory functions of the phase difference @xmath549 . following a number of theoretical proposals [ spivak and khemelnitskii 1982 , altshuler and spivak 1987 , nakano and takayanagi 1991 , takagi 1992 , lambert 1993 , hui and lambert 1993(a ) , hekking and nazarov 1993 ] ,
several experimental realizations of andreev interferometers have been reported [ de vegvar et al .
1994 , pothier et al .
1994 , van wees et al 1994 , dimoulas et al .
1995 , petrashov et al .
1995 , courtois et al 1996 ] .
in addition to interferometers formed when a normal metal makes contact with two superconductors , the conductance of a single exended n - s structure is predicted to be an oscillatory function of the phase gradient across the interface [ cook et al 1995 ] , although no experiments on such phase - gradiometers have been reported to - date .
these papers address a number of issues including the fundamental periodicity of the conductance oscillation , the nature of the zero - phase extremum , the magnitude of the oscillation amplitude and the role of disorder , geometry and dimensionality . in the main ,
experiments have probed either diffusive or ballistic structures , which in - turn can be divided into metallic samples with a conductance @xmath0 much greater than @xmath550 and semiconducting structures with @xmath0 less than or of order @xmath550 . for the former , the ensemble averaged conductance @xmath551 is the relevant quantity and mesoscopic fluctuations are unimportant .
these have confirmed the @xmath30 periodicity for @xmath551 predicted by [ lambert 1993 , hui and lambert 1993 ] and by all subsequent theories , [ hekking and nazarov 1993 , 1994 , zaitsev 1994 , takane 1994 , allsopp et al 1996 , kadigrobov et al 1996 , nazarov and stoof 1996 , volkov and zaitsev 1996 , zagoskin et al 1996 , leadbeater and lambert 1997 ] . for the latter
, mesoscopic fluctuations can dominate and the non - averaged conductance @xmath0 is of interest , which depends in detail on the geometry and impurity realisation of a particular sample .
these confirm the @xmath30 perodicity for the non - averaged conductance predicted initially for diffusive systems by [ altshuler and spivak 1987 ] and for one - dimensional clean systems by [ nakano and takayanagi 1991 and takagi 1992 ] . in this section
we briefly review the various theoretical approaches and comment on their applicability to the experiments .
we start by using quasi - classical theory to compute the ensemble averaged conductance when the mesoscopic normal conductor is in the diffusive transport regime .
as noted in previous sections , the influence of superconductivity on the transport properties of a mesoscopic conductor may change in a qualitative way , depending on the quality of the normal - superconductor interfaces .
consider first the zero temperature conductance of an interferometer comprising a tunnel junction ( labelled as 1 ) connected by diffusive 1-d wires to a fork .
each of the two arms of the fork is a diffusive wire , connected via tunnel junctions ( labelled as 2 ) to infinitely long superconductors .
the conductance of the diffusive wires is assumed to be much greater than that of the tunnel junctions . in this structure
the superconductors play the role of elctrode and the main voltage drop occurs at the tunnel barriers .
the conductance can be analysed using the circuit theory outlined in section iv [ nazarov 1994 , zaitsev 1994 ] , which yields @xmath552 where @xmath553 is the conductance of the tunnel junction ( 1 ) , @xmath554 is the conductance of the tunnel junction ( 2 ) and @xmath27 is the phase difference between the two superconductors . in the limit @xmath555 , this simplifies to @xmath556 whereas if @xmath557 , @xmath558 one can see that for @xmath559 there is a zero - phase minimum , as there is for @xmath560 .
however when @xmath555 , there arises a zero - phase maximum . in all cases the conductance vanishes when the phase difference between the superconductors is @xmath31 .
furthermore , the ratio @xmath561 controls the amplitude of the conductance oscillations which are greater when @xmath562 .
the results of quasi - classical theory can of course be compared with the results of numerical multiple scattering theory .
if the parameters used in the numerical simulations are chosen in such a way to satisfy the assumptions underlying the quasi - classical treatment , then a detailed analysis shows that this is indeed the case [ claughton , raimondi and lambert 1996 ] , although certain features , such as the vanishing of the conductance at @xmath563 , appear to be an artifact of te one - dimensional nature of the analysis leading to equation ( 155 ) .
it is perhaps worth emphasizing that the exact numerical techniques used in [ claughton , raimondi and lambert 1996 ] are not confined to the diffusive regime and an interesting result of these simulations is that when the resistance of a given structure is dominated by tunnel barriers , the diffusive nature of the wires is not relevant and similar results are obtained for both ballistic or diffusive wires .
the geometry is clearly important , but when dissipation occurs along the wire , the almost one - dimensional approximation used in many quasi - classical calculations works well . another structure illustrating the role of finite temperature and voltage
is suggested by the experiment by petrashov et al . [
1995 ] , who measured the conductance of a mesoscopic wire , which makes contact with superconductors at points between the normal electrodes . at zero temperature ,
this structure is predicted to have only small amplitude of osicllation arising from mesoscopic flcutuations , in contrast with the large amplitude oscillations observed experimentally . in the experiment however , a finite voltage drop is distributed along the wire and the temperature is non - zero .
consequently the phase periodic conductance is dominated by an effect similar to the reentrant behaviour discussed in the previous section .
the analysis of this structure has been carried out in detail by [ nazarov and stoof 1996 , volkov , allsopp and lambert 1996 , stoof and nazarov 1996 ] , using both quasi - classical and numerical multiple scattering methods .
having discussed diffusive interferometers , we now examine the clean limit , which was first described in one dimension by [ nakano and takayanagi 1991 and takagi 1992 ] .
an analytic theory of the ballistic limit in higher dimensions has also been developed [ allsopp et al 1996 ] , based on a multiple scattering description of a clean n - s interface . in the absence of disorder , when there is no phase difference between the two superconductors , translational invariance in the direction parallel to the interface ( and transverse to the current flow ) allows one to reduce the two - dimensional system to the sum over many independent one - dimensional channels .
when a phase difference between the superconductors is imposed , inter - channel coupling is introduced , but as shown below , this coupling typically involves only pairs of channels in such a way that an accurate description is obtained by summing over independent * pairs * of coupled channels .
this considerably simplifies evaluation of the boundary conductance @xmath564 where @xmath565 is the andreev reflection probability from channel @xmath93 to channel @xmath60 .
indeed the andreev reflection coefficient is of the form @xmath566 where @xmath567 and @xmath568 is the remaining contribution from inter - channel scattering , @xmath569 .
if channels only couple in pairs , both the off - diagonal scattering and the diagonal scattering will scale as the number of channels .
consider now a normal barrier in the n - region of a n - s interface .
particles ( holes ) impinging on the normal scatterer are described by a scattering matrix @xmath570 , @xmath571 , and those arriving at the n - s interface by a reflection matrix @xmath572 , where @xmath573 the elements of @xmath83 and @xmath572 are themselves matrices describing scattering between open channels of the external leads .
for an ideal interface , where andreev s approximation is valid , @xmath574 and @xmath575 can be neglected and as a consequence , @xmath576 and @xmath577 are unitary and one obtains a generalization of a formula due to beennakker [ 1992 ] @xmath578 with @xmath579 in contrast with the analysis of bennakker [ 1992 ] , where @xmath576 is proportional to the unit matrix , the interference effect of interest here is contained in the fact that @xmath576 induces off - diagonal scattering . substituting @xmath580 into equation ( [ 7.4 ] ) and taking advantage of particle - hole symmetry at @xmath154 , yields @xmath581 where @xmath582 with @xmath583 the transmission matrix of the normal scattering region .
this multiple scattering formula for the boundary conductance is valid in the presence of an arbitrary number of channels and in any dimension
. equation ( [ 7.5 ] ) is very general and makes no assumption about the nature of matrices @xmath577 and @xmath570 . in a two - channel model , @xmath577 is chosen to be an arbitrary two dimensional unitary matrix and since in the absence of disorder , @xmath584 and @xmath585 are diagonal and interchannel coupling arises from @xmath577 only . substituting these matrices into equation ( [ 7.5 ] ) , yields an expression for @xmath580 involving a single phase @xmath293 , whose value is a linear combination of phase shifts due to normal reflection at the barrier , andreev reflection at the n - s interface and the phase accumulated by an excitation travelling from the barrier to the interface . after averaging over the rapidly varying phase @xmath293 ,
the diagonal and off - diagonal terms are determined and found to vary periodically with @xmath27 with period @xmath30 as predicted in one dimension by [ nakano and takayanagi 1991 ] a key new feature of equation ( [ 7.5 ] ) is that the in the absence of the normal barrier , unitarity ensures the the total andreev reflection obtained by summing these two contribution is independent of @xmath27 and therefore the amplitude of conductance oscillation vanishes .
thus to obtain a maximum amplitude of oscillation , the tunnel barrier must be tuned to a non - zero value .
an interesting mechanism capable of producing large conductance oscillations which scale with the number of channels has been proposed by [ kadigrobov et al 1995 ] , based on the simultaneous resonance of andreev levels corresponding to different channels . to illustrate this ,
consider for example the clean s - n - s structure analyzed by [ bardeen and johnson 1972 and kulik 1970 ] . for a long - junction , the andreev levels have energies @xmath586 , \label{bardeen}\ ] ] where @xmath587 is the component of the fermi velocity perpendicular to the n - s interfaces and @xmath588 the phase difference between the superconductors . in general these energies
do not coincide , but when @xmath589 all levels pass through the fermi energy @xmath154 , yielding large - scale oscillations in transport properties . in the calculation of [ kadigrobov et al 1995 ]
, normal scattering arises from the beam splitters which form an integral part of the structure considered and therefore no addition tunnel barriers are needed to break the unitarity condition of [ allsopp et al 1996 ] .
when a superconducting island is added to a normal host , one navely expects that the electrical conductance of the composite material will increase .
in contrast detailed calculations based on the multiple scattering formula [ hui and lambert 1993(b ) , claughton et al 1995 ] predicted that the change @xmath35 in the two probe conductance of a mesoscopic sample , due to the onset of superconductivity can have arbitrary sign .
indeed for a sample with a high enough conductance , a simple theorem [ hui and lambert 1993 ] states that @xmath35 is guaranteed to be negative and furthermore the magnitude of @xmath35 is predicted to scale with @xmath0 . this anomalous proximity effect " occurs in the absence of tunnel barriers and at first sight appears to conflict with the quasi - classical description presented in section iv , which predicts that at zero temperature , the conductance of equation ( [ a22 ] ) , which takes the form @xmath590 is unchanged by the onset of superconductivity . in this section we briefly discuss the origin of this effect and how it can be reconciled with quasi - classical theory . for convenience
we separately discuss the clean , diffusive and almost anderson localised limits .
the landauer formula ( 2 ) predicts that the conductance of a normal ballistic wire is @xmath591 , where @xmath483 is the number of open scattering channels at the fermi energy @xmath592 .
consequently in the absence of spin - splitting , if @xmath592 is varied , @xmath0 exhibits a series of steps of height @xmath29 corresponding to the opening or closing of scattering channels . the effect of inducing a uniform superconducting order parameter of magnitude @xmath593 in such a wire is shown in figure 9 of [ claughton et al 1995a ] . in the vicinity of normal - state steps , the conductance is predicted to be extremely sensitive to the onset of superconductivity . with increasing @xmath593 ,
the steps are destroyed and @xmath0 decreases .
this behaviour is attributed to the breakdown of andreev s approximation , which is most pronounced for scattering channels whose wavevector @xmath594 normal to the n - s interface tends to zero .
such channels occur precisely at a normal - state conductance step . for a clean n - s - n structure ,
provided the superconductor is much longer than the superconducting coherence length @xmath595 , there is negligible quasi - particle transmission through the s region and therefore the total resistance reduces to the sum of two boundary resistances .
since the boundaries are identical , this yields for the total conductance defined in equation ( [ a25 ] ) , @xmath596 and since the system consists of decoupled channels , @xmath191 can be obtained by solving the bogoliubov - de gennes equation at a one dimensional n - s interface . by insisting that scattered wavefunctions and their first derivatives be continuous at the boundary ,
one finds [ claughton et al 1995a ] @xmath597 where @xmath598^{1/2 } } } \label{9.2}\ ] ] and @xmath599 is the longitudinal kinetic energy of a quasi - particle incident along channel @xmath70 .
clearly those channels corresponding to low angle quasi - particles with @xmath600 , possess a small andreev reflection probability @xmath601 and since there is no transmission , the corresponding normal reflection probability @xmath602 approaches unity .
finally one obtains for conductance change @xmath35 due to the switching on of a uniform order parameter in such a clean system , @xmath603= \sum_{n=1}^n { { 1-[1+(\delta_0/\mu_n)^2]^{1/2}}\over { 1+[1+(\delta_0/\mu_n)^2]^{1/2}}}. \label{9.3}\ ] ] it is perhaps worth noting that although the suppression of andreev reflection for low - angle quasi - particles is quite general , the destruction or otherwise of conductance steps in a quantum point contact , depends on the geometry of the contact .
if quasi - particles are adiabatically accelerated before encountering the n - s interface , then andreev reflection need not be suppressed and conductance steps can survive . the criterion for the survival of the @xmath70th step is clearly @xmath604 the results of the previous subsection reveal that @xmath35 is sensitive to fine scale structure in the normal state conductance @xmath42 . for clean systems
this structure takes the form of well - known conductance steps . for strongly disordered systems
, it is known that @xmath42 can exhibit sharp resonances and therefore the behaviour of @xmath0 with increasing @xmath593 is sensitive to such features .
this behaviour is illustrated in figure 10 of [ claughton et al 1995 ] , which shows that negative changes in the conductance with increasing @xmath593 are associated with resonances in the normal - state transport . in view of tunnelling theory
, one might regard the suppression of conductance in an almost insulating n - s structure as unsurprising , in which case the increase in the conductance due to the onset of superconductivity would be regarded as anomalous .
a detailed description of such resonant transport in n - s and n - s - n structures and in resonant interferometers is provided by [ claughton and lambert 1995b ] , which follows an earlier zero - voltage theory [ beenakker 1992 ] and a one - dimensional description of a delta - like potential well [ khlus et al 1994 ] .
the above results show that the conductance of a clean or very dirty conductor can either increase of decrease when superconductivity is induced , depending on the microscopic impurity configuration and on the geometry .
this qualitative behaviour has been observed in experiments by petrashov and antonov [ 1991 ] and petrashov et al [ 1993(b ) ] , which also exhibit the weak magnetic field behaviour shown in fig 12 of [ claughton et al 1995a ] .
nevertheless a quantitative comparison between these calculations and experiments is not possible , because of the differing geometries and multiprobe measurements .
a first attempt at a quantitative comparison has been made recently by wilhelm , zaikin and courtois [ 1997 ] who noted that although the one - dimensional approximation used in sec .
v and vi to study transport properties of n - s structures successfully explains a range of experiments , the experiment by [ petrashov et al 1991 ] requires a multiprobe description .
the key observation is that the kinetic properties of a two - dimensional metallic film , in contact with a superconductor , differ substantially from those of a quasi one - dimensional wire , because of an inhomogenous distribution of the currents in the sample .
although a detailed explanation of the effect can be obtained by solving the equations for the quasi - classical green s function in a two - dimensional structure , it is possible to gain considerable insight by adopting an effective circuit model , which captures the main features . for example consider four wires arranged in a square .
the top two corners of the square ( labelled a and b ) are the voltage probes , while the lower two corners ( labelled c and d ) are the current probes .
let the left - most vertical wire between corners a and c be placed in contact with a superconductor and labelled 1 , the right - most vertical wire between corners b and d labelled 2 and the upper and lower wires labelled 3 and 4 respectively .
then the four - probe conductance reads @xmath605 where @xmath606 is the conductance of wire @xmath60 .
the analysis of sections iv , v and vi shows that the conductance of a wire in contact with a superconductor is changed in several ways .
1 ) the local conductivity is _ increased _ by the presence of a superconductor , but this increase decays over a distance from the superconductor of the order of the coherence length of the normal metal @xmath607 .
2 ) the conductance has a non monotonic behaviour as function of the energy and temperature .
to a first approximation one may consider that when @xmath608 , only the wire attached to the superconductor will be appreciably affected by the presence of the superconductor .
in such a situation the conductances @xmath554 , @xmath609 and @xmath610 are the same as in the normal state , while @xmath553 is _ increased_. in this case , according to equation ( [ fourprobes ] ) , the four - probe conductance is _ decreased _ by superconductivity . on the other hand , at @xmath611 , all conductances increase and so does the four - probes conductance . as
[ wilhelm et al 1997 ] have shown this leads to a much richer structure in the temperature and voltage dependence of a two - dimensional film , compared with that of a one - dimensional wire and to qualitative agreement with experiment .
a key prediction of [ wilhelm et al 1997 ] is that at zero bias and zero temperature this effect should vanish and the conductance should be unchanged by the onset of superconductivity .
in contrast [ seviour et al 1997 ] identify an additional large - scale effect due to weak localisation , which is not contained in standard quasi - classical theory and which will survive at zero - temperature and zero - bias .
we conclude this review by indicating a few lines of future research .
the results of the first experiments in hybrid structures have been successfully explained in terms of almost one - dimensional models .
this has been possible because the physical phenomena responsible for the transport properties of a given structure mainly occurred at the n - s interface , so that the actual geometrical arrangement of current and voltage probes was not important .
this is clearly the case in the presence of low transmitting interfaces , which tend to dominate the overall transport properties .
the improvement of the quality of n - s interfaces betweens separate parts of the hybrid structure and the contact leads makes necessary to take into account the detailed geometrical arrangement of a given measurement setup [ allsopp , hui , lambert and robinson 1994 ] . under this respect ,
the last twelve months have witnessed a considerable experimental effort at paying particular attention to the geometry of a given measurement .
as we have seen in the previous section taking into account the actual disposition of the current and voltage probes has been crucial to describe the effect of the conductance suppression , even though the structure itself has a one - dimensional geometry [ wilheml , zaikin and courtois 1997 ] .
the recent experiment by [ hartog et al 1996 ] has beautifully shown the emergence of new physical effects when the geometry of the experimental setup is considered carefully . a second important semplifying assumption in examining the experiments in hybrid structures consisted in neglecting the effect of electron - electron coulomb interaction .
this is a reasonable assumption in many metallic and semiconducting wires used in the experiments . in these cases ,
a mean - field treatment of electron - electron interaction is sufficient and can be easily incorporated in the effective electronic parameters . in the last decade
there has been however a continuous progress in downsizing fabrication of electronic systems , so that it is now possible to measure transport through almost zero - dimensional structures , or quantum dots . in these systems , the effect of coulomb interaction is no longer negligeable , and various new effects arise .
besides quantum dots , quantum wires also have been a subject of noticeable attention .
it is well known that in one - dimensional systems electron - electron interaction drastically modifies the low energy properties and lead to the concept of luttinger liquid .
quantum wires are believed to be an experimental realization of the luttinger liquid state . as a consequence in the last few years
a considerable theoretical literature has accumulated with the intent of studing phase coherent transport in interacting systems .
it is almost a logical consequence to foresee in the forthcoming years the emergence of a new field of research aimed at combining the fields of mesoscopic superconductivity and strongly interacting systems .
indees already quite a few papers have started to appear [ fisher 1994 , fazio , hekking and odintsov 1995 , maslov , stone , goldbart and loss 1996 , fazio and raimondi 1997 , takane and koyama 1997 ] and the time is mature for new exciting developments . | this article is an overview of recent experimental and theoretical work on transport in phase - coherent hybrid nanostructures , with particular emphasis on dc electrical conduction .
a summary of multiple scattering theory and the quasi - classical methods is presented and comparisons between the two are made .
several paradigms of phase - coherent transport are discussed , including zero - bias anomalies , reentrant and long range proximity effects , andreev interferometers and superconductivity - induced conductance suppression . |
much interest and many questions have been raised by the recent discovery of deep and extremely sharp indentations in the surface phonon spectra of h / mo(110 ) and h / w(110 ) at full hydrogen coverage @xcite ( see fig.[fanom ] ) .
those anomalies are seen at an incommensurate wavevector , @xmath2 , along the @xmath3 $ ] direction ( @xmath4 ) and additionally at the commensurate wavevector @xmath5 at the boundary of the surface brillouin zone ( sbz ) . out of the ordinary rayleigh mode
two simultaneous anomalies develop at those points . one ,
@xmath6 , is extremely deep , and is only seen by helium atom scattering ( has ) .
the other , @xmath7 , is instead soft , and is observed by both has and high resolution electron energy loss spectroscopy ( hreels ) @xcite .
a pronounced but less sharp softening of surface phonons is also seen on the clean ( 001 ) surfaces of w @xcite and mo @xcite .
there , the effect can be explained in terms of a nested structure of the fermi surface @xcite .
however , for the phonon anomalies of h / w(110 ) and h / mo(110 ) such a connection between the electronic structure and the vibrational properties seemed to be non - existent since angular resolved photoemission ( arp ) experiments of the two systems gave no evidence for nesting vectors comparable to the has determined critical wave vectors @xcite .
an alternative model which links the phonon anomalies to the motion of the hydrogen adatoms @xcite has to be ruled out because the has spectra remain practically unchanged when deuterium is adsorbed instead of hydrogen @xcite .
another puzzle is added by the observation of a symmetry loss in the low - energy electron diffraction ( leed ) pattern of w(110 ) upon h - adsorption .
the phenomena may be caused by a h - induced displacement of the top layer w atoms along the @xmath8 $ ] direction @xcite ( see figs.[fadso]b and c ) .
similar studies for h / mo(110 ) do not provide any evidence for a top - layer - shift reconstruction @xcite . within a first - principles approach ,
the goal of our work is to give a comprehensive explanation for the observed anomalous behavior and clarify the confusing picture drawn by the different experimental findings .
we perform density - functional theory ( dft ) calculations using the local - density approximation ( lda ) for the exchange - correlation energy functional @xcite . for the self - consistent solution of the kohn - sham equations a full - potential linearized augmented plane - wave ( fp - lapw ) code @xcite
is employed which we enhanced by the direct calculation of atomic forces @xcite . within a damped molecular dynamics approach
this enables an efficient determination of fully relaxed atomic geometries and frozen - phonon energies . for methodical details
we refer to ref @xcite . in the case of the w surfaces the valence and
semi - core ( core ) electrons are treated scalar ( fully ) relativistically while for mo all electrons are handled non - relativistically .
the in - plane lattice constants @xmath9 and @xmath10 are calculated without including zero - point vibrations .
they are in good agreement with the respective measured bulk lattice parameters at room temperature which are 3.163 and 3.148 for w @xcite and mo @xcite , respectively .
we performed systematic tests comparing the lda and the generalized gradient approximation @xcite exchange - correlation functionals .
for the quantities reported in this paper ( total energy differences of different adsorption sites , bond lengths , etc . ) both treatments give practically the same results .
also , we checked the @xmath11-point convergence : in the case of atomic and electronic structure calculations a two - dimensional linear mesh of 64 @xmath11-points within the @xmath12 sbz gives stable results .
for the evaluation of frozen - phonon energies a set of 56@xmath11-points is employed within the sbz of the enlarged @xmath13 and @xmath14 surface cells .
cccccc @xmath15 & @xmath16 & @xmath17 & @xmath18 & @xmath19 & @xmath20 + ( ) & ( ) & ( ) & ( @xmath21 ) & ( @xmath21 ) & ( @xmath21 ) + + @xmath22 & @xmath22 & @xmath22 & @xmath23 & @xmath24 & @xmath25 + @xmath22 & @xmath22 & @xmath22 & @xmath26 & @xmath27 & @xmath28 + + 0.62 & 1.09 & 0.04 & @xmath29 & + 0.3 & @xmath30 + @xmath31 & @xmath32 & @xmath33 & @xmath34 & @xmath35 & @xmath36 + + @xmath22 & @xmath22 & @xmath22 & @xmath223.6 & @xmath370.2 & @xmath220.3 + @xmath22 & @xmath22 & @xmath22 & @xmath38 & @xmath39 & @xmath40 + + 0.67 & 1.09 & 0.02 & @xmath41 & @xmath25 & @xmath42 + @xmath43 & @xmath44 & @xmath45 & @xmath46 & @xmath47 & [email protected] [ tstructure ] the first step in our study is to determine the atomic structures of the clean and hydrogen - covered ( 110 ) surfaces .
the relaxation parameters presented in table [ tstructure ] are calculated for a nine - layer slab ( the parameters are defined in fig .
[ fadso]d ) .
the results , which are remarkably similar for mo and w , are in excellent quantitative agreement with a recent leed analysis @xcite . as the energetically most favorable hydrogen adsorption site on both w(110 ) and mo(110 )
we identify a quasi - threefold position ( indicated as h in fig.[fadso]a ) .
our investigations also throw light on the suggested model of a h - induced structural change : for both materials the calculated shift @xmath17 is only of the order of 0.01 and thus there is no evidence for a pronounced top - layer - shift reconstruction .
moreover , this subtle change in the surface geometry is unlikely to be resolved experimentally due to zero - point vibrations .
this aspect becomes evident by evaluating the total energy with respect to a rigid top - layer - shift . in fig .
[ fshift ] we present the results of such a calculation for h / w(110 ) and depict the first three vibrational eigenstates obtained from a harmonic expansion of the total energy @xmath49 ; for h / mo(110 ) the energetics is similar .
since we have @xmath50mev at room temperature thermal fluctuations of @xmath17 are of the order of 0.1 .
this is considerably larger than the theoretically predicted absolute value of @xmath17 . the similarities between mo(110 ) and
w(110 ) found for the surface geometries continue when we compare the electronic structures .
for both systems the h adsorption alters the surface potential and induces the shift of a band with @xmath51 character to higher binding energies @xcite . due to this effect
which is illustrated in fig .
[ fband ] the fermi line associated with this band is moved into the band gap of the surface projected band structure , and the respective states become true surface states . the shifted @xmath51 band is characterized by a high density of states at the fermi level .
more important , the new fermi contour shows pronounced nesting features ( see fig . [
ffermi]c ) .
the magnitudes of the calculated nesting vectors and has measured critical wavevectors along @xmath52 and @xmath53 are listed in table [ tnesting ] . for both systems
the agreement between theory and experiment is excellent .
.theoretical fermi - surface nesting vectors compared to critical wavevectors obtained by has and hreels experiments .
[ cols= " < , < , < , < " , ] [ tphonon ] fortunately , at @xmath54 such a calculation is particularly convenient , since at the zone boundary point @xmath55 the rayleigh - wave polarization is purely vertical , and the second layer is immobile by symmetry .
the calculated frequencies for the @xmath55-point and a second zone boundary point , @xmath56 , are presented in table [ tphonon ] . for the latter our results
reproduce the experimentally observed increase of the rayleigh wave frequency as hydrogen is adsorbed . at the @xmath57 point
we find that those stiffening effects are over - compensated : the strong coupling to electronic states at the fermi level leads to a lowering of the phonon energy in good agreement with the experimental results . in fig .
[ fkohn ] we schematically illustrate the mechanism which is responsible for this effect .
it is , in fact , a text - book example of a kohn anomaly due to fermi - surface nesting @xcite : ( a ) within the unperturbated system the ( @xmath58 ) band cuts the fermi level exactly midway between @xmath59 and @xmath55 .
( b ) the @xmath55-point phonon couples to the states at the fermi level and causes the backfolding of the sbz .
( c ) the nuclear distortion associated with the @xmath55-point phonon modifies the surface potential and hence removes the degeneracy at the new zone boundary point @xmath60 .
the occupied states are shifted to lower energies .
this amounts to a negative contribution of the electronic band structure energy to the total energy and thus a lowering of the phonon energy .
a frozen - phonon study for the second nesting vector along @xmath53 is not performed because @xmath2 is highly non - commensurate ; a calculation would be very expensive .
however , since the character of the @xmath51 band does not change when shifting from the @xmath52 to the @xmath53 nesting we expect similar results .
the calculation of the electron - phonon interaction at the mo(110 ) and w(110 ) surfaces and its change due to hydrogen adsorption pinpoints the phonon character of the small anomaly @xmath7 and identifies the interplay between the electronic structure and the vibrational spectra of the transition metal surfaces .
thus , our results which were recently confirmed by bungaro @xcite within the framework of perturbation dft clearly support the interpretation that the small dip observed by both has and hreels is due to a kohn anomaly . for the deep and sharp anomaly @xmath6
the interpretation is less straightforward , particularly since it is only seen in the has spectra .
one needs to understand the nature of rare - gas scattering in order to explain this unusual behavior . in a recent study we found that those scattering processes are significantly more complicated ( and more interesting ) than hitherto assumed @xcite : the reflection of the he atom happens in front of the surface at a distance of 23 .
more important , it is not the _ total _ electron density of the substrate surface which determines the interaction but the electronic wave functions close to the fermi level . in the case of the h / w(110 ) and h / mo(110 ) systems it is thus plausible to assume that the he atom couples directly to the @xmath51 surface states mentioned above and excites electron - hole pairs . in order to analyze the spectrum of those excitations in some more detail we evaluate the imaginary part of the electronic susceptibility @xmath61 within random phase approximation ( rpa ) : @xmath62 with @xmath63 being the @xmath11-space density . in our approach
we use all eigenvalues @xmath64 and occupation numbers @xmath65 obtained via a nine - layer slab calculations . also , we take into account the localization @xmath66 of the respective state at a distance of 2.5 in front of the substrate surface .
the results obtained for mo(110 ) and h / mo(110 ) are presented in fig .
[ fsusc ] : for the clean surface we find that the intensity of the electron - hole excitations decreases continuously as we move away from the @xmath59 point towards the zone boundaries .
this relatively smooth behavior of the susceptibility is modified considerably when h is adsorbed on the clean surface : pronounced peaks appear , in particular one which stretches from the @xmath53 line through the sbz to the symmetry point @xmath55 .
this is in excellent agreement with the has results .
also , there is a peak located at the second half of the @xmath67 symmetry line .
it can not be associated with an anomaly in the has spectrum .
this result may be due to the fact that selection rules for the interaction between the scattering he atoms and the electron - hole excitations are not considered . in view of the simplicity of the approach
, the calculated spectrum of electron - hole excitations however seems to be well - connected to the has measurements .
moreover , it is surprising how detailed the fermi surface nesting is reflected in the susceptibility at such a large distance from the surface where the has scattering process takes place .
this finding supports the earlier suggestion @xcite that the giant indentations , seen only in has , are due to the excitation of electron - hole pairs .
by contrast , in hreels experiments the electrons are directly scattered at the atoms and thus only measure the phonon spectrum .
therefore , the strong anomaly is invisible for hreels .
in conclusion , the major results of our work are as follows : we demonstrate that both the w(110 ) and the mo(110 ) surface do not show a pronounced top - layer - shift reconstruction upon adsorption of hydrogen , a result confirmed by a recent leed analysis @xcite .
also , we give a consistent explanation for the h - induced anomalies in the has spectra of w(110 ) and mo(110 ) .
the h adsorption induces surface states of @xmath51 character that show pronounced fermi - surface nesting .
the soft anomalous dip is clearly identified as a kohn anomaly due to those nesting features @xcite . for an understanding of the huge dip one needs to take into account that the scattering of rare - gas atoms is crucially influenced by interactions with substrate surface wave functions at the fermi level @xcite .
the he atom couples efficiently to the h - induced surface states on h / w(110 ) and h / mo(110 ) .
we therefore conclude that the deep branch of the anomaly is predominantly caused by a direct excitation of electron - hole pairs during the scattering process .
we hope that our first - principles study stimulates some more experimental work : for instance , the liquid - like behavior of the h adatoms observed in the hreels measurements @xcite is still not understood .
one would also like to know whether the has and hreels spectra of the mo@xmath0re@xmath1(110 ) alloy system reveal h - induced anomalies as expected .
finally , we call for a new arp study of the clean and h - covered mo(110 ) and w(110 ) surface and think that scattering experiments with atoms or molecules like ne or h@xmath68 might provide additional insight into the interesting behavior of these surfaces . e.hulpke and j.ldecke , , 2846 ( 1992 ) ; j. electron . spectr
* 64/65 * ( 1993 ) 641 .
j.ldecke , ph.d .
thesis ( universitt gttingen , 1994 ) .
m.balden , s.lehwald , h.ibach , and d.l.mills , ( 1994 ) 854 .
m.balden , ph.d.thesis ( rwth aachen , 1995 ) .
j.ernst , e.hulpke , and j.p.toennies , phys.rev.b * 46 * ( 1992 ) 16081 .
e.hulpke and d .- m.smilgies , phys.rev.b * 40 * ( 1989 ) 1338 .
k.e.smith , g.s.elliott , and s.d.kevan , phys.rev.b * 42 * ( 1990 ) 5385 .
j.w.chung , k.s.shin , d.h.baek , c.y.kim , h.w.kim , s.k.lee , c.y.park , s.c.hong , t.kinoshita , m.watanabe , a.kakizaki , and t.ishii , phys.rev.lett . * 69 * ( 1992 ) 2228 .
k.h.jeong , r.h.gaylord , and s.d.kevan , phys.rev.b * 38 * ( 1988 ) 10302 ; phys.rev.b * 39 * ( 1989 ) 2973 ; j.vac.sci.technol.a * 7 * ( 1989 ) 2199 .
r.h.gaylord , k.h.jeong , and s.d.kevan , phys.rev.lett .
* 62 * ( 1989 ) 2036 .
j.w.chung , s.c.ying , and p.j.estrup , ( 1986 ) 749 .
m.altman , j.w.chung , p.j.estrup , j.m.kosterlitz , j.prybyla , d.sahu , and s.c.ying , j. vac .
technol . a * 5 * ( 1987 ) 1045 .
d.m.ceperley and b.j.alder , ( 1980 ) 566 ; j.p.perdew and a.zunger , ( 1981 ) 5048 .
p.blaha , k.schwarz , p.sorantin , and s.b.trickey , comput .
* 59 * ( 1990 ) 399 ; p.blaha , k.schwarz , r.augustyn , wien93 ( technische universitt wien , 1993 ) .
r.yu , d.singh , and h.krakauer , ( 1991 ) 6411 .
b.kohler , s.wilke , m.scheffler , r.kouba , and c.ambrosch-draxl , comput .
* 94 * ( 1996 ) 31 .
b.kohler , p.ruggerone , s.wilke , and m.scheffler , ( 1995 ) 1387 .
j.s.shah and m.e.straumanis , j.appl.phys.*42 * ( 1971 ) 3288 .
k.w.katahara , m.h.manghnani , and e.s.fisher , j. phys .
f * 9 * ( 1979 ) 773 .
j.p.perdew and y.wang , ( 1992 ) 13244 .
m.arnold , l.hammer , k.heinz , b.kohler , and m.scheffler , to be published .
m.arnold , private communication .
p.ruggerone , b.kohler , s.wilke , and m.scheffler , in : electronic surface and interface states on metallic systems , eds .
e.bertel and m.donath ( world scientific , singapore , 1995 ) p.113 .
m.okada , a.p.baddorf , d.m.zehner , and e.w.plummer , in press , surf .
b.kohler , p.ruggerone , m.scheffler , and e.tosatti , to be published in z. phys .
a.b.migdal , sov .
jetp * 34 * ( 1958 ) 996 .
w.kohn , ( 1959 ) 393 .
c.bungaro , ph.d .
thesis ( sissa , trieste , 1995 ) .
m.petersen , s.wilke , p.ruggerone , b.kohler , and m.scheffler , phys.rev.lett.*76 * ( 1996 ) 995 . | helium atom scattering ( has ) studies of the h - covered mo(110 ) and w(110 ) surfaces reveal a twofold anomaly in the respective dispersion curves . in order to explain this unusual behavior we performed density - functional theory calculations of the atomic and electronic structure , the vibrational properties , and the electronic susceptibility of those surfaces .
our work provides evidence for hydrogen adsorption induced fermi - surface nesting .
the respective nesting vectors are in excellent agreement with the has data and recent angle resolved photoemission experiments of the h - covered alloy system mo@xmath0re@xmath1(110 ) . also , we investigated the electron - phonon coupling and discovered that the rayleigh phonon frequency is lowered for those critical wave vectors compared to the clean surfaces .
moreover , the smaller indentation in the has spectra can be clearly identified as a kohn anomaly . based on our results for the susceptibility and the recently improved understanding of the he scattering mechanism
we argue that the larger anomalous dip is due to electron - hole excitations by the he scattering .
= 10000 |
we would like to thank v. p. frolov and d. n. page for stimulating discussions .
this research was supported by the natural sciences and engineering research council of canada and by the killam trust .
li hawking s w 1984 in _ relativity , groups and topology ii _ les houches session xl ( north holland , amsterdam ) page d n 1984 _ class .
quantum grav . _ * 1 * 417 cornish n j and levin j j 1997 _ phys .
d _ * 55 * 7489 dettmann c p , frankel n e and cornish n j 1994 _ phys .
d _ * 50 * 618 cornish n j and gibbons g w 1997 _ class .
quantum grav . _
* 1865 carter b 1968 _ phys . rev . _ * 174 * 1559 majumdar s d 1947 _ phys . rev . _ * 72 * 390 papapetrou a 1947 _ proc .
irish acad . a _ * 51 * 191 ott e 1993 _ chaos in dynamical systems _ ( cambridge university press , cambridge )
cornish n j 1997 e - print gr - qc/9709036 vilenkin a and shellard e p s 1994 _ cosmic strings and other topological defects _
( cambridge university press , cambridge ) misner c w , thorne k s and wheeler j a 1973 _ gravitation _ ( freeman , san francisko ) larsen a l 1994 _ phys .
d _ * 50 * 2623 de vega h j and egusquiza
i l 1994 _ phys .
d _ * 49 * 763 press w h , teukolsky s a , vetterling w t and flannery b p 1992 _ numerical recipes _ ( cambridge university press , cambridge )
cornish n j and levin j j 1997 _ phys .
* 78 * 998 barnsley m f 1988 _ fractals everywhere _
( ap inc , boston ) frolov v , skarzhinski v , zelnikov a and heinrich o 1989 _ phys .
b _ * 224 * 255 de villier j p and frolov v 1999 _ class .
quantum grav . _
* 2403 page d 1999 _ phys . rev .
d _ * 60 * 023510 | we consider scattering and capture of circular cosmic strings by a schwarzschild black hole .
although being a priori a very simple axially symmetric two - body problem , it shows all the features of chaotic scattering . in particular , it contains a fractal set of unstable periodic solutions ; a so - called strange repellor .
we study the different types of trajectories and obtain the fractal dimension of the basin - boundary separating the space of initial conditions according to the different asymptotic outcomes .
we also consider the fractal dimension as a function of energy , and discuss the transition from order to chaos .
chaos in general relativity and cosmology is by now a well - established subject .
some studies , pioneered by the work of hawking @xcite and page @xcite , have been concerned with the behavior of solutions to the einstein equations themselves ; the most famous example now being the mixmaster universe ( see for instance @xcite , and references given therein ) . in other studies , the subject of interest has been the behavior of test - particle trajectories in black hole spacetimes ( see for instance @xcite ) . in the latter case ,
since point - particle motion is completely integrable in the generic kerr - newman background @xcite , it was necessary to consider quite complicated multi black hole spacetimes ( typically of the majumdar - papapetrou type @xcite ) to obtain chaotic point - particle dynamics . taking into account that these systems represent ( at least ) three - body problems , and comparing with newtonian dynamics , it was certainly no surprise that chaotic dynamics appeared .
( it should be mentioned , however , that in the two black hole case , chaos is really a relativistic effect @xcite . )
the main interest in these systems therefore was also to use and further develop coordinate - invariant descriptions and measures of chaotic behavior @xcite , suitable for general relativity where space and time are not absolute .
in the present paper , we study a conceptually simpler and more symmetric system in general relativity that nevertheless , and maybe somewhat surprisingly , leads to chaotic behavior .
we consider the two - body problem of a circular test - string in the background of a schwarzschild black hole .
the circular string is taken to be coaxial with the black hole and is allowed to oscillate in its plane and to propagate in the direction perpendicular to its plane , as illustrated in fig . [ fig : string ] .
the combined system is therefore axially symmetric .
the physical picture we have in mind is that of a cosmic string nearby an astrophysical black hole . for a gut string with string tension of the order @xmath0 @xcite
, the test - string approximation should be valid even for a string initially up to ( say ) @xmath1 times longer than the black hole horizon radius .
moreover , for a gut string we can use the leading order thickness approximation , where the cosmic string is described by the nambu - goto action @xcite @xmath2}\;,\ ] ] where @xmath3 is the induced metric on the string worldsheet . in this case , the string equations of motion and constraints ( in conformal gauge ) take the standard form @xmath4 where dot @xmath5 and prime @xmath6 denote derivatives with respect to string worldsheet coordinates @xmath7 and @xmath8 . using schwarzschild coordinates for the background metric @xmath9 and parameterizing the circular string by ansatz @xmath10
one finds that the string equations of motion ( [ eq : eom ] ) and constraints ( [ eq : constraint ] ) lead to the following system of ordinary differential equations @xmath11
@xmath12 @xmath13 supplemented by the constraint @xmath14 the integration constant @xmath15 , appearing here , will play the role of external control parameter ( `` order parameter '' ) for the system .
it is equal to the total conserved energy of the string divided by @xmath16 . in the absence of first integrals other than ( [ eq : e ] ) , we are thus dealing with a three - dimensional phase space .
notice also that the worldsheet time @xmath7 is not equal to the proper time .
the worldsheet time @xmath7 is however finite for a string falling into the black hole .
eqs .
( [ eq : t][eq : e ] ) are equivalent to the following hamiltonian system @xmath17 with the constraint @xmath18 .
this is very similar to the hamiltonian describing zero angular momentum photons in the schwarzschild background @xcite , except for the last term in the potential which is due to the string tension .
this is precisely the term that leads to non - integrability and , as we shall see , chaos .
the hamiltonian system ( [ eq : h ] ) has been previously considered and solved in the equatorial plane ( @xmath19 ) @xcite , and some trajectories in the general case were computed numerically in @xcite . in this paper
, we make a more complete analysis of the dynamics associated with the hamiltonian ( [ eq : h ] ) . in particular
, we shall demonstrate the presence of chaos in this simple system . due to the non - integrable nature of the dynamical system ( [ eq : t][eq : e ] ) , the analysis of the string evolution was done numerically .
the string trajectories for various initial conditions were obtained by integrating eqs .
( [ eq : r ] ) and ( [ eq : theta ] ) using the fifth order embedded runge - kutta method with adaptive step size control @xcite .
the constraint ( [ eq : e ] ) was used to independently check for numerical precision .
it is easy to see that there are three possible asymptotic outcomes of the string dynamics .
the string can either fly by the black hole and escape to @xmath20 , or the string can backscatter and escape to @xmath21 , or the string can be captured by the black hole @xmath22 . some examples of string trajectories illustrating these outcomes are shown in fig .
[ fig : outcomes ] , where we plot the string radius @xmath23 ( vertically ) as a function of @xmath24 ( horizontally , string comes in from the left ) .
in the three examples shown , the string is initially collapsed at a position a few horizon radii outside of the black hole .
it then expands and propagates to the right towards the black hole . in fig .
[ fig : outcomes]a , the string passes the black hole but then returns and is captured , @xmath22 . in fig .
[ fig : outcomes]b , the string passes the black hole and escapes to the right , @xmath25 . in fig .
[ fig : outcomes]c , the string passes the black hole , then returns and finally escapes to the left , @xmath26 .
besides the solutions with these three asymptotic outcomes , there is an infinite set of unstable periodic orbits , which separate the solutions with different asymptotic outcomes in the phase space of all solutions .
typical examples of periodic orbits are shown in fig .
[ fig : period ] . besides the ones shown , there are also periodic orbits when the string starts a little further away from the black hole and oscillates a number of times before reaching it .
there are no stable periodic orbits in our system . to get a better understanding of the string dynamics , we now consider a two - dimensional slice of the four - dimensional space of initial conditions in more detail .
it is most convenient to fix the constant `` energy '' @xmath15 and then to impose one more relation between initial values of @xmath27 at @xmath28 . following the procedure of the basin - boundary method for chaotic scattering @xcite
, we then color this two - dimensional slice of initial conditions according to the three different asymptotic outcomes mentioned above . due to numerical reasons ,
the determination of the two asymptotic outcomes corresponding to escape is done not at infinity but at some large , but finite , value of @xmath29 . while in general the string can pass this cutoff radius in one direction but change its mind later , for sufficiently large @xmath29
, this procedure will only lead to a wrong color for very few points , as follows from the asymptotic behavior of the potential in the hamiltonian ( [ eq : h ] ) ( this should be contrasted with the case of the mixmaster universe @xcite , where all the trajectories eventually bounce back ) . in fig .
[ fig : phase ] we show examples of this coloring procedure applied to the phase space of string solutions .
fig .
[ fig : phase]a and fig .
[ fig : phase]b show two different two - dimensional slices of the space of initial conditions . fig .
[ fig : phase]c and fig .
[ fig : phase]d show magnifications of the regions of fig .
[ fig : phase]b and fig .
[ fig : phase]c indicated in the previous figures .
the boundaries between different colors in these figures correspond to the unstable periodic orbits .
the magnifications of these so - called basin - boundaries reveal a fractal structure , see figs .
[ fig : phase]b - d .
this provides the coordinate - invariant indication that the dynamics is in fact chaotic @xcite . to get a quantitative measure we can also determine the fractal dimension .
consider for instance fig .
[ fig : phase]b , which corresponds to the two - dimensional slice in the space of initial conditions , given by @xmath30 the box - counting dimension is defined by @xcite @xmath31 where @xmath32 is the number of squares of side length @xmath33 needed to cover the basin - boundary .
a square should be counted only if it contains at least two different colors , otherwise it is not part of the basin - boundary .
the result of this counting is shown as a plot of @xmath34 versus @xmath35 in fig . [
fig : scale ] .
the straight line is a least - square fit to the data points and shows that the dependence is indeed linear over the very wide range of resolution .
the fractal dimension of fig .
[ fig : phase]b , calculated from the highest - resolution grid of @xmath36 points , is @xmath37 the non - integer value shows in a coordinate - invariant way @xcite that the basin - boundary is indeed a fractal ; a so - called strange repellor .
the error in the result is due to the finite size of the numerical grid and slow convergence rate of definition ( [ eq : fdim ] ) , and can be estimated by examining @xmath38 calculated at lower resolutions .
it is interesting to also examine the fractal dimension as a function of `` energy '' @xmath15 .
we considered the slice ( [ eq : slice ] ) in the energy range @xmath39 $ ] and computed the fractal dimension ( [ eq : fdim ] ) of the basin - boundary ( corresponding to fig .
[ fig : phase]b and its analog for different energies ) as a function of energy @xmath15 .
the result is shown in fig .
[ fig : energy ] , where we plot the fractal dimension @xmath38 as a function of @xmath40 . for low energies , @xmath41 ,
the dynamics is completely regular in the sense that there is only one asymptotic outcome namely capture . for slightly larger energies , with values of @xmath15 between @xmath42 and @xmath43 ,
escape becomes possible but the different asymptotic outcomes are connected by `` regular transitions '' , corresponding to one - dimensional curves in the space of initial conditions ( and thus corresponding to fractal dimension @xmath44 ) .
however , at energy @xmath45 , the picture changes dramatically from regular to highly chaotic dynamics . in a very narrow energy range , the fractal dimension changes abruptly from @xmath44 to approximately @xmath46 , whereafter it increases slightly for higher energies .
it must be stressed that the numerical values given above for energies separating the different `` phases '' are slice - dependent . the phenomenon explained and demonstrated by fig .
[ fig : energy ] is however generic . in conclusion ,
we have shown in a coordinate invariant way that the axially symmetric system of a circular test - string in the schwarzschild black hole background is chaotic .
chaos sets in abruptly at a certain `` critical '' value of the external control parameter , which is related to the conserved energy of the string .
below this critical energy , the dynamics is regular . above the critical energy ,
the dynamics is highly irregular and chaotic . in all fairness
, it should be mentioned that we have neglected some physical effects , including the backreaction and gravitational radiation of the cosmic string .
however , the system considered here represents the simplest and most symmetric example of string dynamics in black hole spacetimes and therefore suggests quite generally that string dynamics in black hole spacetimes is chaotic ( notice however that the problem of _ stationary _ strings in black hole spacetimes is completely integrable @xcite ) .
it would thus be interesting to consider other dynamical string configurations as well , and we notice that there has recently been some interest in the scattering and capture of open strings by a black hole ; see refs .
@xcite and references given therein . |
many surveys have been carried out to chart the positions of galaxies in large regions of the universe around us , and many more surveys which go deeper into the universe are currently underway or are planned for the future .
these surveys give us detailed information about the distribution of matter in the universe , and identifying the salient features that characterize this distribution has been a very important problem in cosmology .
the statistical properties , the geometry and the topology are some of the features that have been used to characterize the distribution of galaxies , and a large variety of tools have been developed and used for this purpose .
the correlation functions which characterize the statistical properties of the distributions have been widely applied to quantify galaxy clustering . of the various correlation functions ( 2-point , 3- point ,
etc ... ) the galaxy - galaxy two point correlation function @xmath6 is very well determined on small scales ( peebles 1993 and references therein ) and it has been found to have the form @xmath7 this power - law form of the two point correlation function suggests that the universe exhibits a scale invariant behaviour on small scales @xmath8 . the two point correlation function becomes steeper at larger scales @xmath9 .
it is , however , not very well determined on very large scales where the observations are consistent with the correlation function being equal to zero . the standard cosmological model and the correlation function analysis are both based on the underlying assumption that the universe is homogeneous on very large scales and the indication that the correlation function vanishes at very large scales is consistent with this .
fractal characterization is another way of quantifying the gross features of the galaxy distribution .
fractals have been invoked to describe many physical phenomena which exhibit a scale invariant behaviour and it is very natural to use fractals to describe the clustering of galaxies on small scales where the correlation function analysis clearly demonstrates a scale invariant behaviour .
coleman and pietronero ( @xcite ) applied the fractal analysis to galaxy distributions and concluded that it exhibits a self - similar behaviour up to arbitrarily large scales .
their claim that the fractal behaviour extends out to arbitrarily large scales implies that the universe is not homogeneous on any scale and hence it is meaningless to talk about the mean density of the universe .
these conclusions are in contradiction with the cosmological principle and the entire framework of cosmology , as we understand today , will have to be revised if these conclusions are true . on the other hand ,
several others ( @xcite ) have applied the fractal analysis to arrive at conclusions that are more in keeping with the standard cosmological model .
they conclude that while the distribution of galaxies does exhibit self similarity and scaling behaviour , the scaling behaviour is valid only over a range of length scales and the galaxy distribution is homogeneous on very large scales .
various other observations including the angular distribution of radio sources and the x - ray background testify to the universe being homogeneous on large scales ( @xcite ; @xcite ) .
recent analysis of the eso slice project ( @xcite ) also indicates that the universe is homogeneous over large scales .
the fractal analysis of volume limited subsamples of the ssrs2 ( @xcite ) studies the spatial behaviour of the conditional density at scales up to @xmath10 .
their analysis is unable to conclusively determine whether the distribution of galaxies is fractal or homogeneous and it is consistent with both the scenarios .
a similar analysis carried out for the apm - stromlo survey ( @xcite ) seems to indicate that the distribution of galaxies exhibits a fractal behaviour with a dimension of @xmath11 on scales up to @xmath12 . in a more recent paper ( @xcite )
the fractal analysis has been applied to volume limited subsamples of the las campanas redshift survey .
this uses the conditional density to probe scales up to @xmath13 .
they find evidence for a fractal behaviour with dimension @xmath14 on scales up to @xmath15 .
they also conclude that there is a tendency to homogenization on larger scales ( @xmath16 ) where the fractal dimension has a value @xmath17 , but the scatter in the results is too large to conclusively establish homogeneity and rule out a fractal universe on large scales .
in this paper we study the scaling properties of the galaxy distribution in the las campanas redshift survey ( lcrs ) ( @xcite ) .
this is the deepest redshift survey available at present . here
we apply the multi - fractal analysis ( @xcite ) which is based on a generalization of the concept of a mono - fractal . in a mono - fractal
the scaling behaviour of the point distribution is the same around each point and the whole distribution is characterized by a single scaling index which corresponds to the fractal dimension . a multi - fractal allows for a sequence of scaling indices known as the multi - fractal spectrum of generalized dimensions .
this allows for the possibility that the scaling behaviour is not the same around each point .
the spectrum of generalized dimensions tells how the scaling properties of the galaxy distribution changes from the very dense regions ( clusters ) to the sparsely populated regions ( voids ) in the survey . in this paper
we compute the spectrum of generalized dimensions ( @xmath18 _ vs _
@xmath19 ) by calculating the minkowski - bouligand dimension ( @xcite ) for both volume limited and magnitude limited subsamples of the lcrs .
we also investigate how the spectrum of generalized dimensions depends on the length scales over which it is measured and whether the distribution of galaxies in the lcrs exhibits homogeneity on very large scales or if the fractal nature extends to arbitrarily large scales . .
we next present a brief outline of the organization of the paper .
section [ gendim ] describes the method we adopt to compute the spectrum of generalized dimensions . in section [ survey ]
we describe the basic features of the lcrs and discuss the issues related to the processing of the data so as to bring it into a form usable for our purpose .
section [ analysis ] gives the details of the method of analysis specifically in the context of lcrs .
the discussion of the results are presented in section [ results ] and the conclusions in section [ conc ] .. in several parts of the analysis it is required to use definite values for the hubble parameter @xmath20 and the decceleration parameter @xmath21 , and we have used @xmath22 and @xmath23 .
a fractal point distribution is usually characterized by its dimension and there exists a large variety of ways in which the dimension can be defined and measured . of these possibilities two which are particularly simple and
can be easily applied to a finite distribution of points are the box - counting dimension and the correlation dimension . in this section
we discuss the `` working definitions '' of these two quantities that we have adopted for analyzing a distribution of a finite number of points . for more formal definitions of these dimensions the reader
is referred to borgani ( 1995 ) and references therein .
the formal definitions usually involve the limit where the number of particles tends to infinity and they can not be directly applied to galaxy distributions .
we first consider the box - counting dimension . in calculating the box - counting dimension for a distribution of points , the space is divided into identical boxes and we count the number of boxes which contain at least one point inside them .
we then progressively reduce the size of the boxes while counting the number of boxes with at least one point inside them at every stage of this process .
this gives the number of non - empty boxes @xmath24 as a function of the size of one edge of the box @xmath25 at every stage of the procedure .
if the number of non - empty boxes exhibits a power - law scaling as a function of the size of the box i.e. @xmath26 we then define @xmath27 to be the box - counting dimension . in practice
the nature of the scaling may be different on different length scales and we look for a sufficiently large range of @xmath25 over which @xmath28 exhibits a particular scaling behaviour and we then use equation ( [ eq : gd1 ] ) to obtain the box - counting dimension valid over those scales . so finally we may get more than one value of box - counting dimension for the distribution , each value of the box counting dimension being valid over a limited range of length scales . to compute the correlation dimension for a point distribution with n points we proceed by first labeling the points using an index j which runs from @xmath29 to @xmath30 .
we then randomly select @xmath31 of the @xmath30 points and the index @xmath32 is used to refer to these @xmath31 randomly chosen points .
for every point @xmath32 , we count the total number of points which are within a distance @xmath25 from the @xmath33 point and this quantity @xmath34 can be written as @xmath35 where @xmath36 is the position vector of the @xmath33 point and @xmath37 is the heavy - side function .
@xmath38 for @xmath39 and @xmath40 for @xmath41 .
we next divide @xmath34 by the total number of points @xmath30 to calculate @xmath42 , the probability of finding a point within a distance @xmath25 from the @xmath43 point .
we then average the quantity , @xmath42 , over all the @xmath31 randomly selected centers to determine the probability of finding a point within a distance @xmath25 of another point and we denote this by @xmath44 which is given by , @xmath45 if the probability @xmath46 exhibits a scaling relation @xmath47 we then define @xmath48 to be the correlation dimension . as with the box - counting dimension
, the nature of the scaling behaviour may be different on different length scales and we may then get more than one value for the correlation dimension , each different value being valid over a range of scales .
it is very clear that @xmath44 - which is the probability of finding a point within a sphere of radius @xmath25 centered on another point , is closely related to the volume integral of the two point correlation function . in a situation where the two point correlation function exhibits a power - law behaviour @xmath49 on scales @xmath8
, we expect the correlation dimension to have a value @xmath50 over these scales
. for a mono - fractal the box - counting dimension and the correlation dimension will be the same , and for a homogeneous , space filling point distribution they are both equal to the dimension of the ambient space in which the points are embedded . the box - counting dimension and the correlation dimension quantify different aspects of the scaling behaviour of a point distribution and they will have different values in a generic situation .
the concept of a generalized dimension connects these two definitions and provides a continuous spectrum of dimensions @xmath18 for a range of the parameter @xmath19 .
the definition of the minkowski - bouligand dimension @xmath18 ( @xcite ) closely follows the definition of the correlation dimension the only difference being that we use the @xmath51 moment of the galaxy distribution @xmath52 ( eq . [ eq : gd2 ] ) around any point .
equation ( [ eq : gd3 ] ) can then be generalized to define @xmath53^{q-1 } \label{eq : gd5 } \,.\end{aligned}\ ] ] which is used to define the generalized dimension @xmath54 the quantity @xmath55 may exhibit different scaling behaviour over different ranges of length scales and we will then get more than one spectrum of generalized dimensions each being valid over a different range of length scales . from equations ( [ eq : gd5 ] ) and ( [ eq : gd6 ] ) it is clear that the the generalized dimension @xmath18 corresponds to the correlation dimension at @xmath56 .
in addition @xmath18 corresponds to the box - counting dimension at @xmath57 . for a mono - fractal
the generalized dimension is a constant i.e. @xmath58 which reflects the fact that for a mono - fractal the point distribution is characterized by a unique scaling behaviour . for a generic multi - fractal
the values of @xmath18 will be different for different values of @xmath19 .
the positive values of @xmath19 give more weight - age to the over - dense regions .
thus , for @xmath59 , @xmath18 probes the scaling behaviour of the distribution of points in the over - dense regions like inside clusters etc .
the negative values of @xmath19 , on the other hand , give more weight - age to the under - dense regions and , hence , for negative @xmath19 , @xmath18 probes the scaling behaviour of the distribution of points in the under - dense regions like voids . finally it should be pointed out that the minkowski - bouligand generalized dimension @xmath18 is one of the many possible definitions of a generalized dimension . the minimal spanning tree used by van der weygaert and jones ( @xcite )
is another possible method which can be used .
the minkowski - bouligand generalized dimension has the advantage of being easy to compute .
in addition the various selection effects which have to be taken into account when analyzing redshift surveys can be easily accounted for when determining the minkowski - bouligand generalized dimension and hence we have chosen this particular method for the multi - fractal characterization of the galaxy distribution in lcrs ,
the lcrs consists of 6 alternating slices each subtending @xmath60 in right - ascension and @xmath61 in declination , 3 each in the northern and southern galactic caps centered at @xmath62 and @xmath63 respectively .
the survey extends to a redshift of @xmath64 corresponding to @xmath65 in the radial direction .
the survey contains about 24000 galaxies distributed with a mean redshift of @xmath66 corresponding to @xmath67 .
we next elaborate a little on the shape of the individual slices .
consider two cones both with the same axis and with their vertices at the same point .
let the angle between the first cone and the axis be @xmath68 and the second cone and the axis be @xmath69 so that the angle between the two cones is @xmath61 .
next truncate both the cones at a radial distance of @xmath70 from the vertex .
finally , a slice centered at a declination @xmath71 corresponds to a @xmath60 wedge of the region between these two cones .
the effect of the extrinsic curvature of the cones is small for the three northern slices and we have restricted our analysis to only these three slices for which we have neglected the effect of the curvature .
each slice in the lcrs is made up of @xmath72 x @xmath72 fields some of which were observed with a @xmath73 object fibre system and others with a @xmath74 object fibre system . of the three northern slices the one at @xmath75 is exclusively made up of 112 fibre fields while the slice at @xmath76 is mostly 50 fibre , and the slice at @xmath77 has got both 50 and 112 fibre fields . for each field , redshifts were determined for those galaxies which satisfy the magnitude limits and the central brightness limits of the survey .
these limits are different for the 50 fibre and the 112 fibre fields .
in addition , for those fields where the number of galaxies satisfying the criteria for inclusion in the survey exceeded the number of fibres , the redshifts were determined for only a fraction of the galaxies in the field .
this effect is quantified by the `` galaxy sampling function '' @xmath78 which varies from field to field and is around @xmath79 for the 112 fibre fields and around half this number for the 50 fibre fields .
in addition to the field to field variation of the galaxy sampling function there are two other effects which have to be accounted for when analyzing the galaxy distribution .
they are , ( 1 ) .
apparent magnitude and surface brightness incompleteness , and , ( 2 ) .
central surface brightness selection .
these are quantified by two factors @xmath80 and @xmath81 , respectively , which are discussed in detail in lin _ et al . _
the survey data files provide the product of these three factors @xmath82 for each galaxy and the contribution from the @xmath83 galaxy has to be weighted with the factor @xmath84 when analyzing the survey .
the factor @xmath85 discussed above takes into account the effects of the field - to - field sampling fraction and the incompleteness as a function of the apparent magnitude and central surface brightness .
in addition , the selection function @xmath86 has also to be taken into account , and this depends on both the differential luminosity function @xmath87 and the magnitude limits of the survey .
the luminosity function of lcrs has been studied by lin _
( 1996 ) who have determined the luminosity function for different sub - samples of lcrs .
they find that the schechter form with the parameters @xmath88 , @xmath89 and @xmath90 provides a good fit for the luminosity function in the absolute magnitude range @xmath91 .
they have obtained these parameters from the analysis of the combined northern and southern 112 fibre fields and we shall refer to the schechter luminosity function with these set of parameters as the ns112 luminosity function .
the analysis of lin _ et al . _
( 1996 ) shows that this luminosity function can be used for the northern 50 fibre fields in addition to the northern and southern 112 fibre fields , and we have used the ns112 luminosity function for most of our analysis .
lin _ et al . _
( 1996 ) have also separately provided the luminosity function determined using just the northern 112 fibre fields .
this has the schechter form with the parameters @xmath92 , @xmath93 and @xmath94 and we refer to this as the n112 luminosity function .
we have used this in some of our analysis of the @xmath95 slice which contains only 112 fibre fields . the selection function @xmath96 quantifies the fact that the fraction of the galaxies which are expected to be included in the survey varies with the distance from the observer . for a magnitude limited survey
the apparent magnitude limits @xmath97 and @xmath98 can be converted to absolute magnitude limits @xmath99 and @xmath100 at some redshift @xmath101 .
in addition if we impose further absolute magnitude criteria @xmath102 , then the selection function can be expressed as @xmath103}_{max[m_1(z),m_1 ] } \phi(m ) d m { \big / } \int^{m_2}_{m_1 } \phi(m ) d m \ , .
\label{eq : s2}\ ] ] the apparent magnitude limits are different for the 50 and 112 fibre fields and we have used the appropriate magnitude limits and the n112/ ns112 luminosity functions to calculate the selection function at the redshift of each of the galaxies .
this is then used to calculate a weight factor for each of the galaxies , and the contribution of the @xmath83 galaxy in the survey has to be weighed by @xmath104 another effect that we have to correct for arises because of the fact that we would like to treat the distribution of galaxies in each slice as a two dimensional distribution .
each slice consists of galaxies that are contained within a thin conical shell of thickness @xmath105 and we construct a two dimensional distribution by collapsing the thickness of the slice .
the thickness of each slice increases with the distance from the observer and in order to compensate for this effect we weigh each galaxy by the inverse of the thickness of the slice at its red - shift . taking this effect into account the weight factor gets modified to @xmath106 which we use to weigh the contribution from the @xmath107 galaxy in the lcrs .
we should also point out that through the process of flattening the conical slices and collapsing its thickness , the three dimensional galaxy distribution has been converted to a 2-dimensional distribution and the whole of our multi - fractal analysis is for a planar 2-dimensional point distribution . in our analysis
we have considered various subsamples of lcrs all chosen from the 3 northern slices .
in addition to the apparent magnitude limits of the survey we have imposed further absolute magnitude and redshift cutoffs to construct both volume and apparent magnitude limited subsamples whose details are presented in table i.
we first extract various subsamples of lcrs using the criteria given in table i for each of the subsample . for each subsample
we next calculate the weight function @xmath108 ( equation [ eq : s4 ] ) for all the galaxies in the subsample .
in addition the 3-dimensional distribution of galaxies in the sub - sample is converted into a corresponding 2-dimensional distribution using the steps outlined in the previous section and we finally have a collection of @xmath30 galaxies distributed over a region of a plane .
we next choose @xmath31 of these galaxies at random and count the number of galaxies inside a circle of radius @xmath25 drawn around each of these @xmath31 randomly chosen galaxies .
in determining this we use a modified version of equation ( [ eq : gd2 ] ) where each galaxy in the circle has an extra weight factor @xmath109 as calculated in the previous section , i.e. @xmath110 the different moments of this quantity are averaged over the @xmath31 galaxies to obtain @xmath111 defined in equation ( [ eq : gd5 ] ) for a range of @xmath19 .
the exercise is repeated with circles of different radii ( different values of @xmath25 ) to finally obtain @xmath111 for a large range of @xmath25 .
it should be noted that the region from which the @xmath31 points can be chosen at random depends on the size of the circle which we are considering .
for very large values of @xmath25 a large region around the boundaries of the survey has to be excluded because a circle of radius @xmath25 drawn around a galaxy in that region will extend beyond the boundaries of the survey . as a consequence for large values of @xmath25
we do not have many galaxies which can serve as centers , while for small values of @xmath25 there are many galaxies which can serve as centers for circles of radius @xmath25 .
for @xmath25 between @xmath112 to @xmath1 we use @xmath113 which is of the same order as the the total number of galaxies available for use as centers . to estimate the statistical significance of our results at this range of length - scales we have randomly divided the 60 centers into independent groups of 20 centers and repeated the analysis for each of these .
we have used the variation in the results from the different subsamples to estimate the statistical errors for our results on large scales . in the range
@xmath114 we have used @xmath115 which is only a small fraction of the total number of galaxies which can possibly serve as centers which is around 1500 . at this range of length - scales it is possible to choose many independent sets of 100 centers .
we have performed the analysis for a large number of such sets of 100 centers and these have been used to estimate the mean generalized dimension @xmath18 and the statistical errors in the estimated @xmath18 at small scales .
for both the range of length - scales considered we have tried the analysis making changes in the number of centers and we find that the results do not vary drastically as we vary the number of centers used in the analysis .
the value of the generalized dimension @xmath18 is determined for a fixed value of @xmath19 by looking at the scaling behaviour of @xmath111 as a function of @xmath25 ( _ e.g. _ figures [ scaleq0 ] and [ scaleq2 ] ) we have considered @xmath19 in the range @xmath116 . in principle we could have considered arbitrarily large ( or small ) values of @xmath19 also , but the fact that there are only a finite number of galaxies in the survey implies that only a finite number of the moments can have independent information . this point has been discussed in more detail by bouchet _ et.al . _ ( @xcite ) .
in addition to the subsamples of galaxies listed in table 1 , we have also carried out our analysis for mock versions of these subsamples of galaxies .
the mock versions of each subsample contains the same number of galaxies as the actual subsample .
the galaxies in the mock versions are selected from a homogeneous random distribution using the same selection function and geometry as the actual subsample .
we have carried out the whole analysis for many different random realizations of each of the subsamples listed in table i. the main aim of this exercise was to test the reliability of the method of analysis adopted here .
we first discuss our analysis of the mock subsamples .
since the effect of the selection function and the geometry of the slices have both been included in generating these subsamples , our analysis of these subsamples allows us to check how well these effects are being corrected for . in the ideal situation for all the mock subsamples
we should recover a flat spectrum of generalized dimensions with @xmath117 corresponding to a homogeneous point distribution .
the actual results of the multi - fractal analysis of the mock subsamples are presented below where we separately discuss the behaviour of @xmath18 at small scales @xmath118 and at large scales @xmath119 .
the results for mock versions of the subsample d-12.1 are shown in figure ( [ mock12 ] ) .
this is a magnitude limited subsample from a slice that has only 112 fibre fields and it contains the largest number of galaxies .
we get a nearly flat spectrum with @xmath117 corresponding to a homogeneous point distribution at both small and large scales .
similar results are also obtained for mock versions of the other subsamples of the @xmath75 slice . the analysis of mock versions of the subsample d-03.1 which contains both 112 and 50 fibre fields gives a spectrum with a weak @xmath19 dependence ( figure [ mock3 ] ) .
this effect is more noticeable at small scales than at large scales .
the analysis of mock versions of the d-06.1 subsample ( figure [ mock6 ] ) gives similar results at small scales . at large scales we get a nearly flat curve with @xmath120
this subsample d-06.1 has mostly 50 fibre fields and it has around half the number of galaxies as the d-12.1 subsample .
we thus find that the analysis is most effective for the subsample from the @xmath75 slice where @xmath18 shows very little @xmath19 dependence and @xmath121 .
for the other two slices we find a weak @xmath19 dependence with @xmath122 .
this clearly demonstrates that our method of multi - fractal analysis correctly takes into account the different selection effects and the complicated sampling and geometry for all the subsamples that we have considered .
we next discuss our analysis of the actual data .
the analysis of the curves corresponding to @xmath111 versus @xmath25 for the different subsamples shows the existence of two very different scaling behaviour - one at small scales and another at large scales , with the transition occurring around @xmath112 to @xmath5 .
the scaling behaviour of @xmath111 is shown in figure [ scaleq0 ] and figure [ scaleq2 ] for @xmath123 and @xmath56 , respectively for the subsample d-12.1 the other subsamples all exhibit a similar behaviour .
based on this we have treated the scales @xmath124 ( small scales ) and @xmath125 ( large scales ) separately and the multi - fractal analysis has been performed separately for the small and large scales .
figures [ dq12 ] , [ dq3 ] and [ dq6 ] show the spectrum of generalized dimensions @xmath126 at both small and large scales for three of the subsamples .
we find that at small scales the plots of @xmath18 versus @xmath19 for the actual data ( figures [ dq12 ] , [ dq3 ] , [ dq6 ] ) are quite different from the corresponding plots for the mock versions of the data ( figures [ mock12 ] , [ mock3 ] , [ mock6 ] ) .
this clearly shows that the distribution of galaxies is not homogeneous over the scales @xmath127 .
in addition we find that all the subsamples exhibit a multi - fractal behaviour over this range of length - scales .
the interpretation of the different values of the multi - fractal dimension @xmath18 is complicated by the geometry of the survey and we do not attempt this here . at large scales
the behaviour of the generalized dimension @xmath18 is quite different .
for the subsample d-12.1 the spectrum shows a weak @xmath19 dependence ( figure [ dq12 ] ) and @xmath18 shows a gradual change from @xmath128 to @xmath129 as @xmath19 varies from @xmath130 to @xmath131 .
this is quite different from the behaviour at small scales where the change in @xmath18 is larger and more abrupt .
the behaviour of the other subsamples of the @xmath95 slice are similar . for the subsample d-03.1
we find that the spectrum is nearly flat ( figure [ dq3 ] ) with @xmath128 and for d-06.1 ( figure [ dq6 ] ) the spectrum is nearly flat with @xmath120 .
these values are within the range we recover from our analysis of the mock subsamples which are constructed from an underlying random homogeneous distribution of galaxies .
this agreement between the actual data and the random realizations with @xmath132 in all the subsamples shows that the distribution of galaxies in lcrs is homogeneous at the large scales .
the work presented here contains significant improvements on the earlier work of amendola & palladino ( 1999 ) on two counts and these are explained below : + ( 1 ) . unlike the earlier work which has analyzed volume limited subsamples of one of the slices ( @xmath133 ) of the lcrs we have analyzed both volume and magnitude limited subsamples of all the three northern slices of the lcrs .
the magnitude limited samples contain more than four times the number of galaxies in the volume limited samples and they extend to higher redshifts .
this allows us to make better use of the data in the lcrs to improve the statistical significance of the results and to probe scales larger than those studied in the previous analysis .
we have calculated the full spectrum of generalized dimensions which has information about the nature of clustering in different environments .
the integrated conditional density used by the earlier workers is equivalent to a particular point @xmath134 on the spectrum and it does not fully characterize the scaling properties of the distribution of galaxies .
here we present a method for carrying out the multi - fractal analysis of both magnitude and volume limited subsamples of the lcrs .
our method takes into account the various selection effects and the complicated geometry of the survey .
we first apply our method to random realizations of the lcrs subsamples for which we ideally expect a flat spectrum of generalized dimensions with @xmath135 .
our analysis gives a nearly flat spectrum with @xmath136 on large scales .
the deviation from the expected value includes statistical errors arising from the finite number of galaxies and systematic errors arising from our treatment of the selection effects and the complicated geometry .
the fact that the errors are small clearly shows that our method correctly accounts for these effects .
our analysis of the actual data shows the existence of two different regimes and the distribution of galaxies on scales @xmath137 shows clear indication of a multi - fractal scaling behaviour . on large scales
@xmath138 we find a nearly flat spectrum with @xmath139 .
this is consistent with our analysis of the random realizations which have been constructed from a homogeneous underlying distribution of galaxies .
based on the above analysis we conclude that the distribution of galaxies in the las campanas redshift survey is homogeneous at large scales with the transition to homogeneity occurring somewhere around @xmath3 to @xmath5 .
trs would like to thank thank t. padmanabhan , k. subramanian , j. s. bagla , f. s. labini and l. pietronero for several useful discussions .
akg and trs gratefully acknowledge the project grant ( sp / s2/009/94 ) from the department of science and technology , india .
all the authors are extremely grateful to the lcrs team for making the catalogue publicly available .
ddd 9999 amendola l. and palladino e. , 1999 , ap.j .
, in press , astro - ph/9901420 borgani s. , 1995 , _ phys .
rep . _ * 251 * , 1 bouchet f. r. , schaeffer r. and davis m. , 1991 , _ ap j _ , * 383 * , 19 cappi a. , benoist c. , da costa l.n .
, maurogordato s. , 1998 , astro - ph/9804085 coleman p h. and pietronero l. , 1992 , _ phys .
rep . _ * 213 * , 311 .
falconer k , 1990 in _ fractal geometry : mathematical foundations and applications , john wiley .
_ feder j , 1989 , in _ fractals , plenum press
. _ guzzo , 1998 , _ new astronomy _ * 2 * , 517 .
labini f. sylos & m. montuori , 1997 , astro - ph/9711134 labini f s , montuori m. and pietronero l. , 1998 , _ phys .
* 293 * , 61 .
lin , h. , kirshner , r. p. , shectman , s.a . ,
landy , s. d. , oemler , a. , tucker , d. l. and schechter , p. l. 1996 , _ apj _ , * 471 * , 617 .
martinez v j. , 1991 , in _ applying fractals in astronomy , eds .
a. heck and j.m.perdang , springer - verlag , page 135 .
_ martinez v j. and jones b j t. , 1990 , _ mnras _ * 242 * , 517 .
padmanabhan t. , 1992 in _ structure formation in the universe , cambridge university press .
_ peebles p. j. e. , 1998 , preprint , astro - ph 9806201 peebles p.
j. e. , 1993 , in _ principles of physical cosmology , princeton university press _
shectman , s. a. , landy , s. d. , oemler , a. , tucker , d. l. , lin , h. , kirshner , r. p. and schechter , p. l. 1996 , _ apj _ , * 470 * , 172 .
van der weygaert r , and jones b j t. , 1992 , _ phys .
lett . _ * a169 * , 145 .
wu kelvin k. s. , ofer lahav , martin j. rees , , 1998 , preprint , astro - ph/9804062 lllllll & & & absolute&luminosity&number of&vol./mag
. + subsample&@xmath140&z range & magnitude range & function&galaxies & limited + + d-12.1&@xmath14112.0&0.017 - 0.2&@xmath14123.0 - @xmath14117.5&ns112&4458&m + d-12.2&@xmath14112.0&0.017 - 0.2&@xmath14123.0 - @xmath14117.5&n112&4458&m + d-12.3&@xmath14112.0&0.05 - 0.1&@xmath14121.0 - @xmath14120.0&n112&869 & v + d-12.4&@xmath14112.0&0.065 - 0.125&@xmath14121.5 - @xmath14120.5&n112&923&v + d-06.1&@xmath1416.0&0.017 - 0.2&@xmath14123.0 - @xmath14117.5&ns112&2316&m + d-03.1&@xmath1413.0&0.017 - 0.2&@xmath14123.0 - @xmath14117.5&ns112&4055&m + | we have carried out a multi - fractal analysis of the distribution of galaxies in the three northern slices of the las campanas redshift survey .
our method takes into account the selection effects and the complicated geometry of the survey . in this analysis
we have studied the scaling properties of the distribution of galaxies on length scales from @xmath0 to @xmath1 .
our main results are : ( 1 ) the distribution of galaxies exhibits a multi - fractal scaling behaviour over the scales @xmath2 to @xmath3 , and , ( 2 ) the distribution is homogeneous on the scales @xmath4 to @xmath1 .
we conclude that the universe is homogeneous at large scales and the transition to homogeneity occurs somewhere in the range @xmath3 to @xmath5 . |
the u - cas is an ingenious device that hovers in mid - air while spinning .
it is marketed as a kit in japan under the trade name u - cas @xcite , and in the u.s.a . and
europe under the trade name levitron@xmath14 @xcite .
the whole kit consists of three main parts : a magnetized top which weighs about @xmath15 gr , a thin ( lifting ) plastic plate and a magnetized square base plate ( base ) . to operate the top
one should set it spinning on the plastic plate that covers the base .
the plastic plate is then raised slowly with the top until a point is reached in which the top leaves the plate and spins in mid - air above the base for about 2 min .
the hovering height of the top is approximately @xmath16 cm above the surface of the base whose dimensions are about 10 cm @xmath17 10 cm @xmath17 2 cm .
the kit comes with extra brass and plastic fine tuning weights , as the apparatus is very sensitive to the weight of the top .
it also comes with two wedges to balance the base horizontally .
the physical principle underlying the operation of the u - cas relies on the so - called ` adiabatic approximation ' @xcite : as the top is launched , its magnetic moment points _ antiparallel _ to the magnetization of the base in order to supply the repulsive magnetic force which will act against the gravitational pull . as the top hovers , it experiences lateral oscillations which are slow ( @xmath18 hz ) compared to its precession ( @xmath19 hz ) .
the latter itself , is small compared to the top s spin ( @xmath20 hz ) . since @xmath21 the top is considered ` fast ' and acts like a classical spin .
furthermore , as @xmath22 this spin may be considered as experiencing a _ slowly _ rotating magnetic field . under these circumstances the spin precesses around the _ local _ direction of the magnetic field @xmath23 ( adiabatic approximation ) and , on the average , its magnetic moment @xmath24 points _ antiparallel _ to the local magnetic field lines . in view of this discussion , the magnetic interaction energy which is normally given by @xmath25
is now given approximately by @xmath26 .
thus , the overall effective energy ` seen ' by the top is @xmath27 where @xmath10 is the mass of the top , @xmath11 is the free - fall acceleration and @xmath28 is the height of the top above the base . by virtue of the adiabatic approximation , two of the three rotational degrees of freedom
are coupled to the transverse translational degrees of freedom , and as a result the rotation of the axis of the top is already incorporated in eq.([energy ] ) .
thus , under the adiabatic approximation , the top may be considered as a _ point - like _ particle whose only degrees of freedom are translational .
the important point of this discussion is the following : the energy expression written above _ possesses a minimum _ for certain values of @xmath29 .
thus , when the mass is properly tuned , the apparatus acts as a trap , and stable hovering becomes possible .
a detailed description of this device , extending beyond the adiabatic approximation , may be found elsewhere @xcite .
for the purpose of this paper , the adiabatic approximation will suffice .
in this paper we focus on the question : @xmath30 first , we define what do we mean by ` height ' : we assume that the top hovers above some horizontal plane . the height of the top is measured with respect to that plane .
the ` rules of the game ' are , that we can put permanent magnet _ below _ the plane , but never _ above _ this plane . whatever is below this plane ,
will be called ` the base ' .
the method we use to answer the question above will be by getting upper and lower _
bounds _ for this height , denoted by @xmath0 .
we begin by pointing out the factors that limit the hovering height of the u - cas .
first , we set aside the question of stability and assume that the top is guided by a vertical axis , allowing it to move only axially ( up and down ) . in this case ,
the easiest way to increase @xmath13the hovering height , is by using a more powerful magnet for the base .
this technique however , can not be applied indefinitely since there is no way to increase the magnetization of a given substance without limit .
we are therefore forced to limit the strength of @xmath31the magnetization density of the base , to some maximum value , say @xmath9 , and design the shape and magnetization density of the base so as to maximize @xmath13 .
one more factor that limits the levitation height is the amount of magnetic substance , or _
, in our disposal . clearly , the more material we use , the larger is the levitation height that can be achieved .
but still , even if the volume is infinite the levitation height is bounded , since @xmath32 is bounded .
when stability is taken into account , things get more complicated .
stability sets another limitation on the design of the base , in _ addition _ to the limitations that were discussed above : a brief look at eq.([energy ] ) tells us that for the top to hover _ stably _ over the plate , the effective energy @xmath33 should possess a _
minimum_. this means , in particular , that @xmath34 where the derivatives are evaluated at the equilibrium position of the top . since these are _ homogenous _ inequalities , it is clear that the region in space where a minimum _ may _ occur , does not depend on the _ strength _ of the magnetic field . as a consequence ,
the hovering height of the u - cas is _ not _ determined by the strength of the magnetic field but on its _ geometry , _ or alternatively , by the _ shape _ of the base . as an example , it has been shown @xcite that for a vertically magnetized base in the shape of a disk of radius @xmath35 , the range of heights @xmath13 for which stable hovering is possible , is very narrow and is around @xmath36 .
this result agrees roughly with the parameters of the u - cas , for which @xmath37 cm and @xmath38 cm , when @xmath13 is measured from the center - of - mass of the base .
the _ strength _ of the field then comes into play by tuning it in order to achieve equilibrium for a given mass of the top .
equilibrium prevails when the total force on the top vanishes , i.e. when @xmath39 both in the guided case and in the stable case , we see that in order to increase the hovering height of the u - cas , we should design a better base .
we do expect however , that the hovering height can not be increased indefinitely . in connection with the redesigning of the base , we have recently shown @xcite , both theoretically and experimentally , that the use of a vertically magnetized _ ring _ of radius @xmath35 as a base , increases the hovering height by more than three times to about @xmath40!. this increase in height did not came without cost , as the tolerance on the mass of the top became more stringent , being @xmath41 for the ring vs. @xmath42 in the case of the disk @xcite .
all the calculations that we do are outlined in sec.([sec2 ] ) .
we will now describe what we calculate and what is the motivation behind it .
we start in sec.([sec2.1 ] ) , by deriving a close form expression for the first and second derivatives of the magnetic field , in terms of the magnetization density of the base .
the first derivative is the magnetic force on the top , which is used throughout the paper , while the second derivative is exploited when we discuss stability . in sec.([sec2.2 ] ) we consider the problem of maximizing the hovering height in the case where the _ volume _ of the base may be _ infinite _ , and may be _ arbitrarily _ magnetized under the constraint that @xmath43 .
we still _ do not _ require that the top be stable against lateral translations , and assume that it is guided along a vertical axis .
we show that in this case @xmath44 where @xmath5 is a characteristic height for the problems that are discussed in this paper . it is defined as @xmath45 where @xmath7 is the magnetic permeability of the vacuum
, @xmath8 is the magnetic moment of the top , @xmath10 is the mass of the top , and @xmath11 is the free - fall acceleration . in sec.([sec2.3 ] ) we consider the same problem as before but for a base which is _ uniformly _ magnetized along the vertical direction , namely @xmath46 , where @xmath47 is a unit vector in the vertical direction . in this case
we find that @xmath48 in order to arrive to better bounds on @xmath0 , our next step is to limit the amount of material available for the base .
thus , in sec.([sec2.4 ] ) we maximize the hovering height in the case where the base may be _ arbitrarily _ magnetized under the constraint that @xmath43 , and that its _ volume _ @xmath49 is given . here , the result is given in the form of a plot , showing the dependence @xmath50 on @xmath49 . in the limit @xmath51 , we recover the result of sec.([sec2.2 ] ) .
our derivation is also _ constructive _ in the sense that it shows how to _ construct _ the optimal base for a given @xmath49 . in this paper
however , we do not discuss this matter thoroughly due to space shortage . for completeness
, we study in sec.([sec2.5 ] ) the same problem as in sec.([sec2.4 ] ) , but for a _ uniformly _ magnetized base . here , again we find the dependence of @xmath50 on @xmath49 and recover the result of sec.([sec2.3 ] ) .
sec.([sec2.6 ] ) is the first part where stability is added into play : we begin by explaining how to _ test _ for stability of the top under the adiabatic approximation , and study the possible levitation heights of a _ uniformly _ magnetized base whose shape is _ cylindrical _ of volume @xmath49 . as the stability condition is given in the form of inequality relations ( see eq.([stab ] ) )
, it is not suffice to determine uniquely the levitation height .
we therefore choose to study two particular stable points : an _ isotropic _ stable point and the _ highest _ possible stable point .
the meaning of these points would become clear later . in this case
we also give a plot , showing the dependence of @xmath50 on @xmath49 .
this result should also be considered as a lower bound for @xmath0 , since the cylindrical base is a special case of all the base configurations that are possible . in section([sec2.7 ] ) we take one step further , and generalize the result of section([sec2.6 ] ) to the case of a base which may be _ arbitrarily _ magnetized , and look for the _ lowest _ possible stable point . in this case however , we assume that @xmath52 , to ease the solution of the problem . we also show that in this case , where @xmath53 and @xmath52 , the _ lowest _ possible stable point and the _ highest _ one coincide , so this result represent not a _ bound _ for @xmath0 , but is actually @xmath0 itself , for the case @xmath52 . in sec.([sec3 ] ) we discuss interesting aspects of our results .
in particular we estimate the value of @xmath5 for modern permanent magnet materials and calculate the values of the bounds for @xmath0 .
we comment on the implications of our results and discuss other related questions .
a simplified model of the u - cas is shown in fig.([fig1 ] ) .
it consists of a point - like particle of mass @xmath10 and magnetic moment @xmath8 ( pointing downward ) , hovering at a height @xmath13 above the @xmath54 plane . for the moment
, we consider the top as if it was _ guided _ along the @xmath28-axis so that only its vertical motion is allowed .
other degrees of freedom are considered ` frozen ' .
the region @xmath55 may be partially or wholly filled with a magnetized substance ( `` base '' ) , whose magnetization density is denoted by @xmath31 , producing a magnetic field throughout space which we denote by @xmath56 . in what follows ,
we assume that the base ( and hence the magnetic field ) has a cylindrical symmetry around the @xmath28-axis .
thus , along this axis the magnetic field possesses only a @xmath28-component .
we further assume that this component is _ positive _ , namely that the magnetic field along the @xmath28-axis is pointing upward . using eqs.([energy]),([force ] )
we find that , when the top is in equilibrium , the total vertical component of the force on the top vanishes , @xmath57 where the term @xmath58 is the gravitational force pulling the top towards the base , and the term @xmath59 is the magnetic force which pushes the top upward .
note that @xmath60 should be _
since the top is allowed to move only along the @xmath28-axis , we may write @xmath61 ( in what follows we deserve the notation @xmath62 to denote the @xmath28-component of the magnetic field along the @xmath28-axis , namely @xmath63 ) .
now , the magnetic force on the top simplifies to @xmath64 and hence , when the top is in equilibrium at @xmath65 , @xmath66 reciprocity allows us to express @xmath67 in terms of @xmath31 as @xmath68 here , @xmath69 is the field at the point @xmath70 produced by a unit magnitude dipole pointing _
upward _ , located at a height @xmath28 along the @xmath28-axis . taking @xmath71 and @xmath72 as depicted in fig.([fig1 ] ) , we can write @xmath73 as @xmath74 where @xmath75 is a polar unit vector , also defined in fig.([fig1 ] ) , and @xmath7 is the magnetic permeability of the vacuum .
it is worth to note that @xmath76 is nothing but the field produced at @xmath70 by a _
quadruple _ located at a height @xmath28 along the @xmath28-axis .
this quadruple is made out of a pair of identical dipoles : one is located at @xmath28 and pointing downward and the other is at an infinitesimally higher position @xmath77 and pointing upward , with their magnetic moment being @xmath78 . when eq.([eq1.3 ] ) is substituted into eq.([eq1.2 ] ) we find that , at @xmath65 , the field derivative is given by @xmath79 where @xmath80 and where @xmath81 is a polar unit vector orthogonal to @xmath75 , as is shown in fig.([fig1 ] ) .
we would also need the _ second _ derivative of the field , when we discuss stability .
it is given by @xmath82 where @xmath83 note that in both eqs.([eq1.4 ] ) and ( [ eq1.41 ] ) , @xmath84 is defined with respect to a polar coordinate system whose origin is at @xmath65 and _ not _ at @xmath54 .
also note , that in going from eq.([eq1.4 ] ) to eq.([eq1.41 ] ) , one should take into account the dependence of the unit vectors @xmath81 and @xmath85 on @xmath28 .
we consider an infinitesimal volume element at some point @xmath86 within the base .
it is essentially a magnetic dipole whose magnitude is @xmath9 , and whose direction we will now find by the requirement that the magnetic force on the top is maximized .
we have already shown that the _ force _ on the top , contributed by one elemental dipole , is proportional to the magnetic field @xmath73 that would be produced at @xmath86 by a quadruple located at @xmath65 . to make this force maximal we assign to each point @xmath86 in @xmath55 , a magnetization density @xmath31
whose magnitude is @xmath9 and whose direction is _ antiparallel _ to the field line of @xmath73 at that point .
the magnetization density so defined will _
maximize _ the force on the top , and is therefore the requested answer .
mathematically , this procedure amounts to replacing @xmath31 in eq.([eq1.4 ] ) by @xmath87 , where @xmath88 is a unit vector _
parallel _ to @xmath89 , the latter being defined in eq.([n ] ) .
the result is @xmath90 for which the @xmath71 integration is trivial and the @xmath72 integration may be brought to a simpler form by the transformation @xmath91 .
this gives @xcite @xmath92 which together with eq.([eq1.1 ] ) shows that @xmath93 where @xmath94 is the characteristic length in our problem which will reappear in the next sections .
the value of @xmath5 may be interpreted as the _ distance _ between two colinear dipoles , one of them is of strength @xmath8 while the other is of strength @xmath95 , for which the mutual force between them is @xmath96 .
eq.([eq1.6 ] ) shows that even though the magnetization direction is allowed to vary everywhere inside the base , the levitation height @xmath13 is bounded .
eq.([eq1.6 ] ) presents the maximum height that can be accomplished with a given substance provided that the top is not allowed to move laterally .
clearly , it also serve as an _ upper _ bound for @xmath0 , as stability was not considered yet .
moreover , note that we have calculated the maximum magnetic force , which is proportional to @xmath67 .
the magnetic field @xmath97 on the other hand , becomes _ infinite _ at @xmath98 as can be seen by the following simple argument :
since each elemental dipole within the plate contributes a field that goes as @xmath99 and since the volume of integration goes as @xmath100 , the integrand goes as @xmath101 , for which the integral diverges as @xmath102 .
later we show that the divergence of @xmath97 implies that the top _ can not _ be stable when placed in such a point . in any case , it would be impossible to _ spin _ the top .
uniformly magnetized plates are clearly easier to construct . in this section
we find out what upper bound does this restriction sets on the highest levitation point . in another sense ,
the result of this section may also be considered as a _
lower _ bound on the height of levitation ( for a top which is guided ! ) , provided one considers all possible base configurations .
we again consider an infinitesimal volume element within the base .
it is now oriented along the @xmath28-direction and interacting with the top s magnetic dipole .
if this element is exactly below the top ( i.e. it is located somewhere on the negative * * * * @xmath28-axis ) , it exerts a repelling @xmath28-directed force on the top , pushing it away . if the element is not exactly below the top then the nature of this force ( i.e. weather it is repulsive or attractive ) is determined by the angle formed between the direction of the two dipoles ( in this case both point in the @xmath28-direction ) , and the direction defined by the line joining these dipoles .
we call this angle @xmath72 , and measure it with respect to the positive @xmath28-direction , as is shown in fig.[fig1 ] .
there exists a critical angle @xmath103 ( see fig.([fig2 ] ) ) , for which the @xmath28-directed force vanishes such that for @xmath104 the force becomes attractive .
it is therefore useless to put any material within the region @xmath104 and @xmath105 since this would _ reduce _ the magnetic force and decrease the levitation height . to find @xmath103 we eliminate the @xmath28-component of eq.([n ] ) , and find the value of @xmath72 for which it vanishes
we pick the value of @xmath72 that lies between @xmath106 and @xmath107 .
this gives @xmath108 we now evaluate the upward force exerted on the top , by summing the contributions of _ all _ the elements in the base located within @xmath109 . the ( uniform ) magnetization density within this region is taken to be @xmath46 .
using eq.([eq1.4 ] ) , we find that @xmath110\text{.}%\end{aligned}\ ] ] the integration over @xmath71 is again trivial , and we are left with an integration over @xmath72 .
the latter is brought to a simpler form by changing the @xmath72 variable into @xmath111 , hence @xmath112 this result , together with the equilibrium condition eq.([eq1.1 ] ) , suggests that in this case the maximal levitation height possible is @xmath113 \equiv4\pi\left ( \dfrac{3}{5}\right ) ^{5/2}l_{0}\cong3.5l_{0}\text{. } \label{eq1.5}%\ ] ] note , however , that in order to realize the levitation heights found in this section and in the previous one we would need an unlimited supply of magnetic material . in the following sections we find
how good can we do when the _ volume _ of the base is constrained . under a given volume @xmath49
, we find the optimum _ shape _ and _ magnetization _ of the base , that will maximize the levitation height .
we use the lagrange s multipliers method to treat this variational problem @xcite .
first , we parametrize the _ shape _ of the plate : let @xmath114 be the upper integration radius , as is shown in fig.([fig3 ] ) . the lower integration radius is @xmath115 . utilizing the cylindrical symmetry of the problem
, the volume may be written as @xmath116{c}% v=2\pi% % tcimacro{\dint \limits_{\theta_{1}}^{\pi}}% % beginexpansion { \displaystyle\int\limits_{\theta_{1}}^{\pi } } % endexpansion \sin\theta d\theta% % tcimacro{\dint \limits_{-h/\cos\theta}^{r_{e}(\theta)}}% % beginexpansion { \displaystyle\int\limits_{-h/\cos\theta}^{r_{e}(\theta ) } } % endexpansion r^{2}dr\\ = \dfrac{2\pi}{3}% % tcimacro{\dint \limits_{\theta_{1}}^{\pi}}% % beginexpansion { \displaystyle\int\limits_{\theta_{1}}^{\pi } } % endexpansion d\theta\sin\theta\left ( r_{e}^{3}(\theta)+\dfrac{h^{3}}{\cos^{3}\theta } \right ) , \end{array } \label{eq1.7}%\ ] ] where the angle @xmath117 is determined by the condition that @xmath118 and is the value of the angle to the upper - right corner of the base , as is shown in fig.([fig3 ] ) .
the upward force eq.([eq1.4 ] ) , which in this case is given by setting @xmath119 is obtained from @xmath120 according to the lagrange s multipliers method @xcite , the target function that is to be maximized is @xmath121 where @xmath122 is the lagrange s multiplier and @xmath49 is the given volume of the base . the target function is a functional of @xmath114 .
we require that a variation of @xmath123 with respect to @xmath114 would vanish , thus @xmath124 + \frac{1}{\lambda^{4}}\frac{\delta v}{\delta r_{e}(\theta)}=0 . \label{eq1.11}%\ ] ] using eqs.([eq1.8]),([eq1.9 ] ) we find that @xmath125 therefore , the required parametrization @xmath114 is obtained by equating the integrand in eq.([eq1.12 ] ) to zero , giving @xmath126 ^{1/4}. \label{eq1.13}%\ ] ] eq.([eq1.13 ] ) defines the _ shape _ of the optimal base .
we see that it is universal in the sense that as @xmath49 is changed , the optimal new plate s shape is only a scaled version of the original one .
recall however , that @xmath117 is _
different_. the value of @xmath117 , determined by eq.([eq1.8 ] ) , may be combined with eq.([eq1.13 ] ) to read @xmath127 ^{1/4}=f^{-1}\left ( \theta_{1}\right ) , \label{eq1.14}%\ ] ] which implicitly expresses @xmath117 in terms of @xmath128 .
we denote this function by @xmath129 , and write @xmath130 in addition , we use eq.([eq1.13 ] ) to arrive to an explicit expression for the volume @xmath49 , given in eq.([eq1.7 ] ) : @xmath131 where@xmath132{c}% \left [ 3\sqrt{4\cos^{4}\theta+\sin^{4}\theta}\right ] ^{3/4}\\ + \dfrac{x^{3}}{\cos^{3}\theta}% \end{array } \right ) \text{.}%\ ] ] similarly , applying eq.([eq1.13 ] ) to eq.([eq1.9 ] ) yields @xmath133 with the definition@xmath134 \text{.}%\end{aligned}\ ] ] using eq.([eq1.16 ] ) with the equilibrium condition eq.([eq1.1 ] ) , we find that @xmath135 and with few more steps we arrive to @xmath116{c}% \dfrac{h}{l_{0}}=\dfrac{h}{\lambda}s\left ( \dfrac{h}{\lambda}\right ) , \\ \dfrac{v_{0}}{l_{0}^{3}}=g\left ( \dfrac{h}{\lambda}\right ) s^{3}\left ( \dfrac{h}{\lambda}\right ) . \end{array } \label{eq1.18}%\ ] ] we see that @xmath136 depends on @xmath137 through an intermediate variable @xmath128 .
it can be solved numerically by `` running '' over a wide range of @xmath128 and evaluating @xmath136 and @xmath137 for each of its values .
the result is given by the dash - dotted line in fig.([plots ] ) .
note that at the limit @xmath51 we find that @xmath138 , in agreement with the result of section([sec2.2 ] ) .
the asymptotic behavior of @xmath13 as @xmath139 is quite interesting also : this limit may be evaluated from the above equations but it is much simpler ( and more instructive ) to use the following argument : at the low volume limit , the base may be considered as a dipole centered at the origin . such an assumption is valid only if @xmath140 . in this case
it is easy to see that the magnetic force , acting on the top , is just @xmath141 and hence @xmath142 using the definition of @xmath5 we may rewrite the last result as @xmath143 ^{1/4}.\ ] ] we thus conclude that @xmath144 ^{1/4}. \label{eq1.19}%\ ] ] note that since @xmath145 we see that our assumption is confirmed .
the asymptotic line of eq.([eq1.19 ] ) is also plotted in fig.([plots ] ) by the dash - dot - dotted line .
it is conspicuous that it is indeed an asymptotic .
the solution for this case is essentially similar to the solution presented in the previous section .
the only difference is in the form of the magnetization that is used .
here we take @xmath46 instead of @xmath146 .
the relation between @xmath136 and @xmath137 is again given by eqs.([eq1.18 ] ) with the following new definitions : @xmath116{c}% f^{-1}(\theta)\equiv\cos\theta\left [ 3\cos\theta\left ( 5\sin^{2}% \theta-2\right ) \right ] ^{1/4},\\ g(x)\equiv\dfrac{2\pi}{3}% % tcimacro{\dint \limits_{f(x)}^{\pi}}% % beginexpansion { \displaystyle\int\limits_{f(x)}^{\pi } } % endexpansion d\theta\sin\theta\left ( \begin{array } [ c]{c}% \left [ 3\cos\theta\left ( 5\sin^{2}\theta-2\right ) \right ] ^{3/4}\\ + \dfrac{x^{3}}{\cos^{3}\theta}% \end{array } \right ) , \\
s(x)\equiv6\pi% % tcimacro{\dint \limits_{f(x)}^{\pi}}% % beginexpansion { \displaystyle\int\limits_{f(x)}^{\pi } } % endexpansion d\theta\sin\theta\cos\theta\left ( 5\sin^{2}\theta-2\right ) \\ \times\left [ \dfrac{-1}{\left [ 3\cos\theta\left ( 5\sin^{2}\theta-2\right ) \right ] ^{1/4}}-\dfrac{\cos\theta}{x}\right ] .
\end{array}\ ] ] the result of this calculation is given by the dotted line in fig.([plots ] ) .
note that in this case @xmath147 as @xmath51 , which is again in agreement with the result of section([sec2.3 ] ) .
the asymptotic behavior of @xmath13 as @xmath139 is identical to the result that was found in the previous section . in this case
we take the base to be a uniformly magnetized cylinder with a magnetization @xmath46 .
the radius of the cylinder is @xmath35 and its thickness is @xmath148 , with its upper base at the @xmath54 plane , as is shown in fig.([fig4 ] ) .
the magnetic field outside the cylinder is essentially the field outside a solenoid with similar dimensions .
thus , the @xmath28-component of the magnetic field along the @xmath28-axis is given by@xcite : @xmath149 where@xmath150 \text{.}%\ ] ] we also define the function @xmath151 , to be used later , as@xmath152 we now formulate the conditions for the _ stability _ of the spinning top against both axial and lateral translations : stability in the @xmath28-direction is determined by the sign of the ` spring - constant ' in that direction @xmath153 . under the adiabatic approximation ,
the latter is given by the _ curvature _ of the effective energy along that direction .
hence , @xmath154 where @xmath155 is the radial distance , in cylindrical coordinate system , from the @xmath28-axis .
similarly , stability in the lateral direction @xmath156 , is governed by the sign of @xmath157 , \nonumber\end{aligned}\ ] ] where in the last equality use has been made of the cylindrical symmetry of the magnetic field and the fact that @xmath62 and all of its cartesian derivatives are harmonic functions @xcite
. for the spinning top to be stable against translations , both @xmath153 and @xmath158 should be _ positive_. comparing eq.([eq1.31 ] ) to eq.([eq1.30 ] ) , we see that when @xmath159 , then @xmath158 @xmath160 @xmath161 , and therefore one of the pair @xmath162,@xmath163 _ must _ be negative .
thus , when the magnetic field diverges at a point , a top placed at that point can not be stable .
this proves that the highest hovering height , found in section ( [ sec2.2 ] ) , is not under stable conditions .
the restriction that both @xmath153 and @xmath158 should be positive defines a stable _ region _ along the @xmath28 axis . as an example
it can be shown @xcite , that for a base in the shape of a thin disk of radius @xmath164 , the value of @xmath153 is positive whenever @xmath165 , whereas @xmath158 is positive for @xmath166 .
the region of stability in this case is therefore @xmath167
. within that region there exists a point @xmath168 for which @xmath169 . in the case of the disk
it is @xmath170 .
we call this point the _ isotropically stable _ point because the restoring ( stabilizing ) force which acts on a top , which is tuned to hover at @xmath171 is isotropic , depending only on the deviation from the equilibrium position and not on its direction . for a general cylinder
we now consider two distinct situations : the first is the one in which the stable point is _ isotropic _ , i.e. a hovering height @xmath13 for which @xmath172 .
the second is the case where @xmath13 is at the verge of stability in the lateral direction .
this is also the _ highest _ stable point , characterized by @xmath173 . using eqs.([eq1.30]),([eq1.31 ] )
we write each of the two distinct conditions as @xmath174 where @xmath175 for the first situation and @xmath176 for the second . substituting eq.([eq1.27 ] ) into eq.([eq1.32 ] ) defines a functional relationship between @xmath177 and @xmath178 denoted by the function @xmath179,@xmath180 differentiating eq.([eq1.27 ] ) with respect to @xmath28 , setting @xmath65 , and using the equilibrium condition eq.([eq1.1 ] ) , gives @xmath181 where@xmath182 combining it with an expression for the volume , gives @xmath116{c}% \dfrac{v_{0}}{l_{0}^{3}}=\dfrac{\pi r^{2}d}{l_{0}^{3}}=\pi\left ( \dfrac{d}% { r}\right ) \left ( \dfrac{r}{l_{0}}\right ) ^{3}=\pi\dfrac{d}{r}n^{3}\left ( d / r\right ) , \\ \dfrac{h}{l_{0}}=\dfrac{h}{r}\dfrac{r}{l_{0}}=g(d / r)n\left ( d / r\right ) , \end{array } \label{eq1.33}%\ ] ] which expresses the volume @xmath49 and the levitation height @xmath13 in terms of a common variable @xmath178 .
running over @xmath178 , and evaluating the volume and height according to eqs.([eq1.33 ] ) , furnishes the required relation between @xmath136 and @xmath137 .
this plot is shown in fig.([plots ] ) where the solid line corresponds to @xmath176 ( the isotropic case ) and the dotted line corresponds to @xmath175 ( the highest stable point ) .
note that both of these plots are _ not _ monotonically increasing .
they possess a maximum of @xmath13 at some optimal volume @xmath183 . for @xmath175
we find that @xmath184 and @xmath185 , whereas for @xmath176 these are @xmath186 and @xmath187 .
this indicates that using too a much material _
worsen _ the largest height that can be achieved , which is reminiscent of our conclusion of sec.([sec2.3 ] ) .
if the given volume is larger than @xmath183 however , we can always use only an amount of volume equal to @xmath183 and discard the rest of the material .
hence , in principal at least , one _ can _ realize the largest possible height , which is why the plots of @xmath13 vs. @xmath49 had been artificially corrected by assigning the maximum value of @xmath13 for the values of @xmath49 that are larger than @xmath183 .
in this last section we consider the case of an infinite base , which may be arbitrarily magnetized , such that the height of levitation is maximized , yet the top is stable against both axial and lateral translations . in order to solve this problem
, one needs to maximize the magnetic force @xmath188 under the constraints that @xmath153 and @xmath158 , defined in eqs.([eq1.30 ] ) and ( [ eq1.31 ] ) respectively , are _ both _ positive .
the last requirement however , results in a non - linear inequality , which can not be solved analytically .
the method we take here is to use the constraint @xmath189 instead , which marks the lower end of the stability region along the @xmath28-axis .
consider an infinitesimal magnetic dipole at the point @xmath86 within the base with magnetization @xmath9 .
it is situated below the expected hovering position , as is shown in fig.([fig5 ] ) , and its direction makes an angle @xmath190 with the line joining the dipole to the equilibrium position of the top . substituting @xmath191 into eqs.([eq1.4]),([eq1.41 ] ) , gives @xmath192{c}% -6\sin\alpha\sin\theta\cos\theta\\ -3(3\cos^{2}\theta-1)\cos\alpha \end{array } \right\ } , \nonumber\end{aligned}\ ] ] and @xmath193{c}% -9\sin\theta\left ( 5\cos^{2}\theta-1\right ) \sin\alpha\\ -12\cos\theta(5\cos^{2}\theta-3)\cos\alpha \end{array } \right\ } .\nonumber\end{aligned}\ ] ] the target function to be extremized in this case , is given by @xmath194 = \left .
\frac{\partial b_{z}}{\partial z}\right| _ { z = h}+\lambda\left .
\dfrac{\partial^{2}b_{z}}{\partial z^{2}% } \right| _ { z = h } , \label{eq7}%\ ] ] and is a functional of @xmath190 , with @xmath195 being the lagrange s multiplier . since the variation of @xmath123 with respect to @xmath190 must vanish , we find , on substitution of eqs.([eq5 ] ) and ( [ eq6.1 ] ) into eq .
( [ eq7 ] ) , that @xmath190 is given by @xmath196 using eq.([eq7.1 ] ) inside eq.([eq6.1 ] ) , and requiring that @xmath197 , gives the equation for @xmath198 @xmath199{c}% 3\sin\theta\left ( 5\cos^{2}\theta-1\right ) \sin\alpha\\ + 4\cos\theta(5\cos^{2}\theta-3)\cos\alpha \end{array } \right\ } = 0\text { , } \label{eq9}%\ ] ] in which @xmath190 depends on @xmath200 according to @xmath201 note that in eq.([eq9 ] ) the variable @xmath71 has been changed to @xmath202 .
this way the double integration is finite and the singularity in the integrand is eliminated .
the numerical solution of eq.([eq9 ] ) gives @xmath203 using this value inside eq.([eq5 ] ) yields @xmath204 which , together with the equilibrium condition , gives @xmath205 note that , though the gradient of the field is finite , the field itself _ diverges _ , and hence @xmath206 . as @xmath207 in the case we studied here , we conclude that @xmath158 also vanishes .
thus , the lowest stable hovering height and the highest one _ coincide _ , leaving no range of stability . a small relaxation in the conditions however , such as limiting the volume of the material to a finite , though very large value , results in a formation of a finite albeit small , range of stability .
we showed that the maximum levitation height of the u - cas is bounded and is given in terms of a characteristic length @xmath5 , which depends on the properties of the substance that the base and the top are made of .
note also , that @xmath208 where @xmath155 is the mass density of the top , and @xmath209 is its magnetization density .
since @xmath210 , where @xmath211 is the residual induction , and since @xmath211 is related to the energy product @xcite via @xmath212 , we may also write @xmath5 as @xmath213 in this form , @xmath5 can be easily estimated from the knowledge of the energy product of the material .
the best candidates for large hovering heights are the nd - fe - b magnets .
an example of which is vacuumschmelze s vacodym 344 hr @xcite , whose remanence ( @xmath211 ) is @xmath214 kgauss and whose density is @xmath215 gr/@xmath216 .
for this magnet , the magnetization is @xmath217 emu/@xmath216 .
thus , with @xmath218 cm / sec@xmath219 , we find that @xmath220 meter .
this gives , according to eq.([eq11 ] ) , a maximum stable hovering height of about @xmath221 meters!. typical values of the energy product , density and corresponding @xmath5 , for modern commercial magnets available today are listed in table i. [ c]|c|c||c|c|c||c| & & @xmath222[m ] + material & ( @xmath223)@xmath224@xcite & material & ( @xmath223)@xmath224@xcite & @xmath155[kg / m@xmath225 & + fe - nd - b & @xmath226[kj / m@xmath225 & fe - nd - b & @xmath226[kj / m@xmath225 & @xmath227 & @xmath228 + ferrite & @xmath229[kj / m@xmath225 & fe - nd - b & @xmath226[kj / m@xmath225 & @xmath227 & @xmath230 + strnat@xcite & @xmath231[kj / m@xmath225 & strnat@xcite & @xmath231[kj / m@xmath225 & @xmath232 & @xmath233 + according to eq.([eq1.61 ] ) the characteristic length is proportional to the _ energy product_. therefore , the larger the energy product is , the larger @xmath5 will be .
a question is then asked : what is the highest conceivable energy product ?
this question has been already discussed by strnat @xcite , according to which `` ...
it seems reasonable to assume that the best room - temperature energy products will never exceed @xmath234 mgoe '' , which is about @xmath235 joules / m@xmath236 , and is also included in table i. these predictions , however , assume that a way might be found to give a fairly high @xmath237 to any magnetic material .
table ii summarizes the upper and lower bounds for the maximum hovering height under a selected number of constraints .
[ c]|lll| & & @xmath238{c}% upper\\ bound \end{tabular } $ ] + & & + & & + & & + & & + & & + it is important to note , that in the above derivation we assumed that the magnetization vector @xmath239 is _ independent _ of the field .
though this is a very good approximation for modern permanent magnets like rare - earth magnets , its only an approximation even for these . furthermore , the maximum field at the top is limited by its mechanical and magnetic strength : stability considerations show @xcite that the spinning speed of the top while hovering must be greater than @xmath240 where @xmath241 and @xmath242 are the moments of inertia along the principal and secondary axes , respectively . also , the maximum field at the top is limited by its coercivity as the field is _ opposite _ to the magnetization .
thus , in practice , eq.([eq11 ] ) should be considered as an _ upper bound _ for the maximum hovering height of the u - cas . yet another way to increase the hovering height of the top is to use current coil for the base , instead of a permanent magnet @xcite .
the advantage of the coil over the magnetic plate is in the fact that one can raise the top to the levitation point by electrical means instead of raising the top mechanically , as in the permanent magnet case . here
, the hovering height is not limited directly by the strength of the magnetic field , but rather by the amount of power that one can deliver into the coil to overcome electrical resistance .
one might argue that the use of superconducting wires for the coil should lift this constraint , but the hovering height is bounded in this case as well , for if the field increases beyond a critical value , the superconductor goes into its normal phase .
we have shown @xcite , that for a given height of levitation @xmath13 , the _ minimum _ power required for levitation is given by @xmath243 here , @xmath244 is the power in watts , @xmath71 is the resistivity in @xmath245 cm , @xmath11 is the free - fall acceleration in @xmath246sec@xmath247 , @xmath248 is the magnetization per unit mass of the top in emu / gr , and @xmath249 is a number of dimensions amp@xmath219emu@xmath219/erg@xmath219cm@xmath219 , and is determined by the _ shape _ and _ current distribution _ of the coil .
we found that for a rectangular cross - section coil , the minimal value of @xmath251 . for the _ optimal _ coil however , we find that @xmath252 .
the authors thank profs . m. milgrom and m. kugler for helpful discussions , and s. tozik and n. fernik for help in the calculations .
one of the authors ( s. s. ) thanks riken , saitama in japan and in particular dr .
y. kawamura for kindly hosting him during a short stay in japan that triggered this study , as it is then that he became acquainted with the u - cas .
s. gov , s. shtrikman and h. thomas , `` on the dynamical stability of the hovering magnetic top '' , accepted for publication in _
physica d_. a copy may be found in _ http://xxx.lanl.gov/abs/physics/9803020 .
_ p. flanders , s. gov , s. shtrikman and h. thomas , `` on the spinning motion of the hovering magnetic top '' , accepted for publication in _
physica d_. a copy may be found in _
http://xxx.lanl.gov / abs / physics/9803044_. | the u - cas is a spinning magnetized top that is levitated in a static magnetic field .
the field is produced by a permanent magnet base , positioned below the hovering top . in this paper
we derive upper and lower bounds for @xmath0the maximum hovering height of this top .
we show that the bounds are of the form @xmath1 where @xmath2 is a dimensionless number ranging from about @xmath3 to @xmath4 , depending on the constraints on the shape of the base and on stability considerations , and @xmath5 is a characteristic length , given by @xmath6 here , @xmath7 is the permeability of the vacuum , @xmath8 is the magnetic moment of the top , @xmath9 is the maximum magnetization of the base , @xmath10 is the mass of the top , and @xmath11 is the free - fall acceleration . for modern permanent magnets
we find that @xmath12[meter ] , thus limiting @xmath13 to about few meters . 0.25 in _ index terms _ * * u - cas , levitron , magnetic trap , magnetic levitation , hovering magnetic top . |
modeling heterogeneous catalytic processes is an essential tool for catalyst design and for the improvement of operating conditions @xcite .
hierarchical approaches have to be adopted due to the need of information on a wide range of length and time scales , from the electronic structure to the reactor design @xcite .
an important step of this approach is the microkinetic modeling , where microscopic processes such as reaction , diffusion , aggregation , and desorption are described by stochastic rules , providing information on the efficiency of the catalytic process in length scales ranging from a few nanometers to several micrometers .
an important problem to be addressed with these methods is the effect of diffusion of reactants through the interface between the catalyst particle and the support .
several recent experimental papers illustrate these phenomena in catalyzed reactions @xcite as well as in related problems , such as gas adsorption @xcite , where the same materials may be used .
morphological features of the catalyst and support and physicochemical conditions of operation determine the role of diffusion on the performance of the process @xcite . sometimes these phenomena are called capture - zone effects , since a certain region of the support surrounding the particle may increase the effective area for capturing reactants from the gas phase , but there are also cases where a net flux to the support is observed .
when only one phase can adsorb the reactant from the gas , the terms spillover ( reactant flux from the catalyst to the support ) or back spillover ( the opposite flow ) are used @xcite , although recently many authors have extended these terms to reactants adsorbing in both phases @xcite ( and here we will use them to facilitate the discussion of the results ) .
due to the large interest in industry , some models which incorporated the effects of reactant diffusion through the catalyst - support interface were designed for certain applications .
the simplest models are based on rate equations ( mean - field models ) that do not account for the spatial heterogeneity of the media where reactions take place @xcite , or that use some type of approximation to represent that heterogeneity @xcite .
other models represent it through distributions of catalytic sites in lattices .
most of them are designed to describe @xmath4 oxidation in different catalysts and conditions @xcite , and simplify diffusion and adsorption processes of some species ( although other applications have also been proposed @xcite ) . in order to get a deeper insight on the effects of reactant diffusion , cwiklik et al @xcite recently simulated simple reaction - diffusion models in surfaces with catalytic stripes and squares , as well as random distributions of catalytic sites . in certain ranges of parameters
, they showed monotonic dependences of the turnover frequency ( tof ) on diffusion coefficients and reaction rates @xcite .
a number of other papers aim at a full investigation of simple models of reaction and diffusion . in the present context ,
relevant examples are models in lattices with distributions of catalytic sites , i. e. with some type of non - homogeneity of catalytic activity @xcite .
even adsorption - desorption models without surface diffusion show that the correlations in catalyst particle position have nontrivial effects on the tof , independently of adsorbate interactions @xcite .
moreover , models including diffusion in heterogeneous surfaces show that the structures that maximize the efficiency of a catalytic process are highly dependent on the rates of the main microscopic processes @xcite . in the present paper , we will propose a one - dimensional lattice model for unimolecular reactions in a supported catalyst , with reactant diffusion and desorption both in the support and in the catalyst particles .
our aim is to understand the interplay between these physicochemical mechanisms , particle size and catalyst coverage .
the model geometry is equivalent to that of ref .
@xcite , but here we will obtain an analytic solution that facilitates the illustration of different possible outcomes .
we will use scaling concepts to explain the model results , so that this framework can be extended to more complex models and applications to real systems , where numerical solutions are usually necessary . indeed , scaling approaches were already shown to be very useful to understand qualitative trends in reaction - diffusion models @xcite . among our results , we will distinguish conditions to enhance the net reactant flux from the catalyst to the support or vice - versa by varying one of the microscopic rates , and we will discuss the effect of increasing the catalyst particle size . the increase or decrease of the tof will be shown to depend on the relation between diffusion lengths of reactants in the catalyst and in the support .
some of these results reinforce findings of previous works @xcite .
moreover , under reasonable assumptions for temperature - dependent rates , we will show that a remarkable increase in the tof can be obtained if reverse spillover regularly fills the catalytic particles with the reactants adsorbed in the support .
this feature may be observed in a large temperature range , a possibility which is interesting for applications .
the identification of these scenarios is possible because the model accounts for the inhomogeneity of the catalytic system and , consequently , predicts inhomogeneous distributions of reactants , which advances over the mean - field models .
the physicochemical processes involved in the model are illustrated in fig . 1 . )
is represented by white circles , while adsorbed reactant a is represented by black circles .
the @xmath5 axis used in the analytic solution are shown ( for the particle and for the support).,scaledwidth=50.0% ] the catalyst particles are represented by segments of @xmath6 sites in a line , separated by @xmath7 support sites . assuming that @xmath8 is the size of the lattice site , this corresponds to particles of diameter @xmath9 separated by a distance @xmath10 .
the fraction of the support covered by the catalytic material is @xmath11 this lattice structure may be a reasonable description of some model catalysts @xcite , as well as a good approximation to the morphology of catalytic clusters supported in materials with long and narrow pores @xcite .
the particle size is determined by the physico - chemical conditions in which the catalytic material is deposited on the support @xcite .
the fraction of the support covered , @xmath12 , is related to the amount of material used to produce a sample as well as the particle shape , which determines the surface to volume ratio . in oxide supported @xmath13 or @xmath14 catalysts , the particle diameter usually vary from @xmath15 to @xmath16 - depending on the growth conditions and if sintering occurs , sizes of @xmath17 or more can be found .
since @xmath8 is of order of a few angstroms , this typically corresponds to @xmath6 ranging from @xmath18 to @xmath19 lattice sites .
the spacing between the particles has a broad distribution in catalysts supported in porous materials , but in model catalysts they are nearly uniform , usually in the range @xmath20 ( @xmath7 typically between @xmath21 and @xmath22 lattice spacings for most oxide supports ) .
the flux of a single chemical species ( reactant ) towards the surface occurs with rate @xmath23 , which is defined as the number of incoming molecules per site per unit time . in most of this work , @xmath23 will be used to define the time scale of the model , so that other frequencies are calculated relatively to this quantity .
the reactant adsorbs in the ( randomly chosen ) site of incidence if it is empty , either in the catalyst or in the support , otherwise the adsorption attempt is rejected .
sticking coefficients are set equal to @xmath15 , since the effects of different values in the catalyst and in the support can be incorporated in the corresponding desorption rates : @xmath24 and @xmath25 , in the support and in the catalyst particles , respectively ( each rate corresponds to number of events per site per unit time ) .
interaction between the adsorbates is restricted to the excluded volume condition .
adsorbed reactants diffuse with coefficient @xmath26 , which for simplicity is assumed to be the same in the support and in the catalyst .
it means that each reactant attempts to execute @xmath27 random steps to nearest neighbor sites per unit time .
we are assuming that the activation energy for diffusion is the same in those regions , which is certainly not true for a real catalyst .
however , it is not a very restrictive assumption for our study because the results are interpreted in terms of diffusion lengths and scaling ideas are emphasized , allowing a direct extension to cases of different @xmath26 in different regions .
since reactants are always in contact with the support or the catalyst , this is a surface diffusion model , which is a reasonable assumption on metal particles . on the other hand ,
knudsen diffusion may be more realistic for the movement of reactants inside a catalyst pore .
however , this does not represent a limitation because a suitable value of @xmath26 can be chosen , and interpretations based on diffusion lengths are still valid .
the adopted reaction mechanism for reactant @xmath28 and product @xmath29 is @xmath30 with reaction rate @xmath31 assumed to be uniform in the catalytic region ( this rate corresponds to number of events per site per unit time ) . the unimolecular reaction ( [ reaction ] ) may represent an eley - rideal ( er ) mechanism , in which the adsorbed species reacts with another species @xmath32 arriving from the gas phase and forms a volatile product : @xmath33 for instance , an application to @xmath4 oxidation is discussed in refs .
@xcite . in this case ,
@xmath31 not only accounts for activation of the adsorbed reactant but also for the flux of the other reactant from the gas phase .
certainly the assumption that @xmath31 is constant in the whole catalytic region is not realistic because it is well known that different crystalline faces of a metal have distinct catalytic activity .
however , the present assumption is useful for a study which aims at investigating the interplay of many other different physico - chemical processes .
in order to solve the model analytically , we assume that the catalyst particles and the support segments between them are sufficiently large ( @xmath34 , @xmath35 ) , so that a continuous approximation is possible . although @xmath6 and @xmath7 may not be very large in real systems , we will show that the continuous approximation works well even with @xmath6 and @xmath7 of order @xmath36 .
the dimensionless reactant coverages in the catalyst particle and in the support are respectively defined as @xmath37 and @xmath38 . for simplicity
, we use the same variable @xmath5 for position in both regions , with the range @xmath39 in the catalyst and the range @xmath40 in the support .
diffusion , reaction and adsorption - desorption processes of fig .
1 lead to equations for surface coverages ; in the catalyst , we have @xmath41 - ( k_{\rm r } + k_{\rm d}^{\rm c } ) \theta_{\rm c}(x , t ) , \label{eqcat}\ ] ] and in the support we have @xmath42 - k_{\rm d}^{\rm s } \theta_{\rm s}\left ( x , t\right ) .
\label{eqsup}\ ] ] each of these equations is valid in the above defined ranges of @xmath5 .
here we are interested in steady state solutions , where @xmath43 and , consequently , @xmath44 and @xmath45 depend only on @xmath5 . in the catalyst , this gives @xmath46 in the support , an analogous equation ( without the reaction term ) is obtained .
( [ eqcatss ] ) can be easily solved and gives @xmath47 where @xmath48 is a constant to be determined from boundary conditions and @xmath49 analogously , in the support we obtain @xmath50 where @xmath51 is a constant and @xmath52 note that , as expected , diffusion , reaction and desorption rates appear in our results in the form of ratios to the external particle flux @xmath23 .
the calculation of unknown constants in equations such as ( [ solthetacat ] ) and ( [ solthetasup ] ) usually follows from the use of suitable boundary conditions .
however , here diffusion leads to a net flux of reactants from the catalyst to the support or vice - versa , which depends on the competition of all other processes along both regions .
thus , those unknown constants will be determined by matching the gain and loss terms in each region due to all those processes , and the net flux at the interfaces will be obtained from them . in the catalyst ,
the contribution to the loss rate due to diffusion involves the probability of finding a reactant in the edge site of that region and of finding the neighboring support site empty ( due to the excluded volume condition ) .
other contributions come from reaction and desorption along the particle .
thus , the loss in the coverage of the catalytic region per unit time is @xmath53 where @xmath54 is the coverage of the edge site of the support region ( neighbor of the catalyst ) @xmath55 @xmath56 is the coverage of the edge site of the catalytic region ( neighbor of the support ) @xmath57 and @xmath58 is the average coverage of the catalytic region @xmath59 the gain rate , which accounts for flux from the support to the catalyst at the edge sites and for the external flux , is @xmath60 dx ' } \nonumber \\ & = & \frac{2d}{a^2 } \theta^ { * } ( 1 - \theta^{\dagger } ) + l f ( 1 - \bar{\theta_{\rm c } } ) .\end{aligned}\ ] ] analogously , loss and gain terms can be determined for the support region , so that solutions for @xmath48 and @xmath51 are @xmath61 \sinh{\left ( \frac{la}{2\lambda_{\rm c } } \right ) } } , \label{alphac}\end{aligned}\ ] ] and @xmath62 \sinh{\left ( \frac{da}{2\lambda_{\rm s } } \right ) } } .
\label{alphas}\ ] ] the turnover frequency , which is the number of reactions per unit site and unit time , is given by @xmath63 we simulated the discrete model , as defined in sec . 2 ( see also fig .
1 ) , in order to check the accuracy of the analytic solution when @xmath6 and @xmath7 are not very large . typical simulations consisted in 100 realizations of the process in a lattice with @xmath64 sites and catalyst coverage near @xmath65 . at each step , an attempt to deposit a new reactant at a randomly chosen site is done .
subsequently , the numbers of attempts to move reactants , desorb them and perform reactions are chosen proportional to the respective rates ( @xmath26 , @xmath66 , @xmath67 , @xmath31 ) .
simulations begin with an empty lattice and proceed up to a long time after a steady state has been reached , where the coverages of both regions are constant .
we observe that even for @xmath68 the continuous approximation is good .
this is illustrated in fig .
2a and 2b , where we compare the analytic and numerical results for the coverage distribution in the catalyst and in the support , respectively , using @xmath69 and @xmath70 .
slight deviations are only found near the boundaries of those regions , but are always smaller than @xmath71 for small @xmath6 .
the accuracy in the average coverages and in the tof is usually higher . along ( a ) the catalyst and ( b ) the support , for @xmath72 , @xmath73 , @xmath74 , @xmath70 and @xmath69 .
circles are simulation results and the curves are analytical results.,scaledwidth=50.0% ] here the model was solved in the steady state , but it certainly can be extended to other situations , for instance when a finite amount of the reactant flows to the catalyst surface and the tof changes in time .
other possibilities are the explicit incorporation of sticking probabilities and the assumption of different diffusion coefficients in the particles and in the support .
the one - dimensional structure facilitates the solution , while preserving essential ingredients such as the spatial heterogeneity .
in order to understand the role of the different rates from a scaling approach , we consider the diffusion lengths in the catalyst particle and in the support . if the corresponding region is large enough , the diffusion length measures the typical distance a reactant moves on it before reacting or desorbing . on the other hand
, if the length of that region is smaller than the diffusion length , then it is expected that the reactant reaches its border and can flow to a neighboring domain . in the catalyst ,
the average lifetime of an adsorbed species before reacting or desorbing is @xmath75 . during this time
, it executes random walks with diffusion coefficient @xmath26 , thus the average distance it spans is of order @xmath76 .
this is the so - called diffusion length , @xmath77 analogously , a reactant in the support has a typical lifetime @xmath78 , thus the corresponding diffusion length is @xmath79 these expressions can be easily generalized to the case of different diffusion coefficients in the particles and in the support , which means that interpretations based on these quantities have a broader applicability .
we will consider cases where desorption in the catalytic region is very low , i. e. @xmath80 .
the values of all rates presented below are given relatively to the incident flux rate @xmath23 , thus they are all dimensionless ( setting @xmath81 and the value of the parameter @xmath8 , we would get the other ones in si units ) .
the tof is also expressed relatively to @xmath23 , thus it is limited to a maximum tof/@xmath23 @xmath82 . in order to facilitate the presentation of the results ,
hereafter we will use the terms spillover and back spillover to denote a net flux of reactants by diffusion from the catalyst to the support and vice - versa , respectively . as noted above
, the broader use of these terms follows a trend of some recent works @xcite .
first we distinguish the conditions where either spillover or back spillover is dominant as the reactant mobility in the surface increases . in figs .
3a and 3b we show the normalized tof as a function of @xmath27 for several reaction rates , respectively with fractions of support covered @xmath83 and @xmath84 , and the same particle size @xmath85 . ,
@xmath86 and @xmath85 .
fractions of support covered are ( a ) @xmath83 and ( b ) @xmath84.,scaledwidth=50.0% ] even without reactant diffusion ( @xmath87 ) , the tof significantly varies with @xmath31 because a large area of the catalyst may be poisoned for low reaction rates ; for instance , in fig .
3a we have normalized tof @xmath88 and @xmath89 for @xmath90 and @xmath91 , respectively .
indeed , excluded volume effects limit the adsorption process , which is known to be a relevant effect even in mean - field models @xcite .
the increase of diffusion coefficient improves the catalytic process for high reaction rates , since back spillover effects are dominant as @xmath26 increases .
however , with low reaction rates , the opposite effect is observed : reactants more easily leave the particles as @xmath26 increases , going to the support where they rapidly desorb . in fig .
3a , there is no change in the normalized tof as @xmath26 increases for @xmath92 , which corresponds to @xmath93 .
if other values of the rates and other values of @xmath6 are chosen , the same feature is observed for a different value of @xmath31 , but with the same ratio @xmath94 . on the other hand , in fig .
3b ( @xmath95 ) , that feature is observed when @xmath96 .
thus , the critical ratio @xmath94 which separates regimes of rapid and slow reactions depends only on the fraction of support covered @xmath12 ; above ( below ) the critical ratio , back spillover ( direct spillover ) is dominant .
this result can be interpreted as follows : if the relative increase of diffusion length in the support is larger ( smaller ) than that in the catalyst , then more reactants flow towards the particles ( support ) and the tof increases ( decreases ) .
now we consider the effects of particle size .
we consider changes in @xmath6 with fixed @xmath12 , in order to simulate cases where a fixed amount of catalytic material is deposited on the support , but islands of different sizes are formed . under these conditions , it is important to stress that the size of the gaps between particles ( @xmath7 ) also increases when @xmath6 increases ( eq . [ defrho ] ) . in figs . 4a and
4b we show the normalized tof as a function of the particle size @xmath6 for two different fractions of the support covered and various reaction rates .
again we observe regimes of high and low reaction rates , corresponding to high and low ratios @xmath94 . for large reaction rates , decreasing @xmath6 is favorable , since @xmath7 also decreases ( @xmath12 is fixed ) and facilitates the back spillover .
on the other hand , for low reaction rates , the conversion is improved by increasing the particle size , since the reactants spend longer times in larger catalytic regions , which compensates the increased desorption in the support . in other words , the loss due to decreasing back spillover is compensated by a gain in decreasing spillover .
the same ratios @xmath94 , which depend on the fraction of support covered @xmath12 , separate the two regimes . ,
@xmath97 and @xmath98 .
fractions of support covered are ( a ) @xmath84 and ( b ) @xmath99.,scaledwidth=50.0% ] these nontrivial effects of the catalyst geometry may be helpful for catalyst design if one is interested in taking advantage of back spillover , preferrably in the case of rapid reactions , or in reducing the effect of spillover in the case of slow reactions . at this point , it is important to notice that figs .
3 and 4 show that the change in one model parameter may increase the tof by a factor near @xmath18 , which is a remarkable change in the efficiency of the catalytic process . as far as we know , this is the first work that analyzes the conditions for changing the direction of the net reactant flux in the particle - support interface .
the interpretation of results based on diffusion lengths was formerly proposed in refs .
@xcite for models of @xmath4 oxidation in oxide supports .
however , both studies focused on the regime where back spillover is dominant , for instance by assuming infinite diffusion lengths of some species adsorbed in the particles .
that regime was also considered in ref .
@xcite with a mean - field approach that accounts for the different neighborhood of the catalytic sites ( in an approximate form ) . in this case , the approximation is successful because the diffusion lengths are small .
the diffusion coefficient and the reaction and desorption rates are expected to have arrhenius forms as follows : @xmath100 @xmath101 and @xmath102 here , @xmath103 is a frequency and @xmath104 is an activation energy ( @xmath105 ) . assuming that the activation energy for desorption in the catalyst particles is much larger than the other activation energies , we use @xmath106 , which is negligible compared to the other rates in the relevant temperature ranges . for the other activation energies
, we always assume that @xmath107 @xcite .
other reasonable assumptions on the amplitudes of the arrhenius rates facilitate the analysis of the effects of different ranges of energy barriers .
we will consider @xmath108 , @xmath109 , and work with a range of ratios @xmath110 from @xmath36 to @xmath111 .
these relations are reasonable for langmuir - hinshelwood ( lh ) reactions and for @xmath4 adsorption in oxides @xcite , respectively , but we emphasize that these systems are only rough guides to choose working parameters , and may not be viewed as prospective applications - indeed , the unimolecular reaction of our model is representative of er mechanism . in the following ,
we also consider fraction of support covered @xmath84 and @xmath112 ( thus the calculated rates are again ratios to @xmath23 ) .
qualitatively , we expect that each microscopic process will be significantly activated when its rate exceeds the external flux @xmath23 .
however , excluded volume effects lead to surface poisoning when reactions are not frequent ( low temperatures ) , so that other processes can affect the turnover frequency only after reactions are activated .
first we consider the simplest case where reactions are very difficult compared to the other activated processes , i. e. @xmath3 .
an example of slow reaction is the hydrogenation of @xmath4 on @xmath13 @xcite , when compared to the spillover to the @xmath114 support and the formation of @xmath115 there .
5a shows the typical evolution of the tof with temperature for different particle sizes .
when the reactions become more frequent ( @xmath116 , i. e. @xmath117 ) , desorption is already activated and diffusion is very fast .
thus spillover and subsequent desorption of reactants does not allow the increase of the turnover frequency .
instead , the tof begins to increase only when @xmath118 ( @xmath119 ) . at higher temperatures ,
the negative contribution of spillover is more important when the particles are small , so that a slow increase of the tof is observed . on the other hand , this negative contribution is reduced for large particles , and the maximum tof is rapidly attained by increasing the temperature .
the value tof@xmath120 corresponds to the fraction of the surface covered by the catalyst , which is expected at high temperatures because only species adsorbed on the catalyst react . , @xmath121 and @xmath122 ; ( b ) @xmath123 , @xmath124 and @xmath122 ;
( c ) @xmath125 , @xmath124 and @xmath126 . in all cases ,
@xmath127.,scaledwidth=50.0% ] next we consider the case where reactions are easily excited , i. e. @xmath113 .
an example of rapid reaction with er mechanism is @xmath4 oxidation on @xmath128 @xcite , where reaction rates are nearly @xmath21 times larger than back spillover rates of oxygen .
5b and 5c illustrate the case @xmath113 with small and large desorption energies , respectively . in fig .
5b , the tof increases towards a plateau in tof@xmath120 at @xmath129 ) , where @xmath116 . at @xmath130 ,
when diffusion is activated , back spillover leads to a second jump in the tof , which is highly dependent on the particle size . for smaller sizes ( small gaps between particles ) ,
the migration of reactants from the support to the catalyst is easy even for low @xmath26 , thus large conversion rates are rapidly obtained .
the temperature of maximal tof is attained when desorption begins to play a significant role ( @xmath131 ) , independently of particle size . in the case of large gaps between the particles ,
this temperature is still low for back spillover to be efficient , thus only a small peak appears in the tof plot . for further temperature increase ,
@xmath6-dependent results are again obtained .
the diffusion length in the support is @xmath132}$ ] , which decreases with increasing temperature , and the beneficial effect of back spillover ceases when @xmath133 ; this condition is satistified at higher temperatures for smaller @xmath6 , which explains the slower decay of the tof in this case . for these reasons ,
the peak in the tof is high and broad for small @xmath6 , low and narrow for large @xmath6 . in fig .
5c , the main features of fig .
5b are present .
however , since @xmath0 is much larger than the other activation energies , the maximal effect of back spillover ( tof / f @xmath134 ) is observed in a wider temperature range and for all particle sizes shown there .
thus , fig .
5b illustrates typical conditions in which particle size effects are clearer : desorption in the support has higher activation energy than diffusion and reactions , but activation of a process begins while the other processes are not fully activated , so that the diffusion length @xmath135 can not attain large values before desorption is activated .
now we consider cases of intermediate activation energies for reaction , i. e. @xmath136 . @xmath4
oxidation provides several examples with such relation between activation energies ; however , it is important to stress the difference in the usual reaction mechanism ( lh instead of er ) .
an example is @xmath4 oxidation on @xmath137 @xcite : the energy of diffusion of @xmath138 on the support is @xmath139 , while reaction energy is @xmath140 and desorption in the support is @xmath141 ( diffusion on the catalyst is assumed to be fast for the application of a mean - field model in ref .
@xcite ) . in fig .
6a we show the normalized tof as a function of temperature for our model with three different particle sizes , using a set of activation energies previously suggested for @xmath4 oxidation in @xmath142 and @xmath143 @xcite .
6b shows the evolution of the average coverage ( particle plus support ) . , @xmath144 @xmath126 , and @xmath127 .
( b ) temperature dependence of the corresponding average reactant coverage.,scaledwidth=50.0% ] at low temperatures , all the surface is poisoned , and reactions are slow .
increasing the temperature to @xmath145 ( @xmath116 ) , the turnover frequency increases , thus vacancies are created in the particles and in the support , which facilitates the back spillover effects .
note that there is no significant effect of particle size when a large temperature range is scanned , since the diffusion lengths @xmath135 and @xmath146 are very large ( diffusion is highly activated in much lower temperatures ) .
while the tof increases with temperature , the coverage rapidly decreases towards zero because diffusion and reaction are rapid compared to the external flux .
further temperature increase leads to a maximum in the turnover frequency , again when desorption in the support is activated ( @xmath131 ) .
however , in the right side of the peaks of fig .
6a , we observe a size dependence because back spillover ceases only when @xmath133 . at high temperatures ,
a size dependence is also noticeable : small particles lose more reactants by spillover than the large ones , and those reactants easily desorb in the support , which leads to smaller tof . as explained above , the most remarkable effects of particle size
are observed in cases where the activation energies are close to each other .
this is also illustrated in figs . 7a and 7b for @xmath136 , but @xmath1 close to @xmath2 , with different ratios between @xmath147 and @xmath148 . in this case ,
when reactions become more frequent and leave room for new incoming reactants ( @xmath116 ) , the surface mobility is still low ( i. e. @xmath27 is not much larger than @xmath23 ) .
thus , back spillover is significant only for small particles ( and small vacancies between them ) . however ,
if the spacing between particles is large , then slow diffusion is not able to bring all reactants adsorbed in the support to the catalytic region .
7a and 7b also show that these effects appear in a wider temperature range when the amplitudes of the boltzmann factors are closer to each other , since activation of desorption occurs in higher temperatures . , @xmath149 @xmath122 : ( a)@xmath150 ; ( b ) @xmath151.,scaledwidth=50.0% ]
we proposed a model for a unimolecular reaction in a supported catalyst with tunable particle size and fraction of support covered .
it represents important physico - chemical processes , such as diffusion , desorption and the external flux of reactants .
we analyzed the effects of spillover and back spillover on the turnover frequency under different conditions .
first we considered the isolated effects of enhancing surface diffusion ( coefficient @xmath26 ) and increasing catalyst particle size @xmath6 . for rapid reactions ,
the increase of any of those quantities favors back spillover and , consequently , increases the turnover frequency .
this regime is separated from that dominated by spillover by a critical value of diffusion length in the catalyst and in the support , which depends only on the fraction of the support surface covered by the catalyst . in the spillover - dominated regime ( slow reactions ) , increasing @xmath26 or @xmath6 slows down the conversion of the reactants .
subsequently , we considered temperature effects by assuming arrhenius dependence of all physico - chemical rates and reasonable values for the amplitudes in those relations . with activation energy for desorption ( @xmath0 ) in the support larger than that for diffusion ( @xmath1 ) , a peak in the turnover frequency as a function of temperature
is observed for small and intermediate values of reaction activation energy @xmath2 , i. e. for cases where @xmath152 .
significant particle size dependence in the peaks is usually observed when the gaps between those energies are small , so that activation of one process occurs while the other ones are not fully activated yet , and the corresponding diffusion lengths rapidly vary with temperature .
the right side of those peaks show size - dependence under more general conditions . for fixed amount of catalytic material deposited on the support , small particle sizes ( with small distance between them )
allow the turnover frequency to attain high peak values due to the beneficial effect of back spillover , while large particles provide low enhancements of catalytic activity .
finally , in the case of slow reactions ( @xmath3 ) , the tof monotonically increases with temperature , and large particle sizes are more efficient to avoid the negative effects of direct spillover .
the aim of the present work is not the quantitative description of a particular catalytic process , but to discuss the interplay of various physico - chemical mechanisms and conditions in systems where spillover is present .
however , it may be useful in the interpretation of some experimental results and motivate the proposal of extended models for their quantitative description . here
we mention two recent works as examples where the scaling ideas developed above may be relevant .
however , we emphasize the fact that our model can not be directly applied to these systems , thus our results only suggest general guidelines to understand their qualitative behavior . in the first example , piccolo and henry @xcite studied the oxidation of @xmath4 by @xmath153 on @xmath154 and observed a peak in the tof as a function of the temperature .
they found a remarkable increase in the peak height as the particle size was decreased ( even being accompanied by a decrease in the @xmath14 coverage ) , but tiny shifts in the temperature of the maximum .
these results resemble those in figs .
6a , 7a and 7b .
while the size - independent curve shape suggests weak size - dependence of activation energies , the peak height increase suggests that diffusion lengths of reactants were not very large when reactions were activated , so that back spillover ( which increases tof for any type of reaction ) was facilitated with small particles .
thus , differences in activation energies of diffusion , reaction and desorption are probably small , with @xmath0 being larger than the other ones .
the second example is a study of @xmath4 oxidation on @xmath155 on active supports ( i. e. those which adsorb and supply oxygen to the reaction ) , which shows a different trend @xcite : under certain conditions , the tof does not depend on catalyst particle size .
although that reaction involves two species and is probably of lh type , the experimental result suggests that diffusion was highly activated at the working temperature .
in this situation , similarly to our model , diffusion lengths are always very large and the contribution of back spillover does not depend on particle size . recalling the results of models for ethene and acetylene hydrogenation which incorporate spillover effects is also interesting at this point @xcite . in both systems , the hydrocarbon blocks catalytic sites at high pressures , which leads to a discontinuity of the tof when the external reactant flux towards a catalytic site is of the same order of the microscopic reaction rate . in ethene hydrogenation
, this effect is shown to be more intense with small particles @xcite , leading to finite - size effects
. however , diffusion is weakly activated at the working temperature when compared to the other processes ( the amplitudes of boltzmann factors associated to different processes are very different in that case ) , thus there is no significant effect of diffusion in the tof .
this reinforces the conclusion that the scaling approach proposed here can be extended to interpret more complex reaction models .
this is very important in cases where analytic solutions are not feasible because it may help choosing the conditions to perform simulation work .
the authors thank prof .
robert hamers for a critical reading of the manuscript and acknowledge support from cnpq and faperj ( brazilian agencies ) to their simulation laboratory at uff , brazil .
tgm acknowledges a grant from capes ( brazil ) and fdaar acknowledges support from cnpq for his visit to uw - madison . | we study a model for unimolecular reaction in a supported catalyst including reactant diffusion and desorption , using analytical methods and scaling concepts . for rapid reactions , enhancing surface diffusion or increasing particle size favors the flux of reactants to the catalyst particles , which increases the turnover frequency ( tof ) .
the reactant flux towards the support becomes dominant when the ratio of diffusion lengths in the catalyst and in the support exceeds a critical value .
a peak in the tof is obtained for temperature - dependent rates if desorption energy in the support ( @xmath0 ) exceeds those of diffusion ( @xmath1 ) and reaction ( @xmath2 ) .
significant dependence on particle size is observed when the gaps between those energies are small , with small particles giving higher tof .
slow reactions ( @xmath3 ) give tof monotonically increasing with temperature , with higher reactant losses in small particles .
the scaling concepts can be extended to interpret experimental data and results of more complex models . |
the observation of a light quantum is a demanding task since the detector should be able to resolve this tiny amount of energy via the heat absorption or the photo - electric effect and in a noiseless fashion convert it to a measurable electric signal .
in addition to this , quantum features are extremely fragile and easily lost by any experimental imperfections prior to and at the detection .
nonetheless , in the past years , several photo - detector technologies have demonstrated the sensitivity to measure at the single - photon level @xcite . moreover , many quantum optics applications require photon counters with at least partial photon - number resolution @xcite . in such experiments ,
superconducting detectors @xcite as well as visible light photon counters @xcite can be exploited with high efficiencies however , at the cost of having to cool to temperatures of a few kelvin . at room temperature
only a few options are available , for example photo - multiplier tubes @xcite that suffer from large losses or hybrid detectors @xcite that require specific reconstruction techniques . for photon counting tasks the most widely used detectors are avalanche photo diodes or shortly _ click detectors _ that are in geiger mode sensitive to single photons but unable to resolve various photon numbers @xcite .
this is due to the fact that a whole bunch of carriers is released after a photon impinges on the detector in order to gain a detectable electrical signal @xcite .
this technology for the detection of light at the single - photon level is well - established and commercially available . in the region of visible light ,
click detectors can reach high quantum efficiencies and measure with a very low number of spurious counts @xcite .
additionally , in connection to a fiber - optic beam splitter network these detectors have lately been utilized as photon - number resolving detectors via time multiplexing @xcite .
regarding a particular quantum optics application the selection of a suitable detector starts by determining the wavelength range of the investigated radiation . in the range of telecommunication wavelengths , which are ideal for long - distance transmission of light in optical fibers ,
click detectors have suffered from several drawbacks . until recently these detectors have required narrow timegating to suppress spurious counts , in other words ,
these detectors are active only for a short duration of time and the measured signal has to be synchronized with the gate . in a free - running mode first demonstrated for the telecommunication band in refs .
@xcite this synchronization can be omitted @xcite .
the better temporal availability directly offers the possibility to measure at higher rates .
additionally , free - running detectors make time - multiplexing and photon - counting via a fiber network more practicable at telecommunication wavelengths . apart from the wavelength sensitivity and efficiency , the temporal characteristics of a detector such as dead time , timing jitter , as well as the amount of spurious counts in terms of the dark counts and afterpulses are the figures of merit of click detectors @xcite .
while wavelength region and quantum efficiency mostly depend on the semiconductor s properties the above characteristics are also influenced by the detector realization such as the stochastic process describing the carrier avalanche .
however , also the external electronics can have great impact especially on the temporal characteristics of the detector @xcite .
for example the technique used for quenching the avalanche often limits the dead time of the detector and the read - out electronics the timing jitter .
the rate of spurious counts often strongly depends on the bias voltage @xcite , which also sets the internal ( avalanche ) amplification . for a high signal - to - noise ratio we would like a high efficiency and low spurious counts , or in other words the lowest possible noise equivalent power ( nep , or better noise - equivalent photon count rate ) .
the nep has to be controlled , otherwise the detector looses its single - photon sensitivity @xcite . to complicate things ,
these properties are temperature dependent and can thus be manipulated by cooling the detector .
cooling typically increases the detector efficiency at a given dark count rate @xcite . however , care has to be taken since this comes at the cost of an increasing probability of afterpulses @xcite and a shift in the wavelength sensitivity of the detector towards shorter wavelengths . here , we present a highly practical realization of a liquid - nitrogen ( ln@xmath0 ) cooled single - photon sensitive detector at telecommunication wavelengths .
it is based on a commercially available fiber - coupled negative - feedback avalanche diode ( nfad ) ( pna-300 - 1 ) from princeton lightwave .
the integrated feedback resistor makes this diode intrinsically passively self - quenched with a well - defined pulse form and a consistent reactivation time @xcite .
our implementation of deriving an output signal follows yan et al .
our solution of cooling the detector with liquid nitrogen has several advantages .
a wide range of temperatures can be chosen , the system is compact and easily movable . it does not require any bulky cooling apparatus and it efficiently prohibits humidity from condensing on the detector heads with minimal preventions taken .
in contrast to other methods that may necessitate combination of coolers such as thermoelectric cooling and water cooling , ln@xmath0 requires only a single step and can be implemented at low cost in laboratories , where ln@xmath0 is already available .
this paper is organized as follows . in sec .
[ sec:2 ] we present the chosen cooling technique and study its temperature characteristics . in sec .
[ sec:3 ] we investigate the optical properties of our implementation .
we examine both the efficiency of the detector as well as the nep at different temperatures in terms of the dark count rate and show how to suppress the effects of afterpulsing . after finding the working parameters for the most suitable operating point
we apply our detector in a true quantum optics task for coincidence counting . sec .
[ sec:4 ] summarizes our results and highlights the findings of our studies .
our realization of a free - running , single - photon sensitive detector is based on ln@xmath0 cooling . a schematic picture of our implementation having a height of approximately is shown in fig . [ fig:2.1 ] .
we put our fiber - coupled nfads in a capsule placed on top of a long metal rod of stainless steel , which is set into a partially filled dewar of ln@xmath0 such that the detectors sit above the surface of the liquid .
we can pack two nfads with small circuit boards in one capsule including a heating coil placed in between them as depicted by the inlay in fig .
[ fig:2.1 ] .
all wiring is routed via a stainless steel tube to a connection box at the top of the tube .
a vented bushing allows evaporated nitrogen to escape the dewar and holds the construction straight .
we employ simulations with a commercial solver to investigate the heat transport inside the dewar and at its opening .
this is especially important as the nfads have to be protected from condensed water . in order to gain the proper parameters for the simulation
, we make a one day test run , in which the temperature of the detector housing is kept constant at .
we observe an ln@xmath0 evaporation rate of approximately . at this rate
a single dewar fill of yields an operating time of . a small amount of condensed water , which forms during the test run near the vent , can be easily collected with a tissue placed loosely at the neck of the dewar .
the long term test also shows that a heating power of approximately is required to keep the operating temperature stable .
the simulated dewar shape and size correspond to the dimensions of our real dewar with a interior space filled to a specific level with ln@xmath0 .
the boundary conditions are the measured evaporation rate , an environment temperature of and pressure . to simulate the effect of the heating coil , of power
is deposited in the heating area inside the detector capsule .
our simulation in fig .
[ fig:2.2 ] firstly shows as expected that with the chosen parameters a temperature of can be achieved on the horizontal plate in the detector capsule , where the nfad - detectors are attached .
secondly , it can be clearly seen that the heat is transported not only via conduction but also via convection that causes a small eddy inside the dewar .
more interestingly , there is a continuous flow of evaporated nitrogen from the dewar to the outside , which prohibits the condensation of water and the forming of ice on the detector house and the cooling rod .
a similar eddy inside the detector capsule cools and protects the nfads itself .
although the strength of the heat transport is strongly dependent on the amount of ln@xmath0 , a stable operating temperature can be achieved by a commercial temperature controller that regulates the heating power . in a second long test at
we measure the temperature stability of our diodes and find that the operating temperature can be set with a stability of .
in order to determine the most suitable working range of our single - photon sensitive detectors we investigate their main figures of merits . although each individual detector has its own specific characteristics we present here the figures of merit for one single nfad .
we start by investigating the dark count rate with respect to the bias voltage and the detector s temperature .
the electrical response of the nfad is discriminated at a desired threshold level and measured via a multi - channel time - to - digital converter ( tdc ) that serves as a counter of the detected clicks . as the electrical response of the nfads shows a highly complicated pulse shape ,
we choose to measure the dark counts with two different threshold values for the tdc counting the clicks .
our results in fig .
[ fig:3.1 ] show the working range of the detector in terms of the bias voltage and the achieved dark counts at various temperatures .
we further note that the threshold value can be chosen rather freely between two values we used in our measurements as they both deliver the same tendency of the measured counts . for a low and high threshold voltage ( shortly @xmath1 ) at different operating temperatures ( marked with @xmath2 ) from to . ]
apart from controlling the dark counts also afterpulsing that introduces artificial detection events has to be governed . for this purpose , we measure the dark counts in a time - resolved manner with threshold , which is kept at this level in all the coming measurements , and investigate their temporal characteristics .
we label all clicks happening in proximity of another event as afterpulses and record afterpulses until a time given by 10% of the inverse of the raw dark count rate is passed .
we estimate the afterpulsing probability as a ratio of those events to all registered detection events .
our results in fig .
[ fig:3.2](a ) show that the detector is highly influenced by afterpulsing already at and would be practical only at very low dark count rates or bias voltages . in fig .
[ fig:3.2](b ) we demonstrate a typical temporal distribution of the detection events proceeding a registered click , which indicates that there is an increased probability of detecting another event just after some tens of nanoseconds from a previous one .
[ fig:3.2](b ) also indicates that to counter the afterpulsing , a blocking time should be applied , which drastically diminishes the afterpulsing effect , as seen in fig .
[ fig:3.2](a ) . the cumulative probability in fig .
[ fig:3.2](c ) of the temporal distribution [ fig .
[ fig:3.2](b ) ] can be used to quantify the required blocking time .
( 167,167 ) ( 0,0 ) ( 32,152)(a ) ( 200,152)(b ) ( 200,40)(c ) applying blocking means that we let the detector run freely and post - select only those events , before which no event was detected within the selected time frame .
of course , the blocking time limits the useful signal rate to a value given by its inverse .
our investigations of afterpulsing at lower temperatures show a very similar behavior to that recorded at .
however , due to increasing afterpulsing longer blocking times are required to suppress the afterpulsing probability to the same level . this way the detector
can still be used down to temperatures of .
the appropriate blocking time depends on how large an afterpulsing probability can be accepted .
the trade - off between the highest acceptable afterpulsing probability and the shortest blocking time required delivers a clear upper bound for the raw dark count rate at a chosen temperature , below which the detector has to be operated .
next , for extracting the detector s efficiency we use a continuous - wave ( cw ) laser ( from yenista ) with a maximal power of and tunable between as a light source .
the laser beam is attenuated to the few - photon level with a stack of neutral density filters .
the weak light is then coupled via a single - mode fiber coupler to the nfads .
we determine the detector efficiency at different raw dark count rates that are fixed by choosing a proper bias voltage . for cw light
the detector efficiency can be extracted via @xcite @xmath3 in which @xmath4 describes the energy of a photon at , @xmath5 is the mean power of the cw laser and @xmath6 the amount of the attenuation .
in addition to this , @xmath7 represents the detected click rate , @xmath8 is the dark count rate at the selected bias voltage , and @xmath9 the dead time .
we extract the detector s efficiency at different temperatures and at each temperature we adjust the blocking time such that the afterpulsing is suppressed to the same level as with block at .
we determine the rates @xmath7 and @xmath8 from the post - selected data and use the blocking time as @xmath9 . at each chosen dark count rate
we measure the detector s efficiency at different laser powers to ensure that the extracted efficiency remains intact at different photon fluxes and does not vary due to artificial spurious counts .
figs [ fig:3.4](a - d ) illustrate the efficiencies measured at different operating temperatures with respect to the raw dark count rates .
we conclude that moderate efficiencies on the order of a few percent can be achieved for cw light .
nevertheless , the longer blocking times at lower temperatures make the detector useful for applications that require high efficiency but only small count rates , such as triggering on rare or slowly periodic events .
( 165,120 ) ( 0,0 ) ( 22,115)(a ) ( 22,88)(b ) ( 22,63)(c ) ( 22,37)(d ) ( 151,114)(e ) quite as important as the efficiency is the single photon sensitivity of the detector that can be investigated via the noise equivalent power ( nep ) given by @xcite @xmath10 the lower the nep the more sensitive the detector is at the single - photon level .
furthermore , eq . ( [ eq : nep ] ) represents the trade - off between the detector dark - count level and the achieved efficiency . our results in fig .
[ fig:3.4](e ) show a clear improvement in nep when cooling down the detector , nevertheless , remaining rather constant over the range of chosen dark count rates for each temperature .
the values obtained here for the nep follow the tendency of the latest developments reported for avalanche photodetection at telecommunication wavelengths and are gradually approaching those typical in the visible wavelength range at room temperatures although being still one order of magnitude away from what is commercially available for visible light with slight cooling @xcite . in order to be suitable for the investigation of several different kinds of quantum light sources , a detector with a broad spectral band
is highly desired .
since the properties of semiconductor materials are highly dependent on temperature , we investigate the upper cut - off wavelength of our detector that is expected to shift towards shorter wavelengths when the temperature is decreased .
for this purpose we measure the relative spectral sensitivity while trying to keep the raw dark count rates at a constant level between at different temperatures .
as we are only interested in the relative spectral sensitivity , we choose to employ a fixed blocking time of at all temperatures , due to which the afterpulsing probability varied from 9.5(8)% to 38(6)% when cooling the detector from to .
we investigate the relative sensitivity at different wavelengths between and . at each wavelength
we normalize the detector efficiency against that at . from our results in fig .
[ fig:3.5 ] we conclude that in order to gain the best operation cooling down to is desired only if the light source emits radiation well below . near this wavelength cooling below
does not improve detectors performance .
while we have not directly measured the upper cut - off wavelength , the observed behavior is consistent with its shift towards shorter wavelengths .
moreover , we test our detectors with a quantum light source that produces photon pairs , typically called signal and idler , via parametric down - conversion ( pdc ) . in our experiment the type - ii pdc process is pumped with a picosecond laser ( repetition rate , central wavelength , and bandwidth ) .
our photon - pair source , a long periodically - poled lithium niobate waveguide , is kept in an oven at a temperature of to reach the desired wavelength region . for coupling the light aspheric lenses
are employed and after passing through the waveguide the pump beam having cw - equivalent power of approximately is separated from signal and idler with a dichroic mirror and a long pass filter .
another dichroic filter is used to block background radiation below wavelengths of .
thereafter , signal and idler are separated in a polarizing beam splitter , coupled to single mode fibers with fixed fiber collimators , and send to the nfads kept at . the electric signal of nfads is discriminated and registered with the tdc . in order to measure the arrival times of signal and idler in a time - resolved manner , the laser clock divided by a factor of 128 is used for synchronization .
( 170,170 ) ( 0,0 ) ( 35,161)(a ) ( 35,73)(b ) the nfads are driven with rather low bias voltages such that the raw dark count rates are approximately and for the two detectors .
our results for the raw single counts are shown in fig .
[ fig:3.6](a ) in time resolved manner , which indicates that a large background is detected .
the raw single - count rates measured at the two detectors are and . we then post - selected the recorded data by employing a blocking time of approximately . as seen in fig .
[ fig:3.6](b ) this drastically reduces the background of afterpulses and other spurious counts .
further , by applying narrow time gates of , the post - selected time - gated single - count rates of and are extracted for the two detectors .
this way , the rest of the remaining background can be efficiently suppressed .
we wait for afterpulses until a time given by 10% of the inverse of the post - selected time - gated single - count rate has passed that is approximately and estimate at the both detectors the afterpulsing probabilities of @xmath11 and @xmath12 .
considering coincidence counting , we first investigate the raw coincidence rate in time - resolved manner as shown in fig
. [ fig:3.7 ] , from which we infer the rate of . by applying a tight time gate
, the coincidence count rate drops to , and finally by applying blocking it reduces to , which seem to be , on the other hand , rather robust against the afterpulsing effects . finally , we apply the nfads in a true state characterization task .
we extract the klyshko efficiency of our pdc process by comparing the number of coincidence counts to single counts , after the reduction of accidental counts from the observed coincidences .
this delivers for signal and idler the values @xmath13 and @xmath14 , in which we include only statistical errors .
the coincidence to accidentals ratio of our source is @xmath15 , where the amount of accidentals is estimated via the measured single - count rates .
the state preparation with a fast pump in combination with the free - running detection allows us to drive the source in the extremely low gain regime that is necessary for applications relying on true photon - pair sources .
free - running single - photon sensitive detectors for the telecommunication band are of great importance for many quantum optics applications . still today , most of these are plagued with low detection efficiencies or require special conditions to work .
we implemented a highly practical and flexible cooling with liquid nitrogen for two fiber - coupled , commercially available nfads . with our temperature control extremely stable operation of the detector
is possible .
we investigated several figures of merit of our detector . in order to work at its best performance ,
we investigated the dark count level and afterpulsing probability . due to the large afterpulsing we introduced a blocking time that enables us still to measure with a relative high single - photon sensitivity and a moderate detection efficiency .
further , we investigated the spectral boundaries of our detector at the telecommunication wavelengths .
our results deliver a set of working parameters for the most suitable operation of such a detector in a quantum optics application .
finally , we showed that by applying a blocking time our detector is indeed applicable for time - resolved photon - counting tasks of pulsed light . despite the drawbacks , our detector shows a great potential , and it can be useful for example in coincidence counting , or even in more sophisticated photon counting applications via time - multiplexing in a fiber network .
we thank henning weier ( qutools ) the support with discriminator boards required for operating our nfads .
additionally , we thank armin sailer and gerhard hendl for helping with the mechanical and electrical construction of the detector and raimund ricken for the assistance with the waveguide fabrication . this work was supported in part by the european research council ( erc ) through project ensena ( 257531 ) and the austrian science fund ( fwf ) through project no . | the measurement of light characteristics at the single- and few photon level plays a key role in many quantum optics applications .
often photodetection is preceded with the transmission of quantum light over long distances in optical fibers with their low loss window near .
nonetheless , the detection of the photonic states at telecommunication wavelengths via avalanche photodetectors has long been facing severe restrictions .
only recently , demonstrations of the first free - running detector techniques in the telecommunication band have lifted the demand of synchronizing the signal with the detector .
moreover , moderate cooling is required to gain single - photon sensitivity with these detectors . here
we implement a liquid - nitrogen cooled negative - feedback avalanche diode ( nfad ) at telecommunication wavelengths and investigate the properties of this highly flexible , free - running single - photon sensitive detector .
our realization of cooling provides a large range of stable operating temperatures and has advantages over the relatively bulky commercial refrigerators that have been used before .
we determine the region of nfad working parameters most suitable for single - photon sensitive detection enabling a direct plug - in of our detector to a true photon counting task . |
according to increasing astrophysical indicia , the evolution of the universe seems to be largely governed by dark energy with negative pressure together with pressureless cold dark matter ( see @xcite for the latest review ) in a two to one proportion . however , little is know about the origin of either component , which in the standard cosmological model would play very different roles : dark matter would be responsible for matter clustering , whereas dark energy @xcite would account for accelerated expansion . several candidates for dark energy haven proposed and confronted with observations : a purely cosmological constant , quintessence with a single field ( see @xcite for earliest papers ) or two coupled fields @xcite , k - essence scalar fields , and phantom energy @xcite .
interestingly , a bolder alternative presented recently suggests that an effective dark energy - like equation of state could be due to averaged quantum effects @xcite .
the lack of information regarding the provenance of dark matter and dark energy allows for speculation with the economical and aesthetic idea that a single component acted in fact as both dark matter and dark energy .
the unification of those two components has risen a considerable theoretical interest , because on the one hand model building becomes considerable simpler , and on the other hand such unification implies the existence of an era during which the energy densities of dark matter and dark energy are strikingly similar .
one possible way to achieve that unification is through a particular k - essence fluid , the generalized chaplygin gas @xcite , with the exotic equation of state @xmath1 where constants @xmath2 and @xmath3 satisfy respectively @xmath4 and @xmath5 . using the energy conservation equation and the einstein equation @xmath6 one obtains the evolution @xmath7 where as usual @xmath8 is the scale factor , @xmath9 and @xmath10 is an integration constant .
this model interpolates between a @xmath11 evolution law at early times and @xmath12 at late times ( i.e. the model is dominated by dust in its early stages and by vacuum energy in its late history ) . in the intermediate regime
the matter content of the model can be approximated by the sum of a cosmological constant an a fluid with a soft equation of state @xmath13 .
the traditional chaplygin gas @xcite corresponds to @xmath14 ( stiff equation of state ) .
another possibility which has emerged recently is the modified chaplygin gas ( mcg ) @xcite .
it is characterized by @xmath15 with @xmath16 a constant .
alternatively , such evolution can be seen as coming from a modified gravity approach , along the lines of the dvali - gabadadze - porrati @xcite , cardassian @xcite and dvali - turner @xcite models . in those works the present acceleration of the universe
is not attributed to an exotic component in the universe , but to modifications in gravitational physics at subhorizon scales . following the proposal by @xcite , an evolution like that arising from ( [ eqstate ] ) in standard gravity ,
could alternatively be obtained in the modified gravity picture for a pure dust dust configuration under the modification @xmath17 where @xmath18 . modified chaplygin cosmologies with @xmath19 are transient models which interpolate between a @xmath20 evolution law at early times and a de sitter phase at late times , but interestingly the matter content at the intermediate stage is a mixture of dust and a cosmological constant .
the sound speed for the modified chaplygin gas @xcite becomes [ cs ] c_s^2=. the observational tests of traditional and generalized chaplygin models are numerous .
several teams have analyzed the compatibility of those models with the cosmic microwave background radiation ( cmbr ) peak location and amplitude @xcite , supernovae data @xcite and gravitational lensing statistics @xcite .
the main results of those papers can be summarized as follows : models with @xmath21 and some small curvature ( positive or negative ) are favored over the @xmath0cdm model , and chaplygin cosmologies are much likelier as dark energy models than as unified dark matter models . in what regards the modified chaplygin gas
, it has only been tested observationally in @xcite , using the most updated and reliable compilation of supernovae data so far : the gold dataset by riess et al .
@xcite . by means of a statistical test which depends no only on @xmath22 ( as in usual procedures ) bu also on the number of parameters of the parametrization of the hubble factor as a function of redshift
, it was concluded that the modified chaplygin gas cosmologies give better fits than usual and generalized chaplygin cosmologies . in this paper
we shall be concerned with the evolution of large - scale inhomogeneities in modified chaplygin cosmologies .
this is an issue of interest because candidates for the dark matter and dark matter unification will only be valid if they ensure that initial perturbations can evolve into a deeply nonlinear regime to form a gravitational condensate of super - particles that can act like cold dark matter . here
we will follow the covariant and sufficiently general zeldovich - like non - perturbative approach given in @xcite , because it can be adapted to any balometric or parametric equation of state .
our results indicate that our model fits well in the standard structure formation scenarios , and we find , in general , a fairly similar behavior to generalized chaplygin models @xcite
for the modified chaplygin gas described by eqs .
( [ eqstate ] ) and ( [ effecden])the effective equation of state in the intermediate regime between the dust dominated phase and the de sitter phase can be obtained expanding eqs .
( [ effecden ] ) and ( [ eqstate ] ) in powers of @xmath23 , we get @xmath24 which corresponds to a mixture of vacuum energy density @xmath25 , presureless dust and other perfect fluids which dominate at the very beginning of the universe . in the intermediate regime
the modified chaplygin gas behaves as dust at the time where the energy density satisfies the condition @xmath26 . at very early times
the equation of state parameter @xmath27 becomes @xmath28 so that for very large @xmath29 the dust - like behavior is recovered .
the next step is to investigate what sort of cosmological model arises when we consider a slight inhomogeneous modified chaplygin cosmologies . for a general metric @xmath30 , the proper time @xmath31 , and @xmath32 as the determinant of the induced @xmath33-metric ,
one has _ i j = g_i0g_j0 g_00 - g_ij .
[ inducedmetric ] in the first approximation it will be interesting to investigate the contribution of inhomogeneities introduced in the modified chaplygin gas through the expression = ( a + ) ^ /(-1 ) .
[ genevoldeninh ] the latter result suggests that the evolution of inhomogeneities can be studied using the zeldovich method through the deformation tensor @xcite : d_i^j = a(t ) ( _ i^j - b(t ) ^2 ( ) q^i q^j ) , [ defromationt ] where @xmath34 parametrizes the time evolution of the inhomogeneities and @xmath35 are generalized lagrangian coordinates so that _ i j = _
mnd_i^md_j^n , [ inducedmetric1 ] and @xmath36 is a perturbation h = 2 b(t ) _ ,
[ hevol ] hence , using the equations above and eqs .
( [ rhoapprox ] ) and ( [ papprox ] ) , it follows that @xmath37 where @xmath38 is given by eq .
( [ eqstate ] ) and the density contrast @xmath39 is related to @xmath36 through = h 2 ( 1 + w ) , [ delta ] where @xmath40 . finally , after some algebra we get |p=|(w + ) .
[ omega ] now , the metric ( [ inducedmetric1 ] ) leads to the following @xmath41 component of the einstein equations : - 3 a + 1 2 + h = 4 g |(1 + 3w + ( 1 + ) ) , [ einstein ] where the unperturbed part of this equation corresponds to the raychaudhuri equation - 3 a = 4 g |(1 + 3 w ) .
[ raychaudhuri ] using the friedmann equation for a flat spacetime @xmath42 , one can rewrite eq .
( [ einstein ] ) as a differential equation for @xmath43 : a^2 b
+ ( 1 - w ) a b - ( 1 + w ) ( 1 + ) b = 0 , [ bprimes ] where the primes denote derivatives with respect to the scale - factor , @xmath8 . an expression for @xmath44 as a function of the scale - factor can be derived from eqs .
( [ eqstate ] ) and ( [ effecden ] ) : @xmath45 the latter must be conveniently recast in terms of the fractional vacuum and matter energy densities .
this can be done by using @xmath46 combined with @xmath47 where @xmath48 and @xmath49 are , respectively , the current value of the hubble and scale factor , and @xmath50 and @xmath51 are , respectively , the fractional vacuum and matter energy densities today . setting @xmath52 we obtain @xmath53 and consistently @xmath54 for the modified chaplygin gas for @xmath55 ( continuous lines ) as compared with @xmath0cdm ( dashed line ) and cdm ( dashed - dotted line ) .
lower curves correspond to higher values of @xmath29.,title="fig:",width=264,height=170 ] ( -222,62)@xmath56 ( -90,-6)@xmath8 we have used this expression to integrate eq . ( [ bprimes ] ) numerically , for different values of @xmath29 , and taking @xmath57 and @xmath58 @xcite .
we have set @xmath59 for matter - radiation equilibrium ( while keeping @xmath60 at present ) , and our initial condition is @xmath61 .
our results are shown in figures [ evbsmall ] and [ evblarge ] .
we find that modified chaplygin scenarios start differing from the @xmath0cdm only recently ( @xmath62 ) and that , in any case , they yield a density contrast that closely resembles , for any value of @xmath63 , the standard cdm before the present .
notice that @xmath0cdm corresponds effectively to using eq .
( [ limit ] ) and removing the factor @xmath64 in eq .
( [ bprimes ] ) .
figures [ evbsmall ] and [ evblarge ] show also that , for any value of @xmath29 , @xmath43 saturates as in the @xmath0cdm case . for the modified chaplygin gas for @xmath65 ( continuous lines ) as compared with @xmath0cdm ( dashed line ) and cdm ( dashed - dotted line ) .
lower curves correspond to higher values of @xmath29.,title="fig:",width=264,height=170 ] ( -224,62)@xmath56 ( -90,-6)@xmath8 for the modified chaplygin gas for @xmath66 ( continuous lines ) as compared with @xmath0cdm ( dashed line ) .
lower curves correspond to higher values of @xmath29.,title="fig:",width=264,height=170 ] ( -228,62)@xmath67 ( -90,-6)@xmath8 in what regards the density contrast , @xmath39 , using eqs .
( [ hevol ] ) , ( [ delta ] ) and ( [ wparam ] ) one can deduce that the ratio between this quantity in the modified chaplygin and the @xmath0cdm scenarios is simply given by = b_mchap b_cdm , [ ratio ] and its behavior is depicted in figure [ ratiofig ] .
we find that it asymptotically evolves to a constant value . now , in figure [ contrast ] , we have plotted @xmath39 as a function of @xmath8 for different values of @xmath29 .
as happens in the the traditional @xcite and generalized chaplygin models , in our models the density contrast decays for large @xmath8 also . for the modified chaplygin gas for @xmath68 ( continuous lines ) as compared with @xmath0cdm ( dashed line ) .
lower curves correspond to higher values of @xmath29.,title="fig:",width=264,height=170 ] ( -228,62)@xmath69 ( -90,-6)@xmath8
using a zeldovich - like approximation , we have studied the evolution of large - scale perturbations in a recently proposed theoretical framework for the unification of dark matter and dark energy : the so - called modified chaplygin cosmologies @xcite , with equation of state @xmath70 with @xmath71 .
this model evolves from a phase that is initially dominated by non - relativistic matter to a phase that is asymptotically de sitter .
the intermediate regime corresponds to a phase where the effective equation of state is given by @xmath72 plus a cosmological constant . we have estimated the fate of the inhomogeneities admitted in the model and shown that these evolve consistently with the observations as the density contrast they introduce is smaller than the one typical of cdm scenarios . on general grounds ,
the pattern of evolution of perturbations follows is similar to the one in the @xmath0cdm models and in generalized chaplygin cosmologies , and therefore our represent plausible alternatives alternatives as usual , in modified chaplygin cosmologies , the equation of state parameter @xmath44 can be expressed in terms of the scale factor and a free parameter @xmath29 , and the value of the latter can be chosen so that the model resembles as much as desired the @xmath0cdm model .
it would be very interesting to deepen in the comparison between modified and generalized chaplygin models , particularly from the observational point of view ( as already done in @xcite ) .
it would also be worth generalizing our study by going beyond the zeldovich approximation , to incorporate the effects of finite sound speed .
this can be done by generalizing the spherical model to incorporate the jeans length as in @xcite .
alternatively , following @xcite one could investigate whether the modified chaplygin admits an unique decomposition into dark energy and dark matter , and if that were the case then study structure formation and show that difficulties associated to unphysical oscillations or blow - up in the matter power spectrum can be circumvented .
we hope this will be addressed in future works .
is partially funded by the university of buenos aires under project x224 , and the consejo nacional de investigaciones cientficas y tcnicas under proyect 02205 .
is supported by the university of the basque country through research grant upv00172.310 - 14456/2002 and by the spanish ministry of education and culture through research grant fis2004 - 01626 .
b. ratra and p.j.e .
peebles , ( 1988 ) 3406 ; ( 1988 ) 117 ; c. wetterich , ( 1988 ) 668 ; j. a. frieman , c. t. hill , a. stebbins and i. waga , phys . rev .
* 75 * ( 1995 ) 2077 ; k. coble , s. dodelson and j. a. frieman , phys .
d * 55 * ( 1997 ) 1851 ; m. s. turner and m. j. white , phys .
d * 56 * ( 1997 ) 4439 ; p. g. ferreira and m. joyce , phys .
* 79 * ( 1997 ) 4740 ; p. g. ferreira and m. joyce , phys .
d * 58 * ( 1998 ) 023503 ; e. j. copeland , a. r. liddle and d. wands , phys .
d * 57 * ( 1998 ) 4686 ; r.r .
caldwell and r. dave , p.j .
steinhardt , ( 1998 ) 1582 ; p.g .
ferreira and m. joyce , ( 1998 ) 023503 ; l. m. wang and p. j. steinhardt , astrophys .
j. * 508 * ( 1998 ) 483 ; i. zlatev , l. wang , and p.j .
steinhardt , ( 1999 ) 986 ; p. j. steinhardt , l. m. wang and i. zlatev , phys .
d * 59 * ( 1999 ) 123504 ; i. zlatev and p. j. steinhardt , phys .
b * 459 * ( 1999 ) 570 ; p. bintruy , ( 1999 ) 063502 ; j.e .
kim , 9905 ( 1999 ) 022 ; m.c .
bento and o. bertolami , ( 1999 ) 1461 ; j.p .
uzan , ( 1999 ) 123510 ; t. chiba , ( 1999 ) 083508 ; l. amendola , ( 1999 ) 043501 .
y. fujii , ( 2000 ) 023504 ; a. masiero , m. pietroni and f. rosati , ( 2000 ) 023504 ; m.c .
bento , o. bertolami , and n. santos , ( 2002 ) 067301 .
r. r. caldwell , phys .
b * 545 * ( 2002 ) 23 ; a. e. schulz and m. j. white , phys .
d * 64 * ( 2001 ) 043514 ; b. mcinnes , jhep * 0208 * ( 2002 ) 029 ; r. vaas , bild .
* 2003n8 * ( 2003 ) 52 ; r. r. caldwell , m. kamionkowski and n. n. weinberg , phys .
* 91 * ( 2003 ) 071301 ; j. g. hao and x. z. li , phys .
d * 67 * ( 2003 ) 107303 ; g. w. gibbons , hep - th/0302199 ; x. z. li and j. g. hao , phys .
d * 69 * ( 2004 ) 107303 ; s. nojiri and s. d. odintsov , phys .
b * 562 * ( 2003 ) 147 ; s. nojiri and s. d. odintsov , phys .
b * 565 * ( 2003 ) 1 ; a. yurov , astro - ph/0305019 ; p. singh , m. sami and n. dadhich , phys .
d * 68 * ( 2003 ) 023522 ; j. g. hao and x. z. li , phys .
d * 68 * ( 2003 ) 043501 ; p. f. gonzlez - daz , phys .
d * 68 * ( 2003 ) 021303 ; s. nojiri and s. d. odintsov , phys .
b * 571 * ( 2003 ) 1 ; m. p. dabrowski , t. stachowiak and m. szydlowski , phys .
d * 68 * ( 2003 ) 103519 ; d. j. liu and x. z. li , phys .
d * 68 * ( 2003 ) 067301 ; l. p. chimento and r. lazkoz ,
* 91 * ( 2003 ) 211301 ; j. g. hao and x. z. li , phys .
d * 70 * ( 2004 ) 043529 ; e. elizalde and j. q. hurtado , mod .
a * 19 * ( 2004 ) 29 ; h. stefancic , phys .
b * 586 * ( 2004 ) 5 ; v. b. johri , phys .
d * 70 * ( 2004 ) 041303 ; j. m. cline , s. y. jeon and g. d. moore , phys .
d * 70 * ( 2004 ) 043543 ; m. sami and a. toporensky , mod .
. lett .
a * 19 * ( 2004 ) 1509 ; h. q. lu , hep - th/0312082 ; h. stefancic , eur .
j. c * 36 * ( 2004 ) 523 ; p.
f. gonzlez - daz , phys .
b * 586 * ( 2004 ) 1 ; y. s. piao and y. z. zhang , astro - ph/0401231 ; p. f. gonzlez - daz , phys .
d * 69 * ( 2004 ) 063522 ; m. szydlowski , w. czaja and a. krawiec , astro - ph/0401293 ; e. babichev , v. dokuchaev and y. eroshenko , phys .
* 93 * ( 2004 ) 021102 ; j. m. aguirregabiria , l. p. chimento and r. lazkoz , phys . rev .
d * 70 * ( 2004 ) 023509 ; j. cepa , astro - ph/0403616 ; j. g. hao and x. z. li , phys.lett . b*606 * ( 2005 ) 7 ; z. k. guo , y. s. piao and y. z. zhang , phys .
b * 594 * ( 2004 ) 247 ; m. bouhmadi - lopez and j. a. jimnez madrid , astro - ph/0404540 ; v. faraoni , phys . rev .
d * 69 * ( 2004 ) 123520 ; e. elizalde , s. nojiri and s. d. odintsov , phys .
d * 70 * ( 2004 ) 043539 ; j. e. lidsey , phys .
d * 70 * ( 2004 ) 041302 ; m. g. brown , k. freese and w. h. kinney , astro - ph/0405353 ; l. p. chimento and r. lazkoz , astro - ph/0405518 ; p. f. gonzlez - daz and j. a. jimnez - madrid , phys .
b * 596 * ( 2004 ) 16 ; p. f. gonzlez - daz , hep - th/0408225 ; a. vikman , phys.rev .
d * 71 * ( 2005 ) 023515 ; r. koley and s. kar , mod .
a * 20 * ( 2005 ) 363 ; p. f. gonzalez - daz and c. l. siguenza , nucl .
b * 697 * ( 2004 ) 363 ; p. x. wu and h. w. yu , astro - ph/0407424 ; w. fang , h. q. lu , z. g. huang and k. f. zhang , hep - th/0409080 ; l. amendola , phys.rev.lett .
* 93 * ( 2004 ) 181102 ; y. h. wei , gr - qc/0410050 ; s. nesseris and l. perivolaropoulos , phys.rev .
d * 70 * ( 2004 ) 123529 ; r. j. scherrer , phys . rev .
d * 71 * ( 2005 ) 063519 ; w. hu , phys .
d * 71 * ( 2005 ) 047301 ; e. majerotto , d. sapone and l. amendola , astro - ph/0410543 ; z. k. guo , y. s. piao , x. zhang and y. z. zhang , phys .
b * 608 * ( 2005 ) 177 .
v. k. onemli and r. p. woodard , class .
* 19 * ( 2002 ) 4607 ; v. k. onemli and r. p. woodard , phys .
d * 70 * ( 2004 ) 107301 .
m. c. bento , o. bertolami and a. a. sen , phys .
d * 66 * ( 2002 ) 043507 .
a. kamenshchik , u. moschella , v. pasquier , ( 2001 ) 265 .
l. p. chimento , phys .
d * 69 * ( 2004 ) 123517 .
@xcite t. barreiro and a. a. sen , phys .
d * 70 * ( 2004 ) 124013 .
g. r. dvali , g. gabadadze and m. porrati , phys .
b * 485 * ( 2000 ) 208 ; c. deffayet , phys .
b * 502 * ( 2001 ) 199 .
k. freese and m. lewis , phys .
b * 540 * ( 2002 ) 1 ; p. gondolo and k. freese , phys .
d * 68 * ( 2003 ) 063509 .
g. dvali and m. s. turner , astro - ph/0301510 . t. barreiro and a. a. sen , phys .
d * 70 * ( 2004 ) 124013 .
m. c. bento , o. bertolami and a. a. sen , phys .
d * 67 * ( 2003 ) 063003 ; phys . lett .
b * 575 * ( 2003 ) 172 ; gen .
* 35 * ( 2003 ) 2063 ; d. carturan and f. finelli , phys .
d * 68 * ( 2003 ) 103501 ; l. amendola , f. finelli , c. burigana and d. carturan , jcap * 0307 * ( 2003 ) 005 .
m. c. bento , o. bertolami , n. m. c. santos and a. a. sen , phys .
d * 71 * ( 2005 ) 063501 ; o. bertolami , a. a. sen , s. sen and p. t. silva , mon . not .
* 353 * ( 2004 ) 329 ; r. j. colistete and j. c. fabris , astro - ph/0501519 ; r. j. colistete , j. c. fabris and s. v. b. goncalves , astro - ph/0409245 ; r. j. colistete , j. c. fabris , s. v. b. goncalves and p. e. de souza , int .
j. mod .
d * 13 * ( 2004 ) 669 ; m. biesiada , w. godlowski and m. szydlowski , astro - ph/0403305 ; m. makler , s. quinet de oliveira and i. waga , phys . lett .
b * 555 * ( 2003 ) 1 ; j. c. fabris , s. v. b. goncalves and p. e. d. souza , astro - ph/0207430 ; a. dev , d. jain and j. s. alcaniz , phys .
d * 67 * ( 2003 ) 023515 ; v. gorini , a. kamenshchik and u. moschella , phys .
d * 67 * ( 2003 ) 063509 j. s. alcaniz , d. jain and a. dev , phys .
d * 67 * ( 2003 ) 043514 ; r. bean and o. dore , phys .
d * 68 * ( 2003 ) 023515 .
p. t. silva and o. bertolami , astrophys .
j. * 599 * ( 2003 ) 829 ; a. dev , d. jain and j. s. alcaniz , astron .
astrophys .
* 417 * ( 2004 ) 847 . | we extend the homogeneous modified chaplygin cosmologies to large - scale perturbations by formulating a zeldovich - like approximation .
we show that the model interpolates between an epoch with a soft equation of state and a de sitter phase , and that in the intermediate regime its matter content is simply the sum of dust and a cosmological constant .
we then study how the large - scale inhomogeneities evolve and compare the results with cold dark matter ( cdm ) , @xmath0cdm and generalized chaplygin scenarios .
interestingly , we find that like the latter , our models resemble @xmath0cdm . |
in traditional data centers , applications are tied to specific physical servers that are often over - provisioned to deal with upper - bound workload .
such configuration makes data centers expensive to maintain with wasted energy and floor space , low resource utilization and significant management overhead . with virtualization technology ,
today s cloud data centers become more flexible , secure and provide better support for on - demand allocating . the definition and model defined by this paper
are aimed to be general enough to be used by a variety of cloud providers and focus on the infrastructure as a service ( iaas ) .
cloud datacenters can be a distributed network in structure , containing many compute nodes ( such as servers ) , storage nodes , and network devices .
each node is formed by a series of resources such as cpu , memory , and network bandwidth and so on , which are called multi - dimensional resources ; each has its corresponding properties . under virtualization
, cloud data centers should have ability to migrate an application from one set of resources to another in a non - disruptive manner .
such ability is essential in modern cloud computing infrastructure that aims to efficiently share and manage extremely large data centers .
reactive migration of vms is widely proposed for load balance and traffic consolidation .
one key technology playing an important role in cloud data centers is load balance scheduling .
there are quite many load balance scheduling algorithms .
most of them are for traditional web servers but do not consider vm reservations with lifecycle characteristics .
one of the challenging scheduling problems in cloud data centers is to consider allocation and migration of reconfigurable vms and integrated features of hosting pms .
the load balance problem for vm reservations considering lifecycle is as follows : given a set of @xmath2 identical machines ( pms ) @xmath3 and a set of @xmath4 requests ( vms ) , each request [ @xmath5 , @xmath6 , @xmath7 , has a start - time ( @xmath5 ) , end - time ( @xmath6 ) constraint and a capacity demand ( @xmath8 ) from a pm , the objective of load balance is to assign each request to one of pms so that the loads placed on all machines are balanced or the maximum load is minimized .
this problem is not well studied yet in the open literatures .
the major contributions of this paper are : * providing a modeling approach to vm reservation scheduling with capacity sharing by modifying traditional interval scheduling problem and considering life cycles characteristics of both vms and pms . * designing and implementing load balancing scheduling algorithms , called prepartition for both offline and online scheduling which can prepare migration in advance and set process time bound for each vm on a pm . * deriving computational complexity and quality analysis for both offline and online prepartition . * providing performance evaluation of multiple metrics such as average utilization , imbalance degree , makespan , time costs as well as capacity@xmath1makespan by simulating different algorithms using trace - driven and synthetic data .
the remaining parts of this paper are organized as follows : section 2 discusses the related work on load balance algorithms .
section 3 introduces problem formulation .
section 4 presents prepartition algorithm in details as well as offline and online algorithms are described and compared .
performance evaluations of different scheduling algorithms are shown in section 5 . finally in section 6 , a conclusion is given .
a large amount of work has been devoted to the schedule algorithms and can be mainly divided into two types : online load balance algorithms and offline ones .
the major difference lies in that online schedulers only know current request and status of all pms but offline schedulers know all the requests and status of all pms .
andre et al.@xcite discussed the detailed design of a data center .
armbrust et al.@xcite summarized the key issues and solutions in cloud computing .
foster et al.@xcite provided detailed comparison between cloud computing and grid computing .
buyya et al.@xcite introduced a way to model and simulated cloud computing environments .
wickremasinghe et al.@xcite introduced three general scheduling algorithms for cloud computing and provided simulation results .
wood et al.@xcite introduced techniques for virtual machine migration with spots and proposed a few reactive migration algorithms . zhang@xcite compared major load balance scheduling algorithms for traditional web servers .
singh et al.@xcite proposed a novel load balance algorithm called vectordot which deals with hierarchical and multi - dimensional resources constraints by considering both servers and storage in a cloud .
arzuaga et al.@xcite proposed a quantifying measure of load imbalance on virtualized enterprise servers considering reactive live vm migrations .
galloway et al . in @xcite introduced an online greedy algorithm , in which pms can be dynamic turned on and off but the life cycle of a vm is not considered .
gulati et al @xcite presented challenge issues and distributed resource scheduling ( drs ) as a load balance scheduling for cloud - scale resource management in vmware .
tian et al.@xcite provided a comparative study of major existing scheduling strategies and algorithms for cloud data centers .
sun et al.@xcite presented a novel heuristic algorithm to improve integrated utilization considering multi - dimensional resource .
tian et al.@xcite designed a toolkit for modeling and simulating vm allocation , @xcite@xcite introduced a dynamic load balance scheduling algorithm considering only current allocation period and multi - dimensional resource but without considering life - cycles of both vms and pms .
li et al.@xcite proposed a cloud task scheduling policy based on ant colony optimization algorithm to balance the entire system and minimize the makespan of a given task set .
hu et al.@xcite stated an algorithm named genetic , which can calculate the history data and current states to choose an allocation .
most of existing research does not consider fixed interval constraints of vm allocation .
knauth et al.@xcite introduced energy - efficient scheduling algorithms applying timed instances that have a priori specified reservation time of fixed length , these assumptions are also adopted in this paper .
most of existing research considers reactive vm migrations as a mean for load balance in data centers . to the best of our knowledge ,
proactive vm migration by pre - partition has not been studied yet in the open literatures .
it is one of major objectives in this paper .
in this paper we consider vms reservation and model the vm allocations as a modified interval scheduling problem ( misp ) with fixed processing time . more explanation and analysis about traditional interval scheduling problems with fixed processing time can be found in @xcite and references there in .
we present a general formulation of modified interval - scheduling problem and evaluate its results compared to well - known existing algorithms .
there are following assumptions : + 1 ) all data are deterministic and unless otherwise specified , the time is formatted in slotted windows .
we partition the total time period [ 0 , t ] into slots with equal length @xmath9 , the total number of slots is @xmath10=@xmath11 .
the start time @xmath5 and finish time @xmath6 are integer numbers of one slot .
then the interval of a request can be represented in slot format with ( start - time , finish - time ) .
for example , if @xmath12=5 minutes , an interval ( 3 , 10 ) means that it has start time and finish time at the 3rd - slot and 10th - slot respectively .
the actual duration of this request is ( 10 - 3)@xmath135=35 minutes .
+ 2 ) for all vm reservations , there are no precedence constraints other than those implied by the start - time and finish - time .
+ 3 ) the required capacity of each request is a positive real number between ( 0,1 ] .
notice that the capacity of a single physical machine is normalized to be 1 and the required capacity of a vm can be 1/8 , 1/4 or 1/2 or other portions of the total capacity of a pm .
this is consistent with widely adopted practice in amazon ec2 @xcite and @xcite .
+ a few key definitions are explained as follows : _ definition 1 .
traditional interval scheduling problem ( tisp ) with fixed processing time _ : a set of requests @xmath141 , 2,@xmath15 , @xmath16 where the @xmath17-th request corresponds to an interval of time starting at @xmath5 and finishing at @xmath6 , each request needs a capacity of 1 , i.e. occupying the whole capacity of a machine during fixed processing time . _ definition 2 .
interval scheduling with capacity sharing ( iswcs ) _ : the only difference from tisp is that a resource ( to be concrete , a pm ) can be shared by different requests if the total capacity of all requests allocated on the single resource at any time does not surpass the total capacity that the resource can provide . _
definition 3 .
sharing compatible intervals for iswcs _ : a subset of intervals with total required capacity not surpass the total capacity of a pm at any time , therefore they can share the capacity of a pm . in the literature
, the makespan is used to measure the load balance , which is simply the maximum total load ( processing time ) on any machine .
traditionally , the makespan is the total length of the schedule . in view of the problem in iswcs for vm scheduling
, we redefine the makespan as capacity@xmath1makespan .
_ definition 4 .
capacity@xmath1makespan of a pm @xmath17 _ : in any allocation of vm requests to pms , let @xmath18 denote the set of vm requests allocated to machine @xmath19 .
under this allocation , machine @xmath19 will have total load equal to the sum of product of each required capacity and its duration ( called capacity@xmath1makespan , i.e. , cm for abbreviation in this paper ) , as follows : @xmath20 where @xmath21 is the capacity requests of @xmath22 from a pm and @xmath23 is the span of request @xmath24 ( i.e. , the length of processing time of request @xmath24 ) .
therefore , the goal of load balancing is to minimize the maximum load ( capacity@xmath1makespan ) on any pm .
some other related metrics such as average utilization and makespan are also considered and will be explained in the following section .
assuming there are @xmath2 pms in data centers , the problem of iswcs load balance in it therefore can be formulated as : * @xmath25 * where @xmath21 is the capacity requirement of vm @xmath24 and the total capacity of a pm @xmath17 is normalized to 1 .
the condition 1 ) shows the sharing capacity constraint and condition , 2 ) is for the interval constraint of vm reservations . _
theorem 1 : the offline scheduling problem of finding an allocation of minimizing the makespan in general case is np - complete . _ + the proof can be found in @xcite and is omitted here . in this section ,
a few metrics closely related to iswcs load balance problem will be presented .
some other metrics can be found in @xcite .
+ 1 ) pm resource : + @xmath26 , @xmath17 is the index number of pm , @xmath27,@xmath28,@xmath29 are the cpu , memory , storage capacity of that a pm can provide .
+ 2 ) vm resource : + @xmath30 , @xmath24 is the vm type i d , @xmath31 are the cpu , memory , storage requirements of @xmath22 , @xmath32 are the start time and end time , which are used to represent the life cycle of a vm .
+ 3 ) time slots : we consider a time span from 0 to @xmath33 be divided into slots with same length .
the @xmath4 slots can be defined as @xmath34 $ ] , each time slot @xmath35 means the time span @xmath36 .
+ 4 ) average cpu utilization of @xmath19 during slot 0 and @xmath37 is defined as : @xmath38 where @xmath39 is the average cpu utilization during slot @xmath35 .
average memory utilization ( @xmath40 ) and storage utilization ( @xmath41 ) of both pms can be computed in the same way .
similarly , average cpu ( memory and storage ) utilization of a vm can be computed .
+ 5 ) makespan : the total length of a schedule for a set of vm reservations , i. e. , the difference between the start - time of the first job and the finishing time of the last job .
+ 6 ) the capacity@xmath1makespan ( cm ) of all pms : can be formulated as : @xmath42 + from these equations , we notice that life cycle and capacity sharing are two major differences from traditional metrics such as makespan which only considers process time ( duration ) . traditionally longest process time first ( lpt ) @xcite is widely used for load balance of offline multi - processor scheduling .
reactive ( post ) migration of vms is another popular way of load - balancing .
_ however , reactive migration has difficulty to reach predefined load balance objectives , and may cause interruption and instability of service and other associated costs_. by considering both fixed process intervals and capacity sharing properties in cloud data centers , we propose new offline and online algorithms as follows .
for a given set of vm reservations , let us consider there are @xmath2 pms in a data center and denote opt as the optimal solution for a given set of @xmath43 vm reservations .
firstly define @xmath44 @xmath45 is a lower bound on opt .
algorithm 4.1 shows the pseudocodes of prepartition algorithm .
the algorithm firstly computes balance value by equation ( 7 ) , defines partition value ( @xmath10 ) and finds the length of each partition ( i.e. @xmath46 , which is the max time length a vm can continously run on a pm ) .
for each request , preparition equally partitions it into multiple @xmath46 subintervals if its cm is larger than @xmath46 , and then finds a pm with the lowest average capacity@xmath1makespan and available capacity , and updates the load on each pm .
after all requests are allocated , the algorithm computes the capacity@xmath1makespan of each pm and finds total partition ( migration ) numbers . for practice
, the scheduler has to record all possible subintervals and their hosting pms of each request so that migrations of vms can be conducted in advance to reduce overheads .
initialization : computing the bound @xmath45 value and set the partition value @xmath10 ; sort all intervals in decreasing order of @xmath47 , break ties arbitrarity ; let @xmath48 denote the intervals in this order ; compute cm of each pm and total partitions _ theorem 2 : the computational complexity of prepartition algorithm is @xmath49 using priority queue data structure where @xmath4 is the number of vm requests after pre - partition and @xmath2 is total number of pms used_. + proof : the priority queue is designed such that each element ( pm ) has a priority value ( average capacity@xmath1makespan ) , and each time the algorithm needs to select an element from it , the algorithm takes the one with the highest priority ( the smaller average capacity@xmath1makespan value is , the higher priority it is ) .
sorting @xmath4 numbers in a priority queue takes @xmath50 time and a priority queue performs insertion and the extraction of minima in @xmath51 steps ( detailed proof of the priority queue is shown in @xcite ) .
therefore , by using priority queue or related data structure , the algorithm can find a pm with the lowest average capacity@xmath1makespan in @xmath52 time .
altogether , for @xmath4 requests , prepartition algorithm has time complexity @xmath49 .
_ theorem 3 : the approximation ratio of prepartition algorithm is @xmath53 regarding the capacity_makespan where @xmath54=@xmath55 _ and @xmath10 is the partition value ( a preset constant ) . +
proof : this is because that each request has bounded capacity@xmath1makespan by pre - parition based on ideal lower bound @xmath45 .
we sketch the proof as follows .
each job has start - time @xmath5 , end - time @xmath6 and process time @xmath56=@xmath6-@xmath5 .
consider the last job to finish ( after scheduling all other jobs ) and suppose this job starts at time @xmath57 .
all the machines must have been fully loaded up to capacity_makespan @xmath58 , which gives @xmath59opt . since , for all jobs , we have @xmath60 opt ( by the settting of prepartition algorithm in equation ( 7 ) ) , this job finishes with load @xmath58+@xmath54opt .
hence , the schedule with capacity_makespancan be no more than @xmath58+@xmath54 opt @xmath61 ( 1+@xmath54)opt , this finishes the proof .
+ for online vm allocations , scheduling decisions must be made without complete information about the entire job instances because jobs arrive one by one .
we extend the offline prepartition algorithm to online scenario as prepartitionon .
let us consider there are @xmath2 pms and @xmath62 vms ( including the one just came ) in a data center .
firstly define @xmath63 @xmath64 is called dynamic balance value , which is one half of the max capacity@xmath1makespan of all current pms or the ideal load balance value of all current pms in the system , where @xmath62 is the number of vms requests already arrived .
notice that the reason to set @xmath64 as one half of the max capacity@xmath1makespan of all current pms is to avoid large requests may cause imbalance in some cases .
algorithm 4.2 shows the pseudo codes of prepartitionon algorithm . since in online algorithm ,
the requests come one by one , the system can only capture the information of arrived requests .
when a new request comes into the system , the algorithm computes dynamic balance value by equation ( 8) . to be noticed , @xmath62 represents the number of requests already arrived , and @xmath2 represents the number of pms in use . after the dynamic balanced value ( @xmath64 )
is computed , then the initial request is partitioned into several requests ( segments ) based on the partition value @xmath10 . in these partitioned requests ,
the first one would be executed instantly , which will be allocated to the pm with the lowest capacity_makespan , while others would be put back into the queue waiting to be executed . then the algorithm picks up the next arrived request to follow the same partition and allocation process .
after all requests are allocated , the algorithm computes the capacity@xmath1makespan of each pm and find the total partition numbers for @xmath4 requests .
since the number of partitions and segments of each vm request are known at the moment of allocation , the system can prepare vm migration in advance so that process time and instability of migration can be reduced .
initialization : set the partition value k , total partition number @xmath65=0 ; compute @xmath66 of each pm and output total number of partitions @xmath65 _ theorem 4 : the competitive ratio of prepartitionon is @xmath67 _ regarding the capacity_makespan .
+ proof : without loss of generality , we label pms in order of non - decreasing final loads in prepartitionon .
denote @xmath68 and and @xmath69 respectively as the optimal load balance value of corresponding offline scheduling and load balance value of prepartitionon for a given set of jobs @xmath70 , respectively .
then the load of @xmath71 defines the capacity@xmath1makespan . the first @xmath72 - 1 ) pms each process a subset of the jobs and then experience a ( possibly none ) idle period .
all pms together finish a total capacity@xmath1makespan @xmath73 during their busy periods .
consider the allocation of the last job @xmath24to pm@xmath74 . by the scheduling rule of prepartitionon ,
pm@xmath74 had the lowest load at the time of allocation .
hence , any idle period on the first ( @xmath2 - 1 ) pms can not be bigger than the capacity@xmath1makespan of the last job allocated on pm@xmath74 and hence can not exceed the maximum capacity@xmath1makespan divided by @xmath10 ( partition value ) , i.e. , @xmath75 .
we have + @xmath76 which is equivalent to @xmath77 which is @xmath78 note that @xmath79 is the lower bound on @xmath80 because the optimum capacity@xmath1makespan can not be smaller than the average capacity@xmath1makespan on all pms . and @xmath81 since the largest job must be processed on a pm .
we therefore have @xmath82 .
_ theorem 5 : the computational complexity of prepartitionon is @xmath49 using priority queue data structure , where @xmath4 is the number of vm requests after pre - partition and @xmath2 is the total number of pms used_. + proof : the proof is exactly the same as in the proof for theorem 2 , we therefore omit it .
.8 types of virtual machines ( vms ) in amazon ec2 [ cols="<,<,<,<",options="header " , ] + in this part , we will present the simulation results between prepartition algorithms and other existed algorithms . to achieve this goal
, we used a java simulator cloudsched ( see tian et al .
@xcite ) . for simulation , to be realistic and reasonable , we adopt data both from normal distribution and lawrence livermore national lab ( llnl ) trace , see @xcite for detailed introduction about the trace . all simulations are conducted on a computer configured with intel i5 processor at 2.5ghz and 4 gb memory .
all vm requests are generated following normal distribution . in offline algorithm comparisons , round - robin ( rr ) algorithm , longest process time ( lpt ) algorithm and post migration algorithm ( pmg )
are implemented .
\1 ) round - robin algorithm ( r - r ) : a traditional load balancing scheduling algorithm by allocating the vm requests in turn to each pm that can provide required resource .
+ 2 ) longest processing time first ( lpt ) : it sorts the vm requests by processing time in decreasing order firstly . then allocating the requests in that order to the pm with the lowest load . in this paper ,
the lowest load means the lowest capacity@xmath1makespan of all pms .
+ 3 ) post migration algorithm ( pmg ) : firstly , it processes the requests in the same way as lpt does .
then the average capacity@xmath1makespan of all jobs is calculated .
the up - threshold and low - threshold of the capacity@xmath1makespan for the post migration are calculated through the average capacity@xmath1makespan multiplied by a factor ( in this paper we set the factor as 0.1 , so the up - threshold is average capacity@xmath1makespan multiplied by 1.1 and the low - threshold is multiplied by 0.9 ) .
off course the factor can be set dynamically to meet different requirements ; however , the larger the factor is , the higher imbalance is .
a migration list is formed by collecting the vms taken from pms with capacity@xmath1makespan higher than the low - threshold .
the vms would be taken from a pm only if the operation would not lead the capacity@xmath1makespan of the pm to be less than the low threshold .
after that , the vms in the migration list would be re - allocated to a pm with capacity@xmath1makespan less than the up - threshold .
the vms would be allocated to a new pm only if the operation would not lead the capacity@xmath1makespan of the pm to be higher than the up - threshold
. there may be still some vms left in the list , finally the algorithm allocates the left vms to the pms with the lowest capacity@xmath1makespan until the list is empty .
+ in this paper , we adopt the amazon ec2 configuration of vms and pms as shown in table 1 and 2 .
note that one compute unit ( cu ) has equivalent cpu capacity of a 1.0 - 1.2 ghz 2007 opteron or 2007 xeon processor @xcite .
_ observation 1 .
pmg is a best - effort trial heuristic for load balance .
it does not guarantee a bounded or predefined load balance objective .
this is validated in the following performance evaluation section . _ as for realistic data , we adopt the log data at lawrence livermore national lab ( llnl ) @xcite .
the log contains months of records collected by a large linux cluster and has characteristics consistent with our problem model .
each line of data in that log file includes 18 elements , while we only need the request - id , start - time , duration and number of processors ( capacity demands ) in our simulation .
we convert the units from seconds in llnl log file into minutes , because we set 5 minutes as a time slot length mentioned in previous section .
fig.1 and fig.2 show the average utilization , imbalance degree , makespan and capacity@xmath1makespan comparison for different algorithms with llnl data trace .
from these figures , we can notice that prepartition algorithm has better performance than other algorithms in average utilization , imbalance degree , makespan , capacity@xmath1makespan .
prepartition algorithm has 10@xmath0 - 20@xmath0 higher average utilization than pmg and lpt , and 40@xmath0 - 50@xmath0 higher average utilization than random - robin ( rr ) .
prepartition algorithm has 10@xmath0 - 20@xmath0 lower average makespan and capacity@xmath1makespan than pmg and lpt , 5% imbalance degree than lpt and 40@xmath0 - 50@xmath0 lower average makespan and capacity@xmath1makespan than random - robin ( rr ) . with the partition value @xmath83
, pmg algorithm has a quite similar value in imbalance value , so besides the above evaluations , we also vary the partition number @xmath10 from 4 , 8 to 10 to compare the imbalance degree affects . in figure 3 , we can notice that larger @xmath10 value will induce a lower imbalance degree .
similarly , with a larger value , larger average utilization , lower makespan and capacity_makespan can be acquired . however , increasing the @xmath10 value will bring side - effects .
the dominant one is running time , in fig .
4 , we compare the time costs under different partition value @xmath10 , prepartition algorithm with @xmath84 costs about 10% more running time , and with @xmath85 it costs 15% more running time than prepartition algorithm with @xmath83 on the average . it is easy to understand that a larger @xmath10 value will produce a better load balance , which leads to more partitions , and more partitions need more time to proceed . + _ observation 2 .
whatever numbers of migrations to taken , post migration algorithm ( pmg ) just can not achieve the same level of average utilization , makespan and capacity@xmath1makespan as prepartition does_. this is because that prepartition works in a much more refined and desired scale by prepartition based on reservation data while pmg is just a best - effort trial by migration .
we set 5 minutes as a slot , so 12 slots are for an hour , 288 slots are for a day .
all requests satisfy the normal distribution , with parameters mean @xmath86 and standard deviation @xmath87 as 864 ( three days ) and 288 ( one day ) respectively .
after requests are generated in this way , we start the simulator to simulate the scheduling effects of different algorithms and comparison results are collected . for collecting data , we firstly fix the @xmath10 value of prepartition algorithm as 4 ; different types of vms wit equal probabilities
. then we change the vms numbers from @xmath88 , @xmath89 , @xmath90 and @xmath91 to trace the tendency .
each set of data is the average values of 10-runs .
fig.5 to fig.6 show the average utilization , makespan and capacity@xmath1makespan comparison of different algorithms respectively . from these figures
, we can notice that prepartition algorithm has 10@xmath0 - 20@xmath0 higher average utilization than pmg and lpt , and 40@xmath0 - 50@xmath0 higher average utilization than random - robin ( rr ) ; prepartition algorithm has 8@xmath0 - 13@xmath0 lower average makespan and capacity@xmath1makespan than pmg and lpt , and 40@xmath0 - 50@xmath0 lower average makespan and capacity@xmath1makespan than random - robin ( rr ) .
we can also notice that the pmg algorithm can improve the performance of lpt algorithm .
lpt algorithm is better than r - r algorithm .
similar results are observed for the comparison of makespan .
the performance improvement of mig algorithm is obtained from the extra migration operations .
the vm migration enables a better load balance . in this part
, we will present the simulation results between prepartitionon algorithm and other three existed algorithms .
random , round - robin , online resource scheduling algorithm ( olrsa ) @xcite and prepartitionon algorithm are implemented to compare : + 1 ) random algorithm : a scheduling algorithm that randomly allocates the requests to a pm that can provide required resource .
+ 2 ) round - robin algorithm(r - r ) : a traditional load balancing scheduling algorithm by allocating the vm requests in turn to each pm that can provide required resource .
+ 3 ) olrsa algorithm : an online scheduling algorithm , it computes the capacity@xmath1makespan of each pm and sort the pm by capacity@xmath1makespan in descending order .
this algorithm always allocates the request to the pm with the lowest capacity_makespan and required resource .
+ for realistic data , we utilize the log data at lawrence livermore national lab ( llnl)@xcite because the data is suitable for our research problem . fig .
7 to fig . 8
illustrate the comparisons of the average utilization , imbalance degree , makespan , capacity_makespan . from these figures
, we can notice that prepartitionon shows the highest average utilization , lowest imbalance degree , and lowest makespan .
as for capacity_makespan , olrsa has been proved much better performance compared with random and round - robin algorithms , and prepartitionon still improves 10@xmath0 - 15@xmath0 in average utilization , 20@xmath0 - 30@xmath0 in imbalance degree , and 5@xmath0 to 20@xmath0 in makespan than olrsa .
we set 5 minutes as a slot , so 12 slots are for an hour , 288 slots are for a day .
all requests satisfy the normal distribution , with parameters mean @xmath86 and standard deviation @xmath87 as 864 ( 3 days ) and 288 ( 1 day ) .
we set that different types of vms have equal probabilities , then we change the requests generation approach to produce different size of requests to trace the tendency . from fig .
9 to 10 , we can see that prepartitionon has better performance in average utilization , imbalance degree , makespan and capacity_makespan . comparing to olrsa , prepartitionon still improves about 10@xmath0 in average utilization , 30@xmath0 - 40@xmath0 in imbalance degree , 10@xmath0 - 20@xmath0 in makespan , as well as 10@xmath0 - 20@xmath0 in capacity_makespan .
it is apparent that large @xmath10 values may bring side effects since it will need more number of partitions . in fig .
11 , we compare the time costs ( simulated with llnl data and the time unit is mini second ) under different partition value @xmath10 , prepartitionon algorithm with @xmath92 takes about 10% less running time than that with @xmath10=4 , and @xmath93 takes 15% less running time than that with @xmath83 .
it is easy to understand that a larger @xmath10 value will produce a better load balance with longer process time .
we also observe that larger @xmath10 value will induce a lower capacity_makespan value .
similarly , with a larger @xmath10 value , larger average utilization , lower imbalance degree and makespan are obtained .
in this paper , to reflect the feature of capacity sharing and fixed interval constraint of vm scheduling in cloud data centers , we propose new offline and online load balancing algorithms .
theoretically we prove that offline prepartition is a ( 1+@xmath54)-approximation where @xmath54=@xmath55 and @xmath10 is a positive integer . by increasing @xmath10
it is possible to be very close to optimal solution , i.e. , by setting @xmath10 value , it is also possible to achieve predefined load balance goal as desired because offline prepartition is a ( 1+@xmath55)-approximation and online prepartition ( prepartitionon ) has competitive ratio @xmath67 .
both synthetic and trace driven simulations have validated theoretical observations and shown prepartition algorithm has better performance than a few existing algorithms at average utilization , imbalance degree , makespan , and capacity@xmath1makespan both for offline and online algorithms .
there are still a few research issues can be considered : * making suitable choice between total partition numbers and load balance objective .
prepartition algorithm can achieve desired load balance objective by setting suitable @xmath10 value .
it may need large number of partitions so that the number of migrations can be large depending on the characteristics of vm requests .
for example in ec2 @xcite , the duration of vm reservations varies from a few hours to a few months , we can classify different types of vms based on their durations ( capacity@xmath1makespans ) firstly , then applying prepartition will not have large partition number for each type . in practice
we need analyzing traffic patterns to make the number of partitions ( premigrations ) reasonable so that the total costs , including running time and migrations , are not very high . *
considering heterogeneous configuration of pms and vms .
we mainly consider that a vm requires a portion of total capacity from a pm .
this is also applied in ec2 and knauth et al .
when this is not true , multi - dimensional resources such as cpu , memory and bandwidth etc .
have to be considered together or separately in the load balance , see @xcite and @xcite for a detailed discussion about considering multi - dimensional resources . *
considering precedence constraints among different vm requests .
in reality , some of vm reservations may be more important than others , we should extend current algorithm to consider this case .
this research is partially supported by china national science foundation ( cnsf ) with project i d 61450110440 .
partial results , especially the offline scheduling algorithm of this paper was presented in the conference of icc 2014 , sydney , australia@xcite .
99 l. andre , et al . , the datacenter as a computer : an introduction to the design of warehouse - scale machines , ebook . 2009 .
m. armbrust et al . , above the coulds : a berkeley view of cloud computing , technical report , 2009 .
e. arzuaga , d. r. kaeli , quantifying load imbalance on virtualized enterprise servers , in the proceedings of wosp / sipew 10 , january 28 - 30 , 2010 , san jose , california , usa .
r. buyya , r. ranjan and r. n. calheiros , modeling and simulation of scalable cloud computing environments and the cloudsim toolkit : challenges and opportunities , proceedings of the 7th high performance computing and simulation conference ( hpcs 2009 , isbn : 978 - 1 - 4244 - 4907 - 1 , ieee press , new york , usa ) , leipzig , germany , june 21 - 24 , 2009 .
e. g. coffman jr . , m. r. garey , and d. s. johnson , bin - packing with divisible item sizes , j. complexity , 406 - 428 , 3(1987 ) .
i. foster , y. zhao , i. raicu , s. lu , cloud computing and grid computing 360-degree compared , ieee international workshop on grid computing environments ( gce ) 2008 , co - located with ieee / acm supercomputing 2008 .
j. m. galloway , k. l. smith , s. s. vrbsky , power aware load balancing for cloud computing , proceedings of the world congress on engineering and computer science 2011 vol i wcecs 2011 , october 19 - 21 , 2011 .
r. l. graham _
bounds on multiprocessing timing anomalies _ , siam journal on applied mathematics , vol.17 , no.2 , pp.416 - 429 , 1969 .
a. gulati , g. shanmuganathan , a. holler , i. ahmad , cloud - scale resource management : challenges and techniques , vmware technical journal , 2011 .
j. hu ; j.a gu ; g. sun , et al . , a scheduling strategy on load balancing of virtual machine resources in cloud computing environment , parallel architectures , algorithms and programming ( paap ) , 2010 third international symposium on , pp.89 - 96 , 18 - 20 dec .
t. knauth , c. fetzer , energy - aware scheduling for infrastructure clouds , in the proceedings of cloudcom 2012 .
j. kleinberg , e. tardos , algorithm design , pearson education inc .
k. li , g. xu , g. zhao , et al . , cloud task scheduling based on load balancing ant colony optimization , chinagrid , sixth annual chinagrid conference , pp.3 - 9 , 2011 .
a. singh , m. korupolu , d. mohapatra , server - storage virtualization : integration and load balancing in data centers , international conference for high performance computing , networking , storage and analysis , 2008 .
x. sun , p. xu , k. shuang , et al .
, multi - dimensional aware scheduling for co - optimizing utilization in data center , china communications 8(6 ) , pp.19 - 27 , 2011 .
w. tian , adaptive dimensioning of cloud data centers : in the proceeding of the 8th ieee international conference on dependable , automatic and secure computing , 2009 .
w. tian , y. zhao , m. xu , y. zhong , x. sun , a toolkit for modeling and simulation of real - time virtual machine allocation in a cloud data center .
automation science and engineering , ieee transactions on , 2015 , 12(1 ) : 153 - 161
. w. tian , y. zhao , y .
zhong , m. xu , c. jing , dynamic and integrated load - balancing scheduling algorithms for cloud datacenters , china communications , , vol . 8 issue ( 6 ) : 117 - 126 , 2011 .
w. tian , x. liu , c. jin , y. zhong , lif : a dynamic scheduling algorithm for cloud data centers considering multi - dimensional resources , journal of information @xmath94 computational science , aug . 12 , 2013 .
w. tian , m. xu , y. chen , et al .
prepartition : a new paradigm for the load balance of virtual machine reservations in data centers[c ] , communications ( icc ) , 2014 ieee international conference on .
ieee , 2014 : 4017 - 4022 .
b. wickremasinghe et al . , cloudanalyst : a cloudsim - based tool for modelling and analysis of large scale cloud computing environments , proceedings of the 24th ieee international conference on advanced information networking and applications ( aina 2010 ) , perth , australia , april 20 - 23 , 2010 .
t. wood , et .
black - box and gray - box strategies for virtual machine migration , in the proceedings of symp . on networked systems design and implementation ( nsdi ) , 2007 .
w. zhang , research and implementation of elastic network service , phd dissertation , national university of defense technology , china ( in chinese ) 2000 .
m. xu , w. tian , an online load balancing algorithm for cloud data centers considering real - time virtual machine allocation with fixed process intervals , journal of information and computational science , vol .
11 , pp . 989 - 1001 , 2014 .
h. zheng , l. zhou , j. wu , design and implementation of load balancing in web server cluster system , journal of nanjing university of aeronautics @xmath94 astronautics , vol.38 no . 3 jun . 2006 .
hebrew university , experimental systems lab , www.cs.huji.ac.il/labs/parallel/workload , 2013 .
amazon , amazon elastic compute cloud , http://aws.amazon.com/ec2/ , 2013 . | it is significant to apply load - balancing strategy to improve the performance and reliability of resource in data centers .
one of the challenging scheduling problems in cloud data centers is to take the allocation and migration of reconfigurable virtual machines ( vms ) as well as the integrated features of hosting physical machines ( pms ) into consideration . in the reservation model ,
the workload of data centers has fixed process interval characteristics . in general ,
load - balance scheduling is np - hard problem as proved in many open literatures .
traditionally , for offline load balance without migration , one of the best approaches is lpt ( longest process time first ) , which is well known to have approximation ratio 4/3 . with virtualization , reactive ( post ) migration of vms after
allocation is one popular way for load balance and traffic consolidation .
however , reactive migration has difficulty to reach predefined load balance objectives , and may cause interruption and instability of service and other associated costs . in view of this , we propose a new paradigm , called prepartition , it proactively sets process - time bound for each request on each pm and prepares in advance to migrate vms to achieve the predefined balance goal .
prepartition can reduce process time by preparing vm migration in advance and therefore reduce instability and achieve better load balance as desired .
trace - driven and synthetic simulation results show that prepartition for offline scheduling has 10@xmath0 - 20@xmath0 better performance than the well known load balancing algorithms with regard to average utilization , imbalance degree , makespan as well as capacity@xmath1makespan .
we also apply the prepartition to online ( prepartitionon ) load balance and compare it with existing online scheduling algorithms , in which prepartitionon can improve 8%-20% performance with regard to average cpu utilization , imbalance degree , makespan as well as capacity@xmath1makespan .
both theoretical and experimental results are provided .
cloud computing , physical machines ( pms),virtual machines ( vms),reservation model , load balance scheduling |
planetary nebulae ( pne ) are suitable sites to study the evolution of dust grains since their progenitors , asymptotic giant branch ( agb ) stars , are among the most efficient sources of circumstellar dust in the galaxy @xcite .
these stars lose large amounts of material and create circumstellar envelopes with both low temperatures ( below 1500 k in the regions of dust formation ) and high densities ( @xmath3 10@xmath4 @xmath5 ) , where dust grains are efficiently formed . in @xcite , we studied iron depletions into dust grains in a sample of 33 galactic disk pne .
iron is an important contributor to the mass of dust grains @xcite , and its gaseous abundance can be used as a proxy for the amount of depletion of other elements .
we found low iron gaseous abundances in all the objects , suggesting that more than 90% of their iron atoms are located in grains .
we also found that the range of iron depletions covers about two orders of magnitude , which may be a consequence of differences in the formation and evolution of the grains in the different nebulae .
we did not find any obvious correlation between iron abundances and parameters related to the nebular age ( such as the nebular radius or the surface brightness ) or to the dominant chemistry in the nebulae ( c - rich or o - rich ) that could provide clues on the reasons behind the different depletions .
however , the information available for the studied sample was scarce .
on the other hand , the dust features that appear in the infrared spectra of pne provide information on the composition of their dust grains . in our galaxy ,
pne with crystalline silicates , many of them also with pahs , are the most numerous @xcite .
besides , a general correlation has been found between infrared features and c / o abundance ratios : magellanic clouds pne with silicates have c / o @xmath0 , and those with carbonaceous dust have c / o @xmath6 , although for galactic pne the situation is less clear @xcite .
the relative abundances of carbon and oxygen in the envelope of an agb star depends on the nucleosynthesis processes occurring in the inner regions , which in turn depends on the mass of the progenitor star and on its initial metallicity . in our galaxy ,
we expect stars to be born with an almost solar c / o ratio ( c / o @xmath7 , @xcite ) , that may change during the evolution of the star depending on its initial mass .
theoretical models ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ) predict that the less massive progenitors ( m@xmath8 @xmath92 m@xmath10 ) will remain o - rich during their whole evolution .
stars with masses in the range @xmath3 24 m@xmath10 may evolve from o - rich to c - rich due to the third dredge - up process that enriches the surface of the agb star with carbon . in the most massive progenitors ( m@xmath8 @xmath11 45 m@xmath10 ) , the hot bottom burning process occurs efficiently and , since it converts carbon into nitrogen via the cn cycle , it can prevent the formation of a c - rich star .
the value of c / o in a dust - forming stellar envelope defines whether the dust particles will be c - rich or o - rich .
if carbon is more abundant than oxygen , all the oxygen atoms will be trapped in co molecules and carbon atoms will be available to form c - rich compounds , such as graphite , amorphous carbon , or silicon carbide ( sic ) .
if oxygen is more abundant than carbon , all the carbon atoms will be trapped by co and the remaining oxygen atoms will form o - rich grains like silicates and oxides .
oxygen rich nebulae could have iron deposited in metallic iron grains , silicates and oxides , whereas c - rich nebulae are expected to have their iron atoms in the form of metallic grains , fe@xmath12c , fesi , fes , and fes@xmath13 @xcite . therefore , iron depletion factors could be different in c - rich and o - rich pne .
the c - rich or o - rich chemistry may be studied from the c / o abundance ratios or from the infrared dust features , and we expect both to be related . in principle , c - rich features such as sic are expected in pne with c / o @xmath6 ( but see * ? ? ?
* ) whereas o - rich features like silicates are expected in pne with c / o @xmath0 .
_ iso _ and _ spitzer _ telescopes allow us to detect and identify several dust species in the spectra of pne and therefore , to classify pne as c - rich or o - rich according to the observed features . here
, we extend the work performed in @xcite by studying a sample of galactic pne that nearly doubles the previous one and also by analyzing possible correlations between their iron depletions , c / o abundance ratios , and infrared dust features .
the paper is organized as follows . in section 2
we describe our sample of galactic pne , whereas in section 3 we present the calculation of electron temperatures and densities for all the objects .
section 4 contains an extensive discussion of the iron abundance and depletion factor calculations .
section 5 details the calculation of c / o abundance ratios from collisionally excited lines ( cels ) and recombination lines ( rls ) , and the comparison between these two estimates .
the compilation of infrared dust features observed in the sample is presented in section 6 . in section 7
we discuss our results and study the correlation between iron abundances , c / o values , and dust features in our sample of pne .
our conclusions are presented in section 8 .
our sample consists of 56 galactic pne with published spectra of relatively high spectral resolution and deep enough to detect the lines we need to calculate physical conditions and iron abundances .
the sample includes 28 galactic pne studied in @xcite , but we exclude nebulae with electron densities above 25000 @xmath14 since it is difficult to derive reliable physical conditions and chemical abundances in these objects @xcite . besides , we set an upper limit on the @xmath15 intensity ratio of 0.6 , in order to avoid high excitation pne for which the iron correction scheme adopted here may not apply @xcite . the final sample of 56 pne is presented in table [ tab:1 ] .
lllll & @xmath16 & @xmath17 & @xmath18 & 1 + & @xmath19 & @xmath20 & @xmath21 & 2 + & @xmath22 & @xmath23 & @xmath24 & 2 + & @xmath25 & @xmath26 & @xmath27 & 1 + & @xmath28 & @xmath29 & @xmath30 & 1 + & @xmath31 & @xmath32 & @xmath33 & 1 + & @xmath34 & @xmath35 & @xmath36 & 2 + & @xmath37 & @xmath38 & @xmath39 & 2 + & @xmath40 & @xmath41 & @xmath42 & 3 + & @xmath43 & @xmath44 & @xmath45 & 2 + & @xmath46 & @xmath47 & @xmath48 & 4 + & @xmath49 & @xmath50 & @xmath51 & 5 + & @xmath52 & @xmath53 & @xmath54 & 6 + & @xmath55 & @xmath56 & @xmath57 & 6 + & @xmath58 & @xmath59 & @xmath60 & 7 + & @xmath61 & @xmath62 & @xmath63 & 1 + & @xmath64 & @xmath65 & @xmath66 & 8 + & @xmath67 & @xmath68 & @xmath69 & 9 + & @xmath70 & @xmath71 & @xmath72 & 7 + & @xmath73 & @xmath74 & @xmath75 & 1 + & @xmath76 & @xmath77 & @xmath57 & 10 + & @xmath78 & @xmath79 & @xmath80 & 2 + & @xmath81 & @xmath82 & @xmath83 & 1 + & @xmath84 & @xmath85 & @xmath86 & 1 + & @xmath87 & @xmath88 & @xmath89 & 1 + & @xmath90 & @xmath91 & @xmath75 & 1 + & @xmath92 & @xmath93 & @xmath94 & 1 + & @xmath95 & @xmath96 & @xmath97 & 10 + & @xmath98 & @xmath99 & @xmath100 & 1 + & @xmath101 & @xmath102 & @xmath103 & 1 + & @xmath104 & @xmath105 & @xmath106 & 1 + & @xmath107 & @xmath108 & @xmath42 & 1 + & @xmath109 & @xmath110 & @xmath111 & 6 + & @xmath112 & @xmath113 & @xmath114 & 5 + & @xmath115 & @xmath116 & @xmath117 & 7 + & @xmath118 & @xmath119 & @xmath120 & 6 + & @xmath121 & @xmath122 & @xmath123 & 6 + & @xmath124 & @xmath125 & @xmath126 & 7 + & @xmath127 & @xmath128 & @xmath129 & 6 + & @xmath130 & @xmath131 & @xmath132 & 6 + & @xmath133 & @xmath134 & @xmath106 & 11 + & @xmath135 & @xmath136 & @xmath120 & 5 + & @xmath137 & @xmath138 & @xmath139 & 1 + & @xmath140 & @xmath141 & @xmath142 & 12 + & @xmath143 & @xmath131 & @xmath139 & 1 + & @xmath144 & @xmath145 & @xmath139 & 5 + & @xmath146 & @xmath147 & @xmath120 & 1 + & @xmath118 & @xmath131 & @xmath148 & 5 + & @xmath149 & @xmath150 & @xmath151 & 5 + & @xmath152 & @xmath153 & @xmath154 & 2 + & @xmath155 & @xmath156 & @xmath157 & 6 + & @xmath158 & @xmath159 & @xmath132 & 5 + & @xmath160 & @xmath161 & @xmath162 & 5 + & @xmath163 & @xmath138 & @xmath164 & 2 + & @xmath165 & @xmath166 & @xmath167 & 5 + & @xmath168 & @xmath169 & @xmath170 & 1
we derive electron densities and temperatures , @xmath171 and @xmath172 , using the routine _ temden _ from the package _ nebular _ @xcite in iraf . for each object , we calculate two temperatures from the observed intensity ratios [ ] @xmath173 and [ ] @xmath174 to characterize the low- and high - ionization regions , respectively . we also derive an average density from the available diagnostic ratios among the following : [ ] @xmath175 , [ ] @xmath176 , [ ] @xmath177 , and [ ] @xmath178 .
we use the default atomic data in iraf except for cl@xmath179 and o@xmath180 : for these ions we find densities in better agreement with those obtained from the other diagnostic ratios using the collision strengths of @xcite and @xcite and the transition probabilities of @xcite and @xcite .
@xcite found that a significant fraction of the intensity of the [ ] @xmath1815755 line can arise from recombinations of n@xmath179 .
if this effect is not taken into account , the value of @xmath172 [ ] can be overestimated in those objects where n@xmath179 is an important contributor to the total abundance of nitrogen .
the contribution of recombination to [ ] @xmath1815755 can be estimated using the expression derived by @xcite , which depends on the n@xmath179 abundance .
this ionic abundance can be obtained either from infrared / ultraviolet collisionally excited lines or from optical recombination lines , leading in each case to different results .
thus , the correction is somewhat uncertain and we do not apply it here .
however , we estimate an upper limit of the effect it could have on our results by using the highest possible values for the n@xmath179 abundances , which are those implied by recombination lines . we find
that in many of our objects @xmath172 [ ] would change by less than 10% .
the objects with the largest differences in @xmath172 [ ] are dddm 1 , ic 3568 , ic 4699 , m 3 - 32 , and ngc 3242 , where they reach 3050% .
for these objects , this implies values of fe / o up to 0.4 dex lower or up to 0.2 dex higher , depending on which ionization correction approach we use ( see [ sec : tot_ab ] ) .
hence , the iron abundances derived for these 5 pne have this additional amount of uncertainty .
we calculate the uncertainties of the physical conditions and the ionic and total abundances of o and fe using montecarlo simulations , which assume that the errors in the line intensities follow gaussian distributions .
the line intensity errors are those given in the references listed in table [ tab:1 ] .
each montecarlo run samples the gaussian distributions of all the lines in order to provide new values for @xmath172 , @xmath171 , and all the ionic and total abundances described in
[ sec : ionic_ab ] and [ sec : tot_ab ] below .
the errors listed for the calculated quantities define a confidence interval of 68@xmath182 ( equivalent to one standard deviation in a gaussian distribution ) .
table [ tab:1 ] presents the physical conditions and their associated uncertainties for our sample pne .
the values are similar to the ones obtained in @xcite for the objects in common .
the highest differences are for the values of @xmath171 , and there are two reasons for that .
first , we use here one more diagnostic ratio for the density , [ ] @xmath175 .
second , whereas @xcite derive the final @xmath171 as the weighted mean of the values implied by the available diagnostic ratios , here the final @xmath171 is the median value of the density distribution obtained from averaging the @xmath171 values calculated in each run of the montecarlo simulation .
one of the objects with significant differences is ngc 3587 , where @xcite used the [ ] density diagnostic and found @xmath171 @xmath183 @xmath14 , whereas here we also use the [ ] diagnostic and find @xmath171@xmath184 @xmath14 .
this change in @xmath171 implies a change in @xmath172 [ ] from 9800 k to 10900 k. in general , @xmath172differences are smaller , around 100 k in most of the objects .
the new values derived for the physical conditions imply oxygen abundances that are within @xmath1850.15 dex of the old values .
these differences illustrate the abundance uncertainties introduced by the approach chosen to perform the calculations .
the total iron abundances of the sample pne are calculated from the fe@xmath179 abundance ( and fe@xmath186 in some cases ) along with ionization correction factors ( icfs ) based on the ionic and total oxygen abundances . on the other hand ,
the icf we use for oxygen is based on the he@xmath186 and he@xmath179 abundances .
hence , we derive all these ionic abundances using the densities and temperatures from table [ tab:1 ] , in particular @xmath172 [ ] for o@xmath186 , fe@xmath186 , and fe@xmath179 , and @xmath172 [ ] for o@xmath179 , he@xmath186 , and he@xmath179 . the values of o@xmath186/h@xmath186 and o@xmath179/h@xmath186 are calculated using the intensities of the [ ] @xmath187 and [ ] @xmath188 lines with respect to h@xmath189 and the routine _ ionic _ of the package _ nebular _ in iraf with the changes in atomic data mentioned in [ sec : phys_cond ] .
the values of he@xmath186/h@xmath186 are based on @xmath1816678 and the calculations of @xcite . in dddm 1 and ic 2165
we replace our computed he@xmath186 abundances ( he@xmath186/h@xmath190 and 0.042 ) with the average of the he@xmath186 abundances derived from the @xmath181@xmath1814471 , 5876 lines by @xcite and @xcite , respectively .
these authors suggest that the low values implied by the @xmath1816678 line could be due to underlying absorption . the values of he@xmath179/h@xmath186
are derived using @xmath1814686 and the emissivities of @xcite .
the emissivities are also from @xcite .
we estimate the fe@xmath186 abundances for eleven pne using the emissivities of @xcite and [ ] @xmath191 , one of the few [ ] lines which are not affected by fluorescence @xcite . in some cases where this line is not available we use instead [ ] @xmath192 and @xmath193 , as found by @xcite in several regions .
the values of fe@xmath186/fe@xmath179 for these eleven pne go from 0.06 ( m 3 - 7 ) to 0.80 ( ngc 6565 and ngc 6741 ) .
only five of the sample pne would require a value for the fe@xmath186 abundance in the ionization correction scheme described below , in
[ sec : tot_ab ] .
the two of them where it is available show that its contribution is not likely to be critical , since it only increases the value of fe / h by 0.08 dex ( in ngc 40 ) and 0.17 dex ( in ic 418 ) .
some of the calculated values of fe@xmath186/h@xmath186 are also used in the calculations described in [ withouticf ] .
the fe@xmath179 ionic abundances are derived by solving the equations of statistical equilibrium @xcite for a 34-level model atom , using the collision strengths of @xcite and the transition probabilities of @xcite .
we used pyneb @xcite to compare the results obtained with these atomic data with those implied by the recent calculations of @xcite .
in the three nebulae with the largest numbers of [ ] lines ( cn 1 - 5 , m 2 - 4 , and m 2 - 6 ) , we find differences lower than 5% in the final fe@xmath179 abundances , but the older atomic data lead to a better agreement in the fe@xmath179 abundances derived from all the available [ ] lines , with standard deviations of 32%/39% for the old / new atomic data
( cn 1 - 5 ) , 12%/25%(m 2 - 4 ) , and 23%/25% ( m 2 - 6 ) .
we performed a similar comparison using 14 out of the 15 [ ] lines measured by @xcite in the orion nebula ( we excluded the line at 4926 because it leads to much higher abundances with both sets of atomic data ) .
we find that the new atomic data lead to a fe@xmath179 abundance that is 15% lower , whereas the standard deviation changes from 15% to 30% .
hence , we perform our calculations using the older set of atomic data .
we use up to 9 [ ] lines for each object , excluding those suspected to be contaminated with nearby recombination lines .
for the eight pne in the sample with no [ ] lines in their spectra , we estimate upper limits to their fe@xmath179 abundances using the intensities reported for @xmath1814658 or @xmath1814661 .
these lines are close in wavelength to [ ] @xmath1814658 , the brightest [ ] line for the physical conditions of our pne .
we use @xmath1814658 for ic 4406 , ic 1747 , jner 1 , m 1 - 42 , m 3 - 29 , and ngc 7026 , and @xmath1814661 for h 1 - 41 and m 3 - 32 . some pne show lines from higher iron ionization states ; this will be discussed in [ withouticf ] .
table [ tab:2 ] presents all the computed ionic abundances and their uncertainties .
the last column shows the number of [ ] lines used to derive the fe@xmath179 abundance in each object .
the pne where only upper limits are available are distinguished either with `` 1 ? '' , when the @xmath1814658 line was used ( note that this line could be a misidentification of [ ] @xmath1814658 ) , or with `` '' when @xmath1814661 was used .
lllllllc cn 1 - 5 & @xmath194 & & @xmath195 & @xmath196 & 5.23 & @xmath197 & 8 + cn 3 - 1 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & & @xmath202 & 3 + dddm 1 & @xmath203 & & @xmath204 & @xmath205 & & @xmath206 & 4 + h 1 - 41 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & & @xmath211 & + h 1 - 42 & @xmath212 & @xmath213 & @xmath214 & @xmath215 & 4.50 & @xmath216 & 3 + h 1 - 50 & @xmath217 & @xmath218 & @xmath219 & @xmath220 & & @xmath221 & 1 + hu 1 - 1 & @xmath222 & @xmath223 & @xmath224 & @xmath225 & & @xmath226 & 1 + hu 2 - 1 & @xmath227 & @xmath228 & @xmath229 & @xmath230 & & @xmath231 & 4 + ic 418 & @xmath232 & & @xmath233 & @xmath234 & 3.89 & @xmath235 & 5 + ic 1747 & @xmath236 & @xmath218 & @xmath237 & @xmath238 & & @xmath239 & 1 ?
+ ic 2165 & @xmath240 & @xmath241 & @xmath242 & @xmath243 & 3.64 & @xmath244 & 3 + ic 3568 & @xmath207 & @xmath245 & @xmath246 & @xmath247 & & @xmath248 & 2 + ic 4191 & @xmath249 & @xmath250 & @xmath251 & @xmath252 & & @xmath253 & 1 + ic 4406 & @xmath232 & @xmath250 & @xmath254 & @xmath255 & & @xmath256 & 1 ? + ic 4593 & @xmath257 & @xmath258 & @xmath259 & @xmath260 & & @xmath261 & 4 + ic 4699 & @xmath262 & @xmath263 & @xmath264 & @xmath225 & & @xmath265 & 1 + ic 4846 & @xmath207 & @xmath266 & @xmath267 & @xmath268 & & @xmath269 & 2 + ic 5217 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & & @xmath274 & 1 + jner 1 & @xmath275 & @xmath276 & @xmath277 & @xmath278 & & @xmath279 & 1 ?
+ m 1 - 20 & @xmath217 & @xmath280 & @xmath281 & @xmath282 & & @xmath283 & 1 + m 1 - 42 & @xmath284 & @xmath218 & @xmath285 & @xmath286 & & @xmath287 & 1 ?
+ m 1 - 73 & @xmath288 & @xmath289 & @xmath290 & @xmath291 & & @xmath292 & 2 + m 2 - 4 & @xmath194 & & @xmath293 & @xmath294 & & @xmath295 & 9 + m 2 - 6 & @xmath212 & @xmath296 & @xmath297 & @xmath298 & & @xmath299 & 8 + m 2 - 27 & @xmath300 & @xmath301 & @xmath302 & @xmath303 & & @xmath304 & 2 + m 2 - 31 & @xmath305 & & @xmath306 & @xmath220 & & @xmath307 & 1 + m 2 - 33 & @xmath236 & @xmath308 & @xmath309 & @xmath310 & & @xmath311 & 5 + m 2 - 36 & @xmath236 & @xmath312 & @xmath313 & @xmath314 & & @xmath315 & 1 + m 2 - 42 & @xmath212 & @xmath316 & @xmath317 & @xmath318 & & @xmath319 & 2 + m 3 - 7 & @xmath212 & @xmath320 & @xmath321 & @xmath220 & 4.56 & @xmath322 & 2 + m 3 - 29 & @xmath323 & & @xmath324 & @xmath225 & & @xmath325 & 1 ?
+ m 3 - 32 & @xmath326 & @xmath327 & @xmath328 & @xmath238 & & @xmath329 & + mycn 18 & @xmath330 & @xmath331 & @xmath332 & @xmath210 & & @xmath333 & 6 + ngc 40 & @xmath334 & @xmath335 & @xmath336 & @xmath337 & 4.86 & @xmath338 & 5 + ngc 2392 & @xmath339 & @xmath340 & @xmath341 & @xmath342 & & @xmath343 & 5 + ngc 3132 & @xmath249 & @xmath344 & @xmath345 & @xmath346 & & @xmath347 & 6 + ngc 3242 & @xmath348 & @xmath349 & @xmath350 & @xmath351 & & @xmath352 & 2 + ngc 3587 & @xmath353&@xmath354&@xmath355 & @xmath356 & & @xmath357 & 1 + ngc 3918 & @xmath358 & @xmath359 & @xmath360 & @xmath361 & & @xmath362 & 3 + ngc 5882 & @xmath288 & @xmath363 & @xmath364 & @xmath365 & & @xmath366 & 6 + ngc 6153 & @xmath212 & @xmath367 & @xmath368 & @xmath369 & & @xmath370 & 2 + ngc 6210 & @xmath288 & @xmath371 & @xmath372 & @xmath373 & & @xmath374 & 1 + ngc 6439 & @xmath326 & @xmath375 & @xmath376 & @xmath255 & & @xmath377 & 2 + ngc 6543 & @xmath212 & & @xmath378 & @xmath379 & & @xmath380 & 6 + ngc 6565 & @xmath257 & @xmath381 & @xmath382 & @xmath238 & 4.90 & @xmath383 & 7 + ngc 6572 & @xmath236 & @xmath384 & @xmath385 & @xmath386 & & @xmath387 & 7 + ngc 6620 & @xmath388 & @xmath349 & @xmath389 & @xmath390 & 5.31 & @xmath391 & 1 + ngc 6720 & @xmath207 & @xmath392 & @xmath393 & @xmath394 & 4.26 &
@xmath395 & 4 + ngc 6741 & @xmath396 & @xmath397 & @xmath398 & @xmath345 & 5.58 & @xmath399 & 7 + ngc 6803 & @xmath249 & @xmath400 & @xmath401 & @xmath402 & & @xmath403 & 3 + ngc 6818 & @xmath241 & @xmath404 & @xmath405 & @xmath406 & & @xmath407 & 1 + ngc 6826 & @xmath257 & @xmath408 & @xmath409 & @xmath410 & & @xmath411 & 3 + ngc 6884 & @xmath412 & @xmath381 & @xmath413 & @xmath414 & 4.04 & @xmath415 & 6 + ngc 7026 & @xmath249 & @xmath416 & @xmath417 & @xmath418 & & @xmath419 & 1 ? + ngc 7662 & @xmath420 & @xmath421 & @xmath422 & @xmath423 & & @xmath424 & 4 + vy 2 - 1 & @xmath425 & @xmath426 & @xmath427 & @xmath428 & & @xmath429 & 2
the absence of lines in seven pne of the sample indicates that their he@xmath179 abundances are negligible , and since the ionization potentials of o@xmath179 and he@xmath430 are very similar ( 54.93 ev and 54.42 ev , respectively ) , we do not expect an important amount of o@xmath431 in these nebulae .
therefore , in these low - ionization pne , we simply add the ionic abundances of o@xmath186 and o@xmath179 to calculate the oxygen abundance .
for the other pne we use the expression @xmath432 @xcite .
the icfs of oxygen in most of our sample pne have values below 1.2 .
the exceptions are the pne with the highest @xmath15 ratios : ic 2165 ( icf = @xmath433 ) , ngc 3242 ( 1.28 ) , ngc 3918 ( 1.51 ) , ngc 6741 ( 1.40 ) , ngc 6818 ( 1.97 ) , and ngc 7662 ( 1.57 ) .
the differences between the values of o / h derived from equation ( [ eq : icf_o ] ) and those derived with the icf suggested in @xcite are lower than 0.02 dex for most of the nebulae , and reach a maximum value of 0.13 dex for ic 2165 and ngc 6818 .
the differences are small because both icfs lead to similar corrections for low ionization pne , with @xmath434 . in general
, our values of o / h agree within 0.1 dex with those derived in the papers listed in table [ tab:1 ] .
ten pne show differences of up to 0.4 dex , and are those with the highest differences in the adopted physical conditions and in the correction applied for the unobserved ions .
again , this illustrates the uncertainties introduced by the choices one has to make in order to calculate chemical abundances .
the [ ] lines are usually the brightest iron lines in photoionized nebulae . therefore , total iron abundances are generally calculated using @xmath435 abundances and icfs derived from photoionization models .
however the iron abundances computed in this way differ from the ones obtained by adding up the ionic abundances of fe@xmath186 , fe@xmath179 , and fe@xmath436 in some regions and low - excitation pne .
@xcite studied the uncertainties involved in the calculations and proposed that the iron abundances can be constrained using two icfs , one derived from photoionization models : @xmath437 and the other computed from observational data : @xmath438 for low ionization objects , with @xmath439 , equation ( [ eq:2 ] ) should be replaced by @xmath440 since @xmath435 and @xmath441 will dominate the total abundances of fe and o. this scheme provides three values for the iron abundance , each of them corresponding to one of the three possible explanations suggested by @xcite for the iron abundance discrepancy .
the fe / o values obtained from equation ( 2 ) should be used if the collision strengths for fe@xmath436 are wrong . if the collision strengths for fe@xmath179 are incorrect , the best values are those obtained from equation ( 2 ) lowered by 0.3 dex . and third , if the recombination coefficient or the rate of charge - exchange reaction for fe@xmath436 are wrong , the icf from equation ( 3 ) should be used .
if there is a combination of causes for the discrepancy , the real iron abundance for each object will be intermediate between the two extremes given by the three values . since equations ( 2 ) and ( 3 ) lead to different abundance distributions , and since we want to explore possible correlations with other parameters , we keep the individual abundances obtained from each ionization correction scheme , instead of just providing the average value .
table [ tab:3 ] shows the final abundances of oxygen and iron .
the results shown for iron are the values obtained from equations ( [ eq:1 ] ) and ( [ eq:2 ] ) .
the third set of values , needed to constrain the iron abundances in the manner described above , can be obtained by subtracting @xmath442 dex to the values in column ( 3 ) .
taking into account these three values , we constrain the iron abundances to the abundance range shown in column ( 5 ) .
lllll cn 1 - 5 & @xmath443 & @xmath444 & @xmath445 & 6.46.7 + cn 3 - 1 & @xmath446 & @xmath447 & @xmath448 & 5.45.7 + dddm 1 & @xmath449 & @xmath450 & @xmath451 & 6.06.3 + h 1 - 41 & @xmath452 & @xmath453 & @xmath454 & @xmath455 + h 1 - 42 & @xmath456 & @xmath457 & @xmath458 & 5.36.0 + h 1 - 50 & @xmath390 & @xmath459 & @xmath460 & 5.15.6 + hu 1 - 1 & @xmath461 & @xmath462 & @xmath463 & 4.54.8 + hu 2 - 1 & @xmath464 & @xmath465 & @xmath466 & 5.35.6 + ic 418 & @xmath467 & @xmath468 & @xmath469 & 4.14.6 + ic 1747 & @xmath220 & @xmath470 & @xmath471 & @xmath472 + ic 2165 & @xmath473 & @xmath474 & @xmath475 & 5.56.0 + ic 3568 & @xmath476 & @xmath477 & @xmath478 & 5.16.3 + ic 4191 & @xmath428 & @xmath479 & @xmath480 & 5.05.4 + ic 4406 & @xmath481 & @xmath482 & @xmath483 & @xmath484 + ic 4593 & @xmath485 & @xmath486 & @xmath487 & 5.96.4 + ic 4699 & @xmath488 & @xmath489 & @xmath490 & 5.06.0 + ic 4846 & @xmath491 & @xmath492 & @xmath493 & 5.25.8 + ic 5217 & @xmath494 & @xmath495 & @xmath496 & 5.66.5 + jner 1 & @xmath497 & @xmath498 & @xmath498 & @xmath499 + m 1 - 20 & @xmath500 & @xmath501 & @xmath502 & 4.95.4 + m 1 - 42 & @xmath503 & @xmath504 & @xmath505 & @xmath506 + m 1 - 73 & @xmath507 & @xmath508 & @xmath509 & 5.65.9 + m 2 - 4 & @xmath510 & @xmath511 & @xmath512 & 5.96.3 + m 2 - 6 & @xmath513 & @xmath514 & @xmath515 & 5.76.0 + m 2 - 27 & @xmath516 & @xmath517 & @xmath518 & 6.06.4 + m 2 - 31 & @xmath252 & @xmath519 & @xmath520 & 6.26.7 + m 2 - 33 & @xmath521 & @xmath522 & @xmath523 & 5.96.5 + m 2 - 36 & @xmath524 & @xmath525 & @xmath526 & 4.95.3 + m 2 - 42 & @xmath428 & @xmath527 & @xmath528 & 5.86.4 + m 3 - 7 & @xmath529 & @xmath530 & @xmath531 & 6.26.5 + m 3 - 29 & @xmath532 & @xmath533 & @xmath534 & @xmath455 + m 3 - 32 & @xmath535 & @xmath536 & @xmath537 & @xmath538 + mycn 18 & @xmath539 & @xmath540 & @xmath541 & 5.96.2 + ngc 40 & @xmath542 & @xmath543 & @xmath544 & 5.35.6 + ngc 2392 & @xmath545 & @xmath546 & @xmath547 & 6.16.4 + ngc 3132 & @xmath548 & @xmath549 & @xmath550 & 5.25.5 + ngc 3242 & @xmath551 & @xmath552 & @xmath553 & 5.05.9 + ngc 3587 & @xmath554 & @xmath555 & @xmath556 & 6.26.5 + ngc 3918 & @xmath196 & @xmath557 & @xmath558 & 4.95.2 + ngc 5882 & @xmath294 & @xmath559 & @xmath560 & 5.56.3 + ngc 6153 & @xmath365 & @xmath561 & @xmath562 & 5.25.9 + ngc 6210 & @xmath563 & @xmath564 & @xmath565 & 5.25.8 + ngc 6439 & @xmath318 & @xmath566 & @xmath567 & 5.55.8 + ngc 6543 & @xmath568 & @xmath569 & @xmath570 & 5.66.3 + ngc 6565 & @xmath571 & @xmath572 & @xmath573 & 5.76.0 + ngc 6572 & @xmath574 & @xmath575 & @xmath576 & 5.15.6 + ngc 6620 & @xmath577 & @xmath578 & @xmath579 & 5.45.6 + ngc 6720 & @xmath580 & @xmath581 & @xmath582 & 4.95.2 + ngc 6741 & @xmath583 & @xmath584 & @xmath585 & 5.96.2 + ngc 6803 & @xmath586 & @xmath587 & @xmath588 & 5.56.0 + ngc 6818 & @xmath196 & @xmath589 & @xmath333 & 5.55.9 + ngc 6826 & @xmath291 & @xmath590 & @xmath591 & 5.46.0 + ngc 6884 & @xmath592 & @xmath593 & @xmath594 & 5.46.0 + ngc 7026 & @xmath314 & @xmath595 & @xmath596 & @xmath597 + ngc 7662 & @xmath598 & @xmath599 & @xmath600 & 5.56.4 + vy 2 - 1 & @xmath601 & @xmath602 & @xmath603 & 5.35.6 + the ionization correction scheme for iron described above was derived for low - excitation objects . for high - excitation pne , like ic 2165 ( with a high value of he@xmath179/he@xmath180 ) ,
the estimated iron abundances are less reliable .
besides , for those objects with low values of o@xmath180/o@xmath179 , like ic 3568 , our method does not constrain very well the iron abundance , since it remains uncertain by more than an order of magnitude . however , this is the only method that can be applied at the moment to study in a homogeneous way the iron abundances in a large sample of pne .
the main reason is that it is difficult to obtain reliable line intensities for higher ionization states of iron , because they are weak and prone to be blended .
some of the nebulae do have detections of lines from iron ionization states higher than fe@xmath179 , although there are problems with blends and misidentifications .
nevertheless , four of the sample pne , seem to have reliable measurements of lines from most of their iron ionization states . for these nebulae ,
we have derived the iron abundances implied by the sum of the ionic abundances and compared them to those we obtain using the icfs described in [ sec : tot_ab ] .
the abundance of fe@xmath436 is calculated by solving the equations of statistical equilibrium for 33 energy levels using the transition probabilities of @xcite and the collision strengths of @xcite .
for fe@xmath604 we use a 34-level atom with transition probabilities from @xcite and collision strengths from @xcite . for fe@xmath605
we use a 19-level atom with transition probabilities and collision strengths from @xcite .
finally , for fe@xmath606 we use a 9-level atom with transition probabilities and collision strengths from @xcite .
energy levels are those listed in the nist atomic spectra database ; the h@xmath189 emissivities are based on the empirical formula by @xcite .
we use the values of @xmath171 and @xmath172 [ ] listed in table [ tab:1 ] for all these calculations .
in what follows , we use air wavelengths from either the nist database or from the atomic line listpeter / atomic/ ] of peter van hoof .
ngc 6210 , which is a low - excitation pne with @xmath607 , has measurements of [ ] @xmath608 , [ ] @xmath609 , and [ ] @xmath610 .
we think that the last line is a misidentification , since ( 1 ) fe@xmath604 is ionized at 75 ev and this is a low - excitation object , and ( 2 ) @xcite identify the same feature in ngc 6741 , where it leads to an abundance 13 times larger than the one we get from [ ] @xmath611 .
besides [ ] @xmath611 , ngc 6741 ( with @xmath612 ) has measurements of [ ] @xmath609 , and several [ ] lines at 4967.1 , 4972.5 , 5335.2 , 5424.2 , 5426.6 , 5484.8 , 5631.1 , and 5677.0 .
ngc 6741 also has upper limits for the intensities of the [ ] lines at 4906.6 and 5233.8 , which imply @xmath613 , in agreement with the value we estimate by averaging the abundances of fe@xmath179 and fe@xmath614 : @xmath615 .
we suspect that the two [ ] lines identified at 3839.3 and 3895.2 can be affected by blends or misidentifications because they lead to high abundances in this nebula and in ngc 6884 and ic 2165 .
on the other hand , note that fe@xmath616 has an ionization potential of 125 ev and we do not expect it to be significantly ionized in any of our pne .
ngc 6884 ( with @xmath617 ) has measurements of [ ] @xmath618 , [ ] lines at 5335.2 , 5484.8 , and 5631.1 , and [ ] @xmath619 .
there is also an upper limit for the intensity of [ ] @xmath609 , which implies @xmath620 , in agreement with the average of the fe@xmath431 and fe@xmath621 abundances that we adopt : @xmath622 . finally , ic 2165 , our highest excitation object ( @xmath623 ) , has measurements of [ ] @xmath609 , [ ] lines at 4972.5 , 5176.0 , 5335.2 , 5424.2 , 5426.6 , 5484.8 , 5631.1 , and 5677.0 , and [ ] @xmath624 .
the lines identified as [ ] @xmath625 , [ ] @xmath626 , and [ ] @xmath627 are likely blends or misidentifications .
the fe@xmath431 abundance is estimated from the average of the fe@xmath179 and fe@xmath614 abundances .
the results obtained for ngc 6741 and ngc 6884 suggest that in this case this is a reasonable approximation .
we used all the lines listed above which are not affected or suspected of blends and misidentifications to derive ionic abundances in these four pne .
the resulting ionic and total abundances are presented in table [ tab:3b ] .
if we compare the iron abundances in this table with those shown in table [ tab:3 ] , we find that in ngc 6741 all the derived values of fe / h are very similar ( @xmath628 versus @xmath629 in table [ tab:3 ] ) , in ngc 6884 the value of table [ tab:3b ] is intermediate between those of table [ tab:3 ] ( @xmath630 versus @xmath631 ) , in ngc 6210 the iron abundance of table [ tab:3b ] is similar to the upper limit to the iron abundance inferred from table [ tab:3 ] ( @xmath632 versus @xmath633 ) , and in ic 2165 it is close to the lower limit ( @xmath634 versus @xmath635 ) .
this suggests that the process we follow to constrain the iron abundances within the ranges shown in the last column of table [ tab:3 ] is working well .
the fact that we do not find one icf working significantly better than the other might arise from the inherent limitations of all ionization correction schemes , or of this scheme in particular . on the other hand , further measurements of the main iron ionization states in other pne might help in improving this ionization correction scheme or in defining a better one .
in what follows we will check that all our results hold when using any of the iron icfs that we have applied .
llllll & 4.62 & 4.66 & 4.58 & 4.35 & 5.28 + & 5.75 & 4.04 & & & 5.80 + & 5.58 & 5.45 & 5.10 & 4.67 & 6.15 + & 5.21 & 5.04 & 4.77 & 4.67 & 5.65 figure [ fig:1 ] shows the values of fe / o for all the pne as a function of their degree of ionization , given by @xmath636(o@xmath186/o@xmath179 ) .
the upper panel displays the values derived from equation ( [ eq:1 ] ) and the bottom panel shows the values computed with equations ( [ eq:2 ] ) and ( [ eq:3 ] ) . note that the main difference between the results implied by the two icfs lies in the iron abundances of those objects with the lowest values of o@xmath186/o@xmath179 .
the right axis in figure [ fig:1 ] shows estimates of the depletion factors for fe / o .
the depletion of one particular element ( x ) is usually calculated as the difference between the observed and the expected values of : @xmath637=\log(\mbox{x / h})-\log(\mbox{x / h})_{\rm ref}$ ] .
however , we will use fe / o instead of fe / h to calculate depletion factors , since the intrinsic value of fe / o is expected to show less variations from object to object than either fe / h or o / h @xcite . in principle
, we could use a different reference element , but oxygen abundances require small ionization correction factors in most objects and hence are the ones that can be determined in a more reliable way .
we adopted the solar value of @xmath638 @xcite as the expected abundance ratio for our objects .
for the range of metallicities covered by our sample pne , the stellar values of fe / o are found to decrease from solar to 0.40.5 dex below solar in the galaxy @xcite .
this means that the halo pn dddm 1 , that has the lowest metallicity and the highest fe / o abundance ratio in our sample , could have a depletion factor close or equal to zero ( but its infrared spectrum shows the presence of silicates , see
[ sec : dust ] ) .
we should also bear in mind that the oxygen atoms may be depleted into dust grains , and in such case our values of @xmath639 $ ] should be lowered by up to @xmath30.15 dex if oxygen is trapped in oxides and silicates @xcite . on the other hand , in @xcite we found that the oxygen abundances in a group of pne from the solar neighborhood were systematically higher than the ones in nearby regions ( calculated either from collisionally excited lines or recombination lines ) .
we suggested that the difference could be due to oxygen depletion in organic refractory dust components , but another possible explanation for these overabundances is oxygen production in the pn progenitor stars , and if this production is important , it will change significantly the value of fe / o .
however , the amount of oxygen production by agb stars is very uncertain and could be negligible ( see the predictions of models by different authors in * ? ? ?
thus , we present our results for fe / o , but in what follows we will check that our results hold both for the fe / o and fe / h abundance ratios . as we mentioned above , our ionization correction scheme provides three values for the iron abundance , and the real value of fe / o for each object is expected to lie in the range defined by these three values .
we can see from figure [ fig:1 ] that iron abundances are better constrained for objects with a relatively low degree of ionization , @xmath640 , where the three values obtained with the different icfs differ by less than 0.5 dex .
the values of @xmath641 in our sample range from @xmath642 to @xmath643 , which are the most extreme upper and lower limits that we find . in the same way ,
the depletion factors [ fe / o ] range from @xmath644 for ic 418 to @xmath645 for dddm 1 .
even taking into account all the considerations mentioned above that could change the iron depletions factors shown in the right axes of figure [ fig:1 ] , we can conclude that a significant fraction of our sample pne have more than @xmath646 of their iron atoms deposited into dust grains . in agreement with our previous findings in a smaller sample of pne @xcite ,
the range of depletions is high , with differences reaching a factor of @xmath647 .
these differences can be related to the pne ages or grain compositions , maybe reflecting different efficiencies of the grain formation and destruction processes , an issue that we will explore further in the following sections .
as we mentioned in [ sec : intro ] , one can expect a correlation between the c / o abundance ratios and the type of dust grains present in pne .
unfortunately , obtaining accurate values of c / o for pne is not easy ( see , e.g. , * ? ? ?
* ; * ? ? ?
the ionic abundances of carbon and oxygen can be derived from collisionally excited lines ( cels ) or from recombination lines ( rls ) , and the abundances obtained with rls are systematically higher than the ones derived from cels , by factors that are around two for many pne but can reach @xmath370 ( see * ? ? ?
the reason for this discrepancy is still a matter of debate and we do not know which lines lead to more representative abundances in pne .
the abundances derived from cels are highly dependent on the electron temperature , but these lines are brighter than rls and thus , more easily measured .
on the other hand , rls are weakly dependent on physical conditions , but they are faint and may suffer from other problems ( see , e.g , * ? ? ?
* ; * ? ? ?
. one important source of uncertainties in the c / o values derived from cels is the normalization between ultraviolet and optical fluxes since carbon lines are found in the ultraviolet range whereas oxygen lines are better measured in the optical range .
this correction introduces uncertainties in the values of c / o , which are more severe for extended objects , where ultraviolet and optical observations may cover different regions of the nebula .
one advantage of using rls is that this issue does not arise , since they are observed in the optical range .
a further complication is introduced by the icfs needed to account for unobserved ions and estimate the total abundances . for oxygen
we have used here the equation proposed by @xcite to correct for the contribution of o@xmath436 to the total abundance . in the case of oxygen abundances based on rls , only the o@xmath179 abundances can be easily calculated , and we assume that the distribution of all ionization states is equal to the one inferred from cels . as for carbon , one can apply the widely used correction scheme of @xcite , consisting of several icfs which are employed depending on the degree of ionization of the object and the detected ionization stages ( up to four different ions can be observed in pne : c@xmath186 , c@xmath179 , c@xmath436 , and c@xmath604 ) , although certain cases are not covered by this scheme .
therefore , the correction scheme of @xcite produces an inhomogeneous determination of carbon abundance when applied to different objects .
the other method that is frequently adopted to calculate c / o is @xmath648 .
however , this icf generally overestimates the value of c / o , especially in low ionization pne @xcite . here
we calculate c / o in a homogeneous way using the icf derived in @xcite : @xmath649 where @xmath650 .
this icf is valid in the range @xmath651 . in this range , the icf
is expected to reproduce the values of c / o to within @xmath652 dex
. however , the c / o values derived for ngc 40 ( with @xmath653 ) , and for ic 3568 , ic 5217 , ngc 3242 , ngc 5882 , m3 - 32 , ngc 7662 ( all of them with @xmath654 ) are more uncertain . for ngc 40 ,
we estimate a confidence interval of @xmath655 dex ; for the objects with the highest values of @xmath656 , we estimate confidence intervals of @xmath657 dex .
these uncertainties will be taken into account in the forthcoming analysis .
the value of the icf in equation ( 5 ) is close to 1 for most of our objects , where @xmath658 , changing the c@xmath179/o@xmath179 ratio by less than 0.05 dex .
however , in the low ionization pne ic 418 and ngc 40 , the differences between both estimates reach 0.9 dex .
we compared the values of c / o implied by the three different methods described above ( @xmath648 , the set of icfs from @xcite , and the icf from @xcite ) and by the two types of lines we use ( cels or rls ) .
we find that the differences seem to be more related to whether we use cels or rls than to the correction scheme we apply .
hence we will restrict the forthcoming discussion to the values of c / o calculated with equation ( 5 ) , derived either from rls or from cels .
the c@xmath179 abundances were calculated from the fluxes of the ] @xmath659 doublet ( provided in the same papers we use for the optical fluxes , see table [ tab:1 ] ) through the _ iraf _ routine _ ionic _ using the values of @xmath171 and @xmath172 [ ] listed in table [ tab:1 ] .
since the collision strengths for c@xmath179 , derived by @xcite , are available only for @xmath172 @xmath660 k , we extrapolated them to lower temperatures in order to cover the @xmath172 range found for our sample pne .
table [ tab:4 ] shows the values we obtain for c@xmath179/h@xmath186 .
the highest differences between our values of @xmath661 and the ones in the papers listed in table [ tab:1 ] are found for hu 1 - 1 ( 0.35 dex ) and ic 4406 ( 0.23 dex ) .
for the other pne the differences are lower than 0.15 dex .
these disagreements are due to the different physical conditions adopted in the calculations , except for ic 4406 , for which the value given in @xcite is a typo , and their corrected value agrees with the one we derive ( y. tsamis , private communication ) . using the values of o@xmath179 from table [ tab:2 ]
, we calculate the c / o values listed in the third column of table [ tab:4 ] .
because these values are based on cels located in the optical ( [ ] @xmath188 ) and in the ultraviolet ( ] @xmath659 ) , there are several important sources of uncertainty associated to them , namely , the adopted @xmath172 , the extinction correction , the normalization between ultraviolet and optical fluxes , and the aperture correction in extended nebulae .
the uncertainties associated with the normalization of optical and ultraviolet fluxes are difficult to quantify .
the other sources of uncertainty are discussed below .
the correction for interstellar extinction is important , and the extinction law for the bulge pne , needed for some of our sample objects , is uncertain ( see , e.g. , * ? ? ?
* ; * ? ? ?
* and references therein ) .
even a small error in the extinction coefficient , @xmath662 , introduces an uncertainty of @xmath663 dex in the intensity ratio of the ] @xmath1811908 and [ ] @xmath1814959 lines .
however , the main effect of an error in the extinction coefficient will occur through its impact on the derived temperature .
the determination of an accurate value of @xmath172 is a critical factor in obtaining reliable ionic abundances from cels , and the emissivities of ultraviolet cels are more dependent on @xmath172 than those of optical cels . as an example
, consider an error of 500 k , which can arise easily from errors in the extinction correction , the flux calibration or the line measurements that will not necessarily appear in the quoted line intensity errors . at @xmath172 @xmath664
k , this variation introduces a change of @xmath665 dex in c@xmath179/h@xmath186 and a change of @xmath666 dex in o@xmath179/h@xmath186 ; the final effect in the c@xmath179/o@xmath179 ratio will be around 0.1 dex .
aperture corrections are needed in those nebulae that are more extended than the aperture of the international ultraviolet explorer ( @xmath667 arsec@xmath668 ) , or that were observed in the optical with a long slit at a single position ( as opposed to a long slit scanning the whole object ) .
the scale factors applied by the different authors in our sample pne go up to 13 , with ngc 6720 ( angular diameter of 76 arcsec ) the pn with the highest aperture correction @xcite .
we use here the ultraviolet fluxes provided by the different authors , already scaled and normalized by them .
the uncertainties related to this procedure are difficult to estimate , but might reach a factor of 23 for objects like ngc 6720 .
table [ tab:4 ] shows the c@xmath179 abundances and the values of c / o based on cels and the icf of equation ( 5 ) .
we only present in this table the uncertainties associated with the final values of c / o .
we consider an uncertainty of 0.2 dex in @xmath669 , based on the discussion above ( it could be higher for some objects , such as extended nebulae with important aperture corrections ) , which is added quadratically to the uncertainty in the icf computed from equations ( 40 ) and ( 41 ) in @xcite .
lcl|ccl cn 1 - 5 & 9.04 & @xmath670 & 9.08 & 8.90 & @xmath671 + cn 3 - 1 & @xmath672 & @xmath672 & 8.05 & @xmath672 & @xmath672 + dddm 1 & 6.78 & @xmath673 & @xmath672 & 8.43 & @xmath672 + h 1 - 41 & @xmath672 & @xmath672 & 8.57 & 9.18 & @xmath674 + h 1 - 42 & 7.61 & @xmath675 & 8.34 & 8.92 & @xmath676 + h 1 - 50 & 7.82 & @xmath677 & 8.63 & 9.05 & @xmath678 + hu 1 - 1 & @xmath672 & @xmath672 & 8.94 & 8.54 & @xmath679 + hu 2 - 1 & 8.53 & @xmath680 & 8.62 & 8.71 & @xmath681 + ic 418 & 8.35 & @xmath682 & 8.73 & 8.21 & @xmath683 + ic 1747 & 9.04 & @xmath684 & 9.09 & 8.76 & @xmath685 + ic 2165 & 8.28 & @xmath686 & 8.53 & 8.63 & @xmath687 + ic 3568 & 8.05 & @xmath688 & 8.49 & 8.75 & @xmath689 + ic 4191 & @xmath672 & @xmath672 & 8.72 & 9.07 & @xmath690 + ic 4406 & 8.57 & @xmath691 & 8.89 & 8.89 & @xmath692 + ic 4593 & @xmath672 & @xmath672 & 8.63 & 8.65 & @xmath693 + ic
4699 & 8.00 & @xmath694 & 8.72 & 9.15 & @xmath695 + ic 4846 & 7.86 & @xmath696 & 8.16 & 8.76 & @xmath676 + ic 5217 & 8.26 & @xmath697 & 8.36 & 8.65 & @xmath698 + jner 1 & @xmath672 & @xmath672 & @xmath672 & @xmath672 & @xmath672 + m 1 - 20 & @xmath672 & @xmath672 & 8.66 & 8.63 & @xmath699 + m 1 - 42 & 7.92 & @xmath700 & 9.38 & 9.63 & @xmath701 + m 1 - 73 & @xmath672 & @xmath672 & 8.73 & 9.00 & @xmath702 + m 2 - 4 & @xmath672 & @xmath672 & 8.50 & 8.89 & @xmath703 + m 2 - 6 & @xmath672 & @xmath672 & 7.90 & 8.62 & @xmath704 + m 2 - 27 & @xmath672 & @xmath672 & 8.84 & 9.37 & @xmath705 + m 2 - 31 & @xmath672 & @xmath672 & 8.79 & @xmath672 & @xmath672 + m 2 - 33 & 8.59 & @xmath706 & 8.30 & 9.04 & @xmath707 + m 2 - 36 & 8.65 & @xmath708 & 9.36 & 9.51 & @xmath709 + m 2 - 42 & @xmath672 & @xmath672 & @xmath672 & 9.60 & @xmath672 + m 3 - 7 & @xmath672 & @xmath672 & @xmath672 & 9.42 & @xmath672 + m 3 - 29 & @xmath672 & @xmath672 & 8.51 & 9.25 & @xmath710 + m 3 - 32 & 8.42 & @xmath711 & 9.55 & 9.74 & @xmath712 + mycn 18 & @xmath672 & @xmath672 & 8.36 & 8.81 & @xmath713 + ngc 40 & 8.01 & @xmath714 & 8.81 & 8.61 & @xmath715 + ngc 2392 & 7.69 & @xmath716 & @xmath672 & 9.81 & @xmath672 + ngc 3132 & 8.18 & @xmath717 & 8.82 & 8.80 & @xmath718 + ngc 3242 & 8.04 & @xmath719 & 8.79 & 8.78 & @xmath720 + ngc 3587 & @xmath672 & @xmath672 & 8.36 & 9.38 & @xmath721 + ngc 3918 & 8.36 & @xmath722 & 8.70 & 8.75 & @xmath723 + ngc 5882 & 8.11 & @xmath724 & 8.58 & 8.93 & @xmath725 + ngc 6153 & 8.38 & @xmath726 & 9.35 & 9.51 & @xmath727 + ngc 6210 & 7.87 & @xmath728 & 8.80 & 9.53 & @xmath729 + ngc 6439 & 8.32 & @xmath730 & 8.99 & 9.21 & @xmath731 + ngc 6543 & 8.48 & @xmath732 & 8.76 & 9.07 & @xmath733 + ngc 6565 & 8.29 & @xmath734 & 8.65 & 8.85 & @xmath735 + ngc 6572 & 8.77 & @xmath736 & 8.69 & 8.71 & @xmath737 + ngc 6620 & 8.16 & @xmath738 & 8.94 & 9.24 & @xmath739 + ngc 6720 & 8.38 & @xmath740 & 8.94 & 8.90 & @xmath741 + ngc 6741 & 8.40 & @xmath742 & 8.79 & 8.74 & @xmath743 + ngc 6803 & 8.24 & @xmath744 & 8.79 & 9.05 & @xmath745 + ngc 6818 & 8.17 & @xmath746 & 8.66 & 8.57 & @xmath747 + ngc 6826 & 8.40 & @xmath748 & 8.73 & 8.92 & @xmath749 + ngc 6884 & 8.45 & @xmath706 & 8.87 & 8.98 & @xmath750 + ngc 7026 & 8.33 & @xmath751 & 8.93 & 9.07 & @xmath752 + ngc 7662 & 7.99 & @xmath753 & 8.67 & 8.58 & @xmath754 + vy 2 - 1 & 8.61 & @xmath755 & 8.62 & 8.98 & @xmath756 and rls can be measured in the optical range and do not suffer from many of the uncertainties discussed above .
therefore , these lines will provide better estimates of the c / o abundance ratio , or they will do so if they can be considered to sample the abundances of the nebular gas . for example , these emission lines might not be representative of the nebular abundances if they arise mostly from metal - rich inclusions , whose existence in pne has been postulated to explain the discrepant abundances implied by cels and rls ( see , e.g. , * ? ? ?
the c@xmath179/h@xmath186 and o@xmath179/h@xmath186 abundance ratios implied by rls are derived for all objects with available measurements of the required rls using the values of @xmath171 and @xmath172 [ ] in table [ tab:2 ] and the effective recombination coefficients for h@xmath189 of @xcite .
the c@xmath179 abundances are calculated using the @xmath1814267 line and the case b effective recombination coefficients of @xcite .
the o@xmath179 abundances are computed from the total intensity of multiplet 1 of and the recombination coefficients of @xcite .
the multiplet intensity was corrected for the contribution of undetected lines with the formulae derived by @xcite .
typical errors in c@xmath179/h@xmath186 and o@xmath179/h@xmath186 arising from errors in the line intensities are @xmath757 dex and @xmath758 dex , respectively .
final errors in c@xmath179/o@xmath179 due to errors in the line intensities are around @xmath758 dex . if we also consider the uncertainties in the icf , as explained in
[ subsec1:co ] , we obtain the final uncertainties presented in table [ tab:4 ] together with the ionic abundances of c@xmath179 and o@xmath179 obtained from rls
. we found differences between our values and the ones in the reference papers that are in general lower than 0.06 dex but reach 0.17 dex for some nebulae .
the highest differences are found for h 1 - 42 ( with a difference of 0.32 dex in the value of c@xmath179/h@xmath186 ) , m 2 - 42 ( 0.56 dex in o@xmath179/h@xmath186 ) , and m 3 - 29 ( 0.49 dex in o@xmath179/h@xmath186 ) and are not due to differences in the physical conditions used in the calculations .
we do not know the reasons for these discrepancies , but they do not affect our conclusions , since we do not have information about the types of grains present in h 1 - 42 and m 3 - 29 , and we can not calculate the c / o abundance ratio in m 2 - 42 . hence , these pne are not included in the forthcoming analysis .
figure [ fig : c2o2 ] shows a comparison between the values of c / o obtained from cels and rls in those pne where both types of lines are observed .
we also plot in the figure the error bars associated with this abundance ratio for each object . the error bars take into account both the uncertainties in the line intensity ratios ( @xmath759 dex for the values derived from cels and @xmath758 dex for the ones derived from rls ) and the uncertainties in the icf .
although there is general agreement between these two abundance ratios , a result previously found in several studies ( see * ? ?
* and references therein ) , the differences are high for some objects .
the highest differences reach @xmath760 dex and are found for ngc 40 and m 2 - 33 .
ngc 40 is an extended pn ( diameter of around 48 arcsec ) with a wolf - rayet central star , thus , an incorrect aperture correction and/or contamination of the ultraviolet line fluxes with stellar winds could explain the discrepancy .
m 2 - 33 is smaller ( with a diameter of around 5.8 arcsec ) and thus , we expect that the scaling of optical and ultraviolet observations is not that critical .
the discrepancy could also be explained if the @xmath172 is underestimated by @xmath761 k , but we do not expect such a large error in @xmath172 .
we have explored if the differences between the ionic abundance ratios derived from cels and rls are related to parameters such as @xmath172 [ ] , @xmath171 , the extinction coefficient , the nebular size , the type of dust , the abundance discrepancy factor , or the @xmath172 deduced from the recombination continuum balmer discontinuity , but we do not find any obvious correlation . as we mentioned above , the c / o abundance ratios in pne tell us whether the ionized gas is carbon rich or oxygen rich .
we find that both cels and rls imply that 20% of the pne are c rich .
most of the pne in our sample seem to be o - rich objects . according to theoretical models ( see ,
e.g. , * ? ? ?
* ) , these pne evolve either from the lowest mass progenitors ( masses below 1.5 m@xmath762 ) or from the highest mass progenitors ( above 45 m@xmath762 ) , although the details depend on the metallicity and on the model assumptions .
we used the archive of the infrared spectrograph ( irs ; @xcite ) , on board the _ spitzer space telescope _
@xcite , to download the spectra available for our sample pne .
we decided to look for the following dust features : polycyclic aromatic hydrocarbons ( pahs ) , the broad features around 11 and 30 @xmath1 m ( usually associated with sic and mgs , respectively ) , amorphous silicates , and crystalline silicates .
we restricted the analysis to these features because they are relatively easy to identify and are , in principle , reliably associated with either a carbon - rich or an oxygen - rich environment ( we will come back to this point later ) . a detailed analysis of all the dust features present in the pne is beyond the scope of this paper .
we also compiled dust identifications from the literature , mainly from spectra of the _ infrared space observatory _
( _ iso _ ; @xcite ) .
the results are summarized in table [ tab:5 ] and are discussed below .
lllcccccl cn 1 - 5 & @xmath763 $ ] & @xmath764/@xmath765 & @xmath766 & @xmath767 & @xmath767 & @xmath767 & @xmath766 & _ iso _ ( 1 ) , _ spitzer _ ( 2,3 ) , ukirt ( 7 ) + dddm 1 & @xmath768 $ ] & @xmath769/@xmath672 & @xmath767 & @xmath767 & @xmath767 & @xmath766 & @xmath766 & _ spitzer _ ( 3,4 ) + h 1 - 50 & @xmath770 $ ] & @xmath771/@xmath772 & @xmath766 ? & @xmath767 & @xmath767 & @xmath766 & @xmath766 & _ spitzer _ ( 2,3 ) + hu 2 - 1 & @xmath773 $ ] & @xmath771/@xmath772 & @xmath766 & @xmath766 & @xmath672 & @xmath767 & @xmath672 & _ iso _ ( 1 ) , _ spitzer _ ( 3 ) , ukirt ( 6 ) + ic 418 & @xmath774 $ ] & @xmath775/@xmath776 & @xmath766 & @xmath766 & @xmath766 & @xmath767 & @xmath672 & _ iso _ ( 1 , 5 ) , _ spitzer _ ( 3 ) , ukirt ( 6 ) + ic 2165 & @xmath777 $ ] & @xmath778/@xmath779 & @xmath767 & @xmath766 ? & @xmath672 & @xmath767 & @xmath672 & _ spitzer _ ( 3 ) , ukirt ( 6 ) + ic 3568 & @xmath780 $ ] & @xmath781/@xmath782 & @xmath767 & @xmath672 & @xmath672 & @xmath672 & @xmath672 & _ iso _ ( 1 ) + ic 4406 & @xmath783 & @xmath784/@xmath772 & @xmath767 & @xmath672 & @xmath672 & @xmath672 & @xmath672 & _ iso _ ( 1 ) + ic 4846 & @xmath785 $ ] & @xmath786/@xmath787 & @xmath767 & @xmath767 & @xmath767 & @xmath766 ? & @xmath767 ?
& _ spitzer _ ( 3 ) + m 1 - 20 & @xmath788 $ ] & @xmath789/@xmath790 & @xmath766 & @xmath766 & @xmath766 & @xmath767 & @xmath767 & _ spitzer _ ( 2,3,8 ) , ukirt ( 7 ) + m 1 - 42 & @xmath791 & @xmath792/@xmath793 & @xmath766 & @xmath767 & @xmath767 & @xmath767 & @xmath766 & _ iso _ ( 1 ) , _ spitzer _ ( 3 ) + m 2 - 27 & @xmath794 $ ] & @xmath789/@xmath795 & @xmath766 & @xmath767 & @xmath767 & @xmath767 & @xmath766 & _ spitzer _ ( 2,3 ) + m 2 - 31 & @xmath796 $ ] & @xmath789/@xmath672 & @xmath766 & @xmath767 & @xmath767 & @xmath767 & @xmath766 & _ spitzer _ ( 3,9 ) + m 2 - 36 & @xmath797 $ ] & @xmath798/@xmath799 & @xmath767 & @xmath672 & @xmath672 & @xmath672 & @xmath672 & _ iso _ ( 1 ) + m 2 - 42 & @xmath777 $ ] & @xmath789/@xmath672 & @xmath767 & @xmath767 & @xmath767 & @xmath767 & @xmath766 & _ spitzer _ ( 3 ) + mycn 18 & @xmath800 $ ] & @xmath789/@xmath801 & @xmath766 & @xmath767 & @xmath767 & @xmath766 & @xmath766 & _ spitzer _ ( 3 ) + ngc 40 & @xmath802 $ ] & @xmath803/@xmath804 & @xmath766 & @xmath767 & @xmath766 & @xmath767 & @xmath767 & _ iso _ ( 1 , 5 ) , _ spitzer _ ( 3 ) + ngc 2392 & @xmath805 $ ] & @xmath806/@xmath672 & @xmath767 & @xmath767 & @xmath767 & @xmath767 & @xmath767 & _ spitzer _ ( 3 ) + ngc 3132 & @xmath788 $ ] & @xmath801/@xmath799 & @xmath767 & @xmath767 & @xmath767 & @xmath767 & @xmath766 & _ spitzer _ ( 3 ) + ngc 3242 & @xmath807 $ ] & @xmath808/@xmath765 & @xmath767 & @xmath767 & @xmath766 & @xmath767 & @xmath767 & _ spitzer _ ( 3 ) + ngc 3918 & @xmath809 $ ] & @xmath810/@xmath811 & @xmath767 & @xmath767 & @xmath766 & @xmath767 & @xmath672 & _ iso _ ( 1 , 5 , 10 ) , _ spitzer _ ( 3 ) + ngc 6153 & @xmath812 $ ] & @xmath813/@xmath772 & @xmath767 & @xmath672 & @xmath672 & @xmath766 ? & @xmath767 ? & _ iso _ ( 1 , 10 ) + ngc 6210 & @xmath814 $ ] & @xmath815/@xmath816 & @xmath767 & @xmath767 & @xmath767 & @xmath767 ? & @xmath766 & _ iso _ ( 1 ) , _ spitzer _ ( 3 ) + ngc 6439 & @xmath817 $ ] & @xmath818/@xmath819 & @xmath766 ? & @xmath767 & @xmath767 & @xmath767 & @xmath766 & _ spitzer _ ( 3 ) + ngc 6543 & @xmath820 $ ] & @xmath821/@xmath818 & @xmath767 & @xmath672 & @xmath672 & @xmath766 & @xmath766 & _ iso _ ( 1 , 10 ) + ngc 6572 & @xmath822 $ ] & @xmath823/@xmath824 & @xmath767 & @xmath766 ? & @xmath672 & @xmath767 & @xmath767 & _ iso _ ( 1 ) , ukirt ( 6 ) + ngc 6720 & @xmath809 $ ] & @xmath821/@xmath825 & @xmath767 & @xmath767 & @xmath672 & @xmath767 & @xmath672 & _ iso _ ( 1 ) , _ spitzer _ ( 3 ) + ngc 6741 & @xmath826 $ ] & @xmath772/@xmath779 & @xmath766 & @xmath767 & @xmath672 & @xmath767 & @xmath672 & _ iso _ ( 1 ) , _ spitzer _ ( 3 ) , ukirt ( 6 ) + ngc 6818 & @xmath827 $ ] & @xmath828/@xmath829 & @xmath767 & @xmath767 & @xmath767 & @xmath767 & @xmath767 ? & _ spitzer _ ( 3 ) + ngc 6826 & @xmath820 $ ] & @xmath798/@xmath799 & @xmath767 & @xmath767 & @xmath766 & @xmath767 & @xmath672 & _ iso _ ( 1 , 5 ) , _ spitzer _ ( 3 ) + ngc 6884 & @xmath830 $ ] & @xmath803/@xmath831 & @xmath766 & @xmath767 & @xmath672 & @xmath767 & @xmath672 & _ iso _ ( 1 ) , _ spitzer _ ( 3 ) + ngc 7026 & @xmath832 & @xmath833/@xmath834 & @xmath766 & @xmath767 & @xmath767 & @xmath767 & @xmath766 & _ spitzer _ ( 3 ) + ngc 7662 & @xmath835 $ ] & @xmath836/@xmath837 & @xmath767 & @xmath767 & @xmath672 & @xmath767 & @xmath672 & _ iso _ ( 1 ) , _ spitzer _ ( 3 ) we found _ spitzer _ irs spectra for 27 pne of our sample ( we excluded the very noisy spectra of jner 1 and ngc 3587 ) .
the data belong to the following observing programs : i d 45 ( pi : t. roellig ) , i d 93 ( pi : d. cruikshank ) , i d 1406 ( pi : l. armus ) , i d 1427 ( calibration program ) , i d 20049 ( pi : k. kwitter ) , i d 30285 , 40115 ( pi : g. fazio ) , i d 30430 , 40536 ( pi : h. dinnerstein ) , i d 30550 ( pi : j. r. houck ) , i d 3633 ( pi : m. bobrowsky ) , and i d 50261 ( pi : l. stanghellini ) .
the _ spitzer _
spectra of cn 1 - 5 , dddm 1 , h 1 - 50 , ic 4846 , m 1 - 20 , m 2 - 27 , m 2 - 42 , m 2 - 31 , and mycn 18 have been studied already in several works @xcite , where some of the dust features in these objects are identified .
the pne were observed with at least one of the four different modules of irs ( sl , ll , sh , and lh ) and , therefore , the wavelength coverage and the spectral resolution are not the same for all the nebulae .
each module is named by its wavelength coverage and resolution as short - low ( sl , covering the range 5.214.5 @xmath1 m with a spectral resolution @xmath181/@xmath838 = 64128 ) , long - low ( ll : 14.038 and the same spectral resolution as sl ) , short - high ( sh : 9.919.6 @xmath1 m and @xmath181/@xmath838 @xmath3 600 ) , and long - high ( lh : 18.737.2 @xmath1 m and the same spectral resolution as sh ) .
we used the observations performed in the _ staring _ mode in which the object is placed at two different positions ( nods ) in the slit .
we retrieved the data from the _
spitzer_/irs archive and we reduced them following the usual steps .
we started the reduction from the base calibrated data ( bcd ) .
first , we used the idl procedure doall_coads to produce a single coadded image for each nod position and module separately .
each of the low - resolution modules contains two slits ; when one of them is observing the source , the other one is off - source .
we used the off - source spectra to subtract the background in the on - source spectra .
for the high - resolution modules , extra sky images are needed to remove this contribution in the source spectra . in some cases
, there were no available images to remove the sky background , but this is not important for our identification purposes .
rogue pixels were removed with the irsclean tool .
the spectra for each nod position was wavelength and flux calibrated , and extracted with the _ spitzer _ irs custom extractor ( spice ) using one of the two available extractions : point - source aperture or full - slit extraction , depending on the size of the pne .
finally , we used the spectroscopic modeling analysis and reduction tool ( smart , @xcite ) to manually eliminate bad pixels , jumps , and to combine and merge into one final spectra the observations of each module .
high - resolution spectra were smoothed using a box - car algorithm ( of widths 0.06 @xmath1 m and 0.08 @xmath1 m for the sh and lh modules , respectively ) to slightly lower the noise without losing information about the dust features .
different regions of the final _
spitzer_/irs spectra for the 27 sample pne are shown in figures [ fig : dust_2][fig : dust_5 ] . for the convenience of the reader
, we display all the _ spitzer _ spectra available for our objects , although some of them have already been published , and some of the features of interest have been identified in the works we mentioned above .
a total of 18 pne have observations for the whole irs spectral range , 537 @xmath1 m , where many dust features can be observed ( figs .
[ fig : dust_3 ] and [ fig : dust_4 ] ) .
the other nebulae do not have data for the full spectral range that can be covered with _
spitzer_/irs . on the other hand ,
some objects have wavelength ranges covered at both high and low spectral resolutions . in such cases , we show the spectra where the dust emission features can be seen more clearly . using the irs spectra
, we identify the broad feature around 11.3 , associated with sic grains , in hu 2 - 1 , ic 418 , and m 1 - 20 ( figs .
[ fig : dust_2][fig : dust_3 ] ) .
this feature was already detected by @xcite and @xcite in ukirt and iso spectra of these three pne , and also in ic 2165 and ngc 6572 . in the last two pne , the feature is not as evident as for the other objects . in the case of ic 2165 , there are _ spitzer_/irs spectra , and we do not see clear evidence of the feature ( fig .
[ fig : dust_2 ] ) .
therefore , we label the identification as doubtful in this pne .
the 30 @xmath1 m feature , often attributed to mgs grains , is present in the spectra of ngc 3242 and m 1 - 20 ( see fig . [
fig : dust_4 ] ) .
besides , @xcite detected this feature in _ iso _ spectra of ic 418 , ngc 40 , ngc 3918 , and ngc 6826 .
we do not observe the feature in ngc 40 , but this could be due to the different areas observed by _ iso _ ( @xmath3 440 arcsec@xmath668 , covering almost the whole nebula ) and _ spitzer _ ( @xmath3 248 arcsec@xmath668 ) .
there are pah features in the irs spectra of 12 of the 27 pne with available data ( see figs .
[ fig : dust_2][fig : dust_3 ] ) , and we can not rule out their presence in at least 2 of the other pne . some of the pne would require deeper spectra or a higher spectral resolution to reach more definitive conclusions . as an example , @xcite did not find pahs in m 1 - 42 using _ iso _ observations , but we think that the _ spitzer _ spectra of this nebula shows evidence of these features ( see fig .
[ fig : dust_3 ] ) . for this nebula
, a higher spectral resolution would show the presence or absence of pahs more clearly .
the detection of pahs is not clear in some nebulae , in particular those with spectra of low spectral resolution , because of contamination with emission lines ( see fig .
[ fig : dust_5 ] for some identifications ) .
we see that in those pne where pahs are more conspicuous , such as ngc 40 and cn 1 - 5 ( fig .
[ fig : dust_3 ] ) , the broad feature at 11.3 @xmath1 m is the pah feature most easily seen .
hence we used the presence of this broad feature as the main criterion for the detection of pahs .
originally , pahs were thought to appear only in those dust forming sources that are carbon - rich @xcite , but in recent years these molecules have also been observed in oxygen - rich agb and post - agb stars , and in oxygen - rich pne ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
these so - called mixed - chemistry pne show pahs and crystalline silicate features together in their infrared spectra ( see e.g. * ? ? ? * ; * ? ? ?
we will come back to this issue in [ orich ] .
figure [ fig : dust_5 ] shows the 614 @xmath1 m irs spectra of ngc 6720 , which has observations at three different locations .
the upper panel shows the spectrum for a region that includes part of the outer halo of the nebula ; the middle panel shows the spectrum for an aperture that crosses the halo ; and the lower panel shows the spectrum for an aperture that crosses the bright region close to the center of the nebula .
we do not see evidence for pahs , sic , nor amorphous silicates .
we observe the broad features associated with amorphous silicates in 35 pne of the 27 pne with irs/_spitzer _ spectra in the adequate wavelength range ( see figs .
[ fig : dust_3 ] and [ fig : dust_4 ] ) .
dddm 1 , mycn 18 , and h 1 - 50 show the feature at 9.7 @xmath1 m and some indication of the 18 @xmath1 m feature .
we can not rule out the presence of these features in ic 4846 and ngc 6210 .
some pne , like ngc 40 , ngc 2392 , and ngc 6210 , show a small bump around 1718 @xmath1 m , but this feature is likely to be related with problems in the overlap of the sh and lh spectra .
finally , @xcite identified amorphous silicates in the _ iso _ spectra of ngc 6543 and , probably , in ngc 6153 .
we also identify the crystalline silicate features around 23@xmath1 m , 27@xmath1 m , and 33 @xmath1 m in the irs spectra of 12 pne from the sample .
we can not rule out the presence of crystalline silicates in ic 4846 and ngc 6818 due to the low resolution or poor signal - to - noise ratio of their spectra .
@xcite used _ iso _ spectra and detected crystalline silicates in ngc 6543 , whereas they could not rule out their presence in ngc 6153 .
therefore , 1316 nebulae from the 21 pne with available spectra in the required wavelength range show crystalline silicates .
the three pne with amorphous silicates also show crystalline silicates .
six out of the thirteen pne that show silicate features belong to the mixed - chemistry class , objects that show both silicates and pahs .
several scenarios have been proposed to explain the simultaneous presence of silicates and pahs in pne ( see , e.g. , * ? ? ?
one of the preferred explanations is that the silicates were formed and ejected before the star experienced the third dredge - up , when it was still o - rich , whereas pahs form later @xcite . according to this scenario ,
silicates are located in a o - rich disk / torus while pahs occupy a more recent c - rich outflow .
this could explain the mixed chemistry of cn 1 - 5 , where me measure c / o @xmath6 .
however , this explanation is not valid for stars that remain o - rich during their whole evolution , either because the third dredge - up does not occur due to their low initial masses or because the effect of the third dredge - up is counteracted by the hot bottom burning process . in these pne
, it might apply the explanation proposed by @xcite , in which the pahs form in an o - rich and dense torus after the co molecules are dissociated .
the values of @xmath839 in our sample of pne go from 8.02 ( for the halo pn dddm 1 ) to 8.88 ( for ngc 6620 ) .
if we consider the values of o / h found for the pne that have different dust features , they cover the ranges : 8.268.87 ( pahs ) , 8.268.68 ( sic or 30 @xmath1 m feature ) , 8.028.87 ( silicates ) , and 8.458.87 ( mixed - chemistry objects : silicates and pahs ) .
thus the sample pne with silicate dust grains seem to arise from a broad population that includes the halo pn , dddm 1 , pne from the galactic disk , and objects that might belong to the bulge , like h 1 - 50 and m 2 - 31 @xcite .
our pne with sic and the 30 @xmath1 m feature seem to arise from a more homogeneous population in the galactic disk ; only one of them , m 1 - 20 , might belong to the bulge .
this agrees with the results of @xcite , who found that six of their compact galactic pne with c - rich dust features , for which they could estimate the dust temperature , follow a well defined sequence in their plots of dust temperature versus infrared luminosity or physical radii .
@xcite conclude that this is an evolutionary sequence and that the progenitors of these pne covered a narrow range in initial mass and metallicity .
the high oxygen abundances found for the mixed - chemistry pne also agree with the results of @xcite , since they found that these nebulae are concentrated towards the galactic center and are absent from the magellanic clouds .
the range of iron depletions we find in our sample of pne covers about two orders of magnitude , suggesting differences in the formation and evolution of dust grains from one pn to another . as in @xcite , we explored possible correlations between the iron depletions and different nebular and stellar parameters such as the electron density , the nebular radius , the surface brightness , the effective temperature of the central star , or the nebular morphology . in agreement with @xcite , we do not find any obvious correlation .
we investigate here if this large range of iron depletions is related to the type of dust grains found in the pn ( i.e. , c - rich or o - rich dust grains ) or to the dominant chemistry present in the ionized gas ( i.e. , if the computed c / o ratio is lower or greater than one ) . in figure
[ fig : disc_1 ] we display the values of fe / o and the depletion factors for fe / o as a function of the c / o abundance ratios , derived from cels and rls , for all the pne in our sample with available data .
note that the sample pne can have two , one , or no values for c / o ( depending on whether the required cels and rls are measured or not ) and hence not all the pne in table [ tab:3 ] appear in figure [ fig : disc_1 ] , and some objects only appear in the panels at the right ( left ) .
we present the results for the values of fe / o derived with equations ( [ eq:2 ] ) and ( [ eq:3 ] ) , but the values of fe / o implied by equation ( [ eq:1 ] ) lead to similar results .
we also identify in figure [ fig : disc_1 ] the pne showing dust features in their infrared spectra , as well as those with doubtful or no detection of dust features .
note that , given the uncertainties related to the calculation of c / o from emission lines , the type of dust grains ( o - rich or c - rich excluding pahs ) might provide better information on whether the intrinsic value of c / o in the nebular gas is above 1 or not ( unless grain components like sic can form in slightly o - rich conditions , see * ? ? ?
the upper panels of figure [ fig : disc_1 ] show the pah identifications .
the first studies on the relation between pahs and the c / o abundance ratios found a trend of increasing strength of these features as c / o increases ( * ? ? ?
* and references therein ) , and it was suggested that these molecules are only present in pne with c / o @xmath6 .
here we find some pne with intense pah features and low c / o ratios , such as ngc 7026 or ngc 6439 .
besides , pah emission is not restricted to pne with c / o @xmath6 , in agreement with the idea that pah molecules may form both in c - rich and o - rich environments @xcite .
we also see from this figure that pne with pahs are distributed over the whole range of iron depletions , suggesting that the presence of these molecules has no direct relation with the highest and lowest iron depletions found in the ionized gas .
the middle panels of figure [ fig : disc_1 ] identify those pne with sic or the broad feature at 30 @xmath1 m , usually associated with dust grains formed in a c - rich environment .
most of these pne are located close to the region where c / o @xmath6 .
ngc 40 has a c / o value from rls that is not consistent within the errors with c / o @xmath6 , but its c / o value from cels is consistent with c / o = 1 . on the other hand
, @xcite find in their simulations of terrestrial planet formation that solid sic can form when the c / o ratio is above 0.8 . in this case , the range of c / o ratios for nebulae with c - rich dust can be somewhat wider .
the lower panels of figure [ fig : disc_1 ] show the pne with amorphous or crystalline silicates in their spectra .
these dust features seem to be very common in our sample .
all but one of those pne with silicate features have values of c / o that are compatible with an o - rich environment .
cn 1 - 5 is the exception , showing silicates with a c / o value clearly above one , regardless of what lines are used in the abundance determination and regardless of the adopted icf .
one possible explanation for this is that the c@xmath179/o@xmath179 values derived from cels and rls are seriously wrong for this nebula .
another option , already mentioned in [ sec : dust ] , is that the computed c / o value in the ionized gas is correct , and the prominent silicate features are revealing an o - rich past where the silicate grains formed . if we consider as c - rich those pne with sic or the 30 @xmath1 m feature and as o - rich those with silicates , we can see in figure [ fig : disc_1 ] that both c - rich and o - rich pne cover a wide range of iron depletions . however , c - rich nebulae are not present in the region of lowest iron depletions ( highest gaseous iron abundances ) , whereas o - rich pne do not appear at the highest depletions .
the pne with the highest iron depletions that have identifications of dust features are ic 418 and ngc 3242 , two pne with c - rich dust features and no silicates . on the other hand , the pne with the lowest depletions that also have identifications of dust features are cn 1 - 5 , dddm 1 , m 1 - 42 , and ngc 2392 , four pne that show silicates and no c - rich dust features . in order to check
whether this finding is due to small - number statistics , we performed a kolmogorov - smirnov ( ks ) test that compares the distributions of iron depletions for c - rich and o - rich pne ( as defined by the presence of sic or the 30 @xmath1 m feature , versus the presence of silicates ) . using the values of fe / o in figure [ fig : disc_1 ]
we find that the p - value ( the probability of finding different distributions of fe / o for c - rich and o - rich pne if both arise from the same distribution ) is 4% .
we also performed the ks test using the values of fe / o implied by equation ( [ eq:1 ] ) and the values of fe / h implied by both ionization correction schemes , finding probabilities in the range 0.2% to 7% , with the values of fe / h implying the lowest values ( 0.2% and 1% ) .
therefore , the difference is significant at the 0.27% level . an independent way to test this relation of iron depletions with c / o
is by studying the correlation between the iron abundances and the values derived for c / o .
we studied this correlation considering the two values of fe / h and fe / o implied by the two different ionization correction schemes we use , and the values of c / o implied by either cels or rls ( making a total of 8 cases ) .
we find weak spearman rank correlation coefficients , in the range @xmath645 to @xmath840 , and p - values that go from 0.1% ( for iron abundances derived from eq .
( [ eq:2 ] ) and ( [ eq:3 ] ) and c / o values implied by rls ) to 30% .
hence , the relation between iron depletions and c / o abundances , although weak , seems to be significant . this relation could arise from changes in the composition of dust grains for environments characterized by different values of c / o .
our results do not agree with the model results of @xcite for dust production in agb stars .
whereas most of our objects show high iron depletion factors , their calculations only show large degrees of condensation for iron , above 80% , in objects where the value of c / o is extremely close to 1 ( see their figs . a.3 and a.4 ) .
as discussed by @xcite , who find strong depletions of elements like fe and ca in the circumstellar envelope of an agb carbon star , adsorption of metals at the surface of the dust grains may be as important as the initial dust condensation processes considered by the models of dust formation in stellar envelopes .
we have studied a sample of 56 pne , belonging to the galactic bulge , the halo , and the disk .
the nebulae have available optical spectra of good quality and we constrain their iron abundances using [ ] lines and the ionization correction scheme derived by @xcite .
the iron depletion factors we find imply that most of the studied pne have less than 10% of their iron atoms in the gas , the missing iron atoms are presumably condensed into dust grains .
we have calculated the c / o abundance ratios using both optical rls ( for 49 pne ) and ultraviolet and optical cels ( for 39 pne ) .
the differences between the two estimates reach @xmath760 dex in some objects .
most of the sample pne show c / o @xmath0 which , according to theoretical models , indicates that they could either descend from stars with masses below @xmath841 m@xmath762 or from stars with masses above 45 m@xmath762 .
those pne with c / o @xmath6 are probably the result of intermediate mass stars .
we have also studied the infrared spectra of the pne with available data in order to identify the following features : pahs , sic , the broad band at 30 @xmath1 m , and amorphous and crystalline silicates .
we find pahs in 1214 of the 33 pne with infrared spectra .
the presence of these features is not restricted to pne with c / o @xmath6 .
silicates are found in 1316 pne ; sic and/or the 30 @xmath1 m feature , which is also associated with c - rich environments , in 79 objects .
the presence of silicates and sic or the 30 @xmath1 m feature does not agree in all cases with the values of c / o derived from emission lines , maybe reflecting the uncertainties related to the determination of the c / o abundance ratio .
note also that the c / o value we measure can be affected by the depletion of c or o atoms into dust grains .
the oxygen abundances , @xmath839 , found for the pne that have different dust features cover the ranges : 8.268.87 ( pahs ) , 8.268.68 ( sic or 30 @xmath1 m feature ) , 8.028.87 ( silicates ) , and 8.458.87 ( mixed - chemistry objects : silicates and pahs ) .
therefore , the sample pne with silicate dust grains seem to arise from a broad population that includes our halo pn , pne from the galactic disk , and objects that might belong to the bulge .
our pne with sic and the 30 @xmath1 m feature seem to arise from a more homogeneous population in the galactic disk , while the mixed - chemistry objects arise from pne of relatively high metallicities , results that agree with those found by @xcite using measurements of the dust temperature and the distribution in the galaxy of their sample objects .
we find a relation between the iron depletions and the type of chemistry in the nebulae , c - rich or o - rich .
both c - rich and o - rich pne cover a wide range of iron depletions , but the pne with the highest iron depletions have c - rich dust features ( sic and/or the 30 @xmath1 m band ) whereas those pne with the lowest iron depletions have silicates in their infrared spectra . in accordance with this result , we find a weak but significant anticorrelation between the iron abundances and the c / o abundance ratios derived from emission lines .
some kind of correlation is expected from the different molecules and dust compounds that should form in environments with different values of c / o .
we thank the anonymous referee for valuable comments that helped to improved the paper .
we acknowledge support from mexican conacyt grants 50359-f and 131610-f .
we have used nasa s astrophysics data system , and the simbad database operated at cds , strasbourg , france .
this work is based in part on archival data obtained with the _ spitzer _ space telescope , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . | we study the dust present in 56 galactic planetary nebulae ( pne ) through their iron depletion factors , their c / o abundance ratios ( in 51 objects ) , and the dust features that appear in their infrared spectra ( for 33 objects ) .
our sample objects have deep optical spectra of good quality , and most of them also have ultraviolet observations .
we use these observations to derive the iron abundances and the c / o abundance ratios in a homogeneous way for all the objects .
we compile detections of infrared dust features from the literature and we analyze the available _ spitzer_/irs spectra .
most of the pne have c / o ratios below one and show crystalline silicates in their infrared spectra .
the pne with silicates have c / o @xmath0 , with the exception of cn 1 - 5 .
most of the pne with dust features related to c - rich environments ( sic or the 30 @xmath1 m feature usually associated to mgs ) have c / o @xmath2 .
pahs are detected over the full range of c / o values , including 6 objects that also show silicates .
iron abundances are low in all the objects , implying that more than 90% of their iron atoms are deposited into dust grains .
the range of iron depletions in the sample covers about two orders of magnitude , and we find that the highest depletion factors are found in c - rich objects with sic or the 30 @xmath1 m feature in their infrared spectra , whereas some of the o - rich objects with silicates show the lowest depletion factors . |
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.